| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Genomic Medicine Institute (F.W., C.E.), Cleveland Clinic Lerner Research Institute, Cleveland, Ohio 44195; Department of Genetics (C.E.), Case Western Reserve University School of Medicine, Cleveland, Ohio 44106; Human Cancer Genetics Program (M.S.S., F.W., C.E.), Comprehensive Cancer Center (C.E.), Department of Molecular Virology, Immunology, and Medical Genetics (M.S.S., F.W., C.E.), Division of Epidemiology and Biometrics (L.S.), and Department of Physiology and Cell Biology (S.M.J.), The Ohio State University, Columbus, Ohio 43210; Faculdade de Medicina (M.S.S.), Universidade Federal de Minas Gerais, 30130 Belo Horizonte, Brazil; Bolsista do CNPq (M.S.S.), 30130 Belo Horizonte, Brazil; Department of General Surgery and Transplantation (F.W., A.F., C.E.B.), University of Essen, 45122 Essen, Germany; Department of Surgery (J.Z.), Karolinska University Hospital, SE-171 76 Stockholm, Sweden; and Cancer Research U.K. Human Cancer Genetics Research Group (C.E.), University of Cambridge, CB2 1XZ Cambridge, United Kingdom
Address all correspondence and requests for reprints to: Charis Eng, M.D., Ph.D., Genomic Medicine Institute, Cleveland Clinic Lerner Research Institute, 9500 Euclid Avenue, NE-30 (Room NE3-307), Cleveland, Ohio 44195. E-mail: engc{at}ccf.org.
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and MethodsResultsDiscussionReferences |
|---|
Objective: The goal of this study was to specifically address this hypothesis that in thyroid neoplasias loss of imprinted loci becomes enriched during oncogenesis.
Design: In total, thyroid tissue was obtained from 72 patients with thyroid neoplasias comprising 34 follicular thyroid carcinomas (FTCs) and 38 follicular adenomas. We performed PCR-based LOH analysis of DNA from paired normal-tumor samples using 18 markers mapped to imprinted regions (IR) and 13 markers in nonimprinted regions (NIR).
Results: Overall LOH frequencies for the IR markers were 26% for the adenomas and 38% for the carcinomas. In the NIR, the overall LOH frequency was 23 and 26% for adenomas and FTCs, respectively. The difference in LOH frequencies between IR and NIR was statistically significant only for the carcinomas (P = 0.001), although there was a similar trend for the atypical adenomas (ATY, P = 0.06).
Conclusions: Our observations suggest that IRs are more prone to genomic instability in FTCs. The fact that the ATY trended toward differential IR/NIR LOH, similar to FTC, may suggest that loss of IR might be instrumental in the adenoma-carcinoma sequence in thyroid carcinogenesis and that ATY could be an important intermediate in this pathway.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
Despite research that elucidated the activation of oncogenes (RAS, BRAF, RET, NTRK1, MET) and silencing of tumor suppressor genes (TSG) (p53, RASSF1A, PTEN, PPARg, CDK inhibitors) (2, 3), the molecular etiology and pathogenesis of epithelial thyroid carcinoma, in particular follicular thyroid carcinoma (FTC), remain to be fully clarified. For instance, it is not yet clear whether, in a corollary to colon cancer development, the adenoma-carcinoma sequence might exist. Whereas clinical data suggest that the progression of follicular adenoma (FA) to FTC is an extremely rare event, recent data suggest that a subset of FA, the atypical adenomas, characterized by unusual cellular density, mitoses, and a less regular cytological pattern might indeed predispose to malignant transformation (4).
In our previous work using microarray-based expression analyses to delineate genes involved in FTC genesis, we identified the maternally imprinted TSG ARH I to be down-regulated at a very early stage in the malignant transformation of thyroid follicular epithelial cells (5). We showed that the mechanism of ARHI inactivation is due to either allelic loss or hypermethylation of the promoter of the functional, nonimprinted allele (5).
