Integrated methodologies to study human transcriptional regulation from functional genomics data

Abstract : The human body is composed of a huge number of cells acting together in a concerted manner. The current understanding is that proteins perform most of the necessary activities in keeping a cell alive. The DNA, on the other hand, stores the information on how to produce the different proteins in the genome. Regulating gene transcription is the first important step that can thus affect the life of a cell, modify its functions and its responses to the environment. Regulation is a complex operation that involves specialized proteins, the transcription factors. Transcription factors (TFs) can bind to DNA and activate the processes leading to the expression of genes into new proteins. Errors in this process may lead to diseases. In particular, some transcription factors have been associated with a lethal pathological state, commonly known as cancer, associated with uncontrolled cellular proliferation, invasiveness of healthy tissues and abnormal responses to stimuli. Understanding cancer-related regulatory programs is a difficult task, often involving several TFs interacting together and influencing each other's activity. This Thesis presents new computational methodologies to study gene regulation. In addition we present applications of our methods to the understanding of cancer-related regulatory programs. The understanding of transcriptional regulation is a major challenge. We address this difficult question combining computational approaches with large collections of heterogeneous experimental data. In detail, we design signal processing tools to recover transcription factors binding sites on the DNA from genome-wide surveys like chromatin immunoprecipitation assays on tiling arrays (ChIP-chip). We then use the localization about the binding of TFs to explain expression levels of regulated genes. In this way we identify a regulatory synergy between two TFs, the oncogene C-MYC and SP1. C-MYC and SP1 bind preferentially at promoters and when SP1 binds next to C-NIYC on the DNA, the nearby gene is strongly expressed. The association between the two TFs at promoters is reflected by the binding sites conservation across mammals, by the permissive underlying chromatin states 'it represents an important control mechanism involved in cellular proliferation, thereby involved in cancer. Secondly, we identify the characteristics of TF estrogen receptor alpha (hERa) target genes and we study the influence of hERa in regulating transcription. hERa, upon hormone estrogen signaling, binds to DNA to regulate transcription of its targets in concert with its co-factors. To overcome the scarce experimental data about the binding sites of other TFs that may interact with hERa, we conduct in silico analysis of the sequences underlying the ChIP sites using the collection of position weight matrices (PWMs) of hERa partners, TFs FOXA1 and SP1. We combine ChIP-chip and ChIP-paired-end-diTags (ChIP-pet) data about hERa binding on DNA with the sequence information to explain gene expression levels in a large collection of cancer tissue samples and also on studies about the response of cells to estrogen. We confirm that hERa binding sites are distributed anywhere on the genome. However, we distinguish between binding sites near promoters and binding sites along the transcripts. The first group shows weak binding of hERa and high occurrence of SP1 motifs, in particular near estrogen responsive genes. The second group shows strong binding of hERa and significant correlation between the number of binding sites along a gene and the strength of gene induction in presence of estrogen. Some binding sites of the second group also show presence of FOXA1, but the role of this TF still needs to be investigated. Different mechanisms have been proposed to explain hERa-mediated induction of gene expression. Our work supports the model of hERa activating gene expression from distal binding sites by interacting with promoter bound TFs, like SP1. hERa has been associated with survival rates of breast cancer patients, though explanatory models are still incomplete: this result is important to better understand how hERa can control gene expression. Thirdly, we address the difficult question of regulatory network inference. We tackle this problem analyzing time-series of biological measurements such as quantification of mRNA levels or protein concentrations. Our approach uses the well-established penalized linear regression models where we impose sparseness on the connectivity of the regulatory network. We extend this method enforcing the coherence of the regulatory dependencies: a TF must coherently behave as an activator, or a repressor on all its targets. This requirement is implemented as constraints on the signs of the regressed coefficients in the penalized linear regression model. Our approach is better at reconstructing meaningful biological networks than previous methods based on penalized regression. The method is tested on the DREAM2 challenge of reconstructing a five-genes/TFs regulatory network obtaining the best performance in the "undirected signed excitatory" category. Thus, these bioinformatics methods, which are reliable, interpretable and fast enough to cover large biological dataset, have enabled us to better understand gene regulation in humans.

