Abstract

Genome-wide association (GWA) studies revealed a number of single nucleotide polymorphisms (SNPs) significantly associated with type 1 diabetes (T1D). In an attempt to confirm some of these candidate associations, we genotyped 2046 Caucasian patients and 2417 normal controls from the United States for SNPs in five genomic regions. While no evidence was obtained for four genomic regions (rs2929366/ NM_144715 on chromosome 3, rs9127/ Q7Z4C4 on chromosome 5, rs1445898/ CAPSL on chromosome 5 and rs2302188/ NM_033543 on chromosome 19), we provide strong evidence for association between T1D and multiple SNPs in the IFIH1 linkage disequilibrium (LD) block on chromosome 2q. Among the 10 SNPs genotyped for the 2q region, four SNPs located within the IFIH1 gene or at the 5′ region of IFIH1 showed significant association with T1D in the Georgia population [odds ratio (OR) = 1.7–1.9] with the best P -value found at SNP rs1990760 ( P = 8 × 10 −8 and OR = 1.9). Several SNPs outside of the IFIH1 gene also showed significant but weaker associations. Furthermore, IFIH1 gene expression levels in peripheral blood mononuclear cells are significantly correlated with IFIH1 genotypes, and higher IFIH1 levels are found in individuals with the susceptible genotypes ( P = 0.005). Thus, both genetic association and gene expression data suggest that IFIH1 is the most plausible candidate gene implicated in T1D in this LD block.

INTRODUCTION

Mapping complex disease genes has been a daunting task partly due to the large number of genes involved in each disease and the small effect of each gene. Advances in high throughput and affordable genotyping technologies for large numbers of single nucleotide polymorphisms (SNPs) have made it possible for genome-wide association (GWA) that is revolutionizing studies of complex diseases. This approach has recently been attempted for several diseases including type 1 diabetes (T1D), type 2 diabetes and several other common diseases ( 1–7 ). However, an intrinsic problem with GWA is the expected high false-positive rates due to the large number of markers analyzed by GWA ( 8 ). It is therefore essential that the putative associations identified by GWA are confirmed in multiple large cohorts.

The first GWA for T1D ( 1 ) analyzed 6500 nonsynonymous SNPs. Although it did not cover the entire human genome, this interim analysis revealed, outside of the HLA region, 10 SNPs that showed reasonable to strong association evidence ( P = 3.2 × 10 −4 —7.2 × 10 −17 ). Among these top 10 SNPs, four had poor genotyping quality and were not considered to be reliable ( 1 ) and another one is within the PTPN22 gene that is well known to be associated with T1D and other autoimmune diseases in multiple populations ( 9–16 ). Therefore, we evaluate the five other SNPs that have acceptable genotyping quality in the Caucasian population ascertained from the United States and report here highly significant evidence for association between T1D and genetic markers in the IFIH1 linkage disequilibrium (LD) block that contains four genes: fibroblast activation protein, alpha ( FAP ); interferon induced with helicase C, domain 1 ( IFIH1 ); grancalcin ( GCA ); and potassium voltage-gated channel, subfamily H, member 7 ( KCNH7 ). Furthermore, we evaluated and identified a significant genotype–expression correlation between the best candidate gene ( IFIH1 ) and four SNPs within and at the 3′ end of the IFIH1 gene.

RESULTS

In an attempt to confirm the associations suggested by the first GWA study, we initially genotyped five SNPs in five separate genomic regions. The association between T1D and SNP rs1990760 (exon 15 of IFIH1 ) on chromosome 2 was demonstrated in multiple data sets, whereas the associations with the other four SNPs including rs2929366 ( NM_144715 ) on chromosome 3, rs9127 ( Q7Z4C4) on chromosome 5, rs1445898 ( CAPSL ) on chromosome 5 and rs2302188 ( NM_033543 ) on chromosome 19 have only been reported for the GWA data set. In our study population, there was no evidence for association between T1D and these four SNPs ( Supplementary Material, Table S1 ). No further investigation was carried out for these four regions. Consistent with the strong evidence for association for the rs1990760 ( IFIH1 ) SNP in the initial report ( 1 ), the rs1990760 SNP is significantly associated with T1D in the Georgia data set [odds ratio (OR) = 1.8, P = 8 × 10 −8 ] (Table  1 ).

Table 1.

Associations for SNPs in the IFIH1 LD block in the Georgia and Colorado Caucasian populations

Gene SNP Major allele  Allelic freq.
 
BB versus bb
 
BB + Bb versus bb
 
   T1D Ctrl OR P -value  OR P -value  
Georgia 
 KCNH7 rs12185647 0.457 0.455 1.0 0.9765 1.1 0.4312 
 KCNH7 rs2068330 0.674 0.640 1.4 0.0069 1.3 0.0199 
 Intergenic rs984971 0.673 0.637 1.4 0.0043 1.2 0.0129 
 GCA rs3788964 0.867 0.852 1.4 0.178 1.1 0.1414 
 IFIH1 rs3747517 G 0.763 0.735 1.7 0.0006 1.7 0.00062 
 IFIH1 rs1990760 A 0.667 0.604 1.8 8 × 10−8 1.7 0.000001 
 Intergenic rs2111485 G 0.662 0.606 1.7 4 × 10−6 1.5 0.00007 
 Intergenic rs13422767 G 0.858 0.837 1.9 0.0082 1.9 0.0105 
 FAP rs2075302 0.492 0.457 1.3 0.0045 1.2 0.0099 
 DPP4 rs2052401 0.612 0.583 1.4 0.0011 1.5 0.00005 
Denver 
 KCNH7 rs12185647 0.472 0.458 1.1 0.5332 1.1 0.5893 
 KCNH7 rs2068330 0.680 0.661 1.2 0.2856 1.2 0.2502 
 Intergenic rs984971 0.677 0.642 1.4 0.0669 1.4 0.0881 
 GCA rs3788964 0.844 0.854 0.7 0.4557 0.7 0.4721 
 IFIH1 rs3747517 G 0.736 0.738 1.3 0.3424 1.3 0.2056 
 IFIH1 rs1990760 A 0.630 0.608 1.3 0.1822 1.3 0.1444 
 Intergenic rs2111485 G 0.639 0.611 1.3 0.1760 1.5 0.2216 
 Intergenic rs13422767 G 0.831 0.842 1.0 0.9374 1.1 0.8509 
 FAP rs2075302 0.492 0.471 1.2 0.3601 1.3 0.0552 
 DPP4 rs2052401 0.620 0.572 1.5 0.0143 1.5 0.0124 
Gene SNP Major allele  Allelic freq.
 
