Genetical genomics of Th1 and Th2 immune response in a baboon model of atherosclerosis risk factors

Abstract

Objective

CD4⁺ T-cells mediate inflammation in atherosclerosis, but additive genetic effects on associated pathways of Th1 and Th2 immune response have not been described. We sought to characterize heritability, pleiotropy, and QTL effects on the expression of genes implicated in Th1 and Th2 immune response in a baboon model of risk factors for atherosclerosis.

Methods

We employed a maximum likelihood-based variance decomposition approach to estimate additive genetic effects on transcript levels generated from a gene expression profile of lymphocytes in 499 pedigreed baboons maintained on a basal diet. Transcript levels for 57 genes implicated in Th1 and Th2 immune response were selected for analysis based on significant heritability in this profile. Multipoint whole genome scans were conducted on heritable transcript levels to localize QTLs influencing these measures. To evaluate pleiotropic effects on transcript levels, we estimated genetic and phenotypic correlations among transcript measures, and assessed their correspondence using a Mantel test. Network analysis using GeneGo's MetaCore™ software was conducted to characterize known interaction among coded proteins.

Results

Heritabilities for candidate gene transcript levels ranged from 0.092— 0.786 (median h²=0.278, P=4.72× 10^-4). Linkage analyses yielded significant evidence (LOD≥2.73) for 14 eQTLs (LOD score range 2.76—14.87, genome-wide P=4.9 × 10^-2—1.03 × 10^-14). Estimates of genetic correlation supported shared additive genetic effects incorporating all 57 transcripts (null hypothesis of ρ_G= rejected at FDR≤0.05 for 522 of 1,596 estimates), and accounted for most of the observed phenotypic correlation among transcripts (Mantel test, r_[ρ_{P], [}ρ_G]=0.781, P<0.0001). Network analysis revealed direct interactions among 54 of the 57 coded proteins.

Conclusions

We conclude that major genetic effects influence expression levels of multiple genes implicated in Th1 and Th2 immune response. Additionally, we find that expression levels of these candidate genes are characterized by extensive pleiotropy, consistent with known interaction among their coded proteins, many of which are independently associated with atherosclerosis.

Introduction

Considerable evidence from patient-based studies of atherosclerosis and in experimental models of disease indicates that activated CD4⁺ T cells modulate inflammation at multiple stages of plaque development and progression. In emerging lesions, CD4⁺ T cells home to sites of endothelial disturbance and migrate into the arterial intima, where they are activated by specific antigens. These cells are characterized phenotypically and functionally as effector T-helper cells, and serve to activate other immune cells that may continue to drive inflammation in the lesion (reviewed in ¹). Notably, activated T cells are expanded in complicated human atheromata^2,3 and in the periphery of patients presenting with acute coronary syndrome (ACS)^4-6, findings that implicate these cells in the instability that characterizes plaques prone to rupture, with accompanying thrombosis and myocardial infarction.

Pro-inflammatory T-helper 1 (Th1) effector cells are found in emerging (fatty streak) atherosclerotic lesions⁷, and are key mediators of inflammation in the developing plaque. Th1 cells promote inflammation by proliferating in response to antigens present in the lesion (e.g., modified lipoproteins or bacterial and viral antigens), and secreting cytokines that activate macrophages and other cell types¹. Cytokines associated with a Th1 immune response include the signature Th1 cytokine interferon-gamma (IFNγ), TNF family members TNFα and lymphotoxin (TNFβ), and IL-12, IL-15, and IL-18, which may further stimulate Th1 response by increasing IFNγ expression^1,8-9. The production of IFNγ results in a cascade of pro-inflammatory and atherogenic effects. IFNγ stimulates increased expression of cell adhesion molecules on the endothelial cell surface, increasing the attraction of additional T-cells to the developing lesion¹. IFNγ also upregulates scavenger receptor expression on macrophages, and expression of the MHC Class II molecules HLA-DQ on vascular smooth muscle cells and HLA-DR on macrophages and vascular smooth muscle cells, allowing these cells to present antigen to T-cells¹⁰. Moreover, IFNγ upregulates the expression of multiple co-stimulatory molecules required for continued T-cell activation¹. Consistent with these findings, Th1-associated cytokines are found in murine fatty streak lesions (IFNγ and lymphotoxin)^7,11, and Th1 cell numbers are positively associated with size of murine fatty streak lesions⁷.

Evidence suggests that Th1 effector cells increase in number during the transition from fatty streak to a more complicated atheroma, and act to enhance the susceptibility of the lesion to rupture^1,3,12. Significant numbers of T cells are found in the fibrous cap of atherosclerotic plaques, with a substantial proportion bearing markers of activation^3,13-14. Activated T-cells surrounded by IFNγ are found in human atheromata collected at endarterectomy³, increase in frequency with severity of disease², and are located at the point of rupture in lesions resulting in fatal myocardial infarction¹⁴. These findings are consistent with progressive effects of IFNγ, both direct and indirect, that erode the protective fibrous cap present in the lesion. IFNγ directly inhibits vascular smooth muscle cell proliferation and collagen synthesis in the lesion, leading to thinning and instability of the fibrous cap^1,15. IFNγ also upregulates inflammatory molecules (e.g., reactive oxygen species, proteases, coagulation factors, and multiple other pro-inflammatory cytokines and chemokines) in other cell types that also act to degrade the fibrous cap of the lesion¹.