Short-term oleoyl-estrone treatment affects capacity to manage lipids in rat adipose tissue

Background: Short-term OE (oleoyl-estrone) treatment causes significant decreases in rat weight mainly due to adipose tissue loss. The aim of this work was to determine if OE treatment affects the expression of genes that regulate lipid metabolism in white adipose tissue. Results: Gene expression in adipose tissue from female treated rats (48 hours) was analysed by hybridization to cDNA arrays and levels of specific mRNAs were determined by real-time PCR. Treatment with OE decreased the expression of 232 genes and up-regulated 75 other genes in mesenteric white adipose tissue. The use of real-time PCR validate that, in mesenteric white adipose tissue, mRNA levels for Lipoprotein Lipase (LPL) were decreased by 52%, those of Fatty Acid Synthase (FAS) by 95%, those of Hormone Sensible Lipase (HSL) by 32%, those of Acetyl CoA Carboxylase (ACC) by 92%, those of Carnitine Palmitoyltransferase 1b (CPT1b) by 45%, and those of Fatty Acid Transport Protein 1 (FATP1) and Adipocyte Fatty Acid Binding Protein (FABP4) by 52% and 49%, respectively. Conversely, Tumour Necrosis Factor (TNF¿) values showed overexpression (198%). Conclusion: Short-term treatment with OE affects adipose tissue capacity to extract fatty acids from lipoproteins and to deal with fatty acid transport and metabolism.

A genome-wide association study of neuroticism in a population-based sample.

Neuroticism is a moderately heritable personality trait considered to be a risk factor for developing major depression, anxiety disorders and dementia. We performed a genome-wide association study in 2,235 participants drawn from a population-based study of neuroticism, making this the largest association study for neuroticism to date. Neuroticism was measured by the Eysenck Personality Questionnaire. After Quality Control, we analysed 430,000 autosomal SNPs together with an additional 1.2 million SNPs imputed with high quality from the Hap Map CEU samples. We found a very small effect of population stratification, corrected using one principal component, and some cryptic kinship that required no correction. NKAIN2 showed suggestive evidence of association with neuroticism as a main effect (p < 10(-6)) and GPC6 showed suggestive evidence for interaction with age (p approximately = 10(-7)). We found support for one previously-reported association (PDE4D), but failed to replicate other recent reports. These results suggest common SNP variation does not strongly influence neuroticism. Our study was powered to detect almost all SNPs explaining at least 2% of heritability, and so our results effectively exclude the existence of loci having a major effect on neuroticism.

Thermal resonance in signal transmission

We use temperature tuning to control signal propagation in simple one-dimensional arrays of masses connected by hard anharmonic springs and with no local potentials. In our numerical model a sustained signal is applied at one site of a chain immersed in a thermal environment and the signal-to-noise ratio is measured at each oscillator. We show that raising the temperature can lead to enhanced signal propagation along the chain, resulting in thermal resonance effects akin to the resonance observed in arrays of bistable systems.

Feedback control of unstable cellular solidification fronts

We present a feedback control scheme to stabilize unstable cellular patterns during the directional solidification of a binary alloy. The scheme is based on local heating of cell tips which protrude ahead of the mean position of all tips in the array. The feasibility of this scheme is demonstrated using phase-field simulations and, experimentally, using a real-time image processing algorithm, to track cell tips, coupled with a movable laser spot array device to heat the tips locally. We demonstrate, both numerically and experimentally, that spacings well below the threshold for a period-doubling instability can be stabilized. As predicted by the numerical calculations, cellular arrays become stable with uniform spacing through the feedback control which is maintained with minimal heating.