In a subset of the mammalian genes, including some that regulate growth of cells and tissues, only the maternal or paternal copy is functional because of silencing of one of the two alleles. This phenomenon of monoallelic, parent-specific expression of genes is termed genomic imprinting (6, 7). As a consequence of being functionally haploid, loss of heterozygosity (LOH) at an imprinted locus may result in the deletion of the only working copy of an imprinted TSG. This class of genes should therefore be subjected to only one hit inactivation, either by simple mitotic recombination and LOH or by other pathways such as altered DNA methylation (8). Thus, we hypothesized that random genomic events affecting imprinted regions provide the cell with a growth advantage that would promote clonal expansion and neoplasia. So enrichment of LOH at crucial growth-promoting imprinted regions should be identified because they are good candidate tumor pathogenic pathways. We therefore sought to conduct a comparison between the frequency of allelic imbalance on imprinted (IR) and nonimprinted regions (NIR) to address this hypothesis in thyroid neoplasias. If our hypothesis is correct, then we expect the frequency of LOH/allelic imbalance at loci harboring genomically imprinted gene(s) to be higher than that in regions not known to have imprinted genes.
|
Patients and Methods |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
Formalin-fixed, paraffin-embedded thyroid archival tissues, tumor, and matching normal tissue were obtained from 72 patients of western European descent with thyroid neoplasias comprising 34 FTCs and 38 FAs. The FAs comprised nine Hurthle adenomas (HAs), nine atypical adenomas (ATYs), and 20 typical follicular adenomas (TFAs). All samples were acquired as anonymized materials without linked identifiers, under approval from the Human Subjects Protection Committees of the respective institutions. Enrichment of tumor epithelium was achieved by microdissection. Genomic DNA was extracted using QIAamp DNA minikit (QIAGEN, Valencia, CA), according to the manufacturer’s instructions.
Genotyping for LOH analysis
A PCR-based LOH analysis was performed using 18 microsatellite markers mapped to known genomic IRs (Table 1
) and 13 markers, which are located in NIRs (Table 2). PCR was carried out using the multiplex PCR master mix (QIAGEN) in a final volume of 25 µl. After 15 min at 95 C, the reactions were subjected to 40 cycles at 94 C for 30 sec, 55 C for 45 sec, and 72 C for 45 sec followed by 10 min at 72 C.
|
|
After PCR, the amplicons were genotyped with the ABI 377xl or 3700 semiautomated sequencer (Applied Biosystems, Perkin-Elmer Corp., Norwalk, CT). The results were manually evaluated using the GeneScan collection and analysis software (GeneScan; Applied Biosystems, Norwalk, CT). LOH frequency is expressed as a percentage of samples with LOH divided by the total number of informative cases.
Statistical analyses
For global comparisons of LOH frequencies among all markers residing in multiple IRs and those in NIRs, a permutation-based test was performed for each comparison among multiple IRs and NIRs. This test requires fewer assumptions and therefore is more robust than the Fisher’s exact test. The permutation test also takes into consideration that markers mapping to the same region tend to be correlated (i.e. may influence one another). We obtained a reference distribution of the difference in LOH frequency under the null hypothesis by randomly permuting the labels of IRs and NIRs a large number (10,000) of times. A P value was then obtained based on this reference distribution.
For comparisons of LOH frequencies within a single IR with those in markers in physical proximity or when comparing frequencies among small groups of samples, the two-tailed Fisher’s exact test of association was used.
Correlation with global gene-expression data
In total, 17 documented imprinted genes mapped to 14 genomic regions were analyzed in this study (Table 3). We evaluated the expression of these 17 using the Affymetrix (Santa Clara, CA) HG-U133A Genechip platform in a set of 12 FTCs and 12 FAs, which represent an independent series from the current series. A detailed description of the gene expression analysis according to the MIAME criteria is published (10). In short, the cell intensity files were interrogated using the Affymetrix Microarray Suite 5.0 software. The DNA-chip analyzer software (dChip) developed by Li and Wong (www.dchip.org) was used to normalize all arrays to a common array having a median overall brightness by using an invariant set of probes (11). A perfect match/mismatch difference model of the dChip software was then used to compute the model-based expression index (12). We used the modeled gene expression data of 24 total probe sets, representing the 17 imprinted genes, to generate a heat map. In addition, we filtered the 24 probe sets by setting the thresholds to 1.5-fold expressional changes and a P < 0.05 for the difference in expression between FTCs and FAs. Furthermore, we used the multidimensional scaling function of the BRB Array Tools software package (developed by R. Simon and A. Peng) to visualize the naturally arising (unforced) sample classes based on the expression profile of the imprinted genes (see Results).