In agreement with these findings, a dominant Th1 immune response is associated with advanced atherosclerosis in humans and in experimental models of human disease^16-18. Cytokines associated with a Th1 immune response are present in human and murine atheromata (IFNγ, lymphotoxin, IL-12p40, IL-12p70, and IL-18)^{8,11,17,19-20}, and are implicated in disease progression, as evidenced by accelerated atherosclerosis with IL-12 administration in mice²⁰, and association of circulating IL-18 with advanced heart failure in humans²¹. The presence of these cytokines in advanced disease is consistent with the significantly increased numbers of peripheral activated T cells (i.e., CD3⁺CD69⁺, CD3⁺HLA-DR⁺) found in patients with coronary artery disease (CAD), compared to healthy controls⁶. Further, in a finding that underscores the role of Th1 cells in decreasing the stability of advanced lesions, patients presenting with ACS, including unstable angina^4-5 and ST-segment elevation myocardial infarction⁴, exhibited a significant expansion of peripheral Th1 cells (i.e., CD3⁺IFNγ⁺; CD3⁺CD4⁺IFNγ⁺)^4-5 compared to patients with extensive but stable disease, accompanied by corresponding increases in peripheral expression of soluble STAT4 (a Th1-associated transcription factor), IL-2, and IFNγ⁴.

In contrast to the pro-inflammatory Th1 effector cell, the T-helper 2 (Th2) cell has been associated with attenuated inflammation in atherosclerosis. The differentiation of Th2 cells both depends on and results in the further production of IL-4, the signature cytokine of the Th2-mediated response, although Th2 cells also produce IL-5, IL-10, and IL-13²² (reviewed in ²³). Cytokines characteristic of a Th2-mediated response may inhibit production of IFNγ, down-regulate Th1 differentiation, or have other antiinflammatory effects that inhibit Th1 response and atherogenesis^23-24. Evidence consistent with anti-atherogenic effects of Th2 cytokines includes decreased lesion size in mice with administration of IL-4⁷, accelerated atherosclerosis with IL-5 deficiency in mice²⁵, and the association of IL-10 with down-regulation of IFNγ and decreased lesion formation in mice²⁴, and with down-regulation of IL-12 in humans¹⁹. Production of Th2 cytokines in T cell clones from human atheromata is limited, suggesting the relative rarity of Th2 cells in atherosclerotic lesions^16-17.

Although substantial evidence indicates that CD4⁺ effector T cells modulate inflammation in atherosclerosis, the extent of genetic effects influencing these pathways has not been described. In this study, we characterize additive genetic effects on pathways of T-helper cell immune response in the baboon, an important model of atherosclerosis risk factors, a) by reporting significant heritability for transcript levels of 57 candidate genes implicated in multi-cellular pathways of human Th1 and Th2 cell activation, differentiation, and cytokine signaling; b) by localizing quantitative trait loci (QTLs) influencing these candidate gene transcript levels, and c) by assessing the extent of genetic effects shared among transcript levels, and the relationship of these shared genetic effects to observed co-variation among transcripts and coded proteins.

Materials and Methods

Animals

We obtained data for this study from a sample of 499 baboons (Papio hamadryas, 373 females and 126 males) that form part of a large pedigreed and genotyped population used successfully for >30 years to study the interaction of diet and genotype in characterizing risk factors for atherosclerosis. Baboons are maintained outdoors in social groups at the Southwest National Primate Research Center/Southwest Foundation for Biomedical Research (SNPRC/SFBR) in San Antonio, Texas, and range in age from 6 to 34 years, corresponding developmentally to an age range of 18—102 years in humans. Animal care personnel and staff veterinarians provided routine and emergency health care to all animals in accordance with the Guide for the Care and Use of Laboratory Animals. The SFBR facility is certified by the Association for Assessment and Accreditation of Laboratory Animal Care International, and all procedures were approved by the Institutional Animal Care and Use Committee. Baboons sampled in this study had ad libitum access to a low-fat, low cholesterol commercial monkey diet (basal diet, SWF Primate Diet; Harlan Teklad, Madison, WI).

