Abstract
Purpose
Urinary incontinence (UI) and fecal incontinence (FI) are common disorders in women that negatively impact quality of life. In addition to known health and lifestyle risk factors, genetics may play a role. Identification of genetic variants associated with UI and FI could result in better understanding of etiologic pathways and new interventions and treatments.
Methods
We previously generated genome-wide single nucleotide polymorphism (SNP) data from Nurses’ Health Studies participants. The participants provided longitudinal UI and FI information via questionnaires. Cases of UI (n=6120) reported at least weekly UI on a majority of questionnaires (3 or 4 across 12–16 years); controls (n=4811) consistently reported little to no UI. We classified women with UI into stress (n=1809), urgency (n=1942), and mixed (n=2036) subtypes. Cases of FI (n=4247) reported at least monthly FI on a majority of questionnaires; controls (n=11634) consistently reported no FI. We performed a genome-wide association study (GWAS) for each incontinence outcome.
Results
We identified 8 SNPs significantly associated (p<5×10−8) with UI, located in two loci, chromosome 8q23.3 and 1p32.2. There were no genome-wide significant findings for the UI subtype analyses; however, the significant associations for overall UI were stronger for the urgency and mixed subtypes than for stress. While no SNP reached genome-wide significance for FI, four SNPs had p<10−6.
Conclusion
Few studies have collected both genetic data and detailed UI and FI information. This GWAS provides initial evidence of genetic associations for UI and merits further research to replicate our findings and identify additional risk variants.
Keywords: urinary incontinence, fecal incontinence, incontinence, genome-wide association study, GWAS
Introduction
Urinary incontinence (UI) and fecal incontinence (FI) are common disorders affecting women. Based on data from the National Health and Nutrition Examination Survey (NHANES), about 16% and 9% of U.S. women report history of UI and FI, respectively1. Both conditions have a significant negative impact on physical and emotional quality of life2, 3. There are known risk factors for UI and more limited knowledge regarding FI, although age, race, parity, BMI, and some others have been identified4. However, little is known regarding genetic predisposition to incontinence.
Women with UI, including stress, urgency and mixed UI, show familial aggregation. Studies of twins using a Danish population-based twin registry suggest that urgency UI and mixed UI (but not stress UI) have a significant genetic component5. On the other hand, the Swedish twin registry revealed the presence of a strong genetic risk for stress UI6. Genetic loci have been suggested for urgency UI7, 8 and for stress UI9. Limited information exists from epidemiologic or familial aggregation studies on the relationship between genetic influences and risk of FI10, 11.
One approach to identify genetic factors associated with UI and FI susceptibility is to perform genome-wide association studies (GWAS)12. Members of the Pelvic Floor Disorders Network published a published a GWAS of 2,241 cases of urgency UI and 776 controls from the Women’s Health Initiative that identified six loci associated with urgency UI13. Additional GWAS with larger sample sizes are needed to validate these findings, and to study FI and UI across all subtypes. The discovery of genetic variants associated with the risk of UI and FI could lead to better understanding of the pathophysiology of these conditions, and could lead to improved interventions and treatments. The Nurses’ Health Study (NHS) has enabled the identification of genetic loci involved in many other complex human diseases14, 15. In this study, we performed a GWAS using the NHS to identify genetic variants associated with the risk of UI (and its subtypes) and FI in women.
Materials and Methods
Study Population
In 1976, the NHS was initiated with 121,700 nurse participants, 30 to 55 years of age in 1976, who have answered biennial questionnaires (follow up >90% over time) about health and lifestyle16. In addition, 32,826 blood and 29,684 cheek cell samples have been collected since the late 1980s. An additional 116,430 nurses, aged 25 to 42, were recruited in 1989 as a part of Nurses’ Health Study II (NHSII) and have returned biennial questionnaires similar to those used for NHS17. We have collected blood samples from 29,612 women and cheek cell samples from an additional 29,859 women in NHSII.