BB versus bb
 
BB + Bb versus bb
 
   T1D Ctrl OR P -value  OR P -value  
Georgia 
 KCNH7 rs12185647 0.457 0.455 1.0 0.9765 1.1 0.4312 
 KCNH7 rs2068330 0.674 0.640 1.4 0.0069 1.3 0.0199 
 Intergenic rs984971 0.673 0.637 1.4 0.0043 1.2 0.0129 
 GCA rs3788964 0.867 0.852 1.4 0.178 1.1 0.1414 
 IFIH1 rs3747517 G 0.763 0.735 1.7 0.0006 1.7 0.00062 
 IFIH1 rs1990760 A 0.667 0.604 1.8 8 × 10−8 1.7 0.000001 
 Intergenic rs2111485 G 0.662 0.606 1.7 4 × 10−6 1.5 0.00007 
 Intergenic rs13422767 G 0.858 0.837 1.9 0.0082 1.9 0.0105 
 FAP rs2075302 0.492 0.457 1.3 0.0045 1.2 0.0099 
 DPP4 rs2052401 0.612 0.583 1.4 0.0011 1.5 0.00005 
Denver 
 KCNH7 rs12185647 0.472 0.458 1.1 0.5332 1.1 0.5893 
 KCNH7 rs2068330 0.680 0.661 1.2 0.2856 1.2 0.2502 
 Intergenic rs984971 0.677 0.642 1.4 0.0669 1.4 0.0881 
 GCA rs3788964 0.844 0.854 0.7 0.4557 0.7 0.4721 
 IFIH1 rs3747517 G 0.736 0.738 1.3 0.3424 1.3 0.2056 
 IFIH1 rs1990760 A 0.630 0.608 1.3 0.1822 1.3 0.1444 
 Intergenic rs2111485 G 0.639 0.611 1.3 0.1760 1.5 0.2216 
 Intergenic rs13422767 G 0.831 0.842 1.0 0.9374 1.1 0.8509 
 FAP rs2075302 0.492 0.471 1.2 0.3601 1.3 0.0552 
 DPP4 rs2052401 0.620 0.572 1.5 0.0143 1.5 0.0124 

All SNPs are in Hardy–Weinberg equilibrium except rs2052401 in patient population ( P = 0.003). Odds ratios (OR) were calculated for ‘bb’ (minor allele homozygous genotypes) versus ‘BB’ (major allele homozygous genotypes) as well as ‘bb’ versus heterozygous genotypes ‘Bb’ plus ‘BB’ genotype. The major alleles for SNPs with allelic frequencies near 0.5 were designated in such a way that the ORs are in the same direction as the other SNPs. The allelic frequencies (Freq) are presented for the designated ‘major’ allele of each SNP. Data for the IFIH1 SNPs are bolded.

After obtaining strong evidence for association with the rs1990760 SNP in the Georgia population, we analyzed nine additional SNPs in the IFIH1 genomic region (∼450 kb). These SNPs were selected based on their high degrees of polymorphisms and spacing to cover the entire 450 kb region. On the basis of data for the GA population, 8 of the 10 SNPs (SNP 2–9) have some limited degree of LD and they form two LD sub-blocks (Fig.  1 ). The first LD sub-block includes six SNPs of which four are within or at the 3′ end of the IFIH1 gene, whereas the second LD sub-block contains only two SNPs with strong LD ( R2 = 0.96). The two SNPs in the first LD sub-block (rs1990760 and rs2111485) also have strong LD ( R2 = 0.85). The allelic frequencies for each SNP were shown for both T1D and control subjects (Table  1 ). The ORs were calculated using two commonly used approaches. In the first approach, the two homozygous genotypes for each SNP were compared with each other. These analyses revealed that the major allele homozygous genotypes (BB) are associated with higher T1D risk, while the minor allele homozygous genotypes (bb) are associated with lower T1D risk. Eight of the 10 SNPs showed significant associations (Table  1 ). In the second approach, the minor allele homozygous genotypes (bb) were compared with the major allele homozygous and heterozygous genotypes (BB + Bb). These analyses suggest that the heterozygous genotypes also have high risk than the minor allele homozygous genotypes. Both ORs and the P -values indicate that the four SNPs within IFIH1 or at the 3′ side of IFIH1 and in the middle of the genomic region (rs13422767, rs2111485, rs1990760 and rs3747517) (OR = 1.7–1.9) are SNPs that are more strongly associated with T1D than the remaining SNPs in the region (OR = 1.1–1.4) (Table  1 ). The most significant P -values were obtained for two SNPs within the IFIH1 gene ( P = 8 × 10 −8 and 4 × 10 −6 for rs1990760 and rs2111485, respectively) (Fig.  1 ). These SNPs also showed significant associations when the minor allele homozygotes are compared with the other two genotypes (bb versus BB + Bb) (Table  1 ). Our data therefore provide strong evidence for association between the IFIH1 LD block and T1D in the Georgia population.

Figure 1.

Genetic association between T1D and the IFIH1 genomic region in the Georgia population. Shown are P -values, odds ratios and their 95% confidence interval (CI) estimated for the major allele homozygous genotypes (BB) using minor allele homozygous genotypes (bb) as reference. The whole region appears to be in LD. The R2 values are presented between pairs of SNPs.

Figure 1.

Genetic association between T1D and the IFIH1 genomic region in the Georgia population. Shown are P -values, odds ratios and their 95% confidence interval (CI) estimated for the major allele homozygous genotypes (BB) using minor allele homozygous genotypes (bb) as reference. The whole region appears to be in LD. The R2 values are presented between pairs of SNPs.

We next test whether the associations between SNPs in the IFIH1 LD block is dependent on other covariates such as gender of the subjects, age of onset of diabetes or HLA-DQB1 genotypes using logistic regression. As shown in Table  2 , the IFIH1 SNPs are significantly associated with T1D in both early and late onset T1D data sets with similar OR estimates from logistic regression. Furthermore, no significant interaction between SNP and sex for any of the SNPs in the early onset, late onset or combined data sets (Table  2 ), suggesting that the associations are not dependent on sex of the subjects (ORs are not heterogeneous). To assess the potential effect of HLA on the observed associations, SNP/T1D associations were examined for subject groups with different HLA-DQB1 genotypes. All subjects were classified into two subsets, one with the high-risk DQB1 genotypes (0201/0201, 0302/0302 or 0201/0302) and the second subset with low-risk DQB1 genotypes (all others). These analyses revealed highly significant associations in the low-risk HLA subset, but not in the high-risk HLA subset (Table  3 ). Since the P -values for these tests are heavily influenced by the smaller number of control subjects having the high-risk HLA genotypes, we performed heterogeneity tests to determine whether the ORs significantly differ between the high- and low-risk subsets. Although some tests had P -values near 0.05 (Table  3 ), there is no conclusive evidence that the ORs for the SNPs differ between HLA–DQB1 risk subsets in the GA population.

