Anda di halaman 1dari 17

Detemukan Hormon Pengendali Rasa Lapar Senin, Detemukan Gizi.

net 14 November, 2005 oleh: Hormon Pengendali Rasa DITEMUKAN, HORMON PENGENDALI RASA Gsianturi Lapar LAPAR

Para ilmuwan telah menemukan hormon yang berkaitan dengan pengaturan rasa lapar. Obestatin, hormon ini, bergabung dengan beberapa hormon lain yang fungsinya menekan atau meningkatkan nafsu makan seseorang. Majalah Science menuliskan bahwa tim dari Stanford University Amerika melakukan pencarian informasi genetik tentang obestatin ini menggunakan komputer. Diharapkan dalam 5 hingga 10 tahun ke depan penelitian yang dilakukan oleh para ilmuwan dapat menghasilkan sesuatu yang dapat mengontrol nafsu makan. Obestatin ditemukan ketika para peneliti mengamati rangkaian gen dalam tubuh manusia dan hewan, untuk mencari ghrelin, hormon pendorong nafsu makan. Nah, saat itulah mereka menemukan suatu hormon lain - kemudian dinamakan obestatin - terbentuk dari protein yang juga memproduksi ghrelin. Walaupun dibentuk dari protein yang sama, kedua hormon ini memiliki fungsi yang berkebalikan . Obestatin bekerja untuk mengurangi nafsu makan, sedangkan fungsi ghrelin adalah menambahnya. Sebuah teka-teki Ketika tikus-tikus diberi suntikan obestatin pada perut dan otak mereka, porsi makannya menjadi setengah dari apa yang dimakan oleh hewan yang tidak disuntik hormon ini. Tikus-tikus ini juga mengalami penurunan berat badan. Selain mengurangi nafsu makan, obestatin juga memperlambat proses pencernaan dengan memperlambat gerakan makanan dari perut ke usus. Dalam tulisan di Science, kelompok ilmuwan yang dipimpin oleh Dr Aaron Hsueh mengatakan, "Pemahaman yang lebih baik tentang fungsi ghrelin dan obestatin dalam keseimbangan pengendalian energi dan pengendalian berat badan mungkin bisa dipakai untuk mengatasi masalah kegemukan." Prof. Steve Bloom, seorang pakar masalah obesitas dari Imperial College London mengatakan "Ini adalah penelitian yang luar biasa. Ini menunjuk pada bagian lain sistem pengaturan nafsu makan sehingga kita bisa menghindari penggunaan obat-obatan jika ingin langsing." Prof. Bloom menambahkan, penelitian ini menghasilkan penemuan baru untuk mengatur rasa lapar. "Dalam jangka waktu 5 hingga 10 tahun ke depan, kita akan dapat mengatur nafsu makan." Namun, kenyataannya bahwa hormon penekan nafsu makan dibentuk dari sel yang sama dengan hormon penambah nafsu makan adalah sebuah teka-teki bagi para peneliti. (k-1) Sumber: http://www.kompas.co.id/teknologi/news/0511/14/115018.htm 14 November 2005

Disruption of Growth Hormone Receptor Prevents Calorie Restriction from Improving Insulin Action and Longevity
Abstract Top
Most mutations that delay aging and prolong lifespan in the mouse are related to somatotropic and/or insulin signaling. Calorie restriction (CR) is the only intervention that reliably increases mouse longevity. There is considerable phenotypic overlap between long-lived mutant mice and normal mice on chronic CR. Therefore, we investigated the interactive effects of CR and targeted disruption or knock out of the growth hormone receptor (GHRKO) in mice on longevity and the insulin signaling cascade. Every other day feeding corresponds to a mild (i.e. 15%) CR which increased median lifespan in normal

mice but not in GHRKO mice corroborating our previous findings on the effects of moderate (30%) CR on the longevity of these animals. To determine why insulin sensitivity improves in normal but not GHRKO mice in response to 30% CR, we conducted insulin stimulation experiments after one year of CR. In normal mice, CR increased the insulin stimulated activation of the insulin signaling cascade (IR/IRS/PI3K/AKT) in liver and muscle. Livers of GHRKO mice responded to insulin by increased activation of the early steps of insulin signaling, which was dissipated by altered PI3K subunit abundance which putatively inhibited AKT activation. In the muscle of GHRKO mice, there was elevated downstream activation of the insulin signaling cascade (IRS/PI3K/AKT) in the absence of elevated IR activation. Further, we found a major reduction of inhibitory Ser phosphorylation of IRS-1 seen exclusively in GHRKO muscle which may underpin their elevated insulin sensitivity. Chronic CR failed to further modify the alterations in insulin signaling in GHRKO mice as compared to normal mice, likely explaining or contributing to the absence of CR effects on insulin sensitivity and longevity in these long-lived mice.

Introduction Top
A vast majority of the mutations that delay aging and prolong lifespan in the mouse (Mus musculus) either directly or indirectly alter somatotropic and/or insulin signaling [1], [2]. There is considerable intra- and extra-cellular crosstalk between growth hormone (GH), insulin-like growth factor 1 (IGF-1), and insulin in mediating growth and metabolism in mammals [3]. These signaling pathways are highly conserved across phyla; and mutations affecting homologous IGF-1/insulin-like signaling and downstream gene expression increase longevity in yeast (Saccharomyces cerevisiae), worms (Caenorhabditis elegans), and flies (Drosophila melanogaster) [4], [5]. Taken together, the above-mentioned findings posit GH/IGF-1/insulin (and homologous) signaling as one of the key mediators of longevity. Calorie restriction (CR) is the only environmental treatment known to consistently increase average and maximal lifespan and delay aging in organisms ranging from yeast to mammals [6], [7]. In addition to extended longevity and reduced cancer incidence, the most consistent responses to CR in mammals include reductions in peripheral (i.e. blood) insulin, GH, IGF-1, and glucose levels [6], [8]. These biomarkers of a CR response are reported in species ranging from mice to humans [4]. Interestingly, there is considerable phenotypic overlap between mice on CR and many long-lived mutant mice. These similarities include reductions in body weight, body size, neoplastic disease incidence, peripheral GH/IGF1, insulin, and glucose, relative to their respective controls [9][11]. In addition, most of the long-lived mutant mice and mice on long-term CR show improvements in insulin sensitivity, feed efficiency and health-span. These similarities suggest that studying interactions between the life-extending mutations and CR may reveal what pathways and mechanisms are utilized by CR to alter aging. GH receptor/binding protein knockout (GHRKO) mice were developed by targeted disruption of the GHR/GHBP gene [12]. This insertion mutation renders the GH receptor nonfunctional and causes severe GH resistance, leading to suppression of peripheral IGF-1 and insulin levels. In addition, GHRKO mice are markedly insulin-sensitive, tend to be hypoglycemic, have reduced body weight and size, reduced incidence of neoplastic disease and a pronounced increase in all lifespan parameters, including increased mortality rate doubling time (MRDT), as well as median, average and maximal lifespan [10], [13], [14]. We previously reported that in contrast to hypopituitary, GH-deficient Ames dwarf mice [15], GHRKO mice did not receive a CR benefit in insulin sensitivity and most measures of longevity [16]. From a genetic perspective, these findings suggest that CR may work partially via GH signaling to mediate improvements in insulin sensitivity and extended lifespan, especially since it is well known that GH can induce insulin resistance [17][19]. One possible caveat is that when using a single level of CR, it is difficult to conclude whether the mutation ablates the effects of CR at all doses (i.e., flat-lines the doseresponse curve) or if the mutation altered (or shifted) the CR dose response curve [20]. While both outcomes are important, the former would implicate GHR-based signaling as the paramount pathway controlling the effects of CR on aging and lifespan in mice and thus potentially one of the pathways that may be exploited to induce CR-like effects in humans.

