Evaluation of the TREX1 gene in a large multi-ancestral lupus cohort

Abstract

Systemic Lupus Erythematosus (SLE) is a prototypic autoimmune disorder with a complex pathogenesis in which genetic, hormonal and environmental factors play a role. Rare mutations in the TREX1 gene, the major mammalian 3′-5′ exonuclease, have been reported in sporadic SLE cases. Some of these mutations have also been identified in a rare pediatric neurologic condition featuring an inflammatory encephalopathy known as Aicardi-Goutières syndrome (AGS). We sought to investigate the frequency of these mutations in a large multi-ancestral cohort of SLE cases and controls.

Methods

Forty single-nucleotide polymorphisms (SNPs), including both common and rare variants, across the TREX1 gene were evaluated in ∼8370 patients with SLE and ∼7490 control subjects. Stringent quality control procedures were applied and principal components and admixture proportions were calculated to identify outliers for removal from analysis. Population-based case-control association analyses were performed. P values, false discovery rate q values, and odds ratios with 95% confidence intervals were calculated.

Results

The estimated frequency of TREX1 mutations in our lupus cohort was 0.5%. Five heterozygous mutations were detected at the Y305C polymorphism in European lupus cases but none were observed in European controls. Five African cases incurred heterozygous mutations at the E266G polymorphism and, again, none were observed in the African controls. A rare homozygous R114H mutation was identified in one Asian SLE patient whereas all genotypes at this mutation in previous reports for SLE were heterozygous. Analysis of common TREX1 SNPs (MAF >10%) revealed a relatively common risk haplotype in European SLE patients with neurologic manifestations, especially seizures, with a frequency of 58% in lupus cases compared to 45% in normal controls (p=0.0008, OR=1.73, 95% CI=1.25-2.39). Finally, the presence or absence of specific autoantibodies in certain populations produced significant genetic associations. For example, a strong association with anti-nRNP was observed in the European cohort at a coding synonymous variant rs56203834 (p=2.99E-13, OR=5.2, 95% CI=3.18-8.56).

Conclusion

Our data confirm and expand previous reports and provide additional support for the involvement of TREX1 in lupus pathogenesis.

Introduction

Increased expression of interferon regulated genes and disturbance of interferon alpha (IFN-α) homeostasis has a major role in the pathogenesis of the prototypic autoimmune disorder, systemic lupus erythematosus (SLE) (1–3). A perturbation of IFN-α metabolism is also a major pathogenic feature of the inflammatory encephalopathy, Aicardi-Goutières syndrome (AGS) (4). After the discovery of AGS-causing mutations in the TREX1 gene, distinct heterozygous mutations were described in autosomal dominant diseases such as retinal vasculopathy with cerebral leukodystrophy (RVCL) (5), and familial chilblain lupus (FCL) (6). Subsequent studies demonstrate that up to 2% of patients with SLE harbor pathogenic mutations in TREX1 (7). These rare but highly penetrant causative mutations are not detected in genome wide studies and, thereby, may explain part of the missing heritability of lupus as well as provide insights into disease pathogenesis.

The shared TREX1 genetics in clinically distinct human disorders points to a common molecular etiology. Indeed, some AGS individuals also fulfill diagnostic criteria for SLE and have antinuclear antibodies including those with antigenic specificity for ssDNA and dsDNA (8). Likewise, some RVCL patients have autoantibodies to nuclear antigens (unpublished, JPA, PHK).

TREX1 (DNase III) is the major 3′-5′ DNA exonuclease of mammalian cells (9). It has been proposed to have a major role in cell death processes and genomic DNA degradation where it may minimize immune activation by self DNA. It is also a key component of the SET complex, a multitasking protein involved in apoptosis, transcription, nucleosome assembly and histone binding. This complex is normally associated with the endoplasmic reticulum. It is mobilized during the cellular response to oxidative damage and is postulated to participate in the oxidative stress response. A connection between TREX1 and immune activation was initially suggested in the TREX1 null mice that develop an inflammatory myocarditis similar to autoimmune cardiomyopathy and produce type 1 IFN (10, 11). Furthermore, TREX1 deficient cells accumulate single stranded DNA species that may trigger autoimmunity (12).