Table 2.

Associations with SNPs in early onset, late onset and combined cases from Georgia using sex of subjects as a covariate

SNPs  OR
 
P -values
 
 Estimate LCL UCL SNP SNP*sex 
All cases ( n = 1434)  
 rs12185647 1.02 0.83 1.25 0.8617 0.954 
 rs2068330 1.48 1.16 1.88 0.0013 0.568 
 rs984971 1.40 1.11 1.76 0.0042 0.768 
 rs3788964 1.60 0.95 2.69 0.0775 0.417 
 rs3747517 1.78 1.30 2.45 0.0003 0.862 
 rs1990760 1.87 1.49 2.35 6.6E−08 0.694 
 rs2111485 1.68 1.35 2.10 4.3E−06 0.529 
 rs13422767 2.31 1.35 3.95 0.0022 0.482 
 rs2075302 1.33 1.09 1.62 0.0054 0.856 
 rs2052401 1.44 1.16 1.80 0.001 0.104 
Early onset cases ( n = 893)  
 rs12185647 1.04 0.82 1.32 0.7169 0.861 
 rs2068330 1.41 1.06 1.86 0.0163 0.732 
 rs984971 1.33 1.02 1.74 0.0358 0.992 
 rs3788964 1.33 0.75 2.36 0.3347 0.598 
 rs3747517 1.59 1.10 2.29 0.0136 0.377 
 rs1990760 1.84 1.41 2.40 6.6E−06 0.756 
 rs2111485 1.55 1.20 2.00 0.0007 0.637 
 rs13422767 1.99 1.09 3.63 0.0241 0.676 
 rs2075302 1.23 0.98 1.56 0.0774 0.987 
 rs2052401 1.64 1.26 2.13 0.0002 0.087 
Late onset cases ( n = 541)  
 rs12185647 0.98 0.74 1.31 0.9137 0.729 
 rs2068330 1.64 1.16 2.34 0.0057 0.455 
 rs984971 1.56 1.12 2.19 0.0092 0.476 
 rs3788964 – – – – – 
 rs3747517 2.72 1.56 4.75 0.0004 0.120 
 rs1990760 2.01 1.43 2.82 5.3E−05 0.192 
 rs2111485 2.01 1.44 2.82 4.2E−05 0.280 
 rs13422767 – – – – – 
 rs2075302 1.52 1.15 1.99 0.0029 0.776 
 rs2052401 1.22 0.90 1.66 0.1915 0.304 
SNPs  OR
 
P -values
 
 Estimate LCL UCL SNP SNP*sex 
All cases ( n = 1434)  
 rs12185647 1.02 0.83 1.25 0.8617 0.954 
 rs2068330 1.48 1.16 1.88 0.0013 0.568 
 rs984971 1.40 1.11 1.76 0.0042 0.768 
 rs3788964 1.60 0.95 2.69 0.0775 0.417 
 rs3747517 1.78 1.30 2.45 0.0003 0.862 
 rs1990760 1.87 1.49 2.35 6.6E−08 0.694 
 rs2111485 1.68 1.35 2.10 4.3E−06 0.529 
 rs13422767 2.31 1.35 3.95 0.0022 0.482 
 rs2075302 1.33 1.09 1.62 0.0054 0.856 
 rs2052401 1.44 1.16 1.80 0.001 0.104 
Early onset cases ( n = 893)  
 rs12185647 1.04 0.82 1.32 0.7169 0.861 
 rs2068330 1.41 1.06 1.86 0.0163 0.732 
 rs984971 1.33 1.02 1.74 0.0358 0.992 
 rs3788964 1.33 0.75 2.36 0.3347 0.598 
 rs3747517 1.59 1.10 2.29 0.0136 0.377 
 rs1990760 1.84 1.41 2.40 6.6E−06 0.756 
 rs2111485 1.55 1.20 2.00 0.0007 0.637 
 rs13422767 1.99 1.09 3.63 0.0241 0.676 
 rs2075302 1.23 0.98 1.56 0.0774 0.987 
 rs2052401 1.64 1.26 2.13 0.0002 0.087 
Late onset cases ( n = 541)  
 rs12185647 0.98 0.74 1.31 0.9137 0.729 
 rs2068330 1.64 1.16 2.34 0.0057 0.455 
 rs984971 1.56 1.12 2.19 0.0092 0.476 
 rs3788964 – – – – – 
 rs3747517 2.72 1.56 4.75 0.0004 0.120 
 rs1990760 2.01 1.43 2.82 5.3E−05 0.192 
 rs2111485 2.01 1.44 2.82 4.2E−05 0.280 
 rs13422767 – – – – – 
 rs2075302 1.52 1.15 1.99 0.0029 0.776 
 rs2052401 1.22 0.90 1.66 0.1915 0.304 

ORs are estimated using logistic regression with sex as a covariate. P -values for SNPs are for associations between T1D and SNPs. P -values for SNP*sex measure the interaction between sex and SNP association. In the late onset subset, the numbers of subjects are too small for two SNPs (with ‘–’) and the analyses were not carried out. Data for IFIH1 SNPs are bolded.

Table 3.

Associations for SNPs in the IFIH1 region after stratification by HLA DQB1 genotypes

SNP Major allele  Low-risk DQB1
 
High-risk DQB1
 
Het 
  OR P -value  OR P -value  P -value  
Georgia 
 rs12185647 1.1 0.5217 1.0 0.8561 0.6699 
 rs2068330 1.9 0.0001 1.2 0.5748 0.1316 
 rs984971 1.7 0.0006 1.2 0.5422 0.1451 
 rs3788964 1.3 0.372 1.8 0.2607 0.8978 
 rs3747517 G 1.9 0.001 1.5 0.1930 0.4785 
 rs1990760 A 2.1 3 × 10−7 1.4 0.2260 0.0911 
 rs2111485 G 2.0 3 × 10−6 1.3 0.3539 0.0660 
 rs13422767 G 2.4 0.009 1.9 0.1434 0.7216 
 rs2075302 1.5 0.0025 0.8 0.4355 0.0183 
 rs2052401 1.5 0.0032 1.3 0.3518 0.6879 
Denver 
 rs12185647 0.8 0.4484 1.0 0.9700 0.7012 
 rs2068330 1.1 0.6538 1.1 0.7288 0.8183 
 rs984971 1.3 0.2651 1.5 0.1761 0.7881 
 rs3788964 0.8 0.3586 0.8 0.8398 0.7246 
 rs3747517 G 1.2 0.6594 1.5 0.2767 0.6159 
 rs1990760 A 1.1 0.7338 1.5 0.1861 0.4304 
 rs2111485 G 1.1 0.6455 1.5 0.1860 0.4638 
 rs13422767 G 0.8 0.5646 1.6 0.3753 0.2982 
 rs2075302 1.2 0.4581 1.0 0.8833 0.7184 
 rs2052401 1.5 0.0901 1.6 0.1032 0.9014 
SNP Major allele  Low-risk DQB1
 