Measures of transcript abundance

Conservation of focal transcript sequences in human and baboon genome

Due to the current draft stage of the baboon genome, a comprehensive and well-annotated whole transcriptome assay for this species has not yet been developed. Based on our own successful use of human short tandem repeats (STRs) to map the baboon genome, as well as our considerable past success using human assays to measure cardiovascular risk factors in this species^S1-S3, we elected to use the Illumina HumanWG-6 Expression BeadChip to measure whole-genome gene expression in this study. To gauge the likely success of using this oligonucleotide-based assay in baboons, we assessed the extent of mRNA sequence conservation between both species for the set of candidate genes investigated in this study. Sequence archived in GenBank (http://www.ncbi.nlm.nih.gov/genbank/) was available for both human and baboon mRNA representing 6 of the 57 genes (CD28, CD4, CD86, CTLA4, IFNG, and IL12B/IL12p40); these sequences were downloaded and aligned using GeneDoc software²⁶. Additionally, for 45 of the 57 genes, we used the “discontiguous megablast” algorithm with the nucleotide BLAST program (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to interrogate Phase 2 and 3 sequences found in the Papio hamadryas high throughput genomic sequence database (HTGS; http://www.ncbi.nlm.nih.gov/HTGS/) with human mRNA sequence downloaded from GenBank.

In all, 48 of the 57 genes and transcripts investigated were analyzed for expressed sequence identity between baboons and humans, revealing an estimated mean identity of 96% (range 83% - 100%), weighted by number of nucleotides examined. These findings (and others described in the RESULTS section) indicate that DNA sequence is highly conserved between both species, and support the idea that transcript levels measured in this study are a reflection of biological reality, rather than a consequence of mismatches in probe sequence.

Data collection and pre-processing

Lymphocytes were extracted from whole blood collected in EDTA. Total RNA was isolated from these cells using a TRIzol (Invitrogen) RNA extraction protocol with a chloroform interphase separation and subsequent isopropanol and ethanol precipitation. Total RNA samples were air dried, resuspended in RNase-free water and stored at -80°C. Seven microliters of isolated total RNA was further treated to remove globin mRNA with the GLOBINclear® kit (Ambion, USA) according to the manufacturer's protocol. Purified total RNA yield (μg) and purity (260nm:280nm) were determined spectrophotometrically using the NanoDrop ND-1000 (Wilmington, DE). A total of 500ng total RNA was dried using an Eppendorf Vacufuge Concentrator 5301 (Eppendorf, Germany) and stored at -20°C prior to anti-sense RNA (aRNA) synthesis. Anti-sense RNA was synthesized, amplified and purified using the Ambion MessageAmp II Amplification Kit (Ambion, USA). Hybridization of aRNA to Illumina's HumanWG-6 v1.0 Expression BeadChips and subsequent washing, blocking and detection were carried out using Illumina's BeadChip 6×2 protocol²⁷. Baboon lymphocyte samples were scanned on the Illumina BeadArray 500GX Reader using Illumina BeadScan image data acquisition software (version 2.3.0.13). Illumina BeadStudio software (version 1.5.0.34) was used for preliminary data analysis after correcting for background signal noise. Illumina Gene Expression BeadChips have internal control features to monitor data quality, and results of these controls were visualized using Illumina GenomeStudio™ software. This software includes routines for calculating and reporting a detection p-value that represents the confidence a given transcript is expressed above the background defined by the negative control probes included on the BeadChips. Other controls include Illumina-supplied oligos randomly imbedded in the expression chip substrate to ensure adequate and specific hybridization of samples. A positive control RNA sample (Stratagene Universal RNA) was also included in each day's run to assess for abnormalities in the aRNA synthesis, amplification and purification procedures.

Each Illumina HumanWG-6 v1 Expression BeadChip contains 47,289 unique 50-mer oligonucleotides, of which 22,151 (47%) target Reference Sequence²⁸ transcripts and the remaining 25,128 probes (53%) target other, generally less well characterized transcripts, including predicted transcripts. To identify transcripts exhibiting sufficient quantitative expression in baboon lymphocytes to support further analyses, we compared the distribution of expression values for a given transcript to the distribution of the expression values of the controls imbedded in each chip. For each transcript, we conducted a χ² “tail” test to determine if there was a significant excess of samples with values above the 95^th percentile of the null distribution based on manufacturer-provided negative control samples. Using a stringent false discovery rate (FDR) of 5%, we identified 15,683 transcripts that exhibited significant expression in baboon lymphocytes by this criterion.

To detect true biological differences between individual baboons, a series of standardization steps was used to make the expression phenotypes comparable across individuals and across transcripts, resulting in normally distributed expression phenotypes. To minimize the influence of overall signal levels, which may reflect RNA quantity and quality rather than true biological differences between individuals, we first standardized abundance values of all retained transcripts by z-scoring within individuals, using decile percentage bins of transcripts grouped by average log-transformed raw signals across individuals. We followed this with linear regression against individual-specific average log-transformed raw signal and its squared value. Lastly, for each transcript, we directly normalized these residual expression scores by employing an inverse Gaussian transformation across individuals, to ensure that the assumption underlying variance components-based analyses is not violated. This very conservative procedure results in normalized expression phenotypes that are comparable between individuals and across transcripts²⁷. Using the same FDR of 5%, 10,719 of these transcripts (68.3%) exhibited significant heritability.

