Dennis Miedema What Attracts Women Ebook

Spectroscopy 24 (2010) 585–592 585
DOI 10.3233/SPE-2010-0479
IOS Press
Dynamic properties of a reconstituted myelin

sheath
W. Knoll a,b , F. Natali c,∗ , J. Peters a,b,d , R. Nanekar e , C. Wang e and P. Kursula e
a
University Joseph Fourier, Grenoble, France
b
Institut Laue-Langevin, Grenoble, France
c
CNR-IOM, OGG, c/o Institut Laue-Langevin, Grenoble, France
d
Institut de Biologie Structurale, Grenoble, France
e
Department of Biochemistry, University of Oulu, Oulu, Finland
Abstract. Myelin is a multilamellar membrane which, wrapping the nerve axons, increases the efficiency of nervous signal
transmission. Indeed, the molecular components of the myelin sheath interact tightly with each other and molecules on the
axonal surface to drive myelination, to keep both myelin and the axon intact, and to transduce signals from myelin to the axon
and vice versa. Myelin is strongly affected in human demyelinating diseases in both the central and peripheral nervous system
(CNS and PNS, respectively). Despite the presence of a well-defined set of myelin-specific proteins, little is known about the
structure and the dynamics of these proteins, their interactions with the membrane and their influence on myelin stability. We
present here the first neutron scattering results on the dynamics of the myelin sheath in PNS and of the interaction between its
constituents. Specifically, the human P2 protein is shown to stabilize the lipid membrane upon binding to it.
Keywords: Membrane dynamics, neutron scattering, myelin
1. Introduction
Biological membranes serve diverse basic functions in eukaryotic and prokaryotic cells, regulating
the exchange of matter, energy and information with the surrounding environment. In a functional sense,
proteins obviously play a pertinent role in many of these functions. Yet, their lateral organization within
membranes is often highly dynamic, causing their transient localization in different lipid environments
that may affect protein structure, their oligomerization and, thereby, function.
Myelin is the lipid-rich, multilamellar membrane responsible for the speed-up of the nerve impulse
conduction in vertebrates (up to 100 m/s in a healthy person). Myelin is destroyed by autoimmune
processes in human demyelinating diseases, including multiple sclerosis (MS) in CNS and peripheral
neuropathies (PN) in PNS. MS is characterized by patches of demyelination [10] that can be scattered
throughout the white matter of the CNS, but are usually localized in the brain stem, periventricular areas,
the optic nerve and the cervical spinal cord. These multiple scars (or scleroses) cause symptoms specific
to the interrupted signals.
The biochemical composition of myelin includes only a few major proteins [19]; two proteins, the
proteolipid protein (PLP) and the myelin basic protein (MBP) account for 80% of the total CNS myelin
protein. CNS and PNS myelin differ in their protein constituents. MBP is also an abundant component
*
Corresponding author: Dr. F. Natali, CNR-IOM, c/o Institut Laue-Langevin, BP 1566, rue Jules Horowitz, 38042 Grenoble
Cedex 9, France. Tel.: +33 4 76 20 70 71; Fax: +33 4 76 20 76 88; E-mail: natali@ill.fr.
0712-4813/10/$27.50 © 2010 – IOS Press and the authors. All rights reserved
586 W. Knoll et al. / Dynamic properties of a reconstituted myelin sheath
of PNS myelin [11], where it colocalizes with another myelin-specific protein, P2. The P2 protein, while
missing in CNS, represents 1–15% of the total myelin protein in the PNS.
MBP is essential for the formation of CNS myelin, able to interact with a wide range of ligands,
often polyanionic in nature [2]. Due to its high charge and low overall hydrophobicity, there is signifi-
cant intramolecular electrostatic repulsion in MBP; this results in MBP being an extended, intrinsically
unstructured protein in solution [12,21].
P2 is a small folded protein with a high homology to members of the fatty acid binding protein family
[7,18]. It is expressed in the cytoplasmic face of compact myelin in the PNS [9,27], and its lipid-binding
activity suggests it may serve important functions in generating and maintaining the unique lipid com-
position of myelin [28]. P2 is possibly one of the autoantigens in the human disorder Guillain–Barre
syndrome [14,15], an autoimmune disease of the PNS.
Although the main subject of myelin research often concerns its proteins, dynamic changes in protein–
lipid interactions and lipid-based membrane microdomains also have a role in the etiology of MS and
PN. Detailed knowledge of the fine structure and dynamics of myelin at the level of lateral membrane
domain composition, and lipid–protein interactions within such domains, is imperative for understanding
the unknown pathogenic mechanisms of demyelination.
Biochemical and structural studies on myelin flourished during the 1970s and 1980s [3–6,16,26,29],
and correlations have been observed between myelin structural modifications [13,22] and some common
neurological degenerative events of pathological nature in humans. In this context, wide angle X-ray
diffraction studies [5] revealed that the lipid phase transition temperature of MS myelin is 20◦ C lower
than that of normal myelin, indicating differences in the structural organization of the bilayer.
Dynamic properties are of great importance at the molecular level of biology and the marked tem-
perature dependence of the activity of biomolecules reflects their thermal mobility. Dynamic events in
biomolecular systems occur on a very large time-scale, ranging from femtoseconds to nearly seconds.
Within this broad interval, motions occurring in the pico- to nano-second time-scale are of particular
interest and relevance, since they cover the transition region from ‘discrete’ local excitations of small
molecular subunits to slower processes involving collective motions of massive parts of macromolecular
assemblies. This time window is well covered by inelastic and quasielastic neutron scattering, These
techniques can, therefore, play a relevant role in improving the understanding of molecular motions
which affect the functionality of biomolecules.
We have demonstrated, during the last five years [8,23–25], the power of the neutron scattering tech-
nique as an optimal probe for the dynamics of the myelin sheath and its components. Our studies so far
focused on the effect of MBP on the global membrane dynamics of ordered dimyristoyl L-a-phosphatidic
acid (DMPA) assemblies, simulating the myelin sheath in the CNS. The study covered the lipid gel
(phase α) to liquid (phase β) phase transition (Tα→β = 320 K). Strong anisotropy in the dynamic
behaviour depended on the membrane composition, and MBP was found to strongly enhance the out-of-
plane mobility of the lipid molecules in the liquid phase. In this study, we now address the PNS system,
and effects of both MBP and P2 on liposome membranes are investigated.
2. Materials and methods
2.1. Protein purification
Details of the protein purification protocols will be published elsewhere. In brief, His-tagged human
P2 was expressed in E. coli and purified using immobilized metal ion affinity chromatography. The His-
W. Knoll et al. / Dynamic properties of a reconstituted myelin sheath 587
tag was cleaved off with recombinant TEV protease and purification was completed by gel filtration.
The recombinant His-tagged C1 and C8 isoforms of murine 18.5 kDa MBP were expressed in E. coli
and purified by slight modifications of previously published procedures [1]. All protein samples were
lyophilized prior to shipment to the neutron source and final sample preparation.
2.2. Sample preparation and characterization
1,2-dimyristoyl-sn-glycero-3-phosphate (DMPA), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC)

and 1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS) were purchased as dry powders from Lipoid
Company (Germany). After dissolving the lipids in chloroform and subsequent drying with nitrogen
gas, they were solubilised in D2 O. By extruding the lipid solution through a 100 nm pore sized mem-
brane, using the Avanti® Mini-Extruder, unilamellar vesicles were prepared. To compare liposomes of
DMPA and DOPS + DOPC with these liposomes mixed with the MBP or P2 proteins, the proteins were
added to one part of the lipid solution before the extrusion. The hydrodynamic radius (Rh ) of the lipo-
somes, with and without proteins, was estimated by dynamic light scattering (DLS). The experiments
were carried out at the EMBL outstation in Grenoble using the ALV CGS-3 Compact Goniometer. The
obtained correlation intensities were normalized to the measurement of a standard for which toluene is
typically used. The suitable lipid concentration for DLS measurement was found to be of 0.001 mg/ml.
Dust particles were removed by using a filter with 20 µm pores.
2.3. Neutron scattering experiment
Incoherent elastic neutron scattering measurements as a function of temperature were performed on

the thermal (λ = 2.23 Å) high-energy resolution backscattering spectrometer IN13 (Institut Laue-
Langevin, Grenoble, France), characterized by a very large momentum transfer range (0.2 < Q <
4.9 Å−1 ) with a good and nearly Q-independent energy resolution (8 µeV FWHM). IN13, therefore,
allows accessing the space and time windows of 1–6 Å and 0.1 ns respectively.
The elastic scattering intensities (Iel (Q) = S(Q, ω = 0)), properly corrected for the empty sam-
ple holder signal, were normalized with respect to a vanadium scan (typically used as a standard), to
compensate for spurious background contributions and detector efficiency. The sample mass and thick-
ness were suitably chosen to optimize the compromise between good signal-to-noise ratio and multiple
scattering contributions. For this purpose, a transmission of approximately 90% was guaranteed using
100 mg of powder per sample allocated in 1.5 mm thick aluminum sample holders.
3. Results and discussion
The temperature region (278 < T < 313 K) was chosen in order to (1) avoid any Bragg reflections
arising from ice assigned to free water molecules in- and outside the liposome (T > 273 K) and (2) pre-
vent P2 from denaturation (T < 320 K). In order to investigate how the membrane dynamics are affected
by the presence of P2 in both the gel and liquid lipidic phase (which would allow us a direct comparison
to our previous study in CNS), two different liposome compositions were chosen. The DMPA artificial
membrane shows a lipid Lα → Lβ phase transition at T = 320 K. Thus, in the whole 273 < T < 315 K
region, DMPA is in the gel phase. On the other hand, two-component liposomes, made of a mixture
of neutral (DOPC) and charged (DOPS) lipids, at 1:1 w/w ratio, were chosen to investigate the liquid
lipid phase. Indeed, both DOPC and DOPS show a lipid Lα → Lβ phase transition well below room
temperature (DOPC: Tα→β = 253 K; DOPS: Tα→β = 264 K). Thus, the mixed lipid compound is in
the liquid phase in the entire T range investigated here.
While DMPA was chosen to compare results in PNS to our previous investigation in CNS [23–25], the
neutral-charged mixture was chosen to better represent the lipid composition in myelin. The chosen P2
concentration (1.25% of the total sample mass) allows neglecting the scattering arising from the protein
(5% at maximum of the total scattering signal). Thus, any differences in the neutron response can be
unambiguously assigned to changes in membrane dynamics induced by the presence of the P2 protein.
In Fig. 1, we report the incoherent elastic intensity of DMPA liposomes vs. the square of momentum
transfer. The fast decrease of the intensity observed for all the temperatures investigated is a clear sign
of enhanced dynamics that may count also for some additional quasi-elastic components to the elastic
scattering. In particular, two main effects may contribute to the QENS additional terms: (1) confined
water inside the liposomes, and (2) liposome diffusion in the heavy water medium. D2 O was chosen
as solvent, indeed, to minimize the first contribution to the total scattering intensity. However, while
free water outside the liposomes does not participate in our signal since it counts as flat background
in the energy window explored here, the heavy water molecules inside the liposomes may include at
least two different populations: free (at the center of the liposomes) and bound (closer to the lipid polar
heads of the internal leaflet) molecules. The latter may give rise to a quasi-elastic contribution, which
can fall within the energy resolution of the experiment, thus affecting the correct interpretation of the
elastic data, since it results in a spurious decrease of the elastic intensity. Thus, bound water molecules
have to be taken into account for a correct data interpretation. Also liposome diffusion has to be properly
Fig. 1. Normalized elastic intensity of DMPA liposomes vs. T . For clarity, only 3 temperatures are shown (circles: 277 K;
squares: 287 K; triangles: 297 K). Insert: experimental DLS correlation functions of DMPA liposomes measured before and
after the neutron experiments (continuous and dotted lines, respectively).
considered, since at the used lipid concentration (50 mg/ml), the system has to be considered as crowded,
thus speeding down the lipid diffusion constant to values observable on IN13. To identify and quantify
all the different contributions to our elastic data, further QENS experiments are planned on several
complementary instruments, allowing us to span a wide energy range (1–100 µeV).
In order to check the quality of the sample we systematically performed DLS measurements before
and after the neutron experiments. In the insert of Fig. 1, we report, as an example, the experimental
time dependent correlation function G(t) of DMPA liposomes measured before and after the neutron
experiments. The two curves overlap perfectly and can be fitted using a single and narrow liposome
population of hydrodynamic radius Rh = 33 nm and a very low polidispersity (p = 0.98).
The proper protein-liposome binding is currently checked by gel electrophoresis using a sucrose gra-
dient protocol [17].
The influence of the composition of the two protein-free liposome samples used in this work (DMPA
and DOPC + DOPS) is shown in Fig. 2, where the integrated elastic intensities are reported as a function
of the explored T range. In order to gain in statistics, data have been binned over the interesting Q region
where the non-zero intensities are still appreciable (Q2 6 Å−2 ). A remarkable difference is observed,
supporting a damping of membrane dynamics in the DMPA system. This may be assigned to the different
lipid phases explored by the two types of liposomes in the measured T region. As expected, indeed, the
liquid phase at which DOPC+DOPS liposomes are trapped reflects much higher lipid motion. However,
a contribution from a different behavior of the dynamics of the confined water molecules inside the
liposomes cannot be excluded, due to a difference in the charge distribution on the lipid inner surface;
the neutral DOPC lipids behave as spacers, probably causing a change in the lipid polar head-water
molecules interactions.
The mean square displacements (MSD) extracted for the protein-free and P2-bound liposomes are re-
ported in Fig. 3. All data on MBP bound liposomes are here only briefly commented and not shown, even
Fig. 2. Comparison of temperature dependence of the normalized elastic intensity of mono and bi-component liposomes. To
gain in statistics, data are integrated over the Q2 region up to 6 Å−2 (DMPA: filled symbols; DOPC + DOPS: empty symbols).
Fig. 3. Temperature dependence of the mean square displacements of proteolipid complexes compared to their respective
liposomes. Upper panel: DMPA complexes; lower panel: DOPC + DOPS complexes. Empty symbols: liposomes with P2;
filled circles: protein-free liposomes. The k force constants, estimated using the force constant model, are also reported.
though in perfect agreement to previous results [23–25], because of the poor sample quality suggested
by significant changes in the DLS spectra as a function of time. The MSD values have been calculated
from the slope of the elastic intensity vs. Q2 , in the low Q limit, where the Gaussian approximation
is valid. The MSD count for both harmonic and anharmonic contributions of the all proton dynamics.
A detailed description of the MSD extraction protocol may be found elsewhere [20]. The binding of P2
to the liposomes shows the same effect on the two lipocomplexes studied here: P2 is found to enhance
the membrane stability, reducing the lipid dynamics. Using the free constant model successfully intro-
duced by Zaccai in 2000 [30], the effective force constants (k parameters) have been estimated for our
compounds (Fig. 3). The k constants are an indirect measure of the stability of a system; higher k values
refer to a more rigid system, confirming the property of P2 to enhance the membrane stability, which
may play a key role in proper nerve impulse transmission.
4. Conclusions
P2 is a highly abundant protein of the PNS myelin membrane, with putative roles in the formation
and homeostasis of myelin; the relationships between the structure, function, and dynamics of P2 have,
however, been poorly characterized. The results presented here strongly suggest a stabilizing effect in-
duced by the P2 protein on the membrane dynamics, independently from the liposome composition (in
particular from the net charge density of the liposomes) and from the lipidic phase transition in which
the liposomes are trapped (gel phase in DMPA and liquid phase in DOPC + DOPS). One possible phys-
ical explanation for the observed higher rigidity may account for the recent observation (data not yet
published) by molecular dynamics simulation, of the partial disposition of the P2 protein inside the lipid
bilayer and on its surface, affecting in such a way both in plane and out-of-plane lipid motions.
In the near future, QENS experiments are planned to get better insight into the properties exerted by
P2, in particular to map selectively the different lipid motions, which may be affected by the lipid–protein
interaction. Complementary elastic experiments will also soon be performed on the same liposome com-
positions, where the free water molecules will be removed by promoting oriented multibilayer formation,
followed by drying treatment, in order to extend the study of the DOPC + DOPS liposomes below room
temperature, i.e., in the gel phase. This will allow us to discriminate between net charge density, lipid
phase and chemical lipid composition contributions. Taken together, our results elucidate the functional
properties of a quantitatively major component of vertebrate PNS myelin.
Acknowledgements
We would like to express our gratitude to Mrs. Iulia Blesneac, Ms. Niina Torniainen and Ms. Viivi
Majava for their help in sample preparation and characterization. This work has been supported by the
Academy of Finland and the Sigrid Juselius Foundation (Finland).
References
[1] I.R. Bates, P. Matharu, N. Ishiyama, D. Rochon, D.D. Wood, E. Polverini, M.A. Moscarello, N.J. Viner and G. Harauz,
Characterization of a recombinant murine 18.5-kDa myelin basic protein, Protein Expr. Purif. 20 (2000), 285–299.
[2] J.M. Boggs, Myelin basic protein: a multifunctional protein, Cell. Mol. Life Sci. 63 (2006), 1945–1961.
[3] C. Brunner, H. Lassmann, T.V. Waehneldt, J.M. Matthieu and C. Linington, Differential ultrastructural localization of
myelin basic protein, myelin/oligodendroglial glycoprotein, and 2 ,3 -cyclic nucleotide 3 -phosphodiesterase in the CNS
of adult rats, J. Neurochem. 52 (1989), 296–304.
[4] D.L. Caspar and D.A. Kirschner, Myelin membrane structure at 10 Å resolution, Nat. New Biol. 19 (1971), 46–52.
[5] L.S. Chia, J.E. Thompson and M.A. Moscarello, Alteration of lipid-phase behavior in multiple sclerosis myelin revealed
by wide-angle X-ray diffraction, Proc. Natl. Acad. Sci. USA 81 (1984), 1871–1874.
[6] L.S. Chia, J.E. Thompson and M.A. Moscarello, X-ray diffraction evidence for myelin disorder in brain from humans
with Alzheimer’s disease, Biochim. Biophys. Acta 775 (1984), 308–331.
[7] A. Chmurzynska, The multigene family of fatty acid-binding proteins (FABPs): function, structure and polymorphism,
J. Appl. Genet. 47 (2006), 39–48.
[8] A. Deriu and F. Natali, Recent neutron investigation on biomolecular dynamics: from model systems towards complex
macromolecular machines, J. Neutron Research 10 (2002), 115–122.
[9] E.H. Eylar, I. Szymanska, A. Ishaque, J. Ramwani and S. Dubiski, Localization of the P2 protein in peripheral nerve
myelin, J. Immunol. 124 (1980), 1086–1092.
[10] R.J. Franklin, Why does remyelination fail in multiple sclerosis?, Nat. Rev. Neurosci. 3 (2002), 705–771.
[11] B. Garbay, A.M. Heape, F. Sargueil and C. Cassagne, Myelin synthesis in the peripheral nervous system, Prog. Neurobiol.
61 (2000), 267–304.
[12] G. Harauz, N. Ishiyama, C.M. Hill, I.R. Bates, D.S. Libich and C. Farès, Myelin basic protein-diverse conformational
states of an intrinsically unstructured protein and its roles in myelin assembly and multiple sclerosis, Micron 35 (2004),
503–542.
[13] A. Hirano, S.D. Cook, J.N. Whitaker, P.C. Dowling and M.R. Murray, Fine structural aspects of demyelination in vitro.
The effects of Guillain–Barré serum, J. Neuropathol. Exp. Neurol. 30 (1971), 249–265.
[14] R.A. Hughes and D.R. Cornblath, Guillain–Barre syndrome, Lancet 366 (2005), 1653–1666.
[15] R.A. Hughes, R.D. Hadden, N.A. Gregson and K.J. Smith, Pathogenesis of Guillain–Barre syndrome, J. Neuroimmunol.
100 (1999), 74–97.
[16] D.A. Kirschner and A.L. Ganser, Compact myelin exists in the absence of basic protein in the shiverer mutant mouse,
Nature 283 (1980), 207–210.
[17] J. Knol, Membrane reconstitution and functional analysis of a sugar transport system, PhD dissertation, University of
Groningen, Groningen, 1999.
[18] P. Kursula, Structural properties of proteins specific to the myelin sheath, Amino Acids 34 (2008), 175–185.
[19] C. Laule, I.M. Vavasour, S.H. Kolind, D.K. Li, T.L. Traboulsee, G.R. Moore and A.L. MacKay, Magnetic resonance
imaging of myelin, Neurotherapeutics 4 (2007), 460–484.
[20] U. Lehnert, V. Reat, M. Weik, G. Zaccaï and C. Pfister, Thermal motions in bacteriorhodopsin at different hydration levels
studied by neutron scattering: correlation with kinetics and light-induced conformational changes, Biophys. J. 75 (1998),
1945–1952.
[21] V. Majava, C. Wang, M. Myllykoski, S.M. Kangas, S.U. Kang, N. Hayashi, P. Baumgärtel, A.M. Heape, G. Lubec and
P. Kursula, Structural analysis of the complex between calmodulin and full-length myelin basic protein, an intrinsically
disordered molecule, Amino Acids 39 (2010), 59–71.
[22] E.B. Masurovsky, M.B. Bunge and R.P. Bunge, Cytological studies of organotypic cultures of rat dorsal root ganglia
following X-irradiation in vitro. II. Changes in Schwann cells, myelin sheaths, and nerve fibers, J. Cell Biol. 32 (1967),
497–518.
[23] F. Natali, A. Gliozzi, R. Rolandi, P. Cavatorta and A. Deriu, Myelin basic protein reduces molecular motions in DMPA,
an elastic neutron scattering study, Appl. Phys. A 74 (2002), 1582–1584.
[24] F. Natali, A. Relini, A. Gliozzi, R. Rolandi, P. Cavatorta, A. Deriu, A. Fasano and P. Riccio, Protein-membrane interaction:
effect of myelin basic protein on the dynamics of oriented lipids, Chem. Phys. 292 (2003), 455–464.
[25] F. Natali, A. Relini, A. Gliozzi, R. Rolandi, P. Cavatorta, A. Deriu, A. Fasano and P. Riccio, The influence of the lipid–
protein interaction on the membrane dynamics, Physica B 350 (2004), E623–E626.
[26] P. Riccio, A. Tsugita, A. Bobba, F. Vilbois and E. Quagliariello, A new protein of the brain myelin: isolation and chemical
characterization, Biochem. Biophys. Res. Commun. 127 (1985), 484–492.
[27] B.D. Trapp, M. Dubois-Dalcq and R.H. Quarles, Ultrastructural localization of P2 protein in actively myelinating rat
Schwann cells, J. Neurochem. 43 (1984), 944–948.
[28] K. Uyemura, K. Yoshimura, M. Suzuki and K. Kitamura, Lipid binding activities of the P2 protein in peripheral nerve
myelin, Neurochem. Res. 9 (1984), 1509–1514.
[29] C.R. Worthington and T.J. McIntosh, Direct determination of the lamellar structure of peripheral nerve myelin at moderate
resolution (7A), Biophys. J. 14 (1974), 703–729.
[30] G. Zaccai, How soft is a protein? A protein dynamics force constant measured by neutron scattering, Science 288 (2000),
1604–1607.
Spectroscopy 24 (2010) 651–659 651
DOI 10.3233/SPE-2010-0490
IOS Press
Prediction of favourable sites for spin