High-risk DQB1
 
Het 
  OR P -value  OR P -value  P -value  
Georgia 
 rs12185647 1.1 0.5217 1.0 0.8561 0.6699 
 rs2068330 1.9 0.0001 1.2 0.5748 0.1316 
 rs984971 1.7 0.0006 1.2 0.5422 0.1451 
 rs3788964 1.3 0.372 1.8 0.2607 0.8978 
 rs3747517 G 1.9 0.001 1.5 0.1930 0.4785 
 rs1990760 A 2.1 3 × 10−7 1.4 0.2260 0.0911 
 rs2111485 G 2.0 3 × 10−6 1.3 0.3539 0.0660 
 rs13422767 G 2.4 0.009 1.9 0.1434 0.7216 
 rs2075302 1.5 0.0025 0.8 0.4355 0.0183 
 rs2052401 1.5 0.0032 1.3 0.3518 0.6879 
Denver 
 rs12185647 0.8 0.4484 1.0 0.9700 0.7012 
 rs2068330 1.1 0.6538 1.1 0.7288 0.8183 
 rs984971 1.3 0.2651 1.5 0.1761 0.7881 
 rs3788964 0.8 0.3586 0.8 0.8398 0.7246 
 rs3747517 G 1.2 0.6594 1.5 0.2767 0.6159 
 rs1990760 A 1.1 0.7338 1.5 0.1861 0.4304 
 rs2111485 G 1.1 0.6455 1.5 0.1860 0.4638 
 rs13422767 G 0.8 0.5646 1.6 0.3753 0.2982 
 rs2075302 1.2 0.4581 1.0 0.8833 0.7184 
 rs2052401 1.5 0.0901 1.6 0.1032 0.9014 

All subjects in each population were separated into high-risk versus low-risk subsets according to their HLA DQB1 genotypes as discussed in the text. Odds ratios (OR) presented here were calculated for ‘BB’ (major allele homozygous genotypes) versus ‘bb’ (minor allele homozygous genotypes). Het P -values were derived from heterogeneity test using the Breslow–Day test. Data for the IFIH1 SNPs are bolded.

To further assess the difference in association between high- and low-risk HLA subsets, we use a control population that has more subjects with high-risk HLA genotypes. For this purpose, we analyzed a sample set from Colorado that contains an unbiased patient population but a control population that was enriched for HLA high-risk genotypes with the understanding that the HLA bias may have an impact on the association between IFIH1 SNPs and T1D. This data set should allow more reliable estimate of the contribution of the IFIH1 SNPs to T1D risk in the subjects with high risk HLA genotypes. As shown in Table  1 , the SNP frequencies in the entire patient group and control group from the Colorado are similar to those observed in the Georgia data set. Three IFIH1 SNPs (rs3747517, rs1990760 and rs2111485) had similar ORs in the Colorado data set compared with the Georgia data set, although the SNPs did not reach statistical significance (Table  1 ). Logistic regression analyses using sex and age of onset did not find any significant relations and did not significantly improve associations between T1D and the IFIH1 SNPs (data not shown). Stratification by HLA risk revealed that ORs for the IFIH1 SNPs in the high-risk HLA subset is very similar to those observed in the GA data set (Table  3 ). In contrast to the Georgia data set, the IFIH1 SNPs did not show any difference between patients and controls with low-risk HLA genotypes (Table  3 ). This difference between Colorado and Georgia may partially explain why no significant association is observed between IFIH1 SNPs and T1D in the Colorado data set.

As the genetic association data indicated that IFIH1 may be the best candidate gene for this LD block that contains four annotated genes ( FAP , IFIH1 , GCA and KCNH7 ), we further tested this hypothesis using functional studies. We decided to search for potential correlation between IFIH1 genotype and IFIH1 gene expression. Owing to limitation of sample availability, we were only able to examine gene expression in the peripheral blood mononuclear cells (PBMCs) purified from patient and control subjects. Initially, we determined whether these four genes can be detected in PBMCs using real-time RT–PCR and found that the expression of FAP and KCNH7 were below the reliable detection level but IFIH1 and GCA were readily detectable in PBMCs without any stimulation. Therefore, we analyzed the basal expression levels of IFIH1 and GCA in the PBMC of 384 subjects. The samples were analyzed in two different batches, each containing 192 samples. Within each batch, 96 T1D and 96 controls were analyzed. Our primary goal of this experiment was to examine the possible correlation between IFIH1 with a secondary goal to examine potential differences between T1D and control subjects. Using data from both batches, we first used logistic regression to examine the correlation between SNPs and gene expression (both IFIH1 and GCA ) using T1D status as covariate. As no significant interaction was revealed between T1D and SNP/expression correlation, T1D and AbN subjects can be pooled to estimate the differences in IFIH1 or GCA expression between the SNP genotypes. As shown in Fig.  2 , the mean IFIH1 levels in the heterozygous genotypes are between the mean levels in the two homozygous genotypes. Therefore, gene expression differences between SNP genotypes were further examined using a regression model in which the association between gene expression and SNP genotype was additive. The estimated mean expression differences between homozygous genotypes are consistent in the two batches of samples (Fig.  2 ), and the estimates for the combined data set are presented in Table  4 . These analyses indicated that the expression of IFIH1 was significantly higher for the major allele homozygous genotypes (BB), which are associated with higher T1D risk, compared with the minor allele homozygous genotypes (bb), which are associated with lower T1D risk. For example, the mean expression level of IFIH1 was 43% (or 0.57 C T ) higher for subjects with the high-risk GG homozygotes at rs13422767 than subjects with the low-risk homozygous genotype. These results suggest that higher IFIH1 expression may be associated with increased risk for T1D. Indeed, the mean IFIH1 expression was slightly higher in T1D subjects than controls even though the differences did not reach the statistical significance. This potential difference should be further examined in larger data sets. The GCA expression data were analyzed in the same fashion but no significant genotype–expression correlation was found (Table  4 ), further strengthening the argument that IFIH1 may be the best candidate gene. Of course, the lack of gene expression difference alone does not exclude this gene as a candidate gene.

