Genome-Wide Association Analysis Implicates Elastic Microfibrils in the Development of Nonsyndromic Striae Distensae

      Abbreviations

      BMI
      body mass index
      SNP
      single-nucleotide polymorphism

      TO THE EDITOR

      Striae distensae, or stretch marks, are a common skin condition that appear initially as red, and later on as white, lines on the skin. These lines represent scars of the dermis, and are characterized by linear bundles of collagen lying parallel to the surface of the skin, as well as eventual loss of collagen and elastin. Reports differ on the level of fragmentation of elastic fibers (
      • Zheng P.
      • Lavker R.M.
      • Kligman A.M.
      Anatomy of striae.
      ;
      • Sheu H.M.
      • Yu H.S.
      • Chang C.H.
      Mast cell degranulation and elastolysis in the early stage of striae distensae.
      ). Estimates of the prevalence of stretch marks range from 50 to 80% (
      • Atwal G.S.
      • Manku L.K.
      • Griffiths C.E.
      • et al.
      Striae gravidarum in primiparae.
      ;
      • Cho S.
      • Park E.S.
      • Lee D.H.
      • et al.
      Clinical features and risk factors for striae distensae in Korean adolescents.
      ). Although stretch marks are only harmful in extreme cases (
      • Dosal J.
      • Handler M.Z.
      • Ricotti C.A.
      • et al.
      Ulceration of abdominal striae distensae (stretch marks) in a cancer patient—quiz case.
      ), even mild stretch marks can cause distress to the bearer.
      The causes of stretch marks are not well understood. Excessive skin distension (such as that which occurs during pregnancy, growth spurts in puberty, or rapid weight gain), prolonged exposure to cortisol (such as in individuals with Cushing syndrome), and genetics may all have a role (
      • Elsaie M.L.
      • Baumann L.S.
      • Elsaaiee L.T.
      Striae distensae (stretch marks) and different modalities of therapy: an update.
      ). A few monogenic connective tissue diseases, including Marfan syndrome and congenital contractural arachnodactyly, are known to be associated with stretch marks. These syndromes are caused by mutations in genes that encode extracellular matrix proteins (fibrillin-1 and fibrillin-2, respectively) that are part of elastic microfibrils present in skin and other tissues. However, to date, no genetic variants are known to be associated with isolated stretch marks that afflict the general population.
      To identify variants associated with the development of stretch marks, we conducted a genome-wide association analysis of stretch marks in a discovery cohort of 33,930 unrelated 23andMe customers (Supplementary Table S1 online) of European descent. There were a total of 13,068 cases and 20,862 controls. The 18,650 men in the cohort were much less likely to report stretch marks (25% versus 55% of women), which is consistent with other reports (
      • Elsaie M.L.
      • Baumann L.S.
      • Elsaaiee L.T.
      Striae distensae (stretch marks) and different modalities of therapy: an update.
      ). We further evaluated the associated variants in a cohort consisting of 4,967 female 23andMe customers of European descent (disjoint from the first group) who reported on severity of stretch marks during pregnancy (also known as striae gravidarum, a closely related phenotype). See Supplemental Methods for additional details on phenotyping. The protocol for this study was approved by an independent institutional review board (E&I Review Services) and was conducted according to the Declaration of Helsinki Principles; all study participants provided informed consent online, which was recorded in an electronic database. See Supplementary Methods for details on genotyping and imputation. All analyses used logistic (discovery cohort) or linear (pregnancy cohort) regression against imputed allele dosages, controlling for age, population structure (using five principal components), and (except for the pregnancy cohort) sex. Because the prevalence of stretch marks differs between men and women, we checked for but did not observe differing effects for the single-nucleotide polymorphisms (SNPs) in men and women.
      Four regions were significantly (P<5e-8) associated with stretch marks (Table 1, Figure 1, Supplementary Figure S1 online). The most strongly associated SNP in the first region, rs7787362 (P=1.8e-23, odds ratio (OR)=0.84), lies 40kb upstream of the ELN (elastin) gene. It was also associated with striae gravidarum in the pregnancy cohort (P=7e-5, β=-0.072, Supplementary Table S2 online). Elastin is the major component of elastic fibers, which provide reversible extensibility to connective tissue. Mutations in elastin that result in a loss of mature elastin can lead to autosomal dominant cutis laxa (a condition characterized by loose, sagging skin, and higher risks of aortic aneurysm) or supravalvular aortic stenosis (a localized narrowing of the ascending aorta caused by thickening of the aortic smooth muscle layer to compensate for the loss of elastin;
      • Milewicz D.M.
      • Urbán Z.
      • Boyd C.
      Genetic disorders of the elastic fiber system.
