General heart pathology of A/JHD mutant mice
Whole-mount images of hearts from an A/JHD litter are presented in Fig. 1. No indications of HD were observed in three siblings (Fig. 1A-C), whereas signs of HD were noted in three of their littermates, including one heart of approximately normal size with significant atrial and ventricular fibrosis (Fig. 1D) and two with extensive right ventricular dilatation and dramatic right outer wall thinning that approached transparency (Fig. 1E-F). To confirm the extensive fibrosis seen in the myocardium of affected mutants, hearts from mice with HD were stained with Trichrome. A representative affected mutant heart (Fig. 2B) with images taken at several sites (indicated by arrows) and increased magnifications (Fig. 2C-H) and compared to an unaffected A/JHD littermate (Fig. 2A).
Replicating the trait in recombinants
We have sustained the mutant line for more than 20 generations as an A/J inbred strain (verified by MegaMuga SNP analysis). Although mutant mice are inbred, they differ from A/J mice for segregating variants involved in the trait. Because we do not yet know the responsible genes, we have preserved the necessary pool of disease-associated alleles by consistently maintaining about 10 mating pairs. Breeder pairs are selected to have at least one mate that had a parent or a sibling that died of HD, such that mating pairs are parent-offspring, brother-sisters from the same litter, or brother-sisters from different litters of same parents. To map these causal variants, our goal was to reproduce the disease trait dozens of times in recombinant mice generated from at least two different strains. To detect both dominant and recessive linkage, we first used an F2 breeding scheme. For all crosses, proven mutant males were used in the initial outcrosses (F1) for each strain. Then daughters were backcrossed with sires or intercrosses were performed to generate the recombinant populations. Initially, F1 hybrids produced from C57BL/6J (B6) females and proven A/JHD mutant males were intercrossed to generate a cohort of 102 B6.A/JHD (B6.HD) F2 mice (Additional file 1, see 1A); unexpectedly, none of these B6.HD-F2 recombinants developed HD (Table 1). A second B6.HD-F2 population (Additional file 1, see 1B) was generated using different mutant male breeders, but also yielded no mice that developed HD (n=181; Table 1). Compared to the 2-3 genes estimated to be segregating in the A/JHD inbred line, these B6.HD-F2 data (0 affected in 283 F2 mice) suggested a considerably more complex pattern of trait inheritance (i.e., at least 5 variants or other genetic or non-genetic factors).
Table 1. Crosses used to produce mice for EM/HD mapping.
Recombinant Cross
|
# litters
|
♂, affected
|
♀, affected
|
total, affected
|
% affected
|
(B6 x A/JHD) F2
|
10
15
|
54, 0
84, 0
|
48, 0
97, 0
|
102, 0
181, 0
|
0
0
|
(B6 x A/JHD) N2
|
5
|
21, 0
|
29, 1
|
50, 1
|
2.0
|
(S1 x A/JHD) F2
|
32
|
123, 0
|
143, 2
|
266, 2
|
0.75
|
(D2 x A/JHD) F2
(C5 deficient)
|
58
|
262, 9
|
248, 6
|
510, 15
|
2.9
|
(D2 x A/JHD) N2
(C5 deficient)
|
26
|
114, 7
|
122, 13
|
236, 20
|
8.5
|
(SJ x A/JHD) F2
(Dysf deficient)
|
31
|
148, 12
|
139, 10
|
287, 22
|
7.7
|
(SJ x A/JHD) N2
(Dysf deficient)
|
13
|
42, 8
|
56, 9
|
98, 17
|
17.3
|
Our second attempt concurrently generated two distinct recombinant populations to further explore whether the HD trait could be recapitulated in a mapping cross (Additional file 1, see right). For these crosses, we changed the mating scheme for one recombinant population and switched the inbred strain for the second. The first population again used B6 females, but this time in a backcross strategy to potentially double our chances of recapitulating the trait (Additional file 1, see 2A). In this case, a small cohort of B6 backcrosses (B6.HD-N2) was produced by using each mutant A/JHD male (proven sires) for both the initial F1 outcross and the subsequent backcrosses with his daughters; still, just one of 50 (2%) B6.HD-N2 recombinants developed HD (Table 1). Concomitantly, we generated 266 F2 recombinants from mutant A/JHD males and 129S1/SvImJ (S1) inbred females (Additional file 1, see 2B); two S1.HD-F2 mice developed HD (0.75% affected). Thus, although rare, HD was recapitulated in two different inbred strain-pairs and two breeding schemes, demonstrating that mapping the trait might be possible; however, this would require a huge population of recombinant mice to produce the number of affected mice required for linkage analysis. Still puzzling, however, was the discrepant rates of HD seen between these mapping crosses and the inbred A/JHD line.