labelling of proteins
Yevhen Polyhach and Gunnar Jeschke ∗
Laboratory of Physical Chemistry, ETH Zurich, Zurich, Switzerland
Abstract. Electron paramagnetic resonance studies on diamagnetic proteins are based on site-directed spin labelling and require
relatively much effort for engineering cystein point mutations as well as expressing and labelling the protein. Therefore, it
is advantageous to predict those sites and pairs of sites that can provide the most precise and most reliable information on
accessibility and distances. Systematic site scans based on a rotamer library approach for spin label side groups allow for such
predictions. Figures of merit are defined that can be used to rank sites according to their potential suitability in studies of
structure and structural changes. Such site scans can still provide useful results if only a backbone model and the amino acid
sequence of the protein are known.
Keywords: EPR spectroscopy, site-directed spin labelling, membrane proteins, molecular modelling, distance measurements
1. Introduction
Site-directed spin labelling (SDSL) of proteins [12], has recently become a widespread technique, in
particular, in combination with pulsed electron paramagnetic resonance (EPR) techniques for the mea-
surement of distance distributions between two spin labels [13,17,20]. It was early recognized that the
separation between the electron spin and the protein backbone poses problems in precise modelling
of structures and structural changes [3]. A more detailed study based on a molecular dynamics (MD)
approach demonstrated that conformational distribution of the spin label can provide the major contribu-
tion to the width of the measured distance distributions [16]. To model this conformational distribution
and to predict spin-to-spin distance distributions from a given model for the protein structure, we have
developed a rotamer library approach [13] that is similar in spirit to an approach for understanding spin
label dynamics [18,19].
This approach has so far yielded a structural model of the functional dimer of the Na+ /H+ antiporter
NhaA of E. coli [11], which was later confirmed by a mutation and crosslinking study [9] and by electron
crystallography [2]. Furthermore, a coarse-grained model was obtained for a kinked transmembrane he-
lix in the proline/Na+ symporter PutP of E. coli [10], whose full structure remains unknown. In cases of
known structures, satisfying agreement was found between measured distance distributions and distance
distributions predicted from crystal structures by the rotamer library approach [6,8]. A more sophisti-
cated approach to rotamer library generation and computation of the interaction between spin label and
protein has been developed and will be described in detail elsewhere.
*
Corresponding author: Gunnar Jeschke, Laboratory of Physical Chemistry, ETH Zurich, Wolfgang-Pauli-Strasse 10, CH-
8093 Zurich, Switzerland. Tel.: +41 44 632 5702; Fax: +41 44 633 1448; E-mail: gunnar.jeschke@phys.chem.ethz.ch.
652 Y. Polyhach and G. Jeschke / Prediction of favourable sites for spin labelling of proteins
The new rotamer library approach allows for prediction of the conformational distribution of a spin
label with a common personal computer in a computation time between less than a minute and a few
minutes. This opens up the possibility to systematically scan a protein structure for favourable spin
labelling sites and site pairs. Such computational site scans can potentially decrease the expense of
SDSL studies and provide more reliable results. Both very tight sites, where labelling fails or leads to
changes in structure and function and very loose sites, where distance distributions are very broad, can
be avoided. Site scans are most useful in situations where the structure of the studied proteins is known
and either structural changes [6,8] or the relative arrangement of several protein molecules [11] are of
interest.
In this paper we consider which information can be obtained from a computational site scan and how
the sites can be ranked in terms of suitability for certain EPR experiments. In particular, we consider
measurements of spin label accessibility [1] and distance measurements, as well as the study of structural
changes.
2. Information from rotamer attachment to a single site
Our approach for modelling of the conformational distribution of spin labels is based on a library
of about N = 100–200 rotameric states whose relative energies in the absence of interaction with the
protein are known approximately. To identify these states and estimate their relative energies we have
performed MD simulations of the free spin label with the CHARMM27 force field, using a recently
published parametrization for the methanethiosulfonate spin label (MTSL) [18] and our own parame-
trization for the iodoacetamido-proxyl (IA-Proxyl) spin label. Briefly, the states are derived by analyzing
maxima in the population of individual dihedral angles. The internal populations pi of these states can
be obtained by analyzing all frames of the MD trajectory in terms of the closest rotamer. These pop-
ulations are related to relative energies ei of the unconstrained label via the Boltzmann distribution,
pi /pmax = exp(−ei /kT ), where pmax is the population of the most populated rotamer, whose relative
energy is defined as zero, k is the Boltzmann constant and T the simulation temperature.
The relative energies of the rotamers attached at a certain site of the protein are computed by adding
to the ei further energies εi that correspond to the interaction between label and protein. The interaction
energy εi for a particular rotamer is a sum of pairwise Lennard-Jones potential energies between one
atom of the label and one atom of the protein. A distance cutoff is used to speed up computations and
scaling of the van der Waals radii in the Lennard-Jones potentials [4,14] is applied to empirically account
for flexibility of the protein structure. Details of the approach and validation data will be published
elsewhere.
This approach yields, for each site-attached rotamer, a predicted
population πi = pi exp(−εi /kT )/Z
and, for the whole set of rotamers, the partition function Z = pi exp(−εi /kT ). The partition func-
tion Z is a measure for the tightness of the site, with small values of Z corresponding to large positive
interaction energies between label and protein. Very small Z indicate that labelling may fail or, if it
succeeds, is likely to cause a change in protein structure. In predictions, we assume that for the MTS
rotamer library consisting of 210 rotamers Z < 0.05 indicates difficulties in labelling. This threshold
corresponds to a situation where label attachment becomes less favourable by an average strain energy
of approximately 7.5 kJ/mol compared to an unconstrained site.
As a measure for the width of the conformational ensemble, we use the number n of most populated
rotamers that together make up 99.5% of the total population. This number is correlated to the partition
Y. Polyhach and G. Jeschke / Prediction of favourable sites for spin labelling of proteins 653
function Z, however, the correlation is not strict. For instance, a single, low-energy rotamer can lead
to a larger value Z than tens of rotamers that all have high, but similar energies. These two scenarios
correspond, respectively, to a favourable tight site, where the distribution is narrow and energy low, and
an unfavourable tight site, where the distribution is broad and energy high.
From an experimental point of view, the spatial distribution of the midpoint of the nitroxide N–O bond
is of greater interest than the number of populated rotamers. This midpoint is the approximate site of
the electron spin. A broad spatial distribution induces ambiguities in accessibility studies and may cause
broad distance distributions. These questions are now discussed in turn.
3. Site choice for accessibility measurements
The accessibility of the spin label to water or oxygen is most easily interpreted when the spatial
distribution of the label is narrow. This can be quantified by the root mean square deviation of the
midpoint of the N–O group from its mean position,
1/2
σNO = 0.005 nm2 + N/(N − 1) πi [(xi − x)2 + (yi − y)2 + (zi − z)2 ] . (1)
In Eq. (1) xi , yi and zi are

the Cartesian coordinates of the N–O midpoints of the rotamers (I =
1, . . . , N ), while x = πi xi is the mean value of the x coordinates and y and z are defined
analogously. The offset of 0.005 nm2 accounts for the spatial distribution due to nitroxide libration mo-
tion that is expected even if only one rotamer is populated [19].
For MTSL attached to an isolated cystein, i.e., for unconstrained MTSL, we find σNO = 0.64 nm
(see Fig. 1). Note that in special cases slightly larger values may occur for constrained MTSL. This
happens if conformations that are close to the mean conformation are more strongly depopulated than
conformations that are far from the mean conformation. The smaller σNO , the more favourable is the
site for an accessibility study, as long as the partition function Z indicates that the site can be labelled
without structural distortion. In the following, such favourable sites with σNO 0.2 nm are termed
well-constrained sites without collisions (WCSWC).
If the orientation of a membrane protein in a lipid bilayer is known, symmetry suggests a slightly
different figure of merit. Assume without loss in generality that the bilayer normal is the z axis of
the coordinate frame. Spatial distribution within the xy plane has smaller consequences than spatial
distribution along the z axis, since, to a first approximation, the concentrations of water and oxygen in
the bilayer vary exclusively along z. Although presence of the protein itself may induce some variation
in the xy plane, the quantity:
1/2
σNO,z = (0.005/3) nm2 + N/(N − 1) πi (zi − z)2 (2)
may be a better predictor for structure-related changes in accessibility than σNO .
4. Site pair choice for distance measurements
For measurements of the distance rNO,NO between two spin labels, both labels must have a partition
function Z that indicates labelling without structural distortion. Some label pairs can be excluded without
Fig. 1. Visualization of the distribution of midpoints of the N–O bond for completely unrestricted MTSL. The backbone atoms of
the labelled residue, the Cα –Cβ bond (pointing to the back) and the bond from the carbonyl C to the next residue (pointing to the
front, slightly right) are shown as a ball and stick plot. Rotameric states are visualized as magenta spheres at the N–O midpoints
with the radius proportional to the internal population pi . The spatial distribution of the N–O midpoints is visualized as a
transparent green ellipsoid with half axes corresponding to two times the standard deviation of the N–O midpoint coordinates.
(The colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0490.)
detailed analysis of the distribution of conformations, as they correspond to mean distances that are either
too short (rNO,NO < 0.8 nm) or too long (rNO,NO > 7 nm). Although very short distances can be detected
in principle and distance estimates can be obtained from the intensity of half-field transitions [5,7],
interpretation in the presence of conformational distribution of a nitroxide label would be unreliable.
Distances longer than about 7 nm cannot usually be measured in proteins with reasonable precision and
currently available techniques [13].
The σNO values of the two labelled sites are not necessarily a good indicator for the expected width of
the distance distribution. This is because the spatial distribution of the N–O midpoints is not spherical.
The origin of this effect is illustrated in Fig. 1 for MTSL attached to an isolated cystein. For this situa-
tion with an unconstrained conformational space, the inertia tensor of the N–O midpoints, considering
only the internal populations pi , is visualized as a green semi-transparent ellipsoid. The direction of
maximum spatial extent is along the principal axis of this tensor corresponding to the largest principal
value of 0.44 nm2 ; it includes an angle of 70.2◦ with the Cα –Cβ bond and an angle of 27◦ with the
peptide plane. If the spin–spin vector is parallel to this direction, the width of the distance distribution is
maximum. The direction of minimum spatial extent corresponds to the principal axis with the smallest
principal value of 0.27 nm; it includes an angle of 96◦ with the Cα –Cβ bond and an angle of 46◦ with the
peptide plane. If the spin–spin vector is parallel to this direction, the width is minimum. For completely
unrestricted MTSL, the ratio of the maximum and minimum principal value of the tensor of spatial
extension, corresponding to standard deviations of the N–O midpoint coordinates, is 1.6. Note that the
effect may be stronger if a subset of rotamers is selected due to interaction with the protein. For instance,
if two subsets of conformations are populated, the distance distribution is most strongly broadened if the
spin–spin vector is parallel to the vector between the centres of masses of the N–O midpoints of the two
subsets and it is least broadened if the spin–spin vector is perpendicular to this vector.
The value of experimental distance restraints is related to the relative width σr /r of the distance
distribution, where r is the mean distance and σr the standard deviation. This relative width is also
most relevant in pulsed EPR distance measurements, as the effect of distance distribution on the shape
of the dipolar evolution function is proportional to this parameter. Hence, we consider σr /r as the
figure of merit for a pair of sites. Here r is defined as:

r = πi πk rij πi πk (3)
i k i k
with indices i and k running over the N rotamers at the first and second site, respectively, and rij is the
distance between the N–O midpoints for a given combination of rotamers. Likewise, σr is defined as:
1/2
σr = 0.01 nm2 + N 2 /(N 2 − 1) πi πk (rij − r)2 πi πk . (4)
i k i k
5. Studying structural changes
A particular strength of SDSL EPR is the characterization of structural differences between different
states of a biological system, if the structure of only one of these states is known. In many of these
cases there exists an initial hypothesis about the structural change that is to be tested. Such a hypothesis
usually involves relative movements of protein domains or of the individual peptide chains in a complex.
The individual domains or chains can be approximated as rigid bodies. In such cases, the hypothesis can
be translated into an approximate model of the structure of the unknown state and a second site scan can
be performed for this model. If there are several conflicting hypotheses for the structural change, site
scans have to be performed for each corresponding model.
The figures of merit for sites and site pairs are then determined by relative changes between the states.
For instance, in an accessibility study with known orientation of the system with respect to the bilayer,
the mean change Δz in the z coordinate of the N–O midpoints divided by the larger of the two σNO,z
values predicts the potential of the experiment to detect the change.
For distance measurements by pulsed EPR techniques, experimental form factors (dipolar evolution
data after background correction) can be predicted from the simulated distance distribution. In this case,
the figure of merit for a pair of sites is the root mean square deviation between the normalized form
factors for the two states. The larger this deviation is, the easier it is to detect the structural change.
6. Further considerations
The figures of merit that result from computational site scans are not failsafe predictors of favourable
and unfavourable labelling sites. This is, first, because the rotamer library approach provides only an
approximate prediction of the conformational distribution and, second, because other structural aspects
of the sites may influence the suitability for a certain experiment. We consider two examples of such
structural aspects, one for accessibility and one for distance measurements.
If the labelled site is buried in the protein the accessibility to water and oxygen is low and accessibility
changes are expected to be below the detection limit. Such sites can be recognized by inspecting the
predicted favourable sites by protein and label visualization. In a future, more quantitative approach,
accessibility will be predicted by using a model for water and dioxygen distribution in the bilayer and
by considering explicitly that protein atoms exclude water and dioxygen.
If EPR measurements are performed with deuterated buffers, water-exposed spin labels feature much
longer transverse relaxation times T2 than lipid- or detergent-exposed labels [21]. As the reliability
and precision of measured distances increases with increasing T2 , pairs of water accessible sites with
moderately broad distance distributions may be more favourable than lipid-exposed pairs with narrower
distributions. Note also that one water-exposed label is sufficient to measure data with a long maximum
dipolar evolution time, even if the label at the other site has a short T2 .
7. Labelling statistics for two membrane proteins
Computational site scans may provide estimates on the variability of the conformational distribution
of spin labels in proteins and may identify motifs that lead to a well-defined position of the electron
spin. Here we discuss such scans for two membrane proteins (Fig. 2), the leucine transporter LeuTAa of
Aquifex aeolicus (PDB entry 2A65) [22] and the outer membrane protein A of Escherichia coli OmpA
(PDB entry 1QJP) [15]. The two structures were selected because of their relatively high and comparable
resolution of 1.65 Å. Furthermore, LeuTAa is a paradigm for sodium dependent secondary transporters
with an α-helical bundle structure and thus is related to PutP [10,22], while OmpA belongs to the second
main class of integral membrane proteins that have β-barrel structures. We assumed a temperature of
175 K corresponding to solidification of a lipid bilayer environment. Detergent molecules and water
molecules were removed from the structures, as it is assumed that a spin label could displace them.
The site scan of LeuTAa with MTSL predicts that out of the 509 residues resolved in the structure,
120 (24%) correspond to tight sites where spin labelling might fail or distort structure. For further
10 residues, significant population of only one rotamer is observed, whereas the partition function is
at least 0.1 (typically >0.4). A total of 52 WCSWC (about 10%) have σNO 0.2 nm, 48 of those are
situated in helices and only 4 in loops. Of these WCSWC in loops, only one is more than 2 residues
away from the next helix. The WCSWC in helices are distributed over nearly the whole structure, except
for transmembrane helices TM7, TM9 and TM12 and the helical domain of IL5. A particularly high
incidence of 11 WCSWC (33% of all residues) is seen in TM8. Seven of the WCSWC are very close
Fig. 2. Conformational distribution of MTSL constrained by side groups of other residues (rotamer library simulations). Green
ball and stick plots are allowed MTSL conformations, with the transparency encoding population. The N–O nitrogen and
oxygen atoms shown in blue and red, respectively. Red or reddish space-filling plots denote spatially close sidechains that
constrain MTSL. (A) Loose surface site A195R1 in transmembrane domain TM5 of LeuTAa (PDB identifier 2A65). (B) Tight
core site A95R1 in transmembrane domain TM5 of LeuTAa . (C) Tight core site F40R1 in OmpA. (The colors are visible in the
online version of the article; http://dx.doi.org/10.3233/SPE-2010-0490.)
to the substrate binding site and five more are residues that are conserved in related human transporters
[5]; labelling of these residues would certainly affect and possibly block transport.
For all WCSWC, without a single exception, we find that label conformations are restricted by side
groups of some residues that are remote in the sequence, yet spatially close. This indicates that bulky
neighbouring side groups in the context of an isolated helix are not sufficient to strongly constrain MTSL
conformations. Hence, WCSWC cannot be predicted from sequence and secondary structure information
alone. Furthermore, the WCSWC are almost randomly distributed among the types of original residues,
except perhaps for a preference for Phe sites (11 out of 50, 22%) and a rejection of Pro sites (1 out of
24, 4%) and Gly sites (2 out of 45, 4%).
Among the 137 potential sites in OmpA, 17 correspond to tight sites where spin labelling might fail or
distort structure (12%). The lower incidence of such sites compared to LeuTAa may be due to the larger
fraction of surface sites in a β-barrel compared to an α-helical bundle structure. Likewise, the incidence
of WSWC is lower (8 WCSWC, 6%). All WCSWC are found in strands, whereas in the whole protein
30 residues (22%) are situated in loops. The spin-labelled side groups of all WCSWC are oriented
towards the inner of the barrel. Note however that at site F123R1, where the spin label is pointing to the
outside, spatial distribution of the electron spin is only slightly broader with σNO = 0.21 nm, although
as many as 31 rotamers contribute significantly to the distribution at this site.
Consider now the case where no high-resolution structure is available. If a coarse-grained model of the
backbone exists, sidechains can be attached by the SCWRL4 [14] program. We have tested this approach
by performing a site scan for a decoy structure that was obtained from only the backbone coordinates
of LeuTAa by attaching the side chains with SCWRL4. This test also reveals sensitivities of the rotamer
library approach to differences in the conformation of side chains in the vicinity of spin labels.
When comparing the predicted mean coordinates of the N–O midpoints in the site scans on the LeuTAa
X-ray structure and on the structure with SCWRL4 predicted side groups, we find that the deviation is
smaller than 0.1 nm for 268 residues (53%) and smaller than 0.2 nm for 364 residues (72%). For σNO
we find a deviation smaller than 0.1 nm for 455 residues (89%) and smaller than 0.2 nm for 487 residues
(96%).
Larger deviations than 0.2 nm occur mainly for tight sites where significant population is predicted for
only one or two rotamers. As an example we discuss residue 382 for which the maximum deviation of the
mean N–O positions of 0.62 nm and the σNO values deviate by 0.72 nm (Fig. 3). With the LeuTAa X-ray
Fig. 3. Tight site (T382R1) where strongly different rotamer distributions are predicted from the X-ray structure of LeuTAa
(blue stick plots for side groups) and from a structure where all side groups were remodelled by SCWRL4 (red stick plots).
Only one MTSL rotamer (green) is significantly populated in the X-ray structure, while a second one (yellow) appears with the
SCWRL4 prediction. The reason is a different conformation for residue Met 363 (arrow). (The colors are visible in the online
version of the article; http://dx.doi.org/10.3233/SPE-2010-0490.)
structure, significant population of only one rotamer, shown as a green stick plot, is predicted. With
side group conformations predicted by SCWRL4, a second rotamer in a completely different position
is significantly populated (yellow stick plot). This can be traced back to a different prediction for the
conformation of residue Met 363 by SCWRL4 (arrow) which relieves a clash that depopulates this
MTSL rotamer in the X-ray structure. It is likely that the side group of Met 363 would flip on labelling,
i.e., that the prediction from the X-ray structure underestimates the accessible conformational space for
the label in this case.
8. Conclusion
Computational site scans based on the rotamer library approach [13] provide an approximate fast
prediction of the conformational distribution of spin labels. Very tight positions, where labelling is likely
to fail or distort structure, can be revealed and figures of merit of a labelling site can be defined for
different EPR experiments. Such figures of merit should always be critically examined with the intended
experiment in mind, using visualization of the spin label rotamers and their structural context. The
assumption of fixed positions of the side groups of neighbouring residues may lead to an underestimate
of conformational flexibility for tight sites. Work on an approach that allows for side group repacking is
now in progress.
It is expected that systematic computational site scans before mutant engineering and chemical la-
belling will lead to a decrease in the expense of SDSL EPR studies and improve the information content
and reliability of the results, in particular for distance measurements. To facilitate such site scans, the
rotamer library and site scan approaches were implemented as modules of the Matlab-based protein
visualization and modelling program MMM (Multiscale Modelling of Macromolecular Systems). This
open-source program can be downloaded from http://www.epr.ethz.ch/software/index.
Acknowledgement
Financial support by Swiss National Science Foundation is gratefully acknowledged.
References
[1] C. Altenbach, D.A. Greenhalgh, H.G. Khorana and W.L. Hubbel, Proc. Natl. Acad. Sci. 91 (1994), 1667–1671.
[2] M. Appel, D. Hizlan, K.R. Vinothkumar, C. Ziegler and W. Kuhlbrandt, J. Mol. Biol. 386 (2009), 351–365.
[3] P.P. Borbat, H.S. Mchaourab and J.H. Freed, J. Am. Chem Soc. 124 (2002), 5304–5314.
[4] A.A. Canutescu, A.A. Shelenkov and R.L. Dunbrack Jr., Protein Sci. 12 (2003), 2001–2014.
[5] R.E. Coffman and A. Pezeshk, J. Magn. Reson. 70 (1986), 21–33.
[6] C. Dockter, A. Volkov, C. Bauer, Y. Polyhach, Z. Joly-Lopez, G. Jeschke and H. Paulsen, Proc. Natl. Acad. Sci. USA 106
(2009), 18485–18490.
[7] S.S. Eaton, K.M. More, B.M. Sawant and G.R. Eaton, J. Am. Chem. Soc. 105 (1983), 6560–6567.
[8] M. Grote, E. Bordignon, Y. Polyhach, G. Jeschke, H.J. Steinhoff and E. Schneider, Biophys. J. 95 (2008), 2924–2938.
[9] K. Herz, A. Rimon, G. Jeschke and E. Padan, J. Biol. Chem. 284 (2009), 6337–6347.
[10] D. Hilger, Y. Polyhach, H. Jung and G. Jeschke, Biophys. J. 96 (2009), 217–225.
[11] D. Hilger, Y. Polyhach, E. Padan, H. Jung and G. Jeschke, Biophys. J. 93 (2007), 3675–3683.
[12] W.L. Hubbell, D.S. Cafiso and C. Altenbach, Nat. Struct. Biol. 7 (2000), 735–739.
[13] G. Jeschke and Y. Polyhach, Phys. Chem. Chem. Phys. 9 (2007), 1895–1910.
[14] G.G. Krivov, M.V. Shapovalov and R.L. Dunbrack Jr., Proteins 77 (2009), 778–795.
[15] A. Pautsch and G.E. Schulz, J. Mol. Biol. 298 (2000), 273–282.
[16] K. Sale, L. Song, Y.-S. Liu, E. Perozo and P. Fajer, J. Am. Chem. Soc. 127 (2005), 9334–9335.
[17] O. Schiemann and T.F. Prisner, Quater. Rev. Biophys. 40 (2007), 1–53.
[18] D. Sezer, J.H. Freed and B. Roux, J. Phys. Chem. B 112 (2008), 5755–5767.
[19] F. Tombolato, A. Ferrarini and J.H. Freed, J. Phys. Chem. B 110 (2006), 26248–26259.
[20] Y.D. Tsvetkov, A.D. Milov and A.G. Maryasov, Usp. Khim. 77 (2008), 515–550.
[21] A. Volkov, C. Dockter, T. Bund, H. Paulsen and G. Jeschke, Biophys. J. 96 (2009), 1124–1141.
[22] A. Yamashita, S.K. Singh, T. Kawate, Y. Jin and E. Gouaux, Nature 437 (2005), 215–223.
Spectroscopy 24 (2010) 593–599 593
DOI 10.3233/SPE-2010-0489
IOS Press
Biomedical application of Mössbauer