Baboon Pedigree and Whole Genome Linkage Map

Analyses of these traits took advantage of a baboon genome linkage map based on genotype data at nearly 300 microsatellite marker loci (mean inter-marker interval=8.9 cM) from 2,044 baboons placed in a single extended pedigree spanning 6 generations. The physical locations in the human genome for nearly all marker loci in the baboon map are known, facilitating the identification of orthologous chromosomal regions in the two species. Construction of the current baboon linkage map is described in detail elsewhere²⁹, and additional information can be found at the SNPRC website: http://baboon.sfbrgenetics.org/.

Statistical Genetic Methods

Heritability

We conducted all analyses using a maximum likelihood-based variance decomposition approach implemented in the software package SOLAR (Sequential Oligogenic Linkage Analysis Routines). This approach, described in detail elsewhere³⁰ partitions the phenotypic variance in each trait $({\sigma }_p^2)$ into components corresponding to additive genetic effects $({\sigma }_G^2)$, estimated as a function of relatedness among pedigreed baboons, and environmental effects $({\sigma }_E^2)$. After regressing out mean effects of age, age², and sex from standardized measures of transcript abundance, we assessed residuals for departures from multivariate normality³¹, and applied an inverse Gaussian transformation to correct for any such departure. We estimated heritability for these data as the proportion of residual phenotypic variance unexplained by covariates that can be attributed to additive genetic effects $({\text{h}}^2=\frac{{\sigma }_G^2}{{\sigma }_p^2})$.

Pleiotropy

To determine the extent of additive genetic effects shared among transcript levels, we used a bivariate expansion of the variance decomposition model described above to estimate the additive genetic correlation (ρ_G) between all possible trait-pairs (this model additionally provides estimates of environmental correlation (ρ_E) among trait-pairs). We assessed the significance of estimates of ρ_G by means of likelihood ratio tests, which compare the likelihoods of models in which the correlation was estimated to those in which it was constrained to zero (rejection of ρ_G=0 indicates pleiotropy) or to 1 (failure to reject |ρ_G|=1 indicates complete pleiotropy). We corrected for multiple comparisons using the false discovery rate (FDR) of Benjamini and Hochberg³², and considered as significant only those genetic correlations characterized by a FDR ≤ 0.05.

To investigate the extent to which patterns of phenotypic correlation (ρ_P) reflect patterns of genetic correlation between transcript levels for our focal gene set, we first estimated phenotypic correlations among all possible trait-pairs as ${\rho }_P={\rho }_G\sqrt{h_1^2}\sqrt{h_2^2}+{\rho }_E\sqrt{1-h_1^2}\sqrt{1-h_2^2}$, using estimates of trait-specific heritability and genetic and environmental correlations among trait-pairs provided by the bivariate analysis described above. Using the Pearson product moment correlation (r) as an estimate of similarity, we conducted a Mantel test³³ that compared the matrices of these estimated genetic and phenotypic correlations using a routine implemented in XLSTAT (version 2008.6.04; http://www.xlstat.com/). The significance of the estimate of this correlation (for a two-tailed test with α=0.01) was obtained from the distribution of r[ρ_P],[ρ_G] estimated from 100,000 permutations of matrix columns and rows.

Linkage

To assess whether major genetic effects on candidate gene transcript levels could be localized, we conducted univariate multipoint whole genome linkage analyses for each of the focal transcripts. Linkage models based on a variance decomposition approach partition the genetic covariance between relatives into locus-specific and residual genetic effects, using estimates of multipoint identity-by-descent (IBD) allele-sharing. Multipoint IBD allele-sharing among relatives is estimated throughout the baboon linkage map based on genotype data at the microsatellite markers, employing Markov Chain Monte Carlo routines implemented in the computer package Loki³⁴. Based on these estimates of IBD, we tested linkage hypotheses at 1 cM intervals along each chromosome using likelihood ratio tests, and converted the resulting likelihood ratio statistic to the LOD score of classic linkage analysis³⁵. To control for the genome-wide false positive rate, we calculated genome-wide P-values for each LOD score using a modification of a method suggested by Feingold et al.³⁶ that takes into account pedigree complexity and the finite marker density of the linkage map. Accordingly, our threshold for significant evidence of linkage (corresponding to genome-wide α=0.05) was LOD=2.73, while the threshold for suggestive evidence of linkage was LOD=1.50.

Network Analysis

To determine whether relationships among molecules coded by candidate genes were consistent with the patterns of genetic and phenotypic correlation observed in this study, we used the GeneGo MetaCore™ pathways analysis software (v.5.2, Build 17389) to build a network based on curated experimental evidence for relationships among the molecules of interest. Gene symbols for focal transcripts were uploaded as a list to the MetaCore™ portal. Using the direct-interactions network building algorithm, interactions among gene products were diagrammed based on peer-reviewed experimental evidence culled by expert personnel following the GeneGo MetaCore approach (described in detail at http://www.genego.com/metacore.php).

