Letter
1038/nature23017
CRISPRCas encoding of a digital movie into the
genomes of a population of living bacteria
SethL.Shipman1,2,3, JeffNivala1,3, JeffreyD.Macklis2 & GeorgeM.Church1,3
DNA is an excellent medium for archiving data. Recent efforts
have illustrated the potential for information storage in DNA
using synthesized oligonucleotides assembled in vitro16. A
relatively unexplored avenue of information storage in DNA is
the ability to write information into the genome of a living cell
by the addition of nucleotides over time. Using the Cas1Cas2
integrase, the CRISPRCas microbial immune system stores
the nucleotide content of invading viruses to confer adaptive
immunity7. When harnessed, this system has the potential to
write arbitrary information into the genome8. Here we use the
CRISPRCas system to encode the pixel values of black and white
images and a short movie into the genomes of a population of
living bacteria. In doing so, we push the technical limits of this
information storage system and optimize strategies to minimize
those limitations. We also uncover underlying principles of the
CRISPRCas adaptation system, including sequence determinants
of spacer acquisition that are relevant for understanding both
the basic biology of bacterial adaptation and its technological
applications. This work demonstrates that this system can capture
and stably store practical amounts of real data within the genomes
of populations of living cells.
By combining the principles of information storage in DNA with
DNA-capture systems capable of functioning in living cells, we can
create living organisms that capture, store, and propagate information
over time. In prokaryotic viral defence, the CRISPR-associated (Cas)
proteins, Cas1 and Cas2, function as an integrase complex to acquire
nucleotides from invading viruses and store them in the CRISPR
array7,9,10. In previous work, we found that we could direct the system
to acquire synthetic sequences into the CRISPR array if those sequences
are supplied as oligonucleotides8. Using this approach, we showed
simple molecular recordings by supplying different oligonucleotide
sequences over time.
Here we markedly scale up this approach to define the information
capacity that the system can record, with an eye towards future bio-
logical recordings. Rather than arbitrary sequences, we encode real
information (images) and optimize the method of delivery, nucleotide
content of the sequences, and reconstruction method (for which we
use a population of bacteria). In the Escherichia coli type I-E CRISPR
Cas system, DNA from invading viruses is inserted into a genomic
CRISPR array in 33-base units termed spacers11. The sequences from
which spacers are derived are termed protospacers12. We began with an
image (Extended Data Fig. 1a) and stored pixel values in a nucleotide
code, distributed over many individual synthetic protospacer oligo
nucleotides. We electroporated these oligonucleotides into a population
of bacteria, each harbouring a functional CRISPR array and over
expressing the Cas1Cas2 integrase complex, allowing cells to acquire
the oligonucleotides into their genome. We recover the information by
high-throughput sequencing: newly acquired spacers are decoded to
reconstruct the original image.
1
Department of Genetics, Harvard Medical School, 77 Avenue Louis Pasteur, Boston, Massachusetts 02115, USA. 2Department of Stem Cell and Regenerative Biology, Center for Brain Science,
and Harvard Stem Cell Institute, Harvard University, Bauer Laboratory 103, Cambridge, Massachusetts 02138, USA. 3Wyss Institute for Biologically Inspired Engineering, Harvard University,
Cambridge, Massachusetts 02138, USA.
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
doi:10.
We first encoded images of a human hand using two different
pixel-value-encoding strategies: a rigid strategy (handR), in which
4 pixel colours were each specified by a different base (Extended
Data Fig. 1b, c); and a flexible strategy (handF), in which 21 possible
pixel colours were specified by a degenerate nucleotide triplet table
(Fig. 1a, b). To distribute the information across multiple proto
spacers, we gave each protospacer a barcode that defined which pixel
set (denoted as pixet) was encoded by the nucleotides in that spacer.
Four nucleotides define each pixet, and the pixels of a given pixet are
distributed across the image (Fig. 1c, Extended Data Fig. 1d). We
included a protospacer adjacent motif (PAM) on each protospacer,
which increases the efficiency of acquisition and determines orientation
of spacer insertion8,1315. After adding the PAM and pixet, we were left
with 28 bases per protospacer to encode pixel values.
For handR, each of the 28bases encoded a pixel value, thereby distrib-
uting a 4-colour, 5656 pixel image across 112 oligonucleotide proto-
spacers (total information content of 784bytes). For handF, the 28 bases
encoded 9 pixels, each specified by a nucleotide triplet. Specific triplet
combinations were chosen to build sequences that we hypothesized
might increase acquisition efficiencyGCcontent around 50%, no
mononucleotide repeats >3bp, and no internal PAMs. For the handF,
we distributed a 21-colour, 3030 pixel image across 100 protospacers
(total information content of around 494 bytes). Oligonucleotide pro-
tospacers were supplied in a minimal hairpin format (design based on
insights from the crystal structure16,17) to prevent segregation of the two
strands into different cells during electroporation (see Supplementary
Information, Extended Data Fig. 2).
For each image, we electroporated the pooled oligonucleotides
into a population of E. coli containing a genomic CRISPR array and
expressing the Cas1Cas2 integrase18. Cells were then recovered, pas-
saged overnight, and the next day a sample of the genomic CRISPR
arrays were sequenced. Newly acquired spacers were bioinformati-
cally extracted from the arrays, and those that were not derived from
the plasmid or genome were analysed. Pixel values were assigned on
the basis of the most numerous new spacer with a given pixet. Images
reconstructed from the handR and handF images are shown in Extended
Data Fig. 1e and Fig. 1d, respectively.
Using 655,360 reads, around 88% and around 96% of pixet sequences
were accurately recalled from the handR and handF images, respectively.
We found that handF was more resistant to errors by under-sampling
(Fig. 1e, f, Extended Data Fig. 1fg). By electroporating subsets of the
oligonucleotides, we found that the number of reads required to achieve
similar levels of accuracy in recall is linearly related to the number of
oligonucleotides electroporated (Fig. 1g, Extended Data Fig. 1h, i), and
that it took substantially more reads per oligonucleotide protospacer to
reach 80% accuracy from handR (around 1,580 reads per protospacer)
versus handF (around 150 reads per protospacer).
We also sampled time points of the bacterial culture following the
electroporation of handF. Oligonucleotide-derived spacer acquisitions
0 0 M o n t h 2 0 1 7 | VO L 0 0 0 | NAT U R E | 1
 RESEARCH Letter

a b c GC ~50%, no mononucleotide
repeats >3 bp,
Pixet no internal PAMs

se
Encoded image 2nd base

ba
T C A G

d
1st base 1 1 2 3 4 5 6 7 8 9

3r
1 2 3 4 T
5 6 7 8 C 00000001
T 9 10 11 12 A 00 = C 1 4 7
13 14 15 16 G
17 18 19 20 T 01 = T
21 1 2 3 C 10 = A
C 4 5 6 7 A
8 9 10 11 G 11 = G 8 2 5
12 13 14 15 T
16 17 18 19 C
A 20 21 1 2 A
3 4 5 G
6 7 8 9 T 6 9 3
10 11 12 13 C
G 14 15 16 17 A
18 19 20 21 G

d Recalled image e f g

Accurately recalled pixets (%)