GWAS data
GWAS for numerous disease outcomes (e.g. breast cancer, type 2 diabetes) have been conducted in the NHS and NHSII participant samples using nested case-control study design, and secondary analyses have been performed leveraging these genetic data18–20. In total, these GWAS data sets comprise 19,959 women from NHS and 8,488 women from NHSII. These studies were genotyped on different arrays at different genotyping centers (described in Lindstrom et al14; we have included two additional datasets since this publication). Standard quality control filters for call rate, Hardy-Weinberg equilibrium, and other measures were applied to the genotyped SNPs and samples14. Studies were grouped by genotyping platform (see Supplementary Table 1 for samples sizes). Each grouped dataset was imputed using the 1000 Genomes Project ALL Phase 3 Version 5 as the reference panel21, resulting in 1,410,640 variants across the 5 platforms (each SNP is present in 1–5 platforms). We restricted our analysis to participants with self-reported European ancestry. We also removed outliers that were identified as non-white by principal components analysis.
Outcome definitions
On biennial questionnaires, participants in NHS and NHSII have been regularly asked about urine leakage. Specifically, the UI questions are: 1) “During the past 12 months, how often have you leaked urine or lost control of your urine?” with response options of: never, <1/month, 1/month, 2–3x/month, 1/week, almost every day and 2) “When you lose urine, how much usually leaks?” with response options of: a few drops, enough to wet your underwear, enough to wet outer clothing, enough to wet the floor. These questions have been repeated seven times across sixteen years in the NHS cohort starting in 1996, and five times across twelve years in NHSII starting in 2001, providing detailed ascertainment of UI over time. For this analysis, among those with GWAS data, we defined cases of UI (n=6120) as those who reported at least weekly UI on a majority of questionnaires (≥4 in NHS, ≥3 in NHSII). Controls (n=4811) were women with GWAS data who reported never experiencing UI or no more than leaking a few drops less than once a month on all questionnaires to which they responded; all eligible participants must have responded to the majority of questionnaires.
For UI subtypes, we defined stress urinary incontinence (SUI) as a self-report that most leaking episodes related to physical activity, coughing, sneezing, etc; urgency urinary incontinence (UUI) as most leaking episodes related to a feeling of urgency; mixed urinary incontinence (MUI) as leaking episodes equally related to both activity/coughing/etc and urgency. We classified those who met the overall UI definition into SUI (n=1809), UUI (n=1942), and MUI (n=2036) based on the subtype they first reported. The subtype questions were not included on the questionnaires until 2004 in NHS and 2005 in NHSII, so the total number of participants with subtype information is less than the total UI cases. We included all UI controls for the three subtype analyses. We additionally created more stringent subtype definitions, using either the most common/majority subtype reported across all questionnaires (SUI n=1347, UUI n=1860, MUI=1930) or requiring that a participant always reported the same subtype across all returned questionnaires (SUI n=700, UUI n=848, MUI n=676).
Covariates for the UI analyses were included from the questionnaire when a participant first answered the urine leakage questions. These included age (continuous), BMI (<25, 25 to <30, and ≥30), parity (nulliparous, 1–2 births, 3 or more births), and Type II diabetes (yes/no). For those missing BMI (n=8) or parity (n=63), the median category was used.
The 2008, 2010, and 2012 NHS, and the 2011 and 2013 NHSII questionnaires included the question “On average, how many times in the past year have you experienced any amount of accidental bowel leakage?” with the response options of never, <1/month, 1–3/month; 1/week, several times/week, daily. We defined cases of FI (n=4247) among those with GWAS data as those who reported at least monthly FI on a majority of questionnaires (≥2 in NHS, ≥1 in NHSII); controls (n=11634) had GWAS data and consistently reported no FI and replied to the majority of questionnaires.
A flow chart of the participants in each analysis is provided in Supplementary Figure 1. Statistical power for each outcome analysis is in Supplementary Figure 2.
