Background
Vitamin D is of interest in relation to a number of health outcomes, with putative function beyond its classical role in maintaining bone health. The active form of vitamin D, 1,25-dihydroxyvitamin D [1,25(OH)
2D], when bound to the vitamin D receptor (VDR), regulates the expression of genes in many molecular pathways, including inflammation, cell proliferation, cell death, and tissue-remodeling pathways [
1]. Serum 25-hydroxyvitamin D [25(OH)D] is the primary circulating biomarker of vitamin D status, and recent national survey data in the U.S. indicate 32% of Americans are at risk of vitamin D inadequacy or deficiency, defined as 30–49 nmol/L and <30 nmol/L serum 25(OH)D, respectively [
2,
3].
Chronic obstructive pulmonary disease (COPD) is the third leading cause of death in the United States, and is a large and growing burden on health care [
4]. While smoking is the primary risk factor for rapid lung function decline and development of COPD, about 15% of individuals who have never smoked develop COPD and not all smokers succumb, implicating other factors, such as genetic, dietary, and lifestyle factors, in lifetime lung function patterns and disease risk [
5].
Recent evidence indicates that vitamin D, as a steroid hormone capable of influencing gene expression, may be a determinant of lung function [
6]. A cross-sectional study in the National Health and Nutrition Examination Survey (NHANES) III reported a strong positive association between serum 25(OH)D and lung function, with clinically relevant effect sizes for forced expiratory volume in the first second (FEV
1) and forced vital capacity (FVC) [
7]. However, a subsequent cross-sectional study in the U.K. reported no association between serum 25(OH)D and FEV
1[
8]. Causal inferences are limited in the cross-sectional design, effect estimates may be biased by uncontrolled confounders such as physical activity, and, furthermore, comparisons are limited by differences in the range in serum 25(OH)D between studies. Investigations of serum 25(OH)D or high-dose vitamin D supplementation in relation to the risk of exacerbations in COPD patients reported overall null findings [
9,
10]. However, vitamin D supplementation led to a statistically significant reduction in COPD exacerbations in the subgroup with severe vitamin D deficiency at the study baseline (serum 25(OH)D < 10 ng/mL) [
9], underscoring the importance of considering the potential to benefit in studies of nutritional supplementation.
In vitro animal and cell culture studies demonstrate that vitamin D-responsive genes play a role in airway remodeling and inflammation, which are key processes in the pathogenesis of COPD [
11,
12]. However, few studies directly investigate mechanisms for vitamin D’s effect
in vivo, which would strengthen the causal inference of population-level association studies. Furthermore, most experimental work to date has focused on effects of the active metabolite of vitamin D, 1,25-dihydroxyvitamin D. This metabolite is generated in the kidney for systemic circulation, and in many tissues, including lung [
13]. It is not yet established whether the population-level range in serum 25-hydroxyvitamin D, the primary biomarker for vitamin D status in humans, is associated with effects similar to those seen
in vitro for 1,25-hydroxyvitamin D.
We used an interdisciplinary approach to investigate the mechanisms through which vitamin D affects lung function. Genes with in vitro evidence of vitamin D regulation were studied to assess whether serum 25(OH)D concentration was associated with gene expression in lung epithelial tissues sampled from free-living humans. Identified genes were investigated in a study of expression quantitative trait loci (eQTL) in human lung epithelial cells to assess if genetic variation affects gene expression. Also, identified genes were investigated in an epidemiologic cohort study in relation to pulmonary function phenotypes. We hypothesized that serum 25(OH)D affects expression of vitamin D-responsive genes by modulating levels of active 1,25(OH)2D in lung tissue, and that variants in candidate genes directly regulated by 1,25(OH)2D in lung tissue are associated with FEV1 and FEV1/FVC, the key parameters used for COPD diagnosis and staging.