We report here two long-term studies, of the interaction between GHR disruption and CR in male mice that were designed to address the following questions: 1) Does the GHRKO mutation modulate or completely ablate the benefits of CR on lifespan? 2) How does CR improve insulin sensitivity and longevity in normal mice? 3) Why does CR fail to improve insulin sensitivity and longevity in GHRKO mice?

Results and Discussion Top


Longevity characteristics of normal and GHRKO mice on mild CR To determine whether the previously documented differential responses of GHRKO and normal (N) mice to 30% CR may have been limited to this level of dietary restriction, we conducted an additional longevity study using a milder degree of CR in both mice. Every other day (EOD) feeding resulted in approximately a 1015% reduction in the average daily food intake compared to ad libitum (AL) control mice (data not shown) [21], [22]. In normal male mice, EOD feeding led to a 16% increase in median lifespan (N AL = 851 days vs. N EOD = 1010 days, hazard ratio (HR) 2.62 confidence interval (CI) 1.316.03). In contrast, median lifespan of GHRKO mice was not extended (KO AL = 1178 days vs. KO EOD = 1158 days, HR 1.063, CI 0.502.29). Compared to N AL mice, EOD feeding increased overall lifespan in normal mice as evaluated by log-rank analysis (P<0.008); however, there were no significant diet effects in GHRKO mice (P = 0.6727). Thus, a milder and different form of CR that increased longevity in normal control mice failed to further increase the lifespan of long-lived GHRKO mice, corroborating our previous findings using a more severe (30%) traditional CR regimen [16]. Taken together, these studies indicate that GH resistance ablates rather than modulates (or shifts) the benefits of CR on longevity in male GHRKO mice. Effects of long-term 30% CR on metabolic characteristics of normal and GHRKO mice To search for potential mechanisms for the differential responses of GHRKO and normal mice to CR and to further investigate the interaction between CR, GH resistance and insulin sensitivity, 30% CR was initiated at 2 months of age and maintained for one year in normal and GHRKO male mice. In comparison to animals fed AL, CR produced the expected decrease in body weight trajectory in both normal and GHRKO mice (Figure 1A). Furthermore, we fasted mice in all groups overnight to determine body weight in the nonfed state. Compared to AL controls, one year of 30% CR led to reduced fasted body weight in both normal and GHRKO mice (Figure 1B). There were no significant differences in the fasted peripheral glucose levels in any of the groups tested (Figure 1C). In normal mice, CR decreased the level of fasted insulin (P<0.001), while both AL and CR GHRKO mice had reduced insulin levels compared to normal mice on either diet (P<0.001 for each comparison). Calorie restriction did not further decrease fasted peripheral insulin levels in GHRKO mice (Figure 1D).

Figure 1. Metabolic parameters of male GHRKO (KO) and normal (N) mice fed ad libitum (AL) or subjected to 30% calorie restriction (CR) for one year. Panel (A) shows time-course changes in body weight over a one year period. After one year of CR, all groups were fasted overnight and bodyweight (B), peripheral glucose (C), insulin (D), and calculated homeostatic model of assessment (HOMA, E) were determined. Values with unlike superscripts/letters are significantly different (P<0.05). doi:10.1371/journal.pone.0004567.g001 The relationship between peripheral glucose and insulin levels is an important indicator of the overall insulin sensitivity of the animal. One model that utilizes fasted glucose and insulin levels to represent the whole organism's insulin sensitivity is the Homeostatic Model of Assessment (HOMA), with a relatively lower score representing improved insulin sensitivity. Figure 1E shows that in normal mice, one year of CR led to a reduction

in the HOMA score compared to AL controls (P<0.01), while CR did not alter the already improved HOMA score in GHRKO mice. To summarize, CR reduced body weight in both phenotypes and improved fasting insulin sensitivity in normal mice, but failed to further improve the already insulin-sensitive status of GHRKO mice, thus uncoupling the effects of CR on body weight/growth from its effects on insulin sensitivity and lifespan. To further explore the interaction between GH resistance, CR, and insulin sensitivity, we investigated the activation state of the insulin signaling cascade in these groups in response to an exogenous insulin challenge. Effects of CR on the response of normal and GHRKO mice to insulin stimulation The actions of insulin are critical for the regulation of the processes involved in glucose homeostasis. The maintenance of carbohydrate homeostasis involves multiple enzymatic processes, in addition to cellular signaling processes, with liver and muscle playing different roles. To determine the molecular mechanism of differential impact of CR on insulin sensitivity in normal mice and GHRKO mice, we evaluated insulin signal transduction through the IR/IRS/PI3K/AKT1/AKT2 pathway in liver and skeletal muscle in normal and GHRKO mice fed AL or subjected to CR for one year in response to an insulin challenge (10 IU/kg insulin versus saline). Liver responses to insulin stimulation Calorie restriction increased total amount of insulin receptor (IR) protein in normal mice (P<0.03), but did not increase the already-elevated levels of IR in GHRKO mice on both AL and CR diets (Figure 2A). The heightened insulin induced upstream events in GHRKO mice corresponds with previous studies in these insulin sensitive mutants [23], [24]. In response to acute insulin stimulation, normal mice on CR had a 2-fold greater increase in IR activation (as estimated by measuring IR phosphorylation at tyrosine (Tyr) residue 1158 [pY1158] as a proxy for activation) than the normal AL mice, while CR did not increase the already elevated insulin-induced IR activation in GHRKO mice (Figure 2B). Figure 2. Liver insulin signaling cascade total proteins and phospho-proteins in response to insulin stimulation (10 IU/kg) versus saline treated controls in GHRKO (KO) and normal (N) mice fed ad libitum (AL) or 30% calorie restricted (CR) for one year. ELISA with antibodies directed towards total insulin receptor (IR; A), IR pY1158 (B), AKT1 (D), and AKT1 pS473 (E) were performed. Liver homogenates were immunoprecipitated (IP) with anti-p85, separated using SDS-PAGE and the level of nonspecific Tyr phosphorylation was determined at approximately 180 kDa using anti-pY99 (C). Total AKT2 protein (F) and AKT2 pS473/474 (G) was determined from liver protein homogenates first subjected to IP using anti-AKT2. Panel (H) summarizes key insulinstimulated phospho-proteins compared to insulin-stimulated N AL mice. Bands are representative blots from 46 male mice per phenotype/diet/treatment group. Values with unlike superscripts/letters are significantly different (P<0.05). doi:10.1371/journal.pone.0004567.g002 Next, we determined the levels of activation of a major intracellular signaling intermediate in the insulin signaling cascade, namely phosphatidylinositol-3 kinase (PI3K). In normal mice, CR significantly increased insulin-induced p85/IRS pY99 phosphorylation from 1.8fold (N AL) to 2.2-fold in (N CR). However, this parameter was not affected by CR in GHRKO mice (Figure 2C). This finding suggests that while the enhanced amplitude of the insulin signal from the hepatic IR was maintained at the level of PI3K in normal mice on CR, GHRKO mice on either diet did not show the level of activation (as presumed by phosphorylation) that corresponded to the upstream activation of the insulin receptor (Figure 2B). An important downstream target of the insulin/PI3K signaling cascade is protein kinase B (AKT1). Total AKT1 protein in liver did not differ in any of the groups tested (Figure 2D). In normal mice, CR increased the insulin-induced phosphorylation of AKT1 at serine (Ser) 4