Statistical analysis
The RVTests22 program was used to test the association of all SNPs with overall UI, each UI subtype, and FI, based on logistic regression models under an additive genetic model. All models were adjusted for the top 4 principal components to account for population structure. Each genotyping platform grouped dataset was analyzed separately. SNPs with poor imputation quality (R2<0.3) were excluded. We also applied a minor allele frequency threshold to each platform; this varied depending on the number of cases and controls, and was therefore different for each analysis (details provided in Supplementary Table 1). For each outcome, a meta-analysis was conducted to combine the results from each platform using a fixed effects model using METAL23. We considered p<5×10−8 to be genome-wide significant.
We performed the same GWAS for overall UI and the UI subtypes additionally adjusting for known UI risk factors (age, BMI, parity, and diabetes), as described above. For the UI subtypes, we also tested the association of all SNPs with p<1×10−5 from the overall UI analysis regardless of their minor allele frequency to examine the overlap across subtypes. We also performed the same GWAS using the more stringent UI subtype definitions, described above. Q-Q plots for each GWAS are provided in Supplementary Figure 3.
The study protocol was approved by the Institutional Review Board of Brigham and Women’s Hospital.
GTEx analysis
For all SNPs with p<10−5 from the overall UI, UI subtypes, and FI analyses, we evaluated whether the variants were significant cis expression quantitative trait loci (eQTL) based on information from the Genotypes and Tissue Expression (GTEx) database24, 25. A cis eQTL is a variant or locus associated with the expression of a nearby gene. eQTL identification has been previously described in Fagny et al26. Briefly, this analysis leveraged publicly available gene expression data for 19 tissues in GTEx version 6.0 from at least 150 distinct individuals with imputed genetic data. For each of these 19 tissues, we identified SNPs that had a minor allele frequency greater than 0.05 across individuals with gene expression data. We then used Matrix eQTL27 to quantify the statistical association of the expression of autosomal genes with each of these genetic variants. For this analysis, we used a window around the gene of 1 mega-base, and adjusted for sex, age, and the first three principal components obtained from the genotyping data, to determine eQTLs in the GTEx dataset. We examined the GTEx associations for our top (p<10−5) SNPs, and report the significant eQTL associations (p<0.05) between these candidate SNPs and nearby genes.
Results
The number of participants in each UI GWAS and risk factors associated with UI are presented in Table 1. We identified 8 SNPs significantly associated with UI (p<5×10−8). These SNPs are located within two loci, chromosome 8q23.3 (in the TRPS1 gene region) and 1p32.2 (closest gene DAB1). A Manhattan plot of the p-values for genome-wide findings is presented in Figure 1. For the UI subtypes, no SNP reached genome-wide significance (Manhattan plots in Supplementary Figure 4). No genome-wide significant associations were observed when using the more stringent UI subtype definitions with somewhat smaller sample sizes (results not shown).
Table 1.
Characteristics of UI controls, UI cases, and UI cases by subtype
| UI controls | all UI cases | UUI cases | SUI cases | MUI cases | |
|---|---|---|---|---|---|
| N | 4811 | 6120 | 1942 | 1809 | 2036 |
| age, median (IQR) | 54 (49–63) | 61 (53–68) | 63 (55–69) | 57 (51–65) | 62 (54–68) |
| BMI (%) | |||||
| <25 | 2843 (59.1) | 2153 (35.2) | 717 (36.9) | 690 (38.1) | 634 (31.1) |
| 25–<30 | 1321 (27.5) | 2028 (33.1) | 603 (31.1) | 604 (33.4) | 706 (34.7) |
| ≥30 | 647 (13.4) | 1939 (31.7) | 622 (32.0) | 515 (28.5) | 696 (34.2) |
| Parity (%) | |||||
| nulliparous | 687 (14.3) | 493 (8.1) | 135 (7.0) | 178 (9.8) | 156 (7.7) |
| 1–2 children | 2072 (43.1) | 2256 (36.9) | 652 (33.6) | 753 (41.6) | 726 (35.7) |
| ≥3 children | 2052 (42.7) | 3371 (55.1) | 1155 (59.5) | 878 (48.5) | 1154 (56.7) |
| Diabetes (%) | 131 (2.7) | 367 (6.0) | 133 (6.9) | 76 (4.2) | 132 (6.5) |
Figure 1.