Improving the rate of affected recombinant mice.
To explain the disparity between the estimated number of segregating causal mutations in the A/JHD line compared to the recombinants from B6.HD-N2, B6.HD-F2, and S1.HD-F2 crosses, we hypothesized that one or more causal variants contributing to EM/HD must already exist as fixed polymorphism(s) in the A/J inbred strain (and the A/JHD inbred mutant line). Because one or more of these A/J polymorphisms are fixed as homozygous, they are masked when maintaining the A/JHD line with brother-sister and parent-offspring matings. However, these fixed mutations begin to segregate when the A/JHD line is outcrossed with another inbred strain for the mapping studies, which significantly reduces the odds to reestablish the full set of allelic variants needed to cause HD. Given the large difference in the rates of affected offspring between A/JHD and its mapping crosses, it is plausible that more than one latent A/J variant is affecting the HD trait. By choosing a strain with the same or functionally similar mutation as in A/J, we could indirectly test this possibility while still attempting to generate recombinants for mapping. Because of their potential role in EM/HD pathology, hemolytic complement (C5) and dysferlin (Dysf) were prioritized from the list of 14 known mutated disease genes in A/J inbred mice (see: https://www.jax.org/strain/000646; View Genetics).
C5. Two conceivable mechanisms of disease initiation include 1) an unknown pathogen acts as a trigger or 2) an unknown self-protein induces an autoimmune response. C5 has a potential role in both: C5 deficiency increases susceptibility of A/J mice to many infectious agents [3-6] and C5 function and dysfunction are indicated in immunity and autoimmunity [7-11]. From this, we predicted that the effects from the natural C5 mutation in the A/J strain [12] combines with that of the de novo mutation (and possibly other latent mutations) in our A/JHD line to cause disease. To test this hypothesis A/JHD males were mated with DBA/2J (D2) females, an inbred strain that carries the same 2-bp deletion in C5 as does A/J [13]. These matings necessarily produce litters in which all recombinant offspring have homozygous mutant copies of C5 (as do the A/J and A/JHD lines). If C5 deficiency contributes to disease, then the odds are improved that the breeders will carry the required set of mutant alleles and produce more affected recombinant offspring.
D2 mice were bred using F2 and N2 mating schemes (Additional file 1, bottom) to evaluate a potential role of C5 in EM/HD and to produce recombinants for mapping the causal variants of HD. Cohorts of 510 intercross mice (D2.HD-F2) and 236 backcross mice (D2.HD-N2) for a total of 746 D2-derived recombinants, were produced and monitored for HD up to 20-weeks old (Table 1). D2.HD-F2 recombinants developed HD at a rate of 2.9% (15/510), with no sex difference identified (p=0.186). Twenty D2.HD-N2 mice (8.5%) developed HD, with a 1.7-fold difference (p=0.017) between females (13/122; 10.7%) and males (7/112; 6.3%). As expected, because the same mutant male was used for the F1 and then the N2 crosses with daughters, the rates of affected D2.HD-N2 mice were 2.9-fold that of D2.HD-F2 mice (8.5% versus 2.9%). The use of D2 mice increased the percentages of affected recombinants over B6 and S1 mice, consistent with our prediction that the lack of C5 has a direct or indirect role in EM/HD, possibly stemming from a more susceptible immune milieu in the absence of C5.
Dysf. For an additional strategy to improve the rate of generating recombinants developing HD and to identify regions significantly linked to the disease, we tested a parallel hypothesis that the known Dysf gene mutation in the A/J strain (and thus the A/JHD line) contributes to the mutant phenotype. Although the Dysf mutation in the A/J strain appears to be unique to this strain (a 5-6kb ETn retrotransposon inserted into intron 4) [13], the SJL/J (SJ) inbred strain has a splice-site mutation in Dysf, resulting in a markedly decreased protein level [14]. Like the A/J strain, the SJ strain is also considered a naturally occurring animal model for dysferlinopathy [15]. Consequently, all mice deriving from crosses of the SJ and A/JHD strains will carry two mutated copies of the Dysf gene. Like the D2 crosses for mutant C5, an increase in affected SJ-derived recombinants would indirectly support an involvement of mutant Dysf in the development, progression, and/or modification of the pathobiology associated with EM/HD.
