Shared and Distinctive Transcriptomic and Proteomic Pathways in Adult and Juvenile Dermatomyositis

Abstract

Objective

Transcript and protein expression were interrogated to examine gene locus and pathway regulation in the peripheral blood of active adult dermatomyositis (DM) and juvenile DM (JDM) patients receiving immunosuppressive therapies.

Methods

Expression data from 14 DM and 12 JDM patients were compared to matched healthy controls. Regulatory effects at the transcript and protein level were analyzed by multi-enrichment analysis for assessment of affected pathways within DM and JDM.

Results

Expression of 1,124 gene loci were significantly altered at the transcript or protein levels across DM or JDM, with 70 genes shared. A subset of interferon-stimulated genes was elevated, including CXCL10, ISG15, OAS1, CLEC4A, and STAT1. Innate immune markers specific to neutrophil granules and neutrophil extracellular traps were up-regulated in both DM and JDM, including BPI, CTSG, ELANE, LTF, MPO, and MMP8. Pathway analysis revealed up-regulation of PI3K/AKT, ERK, and p38 MAPK signaling, whose central components were broadly up-regulated in DM, while peripheral upstream and downstream components were differentially regulated in both DM and JDM. Up-regulated components shared by DM and JDM included cytokine:receptor pairs LGALS9:HAVCR2, LTF/NAMPT/S100A8/HSPA1A:TLR4, CSF2:CSF2RA, EPO:EPOR, FGF2/FGF8:FGFR, several Bcl-2 components, and numerous glycolytic enzymes. Pathways unique to DM included sirtuin signaling, aryl hydrocarbon receptor signaling, protein ubiquitination, and granzyme B signaling.

Conclusion

The combination of proteomics and transcript expression by multi-enrichment analysis broadened the identification of up- and down-regulated pathways among active DM and JDM patients. These pathways, particularly those which feed into PI3K/AKT and MAPK signaling and neutrophil degranulation, may be potential therapeutic targets.

Graphical Abstract

Introduction.

Dermatomyositis (DM) in adults and children (JDM) is a systemic autoimmune disorder affecting skin and muscle tissues, with proximal weakness and photosensitive rashes (1). Affected muscle tissue often shows elevated levels of MHC-I expression, as well as infiltration of dendritic cells, B cells, T helper cells, and macrophages (1). Myositis autoantibodies define distinct phenotypes, which often involve other organs, such as joints, lungs, heart, and gastrointestinal tract (2).

Both DM and JDM have environmental and genetic causes, including shared HLA class II alleles and immunoregulatory genes, with some distinct allelic risk factors (3, 4). Multiple studies have shown upregulation of type I and II interferon (IFN) response genes (IRGs) in the peripheral blood, muscle, and skin of untreated DM/JDM patients, which correlates with disease activity (5–7). Limited data suggests additional pathways have altered regulation in peripheral blood and muscle tissue, including non-IFN immune response genes, autoantigens, muscle structural proteins, and mitochondrial enzymes (8–12).

Gene expression studies have described shared effects in peripheral blood and target tissues of DM and JDM, although few have studied them in parallel or examined their distinctions (5, 13, 14). Most gene regulation is described only at the transcript level, with protein measurements limited to a targeted subset of genes of interest (6, 8, 15). Few studies have utilized proteomics mass spectrometry to look at biopsied DM tissue, and none have analyzed transcript and protein expression together, nor at sufficient scale for comparative pathway analysis (15, 16).

The present study compares transcript and protein regulation in peripheral blood in DM and JDM to identify novel pathway effects in active patients receiving immunosuppressive and biologic therapies.

MATERIALS AND METHODS

Patients and sample collection.