Proof of concept for microarray-based detection of DNA-binding oncogenes in cell extracts.

The function of DNA-binding proteins is controlled not just by their abundance, but mainly at the level of their activity in terms of their interactions with DNA and protein targets. Moreover, the affinity of such transcription factors to their target sequences is often controlled by co-factors and/or modifications that are not easily assessed from biological samples. Here, we describe a scalable method for monitoring protein-DNA interactions on a microarray surface. This approach was designed to determine the DNA-binding activity of proteins in crude cell extracts, complementing conventional expression profiling arrays. Enzymatic labeling of DNA enables direct normalization of the protein binding to the microarray, allowing the estimation of relative binding affinities. Using DNA sequences covering a range of affinities, we show that the new microarray-based method yields binding strength estimates similar to low-throughput gel mobility-shift assays. The microarray is also of high sensitivity, as it allows the detection of a rare DNA-binding protein from breast cancer cells, the human tumor suppressor AP-2. This approach thus mediates precise and robust assessment of the activity of DNA-binding proteins and takes present DNA-binding assays to a high throughput level.

Defining the genomic signature of the parous breast.

ABSTRACT: BACKGROUND: It is accepted that a woman's lifetime risk of developing breast cancer after menopause is reduced by early full term pregnancy and multiparity. This phenomenon is thought to be associated with the development and differentiation of the breast during pregnancy. METHODS: In order to understand the underlying molecular mechanisms of pregnancy induced breast cancer protection, we profiled and compared the transcriptomes of normal breast tissue biopsies from 71 parous (P) and 42 nulliparous (NP) healthy postmenopausal women using Affymetrix Human Genome U133 Plus 2.0 arrays. To validate the results, we performed real time PCR and immunohistochemistry. RESULTS: We identified 305 differentially expressed probesets (208 distinct genes). Of these, 267 probesets were up- and 38 down-regulated in parous breast samples; bioinformatics analysis using gene ontology enrichment revealed that up-regulated genes in the parous breast represented biological processes involving differentiation and development, anchoring of epithelial cells to the basement membrane, hemidesmosome and cell-substrate junction assembly, mRNA and RNA metabolic processes and RNA splicing machinery. The down-regulated genes represented biological processes that comprised cell proliferation, regulation of IGF-like growth factor receptor signaling, somatic stem cell maintenance, muscle cell differentiation and apoptosis. CONCLUSIONS: This study suggests that the differentiation of the breast imprints a genomic signature that is centered in the mRNA processing reactome. These findings indicate that pregnancy may induce a safeguard mechanism at post-transcriptional level that maintains the fidelity of the transcriptional process.

SNP2GO: functional analysis of genome-wide association studies.

Genome-wide association studies (GWAS) are designed to identify the portion of single-nucleotide polymorphisms (SNPs) in genome sequences associated with a complex trait. Strategies based on the gene list enrichment concept are currently applied for the functional analysis of GWAS, according to which a significant overrepresentation of candidate genes associated with a biological pathway is used as a proxy to infer overrepresentation of candidate SNPs in the pathway. Here we show that such inference is not always valid and introduce the program SNP2GO, which implements a new method to properly test for the overrepresentation of candidate SNPs in biological pathways.

LDL-cholesterol concentrations: a genome-wide association study.