Accordingly, the SJ-derived crosses were used to map the causal HD genes and test a potential contributing role for Dysf in EM/HD; again, the N2 and F2 mating schemes were exploited (Additional file 1, left). Cohorts of 287 intercross mice (SJ.HD-F2) and 98 backcross mice (SJ.HD-N2), for a total of 385 SJ-derived recombinants were produced and monitored up to 20-weeks old for HD (Table 1). SJ.HD-F2 recombinants developed HD at a rate of 7.7% (22/287), with no sex difference observed (females: 10/139; 7.2% and males 12/148; 8.1%; p=0.348). Seventeen SJ.HD-N2 mice (17.3%) developed HD, with a slight yet significant (p=0.029) difference between females (9/56; 16.1%) and males (8/42; 19.0%). Again, the backcross breeding scheme produced HD at more than twice the rate of F2 crosses (17.3% versus 7.7%). Consistent with our hypothesis of a contributing role for Dysf deficiency in EM/HD, the use of SJ-derived crosses dramatically increased the percentages of affected recombinants. Comparing these final crosses to the first 3 attempts (Additional file 1), the SJ.HD-N2 (17/98; 17.3%) improved the rate of HD ~35-fold over the combined B6 and S1 crosses (3/599; 0.5%).
QTL analysis
Assessment of genetic linkage to the EM/HD trait was performed using R/qtl analysis [16, 17] to associate HD (binary trait: present or not by 20-weeks old) with SNPs on the GigaMUGA SNP array generated for all affected and unaffected mice in the analyses. Each cross was analyzed separately and then combined with its alternative mating scheme partner (i.e., D2 crosses or SJ crosses).
D2 crosses. For the D2.HD-N2 analysis, a total of 19 affected (i.e., verified HD) and 6 unaffected (i.e., littermates that did not develop HD by 20-weeks old) backcross mice were used. Similarly, 16 affected and 16 unaffected D2.HD-F2 mice were utilized in a separate QTL analysis (Table 2). Except for the chromosome 2 area around C5, the F2 population will help to rule out genomic regions that are homozygous D2 in affected mice.
Table 2. QTLs identified in R/qtl analyses.
Cross
(#HD / #Unaffected)
|
Chr position
(@ peak SNP)
|
LOD Score
|
% Variance Explained
|
Level of Significance*
|
Haplotype-defined interval
Chr: Mb range (size of interval)
|
D2.HD-N2
(19 / 6)
|
5:147147276
|
3.32
|
43.2
|
suggestive
|
5:143159090–151734385
(8.58 Mb)
|
D2.HD-F2
(16 / 16)
|
–
|
–
|
–
|
–
|
5:144329217–147894899
(3.57 Mb)
|
SJ.HD-N2
(16 / 17)
|
5:148558183
|
4.88
|
47.4
|
highly significant
(Emhd1)
|
5:129525220–151734385
(22.21 Mb)
|
SJ.HD-F2
(22 / 22)
|
–
|
–
|
–
|
–
|
5:145128608–147778059
(2.65 Mb)
|
17:21456192
|
7.51
|
52.9
|
highly significant
(Emhd2)
|
17:15358323–23816187
(8.46 Mb)
|
|
17:35308100
|
6.89
|
49.6
|
highly significant
(Emhd3)
|
17:31471362–46058209
(14.59 Mb)
|
SJ.HD-N2 + F2
(38 / 39)
|
1:176044015
|
4.12
|
21.3
|
significant
|
1:172969406–194625219
(23.59 Mb)
|
|
5:146601763
|
5.79
|
28.6
|
highly significant
(Emhd1)
|
5:143159090–151734385
(8.58 Mb)
|
|
17:17933928
|
6.53
|
35.1
|
highly significant
(Emhd2)
|
same as F2
|
|
17:35308100
|
7.43
|
highly significant
(Emhd3)
|
same as F2
|
All genomic positions are based on mm10 (GRCm38; Build 38). Genotyping of >143,00 SNPs was performed by GeneSeek (Lincoln, NE) using the GigaMUGA SNP panel. Separate QTL analyses were run for each cross using R/qtl [16, 17] after applying argyle routines for quality checks [18]. Combined SJ crosses were analyzed as an F2 model (0,1,2), accepting the limitations. * Significance was determined using 1,000 permutations of the corresponding dataset. QTL intervals were determined using haplotype analysis, in which we identified the proximal and distal crossovers that best correlated the genotype with phenotype.