DM patients (16 DM, 12 JDM) meeting definite EULAR-ACR criteria were enrolled at the National Institutes of Health (NIH) Clinical Center in an institutional review board-approved myositis natural history study (protocol 94E0165) (17). Whole blood samples from healthy subjects were collected under the internal donor program approved by the MedImmune institutional review board and matched to DM patients by sex; race was not available. Eleven healthy pediatric controls enrolled at the NIH Clinical Center and one enrolled at MedImmune were matched to 12 JDM patients by sex and race, after excluding technical outliers. The average difference in age between healthy subjects and DM/JDM patients was 8.0 years for adults and 3.1 years for pediatric subjects. Patients were evaluated using International Myositis Assessment and Clinical Studies (IMACS) Group core set measures of activity (18). Serum and RNA for all subjects were collected in PAXgene tubes (Becton-Dickinson, Palo Alto, CA), processed immediately following blood draw, and then stored at −80°C.

Transcript and protein expression analysis.

Whole blood RNA was examined using Affymetrix Human U133 Plus 2.0, processed in one batch, then normalized by frozen Robust Multiarray Analysis (fRMA) (19). Technical outliers were defined as >2x median absolute deviation (MAD) versus median variability within each patient group and reviewed by mean-difference MA-plots, centered expression heatmaps, and between-groups analysis (BGA) clustering (20, 21). The final cohort consisted of 14 DM and 12 JDM patients, and equal numbers of healthy matched controls.

Protein measurements were obtained using SOMAscan Assay v3.2 (SomaLogic, Boulder, CO) aptamer-based DNA probes. The standard protocol was followed by a mitigation protocol for DNA autoantibodies (22). Protein measurements were log2-transformed, independently quantile-normalized, and filtered as described for Affymetrix data. In total, 20,962 genes were represented by Affymetrix transcriptome array and 1,309 proteins by the SomaLogic array.

Statistical comparisons were applied using limma-3.40.6 unpaired moderated t-tests (23). Significantly altered Affymetrix probes were defined by Benjamini-Hochberg adjusted (BH) P-value <0.05 and absolute fold change $\ge$1.5 versus respective control subjects (24). Significantly altered protein levels were defined by BH-adjusted P-value <0.1 and absolute fold change $\ge$1.3 versus respective controls. Platform identifiers were updated using Bioconductor to current Entrez gene symbols for cross-platform comparisons (25). Results were summarized to the gene locus level using the most statistically significant assay per gene. Heatmaps were produced with ComplexHeatmap (26).

Interferon gene score.

Interferon (IFN) gene scores were calculated as described by Kim et al. using 28 IFN genes on the Affymetrix array (27). IFN gene scores in patients and healthy controls were compared using one-tailed Mann-Whitney-U test and were tested for correlation with disease activity using Spearman’s Rho correlation. “Elevated IFN” was defined as an IFN score $\ge$50 (27).

Pathway enrichment.

Significantly altered gene transcripts and proteins were tested for enrichment of canonical pathways through Ingenuity Pathway Analysis (version 48207413, Aarhus, Denmark, www.ingenuity.com), using a false discovery rate (FDR) adjusted P-value <0.05 and at least four significantly regulated genes (28). Multi-enrichment analysis was performed using multienrichjam to integrate transcript and protein evidence of pathway regulation within and across patient groups (20). A gene-pathway incidence matrix was created using the top 20 canonical pathways and the corresponding genes for each pathway. Hierarchical clustering produced a dendrogram with four pathway clusters, which were used to produce a gene-pathway concept network. Genes in each cluster were functionally annotated with findings from published studies, databases, and known temporal phases of immune responses. ProteinAtlas.org annotations were applied, and expression profiles were downloaded for all human tissues and blood cell types and were made available via Github R package “jmw86069/pajam” (29).

Data availability.

Transcript and protein expression data are available via dbGaP phs003270.v1.p1.

RESULTS

Patient characteristics.

Features were generally similar between DM and JDM patients, including sex, race/ethnicity, disease activity, myositis autoantibodies, and number of immunosuppressive therapies (Table 1). All patients received immunosuppressive therapies, such as prednisone in combination with methotrexate, other DMARDs, and biologic therapies, such as rituximab and IVIG. The median Global Disease Activity was 2.2 on a 10 cm visual analogue scale, and median disease duration was 1.2 and 1.3 years, respectively. Median oral corticosteroid dosage at the time of blood sampling was higher in JDM at 0.45 mg/kg/d, than DM patients at 0.07 mg/kg/d.

Table 1.