The TREX1 gene is located on chromosome 3p21.31 and consists of a single exon encoding a 314-amino acid polypeptide. It has three conserved sequence motifs, Exo I, Exo II and Exo III, which form the catalytic site (Fig. 1). TREX1 has a hydrophobic carboxyl-terminal region, predicted to form a transmembrane helix, likely important in defining its intracellular localization (5). In addition, the TREX1 protein contains a proline-rich sequence that is postulated to participate in protein-protein interactions (13) (Fig. 1).

Fig 1.

Schematic representation of some important disease associated TREX1 mutations. Coding synonymous SNP (green); coding non-synonymous SNP (red). Exo 1, 2 and 3 denote exonuclease domains, PII denotes polyproline II domain, and TM denotes transmembrane domain.

Mutations throughout the TREX1 gene have been identified in patients with several different human diseases (13). These mutations include null alleles, frameshift mutations and non-synonymous changes in the catalytic domains and the C-terminal region. In AGS, most TREX1 mutations are autosomal recessive and diminish the exonuclease activity of the enzyme, in particular a transition of arginine to histidine at position 114 (R114H). A homozygous change at this locus (R114H) has been found to have a major effect on exonuclease enzyme activity (14,15). The heterozygous R114H mutation has been reported in one individual with SLE (7). Other mutations include D200N and D18N, which have been reported in AGS and FCL, respectively, and are inherited in an autosomal dominant manner (Fig. 1). In lupus, most of the mutations reported thus far are heterozygotes and are usually located outside the catalytic domain in the C-terminal region (7). The functional significance of these mutations is unknown. C-terminal frameshift mutations that retain exonuclease activity are observed in SLE and account for all of the mutations in RVCL (7,13).

In this study, we evaluated these aforementioned mutations as well as common tagged SNPs in the TREX1 gene in a large, multi-ancestral cohort of SLE cases and controls. This study is the first to investigate TREX1 in populations with a higher prevalence of lupus including African- Americans and Hispanics. Our results confirm and extend the findings in Caucasians and provide additional support for association of SLE with TREX1 in multiple ancestries.

Results

To determine if TREX1 is associated with SLE, we genotyped 40 SNPs in the TREX1 genomic region including both previously reported rare SNPs and more common tag SNPs that capture most of the variation in this region (Table 2). After removing the outliers and correcting for population stratification, 15864 samples (8372 cases and 7492 controls) were included in the analysis (Table 1). All SNPs were in HWE and passed stringent quality control thresholds (see methods).

Table 2.

Selected SNPs in the TREX-1 gene.