Results of the D2.HD-N2 QTL analysis identified a suggestive linkage on distal chromosome 5 (LOD=3.32; Fig. 3). Details of the peak markers, LOD scores, percent variance explained and QTL intervals for this and all other performed analyses are summarized in Table 2. Initially, the results of the D2.HD-N2 cohort also indicated a suggestive locus on distal chromosome 17 (LOD=3.39; Fig. 3), but this peak was subsequently ruled out after determining the region coincided with residual B6 genome from the founder A.B6 congenic. Specifically, one of the A/JHD breeder males used in the D2.HD-N2 crosses still contained a portion of the original B6 congenic interval. However, since HD occurred in mutant mice after removing the residual B6 interval, linkage to this region (and thus the QTL peak) was ruled out. Interestingly, neither the D2.HD-F2 analysis nor a combined (D2.HD-N2 + D2.HD-F2) analysis identified any locus that reached suggestive linkage (Additional file 2).
SJ crosses. Separate QTL analyses of SJ-derived N2 and F2 recombinants were performed and results are plotted in Fig. 4 and detailed in Table 2. The SJ.HD-N2 recombinants (16 affected and 17 unaffected) found the same distal chromosome 5 QTL identified as a suggestive locus in the D2.HD-N2 population, but in this case, the SJ.HD-N2 cohort was highly significant (LOD=4.88) and explained 47.4% of the trait variance. With this confirmation, we have designated the chromosome 5 linkage as Emhd1, for ‘eosinophilic myocarditis to heart disease’. Interestingly, the SJ.HD-F2 population (22 affected and 22 unaffected) did not identify the QTL on chromosome 5. Instead, a highly significant QTL was found on chromosome 17 (Fig. 4B). A closer look at the peak revealed two closely linked loci (Fig. 4C), including a locus proximal to MHC (LOD=7.51; designated Emhd2) and another region that contained the MHC region (LOD=6.89; named Emhd3). Each locus explained about 35% of the trait variance. However, their proximity to each other, and the fact that all but two mice had both QTLs in heterozygosity indicated the variance explained by these QTLs almost entirely overlapped. Similarly, both chromosome 17 QTLs were consistent with dominant variants, such that only one mutant allele of each is required in the HD trait. There was no difference in survival of affected mice that were heterozygous (n=20) and homozygous mutant (n=2) genotypes for Emhd2-3.
The combined analysis of the 77 SJ-derived N2 and F2 recombinants (38 affected and 39 unaffected) identified all three Emhd peaks (Table 2; Additional file 3). LOD scores were like those from the separate populations, demonstrating the utility of the separate SJ mating schemes to detect recessive versus dominant linkage for this complex trait: the backcrosses identified the recessive Emhd1 on chromosome 5 and the F2 crosses found the dominant-acting Emhd2 and Emhd3 on chromosome 17. The height of the Emhd2 and Emhd3 peaks flipped in the combined analysis compared with the SJ.HD-F2 but were still highly significant. We next tested for a possible interaction between the QTLs using the scantwo function of R/qtl. In a 2-QTL model for the chromosome 5 and chromosome 17 (which was treated as one locus) QTLs, the scantwo LOD score was additive (13.2) with no signs of epistasis (LOD=1.02), and the variance explained (53.6%) increased only slightly. In addition to these three Emhd loci, analysis of this combined SJ.HD-N2 + F2 dataset also revealed a putative significant linkage on distal chromosome 1 (LOD=4.12; Additional file 3A), which was not seen in any of the separate cohorts. However, one can easily see that this peak derives from the smaller peaks at similar positions in the separate analyses (Fig. 4A, 4B); this locus will require further assessment to determine its involvement.