BACKGROUND: LDL cholesterol has a causal role in the development of cardiovascular disease. Improved understanding of the biological mechanisms that underlie the metabolism and regulation of LDL cholesterol might help to identify novel therapeutic targets. We therefore did a genome-wide association study of LDL-cholesterol concentrations. METHODS: We used genome-wide association data from up to 11,685 participants with measures of circulating LDL-cholesterol concentrations across five studies, including data for 293 461 autosomal single nucleotide polymorphisms (SNPs) with a minor allele frequency of 5% or more that passed our quality control criteria. We also used data from a second genome-wide array in up to 4337 participants from three of these five studies, with data for 290,140 SNPs. We did replication studies in two independent populations consisting of up to 4979 participants. Statistical approaches, including meta-analysis and linkage disequilibrium plots, were used to refine association signals; we analysed pooled data from all seven populations to determine the effect of each SNP on variations in circulating LDL-cholesterol concentrations. FINDINGS: In our initial scan, we found two SNPs (rs599839 [p=1.7x10(-15)] and rs4970834 [p=3.0x10(-11)]) that showed genome-wide statistical association with LDL cholesterol at chromosomal locus 1p13.3. The second genome screen found a third statistically associated SNP at the same locus (rs646776 [p=4.3x10(-9)]). Meta-analysis of data from all studies showed an association of SNPs rs599839 (combined p=1.2x10(-33)) and rs646776 (p=4.8x10(-20)) with LDL-cholesterol concentrations. SNPs rs599839 and rs646776 both explained around 1% of the variation in circulating LDL-cholesterol concentrations and were associated with about 15% of an SD change in LDL cholesterol per allele, assuming an SD of 1 mmol/L. INTERPRETATION: We found evidence for a novel locus for LDL cholesterol on chromosome 1p13.3. These results potentially provide insight into the biological mechanisms that underlie the regulation of LDL cholesterol and might help in the discovery of novel therapeutic targets for cardiovascular disease.

Organisation and role of actin microfilaments during cellular growth and morphogenesis in the moss Physcomitrella patens