40,960 20,480 10,240

100 20
Percentage of supplied

Number of reads (103)

oligos accurately

Reads sampled
80
15 recalled 80%
60
10
40
70%
5
20 60%
50%
0 0
10 102 103 104 105 106 0 50 100
655,360 reads
Reads sampled Oligos supplied
h
Days : hours : minutes (post-electroporation)
0:00:05 0:00:10 0:00:20 0:00:40 0:01:20 0:02:40 0:05:20 1:00:00 3:00:00 5:00:00 7:00:00

Arrays expanded with

recalled pixets (%)

100

image oligos (%)

30
100 30 80
Accurately

80 60 20
60 20
40
40 10
10 20
20
0 0 0 0
0 1 2 3 4 5 0 1 3 5 7
Hours (12) (24) (36) (48)
post-electroporation Days post-electroporation (bacterial generations)
F
Figure 1 | An image into the genome. a, Hand image. b, Encoding sampling the sequencing reads. g, Reads required to reach 50%, 60%,
for 21 colours. c, Sequence at top shows the linear protospacer with 70%, and 80% accuracy on a given oligonucleotide set as a function of
pixet code followed by pixel values (distributed across image). Pixet number of oligonucleotides supplied (n=3; linear regression of the 80%
shown under nucleotides, with binary-to-nucleotide conversion. Small curve, R2=0.9975; runs test of the 80% curve, P>0.99). h, Image recall
colourful numbers below protospacer indicate individual pixels boxed at time points after electroporation. i, Quantification of the percentage
on the image. Minimal hairpin protospacer shown on the right. d, One of accurately recalled pixets (in black) and percentage of arrays with
replicate at 655,360 reads. Black shown if no pixel information recovered. oligonucleotide-derived spacers (in red) by time point. Unfilled circles
e, Accurately recalled pixets by read depth. Unfilled circles indicate points represent 3 biological replicates, lines show the mean. Inset graph (left)
from 3 biological replicates, black line shows the mean. f, Result of down- expands first six hours. Statistical details in Supplementary Table 1.

were detectable ten minutes after the electroporation, and peaked at 2h the handR image had, by chance, an overall lower GC% than those
40min, at which point we could first accurately recall the entire image encoding the handF image (41.80.6% versus 50.60.6%), which
(Fig. 1h, i). From this peak to 24h post-electroporation, the percent- may account for the difference in acquisition frequency. For mono-
age of oligonucleotide-expanded arrays declined slightly, then stabi- nucleotide repeats and internal PAMs, we found that pools of proto
lized over the next six days (~48 bacterial generations). Presumably, spacers with substantial numbers of either displayed reductions in total
some cells lose viability following the electroporation and do not acquisitions.
contribute to the population after outgrowth (Extended Data Fig. 3, Despite differences in acquisition frequency between handR and
Supplementary Information includes information about the internal handF, a similar range of acquisition frequencies is apparent among
integrity of the arrays over time1921). individual protospacer sequences from each (Fig. 2a). We compared
The total acquisition frequency was higher for handF than handR, over-represented protospacers with all protospacers and found
explaining the improvement in recall (Extended Data Fig. 4a). To test a signif icant motif present in the final two nucleotides in over-
which of the parameterspercentage of GC content (GC%), absence of represented sequences (Fig. 2b). A similar motif has been previ-
mononucleotide repeats, or lack of internal PAMsaccounted for this ously reported and termed the acquisition affecting motif (AAM)22;
greater acquisition frequency, we designed new sets of oligonucleotide however, the reported sequence of this motif differs from what we
protospacers, systematically testing each parameter (Supplementary find here. Although we found the motif to be composed of slightly
Information, Extended Data Fig. 4bf). GC% had a clear effect on different bases, we believe that these differences probably to arise
acquisition frequency, with reduced acquisition frequency at low GC%. from the fact that we were able to synthetically control for the pres-
This effect was most extreme when pools contained a wide range of ence of the more dominant PAM motif in our sequences, and thus we
GC%. In pools with homogeneous percentage, those over 50% were adopt the previous term AAM. The PAM sequence has been shown
equally effective. Therefore, it is beneficial to limit the range of GC% to function not only in adaptation but also interference23 and, given
and keep the percentage at 50% or higher. The protospacers encoding that it lies outside of the acquired spacer, serves as a mechanism of

2 | NAT U R E | VO L 0 0 0 | 0 0 m o n t h 2 0 1 7
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Letter RESEARCH

a b c d
5
6 6

log-odds of the binomial probability

9.12 Top 10% of acquired spacers
Oligo-derived acquisitions (%)

Seqover
5

with oligo protospacer (%)

4
4 3.45 Seqover-CCT

Arrays expanded
3
3 0
0 100

2 Sequnder 2
handR
handF 3.45
1
Sequnder-TGA 1
0
0 20 40 60 80 100 9.12
Spacer number 0

er

r
de
T

A
(ranked by acquision frequency)

q ov

G
q un
-C

-T
Se

r
er

de
Se
q ov

q un
Se

Se
e f
g Position
7 31 32 33
7 TGA 6 45
TAA 5
6 AGA 40
4
3 35
Acquired spacers (%)

Acquired spacers (%)

5
2 30
4 1
0 25
0 20 40 60
3 20
CCT CCG 15
2 ACG
GTT 10
1
5
0 0
0 10 20 30 40 50 60 G A T C G A T C G A T C
NNN Cartesian product (ranked by acquisition)

Figure 2 | Sequence determinants of acquisition. a, Acquisition comparison (corrected), seqover versus seqover-CCT: P=0.0023, sequnder
frequency for individual protospacers (of oligonucleotide-derived versus sequnder-TGA: P=0.0294). e, NNN-containing oligonucleotide.
acquisitions) for both images, ranked by frequency. Main plot circles f,Acquisition frequency of protospacers containing each NNN Cartesian
represent means.e.m. Smaller inset shows each replicate (n=3). product (of oligonucleotide-derived acquisitions), ranked by frequency
b,pLogo30 of the top 10% of protospacers (all protospacers as background). (n=3). Plots as in a. g, Representation of nucleotides at positions 3133 in
Red line indicates P<0.05. Over-representation is positive, under- acquired spacers from the NNN-containing oligonucleotide (n=3; one-
representation is negative (n=3). c, Sequences designed to test the way ANOVA on effect of nucleotide, position 31: P=0.0006, position 32:
motif. d, Arrays expanded with the sequences indicated in c. Unfilled P=0.0002, position 33: P<0.0001; follow-up Tukeys multiple comparison
circles represent individual replicates. Bars show means.e.m. (n=3; (corrected) see Supplementary Table 1). Plot as in d. *P<0.05. Statistical
one-way ANOVA on effect of oligo: P=0.002; follow-up Sidaks multiple details in Supplementary Table 1.