|
|
Results |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
The markers were chosen after identification of IRs provided by the Catalogue of Imprinted Genes and Parent-of-Origin Effects Database (9). None of these markers has been previously reported to have particularly high degrees of genomic instability in thyroid neoplasias. Thus, if a region happened to be relatively long, it would be represented by more markers. Apart from 11q13, IR markers on chromosome 11 were more represented because there is the greatest number of IRs on that chromosome. We avoided certain chromosomal regions like 2p, 2q, 3p, 10q, and 11q13, which have already been reported to be more frequently subject to allelic loss (13), to avoid potential bias. For our control group of NIRs, we chose 13 markers, some of them in proximity to the IRs. These markers are not known to be imprinted or to present hot spots of genomic instability based on published data.
Using permutation tests for the assessment of LOH frequencies of markers between IRs and NIRs in FTCs, we obtained a highly significant increased LOH frequency in the IRs (38%) than NIRs (26%) (P = 0.0013). In contrast, for FAs, no difference in the LOH frequencies were found when comparing IRs (27%) with NIRs (23%) (P = 0.309) (Tables 1
and 2).
When we evaluated the overall frequency of LOH in adenomas and carcinomas, we found no difference between the two groups in the NIRs (26% FTC vs. 23% FA). However, the average LOH frequency for IR markers demonstrated a significantly higher frequency of LOH for FTCs (38% FTC vs. 27% FA, P = 0.0007). In particular, there were three IR-specific markers that showed statistically significant differences in LOH frequencies between FTCs and FAs: D1S2829 (P = 0.019), D7S2519 (P = 0.02), and D14S1426 (P = 0.038).
Because we evaluated a heterogeneous group of benign follicular neoplasias, we performed subanalyses based on histological type. No differences were identified when comparing LOH frequencies between IRs and NIRs for HAs (P = 0.74) and TFAs (P = 0.9). We observed a trend toward a higher LOH frequency in IRs (35%) than NIRs (27%) for ATYs (P = 0.06, Table 4). However, when comparing each subtype (HA, TFA, and ATY) against each other, we found the ATY samples had a significantly higher occurrence of LOH in the IRs than that of TFAs (35% ATY vs. 23% TFA, P = 0.025) (Table 4 and Fig. 1A). In fact, the overall LOH frequency (LOH/LOH+ROH) had a similar difference between IRs and NIRs for ATYs and FTCs (Fig. 1A). In addition, when we analyzed for LOH differences in the diverse IRs separately, there is a significantly higher LOH frequency in the ATYs (83%), compared with the TFAs (23%) for the 11p15.5 (D11S1984, P = 0.04) and 6q24-q26 regions (D6S409, D6S437, D6S1277, P = 0.04) with LOH frequencies reaching 50% for ATYs and 29% for TFAs. We also observed a trend toward higher LOH frequencies for D1S2829 in ATYs, compared with TFAs (P = 0.08, Table 5).
|
|
|
, B and C). For example, the 11q25 IR, defined by D11S1304 and D11S4463 and 11p15 (D11S1304), showed LOH frequencies greater than 40% for both benign and malignant follicular neoplasias (Fig. 1B). Other regions that also exhibited 40% or greater LOH frequency were 14q32.33 (D14S1426) among FTCs; 6q24.2–26 (D6S409 and D6S437) in the ATYs; and 1p31, 11p15.5, 15q11.2, and 20q13.12 in both the FTC and ATY groups (Fig. 1C).