residue 473 (pS473, a marker of activation of this enzyme) compared to values measured in insulin-stimulated N AL mice (P<0.03, Figure 2E). Insulin-induced pS473 levels of AKT in KO AL and KO CR groups did not significantly differ from the corresponding values in N AL or N CR mice. In the liver of normal mice, CR increased AKT2 total protein compared to AL controls (P<0.03, Figure 2F). GHRKO mice had comparable AKT2 total protein levels when compared to N AL mice. Acute insulin stimulation increased the levels of Ser phosphorylated AKT2 (pS474) in all groups (P<0.001). When compared to phenotype/treatment saline controls within groups, insulin increased the phosphorylation in N AL by 92%, while N CR, KO AL, and KO CR groups had more than a 2 fold increase in insulin stimulated phosphorylation (N CR 2.1 fold, KO AL 2.05 fold, and KO CR 2.1 fold). Hepatic insulin signaling cascade activation in response to insulin stimulation in normal mice on CR and in GHRKO mice on both diets is compared to N AL controls and summarized in Figure 2H. These data suggest that CR increases the levels of proteins involved in insulin signaling transduction and the activation of these proteins in response to insulin in normal mice, but has no effects in GHRKO mice. Curiously, the enhanced activation of the IR in response to insulin stimulation is not propagated downstream to the AKT proteins in GHRKO mice, whereas it is in the normal mice. In an attempt to understand this discrepancy, we explored the regulatory subunits of PI3K in normal and GHRKO mice fed AL or subjected to 30% CR. Determination of the PI3K regulatory subunits p85/p55/p50 To determine the relative protein abundance of the regulatory subunits of PI3K in the liver, we targeted the p85 subunit of PI3K for immunoprecipitation in liver homogenates and subsequently immunoblotted for the p85, p55 and p50 subunits of PI3K (Figure 3). In the liver of normal mice, CR did not alter p85 (Figure 3A), p55 (Figure 3B) or p50 (Figure 3C) total protein concentration. Similar to these findings, CR did not alter the level of these subunits in GHRKO mice, with the exception of a significant reduction in the p55 subunit compared to GHRKO AL controls (P<0.05). It has been shown that an alteration in a single regulatory subunit of PI3K may act to inhibit or bind up free IRS proteins and alter signal transduction [25], [26]. This could explain the partial decrease in signal amplification seen between the activation of the IR and PI3K phosphorylation in mice from the normal CR group. Irrespective of the diet, GHRKO mice had reduced p85, p55, and p50 regulatory subunits compared to normal mice. CR did reduce the p55 subunit in GHRKO mice. We believe this reduction of the regulatory subunits of PI3K in GHRKO mice may underpin the dampened downstream activation (compared to activation of IR) of the PI3K/IRS, AKT1 and AKT2 proteins seen in the insulin stimulation experiments (Figures 2 and 3). Figure 3. Liver PI3K subunit abundance in GHRKO (KO) and normal (N) mice fed ad libitum (AL) or 30% calorie restriction (CR) for one year. Liver protein isolates were immunoprecipitated (IP) with anti-pan-p85. Total p85 (A), 55 (B) and 50 (C) subunits were separated using SDS-PAGE, transferred to nitrocellulose membranes and blotted with anti-pan-p85. Bands are representative blots from 46 male mice per phenotype/diet/treatment group. Values with unlike superscripts/letters are significantly different (P<0.05). doi:10.1371/journal.pone.0004567.g003 Effects of insulin stimulation in the skeletal muscle Skeletal muscle is the major site of insulin-stimulated glucose uptake in the body, as demonstrated by various tissue-specific knockouts of the glucose transporter GLUT4 [27]. In normal mice, CR increased total IR from 6.920.64 to 8.870.47 units/100ug (P<0.02, Figure 4A). In contrast, GHRKO AL mice had elevated muscle IR protein that was reduced by CR (9.010.90 to 6.720.59 units/100ug; P<0.01). Insulin-stimulated phosphorylation of the IR in normal CR mice was increased approximately 3-fold compared to AL controls (P<0.001), while IR phosphorylation in GHRKO mice was unaffected by the diet, and values obtained were comparable to normal AL mice (Figure 4B). This finding was also 5

reported previously [24], and this does not correspond with their heightened insulin stimulated glucose uptake [16] so we therefore further investigated downstream insulin signaling events associated with insulin mediated glucose uptake. Figure 4. Muscle insulin signaling cascade total proteins and phospho-proteins in response to insulin stimulation (10 IU/kg) versus saline treated controls in GHRKO (KO) and normal (N) mice fed ad libitum (AL) or 30% calorie restriction (CR) for one year. Levels of total insulin receptor (IR; A), IR pY1158 (B), AKT1 (D), and AKT1 pS473 (E) were determined using ELISA. Liver homogenates were immunoprecipitated (IP) with anti-p85, separated using SDS-PAGE and the level of non-specific Tyr phosphorylation was determined at approximately 180 kDa (corresponding to IRS proteins) using anti-pY99 (C). Total AKT2 protein (F) and AKT2 pS473/474 (G) was determined from liver protein homogenates first subjected to IP using anti-AKT2. Total GLUT4 protein (H) was determined in muscle homogenates using anti-GLUT4. Panel (I) summarizes key insulinstimulated phospho-proteins compared to insulin-stimulated N AL mice. Bands are representative blots from 46 male mice per phenotype/diet/treatment group. Values with unlike superscripts/letters are significantly different (P<0.05). doi:10.1371/journal.pone.0004567.g004 Calorie restriction produced a significant increase in the PI3K/IRS phosphorylation of IRS associated with PI3K in the muscle of normal mice, yet did not further elevate the already heightened phospho-Tyr levels in the GHRKO mice (Figure 4C). This finding did not coincide with IR activation, where GHRKO mice did not show elevations in IR phosphoylation, yet there was an increase in the level of IRS phosphorylation at Tyr 99. To further probe this finding, we assessed the total protein and phosphor-Ser levels of the downstream targets AKT1, AKT2, and mTOR, as well as the total protein content of GLUT4. AKT proteins in skeletal muscle are critically involved in insulin-stimulated glucose uptake through the trafficking of GLUT4 to the cell surface. In normal mice, CR increased total AKT1 protein (N AL 11.20.94 versus N CR 16.42.0, P<0.01, Figure 4D). In contrast, CR did not increase the already-elevated level of AKT1 total protein in GHRKO mice. Calorie restriction increased insulin-induced AKT1 phosphorylation in normal mice (P<0.01), but it did not further increase the already-elevated AKT1 response to insulin in GHRKO mice (Figure 4E). This data coincide with the increases of p85/IRS activation in muscle homogenates shown in Figure 4C. As mentioned previously, AKT2 has been shown to mediate the stimulatory effects of CR on improved glucose uptake in skeletal muscle [27]. Calorie restriction did not alter muscle AKT2 protein abundance in either phenotype, but total AKT2 protein was elevated in GHRKO AL mice compared to normal AL controls (P<0.05, Figure 4F). It should be noted that normal CR mice had numerically elevated AKT2 values (compared to normal AL mice) that were not significantly different from the elevated values in GHRKO groups. Calorie restriction increased AKT2 phosphorylation in response to insulin stimulation in normal mice, compared to insulin-stimulated AL controls (P<0.01), while GHRKO mice had elevated phosphor-Ser 474 AKT2 levels (P<0.01 for both, Figure 4G), with no further changes in response to CR. The similarity of activated AKT1 and AKT2 between normal mice on CR and GHRKO mice on either diet coincides with the improved insulin sensitivity in these groups compared to normal AL controls, and with our previous results concerning CR effects on longevity [16]. Since the AKT proteins have been shown to regulate insulin-stimulated GLUT4 trafficking, we next determined the level of GLUT4 protein in skeletal muscle whole cell homogenates. Muscle GLUT4 total protein levels were significantly increased by CR in normal mice (P<0.003, Figure 4H). However, CR did not alter total protein levels in GHRKO mice. Both KO AL and KO CR mice had elevated muscle GLUT4 total protein levels compared to N AL controls (P<0.03 and P<0.04, respectively). These findings indicate that normal mice on CR and GHRKO mice on both diets have increased pools of GLUT4 available for recruitment. Figure 4I illustrates/summarizes the insulin signaling activation state of 6