Haplotype analysis and concordance
Using the Excel file of SNP genotypes for all mice with HD and unaffected littermates used in the mapping cohorts, the heterozygous and both homozygous genotypes were color-coded to visually perform haplotype analysis (Additional file 4). This color-coding allowed us to quickly scan the genomewide SNPs, with specific focus on linkage regions, to identify recombinant mice with crossovers that delineated the three QTL intervals. Fig. 5 presents an overview of the haplotype analysis results. For the recessive Emhd1 minimal region of effect, heterozygous SJ.HD-F2 recombinants determined the proximal and distal ends of the QTL, whereas the bracketed intervals of the dominant Emhd2 and Emhd3 QTLs were defined by homozygous SJ alleles in affected SJ.HD-F2 recombinants. Backcrosses, which carry at least one copy of the QTLs from the mutant breeder, are necessarily concordant for a dominant mutation (Additional file 5), and thus are uninformative for defining the QTL intervals of Emhd2 and Emhd3. Concordance for dominant loci was therefore assessed in F2 mapping cohorts.
Emhd1. The single B6.HD-N2 recombinant (1/50) and the two of 226 S1.HD-F2 recombinants that died of HD (Additional file 1) all carried homozygous A/JHD-derived genome across Emhd1. Among the D2- and SJ-derived recombinant populations, haplotype analysis for Emhd1 determined large differences in the size of the chromosome 5 interval (Table 2). Specifically, the Emhd1 interval spanned 22.21 Mb in the SJ.HD-N2 population (n=33), which was reduced to 8.58 Mb in the D2.HD-N2 cohort (n=25), to 3.57 Mb for the D2.HD-F2 recombinants (n=32), and finally down to 2.65 Mb (145.13–147.78 Mb) in the SJ.HD-F2 population (n=44). Of note, while neither F2 cohort identified linkage to Emhd1, individual recombinants of both F2 cohorts significantly refined the QTL interval of their respective N2 populations. The genotypes of all affected SJ.HD-F2 recombinants were concordant with HD across the defined QTL interval (i.e., 22/22 homozygous for mutant A/JHD alleles). However, 3 of 16 D2.HD-F2 affected recombinants were discordant (1=heterozygous, and 2=homozygous D2) for the expected homozygous mutant alleles at Emhd1 (Additional file 5). Among the unaffected SJ recombinant mice in the analysis, Emhd1 showed the expected 1:1 ratios of heterozygous or homozygous alleles in the N2 cohort and ratios of 1:2:1 in the unaffected F2 littermates. These data strongly support that Emhd1 is necessary, but not sufficient to cause EM/HD in this mutant.
Emhd2. Haplotype analysis of the D2 crosses for Emhd2 was not possible, because D2 and A/J inbred strains are identical-by-descent (IBD) for the proximal ~28.65 Mb of chromosome 17. However, haplotype analysis of the affected SJ.HD-F2 recombinants revealed specific crossovers that defined the proximal and distal ends of Emhd2, which confined the Emhd2 to an 8.46 Mb region (15.36–23.82 Mb; Fig. 5, Table 2) on chromosome 17. This interval maps proximal to the MHC region. All affected SJ.HD-F2 mice (22/22) were concordant for a dominant variant (Additional file 5). In contrast, 8/22 unaffected SJ.HD-F2 recombinants were discordant (i.e., no mutant alleles), which is consistent with the expectation of ~1/4 homozygosity (p=0.36) for SJ alleles in F2 crosses.
Emhd3 and putative QTL on chromosome 1. Haplotype analysis of the 22 SJ.HD-F2 affected mice revealed specific crossovers that defined the proximal and distal ends for Emhd3. As illustrated in Fig. 5, the Emhd3 interval was mapped to a 14.59 Mb region (31.47–46.06 Mb; Table 2) on chromosome 17 that includes the major histocompatibility genes and several complement genes of the alternative pathway. All 22 SJ.HD-F2 recombinant mice with HD used in the QTL analysis were concordant for a dominant variant (Additional file 5). In addition, the same 8/22 unaffected SJ.HD-F2 recombinants that were discordant for Emhd2 were also discordant for Emhd3 (i.e., homozygous for SJ alleles throughout both QTL intervals).