Results

Transcript levels of 57 candidate genes were selected for analysis based on the reported role of the coding gene or gene product in pathways of Th1 and Th2 cell activation and function^S4-S22, and on their significant heritability in this study (h² range = 0.092—0.786, corresponding P-value range = 1.30 × 10^-2—3.42 × 10^-23 (Table 1); transcript levels for an additional 53 genes initially examined were not characterized by significant heritability in this study). Employing a conservative FDR threshold of 0.05, bivariate analyses of all pairwise transcript measures produced 522 significant estimates of additive genetic correlation ranging from |0.41 -1.0| among 1,596 possible tests; significant results included interactions among all 57 transcript measures. Of note, this substantial degree of genetic correlation discovered among transcript levels offers further support for the biological reality of the transcript measures in this study, as the independent occurrence of mismatches in probe sequence between genes would likely preclude correlation among transcripts except that due solely to chance. Following the rule of thumb for interpreting linear correlations, these results indicate that from 17-100% of total additive genetic variance in transcript levels is shared among transcripts. As expected, the magnitude of these shared genetic effects contributed to a close correspondence between genetic and phenotypic correlation among transcript measures, to an extent estimated by the Mantel test of association (r_[ρ_{P], [}ρ_G] = 0.781, P<0.0001), indicating that approximately 60% of the variance in these two matrices is shared.

TABLE 1.

Heritability of transcript levels for 57 genes implicated in pathways of Th1 and Th2 immune response: maximum likelihood estimates and associated P-values.

Transcript	h2	P-value	Transcript	h2	P-value
CCR3	0.138	4.26 × 10-3	IL2RB	0.213	1.20 × 10-4
CD28	0.251	7.50 × 10-6	IL4R	0.243	1.36 × 10-4
CD3D	0.278	4.72 × 10-4	IL10	0.125	7.28 × 10-3
CD3E	0.152	1.66 × 10-2	IL10RB	0.166	3.25 × 10-3
CD3G	0.388	5.73 × 10-11	IL12B/IL12p40	0.120	4.76 × 10-2
CD4	0.303	3.00 × 10-5	IL12RB1	0.392	1.37 × 10-13
CD69	0.223	1.92 × 10-3	IL15	0.092	1.30 × 10-2
CD86	0.247	8.00 × 10-7	IL18	0.179	5.60 × 10-3
CMIP	0.388	1.87 × 10-10	IRF1	0.291	1.00 × 10-7
CSF2RB	0.380	2.88 × 10-10	ITGB2	0.291	1.47 × 10-8
CTLA4	0.623	2.29 × 10-16	KLRD1	0.261	2.80 × 10-6
FYN	0.335	4.80 × 10-9	LAG3	0.214	3.98 × 10-5
GRB2	0.288	4.02 × 10-9	LAT	0.143	7.76 × 10-3
HLA-DMA	0.322	1.00 × 10-7	LTA/TNFB	0.345	3.01 × 10-11
HLA-DMB	0.353	1.60 × 10-6	LTB (isoform b)	0.291	4.00 × 10-7
HLA-DOB	0.384	3.53 × 10-8	NFATC1	0.283	1.00 × 10-7
HLA-DPA1	0.265	8.97 × 10-5	NFATC3	0.239	2.00 × 10-7
HLA-DPB1	0.276	3.50 × 10-4	NFKB1	0.418	4.77 × 10-11
HLA-DQA1	0.786	3.42 × 10-23	PTPN11/SHP2	0.129	5.02 × 10-3
HLA-DQA2	0.306	1.00 × 10-7	PTPN22	0.366	4.50 × 10-9
HLA-DQB1	0.508	9.10 × 10-14	STAT1	0.310	9.64 × 10-9
HLA-DRA	0.432	7.98 × 10-15	STAT6	0.210	6.06 × 10-5
HLA-DRB1	0.340	4.00 × 10-7	TNFRSF5/CD40	0.253	8.05 × 10-5
HLA-DRB3	0.335	8.79 × 10-5	TNFSF6/FASL	0.111	3.07 × 10-2
HLA-DRB4	0.283	2.68 × 10-5	TNFSF11/RANKL	0.206	9.35 × 10-5
HLA-DRB5	0.634	3.29 × 10-19	TRIM	0.228	1.16 × 10-4
ICOS	0.221	2.70 × 10-5	WSX1/IL27RA	0.094	2.89 × 10-2
IFNG	0.282	5.88 × 10-3	ZAP-70	0.221	4.33 × 10-5
IFNGR1	0.105	1.64 × 10-2

Consistent with a close relationship between shared genetic and phenotypic effects on transcript levels, we found that published evidence (as curated by MetaCore personnel) supported direct interactions among virtually all proteins coded by the 57 candidate genes (FIGURE 1); CMIP, LAG3, and NFATC3 were added to this network via interactions with intermediary molecules. The greatest number of interactions were observed for IFNγ (37 edges) and IL-10 (21 edges), in agreement with their role as key pro- and anti-inflammatory cytokines in cross-regulatory Th1 and Th2 immune response, respectively. This network was most significantly associated with autoimmune disease (P=1.52× 10^-34), type 1 diabetes (P=7.39× 10^-31), and rheumatoid arthritis (P=3.08×10^-27).