normal mice on CR and GHRKO mice (on both diets) compared to normal AL controls. GHRKO mice did not have elevated IR activation in comparison to their normal littermates, yet downstream PI3K/IRS-1, AKT1, AKT2 and GLUT4 levels or activity were elevated. While CR increased the activation of all proteins tested in normal animals, when compared to AL controls, the effects of CR in GHRKO mice were absent or opposite to those found in normal mice on CR. To further investigate this incongruity, we more closely examined the level of inhibition between the IR and PI3K signaling molecules. Calorie restriction and/or GH resistance decreases mTOR signaling and phosphoSer inhibition of IRS-1 in skeletal muscle A number of recent reports show that Ser/Thr kinases mTOR, JNK, and PKC mediate a negative feedback loop in the insulin signaling cascade by phosphorylating IRS-1 at Ser position 307 [28], [29]. While IRS proteins are normally Tyr-phosphorylated upon IR activation, Ser phosphorylation prevents binding to the beta subunit of the IR and subsequent Tyr phosphorylation, thereby preventing the progression of insulin signal transduction. Since we have previously reported a reduction in JNK activation and in PKC in the skeletal muscle of GHRKO mice [30], we investigated the levels of IRS-1 total protein and phospho-Ser307 IRS-1 in normal and GHRKO mice on both dietary regimens. Total IRS-1 protein levels were not different in any of the groups tested (Figure 5A). Compared to normal mice, GHRKO mice had more than a 50% reduction in the phosphorylation of IRS-1 at Ser 307 (Figure 5B). We believe that the decrease in IRS-1 phosphorylation at this inhibitory site (Ser 307) in GHRKO mice underpins the insulinstimulated elevations in the activation of downstream PI3K/IRS, AKT1 and AKT2 in the absence of increased IR activation upstream. Figure 5. Muscle total IRS-1 protein (A) and IRS-1 pS307 inhibitory phosphorylation (B) in GHRKO (KO) and normal (N) mice fed ad libitum (AL) or 30% calorie restriction (CR) for one year was determined using ELISA. Bars represent 68 animals per phenotype/diet superscripts/letters are significantly different (P<0.05). doi:10.1371/journal.pone.0004567.g005 It was recently reported in another long-lived, insulin-sensitive mouse, the Ames dwarf, that mTOR signaling was reduced in skeletal muscle [30]. Since mTOR has been directly implicated in mediating IRS-1 inhibition through phosphorylation at Ser 307, we determined the level of mTOR total protein and its phosphorylated isoform in normal and GHRKO mice on both diets. In normal mice, CR reduced total mTOR protein compared to AL controls (P<0.01), while CR did not further reduce the already low levels of mTOR in GHRKO mice (Figure 6A). The activation of mTOR through phosphorylation at Ser 2448 followed the same trends as total content of mTOR protein, where CR reduced basal activation (pS2448) of mTOR in normal mice but did not alter the already low levels of activation in GHRKO mice (Figure 6B). We believe that differences in mTOR signaling implied by these results together with reduced JNK and PKC activation previously reported by our laboratory [31] importantly contribute to, and perhaps account for, the decrease in skeletal muscle inhibitory IRS-1 Ser 307 phosphorylation and improved muscle and whole-animal insulin sensitivity in GHRKO mice. Figure 6. Muscle total mTOR protein (A) and mTOR pS2448 (B) in GHRKO (KO) and normal (N) mice fed ad libitum (AL) or 30% calorie restriction (CR) for one year were determined using western blotting. Bars represent 68 animals per phenotype/diet group. doi:10.1371/journal.pone.0004567.g006 Conclusion In the present study, we addressed three major research questions: 1) Does the GHRKO mutation modulate, or ablate, the effects of different intensities of CR on longevity? 2) 7 group. Values with unlike

How does CR improve insulin sensitivity and longevity in normal mice? and 3)Why does CR fail to improve insulin sensitivity and longevity in GHRKO mice? In regard to the first question, we showed that while longevity of normal mice increased as expected in response to various intensities of CR, GHRKO mice did not derive longevity benefits from identical dietary interventions. This flattening of the median lifespan response in GHRKO mice on various levels of food intake shows that, in the absence of a functional GH receptor, CR does not affect longevity of male GHRKO mice. In other words, GH resistance ablates, rather than modulates, the key benefit of CR in these animals. In yeast and invertebrates the interactions between CR (or dietary restriction) and mutations that extend longevity have been explored [reviewed in 32]. While there seems to be homologous insulin signaling pathway mutations that ablate the CR response in yeast [i.e. Sch9, an AKT homologous protein and TOR, 33], mutations in chico (IRS-1 homologous proteins) modulate, rather than block, the CR response in flies [34]. These findings in less complex organisms are confounded by two major complications. First, yeast and invertebrates have not evolved GH signaling pathways. This fact makes it difficult to interpret findings across species. Second, the method by which dietary restriction, or CR is delivered in yeast and invertebrates is quite different than the traditional mammalian model, further complicating interaction studies [32]. In normal mice on CR, there is a major amplification of the insulin signaling cascade, with increases in protein concentration and/or (insulin-stimulated) activation of almost all the cellular mediators of insulin action tested in the present study (summarized in Figure 7). These increases in the insulin signaling cascade in liver and muscle show that the previously documented improvements in whole-animal insulin sensitivity in response to CR reflect heightened insulin cascade action. We previously postulated that improved whole-animal insulin sensitivity is tightly coupled to extended longevity [16]. Figure 7. Summarizes the effects of CR, or lack thereof, in wildtype and GHRKO mice under insulin-stimulated conditions. Solid grey proteins represent stimulatory activation, while solid black proteins are inhibitory. doi:10.1371/journal.pone.0004567.g007 The present results suggest two differentbut not necessarily mutually exclusive answers to our third question: Why does CR fail to increase insulin sensitivity and longevity in GHRKO mice? Since skeletal muscle is the major mediator of post-prandial, insulin-stimulated glucose uptake, we believe our data showing heightened muscle insulin action in both normal mice on CR and GHRKO mice regardless of diet underpin their improved insulin sensitivity (summarized in Figure 7). This convergence between normal mice exposed to long-term CR and innately insulin-sensitive GHRKO mice could explain why CR fails to further improve insulin sensitivity of GHRKO mutants or further extend their already-long lives. Another conclusion can be drawn from the differential effects of CR on the responses to an insulin challenge in the liver's insulin signaling cascade. While normal mice on CR showed improved downstream insulin action through increases in total protein and/or insulin-stimulated activation of several key elements of the insulin signaling cascade in liver, CR failed to produce corresponding changes in GHRKO mice. Additionally, both KO AL and CR groups had reduced downstream insulin-induced activation. This tissue-specific phenotypic dimorphism may have contributed to the absence of CR effects on whole-animal insulin sensitivity and lifespan in GHRKO mice. Collectively, the present study identifies molecular steps of insulin signaling in the liver and in the muscle that are similarly affected by two conditions that extend life, CR in normal animals and deletion of the GH receptor, but are not altered by CR in GHRKO mice in which CR fails to increase insulin sensitivity and therefore longevity. This includes four parameters in liver and 8 in muscle (plus one altered by CR in the opposite directions in the two phenotypes). We suggest that these steps of insulin signaling are involved in the control of whole animal insulin sensitivity and mammalian longevity.