Demographic and Clinical Features of Adult and Juvenile Dermatomyositis Patients.

Feature	DM (n = 14) Median [Q1, Q3] or n (%)	JDM (n = 12) Median [Q1, Q3] or n (%)
Age (years)	52.4 [49.9, 56.9]	10.9 [8.4, 15.4]
Female subjects	12 (85.7%)	6 (50.0%)
Race and ethnicity
African American	4 (28.6%)	1 (8.3%)
Asian	0 (0.0%)	2 (16.7%)
Hispanic	0 (0.0%)	2 (16.7%)
White	10 (71.4%)	7 (58.3%)
Myositis-specific autoantibodies
p155/140 (TIF1)§	4 (28.6%)	3 (25.0%)
MJ (NXP2)	4 (28.6%)	4 (33.3%)
MDA5	1 (7.1%)	3 (25.0%)
Jo-1	2 (14.3%)	0 (0.0%)
Myositis autoantibody negative	3 (21.4%)	2 (16.7%)
Myositis-associated autoantibodies	3 (21.4%)	4 (33.3%)
Disease activity (0–10 cm VAS)	2.2 [0.4, 6.1]	2.2 [1.7, 6.0]
Disease duration (years)	1.3 [0.9, 8.7]	1.2 [0.4, 6.1]
Daily glucocorticoid dose, mg/kg/d	0.07 [0.04, 0.19]*	0.45 [0.19, 0.80]*
Number of additional immunosuppressive therapies**	1.0 [1.0, 1.25]	2.0 [1.0, 2.0]

^§Autoantibodies were indeterminate for one DM patient who tested positive for both anti-p155 (TIF1) and anti-Mi-2 autoantibodies

^*DM vs. JDM p value < 005

^**Additional immunosuppressive/immunomodulatory therapies received at time of blood draw: methotrexate (9 DM, 8 JDM), hydroxychloroquine (4 DM, 6 JDM), pulse methylprednisolone (4 JDM), mycophenolate mofetil (2 DM, 4 JDM), azathioprine (1 DM), rituximab (1 DM), cyclophosphamide (1 JDM), tacrolimus (1 JDM), intravenous immunoglobulin (IVIG) (3 DM, 6 JDM)

Abbreviations: DM: dermatomyositis, JDM: juvenile dermatomyositis, Q1: first quartile; Q3: third quartile, MSA: myositis specific autoantibodies; VAS, visual analog scale

Transcript and protein regulation.

In DM, 380 transcripts and 283 proteins were significantly up- or down-regulated in peripheral blood (Tables ST1, ST2). In JDM, 499 transcripts and 51 proteins were significantly altered in regulation (Tables ST3, ST4). Although the genes assayed by the transcript and protein platforms showed relatively low overlap (6%), significantly altered transcripts were confirmed by protein data at a high rate (28%). The collective transcript and protein evidence indicated altered regulation in 649 gene loci in DM and 545 in JDM.

Combining transcript and protein results at the gene level, 70 differentially expressed gene loci were shared by DM and JDM, and 63 (90%) of these were concordant in direction across DM and JDM (Table 2, Figure S1). Fifty-six of 70 gene loci (80%) shared across cohorts were detected by the same platform, 31 genes at the transcript level, and 25 at the protein level.

Table 2.

Peripheral blood transcripts or proteins shared by adult and juvenile dermatomyositis patients.