Figure 2.

Correlation between IFIH1 expression level and IFIH1 genotypes. Expression levels of IFIH1 are presented for the three genotypes at each of the four IFIH1 SNPs. The mean expression level was arbitrarily set to 1 for the minor allele homozygous genotype at each SNP.

Figure 2.

Correlation between IFIH1 expression level and IFIH1 genotypes. Expression levels of IFIH1 are presented for the three genotypes at each of the four IFIH1 SNPs. The mean expression level was arbitrarily set to 1 for the minor allele homozygous genotype at each SNP.

Table 4.

Estimated differences of IFIH1 and GCA expression according to IFIH1 genotypes and between T1D and control subjects in human PBMCs

   ΔC T between homozygous genotypes
 
ΔC T between AbN and T1D
 
T1D*SNP 
Genes SNP Estimate LCL UCL P -value   Adj. P Estimate LCL UCL P -value  P -value  
IFIH1 rs13422767 0.571 0.172 0.970 0.005 0.011 0.211 0.006 0.416 0.121 0.743 
IFIH1 rs1990760 0.407 0.120 0.694 0.006 0.011 0.213 0.011 0.414 0.267 0.560 
IFIH1 rs2111485 0.398 0.110 0.687 0.007 0.011 0.198 −0.005 0.400 0.226 0.762 
IFIH1 rs3747517 0.409 0.080 0.739 0.015 0.015 0.195 −0.011 0.400 0.203 0.734 
IFIH1 rs3788964 0.533 0.136 0.930 0.009 0.011 0.207 0.006 0.409 0.089 0.946 
GCA rs13422767 0.279 −0.086 0.645 0.133 0.167 0.120 −0.069 0.310 0.331 0.857 
GCA rs1990760 0.213 −0.047 0.473 0.107 0.167 0.135 −0.050 0.320 0.745 0.292 
GCA rs2111485 0.255 −0.004 0.515 0.054 0.167 0.111 −0.073 0.296 0.736 0.425 
GCA rs3747517 0.170 −0.126 0.467 0.259 0.259 0.115 −0.074 0.303 0.584 0.500 
GCA rs3788964 0.320 −0.044 0.685 0.085 0.167 0.119 −0.066 0.305 0.287 0.958 
   ΔC T between homozygous genotypes
 
ΔC T between AbN and T1D
 
T1D*SNP 
Genes SNP Estimate LCL UCL P -value   Adj. P Estimate LCL UCL P -value  P -value  
IFIH1 rs13422767 0.571 0.172 0.970 0.005 0.011 0.211 0.006 0.416 0.121 0.743 
IFIH1 rs1990760 0.407 0.120 0.694 0.006 0.011 0.213 0.011 0.414 0.267 0.560 
IFIH1 rs2111485 0.398 0.110 0.687 0.007 0.011 0.198 −0.005 0.400 0.226 0.762 
IFIH1 rs3747517 0.409 0.080 0.739 0.015 0.015 0.195 −0.011 0.400 0.203 0.734 
IFIH1 rs3788964 0.533 0.136 0.930 0.009 0.011 0.207 0.006 0.409 0.089 0.946 
GCA rs13422767 0.279 −0.086 0.645 0.133 0.167 0.120 −0.069 0.310 0.331 0.857 
GCA rs1990760 0.213 −0.047 0.473 0.107 0.167 0.135 −0.050 0.320 0.745 0.292 
GCA rs2111485 0.255 −0.004 0.515 0.054 0.167 0.111 −0.073 0.296 0.736 0.425 
GCA rs3747517 0.170 −0.126 0.467 0.259 0.259 0.115 −0.074 0.303 0.584 0.500 
GCA rs3788964 0.320 −0.044 0.685 0.085 0.167 0.119 −0.066 0.305 0.287 0.958 

IFIH1 and GCA expression levels were measured by RT–PCR ( 2 ) in 384 subjects (192 T1D and 192 controls). Data were analyzed using logistic regression. T1D*SNP P -values indicate that there is no interaction between the disease phenotype and the SNPs. Therefore, the correlation between SNP genotype and gene ( IFIH1 or GCA ) expression was assessed using both T1D and control subjects. The expression differences between homozygous genotypes are expressed as delta (Δ) C T (cycle threshold) between C T for bb (minor allele homozygotes) and C T for BB (the major allele homozygotes), e.g. ΔC T = C T for bb − C T for BB. Therefore, a positive ΔC T indicate higher expression level for the major allele homozygotes compared with minor allele homozygotes. For example, the mean IFIH1 expression in subjects with rs13422767 GG homozygotes (major allele) is higher (by 0.57 cycle or 2 0.57 = 1.5-fold) than the minor allele homozygotes. Similarly, positive values for ‘ΔC T between AbN and T1D’ indicates higher expression in T1D subjects. LCL, lower confidence limit; UCL, upper confidence limit. Adj. P , adjusted P -value.

DISCUSSION

Since the initial report on IFIH1 association with T1D, significant association has been reported between IFIH1 and two other autoimmune diseases including Grave’s disease ( 17 ) and multiple sclerosis ( 18 ) but no association was observed in two other studies ( 19 , 20 ). However, confirmatory evidence for IFIH1 association with T1D is still lacking despite an independent report in a family-based study that provided marginal association evidence ( P = 0.03) ( 21 ). In this study, we used a relatively large case/control data set with 10 SNPs in the IFIH1 region and we obtained strong evidence for association between T1D and the IFIH1 genomic region. Therefore, our study provides the urgently needed evidence supporting the IFIH1 LD block as a T1D susceptibility interval.

On the basis of our genetic association data (OR and P -value) in the Georgia population, the two SNPs within the IFIH1 coding region (rs1990760 and rs3747517) and two other SNPs (rs13422767 and rs2111485) in the 3′ intergenic region of IFIH1 (23 and 13 kb from the end of IFIH1 gene) had the best P -value and ORs. These data suggest that IFIH1 may be the best candidate gene. To further substantiate this hypothesis, we took an approach with the hope to reveal potential correlation between IFIH1 expression and IFIH1 genotypes. Using real-time RT–PCR, we analyzed the basal expression levels of IFIH1 in the PBMC and revealed a significant correlation between IFIH1 expression level and genotypes at four SNPs within or near IFIH1 . The expression results were obtained in PBMCs without in vitro stimulation and therefore reflect the basal expression levels. After viral infections, the expression level of IFIH1 is significantly up-regulated. It is possible that the IFIH1 expression differences between IFIH1 genotypes are further increased upon up-regulation by viral infection. This hypothesis should be addressed in future studies.