spectroscopy with a high velocity resolution:
Revealing of small variations
M.I. Oshtrakh a,∗ , V.A. Semionkin a,b , O.B. Milder a,c , I.V. Alenkina a,b and E.G. Novikov a,b
a
Faculty of Physical Techniques and Devices for Quality Control, Physical–Technical Department,
Ural Federal University, Ekaterinburg, Russian Federation
b
Faculty of Experimental Physics, Physical–Technical Department, Ural Federal University,
Ekaterinburg, Russian Federation
c
Radio–Technical Department, Ural Federal University, Ekaterinburg, Russian Federation
Abstract. Application of Mössbauer spectroscopy with a high velocity resolution for study different hemoglobins, ferritin,
its models and chicken liver and spleen as well as normal and lymphoid chicken spleen demonstrated revealing of small
variations of hyperfine parameters related to small variations of iron stereochemistry in biomolecules. These data demonstrate
that Mössbauer spectroscopy with a high velocity resolution may be useful in biomedical research to distinguish small variations
of iron-containing proteins in normal and pathological cases.
Keywords: Mössbauer spectroscopy with a high velocity resolution, biomedical applications, small variations
1. Introduction
Mössbauer spectroscopy is a powerful tool which allows us to observe the hyperfine splitting of the
nuclear energy levels as well as changes of energies of the ground and excited states of Mössbauer nuclei
(for instance, 57 Fe, 119 Sn, 197 Au and some other) in the absorption or emission spectrum of γ-rays (de-
tailed description of the Mössbauer effect see, for instance, in [1–3,16]). Iron is one of the most vitally
important metals. A number of proteins contain iron as an active site. Some proteins are responsible for
the iron transport and storage. The functional properties of these proteins are determined in part by the
iron electronic structure and stereochemistry. Therefore, some functional variations of proteins related
to the structural heterogeneity may be reflected by the small changes of the iron electronic structure.
It is well known that several diseases, so-called molecular diseases, are caused or accompanied by the
synthesis of anomalous biomolecules or any other protein biosynthesis disturbance. Some pathological
states of the body are caused by environmental factors that may affect biological molecules, cells and tis-
sues. Therefore, Mössbauer spectroscopy was successfully applied for studying biomolecules containing
iron in normal and pathological cases [4,7,10,11].
*
Corresponding author: M.I. Oshtrakh, Faculty of Physical Techniques and Devices for Quality Control, Physical–Technical
Department, Ural Federal University, Ekaterinburg, Russian Federation. E-mail: oshtrakh@mail.utnet.ru.
594 M.I. Oshtrakh et al. / Biomedical application of Mössbauer spectroscopy with a high velocity resolution
Mössbauer spectroscopy appeared to be useful for studying so-called “drastic” and “small” changes
of the iron electronic structure [6,8,9]. “Drastic” changes imply a change of the valence and/or spin state
of iron resulting from the transformation or destruction of proteins. In this case Mössbauer spectra or
subspectra may be well defined. “Small” changes imply a change of the iron electronic structure without
any change of the valence and spin state resulting from structural modifications of protein. The study
of “small” changes is more complicated and requires a high precision, sensitive and stable Mössbauer
spectrometer. These structural modifications may be related to protein heterogeneity and functional va-
riety in different organs, in human and various animals as well as in normal and pathological subjects
due to their specific structure–function relationship. Small stereochemical variations of iron neighboring
atoms may change energies of the ground and low-lying iron electronic terms and, therefore, change
electric field gradient (EFG) on the 57 Fe. Quadrupole splitting is Mössbauer parameter which depends
on the EFG. Therefore, small variations of iron stereochemistry can be revealed by detecting small
variations of quadrupole splitting in Mössbauer spectra. However, all previous Mössbauer studies were
performed with a low velocity resolution (in 256 or 512 channels) that limited possibilities to reveal
small variations of parameters outside the experimental error (real experimental error for Mössbauer hy-
perfine parameters is ±1 channel in mm/s if calculated error did not exceed experimental one). Recently
new possibilities of Mössbauer spectroscopy with a high velocity resolution (1024–4096 channels) were
shown [12–14]. Therefore, in this work we demonstrate these possibilities in revealing small variations
in various iron-containing species in recent biomedical research.
2. Experimental
Human adult hemoglobin and rabbit hemoglobin in concentrated solutions were obtained from the
Hematological Research Center (Moscow, Russian Federation). Hemoglobin samples were oxygenated
and placed into a special holder with about 2.5 ml of oxyhemoglobin solution. Then samples were im-
mediately frozen with liquid nitrogen and stored in liquid nitrogen. Human liver ferritin in lyophilized
form was obtained from the Russian State Medical University (Moscow, Russian Federation). We used
100 mg of protein for one sample. Samples of washed and lyophilized chicken liver and spleen in nor-
mal case and chicken spleen during lymphoid leukemia were obtained from the Ural State Agricultural
Academy (Ekaterinburg, Russian Federation). It is well known that liver and spleen tissues contain a
large amount of ferritin molecules. We used 1200–1600 mg of lyophilized tissues for one sample. We
also studied industrial samples of ferritin models used for treatment of iron deficiency such as Imferon
(Fisons, UK) in lyophilized form and Maltofer® (Vifor, Inc., Switzerland) tablet. These samples were
prepared with effective thickness about 5 and 10 mg Fe/cm2 , respectively.
Mössbauer spectra were measured using new Mössbauer spectrometric system on the basis of high
precision, sensitive and stable spectrometer SM-2201 with a high velocity resolution (measurements
in 4096 channels) and temperature variable liquid nitrogen cryostat with moving absorber. Detailed
description and characteristics of the system were given in [15]. The ∼1.85 × 109 Bq 57 Co(Cr) source
was used at room temperature. Mössbauer spectra of hemoglobin samples were measured at 90 K, while
spectra of other samples were measured at 295 K. Spectra of samples with high iron content were
measured with a good statistics for presentation in 2048 channels while spectra of samples with low
iron content were measured with statistics for presentation in 1024 channels. Additionally, all measured
spectra were presented with a low velocity resolution in 512 channels for comparison. Spectra were
computer fitted with the least square procedure using UNIVEM-MS program with Lorentzian line shape.
M.I. Oshtrakh et al. / Biomedical application of Mössbauer spectroscopy with a high velocity resolution 595
Mössbauer parameters isomer shift δ, quadrupole splitting ΔEQ , line width Γ, subspectrum area S and
statistical criterion χ2 were determined. Experimental error for determination of the each spectra point
was ±0.5 channel, experimental errors for hyperfine parameters were ±1 channel while that for Γ was
±2 channels. It should be noted that spectrometer characteristics determined an integral velocity error
which was several times lower than a half of channel value in mm/s during spectra measurements using
4096 channels. The values of isomer shift are given relative to α-Fe at 295 K.
The simplest way to demonstrate possibilities of Mössbauer spectroscopy in revealing of small vari-
ations of the hyperfine parameters may be performed using the simplest fitting of measured spectra. In
this case we used one quadrupole doublet fit for all measured spectra. Mössbauer spectra of human adult
and rabbit oxyhemoglobins measured at 90 K and presented in 1024 channels are shown in Fig. 1(a)
and (b). These spectra are similar quadrupole doublets like previous spectra of oxyhemoglobins [5].
Mössbauer hyperfine parameters for the spectra of both oxyhemoglobins presented in 512 channels and
in 1024 channels are shown in the plot of quadrupole splitting and isomer shift (Fig. 1(c)). It is clearly
seen that ΔEQ values for human adult and rabbit oxyhemoglobins distinguish better for the high velocity
resolution spectra. An increase of ΔEQ for rabbit oxyhemoglobin indicates that EFG on the 57 Fe in this
protein also increases. This may be a result of small changes of the energies of low-lying iron electron
terms due to small stereochemical variations. For instance, it is possible some changes in the Fe(II)–O2
bond in rabbit oxyhemoglobin in comparison with that for human adult oxyhemoglobin.
Mössbauer spectra of Imferon, Maltofer® and human liver ferritin measured at room temperature and
presented in 2048 channels are shown in Fig. 2(a)–(c). These spectra demonstrated similar quadrupole
doublets. Differences in absorption effect are related to differences in the effective thickness of these
samples. Mössbauer hyperfine parameters are shown in Fig. 2(d) in the plot of quadrupole splitting and
isomer shift. In this case improvement in velocity resolution leads to revealing small differences of both
ΔEQ and δ. It is well known that the iron cores in ferritins are in the form of ferrihidrite while those in
Imferon and Maltofer® are in the form of β-FeOOH (both are different modifications of hydrous ferric
oxide). An increase of ΔEQ values is related to increase of EFG on the 57 Fe in Imferon and Maltofer® .
A decrease of δ values is related to decrease of the electron density on the 57 Fe. This may be a result
of stereochemical variations of Fe–O bonds such as changes of angles and distances. In fact, these
results demonstrated as rough approximation revealed small structural differences in the iron cores of
human liver ferritin, Imferon and Maltofer® which are related to small variations of Mössbauer hyperfine
parameters.
Mössbauer spectra of chicken tissues demonstrated similar quadrupole doublets; however, absorp-
tion effect was significantly less than for extracted human liver ferritin. Therefore, these spectra were
presented in 1024 channels. Comparison of Mössbauer hyperfine parameters for human liver ferritin,
normal chicken liver and spleen is shown in Fig. 3(a). It is clearly seen that ΔEQ values differ for all
samples while δ value for chicken spleen is slightly higher than that for other samples in the case of a
higher velocity resolution. These data demonstrate structural differences in the iron cores of different
ferritins in addition to different amino-acid and protein subunits compositions. Comparison of Möss-
bauer hyperfine parameters for normal and lymphoid leukemia chicken spleens is shown in Fig. 3(b). It
was interesting to observe small variations of ΔEQ and δ values in case of a high velocity resolution. An
increase of ΔEQ means an increase of EFG on the 57 Fe as a result of symmetry distortion and a weaker
(a) (b)
(c)
Fig. 1. Mössbauer spectra of human adult (a) and rabbit (b) oxyhemoglobins presented in 1024 channels and small variations
of their hyperfine parameters revealed with a high velocity resolution in comparison with a low velocity resolution (c): human
adult oxyhemoglobin – P (high velocity resolution), Q (low velocity resolution) and rabbit oxyhemoglobin – E (high velocity
resolution), F (low velocity resolution). Experimental errors decreased twice for 1024-channel spectra in comparison with
512-channel ones. T = 90 K.
Fe–O bond in the iron core of iron storage proteins in lymphoid leukemia chicken spleen in comparison
with that in normal chicken spleen. A weaker Fe–O bond may be a result of small increase of Fe–O
distance which leads to decrease of the electron density on the 57 Fe and slightly less δ value.
4. Conclusion
The results of Mössbauer study of some iron-containing biomolecules and model compounds demon-
strated new possibilities to reveal small variations of hyperfine parameters in the case of improving
velocity resolution. Small differences of Mössbauer hyperfine parameters for different oxyhemoglobins,
(a) (b)
(c) (d)
Fig. 2. Mössbauer spectra of Imferon (a), Maltofer® (b) and human liver ferritin (c) presented in 2048 channels and small
variations of their hyperfine parameters revealed with a high velocity resolution in comparison with a low velocity resolution (d):
Imferon – 1 (high velocity resolution), 2 (low velocity resolution); Maltofer® – P (high velocity resolution), Q (low velocity
resolution); human liver ferritin – E (high velocity resolution), F (low velocity resolution). Experimental errors decreased
four times for 2048-channel spectra in comparison with 512-channel ones. T = 295 K.
for human liver ferritin, its model compounds and chicken liver and spleen as well as for normal and
lymphoid leukemia chicken spleen indicated that these differences were sensitive for structural varia-
tions in the iron stereochemistry. Therefore, biomedical applications of Mössbauer spectroscopy with a
high velocity resolution may be useful for revealing small variations in iron-containing biomolecules in
(a) (b)
Fig. 3. Variations of Mössbauer hyperfine parameters for human liver ferritin (! – high velocity resolution, " – low velocity
resolution), chicken liver (P – high velocity resolution, Q – low velocity resolution) and chicken spleen (E – high velocity
resolution, F – low velocity resolution) (a) and for normal chicken spleen (E – high velocity resolution, F – low velocity res-
olution) and for lymphoid leukemia chicken spleen (1 – high velocity resolution, 2 – low velocity resolution) (b). Experimental
errors decreased twice for 1024-channel spectra in comparison with 512-channel ones. T = 295 K.
case of molecular pathology. In this case it would be possible to develop diagnostic tests on the basis of
Mössbauer hyperfine parameters.
Acknowledgements
The authors thank Dr. A.L. Berkovsky for hemoglobin samples preparation, Prof. P.G. Prokopenko
for the gift of human liver ferritin and Dr. L.I. Malakheeva for preparation of chicken tissues. This work
was supported in part by the Russian Foundation for Basic Research (Grant # 09-02-00055-a).
References
[1] H. Frauenfelder, The Mössbauer Effect, W.A. Benjamin, New York, 1963.
[2] V.I. Goldanskii and R.H. Herber (eds), Chemical Applications of Mössbauer Spectroscopy, Academic Press, New York
and London, 1968.
[3] P. Gütlich, R. Link and A. Trautwein, Mössbauer Spectroscopy and Transition Metal Chemistry, Springer-Verlag, Berlin–
Heidelberg–New York, 1978.
[4] M.I. Oshtrakh, Biomedical applications of the Mössbauer effect, Hyperfine Interact. 66 (1991), 127–140.
[5] M.I. Oshtrakh, Comparison of human oxyhemoglobin in lyophilized form, red blood cells, and concentrated solution: the
features of Mössbauer spectra and heme iron stereochemistry, J. Inorg. Biochem. 56 (1994), 221–231.
[6] M.I. Oshtrakh, Mössbauer effect in biomedical research: variations of quadrupole splitting in relation to the qualitative
changes of biomolecules, Z. Naturforsch. 51a (1996), 381–388.
[7] M.I. Oshtrakh, Mössbauer spectroscopy of iron containing biomolecules and model compounds in biomedical research,
J. Mol. Struct. 480,481 (1999), 109–120.
[8] M.I. Oshtrakh, Study of the relationship of small variations of the molecular structure and the iron state in iron containing
proteins by Mössbauer spectroscopy: biomedical approach, Spectrochim. Acta – Part A Mol. Biomol. Spectrosc. 60 (2004),
217–234.
[9] M.I. Oshtrakh, The relationship of Mössbauer hyperfine parameters and structural variations of iron containing proteins
and model compounds in biomedical research, Hyperfine Interact. 159 (2004), 337–343.
[10] M.I. Oshtrakh, Mössbauer spectroscopy: application in biomedical research, Hyperfine Interact. 165 (2005), 313–320.
[11] M.I. Oshtrakh, Biomedical applications of Mössbauer spectroscopy, J. Radioanal. Nucl. Chem. 269 (2006), 407–415.
[12] M.I. Oshtrakh, V.A. Semionkin, V.I. Grokhovsky, O.B. Milder and E.G. Novikov, Mössbauer spectroscopy with high
velocity resolution: new possibilities of chemical analysis in material science and biomedical research, J. Radioanal.
Nucl. Chem. 279 (2009), 833–846.
[13] M.I. Oshtrakh, V.A. Semionkin, O.B. Milder and E.G. Novikov, Biomedical application of Mössbauer spectroscopy with
high velocity resolution: preliminary results, in: Proceedings of the International Conference “Mössbauer Spectroscopy
in Materials Science 2008”, Vol. 1070, M. Mashlan and R. Zboril, eds, AIP Conference Proceedings, Melville, NY, 2008,
pp. 122–130.
[14] M.I. Oshtrakh, V.A. Semionkin, O.B. Milder and E.G. Novikov, Mössbauer spectroscopy with high velocity resolution:
new possibilities in biomedical research, J. Mol. Struct. 924–926 (2009), 20–26.
[15] M.I. Oshtrakh, V.A. Semionkin, O.B. Milder and E.G. Novikov, Mössbauer spectroscopy with high velocity resolution:
an increase of analytical possibilities in biomedical research, J. Radioanal. Nucl. Chem. 281 (2009), 63–67.
[16] A. Vertes, L. Korecz and K. Burger, Mössbauer Spectroscopy, Akadémiai Kiadó, Budapest, 1979.
Spectroscopy 24 (2010) 641–649 641
DOI 10.3233/SPE-2010-0488
IOS Press
A validated direct spectrofluorimetric method

for quantification of mirtazapine in human
whole blood
Balraj Saini, Manjula Kaushal and Gulshan Bansal ∗
Department of Pharmaceutical Sciences and Drug Research, Punjabi University, Patiala, India
Abstract. A spectrofluorimetric method for estimation of mirtazapine in human whole blood was developed and validated. The
recovery efficiency of the processing method was 95–98%. The analytical method was linear over drug concentration of 10–
200 ng/ml. The limit of quantification was 10 ng/ml. The method was precise with %RSD for intra-day and inter-day precision
being <3.0 and 1.5, respectively. Excellent recoveries (97.87–99.69%) were achieved during accuracy studies. The method was
robust to small changes in processing method and instrumental parameters. The present method can be employed for direct
fluorimetric determination of mirtazapine in human whole blood during clinical studies.
Keywords: Mirtazapine, spectrofluorimetry, protein precipitation, human blood, validated
1. Introduction
Mirtazapine is a piperazinoazepine-based tetracyclic compound which is used as an antidepressant

in moderate to severe depression. It is classified as an adrenergic and specific serotonergic antide-
pressant [21,24]. Chemically it is 1,2,3,4,10,14b-hexahydro-2-methylpyrazino[2,1-a]pyrido[2,3-c]-2-
benzazepine which exists as a mixture of (−) and (+) enantiomers (Fig. 1) and both have similar phar-
macological activity [1]. It acts by facilitating central serotonergic and nor-adrenergic transmission and
antagonizes postsynaptic 5-HT2A , 5-HT3 and H1 receptors [8]. It is extensively metabolized in liver by
cytochrome P450 isoenzymes to demethylated and hydroxylated metabolites which are pharmacologi-
cally active [1].
Numerous sophisticated analytical methods like HPLC [4,5,15–17,19,23], LC-MS [2,3,6,7,11,18,25],
GC-MS [9,20,22], capillary electrophoresis [13] and enantioselective electrodriven method [14] are
available in literature for quantification of mirtazapine in biological fluids, formulations and in the
presence of other drugs. However, a simple, fast and reliable analytical method has always remained
a method of choice for quantification of drug in biological matrixes to handle a large number of sam-
ples during clinical studies. Further, analysis of a drug in biological matrix is a two stage process, i.e.,
*
Corresponding author: Gulshan Bansal, Department of Pharmaceutical Sciences and Drug Research, Punjabi University,
Patiala, 147001 India. Tel.: +91 175 3046255; Fax: +91 175 2283073; E-mail: gulshanbansal@rediffmail.com.
642 B. Saini et al. / A validated direct spectrofluorimetric method for quantification of mirtazapine
Fig. 1. Enantiomers of mirtazapine.
recovery of drug from the matrix by liquid–liquid extraction, protein precipitation or solid phase mi-
croextraction followed by quantitation of the recovered drug. Various methods for recovery of mirtazap-
ine from different biological matrixes are reported in literature [2–6,9,11,15–19,22,23]. Labat et al. [12]
have reported a spectrofluorimetric method for determination of mirtazapine in tablets. Very recently,
Youssef [26] has also developed a first derivative spectrofluorimetric method for mirtazapine in human
plasma and tablets. However, there is no report of any spectrophometric or spectrofluorimetric method
for quantification of mirtazapine in human whole blood.
Hence, the present study is designed to develop a simple and an efficient method for recovery of
mirtazapine from human blood and to develop and validate a direct and sensitive spectrofluorimetric
method for determination of the recovered drug. In comparison to various sophisticated methods for
quantification of mirtazapine, the present method is capable of quantifying mirtazapine in human whole
blood with excellent accuracy and sensitivity.
2. Experimental
2.1. Materials
Mirtazapine was supplied by Panacea Biotec Ltd. (Lalru, India) as a gift sample. All chemicals used in
the study were of analytical grade. Methanol was procured commercially from Loba Chemie (Mumbai,
India). EDTA, orthophosphoric acid and acetonitrile were procured from s.d. Fine-Chemical Ltd. (Mum-
bai, India). The solutions and reagents were prepared in double distilled water. The blood (heparinized)
was procured from blood bank of Government Medical College and Hospital, Patiala (India).
2.2. Instrumentation
Analysis was performed on spectrofluorimeter fitted with xenon lamp (Model SLI74, Elico, Hyder-
abad, India). Data was processed on Fluorosoft version 2.1 software. The bandwidths and sensitivity
of the spectrofluorimeter was set at 10 and 540, respectively. A cold centrifuge (Perfit, Ambala, India)
was used for centrifugation. Vortex shaker (Popular, Ambala, India) was used for intimate mixing of
solutions. Borosilicate glassware (A class) was used through out the study.
B. Saini et al. / A validated direct spectrofluorimetric method for quantification of mirtazapine 643
2.3. Methods
2.3.1. Optimization of instrumental parameters