      ). ELN is also one of the genes deleted in Williams–Beuren syndrome, whose symptoms can include lax skin and supravalvular aortic stenosis, among others. Duplication of the elastin gene does not clearly lead to any skin phenotype (
      • Merla G.
      • Brunetti-Pierri N.
      • Micale L.
      • et al.
      Copy number variants at Williams-Beuren syndrome 7q11.23 region.
      ), which suggests that rs7787362 may be associated with a decrease in the expression of functional elastin, although we did not find this SNP in a search of eQTL databases.
      Table 1Index SNPs for regions associated with striae distensae at a significance level of P<1e–6
      DiscoveryPregnancy
      SNPChr (pos)
      1 Chromosome (chr) and position (pos) are with respect to build 37.
      Gene
      2 Gene is gene that is the most likely candidate for the association or the association or the closest gene. Whether the single-nucleotide polymorphism (SNP) is upstream (u), downstream (d), or within (i) the gene is indicated in parentheses.
      All
      3 Alleles are major/minor in the context of European ancestry.
      MAF
      4 MAF is minor allele frequency in the entire 23andMe European research cohort (over 120,000 individuals).
      r
      2 Gene is gene that is the most likely candidate for the association or the association or the closest gene. Whether the single-nucleotide polymorphism (SNP) is upstream (u), downstream (d), or within (i) the gene is indicated in parentheses.
      5 r2 is the estimated imputation accuracy.
      P
      6 Associations with a P-value <5e-8 are genome-wide significant, and those with P-values between 1e-6 and 5e-8 are defined as suggestive.
      OR (CI)
      7 For the discovery set (which included both men and women), the odds ratio (OR) plus confidence interval (CI) is with respect to the minor allele and represents the risk of developing stretch marks.
      P
      6 Associations with a P-value <5e-8 are genome-wide significant, and those with P-values between 1e-6 and 5e-8 are defined as suggestive.
      β (CI)
      8 For the pregnancy set (which included only women), the β plus CI is with respect to the major allele, with positive numbers representing an increase in the severity of stretch marks. These tests were run only for SNPs reaching genome-wide significance in the discovery set. SNPs marked with an asterisk are typed by our genotyping array.
      rs7787362*7 (73392603)ELN (u)C/T0.4670.9921.8E-230.84 (0.81−0.87)7e−50.072 (0.053−0.091)
      rs35318931X (38009121)SRPX (i)G/A0.0790.9641.1E-130.82 (0.77−0.86)0.0260.067 (0.033−0.102)
      rs107980361 (186052962)HMCN1 (i)G/C0.4850.9896.9E−101.11 (1.08−1.15)0.06−0.029 (−0.048 to −0.010)
      rs75942202 (643320)TMEM18 (d)A/G0.1940.9469.8E−090.88 (0.84−0.92)0.70−0.067 (−0.192 to 0058)
      chr6:363110476 (36311047)PNPLA1 (d)C/T0.0070.9889.7E−081.80 (1.45−2.23)
      rs39105162 (216303053)FN1 (u)G/A0.2620.9092.7E−071.11 (1.07−1.16)
      rs6203432216 (28535834)NPIPL2 (i)G/A0.3590.9464.7E−071.10 (1.06−1.14)
      1 Chromosome (chr) and position (pos) are with respect to build 37.
      2 Gene is gene that is the most likely candidate for the association or the association or the closest gene. Whether the single-nucleotide polymorphism (SNP) is upstream (u), downstream (d), or within (i) the gene is indicated in parentheses.
      3 Alleles are major/minor in the context of European ancestry.
      4 MAF is minor allele frequency in the entire 23andMe European research cohort (over 120,000 individuals).
      5 r2 is the estimated imputation accuracy.
      6 Associations with a P-value <5e-8 are genome-wide significant, and those with P-values between 1e-6 and 5e-8 are defined as suggestive.
      7 For the discovery set (which included both men and women), the odds ratio (OR) plus confidence interval (CI) is with respect to the minor allele and represents the risk of developing stretch marks.
      8 For the pregnancy set (which included only women), the β plus CI is with respect to the major allele, with positive numbers representing an increase in the severity of stretch marks. These tests were run only for SNPs reaching genome-wide significance in the discovery set. SNPs marked with an asterisk are typed by our genotyping array.
      Figure thumbnail gr1
      Figure 1Manhattan plot depicting single-nucleotide polymorphism (SNP) associations with stretch marks. (a) Genome-wide view of associations. SNPs shown in red are genome-wide significant (P<5e-8). Regions are named with the postulated candidate gene. (b) Detailed view of local region around rs7787362. Colors depict the squared correlation (r2) of each SNP with rs7787362, shown in purple. Gray indicates SNPs for which r2 information was missing.
      The second association, rs35318931 (P=1.1e-13, OR=0.82), is a missense variant (serine to phenylalanine) in the SRPX (sushi-repeat containing protein, X-linked) gene. It is associated with striae gravidarum in the pregnancy cohort (P=0.026, β=-0.067). Very little is known about the function of this gene. The third association, rs10798036 (P=6.91e-10, OR=1.11), which did not reach significance in the pregnancy cohort (P=0.06), is located in the HMCN1 (hemicentin-1) gene. Mutations in HMCN1 have been associated with age-related macular degeneration (