SNP	Position	Minor Allele	Major Allele	Average MAF in European	Average MAF in African	Average MAF in Asian	Average MAF in Hispanic
rs922075	48464402	A	G	0.474	0.2975*	0.3558	0.3395
rs6776700	48471762	A	G	0.4616	0.2729	0.3566	0.2828
rs6442123	48475290	A	G	0.469	0.4512	0.3589	0.3109
rs34426134	48480236	A	G	0.0005568	0.04316	0	0.003878
rs62264269	48480342	A	G	0.00008173	0.0003569	0	0
rs7626978	48480835	C	A	0.003968	0.1098	0	0.01517
rs3135935	48480928	A	G	0	0.01331	0	0.0008929
rs2242150	48480968	A	C	0.4676	0.4552	0.3558	0.311
rs3135936	48481222	0	C	0	0	0	0
rs3135938	48481308	A	G	0.0002393	0.0003521	0	0.0005984
rs36041404	48481452	A	G	0.000555	0.04846	0	0.007139
rs12486046	48482186	G	A	0.03295	0.004374	0.0001983	0.02826
rs3135940	48482579	A	G	0.001928	0.0003521	0.0004002	0.0009063
rs3135941	48482671	G	A	0.1787	0.03012	0.04005	0.1778
rs3135942	48482826	0	G	0	0	0	0
D18N	48483110	0	G	0	0	0	0
rs55938060	48483247	A	G	0.001311	0.0001779	0.0002006	0.0003081
rs3135943	48483256	0	0	NA	NA	NA	NA
R114H	48483399	A	G	0.001053	0.0003602	0.0008094	0.0003021
V122A	48483423	0	A	0	0	0	0
rs56203834	48483454	A	G	0.01349	0.001092	0.001856	0.01587
rs3135944	48483520	G	A	0.0001586	0.02172	0	0.00119
A158V	48483531	0	G	0	0	0	0
rs55786737	48483544	A	G	0.008773	0.003002	0.003843	0.007869
R164X	48483548	0	G	0	0	0	0
rs11797	48483589	A	G	0.4499	0.2701	0.3499	0.279
D200N	48483656	0	G	0	0	0	0
V201D	48483660	0	T	0	0	0	0
G227S	48483737	0	0	NA	NA	NA	NA
R240S	48483778	G	C	0.0000794	0.002451	0	0.0008929
A247P	48483797	C	G	0.00007926	0.0014	0	0
E266G	48483855	G	A	0.002778	0.0005248	0	0.0002976
P290L	48483927	0	0	NA	NA	NA	NA
T303P	48483965	0	A	0	0	0	0
rs3135945	48483970	A	G	0.000794	0.1282	0	0.01013
Y305C	48483972	G	A	0.0001587	0	0	0
G306A	48483975	0	C	0	0	0	0
rs3135946	48484040	G	A	0.003578	0.1126	0	0.01492
rs56112131	48484061	A	G	0	0.0001766	0	0
rs3135947	48484132	0	A	0	0	0	0

^*For this SNP, minor allele for the African populations was allele G.

Table 1.

Demographic distribution of individuals in study.

	European-American Case/Control	Asian Case/Control	African American Case/Control	Gullah* Case/Control	Hispanic (others) Case/Control
Total	3936/3491	1265/1260	1527/1811	152/123	1492/807
Male:	344/1151	99/154	121/574	15/18	127/80
Female	3592/2340	1166/1106	1406/1237	137/105	1365/727

^*The Gullah are African-Americans who live in the Low Country of South Carolina and genetically show a much lower admixture rate with non- African populations than other African-Americans.

Observance of rare polymorphisms (MAF<0.01)

Table 3 shows the rare non-synonymous coding SNPs (MAF <0.01) in the TREX1 gene including some detected only in cases. We observed at least three different types of mutations in patients that were not detected in the corresponding controls (Table 3). In Europeans, 5 heterozygous mutations were detected at Y305C in lupus cases but none were observed in the European controls. In African-American and Gullah patients (together), 5 E266G heterozygous mutations were observed but, again, none were observed in the corresponding controls. In the Asians, one homozygous and two heterozygous mutations were observed in lupus cases for the R114H polymorphism, but none in controls. Table 4 describes the ACR criteria fulfilled by these patients and their serologic profiles. Interestingly, two of these 13 SLE cases were males, of which one carried the homozygous R114H mutation and developed SLE at 14 years of age (Table 4).

Table 3.

Frequency of nonsynonymous coding mutations in lupus cases and controls.