Discussion
Using an interdisciplinary genomics approach we investigated vitamin D and lung outcomes. SGPP2, a phosphatase involved in the sphingosine-1-phosphate signaling pathway, was identified in all stages of the study as a promising candidate gene contributing to vitamin D-mediated associations with lung function. SGPP2 is differentially expressed in vivo in lung epithelial cells by serum 25(OH)D. eQTL analysis demonstrates that sequence variants in SGPP2 are associated with lung cell gene expression. Although the eQTL finding does not prove that vitamin D regulation affects gene expression, the location of associated variants in regulatory regions supports the hypothesis of vitamin D regulation. Furthermore, a group of 3 linked SNPs in the SGPP2 promoter region are associated with lower FEV1, a reduced FEV1/FVC ratio, and a 2–3 fold increased risk of airflow obstruction in African-Americans, suggesting that a causal variant in this region may affect SGPP2 function and/or vitamin D binding, and, consequently, lung outcomes. Additionally, a SNP in SGPP2 is associated with FEV1 in Health ABC European-Americans and SGPP2 variants were also associated with FEV1 in the Framingham Heart Study, confirming effects across racial groups and in two cohort studies. This multi-faceted approach identifies putative mechanistic pathways for observed vitamin D—lung function associations while reducing the chance of false positive results.
SGPP2 plays a key role in the sphingolipid signaling pathway through dephosphorylation of sphingosine-1-phosphate (S1P) to sphingosine, which is then converted to ceramide or back to sphingosine-1-phosphate by other enzymes [
25]. Sphingosine-1-phosphate acts as both an intracellular and extracellular signaling molecule, and regulates critical cell processes including apoptosis, cell growth, and immune function [
25,
26]. Altered sphingolipid concentrations have important ramifications for lung function; ceramide concentrations are elevated in COPD, contributing to lung alveolar destruction [
25]. Little research exists on
SGPP2, although a 2006 paper showed that
SGPP2 is up-regulated in response to inflammatory stimuli in endothelial cells, suggesting a possible role in mediating inflammation in lung tissue [
27]. However,
SGPP2’s biological function to alter sphingosine-1-phosphate concentrations suggests that this gene contributes to the regulation of sphingolipid signaling pathways in lung tissue.
We identified several additional genes, namely
DAPK1, KCNS3, and
FSTL1, and all three had mechanistic links to lung function identified through gene ontology analysis and literature reviews (Additional files
11 and
12). Expression of all three genes was strongly associated with serum 25(OH)D, and variants in these genes were associated with pulmonary function in the Health ABC cohort study. However, variants were not replicated in the Framingham Heart Study, nor were there observed eQTL associations.
DAPK1, which is down-regulated by 1,25(OH)
2D both
in vivo and
in vitro, is a pro-apoptotic kinase linked to cytoskeletal remodeling and regulation of inflammatory gene expression in macrophages [
28,
29]. SNPs in
KCNS3, which encodes a voltage-gated potassium channel protein, were associated with airway hyperresponsiveness in past studies [
30], which is of interest given postulated associations of airways hyperresponsiveness with an accelerated rate of FEV
1 decline and risk of COPD [
31].
FSTL1 up-regulates pro-inflammatory cytokines; in mice, the highest expression level is in lung [
32]. Dexamethasone, which is a glucocorticoid used to treat both asthma and COPD, is associated with expression of both
KCNS3 and
FSTL1; interestingly, there are striking similarities in the effects of dexamethasone and 1,25-dihydroxyvitamin on the expression of these genes. The combination of 1,25-dihydroxyvitamin D with dexamethasone was investigated
in vitro as an anti-inflammatory treatment; our results suggest the strong possibility of synergistic effects for this treatment combination (Additional file
12 for references).
A major strength of this study is that it translates
in vitro animal and cell culture studies to an
in vivo study, and then extends to study population-level SNP associations with lung phenotypes, which are partially replicated in an independent cohort. The multi-stage approach identified
SGPP2 as a promising vitamin D-responsive gene for further study. The demonstration of differential gene expression in lung tissue associated with the physiologic range of 25-hydroxyvitamin D in a diverse sample of free-living humans confirms
in vitro studies, and, while our study does not manipulate vitamin D, the
in vivo evidence of association is novel. The Health ABC population-based cohort study included high-quality spirometry, detailed information on confounding factors such as smoking and population stratification, and comprised 40% African-American participants, thus allowing consideration of this understudied population in genomic research. FEV
1 is a predictor of all-cause mortality [
33], and thus SNP—FEV
1 associations are clinically relevant. Although associations between SNPs and the FEV
1/FVC ratio were also investigated, the associations were not as strong as for FEV
1. Thus, vitamin D may have a stronger association with overall lung health versus the risk of COPD. This study identifies plausible biological mechanisms that support a true effect of vitamin D on lung function, and will help to guide the design and analysis of randomized controlled intervention trials of the role of vitamin D in lung disease.