      • Schultz D.W.
      • Klein M.L.
      • Humpert A.J.
      • et al.
      Analysis of the ARMD1 locus: evidence that a mutation in HEMICENTIN-1 is associated with age-related macular degeneration in a large family.
      ). As there are no other clear gene candidates for the associations at these loci, it is unclear how these regions might be related to the risk of developing stretch marks.
      The final genome-wide significant association, rs7594220 (P=9.8e-9, OR=0.88), is located 3kb downstream of TMEM18 (transmembrane protein 18), which is involved in neural stem cell migration and cancer but has also been associated with obesity and obesity-related traits (
      • Thorleifsson G.
      • Walters G.B.
      • Gudbjartsson D.F.
      • et al.
      Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity.
      ;
      • Elks C.E.
      • Perry J.R.B.
      • Sulem P.
      • et al.
      Thirty new loci for age at menarche identified by a meta-analysis of genome-wide association studies.
      ). This SNP is in linkage disequilibrium with SNPs previously associated with body mass index (BMI), and correcting for BMI weakens this signal, although not the signals for the other three genome-wide significant hits. To further investigate the association between obesity and stretch marks, we looked more closely at 32 SNPs previously associated with BMI. Even after correction for BMI, one SNP was associated with stretch marks, suggesting a potential effect independent of BMI (Supplementary Table S3 online).
      Three additional regions show suggestive evidence of association with stretch marks (Table 1). Of these, rs3910516 is of particular interest as it lies 2.2kb upstream of FN1, which encodes fibronectin, an extracellular matrix protein that binds to collagen and integrins. Skin biopsies from individuals with stretch marks demonstrate reduced expression of fibronectin (
      • Lee K.S.
      • Rho Y.J.
      • Jang S.I.
      • et al.
      Decreased expression of collagen and fibronectin genes in striae distensae tissue.
      ). The SNP (chr6:36311047) downstream of PNPLA1 is also interesting, as mutations in PLPLA1 are associated with autosomal recessive congenital ichthyosis, a rare skin disease (
      • Grall A.
      • Guaguère E.
      • Planchais S.
      • et al.
      PNPLA1 mutations cause autosomal recessive congenital ichthyosis in golden retriever dogs and humans.
      ). None of the suggestive associations were found in the smaller pregnancy cohort. Although we did not observe a significant difference in the effect of these SNPs between men and women in the discovery cohort, it is still possible that striae gravidarum may have a different etiology.
      Given that loose skin is a symptom of syndromes caused by deletion or loss-of-function mutations in ELN, our results support the hypothesis that variations in the elastic fiber component of the skin extracellular matrix contribute to the development of stretch marks. The expression of collagens, elastin, and fibronectin (which was suggestively associated with stretch marks in our analysis) is decreased in striae, which could be linked to the reorganization and overall loss of elastic fibers in skin affected by striae (
      • Lee K.S.
      • Rho Y.J.
      • Jang S.I.
      • et al.
      Decreased expression of collagen and fibronectin genes in striae distensae tissue.
      ;
      • Watson R.E.
      • Parry E.J.
      • Humphries J.D.
      • et al.
      Fibrillin microfibrils are reduced in skin exhibiting striae distensae.
      ). Other variants associated with elastic tissue have been associated with intracranial aneurysm and exfoliation glaucoma (
      • Akagawa H.
      • Tajima A.
      • Sakamoto Y.
      • et al.
      A haplotype spanning two genes, ELN and LIMK1, decreases their transcripts and confers susceptibility to intracranial aneurysms.
      ;
      • Thorleifsson G.
      • Magnusson K.P.
      • Sulem P.
      • et al.
      Common sequence variants in the LOXL1 gene confer susceptibility to exfoliation glaucoma.
      ). The potential effect of genes associated with obesity, both independent of and via changes in BMI, is also an intriguing area for further study. Replicating this work in a more precisely phenotyped population would be a logical next step.
      None of the existing treatments for stretch marks are completely effective in removing stretch marks. Interestingly, most popular treatments including topical treatments and laser treatments focus on stimulating collagen production, rather than elastin production, to improve the appearance of stretch marks, although some also increase elastic fibers (
      • Elsaie M.L.
      • Baumann L.S.
      • Elsaaiee L.T.
      Striae distensae (stretch marks) and different modalities of therapy: an update.
      ). These findings may provide further insight into future methods for the prevention and treatment of stretch marks.