Materials and Methods Top


8

Animals GHRKO and normal mice were produced in the breeding colony at SIU-SOM Springfield, IL which was developed from mice kindly provided by J. J. Kopchick (Ohio University, Athens). Phenotypically normal siblings of GHRKO mice served as controls in these studies. Animals were weaned at 3 weeks of age and placed on Lab Diet Formula 5001 (Ralston Purina, St. Louis, MO). Animals were housed at 2023C on a 12/12 hour light/dark cycle. Sentinel animals were sent for serological testing every 3 months and the results were uniformly negative. These studies were approved by the Laboratory Animal Use and Care Committee of SIU-SOM. Longevity Study The every other day (EOD) feeding paradigm was conducted in normal and GHRKO mice starting at approximately 8 weeks of age. Mice were gradually shifted to EOD by placing them on every third day fast for the first week and then EOD feeding throughout the remainder of the study. The survival of male mice used in this study (approximately 1520 per phenotype/dietary treatment) was evaluated once all four groups reached median lifespan. At the time of analysis, the remaining survivors from this still ongoing study were 6% of the N AL, 43% of the N CR, 41% of the KO AL and 46% of the KO CR mice. Logrank analysis for partial survivorship was conducted. At the time of analysis, all female treatment groups did not yet reach median lifespan. Animals in the longevity study were checked daily for health and survival, and were handled only for cage changes and biweekly body measurements. Animals that appeared near death (listless, unable to walk, and cold to the touch) or had large bleeding neoplastic growth approaching 10% of their body weight were euthanized and date of euthanasia was considered date of death. Metabolic Parameters In the mice on the terminal insulin stimulation study, metabolic parameters were evaluated from blood collected from the orbital plexus after 10 months of 30% CR. Following a 12 hour fast, animals were anesthetized using isoflurane (Abbott Laboratories, Chicago, IL) and glucose levels were determined using a glucometer (Lifescan, Johnson & Johnson, New Brunswick, NJ.) and insulin levels were determined using an enzyme-linked immunosorbant assay (ELISA, Crystal Chem. Inc., Downers Grove, IL). Homeostatic Model of Assessment (HOMA) score was calculated as 22.5/(insulin [mU/L] X glucose [mmol/L]. Insulin Stimulation Study GHRKO mice and phenotypically normal male siblings were gradually calorie restricted starting at 8 weeks of age, by receiving 90% of the amount of food consumed by AL controls in the initial week, 80% the following week, and 70% for the remainder of the study. After one year of 30% CR, mice from all 4 groups (N AL, N CR, KO AL, and KO CR, n = 1518 per group) were anesthetized [ketamine (100 mg/ml):xylazine (20 mg/ml), at a 0.1 ml per 10 grams of body weight] and injected in the inferior vena cava with either 10 IU/kg of porcine insulin (Sigma, St. Louis, MO) or with saline (control). Liver and hind limb muscle tissues were extracted and snap frozen at 1 and 3 minutes, respectively, after injection. Protein and Phosphoprotein Determination After storage at 80C, liver and muscle tissues in T-Per (Pierce, Rockford, IL) with phosphatase inhibitor (Pierce, Rockford, IL) and protease inhibitors (Sigma, St. Louis, MO) at approximately 1 ml per 1 g of tissue. ELISA for insulin (CrystalChem, Downers Grove, IL), IR, IR pY1158, IRS-1, IRS-1 pS312, AKT1 and AKT1 pS473 were used for quantification of cellular proteins in accordance with manufacturer's instructions (Biosource Camarillo, CA). For immunoprecipitation (IP), equal amounts of solubilized protein were incubated at 4C overnight with ~4ug/ml of specific antibody. Polypeptide-Immuneglobin complexes were then collected by incubation with protein A-sepharose beads (Sigma, St. Louis, MO) 9

washed solubilization buffer A, and boiled in Laemmli sample buffer (Biorad, Hercules, CA). Antibodies (Ab) against pan p85, AKT2, AKT (pS473), mTOR, mTOR (pS2448), and GLUT4 (Cell Signaling Technology, Inc. Danvers, MA) were used according to manufacturer's suggestions. Non-specific IRS Tyr phosphorylation of p85 was determined by detecting anti-pY99 (Santa Cruz Biotechnology, Santa Cruz, CA) signal in ~180 kDa region in the anti-pan p85 immunoprecipitates. Immunoblotting was performed according to standard protocol. Protein content in tissue homogenates extracted with T-Per (described above) was determined by the BCA protein assay (Pierce, Rockford, IL). Protein lysates (4060 ug) were subjected to SDS-PAGE using Criterion XT Precast Gels (BioRad, Hercules, CA), transferred to nitrocellulose membranes (BioRad, Hercules, CA) and even loading was verified using MemCode Reversible Protein Stain Kit (Pierce, Rockford, IL). Membranes were blocked for 1 hour at room temperature with either 5% dry skim milk or 3% BSA when determining phosphorylated proteins, in TBST (TBS+0.05% Tween-20). Blots were then washed with TBST and incubated with the primary Ab diluted in the appropriate Ab-specific blocking solution suggested by the manufacturer. ECL Plus chemiluminescent reagent was used for immunodetection (G.E. Healthcare Bio-sciences Corp, Piscataway, NJ). Digital images of blots were captured using a CCD camera (Hitachi Genetic Systems, Ameda, CA) and quantified using GeneTools software (SynGene, Cambridge, England). Statistical analysis Kaplan-Meier survival curves were used for survival analysis using log-rank test to evaluate significance of survivorship between groups. Median lifespan represents the age at which 50% of the population within groups remained alive and was reported with the corresponding hazard ratio (HR) and confidence interval (CI). Analysis of phenotype X diet was used to reveal interactions of characterization factors upon the phenotype under analysis, with subsequent independent Student's t-tests for within group comparisons. Analysis of basal and insulin-stimulated protein concentration across all groups was analyzed with three-way Analysis of Variance (ANOVA) followed by subsequent two-way ANOVA's and Students t-tests within groups when warranted. All statistics were conducted using SPSS version 13.0 (SPSS, Chicago, IL) with = 0.05. All graphs were made using Prism 4.02 (GraphPad Software, San Diego CA.). Alpha is set to 0.05. With the exception of longevity data, all values are reported as meanStandard Error of the Mean (SEM) throughout the figures and text.

