In a bi-racial population of older adults, we have characterized the epidemiologic and genetic associations of sCD14, as well as relationships of this inflammatory biomarker to incident CVD and mortality. Baseline sCD14 was correlated positively with EA ethnicity, female sex, traditional CVD risk factors (smoking, hypertension, diabetes), and other inflammatory biomarkers (IL-6, CRP, fibrinogen), and negatively correlated with BMI. We identified two genome-wide significant sCD14-associated loci, the CD14 structural locus on chromosome 5q21 and a novel missense variant of PIGC, which encodes an enzyme required for the first step in GPI anchor biosynthesis. We also showed that sCD14 predicts incident CVD and mortality in older adults.
Using a combined GWAS and fine-mapping approach, we identified intra- and inter-population allelic heterogeneity at the CD14
locus, including two new CD14
variants associated with lower sCD14 levels. The significant, non-redundant variants at the CD14
locus explain 3.7% of the trait variability in EA versus 9.0% in AA. One of these variants (haplotype 4 tagged by rs5744455 in ) was associated with lower sCD14 in both EA and AA. rs5744455 may represent a functional variant, as this SNP is located about 50 bp upstream of the TSS within an ENCODE DNaseI hypersensitivity site in CD14+ monocytes and promyelocytic cells. DNaseI hypersensitivity sites are characterized by open chromatin and tend to be transcriptionally active. The other newly discovered sCD14-lowering variant was present only among AA (haplotype 5 tagged by rs5744451 in . Of the SNPs associated with haplotype 5, rs5744430 lies within a 700 bp hepatocyte-specific enhancer region located ~6 kb 5′ of the TSS [9
]. rs5744430 is also located within several putatively functional genomic elements indicative of chromatin accessibility, including an ENCODE DNaseI hypersensitivity site, histone methylation and acylation enrichment sites, and hepatocyte-specific transcription factor binding elements for HNF4A- and FOXA1, as determined by genome-wide ChIP-Seq [22
Two other CD14
variants (haplotypes 2 and 3 in ) were associated with higher sCD14 levels. Haplotype 2 (tagged by rs778584 in our data set) contains the previously characterized -159 T allele of rs2569190, which is located in the proximal promoter [23
]. The CD14
-159 T allele polymorphism increases transcription of CD14
in monocytes by decreasing binding affinity to the inhibitory Sp3 subunit of an Sp transcription factor site [15
] and has been associated with higher sCD14 concentration [14
] as well as greater density of mCD14 in some studies [24
] but not in others [25
Despite the greater burden of CVD risk factors and higher levels of other commonly measured inflammation biomarkers among AA, we found that levels of sCD14 were lower in AA than EA. Whether the ethnic differences in sCD14 are due to environmental and/or genetic factors is uncertain. The CD14
gene locus shows evidence of balancing selection, particularly among European populations [27
], which could be due to its impact on the magnitude of host immune response to local pathogens. It is also noteworthy that the two CD14
alleles tagged by rs778584 and rs4914 associated with higher sCD14 are both more common in EA than AA, while the sCD14-lowering haplotype tagged by rs5744451 is common only among AA. Assuming independence of SNPs, and an additive genetic model, differences in allele frequencies between the identified SNPs accounts for approximately 23% of the inter-population differences in sCD14 levels. Therefore, allele frequency differences at the CD14
locus provide some potential explanation for the higher sCD14 levels observed in EA.
CD14 is one of more than 100 cell-surface proteins anchored to the plasma membrane via GPI. Assembly of GPI-linked proteins occurs in the endoplasmic reticulum and involves more than 30 different genes [28
]. The first step of the biosynthesis, the transfer of N-acetylglucosamine from UDP-N-acetylglucosamine to phosphatidylinositol, is mediated by at least three genes (PIGA
). Somatic mutations in the X-linked gene phosphatidylinositol glycan class A (PIGA
) in hematopoietic stem cells cause paroxysmal nocturnal hemoglobinuria (PNH), which manifests as bone-marrow failure, hemolytic anemia, and thrombosis. As a result of the GPI anchoring defect, PNH patients do not express mCD14 on their monocytes, yet have normal plasma levels of sCD14 [6
]. Notably, the Pro266Ser amino acid substitution encoded by PIGC
rs1063412 is predicted to be functionally deleterious by Condel [29
]. The association of higher sCD14 with a functional coding variant of PIGC
suggests that defective GPI anchor synthesis may be a mechanism that results in increased release of intracellular CD14 into the circulation from myeloid cells. While the association reached genome-wide significance in EA and is certainly biologically plausible, verification of the association between PIGC
rs1063412 and sCD14 in an independent EA population is necessary to authenticate the finding.