To determine whether IRs with significant LOH were associated with underexpression of particular genes in those areas, we reinterrogated our previous microarray expression data (9) done with different samples. The heat map shows the modeled expression values of each gene and indicate that FTCs are significantly (P = 0.0047) enriched in the cluster corresponding to underexpressed genes (heat map not shown). Furthermore, multidimensional scaling based on gene expression differences of all imprinted gene mapping to the evaluated loci reveals at least two naturally arising classes (Fig. 2). After we filtered the 24 probe sets by setting the thresholds to 1.5-fold expressional changes and a P < 0.05 for the difference in expression between FTCs and FAs, we identified five genes, which were significantly differentially expressed between FTCs and FAs (Table 3 and Fig. 3). To statistically validate these findings, we performed a random permutation analysis, in which we randomly permuted the labels of FTCs and FAs many times, repeated the gene selection procedure using the same criteria, and recorded the number of genes identified. It demonstrated that these genes were uncovered due to biological relevance and not by random coincidence (i.e. chance).
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
Thyroid cancer is believed to result in part from the accumulation of multiple genetic alterations leading to oncogene overexpression and tumor suppressor loss. Several LOH studies show an inconsistent frequency of LOH, depending on the population, chromosomal regions, and method used (visual vs. automated). One consensus from all these LOH studies is the higher frequency of LOH among FTCs, compared with papillary thyroid carcinomas (13), and for most markers, a higher LOH frequency in FTCs, compared with FAs, with rare exception (16). We have shown that the latter holds with the markers used in this study (Fig. 1A).
In agreement with our hypothesis, we demonstrated that malignant follicular neoplasias are enriched in genomic instability at imprinted loci, compared with their benign counterparts, whereas we showed no differences in LOH frequencies between malignant and benign follicular neoplasias in NIRs. Our data also suggest that the accumulation of these events provides the follicular epithelial cell with a clonal advantage.
Given that the average LOH frequency in IRs was 38%, we operationally defined hot spot LOH regions as those exceeding 40% LOH frequency to focus on regions with presumed greater genomic instability. Several imprinted regions could be considered hot spots. Among the imprinted loci, the highest LOH frequencies, i.e. the hot spots, found in FTCs were on chromosome arms 1p, 11p and q, 14q, 15q, and 20q (Table 1). One of these regions (11q25), which contains a CpG island, represents a hot spot for all benign histological types as well (Table 5), meaning that this is an important region for initial neoplastic transformation but not necessarily malignant transformation. When these FTC-related hot spots were compared with the whole FA group, two hot spots, 1p31 and 14q32, showed a statistically significantly higher LOH frequency in FTCs over FAs (P = 0.019 and 0.038, respectively). The 1p31 region data here validate our previous report on the maternally imprinted TSG ARHI in an extended sample set (5). Furthermore, we have shown for the first time that whereas TFAs showed a frequency of 14% LOH at this 1p31/ARH1 locus, the ATY LOH frequency at this locus was similar to that of FTCs, 60 and 57%, respectively. The imprinted region 14q32 contains the paternally expressed gene iodothyronine deiodinase type III (DIO3) in proximity to the linked yet reciprocally imprinted genes, DLK1 and GTL2. The loss of GTL2 (MEG3) expression has been implicated in tumorigenesis in human pituitary adenomas and may also contribute to other human cancers like neuroblastomas (17, 18). Reinterrogation of our previously reported microarray data (11) reveals that GTL2 and DLK1 are underexpressed in malignant thyroid tumors, compared with the adenomas without statistical significance (Table 3).