Manhattan plot for urinary incontinence. Dashed line denotes p=5×10−8.
When we adjusted the overall UI GWAS for known risk factors, the chromosome 1 and 8 loci were no longer genome-wide significant. However, these remained the top ranked loci (p~2×10−7 and p~5×10−7, respectively), while no new genome-wide significant results were identified. For the UI subtypes, after adjusting for known risk factors the top ranked SNP in the unadjusted SUI analysis became genome-wide significant (rs7607995; p=4.5×10−8; chromosome 2p13.1; WDR54) but no other new significant findings emerged.
Comparing the significant SNPs in the overall UI analysis with the top SNPs in the GWAS for each of the subtypes, there was limited overlap. However, when we included the top-ranked overall UI SNPs (p<10−5) that had initially dropped out of each subtype analysis due to minor allele frequency filtering, we observed that the chromosome 8q23.3 locus was highly significant (p~10−10) in the UUI analysis, but absent from the top ranked SNPs in SUI (p~0.02) and MUI (p~0.0005). For the chromosome 1p32.2 locus, the SNPs were more significant in both MUI (p~5–8×10−6) and UUI (p~1×10−6), but were again less significant in the SUI analysis (p~0.001), potentially indicating some shared etiology for UUI and MUI. As is shown in Figure 2, a larger number of the top overall SNPs are the top SNPs in the UUI and MUI analysis. Of the 624 SNPs with p<10−5, 48 SNPs (6 loci) are top ranked (p<10−5) in the UUI analysis. For MUI, 18 SNPs (5 loci) overlap; 9 of these SNPs (2 loci) are also present in the top ranked UUI SNPs. This is in contrast to the more limited overlap with SUI (8 SNPs; 2 loci), despite similar sample sizes across subtypes.
Figure 2.
Overlap of top (p<10−5) SNPs in the overall UI analysis with top SNPs in the UI subtypes. Number in parentheses is the total number of significant SNPs in that analysis.
No SNP reached genome-wide significance for FI (Figure 3). However, top ranked SNPs, including rs7586405 on chromosome 2 (p=3×10−7) and rs2715291 on chromosome 18 (p=4.4×10−7), may be interesting candidates for future studies.
Figure 3.
Manhattan plot for fecal incontinence. Dashed line denotes p=10−5.
Of all top ranked (p<10−5) SNPs, only rs7607995 (significant SUI SNP after adjusting for known UI risk factors) was significantly associated with gene expression in the GTEx database. This SNP was significantly associated with several genes – LBX2/LBX2-AS1, MOGS, and WDR54 - across several tissues, including thyroid, esophagus, and lung.
Discussion
In this large, well-characterized study population, we identified two loci associated with risk of UI in one of the first GWAS for this outcome. While there were no genome-wide significant associations in unadjusted analysis of the three UI subtypes, a SNP did become significant in the SUI analysis after adjusting for known UI risk factors. Additionally, comparing the overall UI significant hits to the results from each subtype, we observed that one locus was strongly associated with UUI and the other with both UUI and MUI. There was greater overlap of top overall UI results with UUI and MUI subtypes than with SUI. We hypothesize that SUI may be less genetically determined, as the pathophysiology of SUI is believed to be related to the impairment or disruption of the pelvic support structures that maintain the continence mechanism. This additionally may be why the significant association with SUI did not emerge until controlling for known UI risk factors. UUI and MUI on the other hand, are more idiopathic and may have a stronger genetic predisposition.