ABSTRACT The network of actin cytoskeleton is composed of actin filaments (F-actin) that are made by polymerisation of actin monomers and actin binding proteins. It is required for growth and morphogenesis of eukaryotic cells. The labelling of F-actin with constitutively expressed GFP-Talin (Kost et al., 1998) reveals the organisation of cellular actin networks in plants. Due to the lack of information on actin cytoskeleton through gametophytic development of the model moss plant Physcornitrella patens, stable transgenic lines overexpressing GFP-Talin were generated to detect F-actin structures. It is shown that the 35S promoter driven expression is not suitable for F-actin labelling in all cells. When it is replaced by the inducible heat-shock promoter Gmhsp17.3 from soybean, one hour mild heat stress at 37°C followed by recovery at 25°C is enough to induce efficient and transient labelling in all tissues without altering cellular morphology. The optimal observations of F-actin structures at different stages of moss development can be done between 12-18 hours after the induction. By using confocal microscopy, we demonstrate that stellated actin arrays were densely accumulated at the growing tip in regenerating protoplasts, apical protonemal cells and rhizoids and connected with a fine dispersed F-actin mesh. Following three-dimensional growth, the cortical star-like structures are widespread in the meristematic cells of developing bud and young gametophores. On the contrary, undulating networks of actin cables are found at the final stage of cell differentiation. During redifferentiation of mature leaf cells into protonemal filaments the rather stagnant web of actin cables is replaced by diffuse actin meshwork. In eukaryotes, nucleation of the actin monomers prior to their polymerization is driven by the seven-subunit ARP2/3 complex and formins. We cloned the gene encoding the ARP3 subunit of P. patens and generated arp3 mutants of the moss through gene disruption. The knockout of ARP3 affects the elongation of chloronemal cells and blocks further differentiation of caulonemal cells and rhizoids, and the gametophores are slightly stunted compared to wild-type. The arp mutants were created in the heat-shock inducible GFP-Talin strains allowing us to visualise a disorganised actin network and a lack of star-like actin cytoskeleton arrays. We conclude that ARP2/3 dependent nucleation of actin filaments is critical for the growth of filamentous cells, which in turn influences moss colonization. In complementation assays, the overexpression of Physcomitrella and Arab idopsis ARP3 genes in the moss arp3 mutant results in full recovery of wild type phenotype. In contrast the ARP3 subunit of fission yeast is not able to complement the moss arp3 mutant of moss indicating that regulation of the ARP2/3 dependent actin nucleation diverged in different kingdoms. RESUME Le réseau d'actine est composé de filaments de F-actine et d'un ensemble de protéines s'y attachant (Actin binding proteins). Le réseau d'actine est nécessaire à la croissance et à la morphogenèse de toutes les cellules eucaryotes. Chez les plantes, le marquage ainsi que l'étude de l'organisation du réseau d'actine ont été réalisés en utilisant une fusion GFP-Talin (Kost et al., 1998) exprimée sous le control d'un promoteur constitutif. Afin d'étudier les structures F-actine dans les cellules de Physcomitrella Patens et pour combler le manque d'information sur le développement des gamétophores, des lignées transgéniques stables surexprimant GFP-Talin ont été crées. Nous avons démontré que l'utilisation du promoteur 35S est inadéquate pour le marquage complet et homogène des filaments d'actine dans toutes les cellules de P. patens. Par contre, l'utilisation du promoteur inductible Gmhsp17.3 nous a permis de réaliser un marquage transitoire et général dans tous les tissus de la mousse. Une heure de choc thermique à 37°C suivis d'un temps de récupération de 12-18h à 25°C sont les conditions optimales (sans dommages cellulaires) pour l'observation des structures F-actine à différentes étapes de développement de la mousse. En utilisant la microscopie confocale, nous avons observé l'existence de structures F-actine accumulées en forme d'étoiles. Ces structures, qui sont liées au réseau de microfilaments d'actine, ont été observées dans les protoplastes en régénération, les cellules des protonema apicales ainsi que dans les rhizoïdes. En suivant la croissance tridimensionnelle, ces structures en étoiles ont été observées dans les cellules meristématiques des bourgeons et des jeunes gamétophores. Par contre, dans les cellules différentiées ces structures laissent place à des réseaux de câbles épais. Nous avons également remarqué que durant la redifferentiation des cellules foliaires le réseau de câbles de F-actine est remplacé par un réseau de F-actine diffus. Dans les cellules eucaryotes, la nucléation des filaments d'actirie précédant leur polymérisation est contrôlé par sept sous unités du complexe ARP2/3 et par des formines. Nous avons isolé le gène codant pour la sous unité ARP3 de P. patens et nous avons crée des mutants arp3 par intégration ciblée (Knockout). L'élongation des cellules chloronema est clairement affectée dans les mutants arp3. La différentiation des caulonemata et des rhizoïdes est bloquée et les gametophores sont légèrement plus courts comparé au type sauvage. A fin d'étudier l'organisation des filaments d'actines dans les mutants arp3, nous avons aussi réalisé un arp3-knockout dans la lignée Hsp-GFP-Talin. La nouvelle lignée générée nous a permis de visualiser une désorganisation du réseau d'actine et une absence complète de structures de F-actine accumulée en forme d'étoiles. Les résultats obtenus nous amènent à conclure que la nucléation (ARP2/3 dépendante) des filaments d'actine est indispensable à la croissance des cellules filamenteuses. Par conséquent, les filaments d'actine semblent avoir un rôle dans la colonisation des milieux par les mousses. Nous avons également procédé à des essais de complémentation du mutant arp3. La surexpression des gènes ARP3 de Physcomitrella et d'Arabidopsis dans les cellules du mutant arp3 rétabli complètement le phénotype WT. Par contre, le gène ARP3 des levures n'est pas suffisant pour complémenter la même mutation dans les cellules de mousses. Ce résultat démontre que les mécanismes de régulation de la nucléation des filaments d'actine (ARP2/3 dépendante) sont différents entre les différents groupes d'eucaryotes.

Characterization of a genomic signature of pregnancy identified in the breast.