self versus non-self discrimination14. Although the AAM lies within moved the pixet to the final nucleotides of the protospacer, where a
the acquired spacer (and thus could promote self-targeting) and addi- reduced sequence space was employed, limited to the most efficient
tionally lies outside the seed region24, it would still be interesting to eight AAM triplets from Fig. 2f (Fig. 3a). We again used the flexible
directly test whether the presence of an AAM similarly influences 21-colour code from the handF image and chose to encode five frames
interference efficiency. of a galloping mare from Eadweard Muybridges Human and Animal
To test whether the AAM motif that we found is responsible for the Locomotion at 3626 pixels. Frames were each represented by a unique
difference in acquisition efficiency, we tested individual protospacers, oligonucleotide set of 104 protospacers, for an overall information con-
using nucleotides from the over-represented motif, ending in TGA, tent of around 2.6 kilobytes. Pixet codes were reused between frames,
to define one protospacer (seqover), and nucleotides drawn from the and no nucleotides were used to identify frame order. Rather, each
under-represented motif, ending in CCT, to define another (sequnder). frame was electroporated successively over five days into a single
We also swapped the final three nucleotides from these two sequences to population (Fig. 3b). Because new spacers are almost always acquired
create two more protospacers (seqover-CCT and sequnder-TGA) (Fig. 2c). adjacent to the leader sequence in the CRISPR array18, pushing previ-
We found that the final three nucleotides determined acquisition ously acquired spacers away from the leader, the order of frames within
frequency, with TGA yielding high efficiency and CCT yielding the GIF can be reconstructed on the basis of the pairwise order of
low efficiency, regardless of the rest of the sequence content (Fig. 2d). spacers among many individual arrays.
Because these nucleotides are in the loop region of the hairpin Following electroporation, we found that the protospacers were
protospacer, we also tested these sequences as complementary efficiently acquired from each frame, and populated the first three
single-stranded oligonucleotides and found an identical dependence sequenced positions of CRISPR arrays (Fig. 3c). We extracted all new
on the final three nucleotides (Extended Data Fig. 5). spacers from the arrays, then analysed the pixet nucleotides to recover
Because we identified this motif using sequence-constrained pro- the spacers assigned to each unique pixet, but in this case, captured
tospacers, we tested a hairpin protospacer with random nucleotides the five most frequently acquired spacers with each pixetone for
(NNN) in the final three positions (Fig. 2e). Although we observed each frame.
acquisition events with every possible NNN Cartesian product in the To order the frames over time, we used positioning within individ-
three variable nucleotides, their efficiencies varied and allowed us to ual arrays to reconstruct the electroporation order of the protospacers.
define the ideal AAM (Fig. 2f, g). Ordering information can only be recovered from single cells, in
We next applied our better understanding of protospacer sequence which spacers further from the leader within a single array must have
determinants of acquisition to encode multiple images over time within been acquired earlier than spacers closer to the leader in that same
a single population of bacteria, generating a short movie (GIF). We array. However, the GIF information is widely distributed among a

0 0 M o n t h 2 0 1 7 | VO L 0 0 0 | NAT U R E | 3
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter

a GC ~50%, no mononucleotide b Frame1 Frame2 Frame3 Frame4 Frame5

repeats >3 bp,
no internal PAMs Pixet
16 h 2 h 2 h 16 h 2 h 2 h 16 h 2 h 2 h 16 h 2 h 2 h 16 h 2 h 2 h 16 h

1 2 3 4 5 6 7 8 9 1 1 2 3 4 5

1 7 5 3 9 Binary (reversed) c Frame5 d 100 Frame

10,000,000 60 Frame4
Frame3 5

Accurately recalled pixets (%)

Bases Bases Frame2 4
50

Arrays expanded (%)

31 + 32 33 35 Frame1 80 3
00 = C 0000 = TGA Plasmid 2
01 = T 1000 = AGA 40 Genome 1
4 2 8 6 10 = A 0100 = TAA Sample point 60
11 = G 1100 = CAC
30 1 5
0010 = GAC
20 40
1010 = AGC
0110 = GGA 10
1110 = TGG
20
0
L spacer spacer spacer
0
Position 1 Position 2 Position 3 10 102 103 104 105 106 107
Encoded GIF
Reads sampled
e Reads sampled Recalled GIF
f 12 g

log-odds of the binomial probability

Frame 6.83
Top 10%
10 5
5,120

Acquired spacers (%)

4
3 3.45
8 2
1
40,960

4
4,372,489 1,310,720 163,840

2 3.45

0
0 20 40 60 80 100 6.83
Spacer number
(ranked by acquision frequency)
h Frame5
Recalled GIF

1,310,720
Frame4
Frame3
Frame2
Frame1

Figure 3 | Encoding a GIF in bacteria. a, GIF to be encoded (adapted replicates. e, Examples of the result at different sequence depths (see
from Eadweard Muybridge, Human and Animal Locomotion, plate 626, dotted grey lines in d). f, Protospacer acquisition frequency for individual
thoroughbred bay mare Annie G. galloping, Wikimedia Commons), protospacers (of oligonucleotide-derived acquisitions) by frame, ranked
along with an example of one pixet protospacer. b, Schematic of recording by acquisition frequency. Points show 3 biological replicates. g, pLogo30
process. c, Percentage of arrays with expansions in the first three positions, of the top 10% of protospacers (all protospacers as background). Red line
by protospacer origin, at each sample point. Bars show means.e.m. indicates P<0.05. Over-representation is positive, under-representation
(n=3). d, Accurately recalled pixets as a function of reads (on the x axis) is negative. h, Result of electroporating the same oligonucleotides in the
and frame (denoted by colour). Points show individual biological reverse order. Statistical details in Supplementary Table 1.

population of bacteriano individual cell can be used to recon- single-cell storage, and a comparison of information storage in DNA
struct the entire image series. Therefore, we leveraged many single-cell versus silicon2529).
ordering comparisons among the population of bacteria to reconstruct
the entire GIF (see Supplementary Information, Extended Data Fig. 6 Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
for detail). these sections appear only in the online paper.
We found that we could reconstruct each frame and the order of
frames (Fig. 3d), and that increasing read depth aided the accuracy received 22 August 2016; accepted 2 June 2017.
of the reconstruction (to >90% overall accuracy) (Fig. 3e). Despite Published online 12 July 2017.
optimization of the protospacer sequence, we still found a range of
efficiencies between the protospacers of any given frame (Fig. 3f). We 1. Church, G. M., Gao, Y. & Kosuri, S. Next-generation digital information storage
again found a sequence motif at the AAM location, suggesting that we in DNA. Science 337, 1628 (2012).
2. Goldman, N. et al. Towards practical, high-capacity, low-maintenance
allowed for too large a range of nucleotide triplets in the final posi- information storage in synthesized DNA. Nature 494, 7780 (2013).
tion (Fig. 3g) or this range may reflect an inherent competition among 3. Gibson, D. G. et al. Creation of a bacterial cell controlled by a chemically
protospacers, either for Cas1Cas2 or the genomic array. As the pro- synthesized genome. Science 329, 5256 (2010).
4. Clelland, C. T., Risca, V. & Bancroft, C. Hiding messages in DNA microdots.
tospacers themselves contain no code to specify frame position, we Nature 399, 533534 (1999).
tested the robustness of our reconstruction strategy by delivering the 5. Adleman, L. M. Molecular computation of solutions to combinatorial problems.
oligonucleotide frame sets in reverse order. We were able to accurately Science 266, 10211024 (1994).
6. Davis, J. Microvenus. Art J. 55, 7074 (1996).
reconstruct the reversed GIF, demonstrating reconstruction of an oth- 7. Barrangou, R. et al. CRISPR provides acquired resistance against viruses in
erwise ambiguous signal based on time (Fig. 3h). prokaryotes. Science 315, 17091712 (2007).
In summary, we found that not all protospacer sequences are equally 8. Shipman, S. L., Nivala, J., Macklis, J. D. & Church, G. M. Molecular
recordings by directed CRISPR spacer acquisition. Science 353, aaf1175
effective at transferring data into the genome, and for this reason advo- (2016).
cate for the use of a flexible encoding scheme to allow optimization of 9. Amitai, G. & Sorek, R. CRISPRCas adaptation: insights into the mechanism of
sequence content. We found that sequences with controlled GC content, action. Nat. Rev. Microbiol. 14, 6776 (2016).
10. Sternberg, S. H., Richter, H., Charpentier, E. & Qimron, U. Adaptation in
a lack of mononucleotide repeats, and no internal PAMs outperformed CRISPRCas Systems. Mol. Cell 61, 797808 (2016).
those that lacked such optimization. Further, the inclusion of invariant 11. van der Oost, J., Jore, M. M., Westra, E. R., Lundgren, M. & Brouns, S. J.
nucleotides at both the leading (AAG) and trailing (GA) end of the CRISPR-based adaptive and heritable immunity in prokaryotes. Trends
protospacer has large effects on the frequency of acquisition. We were Biochem. Sci. 34, 401407 (2009).
12. Deveau, H. et al. Phage response to CRISPR-encoded resistance in
able to track the presence of 104 separately barcoded sequence elements Streptococcus thermophilus. J. Bacteriol. 190, 13901400 (2008).
over five time points (520 unique sequence elements), yielding confi- 13. Paez-Espino, D. et al. Strong bias in the bacterial CRISPR elements that confer
dence that this system will be capable of recording multidimensional immunity to phage. Nat. Commun. 4, 1430 (2013).
14. Westra, E. R. et al. Type I-E CRISPRCas systems discriminate target from
biological information (see Supplementary Information, Extended Data non-target DNA through base pairing-independent PAM recognition. PLoS
Figs 7, 8 for discussion into error-correction/compression, obstacles to Genet. 9, e1003742 (2013).