A stock solution of mirtazapine (1 mg/ml) in methanol was serially diluted with 0.05 mol/ml or-
thophosphoric acid solution to obtain standard drug solutions having concentrations 1, 2, 5, 10, 20, 50,
100 and 200 ng/ml. The excitation spectra of the standard solutions were recorded and wavelength of
322 nm was selected as excitation wavelength (λex ). The emission spectra of the same standard solu-
tions were recorded taking 322 nm as λex and wavelength of 405 nm was selected as emission wave-
length (λem ). Subsequently, the standard solutions were analyzed on the different excitation and emission
bandwidths (5 and 10) and sensitivities (540, 650 and 750) to select the optimum excitation and emission
bandwidths and sensitivity.
2.3.2. Processing method for recovery of mirtazapine from blood
2.5 ml of acetonitrile was added to 1 ml of human blood initially spiked with 1 ml of drug solution
(100 ng/ml) and mixture was vortexed for 2 min. Methanol (0.5 ml) was added to the mixed contents
and vortexed for 2 min. The contents were centrifuged at 3200 rpm for 10 min in cold centrifuge and
supernatant was read on spectrofluorimeter. The blank was prepared by replacing drug solution with
diluant. Subsequently, amounts of acetonitrile (1.5, 2.0 and 3.0 ml) and methanol (1.0 and 1.5 ml) were
varied to maximize recovery efficiency of the method. The interference due to anticoagulants (EDTA and
heparin) was evaluated by comparing fluorescence intensity of the drug solution spiked in blood (without
as well as with each anticoagulant) with that of the standard solution. The optimized processing method
was finally applied to all standard solutions.
2.3.3. Analytical method validation
The optimized method was validated by evaluating linearity, precision, accuracy and ruggedness in
accordance with the ICH guidelines [10]. Linearity was evaluated using three sets of calibration solu-
tions. The set I comprised standard drug solutions in the concentration range of 10–200 ng/ml. For set II,
each standard solution in the concentration range of 50–1000 ng/ml was spiked in 1 ml of water and sub-
jected to the optimized processing method so that final concentration in the supernatant was in the range
of 10–200 ng/ml, respectively. The set III was prepared similar to set II where water was replaced by
blood. Each set of calibration solutions was prepared and analyzed six times. The data was analyzed by
STATISTICA and Graph Pad softwares. For intra-day precision, three drug concentrations (50, 250 and
1000 ng/ml) were spiked in water as well as blood to obtain the final drug concentrations of 10, 50 and
200 ng/ml, respectively, and each was analyzed six times on the same day. For inter-day precision, the
same drug concentrations were analyzed on three different days. Accuracy was evaluated by calculating
drug concentrations in the fortified blood samples vis-à-vis unfortified blood sample. The unfortified
sample was prepared by spiking blood (1 ml) with equal volume of standard drug solution (100 ng/ml)
and mixing with 1 ml of diluant so that after application of processing method the drug concentration
in the supernatant was 16.7 ng/ml. The fortified samples were prepared by replacing the diluant in un-
fortified sample with each of the three concentrations (100, 200 and 500 ng/ml) so that after subjecting
to processing method, the drug concentrations were fortified by 16.7, 33.3 and 83.3 ng/ml, respectively.
These samples were prepared in triplicate. The accuracy was expressed as percent recovery obtained in
fortified samples with respect to unfortified one. Robustness of the processing method was determined
by evaluating the influence of deliberate but small changes in the optimized processing method and
excitation wavelength on percent recovery of the drug from blood spiked with drug concentration of
50 ng/ml.
The excitation spectrum of standard solution of mirtazapine showed maximum absorbance at 322 nm
which is very close to the reported λex of 328 nm [12]. The emission spectrum of the same standard
solution excited at 322 nm, showed maximum emission at 405 nm which is also very close to the re-
ported λem = 415 nm (Fig. 2). Further, the absorbance at 322 nm and subsequent emission at 405 nm
were found to increase linearly with concentration. Hence, 322 and 405 nm were selected as λex and
λem for the study. A 0.05 mol/ml orthophosphoric acid solution was used as diluant because it showed
insignificant absorbance at 310–330 nm and no emission above 370 nm (Fig. 2). The linear equations
obtained after analyzing standard solutions (1–200 ng/ml) at varied combinations of excitation and emis-
sion bandwidths (Table 1) revealed that minimum intercept and maximum correlation coefficient (r2 )
were observed at bandwidths of 10. Further, based on the linear equations obtained at varied sensitivity
levels, the spectrofluorimeter was set at sensitivity of 540 for further analyses.
3.1. Processing method
It is the most critical part of analytical method development process as it governs recovery efficiency
of the drug from biological fluids and, hence, sensitivity of the analytical method. A spectrofluorimetric
method for determination of mirtazapine in human plasma is available in literature [26] but its validation
is not clearly addressed. Determination of a drug directly in whole blood is although more challenging
Fig. 2. Excitation (Ex) and emission (Em) spectra of 0.050 mol/ml orthophosphoric acid solution (1) and standard solution of
mirtazapine (2).
Table 1
Effect of excitation (Ex) and emission (Em) bandwidths and sensitivity on linearity
Instrument parameters Linear equations
Bandwidth (Ex/Em)
5/5 Y = 3350.74 · X + 23.61; r2 = 0.9980
10/5 Y = 3946.13 · X + 17.36; r2 = 0.9982
10/10 Y = 2826.49 · X + 5.17; r2 = 0.9995
Sensitivity (at 10/10 bandwidths)
540 Y = 2670.41 · X + 8.24; r2 = 0.9992
650 Y = 3453.71 · X + 2.25; r2 = 0.9974
750 Y = 3445.65 · X + 34.51; r2 = 0.9986
800 Y = 5525.15 · X + 58.66; r2 = 0.9967
than in any other biological fluid but it increases the speed of analysis particularly in clinical studies.
Use of varied extractive organic solvents like toluene and methyl tert-butyl ether provided insignificant
recovery which may be attributed to interference by various proteins in blood that were also possibly
extracted by the solvent. Hence, precipitation of the blood proteins was sought in order to improve re-
covery of the drug. A 5% w/v solution of trichloroacetic acid (TCA) resulted in just 5% recovery which
could be attributed to either insufficient protein precipitation or ionization of the drug in acidic pH of
TCA solution. Replacement of TCA with acetonitrile as protein precipitant resulted in 65% recovery.
Addition of methanol after precipitation with acetonitrile increased the recovery to 95–98%. This sig-
nificant increase in recovery may be attributed to greater solubility of the drug in methanol than in
acetonitrile. Further, use of varied amounts of acetonitrile and methanol to maximize the recovery ef-
ficiency revealed that decrease in acetonitrile decreased the recovery while increase in its amount did
not improve the recovery. Increase in methanol also did not improve the recovery. The calibration lines
(n = 3) plotted over a concentration range of 1–200 ng/ml to verify utility of the optimized processing
method indicated that relationship between drug concentration and fluorescence intensity was linear in
the concentration range of 10–200 ng/ml and the three equations were in well agreement with each other
(Y = 3058.65 ± 14.25 · X + 14.28 ± 0.61; r2 = 0.9981 ± 0.0008) which suggested that the processing
method was sufficiently reproducible over a wide concentration range. EDTA was found to decrease
fluorescence intensity of the drug while no interference was observed with heparin. Hence, heparin was
used as anticoagulant during the study to prevent blood coagulation.
3.2. Method validation
The method was significantly linear for the drug in standard solutions (set I), drug spiked in water
(set II) and in blood in the concentration range of 10–200 ng/ml with standard error of estimate 8.07,
3.08 and 9.64, respectively (Table 2). The r2 was more than 0.9988 and p-value was less than 0.0001 for
all three sets. The randomness in residuals along the linear concentration for each set (Fig. 3) revealed
that the observed data fitted well into the theoretical data. The %RSD of intra-day and inter-day precision
were less than 3.0 and 1.5, respectively (Table 2) confirming the method to be sufficiently precise.
Excellent recoveries (97.87–99.69%) were achieved from each fortified sample (Table 2) indicating the
method to be accurate. The limit of quantitation (LOQ) was found to be 10 ng/ml, which indicated
that the method is very sensitive for quantification of trace amounts of the drug. The method was also
found to be robust as insignificant change (97.58–101.42%) in recovery of the drug was observed upon
deliberate changes in λex , amount of methanol or acetonitrile and centrifugation speeds (Table 3).
Table 2
Linearity, precision and accuracy studies
Linearity
Calibration set Calibration equationa Sy · xa,b
(Y = Slope · X ± intercept; r2 )
I Y = 2767.52 ± 49.25 − 0.71 ± 1.56; 8.07 ± 1.71
0.9989 ± 0.0005
II Y = 2654.53 ± 21.47 + 1.54 ± 1.55; 3.08 ± 0.29
0.9998 ± 0.0001
III Y = 3127.28 ± 18.58 − 8.55 ± 1.08; 9.64 ± 0.86
0.9988 ± 0.0002
Precision
Actual concentration (ng/ml) Measured concentration (ng/ml)
Mean ± SD; %RSD
Intra-day (n = 6) Inter-day (n = 3)
10 10.02 ± 0.24; 2.39 10.16 ± 0.11; 1.08
50 49.17 ± 0.56; 1.14 49.34 ± 0.32; 0.65
200 199.38 ± 0.56; 2.81 199.51 ± 0.58; 0.29
Accuracy
Added concentration (ng/ml) Calculated concentration (ng/ml) Recovery (%)
Mean ± SD; %RSD (n = 3)
16.7 16.4 ± 0.23; 1.40 98.20
33.3 33.2 ± 0.57; 1.72 99.69
83.3 81.53 ± 0.57; 0.70 97.87
Notes: a All values given as mean ± SD; b standard error of estimate.
4. Conclusion
A simple protein precipitation method for efficient recovery of mirtazapine from human whole blood
has been developed and analyzed quantitatively by a direct spectrofluorimetric method, which is vali-
dated in accordance with ICH guidelines. The recovery efficiency of the processing method is 95–98%.
The method is linear in the concentration range of 10–200 ng/ml, sufficiently precise, accurate, sensitive
and robust. The present method is suggested to be better than the reported one for determination of the
drug in human plasma owing to ease and speed of analysis. It can be useful as a sensitive and rapid
method for determination of mirtazapine during clinical studies.
Acknowledgements
The authors are thankful to Panacea Biotec Ltd., Lalru (India) for providing the drug as a generous gift
sample. The authors are thankful to Head, Department of Pharmaceutical Sciences and Drug Research,
Punjabi University, Patiala (India) for providing the necessary infrastructure and grant for conduct of
the study. The authors are also thankful to Blood Bank of Government Medical College and Hospital,
Patiala (India) for providing the human whole blood.
Fig. 3. Residual plots of standard drug solutions (set I), drug solutions spiked and processed in water (set II) and in blood
(set III).
Table 3
Robustness studies
Deliberate change Recovery (%)
Optimum parameters 100.14
λex (±5 nm)
317 98.22
327 99.50
Volume of acetonitrile (±0.1 ml)
2.4 100.78
2.6 98.22
1st centrifugation speed (±400)
2800 100.14
3600 101.42
Volume of methanol (±0.1 ml)
0.4 97.58
0.6 98.22
2nd centrifugation speed (±400)
2800 98.86
3600 99.50
References
[1] M.E. Burton, L.M. Shaw, J.J. Schentag and W.E. Evans, Applied Pharmacokinetics & Pharmacodynamics: Principles of
Therapeutic Drug Monitoring, Lippincott Williams & Wilkins, Baltimore, MD, USA, 2006, pp. 791–792.
[2] N. Castaing, K. Titier, M. Receveur-Daurel, M. Le-Déodic, D. Le-bars, N. Moore and M. Molimard, Quantification of
eight new antidepressants and five of their active metabolites in whole blood by high-performance liquid chromatography-
tandem mass spectrometry, Journal of Analytical Toxicology 31 (2007), 334–341.
[3] F.J.M. De Santana and P.S. Bonato, Enantioselective analysis of mirtazapine and its two major metabolites in human
plasma by liquid chromatography–mass spectrometry after three-phase liquid-phase microextraction, Analytica Chimica
Acta 606 (2008), 80–91.
[4] F.J.M. De Santana, E.J. Cesarino and P.S. Bonato, New method for the chiral evaluation of mirtazapine in human plasma
by liquid chromatography, Journal of Chromatography B 809 (2004), 351–356.
[5] S. Dodd, G.D. Burrows and T.R. Norman, Chiral determination of mirtazapine in human blood plasma by high-
performance liquid chromatography, Journal of Chromatography B 748 (2000), 439–443.
[6] B. Doherty, V. Rodriguez, J.C. Leslie, S. McClean and W.F. Smyth, An electrospray ionisation tandem mass spectrometric
investigation of selected psychoactive pharmaceuticals and its application in drug and metabolite profiling by liquid chro-
matography/electrospray ionisation tandem mass spectrometry, Rapid Communications in Mass Spectrometry 21 (2007),
2031–2038.
[7] T. Gunnar, S. Mykkänen, K. Ariniemi and P. Lillsunde, Validated semiquantitative/quantitative screening of 51 drugs in
whole blood as silylated derivatives by gas chromatography–selected ion monitoring mass spectrometry and gas chro-
matography electron capture detection, Journal of Chromatography B 806 (2004), 205–219.
[8] R.E. Hales and S.C. Yudofsky, Textbook of Clinical Psychiatry, American Psychiatric Publishing, Inc., Arlington, VA,
USA, 2003, pp. 1063–1065.
[9] X. Hong, Y. Yao, S. Hong and C. Lei, LC–MS–MS analysis of mirtazapine in plasma, and determination of pharmacoki-
netic data for rats, Chromatographia 68 (2008), 65–70.
[10] ICH, Q2(R1) Validation of analytical procedures: Definitions and terminology, in: International Conference on Harmo-
nization, IFPMA, Geneva, Switzerland, 2005.
[11] H. Kirchherr and W.N. Kuhn-Velten, Quantitative determination of 48 antidepressants and antipsychotics in human serum
by HPLC tandem mass spectrometry. A multi-level, single-sample approach, Journal of Chromatography B 843 (2006),
100–113.
[12] L. Labat, P. Dallet, E. Kummer and J.P. Dubost, Spectrophotometric, spectrofluorimetric, HPLC and CZE determination
of mirtazapine in pharmaceutical tablets, Journal of Pharmaceutical and Biomedical Analysis 28 (2002), 365–371.
[13] R. Mandrioli, V. Pucci, C. Sabbioni, C. Bartoletti, F. Fanali and M.A. Raggi, Enantioselective determination of the novel
antidepressant mirtazapine and its active demethylated metabolite in human plasma by means of capillary electrophoresis,
Journal of Chromatography A 1051 (2004), 253–260.
[14] R. Mandrioli and M.A. Raggi, Electrodriven methods for the enantioseparation of second-generation antidepressant drugs:
an update, Electrophoresis 29 (2007), 252–259.
[15] F.A. Maris, E. Dingler and S. Niehues, High-performance liquid chromatographic assay with fluorescence detection for
the routine monitoring of the antidepressant mirtazapine and its demethyl metabolite in human plasma, Journal of Chro-
matography B 721 (1999), 309–316.
[16] I. Meineke, I. Kress, W. Poser, E. Ruther and J. Brockmoller, Therapeutic drug monitoring of mirtazapine and its metabo-
lite desmethylmirtazapine by HPLC with fluorescence detection, Therapeutic Drug Monitoring 26 (2004), 277–283.
[17] P.E. Morgan, J. Tapper and E.P. Spencer, Measurement of total mirtazapine and normirtazapine in plasma/serum by liquid
chromatography with fluorescence detection, Journal of Chromatography B 798 (2003), 211–215.
[18] C. Pistos, M. Koutsopoulou and I. Panderi, A validated liquid chromatographic tandem mass spectrometric method for
the determination of mirtazapine and demethylmirtazapine in human plasma. Application to a pharmacokinetic study,
Analytica Chimica Acta 514 (2004), 15–26.
[19] T. Romiguieres, F. Pehourcq, M. Matoga, B. Begaud and C. Jarry, Determination of mirtazapine and its demethyl metabo-
lite in plasma by high-performance liquid chromatography with ultraviolet detection. Application to management of acute
intoxication, Journal of Chromatography B 775 (2002), 163–168.
[20] C. Salgado-Petinal, J.P. Lamas, C. Garcia-Jares, M. Llompart and R. Cela, Rapid screening of selective serotonin re-
uptake inhibitors in urine samples using solid-phase microextraction gas chromatography–mass spectrometry, Analytical
and Bioanalytical Chemistry 382 (2005), 1351–1359.
[21] A.F. Schwartzberg, J.O. Cole and C. Debattista, Manual of Clinical Psychopharmacology, American Psychiatric Publish-
ing Inc., Arlington, VA, USA, 2007, pp. 90–95.
[22] W.F. Smyth, J.C. Leslie, S. McClean, B. Hannigan, H.P. McKenna, B. Doherty, C. Joyce and E. O’Kane, The characteri-
zation of selected antidepressant drugs using electrospray ionisation with ion trap mass spectrometry and with quadrupole
time-of-flight mass spectrometry and their determination by high performance liquid chromatography/electrospray ioni-
sation tandem mass spectrometry, Rapid Communications in Mass Spectrometry 20 (2006), 1637–1642.
[23] K. Titier, N. Castaing, E. Scotto-Gomez, F. Pehourcq, N. Moore and M. Molimard, High-performance liquid chromato-
graphic method with diode array detection for identification and quantification of the eight new antidepressants and five
of their active metabolites in plasma after overdose, Therapeutic Drug Monitoring 25 (2003), 581–587.
[24] C. Velazquez, A. Carlson, K.A. Stokes and J.B. Leikin, Relative safety of mirtazapine overdose, Veterinary and Human
Toxicology 43 (2001), 342–344.
[25] S.M.R. Wille, K.E. Maudens, C.H.V. Peteghem and W.E.E. Lambert, Development of a solid phase extraction for 13 ‘new’
generation antidepressants and their active metabolites for gas chromatographic–mass spectrometric analysis, Journal of
Chromatography A 1098 (2005), 19–29.
[26] R.M. Youssef, Determination of mirtazapine in spiked human plasma and tablets by first derivative spectrofluorimetric
method, Saudi Pharmaceutical Journal 18 (2010), 45–49.
Spectroscopy 24 (2010) 577–583 577
DOI 10.3233/SPE-2010-0487
IOS Press
A modular Raman microspectroscopy system

for biological tissue analysis
Shuang Wang a,b,c , Jianhua Zhao a,b , Harvey Lui a,b , Qingli He c and Haishan Zeng a,b,∗
a
Cancer Imaging Department, British Columbia Cancer Agency Research Centre, Vancouver, BC,
Canada
b
Laboratory for Advanced Medical Photonics, Photomedicine Institute, Department of Dermatology
and Skin Science, University of British Columbia and Vancouver Coastal Health Research Institute,
Vancouver, BC, Canada
c
Department of Physics, Northwest University, Xi’an, Shaanxi, China
Abstract. Raman spectroscopy has been used as a sensitive tool for studying biological tissue and evaluating disease. In many
applications, microscopic level resolution spectral analysis is desirable. And this has been performed mostly by expensive com-
mercial confocal micro-Raman systems. In this research, we present a simple method for building an economical and modular
Raman microspectroscopy system that combines a microscope with a Raman spectrometer using an optical fiber bundle. The
bundle with a circular collection end is positioned at an image plane of the microscope to collect Raman signals from the
interested micro-location on the sample. The light delivery end is specially configured so that its 37 fibers are arranged along a
straight line to fit into the spectrometer entrance slit. This configuration improves light collection efficiency and maintains high
spectral resolution. To battle the great background autofluorescence and Raman signals that could originate from the microscope
slides and optics due to the non-confocal set-up of our simplified system, conventional normal-incident illumination is replaced
by oblique illumination at 45◦ degrees and the microscope slides are coated with gold. We demonstrated the usefulness of the
system by measuring micro-Raman spectra from different skin layers on vertical sections of normal skin tissue samples.
Keywords: Microscopy Raman, near infrared fluorescence imaging, gold-coated slide, oblique illumination, tissue Raman
spectroscopy, skin
1. Introduction
Raman microspectroscopy has been well established as a sensitive method for studying cells, tissues
and diseases [8]. Although the native Raman signal from biological materials is somewhat weak due
to the low probability of inelastic scattering events, those Raman scattered signals that can be captured
and recorded provide specific molecular information about the biochemical composition and structure of
cells and tissues [3,9]. Conventional Raman spectroscopy consists of a light source, typically a laser, and
systems for light delivery, collection and signal detection. Recent technical advances in confocal micro-
Raman spectroscopy provide more precise spatial interrogation and become the preferred analysis in
many applications [3,9]. However, this has been performed mostly by expensive commercial confocal
micro-Raman systems. Having less expensive and simpler systems available in most laboratories will
further expand the applications of micro-Raman analysis.
*
Corresponding author: Dr. Haishan Zeng, Cancer Imaging Department, British Columbia Cancer Agency Research Centre,
675 West 10th Avenue, Vancouver, BC, V5Z 1L3, Canada. Tel.: +1 604 675 8083; Fax: +1 604 675 8099; E-mail: hzeng@
bccrc.ca.
578 S. Wang et al. / A modular Raman microspectroscopy system for biological tissue analysis
A major advantage of Raman microspectroscopy for study of biological samples is the minimal sample
preparation required [7]. Cells and tissues are usually grown or placed on an appropriate IR-transparent
substrate such as CaF2 , BaF2 , MgF2 and ZnSe [7,8]. However, these particular materials exhibit aut-
ofluorescence and/or Raman scattering, which can contaminate the desired measurement in the form of
background signals especially for non-confocal configurations. These substrates are also expensive and
have undesirable mechanical properties such as fragility.
We introduce an economical and modular dispersive Raman microspectroscopy system with low back-
ground signal by combining a Raman spectrometer with a microscope through fiber optics and employ-
ing gold-coated microscope slides. By using glass slides coated with a layer of gold for mounting thin
tissue sections and oblique illumination, we are able to avoid most of the unwanted background signal
from the system and the microscope slide. To enhance the collection efficiency of the inherently weak
Raman signals, we packed thirty-seven 40 µm core-diameter fibers into a bundle. Sample Raman spectra
from normal skin sections using this microspectroscopy system are presented.
Raman microspectroscopy system. The Raman microspectroscopy system was developed by modify-
ing a microspectrophotometer system originally constructed for fluorescence measurements [10]. The
setup of the Raman microspectroscopy system is schematically shown in Fig. 1. Its major components
include light sources (a laser and a tungsten white light lamp), an inverted microscope (Diaphot, Nikon)
and a Raman spectrometer.
For measuring Raman spectra, 785 nm near infrared (NIR) light from a fiber-coupled, temperature
stabilized diode laser (BRW-785-1.0-100-0.22-SMA, B&W Tek Inc., Newark, DE, USA) was delivered
to a collimator, passing through a band-pass filter (FF01-785/62-25, Semrock Inc., Rochester, NY, USA)
and then focused by a biconvex lens onto the tissue section at an angle of 45◦ . The Raman signal was
then collected by the subjacent objective, passed through a 830 nm longpass filter (LP02-830RU-25,
Semrock, Rochester, NY, USA) and focused onto the image plane of the camera port of the microscope.
Fig. 1. The setup of the Raman microspectroscopy system.