The objective of this study was to comprehensively compare the genomic profiles in the breast of parous and nulliparous postmenopausal women to identify genes that permanently change their expression following pregnancy. The study was designed as a two-phase approach. In the discovery phase, we compared breast genomic profiles of 37 parous with 18 nulliparous postmenopausal women. In the validation phase, confirmation of the genomic patterns observed in the discovery phase was sought in an independent set of 30 parous and 22 nulliparous postmenopausal women. RNA was hybridized to Affymetrix HG_U133 Plus 2.0 oligonucleotide arrays containing probes to 54,675 transcripts, scanned and the images analyzed using Affymetrix GCOS software. Surrogate variable analysis, logistic regression, and significance analysis of microarrays were used to identify statistically significant differences in expression of genes. The false discovery rate (FDR) approach was used to control for multiple comparisons. We found that 208 genes (305 probe sets) were differentially expressed between parous and nulliparous women in both discovery and validation phases of the study at an FDR of 10% and with at least a 1.25-fold change. These genes are involved in regulation of transcription, centrosome organization, RNA splicing, cell-cycle control, adhesion, and differentiation. The results provide initial evidence that full-term pregnancy induces long-term genomic changes in the breast. The genomic signature of pregnancy could be used as an intermediate marker to assess potential chemopreventive interventions with hormones mimicking the effects of pregnancy for prevention of breast cancer.

Polymorphisms in toll-like receptor 4 and toll-like receptor 9 influence viral load in a seroincident cohort of HIV-1-infected individuals.

OBJECTIVES: Toll-like receptors (TLRs) are innate immune sensors that are integral to resisting chronic and opportunistic infections. Mounting evidence implicates TLR polymorphisms in susceptibilities to various infectious diseases, including HIV-1. We investigated the impact of TLR single nucleotide polymorphisms (SNPs) on clinical outcome in a seroincident cohort of HIV-1-infected volunteers. DESIGN: We analyzed TLR SNPs in 201 antiretroviral treatment-naive HIV-1-infected volunteers from a longitudinal seroincident cohort with regular follow-up intervals (median follow-up 4.2 years, interquartile range 4.4). Participants were stratified into two groups according to either disease progression, defined as peripheral blood CD4(+) T-cell decline over time, or peak and setpoint viral load. METHODS: Haplotype tagging SNPs from TLR2, TLR3, TLR4, and TLR9 were detected by mass array genotyping, and CD4(+) T-cell counts and viral load measurements were determined prior to antiretroviral therapy initiation. The association of TLR haplotypes with viral load and rapid progression was assessed by multivariate regression models using age and sex as covariates. RESULTS: Two TLR4 SNPs in strong linkage disequilibrium [1063 A/G (D299G) and 1363 C/T (T399I)] were more frequent among individuals with high peak viral load compared with low/moderate peak viral load (odds ratio 6.65, 95% confidence interval 2.19-20.46, P < 0.001; adjusted P = 0.002 for 1063 A/G). In addition, a TLR9 SNP previously associated with slow progression was found less frequently among individuals with high viral setpoint compared with low/moderate setpoint (odds ratio 0.29, 95% confidence interval 0.13-0.65, P = 0.003, adjusted P = 0.04). CONCLUSION: This study suggests a potentially new role for TLR4 polymorphisms in HIV-1 peak viral load and confirms a role for TLR9 polymorphisms in disease progression.

No evidence for mutations of CTCFL/BORIS in Silver-Russell syndrome patients with IGF2/H19 imprinting control region 1 hypomethylation.