FIGURE 1.

Relationships among molecules coded by 57 focal genes, organized by cellular compartment. Network analysis was conducted using the direct-interactions building algorithm implemented in MetaCore software. = receptor ligand, = receptor, = G-protein coupled receptor, = binding protein, = complex/group of proteins, = protein kinase, = protein phosphatase, = protein, = transcription factor. Green line = upregulation, red line = down-regulation, blue line = complex or group member.

Table 2 summarizes the results of linkage analyses reaching genome-wide statistical significance, including LOD scores, expression QTL (eQTL) heritability, implicated eQTL regions, and likely mode of regulation. Univariate linkage analyses yielded at least suggestive evidence for linkage (i.e., LOD≥1.50) for 43 of the 57 transcripts (results not shown for eQTLs with 1.50≥LOD≤2.73). Of these 43 eQTLs, 14 were supported by significant evidence for linkage (i.e., LOD≥2.73), with LOD scores ranging from 2.76 to 14.87, with corresponding genome-wide P-values ranging from 4.9× 10^-2 to 1.03 × 10^-14. The eQTL-specific heritabilities (i.e., the proportion of total phenotypic variation exhibited by transcript levels due to additive genetic effects at the eQTL) ranged from 0.228—0.600. For 9 of the 14 eQTLs supported by significant evidence for linkage, the eQTL accounted for all detectable additive genetic variation influencing transcript levels at the cM position corresponding to maximum linkage evidence.

TABLE 2.

Results of linkage analyses describing significant evidence for 14 QTLs influencing expression levels for candidate genes implicated in pathways of Th1 and Th2 immune response. Gene names in bold indicate transcript levels for which the eQTL heritability appears to account for virtually all trait heritability. Abbreviations: eQTL – expression QTL; PHA – Papio hamadryas; HSA – Homo sapiens.

Transcript	LOD	Genome-wide P-value	eQTL heritability	PHA chromosome	HSA 1.5 LOD-drop region	HSA known location	Likely regulation
CSF2RB	2.82	0.04229	0.368	10	22q11.1-22q13.31	22q13.1	Cis
CTLA4	4.79	0.00342	0.406	12	2q32.3-2q36.1	2q33	Cis
FYN	3.46	0.00883	0.331	8	8p11.21-8q22.3	6q21	Trans
HLA-DMA	2.76	0.04898	0.228	18	18p11.22-18q21.32	6p21.3	Trans
HLA-DPA1	3.12	0.02029	0.276	15	9p21.1-9q32	6p21.3	Trans
HLA-DQA1	14.87	1.027 × 10-14	0.600	4	6p22.3-6p12.3	6p21.3	Cis
HLA-DRA	3.11	0.02079	0.314	4	6p12.1-6q24.3	6p21.3	Trans
HLA-DRB1	6.35	7.677 × 10-6	0.382	4	6p22.3-6p12.3	6p21.3	Cis
HLA-DRB3	10.06	9.692 × 10-10	0.399	4	6p22.3-6q12	6p21.3	Cis
HLA-DRB4	5.53	0.00005	0.339	4	6p24.1-6q12	6p21.3	Cis
HLA-DRB5	8.10	1.102 × 10-7	0.475	4	6p22.3-6q11.1	6p21.3	Cis
ITGB2	3.65	0.00555	0.316	10	22q12.3-22q13.2	21q22.3	Trans
NFATC3	3.58	0.00658	0.266	10	22q11.23-22q13.31	16q22.2	Trans
STAT6	4.19	0.00148	0.252	11	12p13.1-12q23.2	12q13	Cis

We examined evidence for likely cis- and trans- effects of the eQTL on each transcript by identifying the region in the baboon genome surrounding the microsatellite markers containing LOD scores ± 1.5 from the maximum LOD score, and locating the corresponding physical region flanked by the same markers in the reference sequence of the human genome. Of the 14 significant eQTLs, 8 localized to regions in the baboon genome orthologous to human genomic regions containing the gene coding for the measured transcript, while 6 eQTLs localized outside such regions.

Discussion

It is clear that for many genes implicated in T-helper immune response, expression levels in lymphocytes vary considerably among baboons, and additive genetic effects contribute to this inter-individual variation in gene expression. This finding is noteworthy because these traits were measured in vivo in healthy, unchallenged animals fed a basal diet consisting of relatively low amounts of fat and cholesterol compared to the average Western diet. These results suggest that genetic variation influencing T-helper effects on inflammation occurs even among healthy animals fed a non-atherogenic, basal diet. We can infer from these results that, to the extent that genetic effects on T-helper immune response influence atherosclerosis, baboons will vary in susceptibility to the development and extent of atherosclerosis. This inference is consistent with the complex nature of the pathology underlying the emergence and progression of atherosclerotic lesions observed in humans.