4 | NAT U R E | VO L 0 0 0 | 0 0 m o n t h 2 0 1 7
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

15. Shmakov, S. et al. Pervasive generation of oppositely oriented spacers during
CRISPR adaptation. Nucleic Acids Res. 42, 59075916 (2014).
16. Nuez, J. K., Harrington, L. B., Kranzusch, P. J., Engelman, A. N. & Doudna, J. A.
Foreign DNA capture during CRISPRCas adaptive immunity. Nature 527,
535538 (2015).
17. Wang, J. et al. Structural and mechanistic basis of PAM-dependent spacer
acquisition in CRISPRCas systems. Cell 163, 840853 (2015).
18. Yosef, I., Goren, M. G. & Qimron, U. Proteins and DNA elements essential for the
CRISPR adaptation process in Escherichia coli. Nucleic Acids Res. 40,
55695576 (2012).
19. Diez-Villasenor, C., Almendros, C., Garcia-Martinez, J. & Mojica, F. J. Diversity of
CRISPR loci in Escherichia coli. Microbiology 156, 13511361 (2010).
20. Weinberger, A. D. et al. Persisting viral sequences shape microbial CRISPR-
based immunity. PLOS Comput. Biol. 8, e1002475 (2012).
21. Held, N. L., Herrera, A., Cadillo-Quiroz, H. & Whitaker, R. J. CRISPR associated
diversity within a population of Sulfolobus islandicus. PLoS One 5, e12988
(2010).
22. Yosef, I. et al. DNA motifs determining the efficiency of adaptation into the
Escherichia coli CRISPR array. Proc. Natl Acad. Sci. USA 110, 1439614401
(2013).
23. Westra, E. R. et al. CRISPR immunity relies on the consecutive binding and
degradation of negatively supercoiled invader DNA by cascade and Cas3.
Mol. Cell 46, 595605 (2012).
24. Semenova, E. et al. Interference by clustered regularly interspaced short
palindromic repeat (CRISPR) RNA is governed by a seed sequence. Proc. Natl
Acad. Sci. USA 108, 1009810103 (2011).
25. Zhirnov, V., Zadegan, R. M., Sandhu, G. S., Church, G. M. & Hughes, W. L.
Nucleic acid memory. Nat. Mater. 15, 366370 (2016).
26. Hsiao, V., Hori, Y., Rothemund, P. W. & Murray, R. M. A population-based
temporal logic gate for timing and recording chemical events. Mol. Syst. Biol.
12, 869 (2016).
27. McKenna, A. et al. Whole-organism lineage tracing by combinatorial and
cumulative genome editing. Science 353, aaf7907 (2016).
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter RESEARCH
28. Frieda, K. L. et al. Synthetic recording and in situ readout of lineage information
in single cells. Nature 541, 107111 (2017).
29. Erlich, Y. & Zielinski, D. DNA Fountain enables a robust and efficient storage
architecture. Science 355, 950954 (2017).
30. OShea, J. P. et al. pLogo: a probabilistic approach to visualizing sequence
motifs. Nat. Methods 10, 12111212 (2013).
Supplementary Information is available in the online version of the paper.
Acknowledgements S.L.S. is a Shurl and Kay Curci Foundation Fellow of the
Life Sciences Research Foundation. The project was supported by grants from
the National Institute of Mental Health (5R01MH103910), National Human
Genome Research Institute (5RM1HG008525), and Simons Foundation Autism
Research Initiative (368485) to G.M.C., the National Institute of Neurological
Disorders and Stroke (5R01NS045523) to J.D.M and an Allen Distinguished
Investigator Award from the Paul G. Allen Frontiers Group to J.D.M. We thank
G. Kuznetsov for comments on the manuscript.
Author Contributions S.L.S. and J.N. conceived the study. S.L.S. designed the
work, performed experiments, analysed data, wrote custom Python analysis
software, and wrote the manuscript with input from J.N., J.D.M. and G.M.C. S.L.S.
J.N., J.D.M. and G.M.C. discussed results and commented on the manuscript.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare competing financial interests:
details are available in the online version of the paper. Readers are welcome to
comment on the online version of the paper. Publishers note: Springer Nature
remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations. Correspondence and requests for materials should be
addressed to G.M.C. (gchurch@genetics.med.harvard.edu).
Reviewer Information Nature thanks R. Barrangou and the other anonymous
reviewer(s) for their contribution to the peer review of this work.
0 0 M o n t h 2 0 1 7 | VO L 0 0 0 | NAT U R E | 5
 RESEARCH Letter
Methods
Data reporting. No statistical methods were used to predetermine sample size.
The experiments were not randomized. Colony counts were performed blind to
experimental condition using Amazons Mechanical Turk, the investigators were
otherwise not blinded to allocation during experiments and outcome assessment.
Bacterial strains and culturing conditions. All experiments were carried out in
BL21-AI E. coli (Thermo Fisher), containing an integrated, arabinose-inducible
T7 polymerase, an endogenous CRISPR array, but no endogenous Cas1 and Cas2.
The strain was authenticated based on the known endogenous CRISPR array and
expression of T7 polymerase with arabinose induction. A plasmid encoding induci-
ble (T7/lac) Cas1+2 (K-strain origin, pWUR1+2 a.k.a. pCas1+2) was transformed
into cells before each experiment. Cells containing the plasmid were maintained in
colonies on a plate at 4C for up to three weeks. Bacterial cultures were used under
antibiotic conditions, and mycoplasma testing was not necessary.
Oligonucleotide protospacer electroporation. Protospacer electroporations were
performed as previously described8. Briefly, after overnight outgrowth from a sin-
gle colony, Cas1 and Cas2 were induced in a 3ml dilution of the culture (containing
80ul of the overnight), and grown at 37C for 2h (l-arabinose 0.2% w/w, Sigma-
Aldrich; isopropyl--d-thiogalactopyranoside 1mM, Sigma-Aldrich). For a given
condition, 1ml of the induced culture was spun down and washed with water three
times at 4C, then resuspended in 50l of a 6.25M solution (unless other concen-
tration is noted) of either a single protospacer or set of multiple protospacers and
electroporated in a 1mm gap cuvette using a Bio-Rad gene pulser set to 1.8kV and
25F. Only those conditions with an electroporation time constant >4.0ms were
carried through to analysis. After electroporation, cells were recovered in 3ml LB
at 37C for 23h, then diluted (50l) into a fresh 3ml culture and grown overnight. the final base was assigned to the least numerous base in the rest of the spacer. We
Cells were collected for analysis the following morning (unless otherwise noted).
For checking the maintenance of the 21-colour image over time, cells were passaged
daily (50l into 3ml) after the first 24h and grown at 30C. To estimate the number
of bacterial generations, we calculated the number of doublings required take the
starting dilution (50l into 3ml) to saturation at 1109cells perml (note that the process was repeated for each frame.
first dilution was not from a saturated culture; in this case empirically determined
cell numbers were taken from Extended Data Fig. 3). The oligonucleotide proto-
spacers used can be found in Supplementary Table 3. To estimate the number of
cells surviving electroporation, cells were serially diluted 1:300 (normalized to
1ml of starting culture), then 1:400 before plating on spectinomycin-containing
plates. The resulting colonies were imaged using a commercial document scanner.
To obtain colony counts blinded to experimental condition, partial or complete
plate images were uploaded to Amazons Mechanical Turk workplace where remote
workers were asked to count the number of dots per image. The answers provided
by 10 workers were averaged for each plate image, from which the colony-forming
units per millilitre of starting volume were calculated.
Analysis of spacer acquisition. To analyse spacer acquisition, bacteria were
lysed by heating to 95C for 5min, then subjected to PCR of their genomic arrays
using primers that flank the leader-repeat junction and additionally contain
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Illumina-compatible adapters. Libraries of up to 96 dual-indexed samples were
sequenced on a MiSeq sequencer (Illumina) to read up to three spacer positions in
from the leader on each array. Spacer sequences were extracted bioinformatically on
the basis of the presence of flanking repeat sequences, and compared against pre-
existing spacer sequences to determine the percentage of expanded arrays and the
position and sequence of newly acquired spacers. New spacers were blasted (NCBI)
against the genome and plasmid sequences to determine the origin of the proto-
spacer, with those sequences not derived from the genome or plasmid assumed to
be oligonucleotide-derived. As each cell contains a single array, the read depth is
roughly equivalent to the number of cells analysed (including both expanded and
unexpanded arrays). This and all subsequent image analysis was performed using
custom written scripts in Python.
Image coding and decoding. Image protospacer sets were created using a custom
Python script to first open and read the pixel values of a previously created image.
Each protospacer was given a pixet code by a binary-to-nucleotide conversion,
and populated by nucleotides encoding the pixel values according to the scheme
detailed in the text. For the single images, the pixet code was interleaved between
ascending and descending numbers to introduce more sequence diversity in neigh-
bouring pixet protospacers. In the case of the flexible code used in Figs 2 and 3,
the protospacer was built sequentially. For each new pixel value the three possible
nucleotide sequences were ranked according to which triplet would best push
GC% of the resulting sequence towards 50%, then tested for whether the addition
of the triplet would create either an internal PAM or a mononucleotide repeat
>3. If such a situation was created, the next triplet in the list was tested until an
acceptable triplet was identified (Extended Data Fig. 8ad). For the handF image,
did not attempt to actively exclude sequences that matched the plasmid or genome,
as this would be an exceedingly unlikely event given our library sizes. Finally, the
sequences were re-formatted to match the minimal hairpin structure and written
to a spreadsheet for synthesis by Integrated DNA Technologies. For the GIF, this
To reconstruct the single images, newly acquired oligonucleotide-derived
spacers (plasmid- and genome-derived spacers were set aside before this analysis)
were ranked according to frequency of acquisition, then the most frequent spacer
sequences for each pixet (by the reversed nucleotide to binary conversion) were
assigned to that pixet. Pixel values were extracted from the remaining spacer
sequence according to the schemes outlined in the text, and figures and used to
populate an image. The more complicated reconstruction of the GIF is described
in detail in Supplementary Information as are the calculations of information
content.
Statistics. A list of statistical tests can be found in Supplementary Tables 1 and 2.
Data availability. The data that support the findings of this study are available
from the corresponding author upon reasonable request.
Code availability. Custom code used in this study can be accessed at https://github.
com/churchlab/crispr-images.