SNP*	Position	Minor allele	Major allele	EA Case (AA/AB/BB)	EA Control (AA/AB/BB)	AA Case	AA Control	Gullah Case	Gullah Control	Asian Case	Asian Control
D18N	48483110	0	G	0/0/3921	0/0/3479	0/0/1524	0/0/1801	0/0/152	0/0/123	0/0/1262	0/0/1257
R114H	48483399	A	G	0/9/3856	0/5/3389	0/1/1511	0/1/1734	0/0/146	0/0/120	1/2/1232	0/0/1236
V122A	48483423	0	A	0/0/3906	0/0/3456	0/0/1509	0/0/1800	0/0/152	0/0/123	0/0/1262	0/0/1259
A158V	48483531	0	G	0/0/3932	0/0/3480	0/0/1526	0/0/1800	0/0/152	0/0/123	0/0/1263	0/0/1257
R164X	48483548	0	G	0/0/3928	0/0/3487	0/0/1527	0/0/1807	0/0/151	0/0/123	0/0/1262	0/0/1257
D200N	48483656	0	G	0/0/3920	0/0/3489	0/0/1527	0/0/1807	0/0/151	0/0/123	0/0/1263	0/0/1258
V201D	48483660	0	T	0/0/3935	0/0/3490	0/0/1525	0/0/1809	0/0/152	0/0/123	0/0/1264	0/0/1260
R240S	48483778	G	C	0/1/3919	0/0/3491	0/7/1519	0/11/1798	0/3/149	0/1/122	0/0/1265	0/0/1260
A247P	48483797	C	G	0/0/3936	0/1/3489	0/3/1524	0/7/1803	0/0/152	0/2/121	0/0/1264	0/0/1260
E266G	48483855	G	A	0/22/3909	0/19/3466	0/4/1523	0/0/1811	0/1/151	0/0/123	0/0/1265	0/0/1260
T303P	48483965	0	A	0/0/3918	0/0/3458	0/0/1511	0/0/1789	0/0/151	0/0/123	0/0/1263	0/0/1260
Y305C	48483972	G	A	0/5/3924	0/0/3489	0/0/1527	0/0/1811	0/0/152	0/0/123	0/0/1265	0/0/1260
G306A	48483975	0	C	0/0/3936	0/0/3491	0/0/1527	0/0/1811	0/0/152	0/0/123	0/0/1265	0/0/1260

^*Non-Synonymous SNPs with MAF <1%.

Table 4.

Phenotypic characterization of 13 SLE patients with nonsynonymous mutations.

	SLE Cases
	1	2	3	4	5	6	7	8	9	10	11	12	13
Detected Mutations	R114H (HZ)*	R114H (HTZ)**	R114H (HTZ)	E266G (HTZ)	E266G (HTZ)	E266G (HTZ)	E266G (HTZ)	E266G (HTZ)	Y305C (HTZ)	Y305C (HTZ)	Y305C (HTZ)	Y305C (HTZ)	Y305C (HTZ)
Gender	M	F	F	F	F	F	F	M	F	F	F	F	F
Race	Asian	Asian	Asian	African-American	African-American	African-American	African-American	African-Gullah	European	European	European	European	European
Age of SLE onset	14	18	34	16	36	22	50	43	37	27	N/A	31	37
Malar Rash	+	-	-	-	-	-	-	-	-	+	-	+	+
Discoid Rash	+	-	-	-	-	-	+	-	-	+	-	-	-
Photosensitivity	+	-	-	+	-	-	-	+	+	+	+	+	-
Oral ulcers	-	+	-	+	-	-	-	-	-	+	+	-	-
Arthritis	-	-	+	-	+	-	+	+	+	+	+	+	-
Serositis	-	+	-	-	+	-	+	+	-	+	+	+	+
Nephritis	-	+	+	+	-	+	+	-	-	-	-	-	+
Neurologic	-	-	-	-	+	-	-	-	-	+	-	+	-
Hematologic	-	-	-	+	-	+	-	-	-	+	+	+	+
Immunologic	+	+	+	+	+	+	+	+	+	+	+	+	+
Anti-dsDNA	+	-	+	+	-	+	-	+	+	-	N/A	N/A	N/A
Anti-Sm	-	-	-	-	+	+	-	-	-	-	N/A	N/A	N/A