S. Wang et al. / A modular Raman microspectroscopy system for biological tissue analysis 579
A customized optical fiber bundle (FiberTech Optica, ON, Canada) consisting of thirty-seven 40 µm
optical fibers was used to enhance the collection efficiency of the inherently weak Raman signals. At
the collection end connecting to the microscope, the fibers were packed into a 0.385 mm diameter cir-
cular area, which also defined the detection spot on the sample, while in the light delivery end con-
necting to the spectrometer, they were packed into two paralleled columns to fit into the entrance slit
of the spectrograph. One column contains 18 and the other contains 19 of the 40 µm core-diameter
optical fibers. The spectrometer is equipped with an NIR-optimized back-illuminated deep-depletion
CCD array (Spec-10:100BR/LN, Princeton Instruments, Trenton, NJ, USA) and a transmissive imaging
spectrograph (HoloSpec-f/2.2-NIR, Kaiser Optical, Ann Arbor, MI, USA) with a holographic grating
(HSG-785-LF, Kaiser Optical, Ann Arbor, MI, USA). The CCD has a 16 bit dynamic range and is liquid
nitrogen cooled to −120◦ C. The spectral resolution of the system is 8 cm−1 . The spectrometer is con-
trolled by a computer (Optiplex 755, Dell, TX, USA) through a USB 2.0 interface. The wavelength was
calibrated using an Hg–Ar lamp (HG-1, Ocean Optics Inc., Dunedin, FL, USA) and the spectral response
of the system was calibrated using an NIST traceable tungsten calibration lamp (RS-3, EG&G Gamma
Scientific, San Diego, CA, USA) [11]. The data acquisition was implemented with WinSpec software
package (drivers from Roper Scientific, Trenton, NY, USA). The dark noise of the CCD detector was
measured and removed from each acquired spectrum. The integration time for each spectrum is 30 s.
White light illumination was used for conventional microscopic imaging, which can be seen through
the eyepiece. To locate the spot for Raman spectra measurement, the collection optical fiber bundle was
disconnected from the spectrometer and connected to an auxiliary white light source. The light from the
fiber tips was focused on the sample, which was visible through the eyepiece as the image of the fiber
bundle end face. Based on the property of light reversibility, only the signal from this bright spot could
be collected by the fiber bundle for spectral analysis. This bright spot was therefore used to localize the
desired spot (micro-location) for spectral measurement [10]. For the measurements shown in this paper,
the spot size was about 15.4 µm in diameter for a 10× objective and a 0.385 mm diameter collection
fiber bundle.
The system also provides for near-infrared fluorescence imaging of the sample. If the movable mirror
M2 is inserted into the light path in the microscope system (Fig. 1), the fluorescence signal (including
the associated weak Raman signal) from the sample will be deflected and form an image on an NIR-
enhanced CCD camera (Alta® U1, Apogee Instruments Inc., Roseville, CA, USA) through the video
port on the microscope. Fluorescence images were acquired by the computer for storage and/or further
processing. Also, the visible image of the sample could be acquired with a visible light source in the
same way.
Microscope slide background fluorescence. For micro-Raman measurements and NIR fluorescence
imaging, it is critical to choose a substrate (slide) material that is relatively free of intrinsic fluorescence
and Raman scattering because these background signals from the slide may contaminate the Raman
spectra and the fluorescence images. We measured a standard glass slide (Cat. No. 12-544-1, Fisher
Scientific, Pittsburgh, PA, USA), a cover slip (Cat. No. 12-545E, Fisher Scientific, Pittsburgh, PA, USA)
and a BaF2 slide. All presented significant fluorescence and Raman background as shown in Fig. 2. To
reduce the background fluorescence and Raman signals, the glass slide coated with a layer of 120 nm
thick pure gold (Biogold® 63479-AS, Electron Microscopy Sciences, Hatfield, PA, USA), was utilized in
our system and shows near zero background fluorescence and Raman emitted. In addition to the fact that
gold exhibits no NIR fluorescence, the other advantage of this special preparation is that the majority
of excitation light is specularly reflected away from the optical axis by the surface gold layer and is
therefore not collected by the objective (Fig. 1). This helps to reduce the background fluorescence and
Fig. 2. Fluorescence and Raman background spectra of standard glass microscope slide, standard microscope cover slip, BaF2
slide and gold-coated glass slide.
Raman signals from the objective and other optical components of the microscope that could otherwise
be excited by the laser light. Furthermore, when the excitation light is focused onto the tissue, part of
it is absorbed and scattered by the skin’s molecules and the rest is transmitted through the thin tissue
sample. Since most of the transmitted laser beam is subsequently reflected back by the gold layer on the
slide, the absorption and scattering occur twice for this arrangement. This optical tissue path doubling
scheme increases the inelastic scattering events arising within the skin sample.
Sample preparation. Unstained and fresh normal skin sections for Raman microspectroscopy measure-
ment and NIR autofluorescence imaging were obtained from facial surgery. Immediately after excision,
the sample was put into a capsule and snaps frozen with liquid nitrogen. Vertical sections of 20 µm
thicknesses were cut with a microtome at −20◦ C and mounted onto gold-coated glass slides without
cover slips. The frozen sections were kept in −80◦ C until usage. Before the measurements, the samples
were acclimated to room temperature.
The pure Raman spectra of the skin samples were obtained after fluorescence background subtraction
from the measured raw spectra using the Vancouver Raman Algorithm we developed previously [11].
Figure 3 shows Raman spectra of different skin layers along with the autofluorescence image of the
normal skin sample. This image was used to identify various skin structures. Prominent spectral features
in the range of 800–1800 cm−1 are the major vibration bands around 1650, 1445, 1311, 1269, 1080,
1002 and 855 cm−1 , which can all be found from the in vivo skin Raman spectra [11]. The strong
Fig. 3. NIR autofluorescence image of normal facial skin tissue section and Raman spectra measured from different skin
layers on the section. Several spectral differences among the major layers of skin (stratum corneum, epidermis, dermis and
subcutaneous fat) are apparent due to intrinsic differences in skin layer composition. Integration time is 30 s for each spectrum.
Raman bands at 1650 and 1269 cm−1 are assigned to protein vibration modes involving amide bonds
(amide I and amide III), and mainly originate from keratin which is the most abundant protein in stratum
corneum [2]. The next strongest band is at 1445 cm−1 , assigned to the CH2 deformation of proteins
and lipids. Another strong band centered at 1301 cm−1 was assigned to a twisting deformation of the
CH2 methylene groups of intracellular lipid acyl moieties. The region from 1000 to 1150 cm−1 in the
Raman spectra contains information on the hydrocarbon chain, while the peak at 1080 cm−1 could be
due to a random conformation vibration mode. Also, bands around 855 cm−1 and 945 cm−1 could be
observed presumably from collagen. The Raman spectra of normal skin sections obtained in this study
are essentially similar to those reported in the literature [1,4,5].
As shown in Fig. 3, the Raman spectra obtained from vertical skin sections reveal significant dif-
ference between skin layers. The Raman from the stratum corneum is weak because it is a thin layer.
Nevertheless most of them were in agreement with results obtained from FT-Raman measurements on
isolated stratum corneum [1,5]. The major variations in the spectra were seen in the relative band in-
tensities, notably around 850–1200 cm−1 and 1250–1350 cm−1 . One of the most striking differences
between the spectra of the stratum corneum and the underlying epidermis was observed at 1062, 1126
and 1296 cm−1 . All these three bands were strong features in the Raman spectrum of ceramide, which
represents the most abundant class of stratum corneum lipids [4]. The prominent vibrational features of
human skin, such as the ν(C=O) amide I band at 1650 cm−1 , the δ(CH3 ) and δ(CH2 ) scissoring mode at
1445 cm−1 , the CH2 deformation at 1301 cm−1 and the ν(CN) and δ(NH) amide II bands at 1269 cm−1 ,
can be clearly discerned in different skin layers. It clearly illustrates that skin Raman properties can be
easily studied using tissue sections with low background signals by our Raman microspectroscopy sys-
tem. The results obtained will help explain the biophysical basis for the in vivo skin spectra from clinical
measurements, which is an integral contribution of different skin layers.
To the best of our knowledge, this is the first report of microscopic NIR autofluorescence images of
the skin. It clearly outlined the NIR fluorophore distributions in the skin. The image shown in Fig. 3
suggests that there are considerable NIR autofluorescence emission from the stratum corneum and the
dermis, while the epidermis layer exhibits minimum signals. We are planning to study the microscopic
NIR autofluorescence properties of various skin lesions. The results obtained will be used to understand
the microscopic origins and biophysical basis of many interesting in vivo NIR autofluorescence images
we obtained recently in our clinical studies [6].
4. Conclusions
We have successfully built an economical and modular Raman microspectroscopy system that com-
bines a microscope with a Raman spectrometer using an optical fiber bundle. The system has three
unique features: (1) a round-to-linear fiber bundle configuration is used to improve Raman signal col-
lection efficiency, while also maintain high spectral resolution; (2) the use of oblique illumination and
(3) the use of gold-coated microscope slide effectively eliminated the background autofluorescence and
Raman signals that could originate from the microscope slides and optics due to the non-confocal set-
up of our simplified system. We demonstrated the usefulness of the system by successfully measuring
micro-Raman spectra from different skin layers on vertical sections of normal skin tissue samples. Us-
ing the system we also captured microscopic near infrared autofluorescence images of the skin for the
first time. These microscopic spectral and imaging data will be very helpful for understanding the mi-
croscopic origins and biophysical bases of our in vivo clinical Raman spectra and NIR autofluorescence
images.
Acknowledgements
This work was supported by the Canadian Cancer Society, the Canadian Dermatology Foundation,
the VGH & UBC Hospital Foundation “In It for Life Fund” and the BC Hydro Employees Community
Services Fund.
References
[1] A.N.C. Anigbogu, A.C. Williams, B.W. Barry and H.G.M. Edwards, Fourier transform Raman spectroscopy of interac-
tions between the penetration enhancer dimethyl sulfoxide and human stratum corneum, International Journal of Phar-
maceutics 125 (1995), 265–282.
[2] B.W. Barry, H.G.M. Edwards and A.C. Wiliams, Fourier transform Raman and infrared vibrational study of human skin:
Assignment of spectral bands, Journal of Raman Spectroscopy 23 (1992), 641–645.
[3] P.J. Caspers, G.W. Lucassen, E.A. Carter, H.A. Bruining and G.J. Puppels, In vivo confocal Raman microspectroscopy of
the skin: noninvasive determination of molecular concentration profiles, Journal of Investigative Dermatology 116 (2001),
434–442.
[4] P.J. Caspers, G.W. Lucassen, R. Wolthuis, H.A. Bruining and G.J. Puppels, In vitro and in vivo Raman spectroscopy of
human skin, Biospectroscopy 4 (1998), S31–S39.
[5] H.G.M. Edwards, A.C. Williams and B.W. Barry, Potential applications of FT-Raman spectroscopy for dermatological
diagnostics, Journal of Molecular Structure 347 (1995), 379–387.
[6] X. Han, H. Lui, D.I. McLean and H. Zeng, Near-infrared autofluorescence imaging of cutaneous melanins and human
skin in vivo, Journal of Biomedical Optics 14 (2009), 0240179.
[7] C. Krafft and V. Sergo, Biomedical applications of Raman and infrared spectroscopy to diagnose tissues, Spectroscopy 20
(2006), 195–218.
[8] I. Notingher and L.L. Hench, Raman microspectroscopy: a noninvasive tool for studies of individual living cells in vitro,
Expert Review of Medical Devices 3 (2006), 215–234.
[9] M.A. Short, H. Lui, D. McLean, H. Zeng, A. Alajlan and X.K. Chen, Changes in nuclei and peritumoral collagen within
nodular basal cell carcinomas via confocal micro-Raman spectroscopy, Journal of Biomedical Optics 11 (2006), 034004-
1–034004-9.
[10] H. Zeng, C.E. MacAulay and D.I. Mclean, Novel microspectrophotometer and its biomedical applications, Optical Engi-
neering 32 (1993), 1809–1814.
[11] J. Zhao, H. Lui, D.I. McLean and H. Zeng, Integrated real-time Raman system for clinical in vivo skin analysis, Skin
Research and Technology 14 (2008), 484–492.
Spectroscopy 24 (2010) 629–639 629
DOI 10.3233/SPE-2010-0485
IOS Press
Non-invasive blood glucose measurement by

near infrared spectroscopy: Machine drift,
time drift and physiological effect
Simon C.H. Lam ∗ , Joanne W.Y. Chung, K.L. Fan and Thomas K.S. Wong
School of Nursing, The Hong Kong Polytechnic University, Hong Kong, China
Abstract. The aim of this paper is to evaluate development of the non-invasive blood glucose measurement of near infrared
(NIR) spectroscopy. The results showed that NIR spectroscopy might obtain glucose concentration of up to 200% difference
under a same environmental condition with two months apart due to time and machine drifts. These effects can restrict the
development of the non-invasive blood glucose measurement. Partial least square (PLS) regression was used, which showed
advantage over using simple absorbance for glucose concentration. Non-invasive blood glucose measurement of health subjects
(non-diabetics) was also investigated. The results showed that R correlation coefficient of prediction (Rp ) was 0.48 and root
mean square of prediction (rmsep) was 1.34 mmol/l. The error was mainly due to the physiological effect of different subjects.
Keywords: Spectroscopy, machine drift, time drift, physiological effect, PLS, PDS, Beer’s law
1. Introduction
Diabetes mellitus (DM) is a chronic and incurable disease caused by the malfunction of insulin produc-
tion which affects the blood glucose [1,8,10] in the body, normally leading to hyperglycemia. Prolonged
hyperglycemia causes damage to nerves and blood vessels [1,8,10], which may cause many complica-
tions. Thus, the role of blood glucose monitoring and measurement becomes very important for diabetic
therapies.
1.1. Diabetes
The number of diabetics is increasing. According to World Health Organization (WHO), there were
171 million people with DM worldwide [6,11]. In the year 2030, the number of diabetics may reach to
366 million, of which 90–95% will suffer from Type 2 diabetes [6,9]. The highest prevalence was in the
developing countries, India and China (20.8 million) in the year 2000 [6]. It is predicted that the number
of the diabetics will significantly increase to 42.3 million in China after 30 years [6].
Finger-prick glucose measurements are common and can provide reliable glucose readings. They do
not continuously monitor blood glucose and therefore cannot present a precise glucose trend. In addition,
since blood samples are needed for the glucose measurement, this repeated sampling causes pain and
damage to the skin tissue, which may raise the chance of infection.
*
Corresponding author: Simon C.H. Lam, School of Nursing, The Hong Kong Polytechnic University, Hong Kong, China.
Tel.: +852 8210 0441; Fax: +852 8210 0517; E-mail: simon-lam@live.com.
630 S.C.H. Lam et al. / Non-invasive blood glucose measurement by near infrared spectroscopy
A non-invasive approach is the painless method for blood glucose measurement. However, no non-
invasive blood glucose meter currently in use has provided consistently reliable results, due to many
interfering non-specific physiological signals during the measurement (especially for long-term mea-
surement). According to Danzer et al. [3], the long-term unreliability might be due to the physiological
differences amongst subjects and time drift factors. Thus, diabetic patients still need to rely on painful
finger-prick sampling for daily glucose measurement.
2. Experiment and clinical trial
An experiment by using glucose solutions (see Fig. 1) and a clinical trial with healthy subjects had
been conducted.
2.1. Glucose solutions
Control Development’s NIR spectrometer of NIR-128L-1.7-USB, with the wavelength range of 905–
1701 nm was used. D-Glucose dissolved in deionised water was used for preparing different glucose
concentrations. A thermo-resistant petri dish made of polystyrene (PS) was used as the container because
of its high temperature resistance and good optical clarity.
The known concentrations of solutions were measured by NIR spectrometer in a room with a temper-
ature range of 24–25◦ C and a relative humidity of 55–60%. Four sets of glucose solution measurements,
each set with five concentrations (0, 125 250, 500 and 1000 mmol/l), were measured on three separate
days by NIR spectroscopy. The average of the NIR spectra was recorded from 64 readings per sample.
Calibration against the background and the reference (under tungsten halogen light) was carried before
each glucose measurement.
2.2. Clinical trial
Thirty-six healthy subjects voluntarily joined the clinical trial and had been fasting overnight before
the first NIR and the first finger-prick glucose measurement in the morning. The subject group comprised
eight males and 28 females, all of whom were non-diabetic.
Finger pricks and NIR measurements were performed at the preprandial (fasting stage) and again at
the postprandial (after had breakfast) stage. The whole procedure for data capture (including NIR spec-
tra, finger surface temperature and finger-prick glucose levels) took approximately 3–5 min, per patient.
Each subject was provided the same breakfast at the same quantity after the preprandial measurement.
A sucrose candy had also been provided for each subject 15–30 min before the postprandial measure-
ments. This helped to increase their respective glucose level, thus the glucose levels would be varied.
Fig. 1. Block diagram – glucose concentration analysis.

S.C.H. Lam et al. / Non-invasive blood glucose measurement by near infrared spectroscopy 631
3. Analytical methods
According to Beer’s law, the absorbance is related to the concentration. Thus, based on Beer’s law,
a single wavelength was selected for the glucose concentration evaluation. Another technique of partial
least square (PLS) prediction [4,7] was also applied for the analysis. One group of data from two different
experiments were used for calibrations, another one was for prediction and validation purposed as shown
in Fig. 2.
The spectra obtained were separated into two groups. One group (Spectra A) was used for calibration
with the corresponding glucose concentrations and formed the calibration model (Model) shown on
Eqs (1) and (2).
g1 = M1 X1 + e1 , (1)
where X1 – Spectra A, e1 – error 1, g1 – corresponding glucose concentrations, M1 – calibration model

built.
Then, Spectra B were substituted to M1 by PLS.
g = M1 X2 + e2 , (2)
where X2 – Spectra B, e2 – error 2, g – predicted glucose concentrations.

The predicted glucose concentrations g were then validated with g2 , where g2 was the glucose con-
centration corresponding to Spectra B.
4. Results and discussions
4.1. Glucose solutions
The results showed that pure water provided a higher absorbance spectrum. Lower absorbance spectra
represented higher glucose concentrations as shown in Fig. 3. The wavelength 1180 nm was chosen for
the analysis, as it showed the most clearly distinguished absorbance and was also located in the region
of second overtones [5].
Fig. 2. Block diagram – PLS prediction process of glucose solution concentration.

Fig. 3. NIR spectra – glucose solutions, absorbance against wavelength (nm). Day 1 under room condition. (Colors are visible
in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
4.2. Glucose solutions measurement on the same day
Different batches of glucose solutions were measured within one hour by NIR spectroscopy. These
measurements were used to check the reliability of the NIR measurement within a short period of time,
while trying to avoid machine drift and time drift. The glucose concentration was first plotted against
the NIR absorbance spectra.
4.3. Single wavelength absorbance vs. glucose concentration
Inverse linearity was observed with absorbance spectra at wavelength 1180 nm, where the lower ab-
sorbance indicated a higher glucose concentration. The best fitted lines, shown in Fig. 4, can be repre-
sented by the following linear equations, respectively.
Fig. 4. Absorbance at 1180 nm against glucose concentration, separated experiment in one hour; fit of Day 3a is close to fit of
Day 3b. (Colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
Fit 3a:
y = −5.19e−5 x + 0.47 (R = 0.91 and RMSE = 0.0108 AU or 207 mmol/l). (3)
Fit 3b:
y = −4.84e−5 x + 0.46 (R = 0.87 and RMSE = 0.0098 AU or 204 mmol/l), (4)
where y – absorbance units, x – glucose concentrations, R – R correlation coefficient, RMSE – root

mean square error.
Figure 4 showed that the fit of 3a and 3b were close to each other. This showed that the results of
the two experiments were nearly repeated. This also demonstrated that glucose concentrations could
be measured via transmittance NIR spectroscopy, where the absorbance (AU) is related to the glucose
concentrations.
4.4. Glucose concentration predictions by PLS
PLS is another prediction technique mainly used for spectroscopic application. The spectral results
were analyzed by PLS. The known glucose concentrations were calibrated with the measured spectra
from Day 3a to form a model and then the measured spectra of Day 3b (after one hour) were predicted
through PLS. Figure 5 shows the scatter plot of the calibration results and the prediction results of the
glucose concentration within one hour.
These results showed the R correlation coefficient of the prediction Rp = 0.92 and root mean square
error rmse = 158 mmol/l. The relationship between spectral signals (or the absorbance unit under
Fig. 5. PLS prediction scatter plot, separated experiment in the same day within one hour; cross (×) – calibration; hollow (◦) –
prediction. (Colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
tungsten halogen at 1180 nm wavelength) and the glucose concentration was linear. The major predicted
error occurred in the low glucose range.
PLS provides another prediction technique for the glucose concentrations by using NIR spectroscopy.
The poorest prediction was located in the pure water region. PLS prediction provided partly poorer
results when compared with the direct absorbance to the glucose concentration analysis. However, the
prediction made through PLS was still better than just using the absorbance unit. Although glucose
concentration is proportional to the NIR absorbance of a single wavelength, the rate of error may be
higher when only a single wavelength is used. Alternatively, PLS was applied by using the whole range
of the spectra for the prediction. The error may be relatively small since used whole range spectra may
cause to average the error.
4.5. Glucose solution measurements on different days
Different batches of known glucose solutions were measured on different days (two month apart)
by the same NIR spectrometer. The glucose concentration was plotted against NIR absorbance for the
evaluation. It was also analyzed by PLS.
4.6. Single wavelength absorbance versus glucose concentration on separate days
Absorbance of the single wavelength at 1180 nm was plotted against the corresponding glucose con-
centration, as shown in Fig. 6. The results showed that the data of Day 1 is about 0.06 AU (potentially a
very significant error) higher than the data of Day 2. Their fitting equations are shown below.
Fit 1:
y = −3.32e−5 x + 0.47 (R = 0.99 and RMSE = 0.0020 AU or 60 mmol/l). (4-5)

Fig. 6. Absorbance at 1180 nm against glucose concentration, two experiments in two months separation (with two preparations
for same concentrations). (Colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
Fit 2:
y = −3.23e−5 x + 0.41 (R = 0.91 and RMSE = 0.0052 AU or 163 mmol/l), (4-6)
where y – absorbance units, x – glucose concentrations, R – R correlation coefficient, RMSE – root

mean square error.
0.06 AU represents 2000 mmol/l. This showed that the results might not be reproducible under a single
wavelength measurement after two months. This was due to machine drift and time drift. Even though
there was only one variable – glucose, the machine might face micro-deviation among the different inner
parts which accumulated over a period of time.
4.7. Spectra prediction versus glucose concentration by PLS on separate days
Preprocess of Piecewise Direct Standardization (PDS) was used to regulate the spectral drifts from dif-
ferent days. These preprocessed data were then calibrated and predicted by PLS regression. The results
were then scatter plotted, as shown in Fig. 7.
The results showed that Rp = 0.97 and rmsep = 144 mmol/l. A major error also occurred in the low
glucose concentration region, particularly for the pure water measurement. Table 1 shows the summary
of R and rmse (Rp and rmsep) by absorbance and PLS spectral analysis, respectively.
5. Error and noise
There were possible measurement errors, or noises, of the glucose solution. The experiment was car-
ried out under the exposure to the surrounding environment. The results might not be affected by the
Fig. 7. PLS prediction scattered plot, separated experiment in different days in two month: cross (×) – calibration; hollow (◦) –
prediction. (Colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
Table 1
R correlation of coefficient and rmse (root mean square error) comparison; absorbance of Day 1 > 0.06 AU of absorbance of
Day 2
Analysis Day
Under same day Under different days