Extended Data Figure 1 | Recording images into the genome. a, Pixel
values are encoded across many protospacers, which are electroporated
into a population of bacteria that overexpress Cas1 and Cas2 to store
the image data. These bacteria can be archived, propagated, and
eventually sequenced to recall the image. b, Initial image to be encoded.
c, Nucleotide-to-colour encoding scheme. d, Example of the encoding
scheme. Sequence at top shows the protospacer linear view with pixet
code (specifying a pixel set) followed by pixel values, which are distributed
across the image. Pixet number is shown under the pixet nucleotides, with
the binary-converted pixet and binary-to-nucleotide conversion reference
below that. Small numbers (in colour) below the protospacer indicate
individual pixels, identified by boxes on the image. Protospacer in minimal
hairpin format for electroporation is shown on the right. e, Results of one
replicate at a depth of 655,360 reads. White is shown if no information
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter RESEARCH
was recovered about the pixel value (owing to a pixet protospacer not
being recovered after sequencing). f, Percentage of accurately recalled
pixets as a function of read depth. Unfilled circles indicate points derived
from 3 biological replicates. The black line is the mean of the replicates.
g,Examples of the images that result from down-sampling the sequencing
reads. h, Effect of supplying fewer oligonucleotides on recall accuracy as
a function of reads sampled when smaller pools of oligonucleotides are
supplied and recalled. Individual points show 3 biological replicates, lines
are the means of the replicates. i, Number of reads required to reach 50%,
60%, 70%, and 80% accuracy on a given oligonucleotide set as a function
of oligonucleotides supplied (n=3; linear regression of the 80% curve,
R2=0.9466; runs test of the 80% curve, P>0.99). Additional statistical
details in Supplementary Table 2.
 RESEARCH Letter

a b
16
Arrays expanded, oligo derived (%)

14 1.6
12
1.4
10

Arrays expanded, oligo derived (%)

8
1.2
6
4 1
2
0.8
0 No Full Top Bottom
PAM PAM PAM PAM
No PAM
0.6
Full PAM

0.4
Top PAM

Bottom PAM 0.2

0
c 4
Arrays expanded, oligo derived (%)