^*HZ=Homozygous,

^**HTZ=Heterozygous,

Analyses of polymorphisms with MAF >0.01 and haplotype structure

In addition to these rare SNPs mentioned above, we also evaluated common and tag SNPs that capture most of the variation in this region, as well as SNPs with MAF >1%. TREX1 is a small gene with less than 1000 base pairs and one coding exon (Fig 1). Since it is closely linked to the ATRIP (ATR interacting protein) gene, we selected common SNPs to cover both (shown in Table 2). Figures 2a and 2b demonstrate the haplotype structure of this genomic region in European and African cases and controls using SNPs with MAF>1%. There is a high correlation among all of these common SNPs, especially in the European population (r-squared > 0.9). Also, more SNPs are polymorphic (MAF >0.01) in the African-American population than in other racial groups (see Table 2).

Fig 2.

Fig 2a and 2b: European-American (2a) and African-American (2b) haplotype block structure. Blocks connecting SNP pairs are shaded according to the strength of the linkage disequilibrium between the SNPs, from 0.0 (white) to 1.0 (bright red), as measured by the disequilibrium coefficient D′.

Case-control association analysis

In a case control association study, we did not observe significant associations with any of the selected SNPs using the presence of SLE as a phenotype in any of the racial groups. Because of similarities between lupus and neurologic conditions such as AGS, we hypothesized that lupus patients with neurologic manifestations might be enriched for risk alleles in the TREX1 gene. Indeed, in the European population the presence of neurologic manifestations (ACR criteria), especially the presence of seizures (79 European cases), produced significant results at multiple common SNPs when compared with healthy controls (Table 5). The haplotype risk alleles (AAAAAA) (Fig 2a, Table 5) were relatively common in the European population with a frequency of 58% in lupus cases compared to 45% in normal controls (p=0.0008, FDR q=0.007, OR=1.73, 95% CI=1.25-2.39). In addition, in a case only study where these 79 SLE cases with seizure manifestations were compared to 2405 lupus patients with no previous neurologic findings, similar haplotype association results were obtained (p=0.0008, FDR q=0.003, OR=1.75 95% CI=1.25-2.37). Since neurologic manifestations in SLE also correlate with previous cerebrovascular accidents and the presence of antiphospholipid syndrome, we evaluated these sub-groups separately but the results were not significant (data not shown). There was also no evidence of associations with other ACR criteria, gender or age of onset in any population. Of note, all 5 European patients with a mutation at Y305C also had at least one copy of this risk haplotype (2 homozygous and 3 heterozygous patients for the risk allele).

Table 5.

Case control association results in European SLE cases with presence of seizure (79 cases) compared to controls. The bold SNPs represent the common risk haplotype (AAAAAA).

SNP	BP	Minor Allele	Cases	Controls	Major Allele	CHISQ	P	FDR q	OR	L95	U95
rs922075	48464402	A	0.5949	0.4718	G	9.387	0.002185	0.007	1.644	1.192	2.267
rs6776700	48471762	A	0.5921	0.4622	G	10.09	0.001491	0.007	1.689	1.218	2.343
rs6442123	48475290	A	0.6053	0.4674	G	11.34	0.000757	0.004	1.747	1.258	2.427
rs2242150	48480968	A	0.5886	0.4662	C	9.301	0.002291	0.007	1.638	1.189	2.257
rs3135941	48482671	G	0.1429	0.1856	A	1.83	0.1761	0.23	0.7312	0.4637	1.153
rs11797	48483589	A	0.5823	0.4485	G	11.16	0.000835	0.005	1.714	1.245	2.36
rs3135945	48483970	A	0.01266	0.000287	G	42.15	8.46E-11	1.69E-09	44.65	6.25	319