R RMSE (mmol/l) R RMSE (mmol/l)
Absorbance 0.91 207 0.99 (Day 1) 60 (Day 1)
0.87 204 0.91 (Day 2) 163 (Day 2)
PLS (prediction) Rp = 0.92 rmsep = 158 Rp = 0.97 rmsep = 144
surrounding florescent light which exists at 190 nm to 650 nm, while NIR is ranged from 730 nm to
2500 nm. The indium gallium arsenide (InGaAs) detector (NIR detector could only sense within NIR
region) of the spectrometer was not affected by the surrounding visible light. However, NIR might have
derived from other NIR environmental sources, such as temperature variations, if some of the other
equipment present generated an NIR wavelength range.
In addition, the resolution of the spectrometer is 6.22 nm, thus, the specificity of the absorbance
at 1180 nm could be 1180 ± 6.22 nm. The results might not really represent the absorbance of the
exact corresponding wavelength. However, NIR overtones always overlaps to multiple wavelengths or
a certain range of the spectrum (83). The separation of each measurement pixel forms the spectrum by
interpolation. This indicates that the resolution might not provide a significant effect to the absorbance.
Unless there was a particular wavelength that had fallen in between two pixels, error might happen.
However, as NIR overtones always cover multiple wavelengths, the resolution of the absorbance was
enough to cover the measurement.
Nevertheless, the measurement area in the experiment was relatively large within the petri dish, and
it might affect the results, particularly on the low glucose concentration. The glucose particles might
not be equally distributed among the large area. The error could be relatively large due to the outside
environment and the glucose preparation error, particularly in low glucose concentration areas.
5.1. Clinical trial
A total of 108 sets of spectra were obtained. Thirty-six sets of spectra were obtained before breakfast.
Seventy-two sets of spectra (where one additional set of spectra was obtained from each subject in
postprandial measurement) were obtained after breakfast. Figure 8 shows a spectral plot of absorbance
against wavelength. The spectra were similar to those shown in Fig. 3. Overtones also occurred on
980 nm and 1200 nm. PLS was applied for the prediction (each third of the data of the spectra and the
corresponding glucose readings were used for prediction, while the remaining sets of data were used for
calibration). Results are presented in the scatter plot shown in Fig. 9.
Figure 9 showed that both the calibrations and predictions were located within the clinical, acceptable
regions, A and B. The R correlation coefficient of the calibration (Rc ) was 0.85 and the root mean square
Fig. 8. Absorbance against wavelength, healthy subject. (Colors are visible in the online version of the article; http://dx.doi.org/
10.3233/SPE-2010-0485.)
Fig. 9. Scatter plot of predicted against finger-prick blood glucose level with simple mean-centred preprocess and PDS pre-
process, for full wavelength and non-arranged data: (×) – calibration, (◦) – prediction. (Colors are visible in the online version
of the article; http://dx.doi.org/10.3233/SPE-2010-0485.)
error of calibrations (rmsec) was 0.57 mmol/l. The R correlation coefficient of the prediction (Rp ) was
0.48. The root mean square error of prediction (rmsep) was 1.34 mmol/l.
Most of the prediction spots fell in Region A, where errors were within 20%. Regions A and B were
clinically acceptable regions as stated by Clarke Grid Error plot [2]. This demonstrated that NIR spectra
and the blood glucose levels of non-diabetics were correlated after applying PLS for the prediction.
The results showed that the study of non-diabetics provided a positive sign for non-invasive blood
glucose prediction. However, high variation as shown in Fig. 8 may lead to difficulties for blood glu-
cose prediction. The spectral variation might indicate that higher noises or uncertainty of signals might
be obtained in addition to the necessary signals. This might be due to time drift, machine drift and
physiological differences in a long-term clinical measurement.
The physiological influence from the human body was one of the key effects in relation to non-invasive
blood glucose measurement. Time factors have led to uncertain physiology, which could affect the pre-
diction. The longer time span between each measurement, the more uncertainties due to the physiology
occur; thus, the poorer the calibration and the prediction obtained. This is because of the higher chance of
the reaction of the human body to both internal and external environmental changes. Therefore, the study
of healthy subjects in short-term might have provided better predictions due to the shorter time span and
relatively fewer physiological uncertainties. However, Rp and rmsep were 0.48 and 1.34 mmol/l, respec-
tively. This was because of physiological uncertainties between subjects.
These two studies showed that ‘time and machine drifts’ and ‘physiological effect’ could affect in vivo
blood glucose prediction. However, as mentioned, the glucose solution experiment showed that machine
drift and time drift could be reduced by using PLS with preprocess. Thus, the error might be due to
physiological differences between subjects.
Furthermore, the results of this clinical trial also showed that the larger variation caused larger error
for PLS predictions. The large variation might have been due to the physiological influences, including
temperature. Thus, NIR measurement also needed to be improved by reducing the variation during the
measurements.
6. Conclusions
Results showed that higher glucose concentration, the lower the spectral absorbance was in the spec-
trum. However, NIR spectroscopic measurements were deviated due to machine drift and the time drift
after two months. This caused errors in the NIR spectroscopic measurements. These drifts could be
reduced or eliminated by PLS technique. NIR spectroscopic measurement under PLS analysis showed
promising results on the glucose solution test in both short- and long-term experiments.
However, physiological effects are unpredictable and varied from time to time due to the human body’s
reaction to the environment. This led to the most uncertainty during the clinical trial. Thus, considering
the physiological influence for NIR non-invasive blood glucose measurements would be needed.
References
[1] ADA, All about diabetes, American Diabetes Association, 2006, available at: http://www.diabetes.org/about-diabetes.jsp.
[2] W.L. Clarke, D. Cox, L.A. Gonder-Frederick, W. Carter and S.L. Pohl, Evaluating clinical accuracy of systems for self-
monitoring of blood glucose, Diabetes Care 10 (1987), 622–628.
[3] K. Danzer, C. Fischbacher, K.U. Jagemann and K.J. Reichelt, Near-infrared diffuse reflection spectroscopy for non-
invasive blood-glucose monitoring, IEEE 12(2) (1998), available at: http://www.ieee.org/organizations/pubs/newsletters/
leos/apr98/monitoring.htm.
[4] S. de Jong, SIMPLS: An alternative approach to partial least squares regression, Chemometrics and Intelligent Laboratory
Systems 18 (1993), 251–263.
[5] B. Jorgensen and Y. Goegebeur, Multivariate Data Analysis and Chemometrics, Department of Statistic, University of
Southern Denmark, Denmark, 2006.
[6] W. Sarah, R. Sicree, G. Roglic, H. King and A. Green, Global prevalence of diabetes: Estimates for the year 2000 and
projections for 2030, Diabetes Care 27 (2004), 1047–1053.
[7] R.D. Tobias, An introduction to partial least squares regression, SAS Institute Inc., Cary, NC, USA, 1998, available at:
http://support.sas.com/rnd/app/papers/pls.pdf.
[8] WHO, Diabetes mellitus, World Health Organization Media Center, April 2002, Fact Sheet No. 138.
[9] WHO, Screening for type 2 diabetes, World Health Organization, 2003.
[10] WHO, Diabetes programme: What is diabetes?, World Health Organization, 2006, available at: http://www.who.int/
diabetes/BOOKLET_HTML/en/index4.html.
[11] WHO, Diabetes programme: A rising global burden, World Health Organization, 2006.
Spectroscopy 24 (2010) 621–628 621
DOI 10.3233/SPE-2010-0484
IOS Press
Azure dyes as new photosensitizer

prototypes to application in photodynamic
therapy against Candida spp.
Leonardo M. Moreira a , Juliana P. Lyon b , Suellen M.S. Tursi c , Isis Trajano c ,
Monalisa P. Felipe c , Maricília S. Costa c , Máira R. Rodrigues d , Lúcia Codognoto e and
Hueder P.M. de Oliveira e,∗
a
Departamento de Engenharia de Biossistemas, Universidade Federal de São João Del Rei, São João
del Rei, MG, Brazil
b
Departamento de Ciências Naturais, Universidade Federal de São João Del Rei, São João del Rei,
MG, Brazil
c
Instituto de Pesquisa e Desenvolvimento, Universidade do Vale do Paraíba, São José dos Campos, SP,
Brazil
d
Departamento de Ciência e Tecnologia, Universidade Federal Fluminense, Polo Universitário de Rio
das Ostras, Rio das Ostras, RJ, Brazil
e
Universidade Camilo Castelo Branco, São José dos Campos, SP, Brazil
Abstract. Infections caused by Candida albicans are of increasing concern, especially considering immunodepressed patients.
The toxicity of most antifungal agents, the great number of cases with recidives, as well as the emergence of resistant samples
has provoked the evaluation of new forms of therapy. In this context, the photodynamic therapy (PDT) presents auspicious
antimicrobial properties, stimulating the development of trials employing several kinds of photosensitizers. In the present work,
the application of different kind of Azure dyes as photosensitizer in PDT against C. albicans was evaluated through instrumental
measurements of electronic spectroscopy. In fact, the values of optical density were a precise indicator of the growth inhibition
of the microorganisms. Indeed, Azures are phenothiazinium derivatives that constitute a very relevant class of compounds with
several biomedical applications, such as photoantimicrobial therapy against local bacterial infection, tuberculosis, trypanosomi-
asis, malaria, Rickettsia, yeasts, viral infection n and cancer. Azure A, Azure B, Azure A thiocyanate, Azure B BF4 , Azure A
eosinate are the dyes tested against C. albicans. The results denoted completely distinct behaviors to the different types of
Azure compound evaluated in this work. In fact, Azure A and Azure A eosinate presented significant results when irradiated
with 56 J/cm2 , since the growth inhibition of C. albicans reached approximately 60%. This Azure compounds have significant
potential to be employed as photosensitizer (PS) in PDT, especially in cases of mucocutaneous candidosis. The spectroscopic
evaluation was very effective to the detection of slight alterations in the growth of the microorganisms, denoting that this kind
of analysis is an excellent alternative to determine growth inhibition of Candida albicans. The experimental data are discussed
in details in agreement with recent results from literature.
Keywords: Photodynamic therapy (PDT), photosensitizer (PS), Azures, Candida albicans
*
Corresponding author: Hueder P.M. de Oliveira, Universidade Camilo Castelo Branco/Rod. Presidente Dutra Km 139 –
Eugênio de Melo, São José dos Campos, SP 12247-004, Brazil. Tel./Fax: +55 12 3905 4648 – R 209; E-mails: huederpaulo@
yahoo. com.br, leonardomarmo@gmail.com.
622 L.M. Moreira et al. / Azure dyes as new PS prototypes to application in PDT against Candida spp.
1. Introduction
C. albicans is commensal yeast present in about 20% of healthy individuals [10]. However, some
factors are predisposing for the development of candidosis, such as immunodeficiencies, malignant dis-
eases, radiotherapy, long-term treatment with antibacterial drugs and age extremes [13]. Moreover, both
cutaneous and mucocutaneous candidosis may appear in healthy individuals with local predisposing
conditions, such as denture wearing, moist wrinkles in skin and altered local pH [14].
Amphotericin B and azolics, such as ketoconazole, fluconazole and itraconazole, have been considered
efficient in the treatment of both systemic and local candidosis. However, the prophylactic use of azolic
drugs in immunodepressed patients led to the selection of resistant strains [16]. Besides, the increase
in the number of infections caused by Candida species other than C. albicans also contributed to the
emergence of isolates that are resistant to amphotericin B and fluconazole [17]. It is also important to
notice that the treatment with antifungal drugs may be prolonged and recidives are very common.
Photodynamic therapy (PDT) is a promising therapeutic treatment for various tumors and nonmalig-
nant diseases. The treatment requires a dye and a light source. The light activates a dye deposited on
the target organism and sensitizes it, transforming the oxygen at molecular level into reactive oxygen
species (ROS), such as singlet oxygen, which is cytotoxic to microorganisms [2,6,8,9,24]. Photodynamic
therapy has been originally developed for the treatment of malignant diseases. However, other applica-
tions have emerged. Photodynamic antimicrobial chemotherapy (PACT) is an alternative modality of
treatment for microbial infections. The therapy consists in applying a photosensitizer and a light source,
causing selective microbial death [11]. PACT has been successfully employed against leishmaniasis [1,
4], trichomoniasis [18], as well as in the destruction of gram-positive and gram-negative bacteria [22,
24]. PACT has been also tested against yeasts and dermatophytes [3,15].
In this context, the development of new or coadjutants treatments for superficial candidosis has a great
relevance. It is important to notice that infections produced by C. albicans, mainly those that occur in
immunodepressed patients, constitute a focus of interest to several groups. In fact, the development of
new treatments is a point of great relevance due to the appearance of resistant biological samples. Several
works have evaluated the use of photosensitizers and light to kill yeasts and molds [7]. However, there
are few studies regarding the physicochemical properties necessary in a photosensitizer to be effective
in photodynamic antifungal chemotherapy.
The present work evaluated the employment of Azures (Azure A, Azure B, Azure A thiocyanate,
Azure B BF4 and Azure A eosinate, see Fig. 1) as photosensitizers to be employed in PDT against C. al-
bicans. Azures are phenothiazinium derivatives that constitute a very relevant class of compounds with
biomedical applications. The phenothiazinium dyes are known to localize in the plasma membrane of
yeasts. Consequently, this is the cellular structure damaged upon illumination and it has been proposed
that the increased permeability resulting from such damage is the reason for cell death [7]. These com-
pounds have also been recently employed in a great number of studies with biological and chemical
relevance, such as development of new analytical methodologies aiming to detect compounds of envi-
Fig. 1. Structures of Azures A and B.

L.M. Moreira et al. / Azure dyes as new PS prototypes to application in PDT against Candida spp. 623
ronmental importance [12] as well as evaluations focused on the biochemical mechanisms of Alzheimer
disease [20]. Phenothiazinium salts offer more scope in terms of the therapy of disease states than other
dye types, including that employed as photosensitizers (PS). Thus, dyes as methylene blue and tolui-
dine blue O have been lead compounds in drug research against local bacterial infection, tuberculosis,
trypanosomiasis, malaria, Rickettsia illness, yeast infection, viral blood colonization and cancer [23]. In
this work, instrumental measurements of optical density were employed in order to determine the level
of growth inhibition of C. albicans. Indeed, the electronic absorbance in the spectral region of visible
was quite able to identify modifications in the presence of the microorganism. The results obtained in the
present article indicate that these compounds present significant potential to be used as photosensitizer
agents in PDT. The data are evaluated and discussed in details in agreement with recent reports from
literature.

Initially, a standardized suspension (106 viable cells ml−1 ) of C. albicans ATCC 10231 was prepared
as described in the literature [19]. A Candida suspension ((1–5) × 105 cells ml−1 ) was seeded and
incubated in the dark for 5 min at room temperature in the presence of different concentrations of phe-
nothiazinium derivatives Azure A, Azure B, Azure A thiocyanate, Azure B BF4 − and Azure A eosinate,
ranging from 0.01 to 0.5 mg · ml−1 , in a final volume of 0.2 ml. Cells incubated in sterile physiologi-
cal solution alone were included as a control. After this period, the covers of the 96-well plates were
removed, and the plates were illuminated with the appropriate light at room temperature, according to
the method described by Souza et al. [19]. The light source used was a diode laser InGaAlP (Photon
Laser, DMC, São Carlos, Brazil), with output power of 35 mW and wavelength of 684 nm. The energy
dose was of 28 or 56 J/cm2 , varying the time of irradiation. Aliquots of 50 µl were taken before and
after illumination so that we could determine both the number of colony forming units (CFUs). The
contents of the wells were properly homogenized before being sampled. To determine the number of
CFUs, we diluted aliquots 1000-fold in sterile physiological solution and spread them evenly on a Petri
dish containing Sabouraud dextrose agar medium. The colonies were incubated at 37◦ C for 48 h, and
the number of colony forming units per milliliter (CFUs ml−1 ) was determined. The experiments were
performed in the dark and under aseptic conditions. In the spectroscopic analysis, it was evaluated the
optical density at 570 nm, which is an interesting wavelength as function of the absorption of the several
species of aggregates, such as dimmers, trimers and oligomers that present wavelength of maximum ab-
sorption blue-shifted in relation to the monomeric form. In this way, even with significant aggregation,
a representative excitation of the photosensitizers would occur, favoring a higher photodynamic action.
Furthermore, in this wavelength (570 nm), the absorption of dyes inherent to the biological medium,
such as hemoglobin and melanin, is significantly low, implying that the competition by the photons
in the excitation process between Azures and biological dyes would be decreased. For consequence,
a suitable excitation would be obtained independently of the level of aggregation of the Azures. All the
experiments were made in triplicate, being that it was considered the medium value between the three
analysis.

Figure 2 presents the results regarding the effect of Azure B on Candida yeasts. It is possible to note
that Azure B does not present an effective inhibition of C. albicans growth.
Fig. 2. Azure B effect before (◦) and after the diode laser irradiation with DE of 28 J/cm2 (•) and 56 J/cm2 ( ⊗). Candida
albicans suspensions ((1–5) × 105 cells ml−1 ) were irradiated with diode laser (684 nm).
Fig. 3. Azure A effect before (◦) and after the diode laser irradiation with DE of 28 J/cm2 (•) and 56 J/cm2 ( ⊗). Candida
albicans suspensions ((1–5) × 105 cells ml−1 ) were irradiated with diode laser (684 nm).
Figure 3 demonstrates the data related to the Azure A effect on C. albicans cultures. Differently of the
Azure B behavior, it is detected an effective action of Azure A, when submitted to the radiation. In fact,
Azure A reduced the growth of C. albicans yeasts in approximately 30% after an irradiation 28 J/cm2
and near 50–60% subsequently to the irradiation 56 J/cm2 .
Figure 4 presents the behavior of Azure A thiocyanate. Interestingly, this compound demonstrates
an intermediary action, when compared with Azures A and B, since an irradiation 28 J/cm2 does not
precludes the growth of C. albicans, which only occurs with 56 J/cm2 .
Similarly to Azure B, Azure B TF4 does not inhibit the growth of the yeasts with both radiation
conditions evaluated in the present work, in agreement with the respective plot (Fig. 5).
Fig. 4. Azure A thiocyanate effect before (◦) and after the diode laser irradiation with DE of 28 J/cm2 (•) and 56 J/cm2 ( ⊗).
Candida albicans suspensions ((1–5) × 105 cells ml−1 ) were irradiated with diode laser (684 nm).
Fig. 5. Azure B tetrafluoroborate effect before (◦) and after the diode laser irradiation with DE of 28 J/cm2 (•) and 56 J/cm2
( ⊗). Candida albicans suspensions ((1–5) × 105 cells ml−1 ) were irradiated with diode laser (684 nm).
On the other hand, the Azure A eosinate behavior, presented in Fig. 6, is very close to that obtained
with Azure A, being characterized by 20% of growth inhibition with 28 J/cm2 and around 50–60% with
56 J/cm2 .
The present results demonstrate that the photodynamic action of distinct types of Azure (Azure A,
Azure B, Azure A thiocyanate, Azure B BF4 and Azure A eosinate) compounds is quite different regard-
ing the inhibition of C. albicans growth. Indeed, some phenothiazinium derivatives, such as Azure B and
Azure B BF4 , are not effective to inhibit C. albicans growth under both conditions of irradiation analyzed
in this work. However, Azure A and Azure A eosinate presented auspicious results when irradiated with
56 J/cm2 . These Azure compounds have significant potential to be employed as PS in PDT, especially
Fig. 6. Azure A eosinate effect before (◦) and after the diode laser irradiation with DE of 28 J/cm2 (•) and 56 J/cm2 ( ⊗).
Candida albicans suspensions ((1–5) × 105 cells ml−1 ) were irradiated with diode laser (684 nm).
in cases of superficial Candida infection, such as oral and vaginal candidosis, as well as nail and skin
infections. In fact, it is well known that the accessibility to the biological region affected is a funda-
mental pre-requisite to the effectiveness of PDT. This occurs because the local concentration of excited
species (photosensitizer) depends directly on the quantum yield and this phenomenon is consequence of
the light intensity in contact with the photosensitizer.
The phenothiazinium dyes have simple tricyclic planar structures, typically cationic in nature. The
most widely used compounds are methylene blue (MB) and toluidine blue O (TBO). Both are efficient
producers of singlet oxygen and the maximum absorption wavelength in water is 656 nm for MB and
625 nm for TBO, respectively [7].
According to Wainwright et al. [21], cationic photosensitizers are more efficient than their neutral or
anionic compounds in the photodynamic antimicrobial therapy. This information could be associated to
the fact that Azure A presents effective action against C. albicans in opposite to the Azure B behavior. In
fact, the molecular structure of Azure B presents a methyl radical as substituent group of the hydrogen
found in the position called “R3” in Azure A. Considering that this compound present a cationic charge,
the higher electronic density of Azure B could, through inductive effect, to decrease the intensity of this
positive charge. This fact is plausible due to the higher electronic density of the methyl group when
compared with the hydrogen. fungi present much more complex targets than bacteria. Yeasts possess a
thick external wall composed of a mixture of glucan, mannan, chitin and lipoproteins and separated from
the plasma membrane by a periplasmic space. Uptake of exogenous substances by fungi is generally
adversely affected by lipophilicity and positively affected by hydrophilicity and the presence of charged
groups [7].
In this context, it is relevant to note that the higher polarity of Azure A when compared with Azure B
must to favor the occurrence of higher number of hydrogen bonds with the chemical neighborhood
in aqueous medium. Thus, two important effects can be inferred of this significant physical–chemical
difference. The first one is the lower tendency of aggregation of Azure A in comparison with Azure B,
which is a well-known phenomenon associated to a decrease of the photodynamic efficacy. The second
one would be the higher tendency of Azure A to interact with the external membranes of Fungi due to its
higher polarity, which must to facilitate the interaction with the polar groups of the Candida membranes.
Donnelly et al. [7] inferred that there should be no difference in susceptibility to PACT between
organisms resistant or susceptible to conventional antifungals. This occurs because the oxidation caused
by the photosensitizer is non-specific and there are no cellular defenses against it. Besides, it is unlikely
that fungi could readily evolve resistance to singlet oxygen and mutagenesis has never been associated
to PDT.
C. albicans, as well as other yeasts are more resistant to PACT than gram-positive bacterial cells, ne-
cessitating higher drug and light doses [26]. This is probably due to the presence of a nuclear membrane,
to the greater cell size and the reduced number of targets for singlet oxygen per unit volume of cell [26].
Several studies have reported the effectiveness of PACT against Candida species [3,5,7]. However,
they are considerably less susceptible to photodynamic killing than several bacteria [25]. In fact, doses
of TBO as high as 2.0 mg · ml−1 have been required to induce high levels of inhibition (499%).
In conclusion, it is possible to infer that it would be plausible to obtain a higher inhibition of C. albi-
cans growth with Azure A and Azure A eosinate employing a more intense source of light or a higher
concentration of these phenothiazinium derivatives. This new trials are being developed and soon will
be published.
4. Conclusions
The present article evaluates a series of Azures compounds in order to identify new prototypes of
photosensitizer (PS). The results demonstrated that the different phenothiazinium derivatives have quite
distinct inhibitory behaviors regarding the C. albicans growth. The more effective results were obtained
with Azure A and Azure A eosinate. On the other hand, it is clear that some phenothiazinium compounds,
such as Azure B and Azure B BF4 , do not present real possibilities of application as photosensitizers.
In fact, the higher polarity of the molecular structure of Azure A when compared with Azure B must
be a decisive factor to propitiate this more effective photodynamic result. In this way, new tests are
necessary varying concentration and light intensity to obtain a more complete analysis of the potential
as photosensitizer of Azure A and Azure A eosinate, which were the more effective compounds analyzed
in this manuscript. Furthermore, it is important to notice that the spectroscopic measurements of optical
density was quite capable to determine the growth inhibition of C. albicans, and, in this way, can be
considered an interesting alternative in comparison with conventional methodologies in order to develop
microbiological analyzes.
Acknowledgements
The authors thank to FAPESP for financial grant (J.P. Lyon, 2007/07162-5; M.R. Rodrigues,
2002/00272-6; H.P.M. de Oliveira, 06/56701-3; L. Codognoto, 08/50588-6) and CNPq (H.P.M. de
Oliveira, 479655/2008-1). Thanks to A. Lima for assistance in the laboratory.
References
[1] A. Asilian and M. Davami, Comparison between the efficacy of photodynamic therapy and topical paromomycin in the
treatment of Old World cutaneous leishmaniasis: a placebo-controlled, randomized clinical trial, Clin. Exp. Dermatol. 31
(2006), 634–637.
[2] M. Bhatti, A.M. Robert, S. Meghji, B. Henderson and M. Wilson, Effect of dosimetric and physiological factors on the
lethal photosensitization of Porphyromonas gingivalis in vitro, Photochem. Photobiol. 65 (1997), 1026–1031.
[3] J.M. Bliss, C.E. Bigelow, T.H. Foster and C.G. Haidaris, Susceptibility of Candida species to photodynamic effects of
photofrins, Antimicrob. Agents Chemother. 48 (2004), 2000–2006.
[4] W.C. Bristo, R. Hudson, T.A. Paget and R.W. Boyle, Potential of cationic porphyrins for photodynamic treatment of
cutaneous leishmaniasis, Photodiag. Photodyn. Ther. 3 (2006), 162–167.
[5] S.C. de Souza, J.C. Junqueira, I. Balducci and C.Y. Koga-Ito, Photosensitisation of different Candida species by low
power laser light, J. Photochem. Photobiol. B 83 (2006), 34–38.
[6] J. Dobson and M. Wilson, Sensitization of oral bacteria in biofilms to killing by light from a low-power laser, Arch. Oral
Biol. 37 (1992), 883–887.
[7] R.F. Donnelly, P.A. McCarron, M.M. Tunney and A.D. Woolfson, Potential of photodynamic therapy in treatment of
fungal infections of the mouth. Design and characterization of a mucoadhesive patch containing toluidine blue O, J. Pho-
tochem. Photobiol. B 86 (2007), 59–69.
[8] A.S. Garcez, F.R. Souza, S.C. Núñez, J.M. Kather and M.S. Ribeiro, Terapia fotodinâmica em odontologia – laser de
baixa potência para redução microbiana, Rev. APCD 57 (2003), 223–226.
[9] L. Gonçalves, Efeito de fotoativadores utilizados na irradiação laser intracanal, MSc dissertation (Mestrado em Odontolo-
gia), Universidade de São Paulo, 2005.
[10] J.P. Lyon, S.C. Costa, V.M.G. Totti, M.F.V. Munhoz and M.A. Resende, Predisposing conditions for Candida spp. carriage
in the oral cavity of denture wearers and individuals with natural teeth, Can. J. Microbiol. 52 (2006), 462–467.
[11] L.M. Moreira, F.V. Santos, J.P. Lyon, M.M. Costa, C.P. Soares and N.S. Silva, Photodynamic therapy (PDT): porphyrins
and phthalocyanines as photosensitizers (PS), Aust. J. Chem. 61 (2008), 741–754.
[12] B. Narayana, T. Cherian and A. Facile, Spectrophotometric method for the determination of periodate using Azure B,
J. Braz. Chem. Soc. 1 (2005), 978–981.
[13] P.M. Pinto, R.C.B.W. Oliveira, J.P. Lyon, V.F. Cury, R.R. Arantes, C.Y. Koga-Ito and M.A. Resende, In vitro antifungal
susceptibility of clinical isolates of Candida spp. obtained from patients with different predisposing factors to candidosis,
Microbiol. Res. 163 (2008), 579–585.
[14] M.A. Resende, L.V.N. Franco, R.C.B.W. Oliveira, C.Y. Koga-Ito and J.P. Lyon, Prevalence and antifungal susceptibility
of yeasts obtained from the oral cavity of elderly individuals, Mycopathologia 162 (2006), 39–44.
[15] C. Romagnoli, D. Mares, G. Sacchetti and A. Bruni, The photodynamic effect of 5-(4-hydroxy-1-butinyl)-2,2-biethienyl
on dermatomycetes, Mycol. Res. 102 (1998), 1519–1524.
[16] P. Saballs, J.M.T. Rodrigues and M. Salvadó, La candidemia en el síndrome de inmunodeficiencia adquirida. Estudio
retrospectivo de nueve casos, Rev. Iberoam. Micol. 17 (2000), 2–5.
[17] D.J. Sheehan, C.A. Hitchcock and C.M. Sibley, Current and emerging azole antifungal agents, Clin. Microbiol. Rev. 12
(1999), 40–79.
[18] N.S. Silva, C.M. Ribeiro, A.H.A. Machado and C. Pacheco-Soares, Ultrastructural changes in Tritrichomonas foetus after
treatments with AlPcS4 and photodynamic therapy, Vet. Parasitol. 146 (2007), 175–181.
[19] S.C. Souza, J.C. Junqueira, I. Balducci, C.Y. Koga-Ito, E. Munin and A.O.C. Jorge, Photosensitization of different Can-
dida species by low power laser light, J. Photochem. Photobiol. B 12 (2006), 34–38.
[20] S. Taniguchi, N. Suzuki, M. Masuda, S. Hisanaga, T. Iwatsubo, M. Goedert and M. Hasegawa, Inhibition of heparin-
induced tau filament formation by phenothiazines, polyphenols, and porphyrins, J. Biol. Chem. 280 (2005), 7614–7623.
[21] M. Wainwright, Photodynamic antimicrobial chemotherapy, J. Antimicrob. Chemother. 42 (1998), 13–19.
[22] M. Wainwright, Phenotiazinium photosensitizers: V. Photobactericidal activities of chromophore-methylated phenoti-
azinium salts, Dyes Pigm. 73 (2007), 7–12.
[23] M. Wainwright and R.M. Giddens, Phenotiazinium photosensitizers: choices in synthesis and applications, Dyes Pigm.
57 (2003), 245–257.
[24] I.C.J. Zanin, A. Brugnera Júnior and R.B. Gonçalves, Susceptibility of Streptococcus mutans biofilms to photodynamic
therapy: an in vitro study, J. Antimicrob. Chemother. 56 (2005), 324–330.
[25] B. Zeina, J. Greenman, D. Corry and W.M. Purcell, Cytotoxic effects of antimicrobial photodynamic therapy on ker-
atinocytes in vitro, Braz. J. Dermatol. 146 (2002), 568–573.
[26] B. Zeina, J. Greenman, W.M. Purcell and B. Das, Killing of cutaneous microbial species by photodynamic therapy, Braz.
J. Dermatol. 144 (2001), 274–278.
Spectroscopy 24 (2010) 609–619 609
DOI 10.3233/SPE-2010-0483
IOS Press
Direct identification of potato’s fungal