0
0.195
0.39
0.781
1.56
3.125
6.25
12.5
25
50

Oligo concentration
Extended Data Figure 2 | Testing a minimal hairpin protospacer.
a,Percentage of arrays expanded with oligonucleotide-supplied spacers
following electroporation of the sequences indicated below, aimed at
testing PAM inclusion on both the top and bottom strands. Unfilled
circles indicate biological replicates, bars are means.e.m (n=3; one-
way ANOVA: P<0.0001; follow-up Dunnetts multiple comparison
(corrected), no PAM versus full PAM: P=0.0001, no PAM versus bottom
PAM: P=0.0002). *P<0.05. Oligonucleotides supplied at 3.125M each.
b, Percentage of arrays expanded with oligonucleotide-supplied spacers
following electroporation of the sequences indicated to the left, right,
and below aimed at finding a minimal functional hairpin protospacer.
Unfilled circles indicate individual biological replicates, bars are
means.e.m (n=4; one-way ANOVA effect of protospacer: P>0.05).
Oligonucleotides supplied at 3.125M. c, Percentage of arrays expanded
following electroporation of different concentrations of the minimal
hairpin oligonucleotide protospacer (n=1). Additional statistical details
in Supplementary Table 2.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Letter RESEARCH

500
CFU per ml of starting culture Electroporated with:
Oligo
400 H2O
Surviving Electroporation:
300 64.45 11.2 (X 106)
(X106)

112.37 23.2 (X 106)

200

100

0
re

ry
n
as

tio
ltu

ve
-w

ra

co
Cu

st

po

re
Po
ng

ro

1h
ti

ct
ar

le
St

-e
st
Po

Extended Data Figure 3 | Cell surviving electroporation. Colony-

forming units per millilitre of starting culture before beginning
electroporation, after pre-electroporation washes, immediately post-
electroporation, and after 1h of recovery. Cells in red were electroporated
with a minimal hairpin oligonucleotide, those in blue were electroporated
in water alone. Unfilled circles represent individual biological replicates
(n=3), filled circles are means.e.m.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter

a b Single Pool Subpooled

15 25
1.0

by Oligo Protospacer Subpool (%)

Arrays expanded per oligo (%)
12 20
0.8

Arrays expanded (%)

Arrays expanded
9 15
0.6

6 0.4 10

3 0.2 5

0 0.0 0
0 25 50 75 100 0 25 50 75 100
ha R
F
nd
nd

GC Percentage
ha

c d Spacer Number (ranked by acquision frequency)

35 0
Individually electroporated oligos 20 40 60 80 100
by Oligo Protospacer (%)

30 -5

of most stable structure

Arrays expanded

25 -10 handR
20 G (Kcal/mol) -15 handF
15 -20

10 -25

5 -30
0 -35
0 25 50 75 100
GC Percentage

e Single Pool Subpooled f Single Pool Subpooled

0.5 30 0.6 30
by Oligo Protospacer Subpool (%)

by Oligo Protospacer Subpool (%)

25 25
Arrays expanded per oligo (%)
Arrays expanded per oligo (%)

0.4
20 0.4 20
Arrays expanded

Arrays expanded

0.3
15 15
0.2
10 0.2 10
0.1
5 5

0.0 0 0.0 0
3 4 6 8 3 4 6 8 0 1 3 5 0 1 3 5
Mononucleotide Repeat Number Internal PAMs
Extended Data Figure 4 | Optimization of protospacer sequence Tukeys multiple comparison (corrected), see Supplementary Table2).
parameters. a, Comparison of the percentage of arrays that were d,Gibbs free energy of minimal hairpin protospacers structures for
expanded after encoding handR and handF images (n=3). b, Percentage each of the images, with protospacers ranked by overall acquisition
of arrays expanded per oligonucleotide (single pool) or per subpool frequency (n=3; linear regression, handR: P=0.0089, handF: P=0.0004).
(subpooled) across a range of GC percentages. Unfilled black circles to e,Percentage of arrays expanded per oligonucleotide (single pool) or per
the left represent individual oligonucleotide protospacer sequences (three subpool (subpooled) with different numbers of mononucleotide repeats
biological replicates each), while black line shows means.e.m. Unfilled (n=3; one-way ANOVA on effect of mononucleotide repeats, single pool:
red circles to the right represent individual biological replicates. Bars P=0.3843, subpooled: P=0.0015; follow-up testing with Tukeys multiple
are means.e.m (n=3; one-way ANOVA on effect of GC percentage, comparison (corrected), see Supplementary Table 2). Panel attributes as
single pool: P<0.0001, subpooled P=0.0011; follow-up testing with in b. f, Percentage of arrays expanded per oligonucleotide (single pool) or
Tukeys multiple comparison (corrected), see Supplementary Table 2). per subpool (subpooled) with different numbers of internal PAMs (n=3;
c, Percentage of arrays expanded per oligonucleotide electroporated one-way ANOVA on effect of internal PAMs, single pool: P=0.0565,
individually across a range of GC percentages. Unfilled red circles are subpooled: P=0.0052; follow-up testing with Tukeys multiple comparison
individual biological replicates. Bars show means.e.m (n=3; one-way (corrected), see Supplementary Table 2). Panel attributes as in b. *P<0.05.
ANOVA on effect of GC percentage: P=0.0001; follow-up testing with Additional statistical details in Supplementary Table 2.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Letter RESEARCH

14

12
seqover

by Oligo Protospacer (%)

10
seqover (CCT):

Arrays expanded
8

6
sequnder
Complementary
4 oligos
sequnder (TGA):
2 Oligo
Hairpins
0

er

r
de
CT

GA
q ov

q un
-C

-T
se

er

r
de
se
q ov

q un
se

Extended Data Figure 5 | Effect of the 3 motif on protospacer
acquisition when supplied as two complementary oligonucleotides.
Individual sequences designed to directly test the motif identified in Fig. 2b
shown to the left. To the right, percentage of arrays expanded following
electroporation of the sequences indicated as two complementary
oligonucleotides (in dark red), rather than a minimal oligonucleotide
se
hairpin (shown for comparison in pink). Unfilled circles indicate
individual biological replicates. Bars show means.e.m. (n=3; one-way
ANOVA on effect of oligonucleotide: P=0.0041; follow-up testing
with Sidaks multiple comparison (corrected), seqover versus seqover-CCT:
P=0.0103, sequnder versus sequnder-TGA: P=0.0081). *P<0.05.
Additional statistical details in Supplementary Table 2.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter

a Within a pixet

u
t 1
0
1
0
1
1
1
1
1
1
1
1
1
1
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Leader E D C B A
G/P A G/P
1 0 1 1 1 1 1 1 1 1 1 1 0 1 1
1 1 0 1 1 1 1 1 1 1 1 1 1 0 1
0 1 1 1 0 1 1 1 1 1 1 1 0 1 1
1 0 1 1 0 1 1 1 1 1 1 1 0 1 1
0 1 0 1 1 1 1 1 1 1 1 1 1 0 1

r p v
1 0 1 0 1 1 1 1 1 1 1 1 0 1 1
1 1 0 1 1 0 1 1 1 1 1 1 1 0 1

Order Permutations Within Pixet (120)

1 1 1 0 1 1 0 1 1 1 1 1 1 1 0
0 0 1 1 0 1 1 1 1 1 1 1 0 1 1
0 1 1 0 0 1 1 1 1 1 1 1 0 1 1
0 1 0 1 1 1 1 0 1 1 1 1 1 0 1

