Rare SLC13A1 variants associate with intervertebral disc disorder highlighting role of sulfate in disc pathology

Abstract

Back pain is a common and debilitating disorder with largely unknown underlying biology. Here we report a genome-wide association study of back pain using diagnoses assigned in clinical practice; dorsalgia (119,100 cases, 909,847 controls) and intervertebral disc disorder (IDD) (58,854 cases, 922,958 controls). We identify 41 variants at 33 loci. The most significant association (OR_IDD = 0.92, P = 1.6 × 10^-39; OR_dorsalgia = 0.92, P = 7.2 × 10^-15) is with a 3'UTR variant (rs1871452-T) in CHST3, encoding a sulfotransferase enzyme expressed in intervertebral discs. The largest effects on IDD are conferred by rare (MAF = 0.07 - 0.32%) loss-of-function (LoF) variants in SLC13A1, encoding a sodium-sulfate co-transporter (LoF burden OR = 1.44, P = 3.1 × 10^-11); variants that also associate with reduced serum sulfate. Genes implicated by this study are involved in cartilage and bone biology, as well as neurological and inflammatory processes.

Fig. 1. Manhattan plots showing results for meta-analyses of Intervertebral disc disorders (M51), dorsalgia (M54) and lumbar discectomy (LDHsurg).

The P values (−log 10) from meta-analyses of the studied phenotypes are plotted (y-axis) against their respective positions on each chromosome (x-axis). a Intervertebral disc disorders IDD (M51), additive model (four cohorts; 58,854 cases, 922,958 controls), (b) Dorsalgia (M54), additive model (four cohorts, 119,110 cases, 909,847 controls), and (c) severe lumbar IDD defined by surgery (LDHsurg) (three cohorts; 9188 cases, 780,233 controls). P values are two sided and derived from a likelihood-ratio test. The gray and black dots represent SNPs not reaching genome-wide significance threshold weighted for variant impact. The yellow dots represent genome-wide significant SNPs and the red dots represent genome-wide significant SNPs with moderate or high impact (Methods).

Fig. 2. Genes likely to associate with (a) IDD and (b) Dorsalgia.

Sequence variants associated with (a) IDD and (b) Dorsalgia for which functional evidence supports implication of genes in back pain. The variants listed are either protein-coding variants or affect mRNA expression (top cis-eQTL) as depicted by gray boxes (lof loss-of-function) (Supplementary Data 5–8). *Variants also associated in cis with mRNA of other genes (Supplementary Data 7,8). The meta-analyses were performed using logistic regression, the risk (odds ratio OR in yellow) of (a) IDD and (b) Dorsalgia are here shown for the risk-increasing allele and significance in blue. COL11A1 and GSDMC are not included in the figure as evidence for their association with back pain was derived differently as described in results.

Fig. 3. Mendelian randomization (MR) analyses of the genetic relationship between IDD and dorsalgia in terms of causality.

a shows effects of variants associating with Dorsalgia at genome-wide significance, on IDD and dorsalgia. b shows the effects of variants associating with IDD at genome-wide significance, on IDD and dorsalgia. Effects are expressed as logarithms of odds ratios (log(OR)) and black crosses indicate 95% confidence intervals (CI) around effects. To avoid sample overlap, exposure effects are from the cohorts from Iceland (IS), Denmark (DK) and Finland (FIN), while outcome effects are from UK-Biobank. The dashed blue lines show the linear regression fit through the origin, weighting variants according to the square of the standard error of their effect estimates (also known as inverse-variance weighted, IVW)^a)). The IVW-MR method is a multiplicative random effects model, where the test statistic is from a t-distribution, the test is two sided. No multiple comparison adjustments were made. The dashed red lines show the weighted linear regression fit not constrained to go through the origin (also known as MR Egger^b)). For the IDD variants, the slopes of both regression lines are different from zero. For the dorsalgia variants, the slope of the regression line (IVW) through the origin is different from zero, but not the slope of the unconstrained regression line (MR Egger). Further, the effects of the dorsalgia variants deviate substantially more from the regression lines than the IDD variants. These results are not sensitive to outlier removal (Methods, Supplementary Fig. 3).