One of the genome-wide significant loci for overall UI was located on chromosome 8q23.3, in the TRPS1 gene region. TRPS1 is a transcription factor that regulates genes involved with cartilage differentiation and proliferation, and additionally regulates the mesenchymal-epithelial transition during nephron formation and ureteric bud branching during renal development28, 29. The other significant locus is at 1p32.2, a few hundred kilobases away from the gene DAB1. DAB1 regulates the Reelin pathway, which is responsible for neuronal positioning during brain formation. While neither of these may be obviously linked with UI, both bone (through pelvic support) and brain (bladder control) may play a significant role in the continence mechanism. However, in the GTEx analysis, the significant variants at these loci were not found to be significant eQTL.
Despite including fairly large sample sizes for the outcomes of UI and FI, the power of this study is limited, particularly to examine less common variants and to detect small effect sizes. Obtaining additional samples from studies with detailed clinical information is crucial to both confirm our findings and to discover additional candidate variants. However, to our knowledge, the study from the Women’s Health Initiative (WHI) is the only previously published GWAS of UI, and it focused only on UUI13. Our current study did not replicate any of the significant UUI findings from this previous study in either the UUI or overall UI analysis (results not shown). The WHI GWAS had a less stringent case definition (ever reported leaking at least monthly) than the current study (at least monthly UI regularly reported on multiple questionnaires). This difference, as well as chance, could explain why we were unable to replicate the previous UUI findings.
Conclusions
In conclusion, our findings suggest that there are potentially genetic variants associated with UUI and MUI, and possibly SUI. If confirmed, genetic variants associated with the risk of UI and FI could improve the understanding of the pathophysiology of disease and potentially aid in the development of interventions and treatments. In particular, MUI is currently poorly understood, difficult to treat, and associated with the worst quality of life outcomes30, thus further research could be highly important.
Supplementary Material
Funding:
This study was supported by R21HD089502. The Nurses’ Health Studies are supported by UM1CA186107, R01CA49449, UM1CA176726, and R01CA67262. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Glossary
- SNP
single nucleotide polymorphism
- GWAS
genome-wide association study
- NHS
Nurses’ Health Study
- NHS II
Nurses’ Health Study II
- UI
urinary incontinence
- FI
fecal incontinence
- SUI
stress urinary incontinence
- UUI
urgency urinary incontinence
- MUI
mixed urinary incontinence
References
- 1.Nygaard I, Barber MD, Burgio KL et al. : Prevalence of symptomatic pelvic floor disorders in US women. JAMA, 300: 1311, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Markland AD, Richter HE, Kenton KS et al. : Associated factors and the impact of fecal incontinence in women with urge urinary incontinence: from the Urinary Incontinence Treatment Network’s Behavior Enhances Drug Reduction of Incontinence study. Am J Obstet Gynecol, 200: 424e1, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Fialkow MF, Melville JL, Lentz GM et al. : The functional and psychosocial impact of fecal incontinence on women with urinary incontinence. Am J Obstet Gynecol, 189: 127, 2003 [DOI] [PubMed] [Google Scholar]
- 4.Matthews CA, Whitehead WE, Townsend MK et al. : Risk factors for urinary, fecal, or dual incontinence in the Nurses’ Health Study. Obstet Gynecol, 122: 539, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rohr G, Kragstrup J, Gaist D et al. : Genetic and environmental influences on urinary incontinence: a Danish population-based twin study of middle-aged and elderly women. Acta Obstet Gynecol Scand, 83: 978, 2004 [DOI] [PubMed] [Google Scholar]
- 6.Wennberg AL, Altman D, Lundholm C et al. : Genetic influences are important for most but not all lower urinary tract symptoms: a population-based survey in a cohort of adult Swedish twins. Eur Urol, 59: 1032, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Norton P, Milsom I: Genetics and the lower urinary tract. Neurourol Urodyn, 29: 609, 2010 [DOI] [PubMed] [Google Scholar]