We also evaluated SLE samples for an association with autoantibodies (anti-Ro, anti-La, anti-RNP, anti-Sm and anti-dsDNA). In the Asian population, 567 SLE patients lacking anti-Ro antibodies were less likely to carry the same common haplotype mentioned above (AAAAAA) compared to 1260 healthy controls (32% in cases versus 36% in controls [p=0.01, FDR q=0.03, OR=0.82 95% CI 0.71-0.96]). Interestingly the same haplotype was observed less frequently among 260 Asian patients with positive anti-nRNP compared with 1260 healthy controls (30% in cases with positive anti-nRNP versus 36% in controls [p=0.003, FDR q=0.008, OR=0.75, 95% CI=0.61-0.92]). In a case only study, these comparisons were not statistically significant, although the frequency of this haplotype was more frequent in Asian cases with positive anti-Ro (335 cases) in comparison to those lupus cases with negative anti-Ro (567 cases) (36% vs. 32% (p=0.22)) and for anti-nRNP autoantibody, less frequent in cases with positive anti-nRNP compared to cases with negative anti-RNP (30% for positives (260 cases) versus 35% for negatives (789 cases) (p=0.05, FDR q= 0.12)).

In addition, the presence of anti-nRNP in the European population was strongly associated with another coding synonymous SNP rs56203834 that was extremely rare in the Asian population. In this subgroup, 269 European cases positive for anti-nRNP were compared with European controls. The MAF for this SNP was only 1% in controls (Table 2) but 5% in European cases with positive anti-nRNP (p=2.99E-13, FDR q=5.97E-12, OR=5.2, 95% CI=3.18-8.56). Furthermore, in case only study, if these 269 cases were compared with 1413 SLE patients with negative anti-nRNP again similar results were obtained (p= 2.74E-06, FDR q= 2.73E-05, OR=3.33 95% CI 1.95-5.65). In the Hispanic population suggestive results were also obtained but the number of samples for subset analysis was small.

Discussion

To explore the frequency of mutations in the TREX1 gene and their relationship to SLE, we evaluated 40 SNPs, including rare variants, previously reported to be associated with several human diseases in large multi-ethnic sample populations. The large sample size of cases and controls (>15800) provide enough power to evaluate rare variations in TREX1. One important mutation described in AGS (R114H) is on the dimer interface of the protein (14,15). Homozygous mutations at this position (i.e TREX1 ^R114H/R114H) degrade dsDNA 300-fold less efficiently than TREX1 wild type (16). Although AGS cases are associated with homozygous R114H mutations, their parents, who are heterozygous carriers of the mutation, have no abnormal phenotypes reported (17). A heterozygous mutation at this location has been reported in one European SLE case (7). In our European population, we identified 9 SLE cases with heterozygous mutations at this locus and 5 heterozygotes in the European healthy controls (Table 3). In the Asian population, on the other hand, we identified one homozygous SLE case. In addition, two heterozygous SLE cases were detected in the same population, but none in Asian controls. The homozygous case was a male lupus patient with early onset lupus, positive anti-dsDNA antibody and predominant skin manifestations, but no neurological manifestations. The latter is intriguing since all previously reported homozygous R114H mutations have been in AGS patients, who invariably have had central nervous system disease in early childhood. (Table 4).

We identified 5 European lupus cases with heterozygous mutations at another variant (Y305C) but none in controls. This is a missense (coding) mutation located outside of the catalytic domain of TREX1. This polymorphism was previously reported in one European lupus patient (7). None of these 5 patients were carriers for R114H mutant alleles while all of them were carriers for the common risk haplotype (AAAAAA) (2 homozygotes and 3 heterozygotes). This suggests a correlation of this common risk haplotype with coding mutation at Y305C that could be functionally important.

In the African population, we also identified 5 lupus patients with the E266G mutation, which was absent in African controls (Fig 1). This mutation was detected in Europeans but as previously reported in European cohorts, there was no significant difference in cases and controls (Table 3) (7). This association has not been previously reported in the African population. Overall the coding mutation frequency in our SLE cases was ∼0.5%.