Haplotype analysis was also used to determine the interval size of the putative linkage on chromosome 1 (Table 2). Because the genotypes were consistent with a dominant variant (backcrosses are uninformative), only the F2 data were used. All 22 SJ.HD-F2 affected mice were concordant for a 23.59 Mb interval ranging from ~171.04–194.63 Mb on distal chromosome 1. Interestingly, the D2.HD-F2 cohort was highly discordant (i.e., up to 10/16 affected F2 mice were homozygous for D2 alleles across all (6/16) or a large portion (4/16) of this putative chromosome 1 interval, with at least 7/16 affected mice carrying homozygous D2 alleles at every SNP across the interval). This discordance could indicate the chromosome 1 linkage supported by SJ.HD-F2 recombinants was a false positive. However, it is also possible that this putative QTL represents a distinction in the set of contributing causal variants for HD between the SJ and D2 crosses.
Positional candidate genes
Using the Genes and Markers query form in MGI (The Jackson Lab; http://www.informatics.jax.org/marker), we generated lists of all ‘gene feature types’ mapping to the genetic intervals of the three Emhd and putative chromosome loci, as determined by haplotype analysis. After removing transcription start sites and CpG island sites, lists of the remaining genetic elements mapping to each of the haplotype-derived QTL intervals were generated, and results were itemized into 8 general categories for reference (Additional file 6). Using literature searches for keywords like heart disease, heart failure, immune dysfunction (including autoimmunity), inflammation, chemokine and cytokine changes, eosinophil pathobiology or coagulopathy, biologically-relevant positional candidates were identified from the known protein coding genes mapping to the highly significant Emhd loci. These positional candidate genes were further evaluated through the International Mouse Phenotyping Consortium database (www.mouse phenotype.org) to determine possible gene associations with disease. Top candidates are listed in Table 3 and represent potential starting points for follow-on studies.
Table 3. Positional Candidate Genes for Emhd1-3.
Gene
|
Name
|
Position Bld38
|
Example of Possible Biological Relevance
|
Emhd1 (Chr5: 145128608 – 147778059): 46 known protein-coding genes
|
Arpc1b
|
Actin related protein 2/3 complex, subunit 1B
|
145,114,256-
145,128,186
|
deficiency associated with severe inflammation, immunodeficiency, and cardiac vasculitis [19]
|
Cdk8
|
Cyclin-dependent kinase 8
|
146,231,230-
146,302,874
|
ectopic expression of Cdk8 induces eccentric hypertrophy and heart failure [20]
|
Flt1
|
FMS-like tyrosine kinase 1
|
147,562,196-
147,726,005
|
increased cardiac remodeling in cardiac-specific Flt1 receptor knockout mice [21]
|
Flt3
|
FMS-like tyrosine kinase 3
|
147,330,741-
147,400,489
|
activation improves post-myocardial infarction remodeling by protective effect on cardiac cells [22]
|
Lnx2
|
Ligand of numb-protein X 2
|
147,016,655-
147,076,572
|
affects T-cell-mediated immune responses by regulating level of the T-cell co-receptor, CD8a [23]
|
Emhd2 (Chr17: 15358323 – 23816187): 61 known protein-coding genes
|
Dll1
|
Delta-like canonical Notch ligand 1
|
15,367,354-
15,376,048
|
increased serum levels associated with diastolic dysfunction, reduced exercise capacity, and adverse outcome in chronic heart failure [24]
|
Ppp2r1a
|
Protein phosphatase 2, regulatory subunit A, alpha (PP2A)
|
20,945,454-
20,965,905
|
requisite for the function of regulatory T cells [25]
|
Tnfrsf12a
|
TNF receptor superfamily, member 12a (fn14)
|
23,675,445-
23,677,449
|
a novel role in the development of cardiac dysfunction and failure [26]
|
Emhd3 (Chr17: 31471362 – 46058209): 261 known protein-coding genes
|
C2, C4a, C4b, Cfb
|
Complement components 2, 4a, 4b, and factor b (alternative pathway)
|
between
34.728-34.882
|
the alternative complement pathway is dysregulated in patients with chronic heart failure [27]
|
H2 genes
|
~3 dozen MHC genes
|
between
33.996-37.275
|
a humanized HLA-DR4 mouse model for autoimmune myocarditis [28]
|
Tnf
|
Tumor necrosis factor
|
35,199,381-
35,202,007
|
cardiac-specific overexpression causes lethal myocarditis in transgenic mice [29]
|
Tnfrsf21
|
TNF receptor superfamily, member 21
|
43,016,555-43,089,188
|
KO mice show increased Th2 immune responses to T-dependent and -independent antigens [30]
|