Since IFIH1 expression level is correlated with genotypes at four SNPs, an interesting question is which SNP(s) might be responsible for the expression difference. This is a difficult question because the SNPs are in LD and we have not identified/analyzed all SNPs in the region. Two of the four SNPs (rs13422767 and rs2111485) are located at the 3′ end of the IFIH1 gene (23 and 13 kb from the end of the 3′-UTR of the IFIH1 gene, respectively). These SNPs do not change any known transcription factor binding sites and it is unknown whether they contribute to the regulation of IFIH1 expression. The two other SNPs (rs1990760 and rs3747517) are both non-synonymous polymorphism within the IFIH1 coding region. The rs1990760 SNP is located within an HNF-3b binding site and rs3747517 is within an AP-1 binding site. These SNPs are located in the coding region that are 45–50 kb from the start codon and it remains to be determined whether these SNPs may play a role in regulating IFIH1 expression.

IFIH1 is a functionally plausible candidate gene for T1D. The innate immune system has developed two pathways for the recognition of dsRNA produced during viral replication or directly from dsRNA viruses. The widely known pathway for dsRNA is TLR3, which allows cells to detect dsRNA that is phacocytosed from the extracellular space and allow the detection of dsRNA viruses that are internalized from the extracellular space through receptor-mediated endocytosis. The second pathway for detecting dsRNA is mediated by cytoplasmic dsRNA sensors (IFIH1 and RIG-I), which allow cells to directly detect intracellular viral infection. IFIH1 and RIG-I are cytoplasmic helicases. IFIH1 or RIG-I signaling activates interferon-regulatory factors (IRFs) and other transcription factors (NFkB) that induce INF-α/β and interferon-inducible genes. The dsRNA recognition pathways have overlapping but non-redundant functions. Some viruses and dsRNA analogs (including polyI:C) may be detected by both TLR3 and cytoplasmic RNA helicase pathways, while others may be sensed primarily through one pathway. The two helicases also discriminate among different dsRNA ligands, even though they share similar signaling features and structural homologies. The signaling pathways have profound impact on the production of the types and amounts of cytokines, surface molecules and DC function. For example, IFIH1−/− mice treated with polyI:C have decreased levels of the proinflammatory cytokine IL-6 and monocytes chemotractant protein 1 (MCP-1), in addition to abrogation of type I IFN production ( 22 , 23 ). One recent report also suggested that signaling through cytoplasmic helicases/IFIH1 resulted in markedly higher capacity of human monocyte-derived DC in stimulating CD4 + T-cell proliferation ( 24 ). These results suggest that signaling through IFIH1 may result in immunogenic dendritic cells that may promote the development of diabetes.

The IFIH1 expression pattern observed in our study is consistent with the potential pathogenic mechanism. Individuals with IFIH1 susceptible genotypes are associated with higher IFIH1 levels while the protective genotypes are associated with lower IFIH1 levels. In individuals with higher IFIH1 levels, infections by viral triggers such as Coxsackie viruses may be primarily recognized by the IFIH1 pathway, leading to enhanced stimulating capacity of dendritic cells and production of proinflammatory cytokines. This hypothesis is highly plausible but requires further functional prove. If our hypothesis is proved true, IFIH1 would provide a molecular link between genetic susceptibility, viral infections and the innate immune response, which are all believed to contribute to T1D pathogenesis.

MATERIALS AND METHODS

Study populations

The main study population consists of Caucasian T1D patients and healthy controls from Georgia. The data set includes 1434 patients and 1865 controls. All diabetic patients were diagnosed using the criteria of the American Diabetes Association. Most of the cases (893) developed type 1 diabetes at or before the age of 17 years and this group of patients were referred to as early onset patients. A good proportion of patients (541) developed diabetes after 17 years of age and these patients were referred to as late onset diabetes for data analysis purposes. The controls are normal subjects without a family history for T1D, and they are ascertained from the same geographic area as the patients. We have attempted to include only those subjects who have genetic heritage of European Caucasians based on the family history of the subjects. The second data set includes 612 non-Hispanic white (NHW) patients with T1D (526 early onset and 86 late onset) seen at the Barbara Davis Center in Denver and 552 NHW healthy control subjects from the same geographic area. The Colorado and Georgia data sets were analyzed separately due to ascertainment difference and potential geographic differences.

SNP genotyping methods

The SNPs were genotyped mainly using the TaqMan-assay method ( 25 ). All assays (primers and probes) used in this study were designed and validated by Applied Biosystems. Amplification reactions were performed in a 5 µl final volume in optical 384-well plates. PCR was carried out with 2 min at 50°C, 10 min at 95°C followed by 40 cycles of 15 s at 95°C and 1 min at 60°C using an ABI9700 real-time PCR system (Applied Biosystems). To validate the TaqMan assays, five SNPs were also genotyped using amplified restriction fragment length polymorphism. The PCR amplification mixture contains 2 µl (20 ng) of genomic DNA, 1.5 µl of 10× PCR buffer, 1.5 µl of dNTP (1.25 mmol/l), 0.75 µl of DMSO, 0.3 µl of each primer (20 pmol/µl) and 0.1 µl of Taq polymerase in a 12-µl reaction volume. Samples were initially denatured for 2 min at 94°C followed by 35 cycles of 94°C for 30 s, 30 s at 58–60°C, and 30 s at 72°C and an additional extension of 2 min at 72°C. Amplified products (12 µl) were digested using 10 units of enzyme. Digested products were electrophoresed on a 2.5–3.5% agarose gel.