AGS, an autosomal disease (usually recessive but rare dominant case) characterized by progressive encephalopathy of early onset, basal ganglia calcifications and chronic cerebrospinal fluid (CSF) lymphocytosis, is also associated with increased levels of IFNα in the CSF. SLE patients likewise may have high level of IFNα and neurologic manifestations similar to AGS, including seizures. Indeed, neuroimaging in patients with SLE may show calcifications, white matter changes, and atrophy, as typically observed in patients with AGS (18-20). Given these overlapping phenotypes, one may speculate that a common pathogenic mechanism underlies the neurologic phenotype of AGS and cerebral lupus.

In a recent study, up to 60% of patients with AGS presented with clinical findings such as skin rash, arthritis, oral ulcer as well as laboratory findings commonly seen in patients with lupus including ANAs, reduced complement levels, thrombocytopenia, leukocytopenia. Furthermore, if seizures are taken into account, up to 75% of those patients with AGS showed manifestations of lupus (21). Because of these similarities, Vries et al. sequenced genomic DNA of 60 European lupus patients with neurologic manifestations for exonic TREX1 mutations and identified a novel R128H mutation in one of these SLE patients (22). Brain magnetic resonance imaging (MRI) of this patient showed generalized atrophy, extensive symmetric cerebral white matter hyperintensities and cerebellar infarcts, without evidence for ischemia. This rare mutation is located within the highly conserved second exonuclease domain (ExoII) of the TREX1 gene.

In our study we also tested this hypothesis and identified a relatively common risk haplotype in the TREX1 gene among European lupus patients with seizure (58% in SLE cases with seizure compare to 45% in normal controls). The frequency of this haplotype in normal European controls was consistent with the HapMap data for the CEU study population (45%). This haplotype spans 19 kb and covers both TREX1 and ATRIP genes. These two genes are closely linked and some mRNAs encode ATRIP and TREX1 in different reading frames. In fact, these two genes previously were considered a single gene (NCBI-35). ATRIP is an essential component of the DNA damage checkpoint, and binds to ssDNA and interacts with proteins such as ataxia telangiectasia, Rad3 related protein, and breast-ovarian cancer susceptibility 1 (BRCA1) (23). It has central role in checkpoint activation in response to DNA damage and is important for chromosomal stability. Because of the high LD between SNPs in this risk haplotype in Europeans (Fig 2a), conditional analyses were not conclusive; however, analyses on this haplotype suggest that rs11797, a common synonymous SNP located in the exonic region of TREX1, can better explain the whole association in this haplotype and therefore the effect likely originates from the TREX1 gene.

There was also evidence for association of certain SNPs with the presence of autoantibodies in SLE. In particular, there was a positive association with the presence of anti-nRNP in Europeans for SNP rs56203834 (p=2.99E-13, FDR q=5.97E-012, OR = 5.2, 95% CI=3.18-8.56). This is another synonymous SNP located in the TREX1 exon and is less than 60 base pairs away from R114H (Fig. 1). Because of low MAF in this SNP, no clear correlation between this SNP and the common risk haplotype for neurologic manifestation in European can be detected, and in fact all available European cases with seizure manifestations were homozygous for the major wild type allele for this SNP, suggesting that these two effects might be independent. This SNP was extremely rare in the Asian and African populations while in the Hispanic population the result was suggestive.

In regard to the Asian population, the common haplotype that was associated with neurologic manifestations in Europeans was also more frequently seen in Asian patients with this phenotype (42% in cases compare to 38% in controls, p=NS); although it was not significant. As described in the Results, in the Asian population, this haplotype was less frequently observed with the presence of anti-nRNP or in patients with absence of anti-Ro antibodies. In fact this negative correlation with anti-Ro in TREX1 has been previously reported in the same population (24). The reason for this opposite association between anti-Ro and anti-nRNP in Asian with the same haplotype is not clear, and requires additional confirmation. Since SLE is an extremely heterogenous disease, this could be partly related to different disease manifestations. In complex diseases such as SLE, many subtle inherited elements could directly or indirectly affect these autoantibodies with subsequent clinical sequel. Some autoantibodies associated with lupus tend to cluster together and usually result in specific clinical manifestations. For example, the association of anti-Ro with secondary Sjogren's syndrome or leukopenia, anti-RNP with Raynaud's phenomenon, and anti-dsDNA with nephritis have been noted and replicated in many studies. In addition, anti-Ro antibodies have been reported as one of the independent predictors of neurologic damage in lupus (25). This correlation could explain our results in regard to the risk haplotype and anti-Ro antibodies.