Up-regulated in DM1					Tx down-regulated in DM and JDM					Tx or protein down-regulated in DM and JDM
Gene2	DM Tx3	DM Prot	JDM Tx4	JDM Prot	Gene	DM Tx	DM Prot	JDM Tx	JDM Prot	Gene	DM Tx	DM Prot	JDM Tx	JDM Prot
MCL1	2.4	1.9		1.4	AKAP13	−2.1	..	−1.5	..	BDNF		−1.7		−1.4
GDF15		1.7		2.0	AUTS2	−2.0	..	−2.4	..	CD200R1		−1.3		−1.5
H3C3		2.6		1.8	CBLB	−1.7	..	−1.7	..	CHL1		−1.4		−1.5
LGALS9		1.6		1.8	CXCR3	−1.5	..	−1.7	..	CNTN1		−1.5		−1.5
OAS1		1.6		2.4	EOMES	−2.2	..	−2.2	..	COLEC12		−1.4		−2.1
PFDN5		2.0		1.4	FEZ1	−1.7	..	−1.8	..	CST6		−1.9		−1.6
PGAM1		2.8		3.3	GZMK	−3.6	..	−2.9	..	FAP		−1.4		−2.0
AZU1		2.3	2.9		IGH	−4.2	..	−3.0	..	IDS		−1.5	−1.6	−1.4
BPI		3.1	3.1		IL2RB	−1.8	..	−1.8	..	KIT		−2.2		−1.9
CTSG		1.6	2.6		KIR3DL3	−1.5	..	−1.6	..	L1CAM		−1.5		−1.7
ELANE		1.7	3.1		MIAT	−2.1	..	−2.2	..	MRC2		−1.4		−1.7
H2AW		8.5	1.7		NCALD	−1.5	..	−1.5	..	PPBP		−1.4		−1.4
LTF		2.1	6.0		PRF1	−1.8	..	−1.8	..	RET		−2.2		−2.5
MMP8		1.8	5.2		RHOBTB3	−1.5	..	−1.7	..	SHBG		−2.0		−2.8
MPO		1.8	2.0		RUNX3	−1.7	..	−1.7	..	SLITRK5		−1.6		−1.8
					SEC14L1	−1.7	..	−1.6	..	SPINT1		−1.9		−1.5
					SIGLEC17P	−1.6	..	−1.6	..	THBS1		−1.6		−1.4
APLP2	1.6	..	−1.7	..	SMAD7	−1.6	..	−1.9	..	TNFRSF17		−1.5		−1.7
FKBP1A	1.7	..	−1.9	..	SYTL2	−2.2	..	−1.8	..	CCL5		−1.5	−2.4
LAMP1	1.6	..	−1.6	..	TRA2A	−1.8	..	−1.5	..	TGFBR3		−1.4	−2.3
AKR7A2		1.5	−1.8		TXNDC5	−1.7	..	−2.0	..
FGR		2.2	−1.6		XCL2	−1.8	..	−1.7	..
HSP90AB1		2.1	−1.7		YME1L1	−1.9	..	−2.8	..
YWHAZ		1.4	−1.8		IGHG3	−3.4		−3.6
					KIR3DL2	−2.1		−2.2
					MED1	−2.0		−1.6
					SLAMF7	−1.6		−1.9
					XCL1	−1.8		−1.7

¹Numerical values indicate the fold change, empty values indicate no significant change detected, and '..' indicates no data available.

²Entrez gene symbols are listed.

Abbreviations: DM = dermatomyositis; Tx = transcript; Prot = protein; JDM = juvenile dermatomyositis.

Of 163 secretory proteins with altered expression, 75 of 163 (46%) were detected by transcriptomics, and 101 of 120 measured (84%) were detected by proteomics, with 13 of these results shared. In contrast, 69 of 73 (95%) altered transcription factors (TFs) were detected by transcriptomics, while proteomics detected changes in four of only five TFs that it measured. Thus, changes in secreted proteins were detected at a higher rate by proteomics than transcriptomics, while changes in TFs were detected similarly on both, although with much lower coverage of TFs by proteomics.

Transcript and protein up-regulation correlated with specific expression in monocytes, granulocytes, and myeloid dendritic cells, while down-regulation correlated with T, B, and NK cells (Figure S2). These cell type trends were evident in flow cytometry analysis of purified peripheral blood mononuclear cells (data not shown).

Type I IFN gene expression was not significantly elevated by the 28 IFN-related gene score (Figure S3). Elevated expression was seen in a subset of the 28 IFN signature genes, including CXCL10, ISG15, OAS1, CLEC4A, and STAT1 (27). The mean IFN gene score was 7.9 ± 28.6 in DM patients vs. 0.0 ± 24.0 in healthy adult controls (p=0.16), and 14.1 ± 73.8 in JDM patients vs. 0.0 ± 18.2 in juvenile controls (p = 0.86). Elevated IFN gene scores were present in two DM and JDM patients each, one healthy adult control, and none of the pediatric controls. The correlation between IFN gene scores and disease activity was not significant in DM (r=0.30, p=0.29) or JDM patients (r=0.20, p=0.54).