      ACKNOWLEDGMENTS

      We thank the customers of 23andMe who donated their time and data to this research project. We also thank Joanna Mountain and David Hinds who provided useful comments on the manuscript and all the employees of 23andMe who made this research possible.

      SUPPLEMENTARY MATERIAL

      Supplementary material is linked to the online version of the paper at http://www.nature.com/jid

      REFERENCES

        • Akagawa H.
        • Tajima A.
        • Sakamoto Y.
        • et al.
        A haplotype spanning two genes, ELN and LIMK1, decreases their transcripts and confers susceptibility to intracranial aneurysms.
        Hum Mol Genet. 2006; 15: 1722-1734
        • Atwal G.S.
        • Manku L.K.
        • Griffiths C.E.
        • et al.
        Striae gravidarum in primiparae.
        Br J Dermatol. 2006; 155: 965-969
        • Cho S.
        • Park E.S.
        • Lee D.H.
        • et al.
        Clinical features and risk factors for striae distensae in Korean adolescents.
        J Eur Acad Dermatol Venereol. 2006; 20: 1108-1113
        • Dosal J.
        • Handler M.Z.
        • Ricotti C.A.
        • et al.
        Ulceration of abdominal striae distensae (stretch marks) in a cancer patient—quiz case.
        Arch Dermatol. 2012; 148: 385-390
        • Elks C.E.
        • Perry J.R.B.
        • Sulem P.
        • et al.
        Thirty new loci for age at menarche identified by a meta-analysis of genome-wide association studies.
        Nat Genet. 2010; 42: 1077-1085
        • Elsaie M.L.
        • Baumann L.S.
        • Elsaaiee L.T.
        Striae distensae (stretch marks) and different modalities of therapy: an update.
        Dermatol Surg. 2009; 35: 563-573
        • Grall A.
        • Guaguère E.
        • Planchais S.
        • et al.
        PNPLA1 mutations cause autosomal recessive congenital ichthyosis in golden retriever dogs and humans.
        Nat Genet. 2012; 44: 140-147
        • Lee K.S.
        • Rho Y.J.
        • Jang S.I.
        • et al.
        Decreased expression of collagen and fibronectin genes in striae distensae tissue.
        Clin Exp Dermatol. 1994; 19: 285-288
        • Merla G.
        • Brunetti-Pierri N.
        • Micale L.
        • et al.
        Copy number variants at Williams-Beuren syndrome 7q11.23 region.
        Hum Genet. 2010; 128: 3-26
        • Milewicz D.M.
        • Urbán Z.
        • Boyd C.
        Genetic disorders of the elastic fiber system.
        Matrix Biol. 2000; 19: 471-480
        • Schultz D.W.
        • Klein M.L.
        • Humpert A.J.
        • et al.
        Analysis of the ARMD1 locus: evidence that a mutation in HEMICENTIN-1 is associated with age-related macular degeneration in a large family.
        Hum Mol Genet. 2003; 12: 3315-3323
        • Sheu H.M.
        • Yu H.S.
        • Chang C.H.
        Mast cell degranulation and elastolysis in the early stage of striae distensae.
        J Cutan Pathol. 1991; 18: 410-416
        • Thorleifsson G.
        • Magnusson K.P.
        • Sulem P.
        • et al.
        Common sequence variants in the LOXL1 gene confer susceptibility to exfoliation glaucoma.
        Science. 2007; 317: 1397-1400
        • Thorleifsson G.
        • Walters G.B.
        • Gudbjartsson D.F.
        • et al.
        Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity.
        Nat Genet. 2009; 41: 18-24
        • Watson R.E.
        • Parry E.J.
        • Humphries J.D.
        • et al.
        Fibrillin microfibrils are reduced in skin exhibiting striae distensae.
        Br J Dermatol. 1998; 138: 931-937
        • Zheng P.
        • Lavker R.M.
        • Kligman A.M.
        Anatomy of striae.
        Br J Dermatol. 1985; 112: 185-193