One region, 11p15 IR, caught our attention for harboring the highest frequency of LOH in ATYs (83%) with 42% LOH in FTCs (42%) but 23% LOH frequency in TFAs. This region is also frequently lost in various neoplasias including breast cancer (19), and it is enriched with several imprinted genes, including CDKN1C, KCNQ1, IGF2, IGF2-AS, H19, and INS. The analysis of our microarray experiments reveals a significant underexpression of CDKN1C (–2.96-fold, P = 0.0084) in FTCs, compared with FAs (Table 3). CDKN1C (or p57, Kip2) is a negative regulator of cell proliferation. Mutations of this gene are implicated in a broad variety of sporadic cancers and the overgrowth syndrome, Beckwith-Wiedemann, suggesting that it is an important tumor suppressor candidate (20). Another gene underexpressed in this region is IGF2. The imprinted genes, DLK1 and GTL2, have similarities to IGF2 and H19, respectively. Both GTL2 and H19 are maternally expressed RNAs with no protein product displaying paternal allele promoter region methylation, and DLK1 and IGF2 are both paternally expressed. The IGF2-linked H19 gene is reciprocally imprinted with the maternal allele being active and the paternal allele inactive. It has been suggested that H19 has a tumor suppressor activity (21). Insulin and IGFs are major determinants of proliferation and apoptosis, thereby playing a significant role in carcinogenesis. Epidemiological evidence associates high levels of insulin and IGFs with an increased risk of cancer (22). In our microarray data, the expression level of INS was increased in the FTCs but not significantly.
Another FTC /ATY hot spot IR of interest when comparing our LOH data with the previous microarray expression data are the IR 15q11-q13. This is a well-studied IR with several imprinted genes including the Prader-Willi (PWS) and Angelman syndrome (AS) genes, IPW, SNURF, SNRPN, UBE3A, ATP10A, NDN, and ZNF127. PWS and AS are caused by the loss of function of imprinted genes in proximal 15q11-q13. In approximately 2–4% of patients, this loss of function is the result of an imprinting defect. In some cases, the imprinting defect is the result of a parental imprint-switch failure caused by a microdeletion of the imprinting center (23). Small nuclear ribonucleoprotein N (SNRPN) gene encodes SmN spliceosomal protein. This genetic locus appears to have a key role in the cis regulation of imprinting throughout chromosome 15q11-q13 because microdeletions that remove the 5' end of the SNRPN gene are found in a subset of PWS patients in which the paternally inherited 15q11-q13 is otherwise intact but inappropriately bears a maternal epigenotype (23). SNURF lies upstream of the SNRPN and forms with the first an unusual bicistronic gene structure (24). The SNURF-SNRPN locus adds yet another layer of complexity to this locus that is potentially central to PWS and the evolution of imprinting in chromosome 15q11-q13. The bicistronic Snurf-Snrpn are translated in normal but not PWS human and mouse tissues and cell lines (24). There appears to be an increased risk of myeloid leukemias, but not other cancers, among persons with PWS (25). Ubiquitin protein ligase E3A [UBE3A (human papilloma virus E6-associated protein)] encodes an E3 ubiquitin-protein ligase, part of the ubiquitin protein degradation system. Maternally inherited deletion of this gene causes AS. The protein also interacts with the E6 protein of human papillomavirus types 16 and 18, resulting in ubiquitination and proteolysis of tumor protein p53 (26). NDN is an intronless gene and is located in the PWS deletion region. It is an imprinted gene and, like ZNF127, is expressed exclusively from the paternal allele. Murine models suggest that ZNF127 may suppress growth in postmitotic neurons (27).
Reinspection of our microarray expression data revealed that four of these genes show statistically significant underexpression in FTCs, compared with FAs: SNURF, SNRPN, NDN, and IPW (Table 3). In the comparison between FAs and normal thyroid for these genes, we also found that three of these genes, UBE3A, ATP10A, and IPW, were underexpressed in FAs, compared with normal tissue (P < 0.05). This finding might explain the reason we also found a high LOH ratio for the typical adenomas in this region for one of the markers that also substantiate results of previous studies that used comparative genomic hybridization showing frequent allelic loss in FAs in the 15q13 region (28). Thus, our current observations together with our microarray expression studies suggest that the 15q11-q13 and its imprinted genes, IPW, UBE3A, and ATP10A, play a role in neoplastic initiation from normal to adenoma (Fig. 4). The imprinted genes ARH1, IPW, CDKN1C, SNURF, SNRPN, and NDN differentiate TFAs from ATYs and FTCs, suggesting their role in the very earliest step of malignant transformation (Fig. 4).