phyto-pathogens by Fourier-transform
infrared (FTIR) microscopy
V. Erukhimovitch a , L. Tsror (Lahkim) b , M. Hazanovsky b and M. Huleihel c,∗
a
Analytical Equipment Unit, Ben-Gurion University of the Negev, Beer-Sheva, Israel
b
Department of Plant Pathology, The Institute of Plant Protection,
Agricultural Research Organization, Gilat Experiment Station, M.P. Negev, Israel
c
Department of Virology and Developmental Genetics, Faculty of Health Sciences,
Ben-Gurion University of the Negev, Beer-Sheva, Israel
Abstract. Fungi are considered as serious pathogens to many plants and can cause a severe economic damage. The available
methods for identification of fungi are time consuming and not always very specific. In the present study we examined the
potential of FTIR microscopy for direct detection and identification of different fungal potato pathogens on the surface of
potato tubers. Unique spectral bands for each of the examined fungal pathogens appeared in the spectra of naturally infected
potatoes. These results strongly support the potential of FTIR microscopy for successful detection and probably discrimination
between different fungal pathogens directly from the infected tissue.
Keywords: Fungal detection, fungi, FTIR microscopy, spectral peaks, potato
1. Introduction
Fungal pathogens cause a severe disease to various plants resulting a serious damage to large number
of crops with a significant negative feedback on economy [13]. Early and reliable diagnosis methods
allow the monitoring of the pathogen and enable farmers to administer suitable management strategies
in a timely fashion and would greatly increase the effectiveness of the treatment [13]. There are many
methods of identification of fungal phyto-pathogens, but part of them either are not fast, or reliable
enough and many are expensive. Culturing of the pathogens on selective nutrient media, and morphol-
ogy examination of the fungal colony using an optical microscopy [11] is among the current classical
methods used for detection and identification of fungi. These procedures are time consuming, not spe-
cific and at times difficult to interpret. The enzyme-linked immunosorbent assay (ELISA) has been used
as a diagnostic tool in medicine as well as in plant pathology, however, the most controversial aspect
of this test is determining the “cut-off” point between a positive and negative result. Polymerase chain
reaction (PCR) method is considered as a rapid and sensitive for the detection and identification of some
*
Corresponding author: M. Huleihel, Department of Virology and Developmental Genetics, Faculty of Health Sciences, Ben-
Gurion University of the Negev, Beer-Sheva, Israel. Tel.: +972 8 6479867; Fax: +972 8 6479867; E-mail: mahmoudh@bgumail.
bgu.ac.il.
610 V. Erukhimovitch et al. / Direct identification of potato’s fungal phyto-pathogens by FTIR microscopy
plant pathogenic fungi [3,6,25]. Although this method is promising, it is not yet applied in a large-scale
and still with high expenses.
Fourier transform infrared (FTIR) spectroscopy is one of the methods that have been successfully used
for detecting and identifying of cancer cells [12,22], cells infected with viruses [23] and microorganisms
including some fungi [7,8,10,18–21]. Some of these studies with microorganisms showed that discrimi-
nation was possible not only at the genus level, but also at the species level [2,15]. The vast majority of
these studies were conducted with bacteria and very few studies involved fungi. The potential of FTIR
spectroscopy as a tool for detection and identification of fungal pathogens in agriculture promises to be
of a great value because of its sensitivity, rapidity, low expense and simplicity [14]. This together with
the large information already known about spectral peaks obtained from FTIR spectra of living cells [4],
make FTIR spectroscopy as an attractive technique for detection and identification of pathogens. This
technique appears to be also as a very promising tool for the study of fungal metabolism and interactions
with drugs [21,24,25,27].
The present study focuses on investigating the use of FTIR microscopy for rapid discrimination of
three potato fungal pathogens directly on the surface of the intact potato tuber.
2.1. Fungi
In the present study we examined three different potato fungal pathogens [Colletotrichum coccodes
(Col), Rhizoctonia (Rh) and Helminthosporium (Hel)] which cause a serious damage to potatoes world-
wide. Ten pathogens were isolated from naturally infected potato tubers. The fungi were cultured on
potato dextrose agar (PDA) (Difco) and incubated in the dark at 25◦ C for 7 days. Conidia were trans-
ferred to plates of solid SA medium (0.2% sorbose, 1.5% agar, 100 p.p.m. streptomycin sulphate) and
incubated for 24 h at 25◦ C in the dark. Monoconidial cultures were obtained from each isolate (by
micromanipulation) and maintained on Czapek dox agar (CDA) (Difco).
These fungal pathogens were examined by FTIR microscopy both directly from the infected potato
tubers and also from the cultivated isolates.
Ten samples from different infected potato tubers with each of these fungal pathogens were examined.
2.2. Sample preparation
Since ordinary glass slides exhibit strong absorption in the wavelength range of interest to us, we used
zinc selenide crystals, which are highly transparent to IR radiation.
2.3. Preparation of infected potato and disease-free tissue samples
Thin samples (about 20 µm thickness) of samples were scratched from the infected or uninfected areas
on the surface of potato tubers, suspended in 100 µl of distilled sterile H2 O, pelleted by centrifugation
at 1000 rpm for 2 min. Each pellet was suspended with 20 µl of dH2 O and a drop of 1 µl of the ob-
tained suspension was placed in a certain area on the zinc selenide crystal, air dried for 15 min at room
temperature (or for 5 min by air drying in a laminar flow) and examined by FTIR microscopy.
V. Erukhimovitch et al. / Direct identification of potato’s fungal phyto-pathogens by FTIR microscopy 611
2.4. Preparation of purified fungal samples
A small samples of fungal pathogen were obtained from infected areas of potato tubers and grown
on potato dextrose agar as mentioned above. Samples of these fungi were purified from the media by
spinning about 1 ml of medium containing fungi at 2000 rpm for 5 min, washing twice with H2 O and
the pellet was suspended in appropriate volume of H2 O (about 50 µl). A drop of 1 µl of the obtained
suspension was placed on the zinc sellenide crystal, air dried and examined by FTIR microscopy.
2.5. FTIR spectra measurement
FTIR measurements were performed in the transmission mode with a liquid-nitrogen-cooled MCT
detector of the FTIR microscope (Bruker IRScope II) coupled to an FTIR spectrometer (BRUKER
EQUINOX model 55/S, OPUS software). The spectra were obtained in the wavenumber range of 600–
4000 cm−1 . Spectral resolution was set at 4 cm−1 . Baseline correction by the rubber band method and
vector normalization were obtained for all the spectra by OPUS software. Peak positions were deter-
mined by means of a second derivation method by OPUS software. Since the samples to be analyzed
were often heterogeneous, appropriate regions were chosen by FTIR microscopy so as to eliminate dif-
ferent impurities (salts, medium residuals, etc.). The aperture used in this study was 50 µm, since this
aperture gave opportunity to choose appropriate regions only from the epidermis of potato tubes without
including inner parts and still good signal/noise ratio. For each sample, the spectrum was taken as the
average of five different measurements at various sites of the sample. Each experiment with each sample
was repeated 10 times. It is important to mention that there were no significant differences in the spectra
from various sites (SD did not exceed 0.005).
2.6. Cluster analysis
The obtained spectral results of infected and control uninfected potato tissues were classified using
cluster analysis. Cluster analysis was performed according to Ward’s algorithm by OPUS software.
3. Results
3.1. FTIR spectra of purified fungi and control potatoes
The main objective of this study is examining the potential of FTIR-M for identification of fungal
pathogens directly on the potatoes tubers. Since the fungi is growing on the potato surface, part of
the obtained spectra of the infected area will be contaminated with spectral bands of potato sample.
Therefore, it is necessary as a first step to find out the main specific fungal peaks which do not exist in
the potato samples and to focus on such bands when searching for specific biomarkers for the detection
and identification of fungal pathogens directly from potato tissues. So, we first compared the FTIR
spectra of purified fungal pathogens, which were isolated and grown in standard culture medium, with
the spectra of tissue obtained from the epidermis of uninfected tissue. Specific fungal biomarkers for
each examined pathogen, which are significantly distinctive from the spectra of the tissue, will be used
as a good reference for detecting and identifying the fungal pathogen in infected potato tissues. The
results presented in Fig. 1 show the average FTIR spectra of the tested samples. It can be seen that there
are significant differences between the spectra of the purified fungi and the control potato tissue on one
Fig. 1. FTIR spectra of purified examined fungal pathogens grown in standard medium and samples obtained from uninfected
control potatoes. The obtained spectra were examined at various regions as follows: (A) 600–2000 cm−1 ; (B) 920–940 cm−1 ;
(C) 1015–1040 cm−1 ; (D) 1060–1100 cm−1 ; (E) 1220–1340 cm−1 and (F) 1530–1560 cm−1 . Data are means of 10 different
and separate experiments for each sample.
hand and a unique spectrum for each fungi with specific differences between them at various regions
over the spectrum on the other hand.
The major consistent differences between the fungal and potato samples are as follow:
(A) Spectral bands which significantly appeared in control potato samples.
(1) Band at 1033 cm−1 (Fig. 1C). This band is assigned to carbohydrates and completely missing
in all fungal samples.
(2) Bands at 1259 cm−1 , assigned to syringyl ring and C–O stretch in lignin and xylan as reported
previously [14] and at 1330 cm−1 (attributed to amide III) (Fig. 1E) are strong in potato
whereas, they are expressed as a slight shoulders in fungi.
(B) Spectral bands which appeared in all examined fungal pathogens samples while missing or ex-
pressed at very low intensity in control potatoes samples.
(1) Band at 1078 cm−1 (Fig. 1D). This band represents carbohydrates and it seems that fungi
cells contain different carbohydrates compared to epidermal potato cells.
(2) Band at 1542 cm−1 (Fig. 1F). This band is attributed to amide II and significantly appeared in
fungal samples due to the high levels of fungal proteins compared to protein levels in potato
epidermal tissues. It seems that the level of protein in potato epidermal tissue is very low.
This result is in agreement with the results obtained previously by Gordon et al. [9] who
studied the spectral features of fungal corn and with the results of Naumann et al. [21] who
examined fungi in wood.
(C) Spectral bands which are specific to each of the examined fungal pathogens.
(1) Band at 933 cm−1 (assigned to carbohydrates) is specific for Col fungi (Fig. 1B).
(2) Band at 1026 cm−1 (assigned to carbohydrates) appeared only in the spectra of fungal
pathogen Rh (Fig. 1C).
(3) Band at 1543 cm−1 , which is attributed to protein amide II band [26], significantly appeared
only in the spectra of Hel fungi (Fig. 1F).
Potato tissues are highly rich with starch which is considered as the main component of potato tissues
compared to other components such as proteins and lipids. It seems that most of the differences between
the potato epidermal and the fungal spectra are due to the high levels of starch in potato tissue while
starch is missing in fungi. As can be seen in Fig. 2 most of the bands in the spectrum of potato epidermal
tissue are dominated by the spectral bands of the purified starch and might be affected and contributed
by these bands.
These results provide a preliminary indication for possible spectral markers for the identification of
fungi obtained directly from the surface of infected potatoes.
3.2. Examination of potato samples infected with fungi
In the present study we used control uninfected potato tubers and others which are naturally infected
with the various examined fungal pathogens. The identification of these pathogens was done by stan-
dard methods as detailed in Section 2. Samples from the control uninfected potatoes were grown in the
appropriate growth media in order to confirm the absence of fungal infection. For FTIR microscopy
examination, thin potato samples were prepared directly from the surface of control uninfected and in-
fected potatoes as described in M&M. The results presented in Fig. 2 show the FTIR spectra of both
Fig. 2. FTIR spectra in the region of 600–2000 cm−1 of pure starch and samples obtained from control uninfected potato
epidermis. Results are means of 20 different and separate experiments for each sample. The SD for these means was 0.001.
uninfected and infected samples obtained from potatos tubers. Although there is a very high similarity
between the spectra of the infected and the uninfected tissues, specific consistent spectral bands appeared
in the spectra of each of the infected tissues with the different tested fungal pathogens.
These spectral bands seem to be specific to the examined fungus as detailed below:
(1) Two specific bands at 932 and 1537 cm−1 appeared in infected potato samples with Col fungus
while are missing or very weakly expressed in the control uninfected samples and in those infected
with the other fungi strains (Fig. 3C and F respectively, Table 1). These peaks have been shown
above to be specific to the pure Col fungi (Fig. 1B and F).
(2) Band at 1028 cm−1 appeared in infected potato samples with Rh fungi. This peak is missing in
the control uninfected samples and in those infected with the other fungi isolates (Fig. 3D and
Table 1). This band just fits well with the specific peak of the pure Rh fungus mentioned above
(Fig. 1C).
(3) A significant band at 756 cm−1 appeared consistently in all examined infected samples with Hel
fungi (Fig. 3B and Table 1) while it is missing in the control uninfected samples and in those
infected with the other fungi isolates. It is interesting to mention that this peak was not seen in
the pure Hel fungal spectrum. Using these biomarkers, fully and excellent classification between
the control and the infected samples with the different fungal pathogens was obtained by cluster
analysis. It can also be seen that this classification method provides very good accuracy in dif-
ferentiating between the various examined fungal isolates at the different fungal specific bands
mentioned above (Tables 2–4).
Fig. 3. FTIR spectra of control uninfected potato samples and potato samples infected with the various examined
fungal pathogens obtained by scratching technique. The obtained spectra were examined at various regions as fol-
low: (A) 600–2000 cm−1 ; (B) 750–760 cm−1 ; (C) 920–940 cm−1 ; (D) 1020–1040 cm−1 ; (E) 1060–1100 cm−1 and
(F) 1530–1555 cm−1 . Data are means of 10 different and separate experiments for each sample.
Table 1
Peak area of normal uninfected and infected potato samples with the different examined fungal
pathogens at various spectral regions
Peak area at
756 cm−1 932 cm−1 1028 cm−1 1537 cm−1

Control potato 0.0000 0.0001 0.0005 0.0001
Infected with Col 0.0005 0.0220 0.0000 0.0430
Infected with Rh 0.0001 0.0005 0.0350 0.0070
Infected with Hel 0.0680 0.0000 0.0002 0.0000
Note: The results are means of 10 different experiments (SD 5%).
Table 2
Relative identity (determined by cluster analysis) of normal uninfected and infected potato sam-
ples with the different examined fungal pathogens at the region 756 cm−1
Control potato Infected with Infected with Infected with
(%) Col (%) Rh (%) Nel (%)
Control potato 100 98 99 0
Infected with Col 98 99 96 3
Infected with Rh 99 96 100 1
Infected with Hel 0 3 1 100
Table 3
ples with the different examined fungal pathogens at the regions 932 and 1537 cm−1
(%) Col (%) Rh (%) Hel (%)
932 1537 932 1537 932 1537 932 1537

Control potato 100 100 0 0 95 74 100 68
Infected with Col 0 0 100 100 5 19 0 17
Infected with Rh 95 26 5 19 100 94 96 78
Infected with Hel 100 68 0 17 96 93 100 96
Table 4
ples with the different examined fungal pathogens at the regions 1028 cm−1
(%) Col (%) Rh (%) Hel (%)
Control potato 100 95 0 96
Infected with Col 95 100 3 97
Infected with Rh 0 3 99 4
Infected with Hel 96 97 0 100
4. Discussion
In the present study we examined the potential of FTIR microscopy for an easy and rapid detection
and identification of fungal pathogens directly from the surface of infected potatoes, as a representative
model for an in vivo rapid detection of plant fungal pathogens. Fungal pathogens are involved in various
plant diseases which can cause a serious damage to large number of crops [13]. Therefore, early and
reliable diagnosis methods might be critical for controlling the spread of these pathogens [13].
The main routinely used identification method of fungi is based on their physiological properties
which are time consuming, not specific and some times difficult to interpret [11].
The results obtained in this study provide unique, consistent and significant spectral biomarkers for
the identification of infected regions with different fungal pathogens. Two specific bands at 932 and
1537 cm−1 seem to be specific to Col fungi because they were missing or very weakly expressed both
in the control uninfected and in all the other infected samples with the different fungi strains (Fig. 2C
and F, Table 1). These peaks have been shown also to be specific to the pure Col fungi (Fig. 1B and F).
A significant band at 1028 cm−1 appears to be specific to Rh fungi (Fig. 3D and Table 1) and also fits
well with the pure Rh fungi (Fig. 1C). However, a significant band at 756 cm−1 appeared consistently
only in all examined infected samples with Nel fungi (Fig. 3B and Table 1) but it was not seen in the
pure Hel fungi spectrum. Its significant appearance in infected potatoes is probably a result of a specific
enzymatic degradation of potato component (most likely starch).
These specific spectral bands can be considered as good biomarkers for the detection of the appropriate
fungal pathogen isolates. Although it is well known that fungal pathogens have various enzymes that
can specifically digest cellulose, pectin, starch and other potato components [1], it seems that most of
the above discussed spectral bands are not a result of such specific or nonspecific degradation due to
the following reasons: (a) these bands appeared consistently in all examined infected samples of each
pathogen; (b) they are different according to the infecting fungi and specific to that fungi and (c) these
bands (except the band at 756 cm−1 which appeared consistently in all examined infected samples with
Hel fungi) fits very well with specific peak of the pure appropriate fungus (as can be seen in Fig. 1).
Using these biomarkers it was possible to successfully discriminate between the different examined
fungi by cluster analysis (Tables 2–4).
In addition, it is worthwhile to mention that the spectra of the uninfected control potato tubers obtained
in this study were very similar to the spectra of uninfected wood previously published by Naumann et al.
[21]. Whereas, the spectra of potato infected with the examined different fungal pathogens were com-
pletely distinguished from the spectra of infected woods with different fungi as presented by Naumann
et al. [21]. These results together may indicate the potential of FTIR microscopy for rapid and accurate
detection and discrimination between various plant fungal pathogens.
Additionally, the fact that the final results could be obtained from the infected tuber during a very short
time (approximately 30 min) from a small amount of sample and by a simple procedure, support the pos-
sibility of developing FTIR spectroscopy as a reliable method for rapid identification and discrimination
between fungal pathogens.
5. Conclusions
In the present study we examined the potential of FTIR microscopy for an easy and rapid detection
and identification of fungal pathogens directly from the surface of infected potatoes, as a represen-
tative model for in vivo rapid detection of plant fungal pathogens. The results obtained in this study
provide unique, consistent and significant spectral biomarkers for infected region with different fungal
pathogens. Using these biomarkers it is possible to discriminate between the different examined fungi.
Additionally, the fact that the final results could be obtained from the infected tuber during a very short
time (approximately 30 min) from a small amount of sample and by a simple procedure, support the pos-
sibility of developing FTIR spectroscopy as a reliable method for rapid identification and discrimination
between fungal pathogens in plants infections.
References
[1] G.N. Agrios, Plant Pathology, 4th edition, Academic Press, San Diego, CA, 1997.
[2] S.H. Beattie, C. Holt, D. Hirst and A.G. Williams, Discrimination among Bacillus cereus, B. mycoides and B. thuringiensis
and some other species of the genus Bacillus by Fourier transform infrared spectroscopy, FEMS Microbiol. Lett. 164
(1998), 201–203.
[3] P.W.G. Chu, P.M. Waterhouse, R.R. Martin and W.L. Gerlach, New approaches for the detection of microbial plant
pathogens, Genetic Eng. Rev. 7 (1989), 45–61.
[4] M. Diem, S. Boydstom-White and L. Chiriboga, Infrared spectroscopy of cells and tissues: shining light onto a novel
subject, Appl. Spectroscopy 53 (1999), 148–161.
[5] R.K. Dukor, Vibrational spectroscopy in the detection of cancer, in: Handbook of Vibrational Spectroscopy, Wiley, Chich-
ester, UK, 2002, pp. 3335–3345.
[6] D. Errampalli, J. Saunder and D. Cullen, A PCR-based method for detection of potato pathogen, Helminthosporium
solani, in silver scurf infected tuber tissue and soil, J. Microbiol. Methods 44 (2001), 59–68.
[7] V. Erukhimovitch, L. Tsor, M. Hazanovsky, M. Talyshinsky, I. Mukmanov, Y. Souprun and M. Huleihel, Identification of
fungal phyto-pathogens by Fourier-transform infrared (FTIR) microscopy, J. Agric. Technol. 1 (2005) 145–152.
[8] V. Erukhimovitch, L. Tsor, M. Hazanovsky, M. Talyshinsky, I. Mukmanov, Y. Souprun and M. Huleihel, Early and rapid
detection of potato’s fungal infection by Fourier-transform infrared (FTIR) microscopy, Appl. Spectrosc. 61 (2007), 1052–
1056.
[9] S.H. Gordon, R.B. Schudy, B.C. Wheeler, D.T. Wicklow and R.V. Greene, Identification of Fourier transform infrared
photoacoustic spectral features for detection of Aspergillus flavus infection in corn, Int. J. Food Microbiol. 47 (1997),
5267–5272.
[10] S.H. Gordon, R.W. Jones, J.F. McClell, D.T. Wicklow and R.V. Greene, Transient infrared spectroscopy for detection of
toxigenic fungi in corn: potential for on-line evaluation, J. Agric. Food Chem. 47 (1999), 5267–5272.
[11] R.C. Hamelin, P. Berube, M. Gignae and M. Bourassa, Identification of root rot fungi in nursery seedlings by nested
multiplex PCR, Microbiology 62 (1996), 4026–4031.
[12] M. Huleihel, V. Erukhimovitch, M. Talyshinsky and M. Karpasas, Spectroscopic characterization of normal primary and
malignant cells transformed by retroviruses, Appl. Spectroscopy 56 (2002), 640–645.
[13] Y. Katan, Principles in plant pathology, in: J. Rotem, J. Palti and Y. Ben-Yephet, eds, Volcani Center, Bet-Dagan, Israel,
1998, pp. 25–43.
[14] M. Kummerle, S. Scher and H. Seiler, Rapid and reliable identification of food borne yeasts by Fourier-transform infrared
spectroscopy, Appl. Environ. Microbiol. 64 (1998), 2207–2214.
[15] D. Lefier, D. Hirst, C. Holt and A.G. Williams, Effect of sampling procedure and strain variation in Listeria monocy-
togenes on the discrimination of species in the genus Listeria by Fourier transform infrared spectroscopy and canonical
variates analysis, FEMS Microbiol. Lett. 147 (1997), 45–48.
[16] H. Lamprell, G. Mazerolles, A. Kodjo, J.F. Chamba, Y. Noe and E. Beuvier, Discrimination of Staphylococcus aureus
strains from different species of Staphylococcus using Fourier transform infrared (FTIR) spectroscopy, Int. J. Food Microb.
108 (2006), 125–129.
[17] R. Linker and L. Tsror, Discrimination of soil-borne fungi using Fourier transform infrared attenuated total reflectance
spectroscopy, Appl. Spectrosc. 62 (2008), 302–305.
[18] L. Mariey, J.P. Signolle, C. Amiel and J. Travert, Discrimination, classification, identification of microorganisms using
FTIR spectroscopy and chemometrics, Vib. Spectrosc. 26 (2001), 151–159.
[19] K. Maquelin, C. Kirschner, L.P. Choo-Smith, N.A. Ngo-Thi, V. Vreewijk, M. Stammler, H.P. Endtz, H.A. Bruining,
D. Naumann and G.J. Puppels, Prospective study of the performance of vibrational spectroscopies for rapid identification
of bacterial and fungal pathogens recovered from blood cultures, J. Clin. Microbiol. 41 (2003), 324–329.
[20] D. Naumann, The characterisation of microorganisms by Fourier-transform infrared spectroscopy (FT-IR), in: Modern
Techniques for Rapid Microbiological Analysis, W.H. Nelson, ed., VHC, New York, 1991, pp. 43–55.
[21] A. Naumann, M. Navarro-Gonzalez, S. Peddireddi, U. Kues and A. Polle, Fourier transform infrared microscopy and
imaging: Detection of fungi in wood, Fungal Genet. Biol. 42 (2005), 829–835.
[22] B. Rigas, K. LaGuardia, L. Qiao, P.S. Bhandare, T. Caputo and M.A. Cohenford, Infrared spectroscopic study of cervical
smears in patients with HIV: implications for cervical carcinogenesis, J. Lab. Clin. Med. 35 (2000), 26–31.
[23] A. Salman, V. Erukhimovitch, M. Talyshinsky, M. Huleihil and M. Huleihel, FTIR-spectroscopic method for detection of
cells infected with herpes viruses, Biopolymers 67 (2002), 406–412.
[24] G.D. Sockalingum, E.W. Bouhedja, P. Allouch, C. Mandray, R. Labia, J.M. Millot and M. Manfait, ATR–FTIR spectro-
scopic investigation of imipenem-susceptible and resistant Pseudomonas aeruginosa isogenic strains, Biochem. Biophys.
Res. Commun. 232 (1997), 240–246.
[25] M.K. Winson, R. Goodacre, E.M. Timmins, A. Jones, B.K. Alsberg, A.M. Woodward, J.J. Rowland and D.B. Kell, Diffuse
reflectance absorbance spectroscopy taking in chemo-metrics (DRASTIC). A hyperspectral FT-IR based approach to rapid
screening for metabolite overproduction, Anal. Chim. Acta 348 (1997), 273–282.
[26] T.J. White, T. Bruns, S. Lee and J. Taylo, Amplification of direct sequencing of fungal ribosomal RNA genes for phy-
logenetics, in: PCR Protocols: A Guide to Methods and Applications, M. Innis, ed., Academic Press, San Diego, 1990,
pp. 315–327.
[27] W. Zeroual, C. Choisy, S.M. Doglia, H. Bobichon, J.F. Angiboust and M. Manfait, Monitoring of bacterial growth and
structural analysis as probed by FT-IR spectroscopy, Biochim. Biophys. Acta 1222 (1994), 171–178.
Spectroscopy 24 (2010) 601–608 601
DOI 10.3233/SPE-2010-0482
IOS Press
Fourier transform infrared and near-infrared