Fr5 Fr4 Fr3 Fr2 Fr1

1 1 0 1 1 0 1 0 1 1 1 1 1 0 1

w
1 0 1 0 0 1 1 1 1 1 1 1 0 1 1
0 1 1 0 1 1 0 1 1 1 1 1 1 1 0
1 1 1 0 1 1 0 1 0 1 1 1 1 1 0
1 1 0 1 1 1 0 1 1 1 1 1 0 0 1
0 1 0 1 1 0 1 0 1 1 1 1 1 0 1

n l j
0 0 1 0 0 1 1 1 1 1 1 1 0 1 1
0 1 1 0 1 1 0 1 1 0 1 1 1 1 0
1 1 1 0 1 1 0 1 0 0 1 1 1 1 0

G/P B G/P
0 1 0 1 1 1 0 1 1 1 1 1 0 0 1
1 1 0 1 1 0 0 1 1 1 1 1 0 0 1
1 1 0 0 1 1 0 1 1 1 1 1 0 1 0
1 0 0 1 1 0 1 1 1 1 1 0 1 0 1

x
1 0 1 1 1 0 1 1 1 1 1 0 0 0 1
0 1 1 0 1 1 0 1 0 0 1 1 1 1 0
0 1 0 1 1 1 0 0 1 1 1 1 0 0 1
1 1 0 1 1 0 0 0 1 1 1 1 0 0 1

h f d b
1 0 0 1 1 0 1 0 1 1 1 0 1 0 1
1 0 1 1 0 0 1 1 1 1 1 0 0 0 1
0 1 0 0 1 1 0 1 1 1 1 1 0 1 0

y
0 1 0 1 0 1 1 0 1 1 0 1 1 0 1
0 1 1 1 0 1 1 0 1 1 0 1 0 0 1
1 1 0 0 1 1 0 1 0 1 1 1 0 1 0
1 0 1 0 1 1 0 1 0 1 1 0 1 1 0
1 0 1 0 1 1 1 1 0 1 1 0 0 1 0
1 0 0 1 1 0 1 1 0 1 1 0 1 0 1
1 0 1 1 1 0 1 1 0 1 1 0 0 0 1
0 1 0 1 1 0 0 0 1 1 1 1 0 0 1

Leader
1 0 1 1 0 0 1 0 1 1 1 0 0 0 1

E D C B A G/P C G/P
0 1 0 0 1 1 0 1 1 0 1 1 0 1 0
1 1 0 0 1 1 0 1 0 0 1 1 0 1 0
1 0 1 0 1 1 0 1 0 0 1 0 1 1 0
1 0 1 0 0 1 1 1 0 1 1 0 0 1 0
1 0 0 1 1 0 1 0 0 1 1 0 1 0 1

z
1 0 1 1 0 0 1 1 0 1 1 0 0 0 1
0 1 0 1 0 0 1 0 1 1 0 1 1 0 1
0 0 1 1 0 1 1 0 1 1 0 1 0 0 1
0 1 1 0 0 1 0 1 1 0 0 1 1 1 0

g e c a
0 1 1 0 0 1 1 1 1 0 0 1 0 1 0
1 0 1 0 1 0 0 1 0 1 1 0 1 1 0
1 0 1 0 1 0 1 1 0 1 1 0 0 1 0

aa
0 1 0 1 0 1 1 0 1 0 0 1 1 0 1
0 1 1 1 0 1 1 0 1 0 0 1 0 0 1
0 1 0 0 1 1 0 1 0 0 1 1 0 1 0
1 0 0 1 0 0 1 0 1 1 1 0 0 0 1
1 0 1 0 0 1 1 1 0 0 1 0 0 1 0
1 0 1 1 0 0 1 0 0 1 1 0 0 0 1
1 0 1 0 1 0 0 1 0 0 1 0 1 1 0
1 0 1 0 0 0 1 1 0 1 1 0 0 1 0

m k i G/P D G/P
0 0 1 1 0 0 1 0 1 1 0 1 0 0 1
0 1 1 0 0 1 0 1 0 0 0 1 1 1 0
0 0 1 0 0 1 1 1 1 0 0 1 0 1 0
1 1 0 1 1 0 0 1 0 1 1 0 0 0 0
1 0 0 1 1 0 0 1 0 1 1 0 1 0 0

bb
0 1 0 1 0 0 1 0 1 0 0 1 1 0 1
0 0 1 1 0 1 1 0 1 0 0 1 0 0 1
0 1 1 0 0 1 0 0 1 0 0 1 1 1 0
0 1 1 0 0 1 1 0 1 0 0 1 0 1 0
1 0 1 0 0 1 0 1 0 0 1 0 0 1 0
1 0 0 1 0 0 1 0 0 1 1 0 0 0 1

q o
1 0 1 0 0 0 1 1 0 0 1 0 0 1 0

cc
1 1 0 1 1 0 0 0 0 1 1 0 0 0 0
1 0 0 1 1 0 0 0 0 1 1 0 1 0 0
0 0 0 1 0 0 1 0 1 1 0 1 0 0 1
0 0 1 0 0 1 1 1 0 0 0 1 0 1 0
0 1 0 1 1 1 0 0 1 0 0 1 0 0 0
1 1 0 0 1 0 0 1 0 1 1 0 0 0 0
1 0 0 0 1 0 0 1 0 1 1 0 1 0 0
0 0 1 1 0 0 1 0 1 0 0 1 0 0 1

G/P E G/P
0 1 1 0 0 1 0 0 0 0 0 1 1 1 0
0 0 1 0 0 1 1 0 1 0 0 1 0 1 0
0 1 0 1 0 1 0 0 1 0 0 1 1 0 0

s
1 1 0 1 1 0 0 0 0 0 1 0 0 0 0
1 0 1 0 0 0 0 1 0 0 1 0 0 1 0

dd
1 0 0 1 1 0 0 0 0 0 1 0 1 0 0
1 1 0 0 1 0 0 1 0 0 1 0 0 0 0
1 0 0 0 1 0 0 1 0 0 1 0 1 0 0
0 0 1 0 0 1 0 1 0 0 0 1 0 1 0
0 1 0 1 1 0 0 0 1 0 0 1 0 0 0
0 1 0 0 1 1 0 0 1 0 0 1 0 0 0
0 0 0 1 0 0 1 0 1 0 0 1 0 0 1
0 0 1 0 0 1 1 0 0 0 0 1 0 1 0
0 1 0 1 0 0 0 0 1 0 0 1 1 0 0
0 1 0 0 0 1 0 0 1 0 0 1 1 0 0

ordering rules:
1 1 0 0 1 0 0 0 0 0 1 0 0 0 0
1 0 0 0 1 0 0 0 0 0 1 0 1 0 0
0 1 0 1 1 0 0 0 0 0 0 1 0 0 0
0 1 0 0 1 1 0 0 0 0 0 1 0 0 0
1 0 1 1 0 0 1 0 0 0 0 0 0 0 0
0 0 1 0 0 1 0 0 0 0 0 1 0 1 0
0 1 0 1 0 0 0 0 0 0 0 1 1 0 0
0 1 0 0 0 1 0 0 0 0 0 1 1 0 0
0 1 0 0 1 0 0 0 0 0 0 1 0 0 0

u w y aa cc
1 0 0 1 0 0 0 0 0 0 0 0 1 0 0
1 0 0 1 0 0 1 0 0 0 0 0 0 0 0
1 0 1 0 0 0 1 0 0 0 0 0 0 0 0

a>b c>d e>f g>h i>j k>l m>n o>p q>r s>t > > > >
0 1 0 0 0 0 0 0 0 0 0 1 1 0 0
0 0 1 1 0 0 1 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 0 0 1 0 0

