Feasibility of constructing multi-dimensional genomic maps of juvenile idiopathic arthritis

Background Juvenile idiopathic arthritis (JIA) is one of the most common chronic conditions of childhood. Like many common chronic human illnesses, JIA likely involves complex interactions between genes and the environment, mediated by the epigenome. Such interactions are best understood through multi-dimensional genomic maps that identify critical genetic and epigenetic components of the disease. However, constructing such maps in a cost-effective way is challenging, and this challenge is further complicated by the challenge of obtaining biospecimens from pediatric patients at time of disease diagnosis, prior to therapy, as well as the limited quantity of biospecimen that can be obtained from children,particularly those who are unwell. In this paper, we demonstrate the feasibility and utility of creating multi-dimensional genomic maps for JIA from limited sample numbers. Methods To accomplish our aims, we used an approach similar to that used in the ENCODE and Roadmap Epigenomics projects, which used only 2 replicates for each component of the genomic maps. We used genome-wide DNA methylation sequencing, whole genome sequencing on the Illumina 10x platform, RNA sequencing, and chromatin immunoprecipitation-sequencing for informative histone marks (H3K4me1 and H3K27ac) to construct a multi-dimensional map of JIA neutrophils, a cell we have shown to be important in the pathobiology of JIA. Results The epigenomes of JIA neutrophils display numerous differences from those from healthy children. DNA methylation changes, however, had only a weak effect on differential gene expression. In contrast, H3K4me1 and H3K27ac, commonly associated with enhancer functions, strongly correlated with gene expression. Furthermore, although unique/novel enhancer marks were associated with insertion-deletion events (indels) identified on whole genome sequencing, we saw no strong association between epigenetic changes and underlying genetic variation. The initiation of treatment in JIA is associated with a re-ordering of both DNA methylation and histone modifications, demonstrating the plasticity of the epigenome in this setting. Conclusions These findings, generated from a small number of patient samples, demonstrate how multidimensional genomic studies may yield new understandings of biology of JIA and provide insight into how therapy alters gene expression patterns.


106
Gene-environment interactions are thought to mediate many complex human traits, 107 including human diseases [1], [2]. It is becoming increasingly clear that the influences of 108 environment (broadly considered) are mediated through epigenetic changes to DNA and

139
We have therefore explored whether the necessary "genomic maps" might be 140 generated using an approach similar to that used in the ENCODE and Roadmap

153
To accomplish these aims, we started with neutrophils, a cell type that can be safely 154 obtained in relative abundance from children and with which we have considerable experience. We performed genome-wide epigenome analyses using cross-sectional 156 samples to build ENCODE-like epigenetic maps. We then queried whether such maps   initiation of methotrexate, the standard therapy for JIA. Specifically, we studied 6 171 patients with active, untreated disease (referred to as ADU) and 3 patients with active 172 disease who had been started on methotrexate (,these patients are referred to as ADT).

173
Disease activity was assessed using standardized criteria developed by Wallace and 174 colleagues [20]. In addition, we performed whole genome methylation analysis on 5 175 healthy children (HC).

176
ChIP-seq cohort: Five children with JIA were studied: two children had untreated, 177 newly-diagnosed JIA and 3 children were on therapy with methotrexate, studied 6 to 8 178 weeks after the initiation of therapy. We also studied 3 healthy controls.

255
1% formaldehyde in 10 ml PBS for 10 min at room temperature (RT). Crosslinking was 256 quenched by adding 1× glycine and incubation for 5 min at RT. The crosslinked samples were centrifuged at 800 x g for 5 min. The supernatant was discarded, and the pellet 258 was washed two times with cold PBS followed by resuspension in 10 ml ice-cold Buffer

259
A plus DTT, PMSF and protease inhibitor cocktail. Cells were incubated on ice for 10 260 minutes and centrifuged at 800 x g for 5 min at 4°C to precipitate nuclei pellets, which 261 were then resuspended in 10 ml ice-cold Buffer A plus DTT. The nuclei pellet was

400
As noted in our Introduction, our goal in this study was to test the feasibility of generating

442
In addition to corroborating our previous findings, the RNA-Seq studies show 443 that reference "maps" of disease-specific epigenomes will need to be generated for 444 specific disease states (e.g., untreated disease, disease that is active, disease that is in

454
We studied neutrophils collected from children with active, untreated JIA (ADU: 2 455 individuals) and healthy controls (HC: 3 individuals). We first used ChIP-Seq to study 456 two histone marks, H3K27ac and H3K4me1, typically associated with enhancer activity.

457
We binned the human genome, then used edgeR [28] to call H3K4me1 and H3K27ac

513
In contrast, the DMRs identified between ADU and HC were not significantly 514 associated with differential gene expression when considering hyper-and hypo-515 methylated regions separately. We designated the genes that were (-5Kbs, 5Kbs) of 516 TSS or whose gene bodies intersected with DMRs as DMR-associated genes. We 517 identified 160 expressed genes associated with hypermethylated DMRs and 100 518 expressed genes associated with hypomethylated DMRs in the ADU group (S.Table6).

519
However, GO analyses did not reveal any significantly enriched terms in biological 520 processes. Only three ADU hypermethylated DMR-associated genes (ADARB2,

521
CNTNAP3B, and MYOM2) showed significant differential expression in the ADU vs. HC  (Figure 3a). In other words, after 540 treatment, these regions gaining the histone marks in the active disease state changed 541 significantly toward levels identified in HC (Figure 3c, 3d); we designate these as "gain-

547
DERs lost H3K4me1 in ADU then regain it in ADT (Figure 3b). These regions also 548 demonstrated histone marks that were more similar, compared with the gain-then-loss 549 DERs, to HC after the initiation of treatment (Figure 3c, 3d). Only a minority (4 for with the non-ADT groups (Figure 3f). This finding indicates that, after treatment, the 572 histone marks of these regions changed to a unique state other than ADU and HC 573 (Figure 3g, 3h). The expressed genes associated with those ADT unique DERs are 574 displayed in S. Table 9.

575
Taken together, these findings demonstrate that, even for functional signatures such

689
Taken together, these small, independently-acquired data sets do not indicate a strong 690 association between genetic variation (as identified by the WGS data) and epigenetic 691 alterations in JIA neutrophils. It seems likely that specific, directed experiments will be 692 required to identify allelic effects on epigenetic signals at specific genomic locations [38].

695
In the current work, we demonstrate the feasibility of developing informative, multi-696 dimensional genomic maps of JIA to study the interplay between genetics and 697 epigenetics in this common childhood disease. We took the same approach used by the

741
Our study pinpoints one environmental factor that is impacts neutrophil epigenomes: 742 the initiation of effective therapy. This study is the first, to our knowledge, to demonstrate 743 that effective therapy for a chronic human disease alters the epigenome. Furthermore,

744
alterations in the epigenome associated with therapy tended to "correct" the neutrophil 745 epigenome closer to patterns found in the cells of healthy children, as shown in Figure  S3.