Overall, our results with TREX1 indicate a complex relationship between genetic loci, SLE sub-phenotypes and different population ancestry that demands further studies of this gene. Our data, combined with the findings in Trex1-deficient mice which develop lethal autoimmunity accompanied with a high production of type I IFN, suggest that TREX1 is involved in lupus pathogenesis and probably essential for the prevention of autoimmunity.

Materials and Methods

Recruitment and Biological Sample Collection

The participants were enrolled in the Lupus Family Registry and Repository and Lupus Genetics Studies at the Oklahoma Medical Research Foundation (OMRF) as described (26) and by collaborators (27-30). A total of 15864 study participants were used in the current study (Table 1). Protocols were approved by the Institutional Review Boards at each respective institution. Patients met four of the 11 revised 1997 ACR criteria for the classification of SLE (31). Ethnicity was self-reported and verified by principal component and admixture proportion calculations.

Genotyping

This genotyping project was part of the Lupus Large Association Study (LLAS) that involved different investigators and more than 32,000 SNPs. Data were generated using the Illumina iSelect technology at the OMRF. Genotype calls were made using all samples to maximize the accuracy of the cluster plots. Following genotype scoring, SNP clusters were evaluated electronically using the Illumina BEADSTUDIO(r) software package (http://www.illumina.com). Ambiguous SNP clusters were evaluated manually and SNPs with poor cluster characteristics were flagged.

Genotypic data were only used from samples with a call rate >90% (average sample call rate=99.1%) and from SNPs with a call frequency >90% (average SNP call rage=99.0%). Initial QC analyses were performed by plate, by lot of reagents, and by date genotyped to be certain that systematic error did not find its way into our data. A sample report was generated for every sample attempted in the project, including sample barcode, ethnicity, gender, pedigree information, no calls, calls, call rate, genotype frequency and Gencall Score. Any sample with previous genotype data was analyzed for concordance. A summary SNP report was also generated containing chromosome and location, call rate, genotype frequency, and Gencall score.

Statistical Analyses

Testing for association was completed using PLINK (32). Haploview version 4.0 (33) was used to estimate the linkage disequilibrium (LD) between markers and haplotypes in the different racial groups. Conditional haplotype analyses were conducted using the WHAP program version 2.09 (34). To correct for multiple testing, false discovery rate (FDR) methods were used and q values were calculated using PLINK (32). Q values correspond to the proportion of false positives among the results. Thus, q values < 0.05 signify less than a 5% false positive rate and are taken as a measure of significance. For each SNP, missing data proportions for cases and controls, minor allele frequencies, ORs, 95% CI intervals, P values and exact tests for departures from Hardy-Weinberg expectations were calculated. SNPs needed to pass stringent quality control criteria that included: Hardy-Weinberg proportions (HWP) with a p>0.01 in the controls and >0.0001 for cases, total proportion missing <5%, and p>0.05 for differential missingness between cases and controls. Samples with a <90% call rate or increased heterozygosity (>5 std. dev. around the mean) were excluded from the analysis. The remaining samples were then evaluated for duplicates or related individuals. Genetic outliers were removed from further analysis as determined by principal components analysis (35) and admixture proportions calculated using ADMIXMAP. Principal components were calculated using all SNPs and admixture proportions were calculated using 347 AIMs.