|
The sequence of genetic and epigenetic events contribute to the outcome of gene expression and finally to tumorigenesis and should hold true in thyroid neoplasia as well. In our study, we show that the overall effect of genomic imprinting on atypical adenomas and follicular carcinomas is potentially great. Some regions are commonly lost for all types of neoplasias, suggesting that these regions play a role very early, even at the benign neoplasia stage. This might lead to further genomic instability, if another genetic/epigenetic event is superimposed. Taken together, our LOH hot spot data, in conjunction with reinterrogation of our previous microarray data, show that not only are IR LOH prominent in follicular thyroid neoplasia-genesis but also reveal the genes residing within IR that may play a role in the initiation of FAs as well as in the earliest malignant transformation, i.e. to ATYs. It is tempting to speculate that it is these IR-related genes that drive the earliest steps in follicular thyroid neoplasia and transformation and that somatic alterations in the BRAF-RAS pathway subsequently drive the ATY to FTC step.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
First Published Online October 25, 2005
1 M.S.S. and F.W. contributed equally to this work. 
Abbreviations: AS, Angelman syndrome; ATY, atypical adenoma; dChip, DNA-chip analyzer software; FA, follicular adenoma; FTC, follicular thyroid carcinoma; HA, Hurthle adenoma; IR, imprinted region; LOH, loss of heterozygosity; NIR, nonimprinted region; PWS, Prader-Willi syndrome; SNRPN, small nuclear ribonucleoprotein N; TFA, typical follicular adenoma; TSG, tumor suppressor gene; UBE3A, ubiquitin protein ligase E3A.
Received August 19, 2005.
Accepted October 17, 2005.
|
References |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  G. Assie, T. LaFramboise, P. Platzer, and C. Eng Frequency of Germline Genomic Homozygosity Associated With Cancer Cases JAMA, March 26, 2008; 299(12): 1437 - 1445. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. M Cavaco, P. F Batista, C. Martins, A. Banito, F. do Rosario, E. Limbert, L. G Sobrinho, and V. Leite Familial non-medullary thyroid carcinoma (FNMTC): analysis of fPTC/PRN, NMTC1, MNG1 and TCO susceptibility loci and identification of somatic BRAF and RAS mutations Endocr. Relat. Cancer, March 1, 2008; 15(1): 207 - 215. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. T. Netea-Maier, S. W. Hunsucker, B. M. Hoevenaars, S. M. Helmke, P. J. Slootweg, A. R. Hermus, B. R. Haugen, and M. W. Duncan Discovery and Validation of Protein Abundance Differences between Follicular Thyroid Neoplasms Cancer Res., March 1, 2008; 68(5): 1572 - 1580. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Fujarewicz, M. Jarzab, M. Eszlinger, K. Krohn, R. Paschke, M. Oczko-Wojciechowska, M. Wiench, A. Kukulska, B. Jarzab, and A. Swierniak A multi-gene approach to differentiate papillary thyroid carcinoma from benign lesions: gene selection using support vector machines with bootstrapping Endocr. Relat. Cancer, September 1, 2007; 14(3): 809 - 826. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Weber, R. E. Teresi, C. E. Broelsch, A. Frilling, and C. Eng A Limited Set of Human MicroRNA Is Deregulated in Follicular Thyroid Carcinoma J. Clin. Endocrinol. Metab., September 1, 2006; 91(9): 3584 - 3591. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J Di Cristofaro, M Silvy, A Lanteaume, M Marcy, P Carayon, and C De Micco Expression of tpo mRNA in thyroid tumors: quantitiative PCR analysis and correlation with alterations of ret, Braf , ras and pax8 genes. Endocr. Relat. Cancer, June 1, 2006; 13(2): 485 - 495. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