BACKGROUND: Silver-Russell syndrome (SRS) is a genetically and clinically heterogeneous disease. Although no protein coding gene defects have been reported in SRS patients, approximately 50% of SRS patients carry epimutations (hypomethylation) at the IGF2/H19 imprinting control region 1 (ICR1). Proper methylation at ICR1 is crucial for the imprinted expression of IGF2, a fetal growth factor. CTCFL, a testis-specific protein, has recently been proposed to play a role in the establishment of DNA methylation at the murine equivalent of ICR1. A screen was undertaken to assess whether CTCFL is mutated in SRS patients with hypomethylation, to explore a link between the observed epimutations and a genetic cause of the disease. METHODOLOGY/PRINCIPAL FINDINGS: DNA was obtained from 36 SRS patients with hypomethylation at ICR1. All CTCFL coding exons were sequenced and analyzed for duplications/deletions using both multiplex ligation-dependent probe amplification, with a custom CTCFL probe set, and genomic qPCR. Novel SNP alleles were analyzed for potential differential splicing in vitro utilizing a splicing assay. Neither mutations of CTCFL nor duplications/deletions were observed. Five novel SNPs were identified and have been submitted to dbSNP. In silico splice prediction suggested one novel SNP, IVS2-66A>C, activated a cryptic splice site, resulting in aberrant splicing and premature termination. In vitro splicing assays did not confirm predicted aberrant splicing. CONCLUSIONS/SIGNIFICANCE: As no mutations were detected at CTCFL in the patients examined, we conclude that genetic alterations of CTCFL are not responsible for the SRS hypomethylation. We suggest that analysis of other genes involved in the establishment of DNA methylation at imprinted genes, such as DNMT3A and DNMT3L, may provide insight into the genetic cause of hypomethylation in SRS patients.

Comparative genomic and phylogeographic analysis of Mycobacterium leprae.

Reductive evolution and massive pseudogene formation have shaped the 3.31-Mb genome of Mycobacterium leprae, an unculturable obligate pathogen that causes leprosy in humans. The complete genome sequence of M. leprae strain Br4923 from Brazil was obtained by conventional methods (6x coverage), and Illumina resequencing technology was used to obtain the sequences of strains Thai53 (38x coverage) and NHDP63 (46x coverage) from Thailand and the United States, respectively. Whole-genome comparisons with the previously sequenced TN strain from India revealed that the four strains share 99.995% sequence identity and differ only in 215 polymorphic sites, mainly SNPs, and by 5 pseudogenes. Sixteen interrelated SNP subtypes were defined by genotyping both extant and extinct strains of M. leprae from around the world. The 16 SNP subtypes showed a strong geographical association that reflects the migration patterns of early humans and trade routes, with the Silk Road linking Europe to China having contributed to the spread of leprosy.

Genetic Association of Recovery from Eating Disorders: The Role of GABA Receptor SNPs.

Follow-up studies of eating disorders (EDs) suggest outcomes ranging from recovery to chronic illness or death, but predictors of outcome have not been consistently identified. We tested 5151 single-nucleotide polymorphisms (SNPs) in approximately 350 candidate genes for association with recovery from ED in 1878 women. Initial analyses focused on a strictly defined discovery cohort of women who were over age 25 years, carried a lifetime diagnosis of an ED, and for whom data were available regarding the presence (n=361 ongoing symptoms in the past year, ie, 'ill') or absence (n=115 no symptoms in the past year, ie, 'recovered') of ED symptoms. An intronic SNP (rs17536211) in GABRG1 showed the strongest statistical evidence of association (p=4.63 × 10(-6), false discovery rate (FDR)=0.021, odds ratio (OR)=0.46). We replicated these findings in a more liberally defined cohort of women age 25 years or younger (n=464 ill, n=107 recovered; p=0.0336, OR=0.68; combined sample p=4.57 × 10(-6), FDR=0.0049, OR=0.55). Enrichment analyses revealed that GABA (γ-aminobutyric acid) SNPs were over-represented among SNPs associated at p<0.05 in both the discovery (Z=3.64, p=0.0003) and combined cohorts (Z=2.07, p=0.0388). In follow-up phenomic association analyses with a third independent cohort (n=154 ED cases, n=677 controls), rs17536211 was associated with trait anxiety (p=0.049), suggesting a possible mechanism through which this variant may influence ED outcome. These findings could provide new insights into the development of more effective interventions for the most treatment-resistant patients.