Rea-time RT–PCR assay for gene expression

Peripheral blood (2.5 ml for each subject) was immediately preserved in the PaxGene RNA tubes. After 2 h at room temperature, the tubes are frozen at −80°C freezers and total RNA is extracted within a few weeks using a Qiagen kit specially developed for the PaxGene RNA tubes. Extracted RNA was stored at −80°C freezers till use. An aliquot of total RNA (2 µg per sample) were arrayed in 96-well plates and then converted to cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and a PTC-100™ programmable thermal controller (MJ Research, Inc.). The cDNA products are diluted and an aliquot of cDNA equivalent of 10 ng total RNA was used for quantitative real-time PCR performed using ready-to-use TaqMan gene expression assays from the Applied Biosystems. Assays for IFIH1, GCA, FAP and KCNH7 are Hs00223420_m1, Hs 00201854_m1, Hs00990806_m1 and Hs00264454_m1, respectively. The 18S rRNA (Hs 99999901_s1) was used as endogenous control for normalizing RNA concentration. Real-time PCR was performed in 384-well plates with the ABI 7900HT Fast Real-Time PCR System. Standard thermal cycling condition (10 min at 95°C, 40 cycles for 15 s at 95°C, 1 min at 60°C) was used for all genes. All 192 samples were analyzed on the same plate, and each sample was analyzed in duplicate. Cycle threshold (C T ) values for each test gene and 18S rRNA were obtained for each sample using the SDS2.3 and analyzed with RQ Manager 1.2 software. Differences in C T values between a test gene and 18srRNA (ΔC T ) for each sample were calculated and used for statistical analyses.

Statistical analysis for SNP association

Association between each SNP and T1D was assessed by calculating the ORs using two different approaches. In the first approach, the two homozygous genotypes (the minor allele bb homozygotes and major allele BB homozygotes) were compared between T1D and control populations. In the second approach, two of the three genotypes at each SNP locus are collapsed into one group and then compared with the third genotype (e.g. BB + Bb versus bb). P -values were calculated using the χ 2 test. Furthermore, the association between each SNP and phenotype (patient versus control) was tested using logistic regression to fit a linear or additive model, and only the results from the linear models are presented. We also used logistic regression to examine the associations using sex as a covariate. We used the Breslow–Day test to examine heterogeneity in the ORs between low- and high-risk HLA-DQB1 subsets. The early and late onset patients were analyzed separately and jointly to examine the potential influence of age of onset on the SNP associations.

Statistical analysis for correlation between gene expression and SNPs

IFIH1 and GCA expression in PBMCs were assessed in two batches of samples, each containing 192 subjects (96 with T1D and 96 controls). Gene expression differences between SNP genotypes were first examined separately for each of the two batches of samples using a regression model in which the association between gene expression and SNP genotype was additive, which is equivalent to asking whether a trend exists in the association between gene expression and SNP genotype. In such models, the estimated mean for the heterozygote will be equal to the average of the homozygotes. The estimated effects of these models were compared, with the estimates being similar between batches. As such, we used a mixed model analysis of variance in which the additive effect of SNP genotype was one factor and the random effect of batch was the second factor. We also examined the association between gene expression and SNP genotype between T1D subjects and controls using mixed-model analyses of variance. All mixed-model analyses were conducted using the Proc Mixed procedure of SAS, and models without random effects were conducted using the Proc GLM procedure.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online .

FUNDING

This study was partly supported by a grant from the National Institutes of Health (2RO1HD37800, 4R33DK069878, 4R33HD050196) (JXS), P30 DK 57516 (MJR) and Juvenile Diabetes Research Foundation International (JDRF 1-2004-661) (JXS). Purohit S. is recipient of an Advanced Postdoctoral Fellowship from JDRF (10-2006-792).

ACKNOWLEDGEMENTS

We thank a number of physicians (David Brown, N. Spencer Welch, Paul Davidson, Joseph Johnson, David Robertson, Constance Baldwin, Melissa Carlucci, Mark Rappaport, Robert Schultz, Vijayasudha Gunna) and staff members (Kim English, Jessica Leggett, Katherine Guthrie, Melanie Brown, Deana McFeely, Nellie Jenkins, Tami Kinnersley and Ann Simmonds) for their contribution in sample collection. We are very grateful to all patients and other volunteers who participated in this study.

Conflict of Interest statement . None declared.

REFERENCES

1
Smyth
D.J.
Cooper
J.D.
Bailey
R.
Field
S.
Burren
O.
Smink
L.J.
Guja
C.
Ionescu-Tirgoviste
C.
Widmer
B.
Dunger
D.B.
, et al.  . 
A genome-wide association study of nonsynonymous SNPs identifies a type 1 diabetes locus in the interferon-induced helicase (IFIH1) region
Nat. Genet.
 , 
2006
, vol. 
38
 (pg. 
617
-
619
)
2
Sladek
R.
Rocheleau
G.
Rung
J.
Dina
C.
Shen
L.
Serre
D.
Boutin
P.
Vincent
D.
Belisle
A.
Hadjadj
S.
, et al.  . 
A genome-wide association study identifies novel risk loci for type 2 diabetes
Nature
 , 
2007
, vol. 
445
 (pg. 
881
-
885
)
3
Todd
J.A.
Walker
N.M.
Cooper
J.D.
Smyth
D.J.
Downes
K.
Plagnol
V.
Bailey
R.
Nejentsev
S.
Field
S.F.
Payne
F.
, et al.  . 
Robust associations of four new chromosome regions from genome-wide analyses of type 1 diabetes
Nat. Genet.
 , 
2007
, vol. 
39
 (pg. 
857
-
864
)
4
Hinney
A.
Genome wide association (GWA) study for early onset extreme obesity supports the role of fat mass and obesity associated gene (FTO) variants
PLoS One
 , 
2007
, vol. 
2
 pg. 
e1361
 