Role of Haptoglobin in Polycystic Ovary Syndrome (PCOS), Obesity and Disorders of Glucose Tolerance in Premenopausal Women

10

Abstract Top
Background
Hp2 alleles of the haptoglobin chain polymorphism reduce the anti-oxidant properties and increase the pro-inflammatory actions of this acute-phase protein in a gene-dosage fashion. We hypothesized that the haptoglobin polymorphism might contribute to the increased oxidative stress and low-grade chronic inflammation frequently associated with polycystic ovary syndrome, obesity, and abnormalities of glucose tolerance.

Methodology/Principal Findings
Serum haptoglobin and the haptoglobin chain polymorphism were determined in 141 patients with polycystic ovary syndrome and 102 non-hyperandrogenic women. Of the whole group of 243 premenopausal women, 117 were obese and 51 showed abnormal glucose tolerance. Although serum haptoglobin concentrations were similar in PCOS patients and controls, the former presented with an increased frequency of Hp2 alleles (62% vs. 52%, P = 0.023). Circulating haptoglobin levels increased with obesity (P<0.001), yet no association was found between obesity and haptoglobin genotypes. No differences were observed in haptoglobin levels or genotype frequencies depending on glucose tolerance. Fifty percent of the variation in serum haptoglobin concentrations was explained by the variability in serum C-reactive protein concentrations, BMI, insulin sensitivity and haptoglobin genotypes.

Conclusions/Significance
Serum haptoglobin concentrations in premenopausal women are largely dependent on the haptoglobin polymorphism and on the presence of obesity, with insulin resistance and chronic inflammation possibly modulating this relationship. The association of polycystic ovary syndrome with Hp2 alleles suggests that the anti-oxidant and anti-inflammatory properties of haptoglobin may be reduced in these patients.

Introduction Top
Polycystic ovary syndrome (PCOS), a very prevalent syndrome characterized by the association of hyperandrogenism with chronic ovulatory dysfunction in premenopausal women [1], [2], is among the most common metabolic associations of both type 1 [3], [4] and type 2 diabetes mellitus [5]. Hyperinsulinism and insulin resistance are frequent findings in PCOS patients, and these traits have cause-consequence relationships with obesity [6], increased serum ferritin concentrations [7], low-grade chronic inflammation [8], oxidative stress [9], and increased cardiovascular risk [10]. We have shown that serum ferritin concentrations are increased in PCOS patients independently of chronic inflammation, suggesting increased body iron stores in these women very especially when obesity is also present [7]. Furthermore, our recent results strongly suggest that insulin resistance and hyperinsulinism, and not the reduced menstrual losses derived from oligomenorrhea or mutations in the hemochromatosis HFE gene, underlie the origin of these increased body iron stores [11], [12]. But iron metabolism is a complex system in which protection against iron losses and against the oxidative damage caused by free iron radicals are predominant mechanisms [13]. Haptoglobin, a circulating acute-phase protein that is synthesized in the liver under the influence of several cytokines such as TNF- and interleukin-6 [14], ameliorates the iron loss and renal injury that follows the liberation of hemoglobin from erythrocytes during processes of intravascular hemolysis. Haptoglobin accomplishes this function by forming soluble complexes with hemoglobin, thereby reducing the generation of oxygen reactive species and ameliorating oxidative stress [14]. Also, haptoglobin inhibits the synthesis of prostaglandins and the vasodilation mediated by nitric oxide, and exerts immunomodulatory actions by its binding to the CD163 receptor present in macrophages and monocytes [14].

11

The molecular structure of haptoglobin is determined in humans by the existence of a polymorphism affecting the gene on chromosome 16q22 that encodes the chain of this globulin. This polymorphism results in the existence of two different alleles, termed Hp1 and Hp2 [15] that may be explored by DNA genotyping or by plasma gel electrophoresis phenotyping [14]. The Hp2/Hp2 genotype = Hp 2-2 phenotype reduces plasma haptoglobin levels, decreases its hemoglobin-binding capacity and anti-oxidant properties and the secretion of anti-inflammatory cytokines such as IL-10, while increasing immunological reactivity [16]. Furthermore, haptoglobin phenotypes influence serum ferritin concentrations, and increased ferritin concentrations have been described not only in PCOS, but also in insulin resistant disorders such as obesity, type 2 diabetes mellitus and the metabolic syndrome [13]. Men expressing the Hp 2-2 phenotype present with increased serum ferritin concentrations when compared with men expressing the Hp 2-1 and Hp 1-1 phenotypes [17]. Haptoglobin is expressed in adipocytes [18] and, because serum haptoglobin concentrations correlate with increasing body weight, its circulating concentrations are increasingly being considered as a marker of adiposity. Finally, a recent proteomic analysis of serum has identified both the and haptoglobin chains as potential PCOS biomarkers [19], [20]. Therefore, we conducted the present study hypothesizing that haptoglobin and its polymorphism might mediate the associations of PCOS with hyperferritinemia and increased body iron stores, obesity and disorders of glucose tolerance.

Materials and Methods Top


Ethics statement Written informed consent was obtained from all the participants and the study was approved by the Clinical Research Ethics Committee of Hospital Universitario Ramn y Cajal. Participants We studied serum haptoglobin concentrations and haptoglobin genotypes in a large series of obese and non-obese PCOS patients and non-hyperandrogenic controls presenting with or without abnormalities of glucose tolerance. The sample included 141 PCOS patients and 102 non-hyperandrogenic Caucasian women prospectively recruited at the Department of Endocrinology of Hospital Universitario Ramn y Cajal, a tertiary care academic hospital affiliated with the University of Alcal. The diagnosis of PCOS relied on the presence of oligo-ovulation and clinical and/or biochemical hyperandrogenism, after excluding specific etiologies [1], [2], [21]. The particular methods used to establish these criteria have been reported earlier [6]. The non-hyperandrogenic control group was composed of normal weight healthy female volunteers and of consecutive overweight or obese women, who did not have any known metabolic comorbidity, reporting to the clinical practice of the authors solely for treatment of weight excess. Controls were selected in order to be similar in terms of BMI with the patients. None of the controls had signs or symptoms of hyperandrogenism, menstrual dysfunction, or history of infertility. According to their body mass index (BMI), women were classified as being obese (BMI30 kg/m2) or non-obese (BMI<30.0 kg/m2), because only obesity - and not overweight - has demonstrated a negative impact on mortality [22]. None of the participants had a personal history of hypertension, regular alcohol consumption, disorders of glucose tolerance, hyperuricemia, cardiovascular events, sleep apnea, or had received treatment with oral contraceptives, antiandrogens or insulin sensitizers for the previous 6 months. Study protocol and assays Clinical and anthropometrical variables, including the hirsutism score, BMI and waist-tohip ratio were determined as reported previously [6]. Blood samples were obtained between 12