- 8.Eiberg H, Shaumburg HL, Von Gontard A et al. : Linkage study of a large Danish 4-generation family with urge incontinence and nocturnal enuresis. J Urol, 166: 2401, 2001 [PubMed] [Google Scholar]
- 9.McKenzie P, Rohozinski J, Badlani G: Genetic influences on stress urinary incontinence. Curr Opin Urol, 20: 291, 2010 [DOI] [PubMed] [Google Scholar]
- 10.Abramov Y, Sand PK, Botros SM et al. : Risk factors for female anal incontinence: new insight through the Evanston-Northwestern twin sisters study. Obstet Gynecol, 106: 726, 2005 [DOI] [PubMed] [Google Scholar]
- 11.Landefeld CS, Bowers BJ, Feld AD et al. : National Institutes of Health state-of-the-science conference statement: prevention of fecal and urinary incontinence in adults. Ann Intern Med, 148: 449, 2008 [DOI] [PubMed] [Google Scholar]
- 12.Cartwright R, Mangera A, Tikkinen KA et al. : Systematic review and meta-analysis of candidate gene association studies of lower urinary tract symptoms in men. Eur Urol, 66: 752, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Richter HE, Whitehead N, Arya L et al. : Genetic contributions to urgency urinary incontinence in women. J Urol, 193: 2020, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lindstrom S, Loomis S, Turman C et al. : A comprehensive survey of genetic variation in 20,691 subjects from four large cohorts. PLoS One, 12: e0173997, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Townsend MK, Aschard H, De Vivo I et al. : Genomics, Telomere Length, Epigenetics, and Metabolomics in the Nurses’ Health Studies. Am J Public Health, 106: 1663, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Colditz GA, Hankinson SE: The Nurses’ Health Study: lifestyle and health among women. Nat Rev Cancer, 5: 388, 2005 [DOI] [PubMed] [Google Scholar]
- 17.Tworoger SS, Sluss P, Hankinson SE: Association between plasma prolactin concentrations and risk of breast cancer among predominately premenopausal women. Cancer Res, 66: 2476, 2006 [DOI] [PubMed] [Google Scholar]
- 18.Hazra A, Kraft P, Lazarus R et al. : Genome-wide significant predictors of metabolites in the one-carbon metabolism pathway. Hum Mol Genet, 18: 4677, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.He C, Kraft P, Chasman DI et al. : A large-scale candidate gene association study of age at menarche and age at natural menopause. Hum Genet, 128: 515, 2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Coffee, Caffeine Genetics, C., Cornelis MC et al. : Genome-wide meta-analysis identifies six novel loci associated with habitual coffee consumption. Mol Psychiatry, 20: 647, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Genomes Project C, Auton A, Brooks LD et al. : A global reference for human genetic variation. Nature, 526: 68, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zhan X, Hu Y, Li B et al. : RVTESTS: an efficient and comprehensive tool for rare variant association analysis using sequence data. Bioinformatics, 32: 1423, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Willer CJ, Li Y, Abecasis GR: METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics, 26: 2190, 2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Consortium GT, Laboratory DA, Coordinating Center -Analysis Working, G. et al. : Genetic effects on gene expression across human tissues. Nature, 550: 204, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Consortium GT: Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science, 348: 648, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fagny M, Paulson JN, Kuijjer ML et al. : Exploring regulation in tissues with eQTL networks. Proc Natl Acad Sci U S A, 114: E7841, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Shabalin AA: Matrix eQTL: ultra fast eQTL analysis via large matrix operations. Bioinformatics, 28: 1353, 2012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Gui T, Sun Y, Gai Z et al. : The loss of Trps1 suppresses ureteric bud branching because of the activation of TGF-beta signaling. Dev Biol, 377: 415, 2013 [DOI] [PubMed] [Google Scholar]
- 29.Gai Z, Gui T, Muragaki Y: The function of TRPS1 in the development and differentiation of bone, kidney, and hair follicles. Histol Histopathol, 26: 915, 2011 [DOI] [PubMed] [Google Scholar]
- 30.Minassian VA, Devore E, Hagan K et al. : Severity of urinary incontinence and effect on quality of life in women by incontinence type. Obstet Gynecol, 121: 1083, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.