5
Saxena
R.
Voight
B.F.
Lyssenko
V.
Burtt
N.P.
de Bakker
P.I.W.
Chen
H.
Roix
J.J.
Kathiresan
S.
Hirschhorn
J.N.
, et al.  . 
Diabetes Genetics Initiative of Broad Institute of Harvard and MIT, Lund University, and Novartis Institutes of BioMedical Research
Genome-wide association analysis identifies loci for type 2 diabetes and triglyceride levels
Science
 , 
2007
, vol. 
316
 (pg. 
1331
-
1336
)
6
Hakonarson
H.
Grant
S.F.A.
Bradfield
J.P.
Marchand
L.
Kim
C.E.
Glessner
J.T.
Grabs
R.
Casalunovo
T.
Taback
S.P.
Frackelton
E.C.
, et al.  . 
A genome-wide association study identifies KIAA0350 as a type 1 diabetes gene
Nature
 , 
2007
, vol. 
448
 (pg. 
591
-
594
)
7
The Wellcome Trust Case Control Consortium
Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls
Nature
 , 
2007
, vol. 
447
 (pg. 
661
-
678
)
8
Todd
J.A.
Statistical false positive or true disease pathway?
Nat. Genet.
 , 
2006
, vol. 
38
 (pg. 
731
-
733
)
9
Bottini
N.
Musumeci
L.
Alonso
A.
Rahmouni
S.
Nika
K.
Rostamkhani
M.
MacMurray
J.
Meloni
G.F.
Lucarelli
P.
Pellecchia
M.
, et al.  . 
A functional variant of lymphoid tyrosine phosphatase is associated with type I diabetes
Nat. Genet.
 , 
2004
, vol. 
36
 (pg. 
337
-
338
)
10
Zheng
W.
She
J.X.
Genetic association between a lymphoid tyrosine phosphatase (PTPN22) and type 1 diabetes
Diabetes
 , 
2005
, vol. 
54
 (pg. 
906
-
908
)
11
Begovich
A.B.
Carlton
V.E.
Honigberg
L.A.
Schrodi
S.J.
Chokkalingam
A.P.
Alexander
H.C.
Ardlie
K.G.
Huang
Q.
Smith
A.M.
Spoerke
J.M.
, et al.  . 
A missense single-nucleotide polymorphism in a gene encoding a protein tyrosine phosphatase (PTPN22) is associated with rheumatoid arthritis
Am. J. Hum. Genet.
 , 
2004
, vol. 
75
 (pg. 
330
-
337
)
12
Kyogoku
C.
Langefeld
C.D.
Ortmann
W.A.
Lee
A.
Selby
S.
Carlton
V.E.
Chang
M.
Ramos
P.
Baechler
E.C.
Batliwalla
F.M.
, et al.  . 
Genetic association of the R620W polymorphism of protein tyrosine phosphatase PTPN22 with human SLE
Am. J. Hum. Genet.
 , 
2004
, vol. 
75
 (pg. 
504
-
507
)
13
Smyth
D.
Cooper
J.D.
Collins
J.E.
Heward
J.M.
Franklyn
J.A.
Howson
J.M.
Vella
A.
Nutland
S.
Rance
H.E.
Maier
L.
, et al.  . 
Replication of an association between the lymphoid tyrosine phosphatase locus (LYP/PTPN22) with type 1 diabetes, and evidence for its role as a general autoimmunity locus
Diabetes
 , 
2004
, vol. 
53
 (pg. 
3020
-
3023
)
14
Steck
A.K.
Liu
S.Y.
McFann
K.
Barriga
K.J.
Babu
S.R.
Eisenbarth
G.S.
Rewers
M.J.
She
J.X.
Association of the PTPN22/LYP gene with type 1 diabetes
Pediatr. Diabetes
 , 
2006
, vol. 
7
 (pg. 
274
-
278
)
15
Zoledziewska
M.
Perra
C.
Orru
V.
Moi
L.
Frongia
P.
Congia
M.
Bottini
N.
Cucca
F.
Further evidence of a primary, causal association of the PTPN22 620W variant with type 1 diabetes
Diabetes
 , 
2008
, vol. 
57
 (pg. 
229
-
234
)
16
Onengut-Gumuscu
S.
Buckner
J.H.
Concannon
P.
A haplotype-based analysis of the PTPN22 locus in type 1 diabetes
Diabetes
 , 
2006
, vol. 
55
 (pg. 
2883
-
2889
)
17
Sutherland
A.
Davies
J.
Owen
C.J.
Vaikkakara
S.
Walker
C.
Cheetham
T.D.
James
R.A.
Perros
P.
Donaldson
P.T.
Cordell
H.J.
, et al.  . 
Genomic polymorphism at the interferon-induced helicase (IFIH1) locus contributes to Graves’ disease susceptibility
J. Clin. Endocrinol. Metab.
 , 
2007
, vol. 
92
 (pg. 
3338
-
3341
)
18
Martinez
A.
Santiago
J.L.
Cenit
M.C.
de las Heras
V.
de la Calle
H.
Fernandez-Arquero
M.
Arroyo
R.
de la Concha
E.G.
Urcelay
E.
IFIH1-GCA-KCNH7 locus: influence on multiple sclerosis risk
Eur. J. Hum. Genet
 , 
2008
, vol. 
16
 (pg. 
861
-
864
)
19
Zhao
Z.F.
Cui
B.
Chen
H.Y.
Wang
S.
Li
I.
Gu
X.J.
Qi
L.
Li
X.Y.
Ning
G.
Zhao
Y.J.
The A946T polymorphism in the interferon induced helicase gene does not confer susceptibility to Graves’ disease in Chinese population
Endocrine
 , 
2007
, vol. 
32
 (pg. 
143
-
147
)
20
Marinou
I.
Montgomery
D.
Dickson
M.
Binks
M.
Moore
D.
Bax
D.
Wilson
A.
The interferon induced with helicase domain 1 A946T polymorphism is not associated with rheumatoid arthritis
Arthritis Res. Ther.
 , 
2007
, vol. 
9
 pg. 
R40
 
21
Qu
H.Q.
Marchand
L.
Grabs
R.
Polychronakos
C.
The association between the IFIH1 locus and type 1 diabetes
Diabetologia
 , 
2008
, vol. 
51
 (pg. 
473
-
475
)
22
Gitlin
L.
Barchet
W.
Gilfillan
S.
Cella
M.
Beutler
B.
Flavell
R.A.
Diamond
M.S.
Colonna
M.
Essential role of mda-5 in type I IFN responses to polyriboinosinic:polyribocytidylic acid and encephalomyocarditis picornavirus
Proc. Natl Acad. Sci.
 , 
2006
, vol. 
103
 (pg. 
8459
-
8464
)
23
Loo
Y.M.
Fornek
J.
Crochet
N.
Bajwa
G.
Perwitasari
O.
Martinez-Sobrido
L.
Akira
S.
Gill
M.A.
Garcia-Sastre
A.
Katze
M.G.
Gale
M.
Jr
Distinct RIG-I and MDA5 signaling by RNA viruses in innate immunity
J. Virol.
 , 
2008
, vol. 
82
 (pg. 
335
-
345
)
24
Groschel
S.
Piggott
K.D.
Vaglio
A.
Ma-Krupa
W.
Singh
K.
Goronzy
J.J.
Weyand
C.M.
TLR-mediated induction of negative regulatory ligands on dendritic cells
J. Mol. Med
 , 
2008
25
de Kok
J.B.
Wiegerinck
E.T.
Giesendorf
B.A.
Swinkels
D.W.
Rapid genotyping of single nucleotide polymorphisms using novel minor groove binding DNA oligonucleotides (MGB probes)
Hum. Mutat.
 , 
2002
, vol. 
19
 (pg. 
554
-
559
)

Author notes

These authors contributed equally to this study.