days five and 10 of the menstrual cycle, or during amenorrhea after excluding pregnancy, in all the participants. After a three-day 300-g carbohydrate diet and 12-h overnight fasting, samples were obtained early in the morning for the measurement of haptoglobin, ferritin, C-reactive protein, total testosterone, sex hormone-binding globulin and calculated free testosterone [23], androstenedione, estradiol, dehydroepiandrosterone-sulfate, 17hydroxyprogesterone, prolactin, thyrotropin, luteinizing hormone and follicle-stimulating hormone. Complete serum biochemistry and lipid profiles were also obtained. Then, a 75-g oral glucose tolerance test was performed, and samples were obtained for measurement of serum insulin and plasma glucose at 0, 30, 60, 90 and 120 min. Samples were immediately centrifuged, and serum and plasma were separated and frozen at 30 C until assayed. The circulating concentrations of haptoglobin were assayed by a commercial immunonephelometry method (Dade Behring, Marburg, Alemania) calibrated against the international CRM 470 reference material [24] with intraassay and interassay CV of 2.6% and 6.2%, respectively. Haptoglobin genotypes were estimated using the PCR-based technique previously described by Koch et al. [25]. For plasma haptoglobin levels, we used the normal ranges established internationally for each particular haptoglobin genotype/phenotype [14]. Serum ferritin and C-reactive protein concentrations were measured by automated immunochemiluminescence (Immulite 2000 Ferritin and High Sensitivity CRP, Diagnostic Products Corporation, Los Angeles, CA) with lower limit of detection of 0.88 pmol/l and 0.1 mg/l respectively, and intra- and inter-assay coefficients of variation below 10%. The technical characteristics of the assays employed for plasma glucose concetrations, lipid profiles and serum hormone measurements have been described elsewhere [6]. The composite insulin sensitivity index was calculated from the circulating glucose and insulin concentrations during the oral glucose tolerance test according to Matsuda and DeFronzo [26], and women were classified as having normal or abnormal glucose tolerance (impaired fasting glucose, impaired glucose tolerance or diabetes) according to the American Diabetes Association clinical practice recommendations [27]. Statistical analysis The results are reported as meanstandard deviation unless otherwise stated. The association of the haptoglobin genomic variants with PCOS, obesity and abnormalities of glucose tolerance was explored using the 2 or Fisher's exact test as appropriate. The Kolmogorov-Smirnov statistic was applied to continuous variables. Logarithmic or square root transformations were applied as needed to ensure normal distribution of the variables, and parametric tests were applied thereafter. Mathematical transformation was not applied to serum haptoglobin levels, because this variable was normally distributed. General linear models adjusting for age served to evaluate the impact of PCOS on clinical and biochemical variables. General linear models introducing the haptoglobin genotype, obesity and glucose tolerance as independent variables, were used to study their influence on phenotypic variables. A multiple linear regression model with stepwise (probability of F to enter0.05, probability of F to remove0.10) introduction of age, BMI, waist-to-hip ratio, presence or absence of PCOS, serum C-reactive protein, ferritin and free testosterone concentrations, insulin sensitivity index and Hp genotypes (coded as Hp1/Hp1 = 0, Hp2/Hp1 = 1 and Hp2/Hp2 = 2) as independent variables, was used to evaluate the contribution of these variables to the variability in serum haptoglobin concentrations. Analyses were performed using SPSS10 for the Macintosh (SPSS Inc, Chicago, Illinois). P<0.05 was considered statistically significant.

Results Top
Comparison of clinical, biochemical and metabolic variables among patients and controls

13

As expected by design the patients with PCOS and the non-hyperandrogenic controls showed no statistically significant differences in the mean BMI or in the distribution according to the weight classification (Table 1), yet because PCOS patients were younger than controls (25.46.3 vs 31.77.7 yr respectively, P = 0.001), age was introduced as a covariate in the following analyses. Table 1. Distribution of PCOS patients and controls according to their weight classification. doi:10.1371/journal.pone.0005606.t001 PCOS patients presented with increased hirsutism scores, waist-to-hip ratio, fasting insulin and triglycerides, serum concentrations of total and free testosterone, androstenedione, 17hydroxyprogesterone, dehydroepiandrosterone-sulfate and luteinizing-hormone (Table 2). On the contrary, the insulin sensitivity index as well as serum sex hormone-binding globulin and estradiol levels were reduced in the patients. Furthermore, there were no differences in the frequency of smoking or of abnormal glucose tolerance. Table 2. Clinical and biochemical characteristics of PCOS patients and nonhyperandrogenic control women. doi:10.1371/journal.pone.0005606.t002 Haptoglobin and PCOS Serum haptoglobin concentrations were not different among PCOS patients and nonhyperandrogenic controls (Table 2). Furthermore, the frequency of women presenting with serum haptoglobin concentrations above the references values for their haptoglobin genotype was similar in both groups (controls 25.5% vs PCOS patients 26.1%, 2 = 0.010, P = 0.921). After confirming that the haptoglobin genotype was distributed according to HardyWeinberg equilibrium in both PCOS patients and controls (data not shown), we found that the Hp2 allele was more frequent in the former (62% vs 52%, 2 = 5.191, P = 0.023), explaining a borderline significant tendency towards a higher frequency of Hp2/Hp2 genotype, and to a reduced frequency of the Hp1/Hp1 genotype, in PCOS patients compared with the controls (Table 3). Table 3. Distribution of haptoglobin genotypes in the presence or absence of PCOS, obesity, or disordered glucose tolerance. doi:10.1371/journal.pone.0005606.t003 Finally, we confirmed that women with the Hp2/Hp2 genotype had mildly reduced circulating haptoglobin concentrations compared with women presenting the Hp2/Hp1 and Hp1/Hp1 genotypes (Figure 1) yet the 10% difference in the frequencies of haptoglobin genotypes among patients and controls, although statistically significant, was not large enough to cause an actual difference in their mean serum haptoglobin concentrations (Table 2). Figure 1. Serum haptoglobin, C-reactive protein and ferritin concentrations as a function of the haptoglobin genotype, obesity and glucose tolerance. Data are means and 95% confidence intervals. The numbers below the x-axis are the number of subjects in each subgroup. *P<0.05 or less between haptoglobin genotypes; P<0.05 or less between obese and non-obese women. P<0.05 or less between women with normal or abnormal glucose tolerance. doi:10.1371/journal.pone.0005606.g001 Haptoglobin and obesity

14

Serum haptoglobin levels were increased in obese women irrespective of PCOS status (Figure 1). Accordingly, obese women presented with a 5-fold increase in the frequency of women presenting with elevated serum haptoglobin concentrations for their haptoglobin genotype (non-obese 8.7%, obese women 44.1%, 2 = 39.73, P<0.001). Of note, the increase in serum haptoglobin concentrations with obesity was unrelated to the haptoglobin polymorphism, because its genotypes were equally distributed among nonobese and obese women (Table 3). Haptoglobin and abnormalities of glucose tolerance Serum haptoglobin concentrations were not different when women were classified according to their normal or abnormal glucose tolerance (Figure 1), and the frequency of subjects presenting with increased serum haptoglobin levels for their genotype were not different depending on their glucose tolerance (normal glucose tolerance 24.0%, abnormal glucose tolerance 33.3%, 2 = 1.844 P = 0.174). Finally, haptoglobin genotypes were similarly distributed among the groups with normal or abnormal glucose tolerance (Table 3). Impact of the haptoglobin polymorphism, PCOS, obesity and abnormalities of glucose tolerance on serum ferritin and C-reactive protein concentrations The haptoglobin genotype did not influence serum ferritin and C-reactive protein concentrations (Figure 1). Serum ferritin concentrations were higher in PCOS patients when compared with non-hyperandrogenic controls (Table 2), were also higher in obese women when compared with non-obese women (Figure 1), and increased in the women presenting with abnormal glucose tolerance (Figure 1). On the contrary, serum C-reactive protein concentrations increased in obese women compared to their non-obese counterparts and tended to increase (P = 0.079) in the women presenting with abnormal glucose tolerance (Figure 1), but were similar in PCOS and non-hyperandrogenic women (Table 2). Determinants of serum haptoglobin levels in premenopausal women A multiple regression model (R2 = 0.498, F = 58.6, P<0.0001) demonstrated that 50% of the variation in serum haptoglobin concentrations was explained by the variation in serum C-reactive protein concentrations ( = 0.405, P<0.001), BMI ( = 0.251, P<0.001), Hp genotypes ( = 0.207, P<0.001) and insulin sensitivity index ( = 0.167, P = 0.001). Age, PCOS, waist-to-hip ratio, ferritin and free testosterone concentrations were excluded by the stepwise regression model.