Given that the microarray analysis was used exclusively as a candidate screen, limitations including the lack of qPCR confirmation (not possible due to sample volume limitations), use of nominal P values, and the lack of race-stratified analysis (not possible due to sample size limitations) are less of a concern. As expected, the proportion of participants in the race/ethnicity groups varied by tertile of serum 25(OH)D given the role of skin pigmentation in vitamin D synthesis in response to sunlight [
2]. Race may either confound the serum 25(OH)D—gene expression association, or, race may be a causal antecedent variable that, in part, causes serum 25(OH)D concentration and, in turn, differences in gene expression; adjusting for race may be an over-adjustment. Of note, in regressions adjusted for race the regression coefficients for the serum 25(OH)D—gene expression association were similar to unadjusted analyses.
While the studies were all cross-sectional, which limits causal inference, the harmony of findings across different designs partly mitigates this concern. Although it would have been ideal to use the same samples in all studies (that is, expression, eQTL and SNP—lung function studies), practical limitations led to the use of different samples in each phase. Finally, although gene-level replication was observed for SGPP2 and DAPK1, the specific SNPs associated with FEV1 in Health ABC did not reach statistical significance in FHS. We hypothesize that the SGPP2 SNPs identified in the two cohort studies may be tagging the same unknown causal variant(s) or there may be multiple SGPP2 regulatory regions associated with lung function. Additionally, the strongest SNP—lung function associations in Health ABC were in African-Americans, and, because FHS includes only European Americans, the replication was partial. In summary, SNPs in SGPP2 were statistically significantly associated with lung outcomes after FDR multiple testing adjustment and a highly statistically significant lung eQTL was identified for SGPP2; SGPP2 emerged as a clear candidate in all stages of this work.
Acknowledgements
The authors thank Alex Gileta, who contributed to the eQTL analysis during the time he was an undergraduate senior at Cornell University, and a member of the Cassano Research Group. In addition, the authors thank Yael Strulovici-Barel from Weill Cornell Medical College for uploading the gene expression files to the GEO data repository. Finally, the authors thank the participants of the Health ABC study, for giving of their time, and the Health ABC study team, including the coordinating center at UCSF and at the NIA, for all of the infrastructure and support throughout this project.
Funding
This research was supported by National Institutes of Health, National Heart Lung and Blood Institute R03 HL095414 (PAC) and P50 HL084936 (RGC), by RC1-AG035835 (SK, PI; PAC, PI subcontract), and by NRSA Institutional Research Training Grant T32-DK-7158-36 (JGH). This research also was supported by R01‒AG029364, by NIA contracts N01AG62101, N01AG62103, and N01AG62106, and by the Ashken Foundation (RGC). The genome-wide association study was funded by NIA grant R01-AG032098-01A1 to Wake Forest University Health Sciences and genotyping services were provided by the Center for Inherited Disease Research (CIDR). CIDR is fully funded through a federal contract from the National Institutes of Health to The Johns Hopkins University, contract number HHSN268200782096C. This research was supported in part by the Intramural Research Program of the NIH, National Institute on Aging. Research was conducted in part using data and resources from the Framingham Heart Study of the NHLBI of the NIH and Boston University School of Medicine. The analyses reflect intellectual input and resource development from the Framingham investigators participating in the SNP Health Association Resource (SHARe) project. This work was partially supported by the NHLBI's Framingham Heart Study (Contract No. N01-HC-25195) and its contract with Affymetrix, Inc. for genotyping services (Contract No. N02-HL-6-4278). A portion of this research utilized the Linux Cluster for Genetic Analysis (LinGA-II) funded by the Robert Dawson Evans Endowment of the Department of Medicine at Boston University School of Medicine and Boston Medical Center.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All authors satisfy the requirements for authorship and contributorship. BR, RC and PAC designed and conducted the expression study; YL, KL, SK and TH conducted the Health ABC GWAS study, which provided data for this paper; JGH, BR and PAC designed the Health ABC SNP study and JGH and PAC conducted the SNP study; JGH, PAC, JW and GO'C conducted the replication analysis in FHS, and JGH, PAC, JM and CG conducted the eQTL analysis and interpretation. All coauthors read and edited the final manuscript.