V x z bb dd
1 0 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 1 0 0
0 0 0 1 0 0 1 0 0 0 0 0 0 0 0
0 0 1 0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 1 0 0
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0

b Between Pixets

1e 1d 1c 1b 1a 2 1 0 1 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0
2 1 0 0 0 0 0 1 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 0

Fr1 E Fr1 Fr1 D Fr1 Fr1 C Fr1 Fr1 B Fr1 Fr1 A Fr1 2
2
6
1
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
1
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
5
0
0
0
0
0
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
1
0
1
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0

e1 d1 c1 b1 a1
1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 0 0 0
2 1 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 1 0 0 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 0
2 0 1 0 0 0 0 0 1 0 0 0 0 0 1 1 1 0 0 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 0
1 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0

2e 2d 2c 2b 2a
1 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0
2 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0
2 0 0 0 0 0 0 1 1 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0
1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 0 0
2 0 1 0 0 0 0 0 1 0 0 0 0 1 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 1 1 0
2 0 0 0 0 0 1 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 0 0 0
2 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 1 1 0

Order Permutations Within Pixet (120)

1 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0

Fr2 E Fr2 Fr2 D Fr2 Fr2 C Fr2 Fr2 B Fr2 Fr2 A Fr2 2
1
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
0
1
1
0
1
0
1
1
0
1
1
0
1
1
1
1
0
1
1
0
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0

e2 d2 c2 b2 a2
1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 0 1 1 1 0 0 0
1 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 1 1 1 1 1 1 1 1 1 0 0 0
2 0 0 0 1 1 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 1 1 1 1 0 0 1 1 1 1 1 0
2 0 1 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 1 1 0 1 1 1 1 1 0
1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0

3e 3d 3c 3b 3a
2 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0
2 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 0
1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0
1 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 1 1 1 1 1 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 0 0 0

Fr3 E Fr3 Fr3 D Fr3 Fr3 C Fr3 Fr3 B Fr3 Fr3 A Fr3 1
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
1
0
0
0
1
0
1
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
0
0
1
0
0
0
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
0
1
0
0
0
0

e3 d3 c3 b3 a3
1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 1 1 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0

4e 4d 4c 4b 4a
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 0 0 0 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 0 0 0
2 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0
1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0
1 0 0 1 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0

Fr4 E Fr4 Fr4 D Fr4 Fr4 C Fr4 Fr4 B Fr4 Fr4 A Fr4
1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 0 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 1 1 0

e4 d4 c4 b4 a4
1 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1 1 1 0 0 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 0 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 1 1 0
2 0 0 0 0 1 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 0 0
1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 0 1 0 0 0 0 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 1 1 1 1 1 0 1 1 1 0 0 0
1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0

5e 5d 5c 5b 5a
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 0 0 0 0 1 1 1 1 1 0
2 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0
2 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 1 1 1 1 1 0 0 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 0 0 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0

Fr5 E Fr5 Fr5 D Fr5 Fr5 C Fr5 Fr5 B Fr5 Fr5 A Fr5
1 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 0 1 0 0 0 0 0

e5 d5 c5 b5 a5
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0
1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1 0 1 0 0 0 0 0
2 1 0 1 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 0 0 0 0 1 1 1 1 1 0
2 0 0 0 0 0 0 1 1 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0
2 0 0 1 0 0 0 0 0 0 1 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 1 1 1 1 1 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 0 0 0 0 1 1 1 1 1 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 0 1 1 1 0 0 0 1 1 1 1 1 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 1 1 1 0 1 0 0 0 0 0

1e < e1 1d < d1 1c < c1 1b < b1 1a ~ a1

0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 1 0 1 1 1 0 0 0
2 1 0 0 0 0 0 1 0 0 0 0 0 0 1 1 1 1 0 0 0 1 0 0 0 1 1 1 1 1 0 1 1 1 0 0 0
1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 1 1 1 0 1 0 0 0 0 0
1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0

2e < e2 2d < d2 2c < c2 2b ~ b2 2a > a2

1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1 1 1 0 0 1 1 1 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 1 1 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 0
1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0

3e < e3 3d < d3 3c ~ c3 3b > b3 3a > a3

1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 1 1 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 0
2 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0

4e < e4 4d ~ d4 4c > c4 4b > b4 4a > a4

0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
2 1 0 0 1 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
2 0 0 0 1 1 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 0 0 0 0 1 1 1 0 0 0
1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0 1 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 0 0

5e ~ e5 5d > d5 5c > c5 5b > b5 5a > a5

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 1 1 1 0 0 1 0 0 0 0 0
1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 0 0
1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 1 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 1 1 1 0 0 1 0 0 0 0 0

Extended Data Figure 6 | Recall of frame order over time based on
position in the CRISPR array. a, Initial set of rules to test the order
of spacers within a pixet. Every time two spacers from the same pixet
are found in a single array, their relative physical location (with respect
to the leader) is extracted. As is the location of each spacer relative to
spacers drawn from the genome or plasmid (G/P). The actual sequence
of electroporated protospacers should occupy arrays in a predictable
physical arrangement, as described by these ordering rules. Every possible
permutation of spacers within a pixet is tested against each of these rules
and, if a permutation satisfies all the rules, spacers are assigned to frame.
b, Second set of tests to compare between pixets. If no permutation
satisfies all of the tests in a, spacers are compared to previously assigned
spacers from other pixets pairwise when found in the same array. A
larger set of rules will hold true for the actual sequence of electroporated
protospacers when compared against previously assigned spacers. Again,
all possible order permutations are tested, and order is assigned based on
the best overall satisfaction of these ordering rules.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Letter RESEARCH

Error (deviation from intended pixet sequence) Attribution

3.3% Internal nucleotide changes Synthesis, Sequencing, or Mutation

10.7%
Slippage of spacer acquisition start site
12% or integration in reversed orientation Acquisition
74%
No data (no spacer present for pixet)

Acquired spacer is a partial or

Analysis
complete match to E. coli genome

Extended Data Figure 7 | Quantification of errors by source. Includes any instance of a called spacer that does not match the supplied protospacer.

2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Letter
Extended Data Figure 8 | Methods of image encoding for error-
correction. ad, Method used in Fig. 1. a, Triplet code to flexibly specify
21 colours. b, Example of a pixet to be encoded into nucleotide space
with pixel values marked. c, Rules specifying how the protospacer will be
built. d, Example of the build of the protospacer. The AAG introduced
by the addition of pixel 4 is unacceptable and invokes the flexible switch
to another triplet. In a test of the extendibility of this encoding scheme,
we ran three random sets of 100 million different nine-colour orderings
through the sequence build and found that 99.860.07% of colour orders
were able to satisfy the requirements we set out without optimization by
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
hand. ei, Method of alternating clusters for error correction. e, Triplet
assignment to clusters A, B, and X. f, Example of a pixet to be encoded
into nucleotide space with pixel values marked. g, Rules for adding new
triplets in this scheme. h, Example of the build of the protospacer. The
AAG introduced by the addition of pixel 4 is unacceptable and invokes the
flexible switch to cluster X. i, Example of an error signal. jl, Method of
checksum error correction. j, Annotation of protospacer with the addition
of a checksum. k, Annotation of the checksum itself. l, Full protospacer
with checksum implemented.