Discussion Top
To our best knowledge, we here present the first data regarding serum haptoglobin concentrations and its genotypes in patients with PCOS. The most important finding of our present study is that, compared with non-hyperandrogenic women, PCOS patients presented with an increased frequency of the Hp2 allele (and a borderline significant tendency towards an increased frequency of the Hp2/Hp2 genotype). The importance of the association of PCOS with Hp2 alleles is that this polymorphism influences the functional properties of haptoglobin in a way that might contribute to the known clustering of cardiovascular risk factors in PCOS patients [10] even in the absence of actual differences in serum haptoglobin levels. On the one hand, the haptoglobin polymorphism causes a gene-dosage effect resulting in a graded decrease in the haptoglobin ability to bind hemoglobin, in the antioxidant capacity and in prostaglandin inhibition, with the Hp2/Hp2 genotype expressing the weakest protein activities [14]. On the other hand, the haptoglobin polymorphism also results in a graded increase in the angiogenic and inflammatory effects of haptoglobin, with the Hp2/Hp2 genotype expressing the strongest properties [14]. Therefore, the increased frequency of Hp2 alleles in our PCOS patients might also play a role in the previously reported lowgrade chronic inflammation [28] and increased oxidative stress [9] of these women, even when serum haptoglobin concentrations were not altered in these PCOS women. 15

Supporting a role for the haptoglobin polymorphism in the development of atherosclerotic disease and cardiovascular events, the Hp 2-2 phenotype increases the risk for micro- and macrovascular complications in diabetic patients [29] possibly because this phenotype facilitates the generation of reactive oxygen species, and oxidative stress is involved in myocardial damage and atherosclerosis. Furthermore, in the ApoE knock-out murine model of atherosclerosis, Hp2/Hp2 mice obtained by targeted insertion of a murine type 2 Hp allele into the Hp locus by homologous recombination show increased iron deposition, lipid peroxidation and macrophage accumulation within atherosclerotic plaque, increasing the risk of rupture and thrombosis [30]. Interestingly, PCOS patients show early markers of atherosclerosis such as increased carotid intima-media thickness and coronary artery calcification from young ages [31], [32]. Of note, the difference in the frequency of haptoglobin genotypes was actually caused by a relatively low frequency of Hp2 alleles in the controls when compared to that found in the general population of other countries whereas, on the contrary, the frequency found in the patients was similar to that reported in other Caucasians populations. This surprising finding has two possible explanations: first, the frequencies of haptoglobin alleles in the Spanish population has not been established, and may be lower compared with those found in other European controls. Second, our controls were selected as to have no known metabolic or hyperandrogenic disorder, and considering the high prevalences of these conditions in the general population, this selection bias may explain a reduced frequency of Hp2 alleles in them. Nevertheless, the distribution of Hp alleles was in HardyWeimberg equilibrium in both PCOS patients and in controls, ruling out significant population stratification in our series. The chain haptoglobin polymorphism influences circulating haptoglobin concentrations to a lesser degree compared with its impact on the functionality of the protein [14]. This may explain why PCOS patients, despite carrying Hp2 alleles more frequently than controls, did not have reduced serum haptoglobin levels. Also, the lowering effect on serum haptoglobin levels of carrying Hp2 alleles may have been obscured in the PCOS patients presented here because these women were insulin resistant compared with the controls, and insulin resistance actually increases serum haptoglobin levels [33] as our data also demonstrate. Our present results do not suggest that the haptoglobin polymorphism influences the increased serum ferritin concentrations of PCOS patients, in sharp contrast with the findings in male hemochromatosis patients, in whom the Hp 2-2 phenotype increases iron overload and the expression of the hemochromatosis phenotype [17]. Although the ferritin lowering effect of regular menstrual iron losses [34] might contribute to explain this discrepancy, most of the PCOS patients had chronic oligomenorrhea, and our previous studies on PCOS patients suggest that insulin resistance, and not genetic factors, is one of the major determinants of increased serum ferritin concentrations and iron stores in these women [12]. Similarly, the haptoglobin polymorphism did not influence serum C-reactive protein concentrations in our population, yet the possible effect of the polymorphism on local inflammatory mechanisms may have been obscured by the strongest influence of obesity on the serum levels of this inflammatory marker [35]. The impact of weight excess on serum haptoglobin concentrations and the other circulating inflammatory markers studied here merits further comment. Our present results indicate that serum haptoglobin concentration, and the number of women presenting with serum levels above the upper limit of the normal range, increase with obesity. These results confirm the previously reported association of circulating haptoglobin concentrations with BMI [18], [36], which might be dependent on the expression of haptoglobin in adipose tissue [18]. Furthermore, obesity also increases serum C-reactive protein and ferritin concentrations pointing to the existence of chronic inflammation and increased body iron stores in obesity and other disorders characterized by the presence of insulin resistance. Even when serum haptoglobin levels increase with insulin resistance and obesity [18], [33], [36], neither serum haptoglobin levels nor the haptoglobin genotype were related to the development of disorders of glucose tolerance in our series of young premenopausal women. Therefore, haptoglobin does not appear to be involved in the failure of

16

compensatory endogenous hyperinsulinism to maintain glycemia within the normal range, which is the hallmark of abnormal glucose tolerance in insulin resistant subjects. On the contrary, increased body iron stores may have contributed to this failure [13], because serum ferritin levels were clearly increased in the women who had abnormal glucose tolerance in our series. Of note, chronic inflammation might play a similar role, yet the tendency towards an increase in serum C-reactive protein levels in women presenting with abnormal glucose tolerance did not reach statistical significance. Our study was limited, however, by a relatively low statistical power for detection of differences between the small subgroups resulting from categorization by presence or absence of metabolic traits. This limitation was especially evident when considering glucose tolerance, and therefore a possible influence of haptoglobin levels or genotypes on the development of disorders of glucose tolerance cannot be definitely excluded by our present results. Finally, the results of the multiple regression model indicates that serum haptoglobin concentrations are determined in premenopausal women by several factors including BMI, serum C-reactive protein concentrations and the insulin sensitivity index, in addition to haptoglobin genotypes. Overall, the result of the regression analysis suggests that adipose tissue, and also the chronic inflammatory milieu (possibly reflecting the action of inflammatory cytokines such as TNF- and IL-6 on the liver) and the insulin resistant state associated with obesity, up-regulate haptoglobin secretion contributing to its circulating concentrations. On the contrary, the haptoglobin Hp2/Hp2 genotype resulted in a mild decrease in serum haptoglobin concentrations in our premenopausal women, as has been shown in other populations [14]. In summary, the association of PCOS with Hp2 alleles of haptoglobin suggests that the functionality of this protein might be altered in these women, even when its circulating concentrations are not altered because their insulin resistance counteracts the decrease in circulating haptoglobin usually associated with Hp2 alleles. Also, our present results indicate that serum haptoglobin levels of young premenopausal women are mostly dependent on the haptoglobin polymorphism and on obesity, with insulin resistance and chronic inflammation possibly contributing to this association.

17