Based on the corresponding QTL regions inferred in the human genome, the majority of QTLs located in this study contain the gene coding for the focal transcript, a finding consistent with cis-regulation of expression levels for these genes in baboons. Further, for 5 of the 8 apparently cis-regulated QTLs located in this study, the large additive genetic variance estimated at the QTL appears to account for most or all of the total additive genetic effects on levels of the associated transcript. Although there is likely an upward bias in the estimated size of these effects³⁷, these initial results suggest that genetic variation at or near the coding gene exerts significant control over expression levels of many of these candidate genes, a finding in agreement with the evolutionary history of cis-regulation of gene expression in humans³⁸. Interestingly, in agreement with results from a recent genome-wide linkage study of gene expression in human lymphocytes²⁷, among transcript measures in this study with overall heritability estimates ≥0.43, 60% are likely cis-regulated (compared to 50% in humans), and among transcript measures with overall heritability estimated at ≥0.60, 100% (compared to 80% in humans) are likely cis-regulated. This correspondence between findings in baboons and humans provides additional evidence that cis regulation of gene expression is a major determinant of transcript abundance in lymphocytes, and supports a significant degree of genetic similarity between these two species.

The extensive interdependence we observed among candidate gene transcript levels suggests that the similar degree of interaction among coded proteins revealed by network analysis (see Figure 1) may largely reflect shared, heritable variation in the expression of these genes. To assess the relevance of these findings to human atherosclerosis, we searched for published evidence of association in humans or other experimental models for each gene or gene product with lesion characteristics, risk factors for atherosclerosis, or genotypic and clinical correlates of coronary artery disease. We found reports of such associations for 32 of the 57 genes investigated (summarized in Table 3), including findings of effects at the level of the DNA sequence, the gene transcript, and the coded protein. Genetic variants in candidate genes are associated with regulation of blood pressure and susceptibility to hypertension^S23-S24, extent and severity of CAD^S25-S26, carotid artery intima-media thickness^S27, risk of myocardial infarction^S28, or established risk factors for atherosclerosis, including obesity, and CRP and triglyceride levels^S29. Candidate genes and coded proteins are expressed/overexpressed in human and murine atheroma^{8,11,19-20,24,S30-S41}, associated with changes in lesion size and formation^{11,24,S42-S46}, and modulate cytokine levels within lesions^19,S45. Proteins coded by candidate genes are also associated with other pathology implicated in atherosclerosis, including formation of reactive oxygen species^S44, regulation of vascular smooth muscle cell proliferation and vasoconstrictor secretion^S38,S47-S48, and endothelial dysfunction^S31,S49-S52. Candidate gene proteins in both membrane-bound and soluble forms are additionally upregulated in the periphery of patients with CAD^{4,6,21,S37,S52-S58}, particularly in patients presenting with unstable forms of disease (acute coronary syndrome)^4-6. Moreover, it is worth noting that the protein network characterized in this study is significantly associated with autoimmune disease, particularly Type 1 diabetes and rheumatoid arthritis, both diseases featuring a substantially elevated risk of myocardial infarction due to increased atherosclerosis in patients compared to the general population^S59-S60. It appears reasonable to infer from this collective evidence that genetic effects on pathways of Th1 and Th2 immune response influence the development and progression of atherosclerosis.

TABLE 3.

Evidence supporting association of candidate gene expression, protein, or genotypic variation with atherosclerotic lesion characteristics, systemic pathology related to atherosclerosis, risk factors for CAD, or clinical outcomes in coronary heart disease. Abbreviations: CAD – coronary artery disease, ACS – acute coronary syndrome, MI – myocardial infarction, SMC – smooth muscle cells, VSMC – vascular smooth muscle cells, PBMCs – peripheral blood mononuclear cells.