spectroscopic methods for the detection of
toxic Diethylene Glycol (DEG) contaminant
in glycerin based cough syrup
M. Khalique Ahmed a,∗ , Michael P. McLeod b , Jean Nézivar a and Allison W. Giuliani c
a
Department of Science, College of Liberal Education, Lynn University, North Military Trail,
Boca Raton, FL, USA
b
University of Miami, Miller School of Medicine at Florida Atlantic University, Boca Raton, FL, USA
c
Department of Biological Sciences, Florida Atlantic University, Boca Raton, FL, USA
Abstract. Recently there have been reports of the contamination of cough syrups with Diethylene Glycol (DEG). The consump-
tion of such cough syrups has devastating effects on the health. In this paper we report evidence that Fourier transform infrared
(FT-IR) and near-infrared (NIR) spectroscopic techniques are viable, simple, cost effective, rapid and fool proof methods for
the identification and quantification of DEG in glycerin based cough syrups. The FT-IR and NIR spectra of the glycerin based
cough syrup and up to 50:50 mixtures of DEG in cough syrup are recorded. The major peaks in the FT-IR spectrum of the
cough syrup are assigned to the OH stretching (∼3300 cm−1 ), CH stretching (∼2900 cm−1 ), CH bending (1500–1200 cm−1 )
and C–O stretching (1200–900 cm−1 ) vibrational modes. In the FT-IR spectra of the mixtures, DEG contribute distinct peaks
due to the vibrations of the C–O (920 cm−1 ) and OC2 H4 (892 cm−1 ) moieties of its backbone and form the basis of the DEG
detection and quantification. The prominent peaks of the NIR spectra of cough syrup and DEG are assigned to the first over-
tones of OH and CH, and to the combination of OH and CH fundamental vibrations. Both FT-IR and NIR Partial Least Square
(PLS) calibrations produced correlation coefficients of 0.98.
Keywords: Infrared spectroscopy, near-infrared spectroscopy, cough syrup, diethylene glycol, contaminant, detection
1. Introduction
Diethylene Glycol (DEG) is an inexpensive, sweet tasting [1–3], industrial solvent commonly used
in antifreeze solutions [1,4]. It is often substituted for more expensive chemicals [1,4] such as glycerin
and propylene glycol [3,6] in pharmaceutical products such as tooth paste, injectable drugs, fever med-
ication and cough syrup [8]. DEG has a molecular weight of 106.12 g/mol, a boiling point of 245◦ C,
a melting point of −6.5◦ C and a vapor pressure of <0.01 mmHg at 25◦ C [8]. It has been implicated
in a number of serious world-wide contaminations with devastating human consequences; the first well
documented case being the Massergil Tragedy of 1937 in the United States [5,6,9]. This tragedy was
the stimulus behind passing the 1938 Federal Food, Drug and Cosmetic Act requiring product safety
before being marketed [8,9]. Most recently in December 2008, Cable News Network (CNN) reported
*
Corresponding author: M. Khalique Ahmed, Department of Science, College of Liberal Education, Lynn University, 3601
North Military Trail, Boca Raton, FL 33431, USA.
602 M. Khalique Ahmed et al. / Toxic DEG contaminant in glycerin based cough syrup
dozens of children deaths in Nigeria due to DEG poisoning. The symptoms of DEG poisoning include
acute renal failure [1,3,5,6,10], metabolic acidosis [1,5,6], diarrhea [1,6], vomiting [1,3,6], and delayed
(1–2 weeks post ingestion) neurotoxicity [4,9], i.e., optic neuritis, demyelination of peripheral and cra-
nial nerves, and cerebral atrophy. The minimum toxic dose is not readily agreed upon; however, the
median toxic dose for DEG has been reported to be 1.34 ml/kg with a range from 0.22–4.42 ml/kg
[7]. DEG has been postulated to be toxic when converted by alcohol and aldehyde dehydrogenases to
2-hydroxyethoxyacetic acid (HEAA) [5,11]. DEG poisoning is best treated when discovered early [12–
15] in the pathogenesis by implementing the alcohol dehydrogenase inhibitor fomepizole [3,5], ethanol
infusion [2,15] and hemodialysis [3].
In the current study, we report evidence that FT-IR and NIR spectroscopic techniques are viable
screening tools in the identification and quantification of DEG in glycerin based cough syrup.
2. Experimental section
Samples: Glycerin based Robitussin cough syrup was purchased from the local pharmacy. Active
ingredient of the syrup was dextromethorphan HBr (15 mg/5.0 ml) and the major inactive ingredients
were glycerin and water. The DEG was purchased from Sigma Aldrich and had the stated purity of 99%.
The glycerin of 99% purity was purchased from the Fisher Scientific. The cough syrup was spiked with
known amounts of DEG ranging from 3 to 50% DEG. The resultant samples were vortexed to perform
infrared and near-infrared spectroscopic measurements.
FT-IR spectra: FT-IR spectra were recorded on a Bruker Vector 33 spectrometer equipped with a KBr
beam splitter and deuterated triglycine sulfate (DTGS) room temperature detector. The spectra were
recorded in the attenuated total reflection (ATR) mode with a single bounce MIRacle® ATR accessory
from PIKE technologies, Inc. (Madison, WI, USA). The accessory was equipped with a split pea shaped
ZnSe ATR crystal. To record spectra, about 10 µl of sample solution was poured on the ATR crystal and
the single beam spectrum was recorded and then was rationed against the single beam spectrum of the
bare crystal to obtain the absorbance spectrum. Spectra were recorded with a resolution of 4 cm−1 and
64 scans were averaged for each spectrum. The spectrometer’s optics were sealed from the atmosphere
but its compartment was not purged during measurements.
NIR spectra: The near-infrared spectra were recorded in the absorbance mode using 1 mm rectangular
quartz cell. The instrument (Bruker Vector 33) was equipped with the tungsten source, CaF2 beam splitter
and the germanium detector. The spectra were run at the resolution of 4 cm−1 and for each spectrum
64 scans were averaged.
Data processing: The spectra were run and processed with the Bruker OPUS (Optical User Software)
program. The recorded infrared and near-infrared spectra were subjected to the Partial Least Squares
(PLS) analysis using QUANT program of the Bruker Instruments. The procedure produced the actual
and predicted values of DEG in cough syrup.

FT-IR spectra: The infrared spectra of glycerin based cough syrup and of DEG are shown in Fig. 1.
The spectrum of cough syrup is dominated by peaks due to water and glycerin. The band at 3300 cm−1
is due to OH stretching modes of water and glycerin whereas the band at 1650 cm−1 is mainly due to OH
bending mode of water. The other peaks of the cough syrup are mainly due to the presence of glycerin.
The peaks around 2900 cm−1 are due to the CH stretching modes of glycerin whereas the peaks between
M. Khalique Ahmed et al. / Toxic DEG contaminant in glycerin based cough syrup 603
Fig. 1. Infrared spectra of glycerin based cough syrup and diethylene glycol. (The colors are visible in the online version of the
article; http://dx.doi.org/10.3233/SPE-2010-0482.)
Fig. 2. Infrared spectrum of glycerin.
1500–700 cm−1 are due to CH bending (1500–1200 cm−1 ) and C–O stretching (1200–900 cm−1 ) modes
of glycerin. The glycerin peaks in cough syrup were confirmed by running the spectrum of pure glycerin
that is shown in Fig. 2. The assignments of the FT-IR spectra of the cough syrup, glycerin and DEG
are given in Table 1. The closer inspection of the spectra of Fig. 1 revealed absorptions in the region
of 930–870 cm−1 in the DEG spectrum that are absent in the cough syrup spectrum. These absorptions
are possibly from the C–O (920 cm−1 ) and OC2 H4 (892 cm−1 ) moieties of the backbone of DEG. This
assignment is reasonable as glycerin does not have OC2 H4 moiety and that is why 892 cm−1 absorption
is absent in the cough syrup. The 930–870 cm−1 region peaks of DEG were used for the quantization
of DEG in cough syrup through Partial Least Squares (PLS) procedure. Figure 3 shows the sensitivity
Table 1
Assignments of the infrared spectra of the cough syrup, Diethylene Glycol
(DEG) and glycerin. Peak positions are in wavenumbers (cm−1 )
Cough syrup DEG Glycerin Assignment
3306 3340 3295 OH stretch
2938 2924 2932 CH2 antisym. stretch
2890 2870 2879 CH2 sym. stretch
1639 ··· ··· OH bending
1420 1455, 1419 1412 CH2 bending
1356 1362, 1324 1326 CH2 wagging
1150 1234 1209 CCC, COC stretch
1104 1125 1108 C–O stretch
1077 1062 1030 C–O stretch
1028 920 993, 922 COH stretch
··· 892 ··· –OC2 H4
Fig. 3. The infrared spectra showing the sensitivity of the contaminated cough syrup to the presence of varying amounts (0–50%)
of diethylene glycol. (The colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0482.)
of the FT-IR spectra of the contaminated cough syrups to the presence of varying amounts of DEG. As
can be seen the 930–870 cm−1 peaks grow linearly with the concentration of DEG in cough syrup. The
plot of PLS predicted and actual values of DEG from FT-IR spectra is shown in Fig. 4. The agreement
between predicted and actual values is excellent.
NIR spectra: The near-infrared spectra of DEG and the cough syrup are shown in Fig. 5. In both
spectra the bands between 5500–6000 and 6500–7000 cm−1 regions are due to first overtone of CH
Fig. 4. Plot of the actual and predicted values of diethylene glycol in cough syrup from infrared spectroscopy.
Fig. 5. Near-infrared spectra of a typical glycerin based cough syrup and diethylene glycol. (The colors are visible in the online
version of the article; http://dx.doi.org/10.3233/SPE-2010-0482.)
and OH stretching vibrations, respectively. As can be seen in Fig. 5 DEG has distinct absorptions in the
regions of 6000–5500 cm−1 whereas cough syrup absorbs very weakly in this region. The assignments
of the NIR spectra of the cough syrup and DEG are given in Table 2. Figure 6 displays the sensitivity of
the near-infrared spectra of the cough syrup to the presence of DEG. Because of the mutually exclusive
Table 2
Assignments of the near-infrared spectra of the cough syrup and diethylene glycol
Frequency (cm−1 ) Assignment
Cough syrup
5630 First CH (glycerin) overtone
∼6350 CH stretch + OH stretch
6855 First OH overtone
Diethylene glycol
∼5625 First CH overtone
5777 First CH overtone
6347 CH stretch + OH stretch
6772 First OH overtone
8244 Second CH overtone
Fig. 6. Near-infrared spectra showing the sensitivity of the contaminated cough syrup to the presence of varying amounts
(5–50%) of diethylene glycol. (Colors are visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0482.)
Fig. 7. Plot of the actual and predicted values of diethylene glycol in cough syrup from near-infrared spectroscopy. (Colors are
visible in the online version of the article; http://dx.doi.org/10.3233/SPE-2010-0482.)
nature of the 6000–5500 cm−1 spectral range, this region was used in the determination of the DEG in
the cough syrup.
The predicted and actual values of DEG from the near-infrared spectra are shown in Fig. 7 and display
reasonable agreement.
4. Conclusion
The presence of toxic diethylene glycol (DEG) can be successfully detected and quantified in the
glycerin based cough syrup by using FT-IR and NIR spectroscopic techniques. It should be noted that
although FTIR and NIR techniques are not as sensitive as other analytical techniques such as mass
spectrometry and high pressure liquid chromatography (15) but they have the advantage of being rapid
and cost effective. Both FT-IR and NIR are well suited for the initial screening of the cough syrups
before using the most sensitive analytical techniques. The FT-IR and NIR methods are environmentally
friendly as they do not utilize any chemicals. The developed methods are especially suited for the third
world where advanced lab facilities are uncommon and are the regions where most cases of the cough
syrup contaminations have been reported.
References
[1] S. Alfred, P. Coleman, D. Harris, T. Wigmore, E. Stachowski and A. Graudins, Delayed neurologic sequelae resulting
from epidemic diethylene glycol poisoning, Clinical Toxicology 43(3) (2005), 155–159.
[2] B.D. Barr, J.R. Barr, G. Weerasekera, J. Wamsley, S.R. Kalb, A. Sjodin et al., Identification and quantification of diethylene
glycol in pharmaceuticals implicated in poisoning epidemics: an historical laboratory perspective, Journal of Analytical
Toxicology 31 (2007), 295–303.
[3] L.A. Ferrari and L. Giannuzzi, Clinical parameters, postmortem analysis and estimation of lethal dose in victims of a
massive intoxication with diethylene glycol, Forensic Science International 153 (2005), 45–51.
[4] M.T. Gomes, M.I. Veríssimo and J.A. Oliveira, A new method for monitoring the contamination of glycerol with di-
ethylene glycol, Journal of Pharmacy and Pharmacology 51 (1999), 233–236.
[5] P. Hari, Y. Jain and S.K. Kabra, Fatal encephalopathy and renal failure caused by diethylene glycol poisoning, Journal of
Tropical Pediatrics 52(6) (2006), 442–444.
[6] M.J. Hasbani, L.H. Sansing, J. Perrone, A.K. Asbury and S.J. Bird, Encephalopathy and peripheral neuropathy following
diethylene glycol ingestion, Neurology 64 (1992), 1273–1275.
[7] F. Hernandez, M. Ibanez and J.V. Sancho, Fast determination of toxic diethylene glycol in toothpaste by ultra-performance
liquid chromatography – time of flight mass spectrometry, Analytical and Bioanalytical Chemistry 391 (2008), 1021–
1027.
[8] J. Marraffa, M. Holland, C. Stork, C. Hoy and M. Hodgman, Diethylene glycol: widely used solvent presents serious
poisoning potential, Journal of Emergency Medicine 35 (2007), 401–406.
[9] H. Maurer and C. Kessler, Identification and quantification of ethylene glycol and diethylene glycol in plasma using gas
chromatography-mass spectrometry, Archives of Toxicology 62 (1988), 66–69.
[10] H. Maurer, F.T. Peters, L.D. Paul and T. Kraemer, Validated gas chromatographic – mass spectrometric assay for deter-
mination of the antifreezes ethylene glycol and diethylene glycol in human plasma after microwave-assisted pivalylation,
Journal of Chromatography B – Biomedical Sciences and Applications 754 (2001), 401–409.
[11] K.L. O’Brien, J.D. Selanikio, C. Hecdivert, M.F. Placide, M. Louis, D.B. Barr et al., Epidemic of pediatric deaths from
acute renal failure caused by diethylene glycol poisoning, JAMA 279 (1988), 1175–1180.
[12] H.O. Okuonghae, I.S. Ighogboja, J.O. Lawson and E.J. Nwana, Diethylene glycol poisoning in Nigerian children, Annals
of Tropical Paediatrics 12 (1992), 235–238.
[13] L. Schep, R. Slaughter, W. Temple and D. Beasley, Diethylene glycol poisoning, Clinical Toxicology 47(6) (2009), 525–
535.
[14] J.G. Schier, C.S. Rubin, D.M. Miller, D. Barr and M.A. McGeehin, Medication-associated diethylene glycol mass poison-
ing: a review and discussion on the origin of contamination, Journal of Public Health 30 (2009), 127–143.
[15] J. Shin, G. Sachs and J. Kraut, Simple diagnostic tests to detect toxic alcohol intoxications, Translational Research 152
(2008), 194–201.

Dennis Miedema What Attracts Women Ebook

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Dennis Miedema What Attracts Women Ebook

Diunggah oleh

Hak Cipta:

Format Tersedia

Spectroscopy 24 (2010) 585–592 585

Dynamic properties of a reconstituted myelin

2. Materials and methods

2.1. Protein purification

2.2. Sample preparation and characterization

1,2-dimyristoyl-sn-glycero-3-phosphate (DMPA), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC)

2.3. Neutron scattering experiment

Incoherent elastic neutron scattering measurements as a function of temperature were performed on

3. Results and discussion

Prediction of favourable sites for spin

2. Information from rotamer attachment to a single site

3. Site choice for accessibility measurements

In Eq. (1) xi , yi and zi are

may be a better predictor for structure-related changes in accessibility than σNO .

4. Site pair choice for distance measurements

5. Studying structural changes

7. Labelling statistics for two membrane proteins

Financial support by Swiss National Science Foundation is gratefully acknowledged.

Biomedical application of Mössbauer

3. Results and discussion

A validated direct spectrofluorimetric method

Mirtazapine is a piperazinoazepine-based tetracyclic compound which is used as an antidepressant

Fig. 1. Enantiomers of mirtazapine.

2.3.1. Optimization of instrumental parameters

3. Results and discussion

3.1. Processing method

3.2. Method validation

A modular Raman microspectroscopy system

2. Materials and methods

Fig. 1. The setup of the Raman microspectroscopy system.

3. Results and discussion

Non-invasive blood glucose measurement by

2. Experiment and clinical trial

2.1. Glucose solutions

2.2. Clinical trial

Fig. 1. Block diagram – glucose concentration analysis.

where X1 – Spectra A, e1 – error 1, g1 – corresponding glucose concentrations, M1 – calibration model

where X2 – Spectra B, e2 – error 2, g – predicted glucose concentrations.

4. Results and discussions

4.1. Glucose solutions

Fig. 2. Block diagram – PLS prediction process of glucose solution concentration.

4.2. Glucose solutions measurement on the same day

4.3. Single wavelength absorbance vs. glucose concentration

y = −5.19e−5 x + 0.47 (R = 0.91 and RMSE = 0.0108 AU or 207 mmol/l). (3)

y = −4.84e−5 x + 0.46 (R = 0.87 and RMSE = 0.0098 AU or 204 mmol/l), (4)

where y – absorbance units, x – glucose concentrations, R – R correlation coefficient, RMSE – root

4.4. Glucose concentration predictions by PLS

4.5. Glucose solution measurements on different days

4.6. Single wavelength absorbance versus glucose concentration on separate days

y = −3.32e−5 x + 0.47 (R = 0.99 and RMSE = 0.0020 AU or 60 mmol/l). (4-5)

y = −3.23e−5 x + 0.41 (R = 0.91 and RMSE = 0.0052 AU or 163 mmol/l), (4-6)

where y – absorbance units, x – glucose concentrations, R – R correlation coefficient, RMSE – root

4.7. Spectra prediction versus glucose concentration by PLS on separate days

5. Error and noise

Under same day Under different days

5.1. Clinical trial

Azure dyes as new photosensitizer

Fig. 1. Structures of Azures A and B.

2. Materials and methods

3. Results and discussion

Direct identification of potato’s fungal

2. Materials and methods

2.2. Sample preparation