Germline mutations in GPI anchor genes have been identified in congenital autosomal recessive disorders that characterized by a diverse range of phenotypes. To our knowledge, this is the first report of a common, inherited variant of a gene involved in GPI anchor biosynthesis associated with a complex trait. A recent GWAS identified a SNP located about 70 kb downstream of PIGC
associated with waist-hip ratio (WHR) [30
]. However, there is no LD between the WHR-associated rs1011731 variant and the sCD14-associated PIGC
coding variant. In the current study, we were unable to detect an association between the PIGC
locus and sCD14 in AA. Even though the variant is more common in AA, our AA sample size was considerably smaller than our EA sample; therefore, power was only 43% to detect an effect of similar magnitude in AA. Other possible explanations for the lack of replication in AA include (1) a false-positive result (type I error) or overestimation of the effect (“winner’s curse”) in EA; (2) effect modification by another genetic or environmental factor that is differentially expressed between EA and AA; (3) rs1063412 may not be true causal variant and the frequency of the true causal variant differs between EA and AA and/or is not well-tagged in AA.
Our results show that sCD14 levels are associated with both subclinical vascular disease (carotid IMT and ABI) and with risk of future clinical CVD in older adults, independently of other risk factors. The reason for the negative association between sCD14 and BMI is unclear, though the inter-relationships between the CD14 system and obesity appear to be complex. sCD14 can both potentiate and down-regulate responses to LPS. For example, adipose tissue CD14 gene expression is increased in obese humans [31
], while mice harboring CD14 mutations have decreased body fat [31
]. In contrast, administration of exogenous recombinant human sCD14 is associated with decreased visceral fat and adipocyte inflammatory gene expression in high-fat-fed mice. Another possible explanation for the paradoxical association with sCD14 observed in CHS is that lower BMI could constitute a marker for frailty and chronic disease in older adults. Therefore, additional studies of sCD14 and BMI involving younger populations and longitudinal data on BMI may be required.
Prior studies of sCD14 and subclinical atherosclerosis have been inconsistent, though most have included primarily younger and middle-aged subjects [33
]. In a smaller prospective European study of healthy middle-aged men, sCD14 levels did not differ between CHD cases and controls [37
]. Despite the association between sCD14 levels and CVD risk in CHS, we found little evidence of association between CD14
genotype and risk of CHD or stroke, in samples totaling over 4,000 CHD cases and 3,000 stroke cases. These results are consistent with those of a recent meta-analysis of previously published studies (11,813 CHD cases and 6,196 controls) which showed no evidence of association between the CD14
C-159T promoter polymorphism and CHD in Europeans or subjects from India, though a possible association between -159 T and CHD was observed among East Asians [38
Several limitations of the current study should be noted. Replication in an independent sample of EA, and also studies including larger numbers of AA, will be required to validate the PIGC association and to determine whether the finding is ethnicity-specific. The remaining phenotypic variability in sCD14 not ascribed to heritable factors or loci identified in our analyses suggests a contribution of combined effects of rare variants not tagged by current genotyping arrays, non-additive genetic factors, as well environmental factors. Therefore, genotyping arrays of even greater density and/or targeted sequencing of the CD14 locus may ultimately be required to identify the full spectrum of trait-associated variants. Finally, further study of the inverse relationship of sCD14 with body mass, as well as generalization of these association findings to younger adults is needed.
In population studies of older adults, serum levels of CRP and IL-6 predict adverse health outcomes and mortality. Though sCD14 is an acute phase protein, and correlates with CRP and IL-6, we found that sCD14 and IL-6 were strong and independent risk factors for mortality in older adults. Accounting for IL-6 and sCD14, there was no association of CRP or fibrinogen with mortality. sCD14 levels are increased in a number of acute and chronic inflammatory conditions and higher levels have been associated with disease severity [39
], including mortality in Gram-negative sepsis [40
]. Conversely, in mouse models, CD14 deficiency has been associated with resistance to LPS-induced shock [41
], while mice harboring TLR4 and CD14 mutations have increased bone mineral density, decreased body fat, greater insulin sensitivity, live long lives and show no evidence of infection [32
]. Increasing evidence suggests that pro-inflammatory mediators play a direct role in aging-related chronic diseases such as atherosclerosis, diabetes and cancer. Whether higher levels of sCD14 or other inflammatory mediators are directly causal, or indicative of chronic, low-level infection and other immune alterations in older adults, or simply reflect worsening subclinical and clinical disease, is difficult to distinguish. Nonetheless, sCD14 may have potential clinical utility as a marker of aging, disease risk, or disease progression in older adults.