Candidate gene	Published evidence for association of candidate gene or gene product with atherosclerosis (selected references)
CCR3	Overexpressed in human atheromaS30
CD28	Required for T-cell response to plaque antigen(s)S64
CD4	Lineage marker for the majority of T-cells found in human atheroma (reviewed in S65)
CD69	Higher percentages of T-cells with this marker found in patients with ACS and CAD than controls6
CD86	Upregulated in dendritic cells derived from CAD patientsS53
CSF2RB	Codes for receptor subunit detected in cultured SMCs and endothelial and SMCs in human atheromataS31
CTLA4	Increased expression associated with reduced plaque size in HSP60-tolerant LDLr-/- miceS32
FYN	Polymorphisms associated with blood pressureS23
GRB2	Required for oxLDL uptake by macrophages and for lesion formation in miceS42
HLA-DRA	Expression induced in vascular endothelial cells by interferon-γ via reactive oxygen speciesS49
HLA-DRB1	Genotype associated with CAD in ACS patients and in transplantation recipientsS25
ICOS	Expressed abundantly in plaques from ApoE-KO miceS33; increase in atherosclerosis in LDLR-/- mice transplanted with ICOS-deficient marrowS43
IFNG	Implicated in reactive oxygen species formation and lesion development (reviewed in S44); transcript levels are increased in the periphery of ACS patients vs. controls4.
IL4R	Expressed in atherosclerotic plaques in both Ldlr-/- mice and in humansS34
IL10	Inhibits LDL-induced IL-12 release, overexpression decreases lesion size and complexity in LDLr-/- mice (19,24; reviewed in S45)
IL12B/IL12p40	Expressed abundantly in human atherosclerotic plaques19-20; expressed in aortas of apoE-deficient mice20; expression in human endothelial cells attenuated via IL-10S50
IL15	Expressed in human and ApoE-deficient mouse atherosclerotic lesionsS35-S36; co-localized with CD40L-positive T cells in human lesionsS35
IL18	Present in plasma of patients presenting with myocardial ischemiaS54; circulating levels associated with myocardial infarction and plaque areaS55, and severity of heart failure21
IRF1	Protein upregulated by TNFα and IFNγ, stimulating pro-atherogenic endothelin-1 release in human VSMCsS47; mediates apoptosis in VSMCs via PPARγS48
ITGB2	Activation on T cells independently predicts severe CAD and MIS56
LTA/TNFB	Genotype associated with extent of coronary atherosclerosisS26; polymorphism associated with increased intima-media thickness of carotid arteryS27 polymorphisms associated with increased risk of myocardial infarctionS28; present in fatty streak lesions of the aortic sinus, reduced atherosclerosis in LTα -/- mice11
NFATC1	Mediates injury-induced vascular wall remodelingS66
NFATC3	Changes in blood glucose are detected by NFATc3 in arterial smooth muscle, upregulating expression of osteopontin, a promoter of vascular diseaseS67
NFKB1	Implicated in pathways controlling lipid-dependent gene expression and lesion inflammationS68; stimulates upregulation of pro-atherogenic endothelin-1 release from human VSMCsS47; promotor polymorphisms associated with endothelial function in subjects with pre- and stage I hypertensionS24; implicated in inflammatory gene expression and neointima formation following carotid injury in miceS69; increase in NF-κB activation in PBMCs and human carotid lesionsS37
PTPN11/SHP2	Increased expression and SMC proliferation in rat aorta via pro-atherogenic advanced glycation end-product stimulation of NF-κBS38; part of signaling pathway regulating adhesion molecule expression in endothelial cellsS51
PTPN22	Polymorphisms associated with carotid artery intima-media thickness and multiple risk factors for atherosclerosis in a Finnish populationS29
STAT1	Regulates foam cell formation in vitro, and deficiency reduces lesion size in apoe-/- miceS46; stimulates upregulation of pro-atherogenic endothelin-1 from human VSMCsS47
STAT6	Differentially expressed in coronary arteries from patients with ischemic heart disease, upregulated in VSMCs of the same arteriesS39; expression increases in human lesion-derived cells resistant to apoptosisS40
TNFRSF5/CD40	Expressed in human fatty streaks and complicated plaquesS41; upregulated in dendritic cells derived from CAD patientsS53
TNFSF6/FASL	Expression in circulating leukocytes associated with endothelial dysfunction in hyperlipidemic patientsS52; implicated in plaque vulnerability in miceS70; overexpressed in PBMCs and human carotid plaquesS37
TNFSF11/RANKL	Baseline serum levels predict CVD risk over a 15-year period in a human study populationS57; increased T cell expression associated with unstable disease (ACS)S58
ZAP-70	Signal transduction in T-cells via pro-atherogenic lysophosphatidylcholineS71

In this study, we demonstrated that quantitative genetic studies of transcripts implicated in Th1 and Th2 cell activation and function in baboons can yield valuable information about genetic effects on gene expression in associated pathways, including the extent of their interaction, chromosomal location, and likely mode of regulation. Baboons have proven to be valuable models for the study of genetic contributions to known risk factors for human atherosclerosis^S1,S61 due to their substantial genetic and physiological similarity to humans. Many of these physiological similarities are implicated in atherosclerosis, including similarities in cholesterol metabolism^S62, in endothelial cell function^S2, and in features of both fatty streak lesions that may develop under basal conditions, and of the more advanced atherosclerotic lesions that may develop under experimental conditions^S63. These systematic similarities, combined with the ability to control dietary and other environmental variables, are likely to result in increased power to detect genetic effects on traits relevant to human disease. Further genetic studies of potential interaction between T-helper immune response traits and traditional cardiovascular risk factors in baboons may yield valuable insights into the role of adaptive immune response in the development and progression of human atherosclerosis.