From the examination of differential transcript expression in DM patients, 38 functional pathways were identified (Table ST5), and 293 pathways were detected using the protein expression data (Table ST6). Most transcriptome-driven pathways (31 pathways, 82%) were shared with protein pathways. Transcriptomics of the JDM peripheral blood identified 181 pathways (Table ST7), and proteomics identified eight pathways with altered regulation (Table ST8), all shared with transcriptomics pathways. Overall, 135 of 181 JDM pathways (75%) were shared with DM.

DM transcript and protein multi-enrichment.

Multi-enrichment analysis of transcript and protein pathways in DM revealed broad, complementary expression alterations within each pathway (Figures 1, S4). Collective effects were illustrated by seven up-regulated ubiquitination enzymes across the array platforms, with one found at both the transcript and protein level (UBE2I). Similar examples of cross-platform confirmation included two of five up-regulated Bcl2 pathway components, one of five confirmed IL-1 signaling molecules, and one of three down-regulated casein kinase subunits detected by both platforms. Together, the transcript and protein regulation patterns corroborated evidence for the same functional pathways despite relatively low, direct molecular overlap.

Figure 1. Adult dermatomyositis multi-enrichment of transcript and protein pathways.

Pathway clusters (large nodes A, B, C, D) are connected to regulated gene loci (small nodes). Nodes are red for transcript, yellow for protein, red/yellow for both with altered regulation. Heatmaps show patient (column) expression fold change versus healthy controls for each gene locus (row), with black box indicating significant changes. Heatmap column groups are grouped red for transcript, yellow for protein data. Heatmap rows are grouped red for transcript, yellow for protein, red-yellow for both with altered regulation. Heatmap cells are colored by direction: blue for down-regulation, red for up-regulation, white for no change, and grey for no protein data available for the corresponding gene.

Abbreviations: DM, dermatomyositis; tx, transcript

The combined transcript and protein pathway analysis revealed alterations in systemic inflammatory pathways, including innate and adaptive immune responses, chemo-attractants and extravasation factors, intracellular and extracellular signaling, evidence of immunosuppression, and crosstalk with metabolic pathways. Activation of innate and adaptive responses was shown in DM Cluster A by up-regulation of extracellular signaling components CXCL10, CSF2, NAMPT, IL1B, and IL1RN. Altered regulation of signaling via the aryl hydrocarbon receptor, docosahexaenoic acid (DHA), and vitamin D receptor (VDR)/retinoid X receptor (RXR) indicated heightened responsiveness to external stimuli, punctuated with up-regulation of protein ubiquitination and immunoproteasome-specific PSMB8 (β5i) (30). In contrast, up-regulation of sirtuin signaling and several apoptotic BCL-2 components (BID, BAD, BAX, MCL1), together with down-regulation of pro-inflammatory molecules in T cells and NK cells (CCL5, IL6R, KIR2DL5A), indicated an immuno-suppressive late-stage of chronic immune responses.

Four ubiquitous signaling pathways are defined in DM cluster B: NF-κB, PTEN, STAT3 signaling, and IL-15 production. The core genes in these pathways feature several growth factors and growth hormone receptors: GHR, TGFBR3, IGF1R, and EGFR. These proteins were coordinately down-regulated, and were secreted or membrane-bound, suggesting suppressed activity in the involved cell types.

A pro-inflammatory state was observed in DM cluster C, via up-regulation of core signaling components STAT1, STAT3, and eight MAP kinases. These signaling components were accompanied by innate immune signaling genes IL1B, IL1RN, IL1R1, IL1R2, and IL6R, as well as anti-apoptotic components BID and MCL1. Up-regulated pathways included IL-6-type cytokine signaling, acute phase response, Nrf2-mediated oxidative stress response, and glucocorticoid signaling. Activation of these pathways demonstrated an active, inflammatory immune response.

Changes in inflammatory pathways were further shown in DM cluster D, including uniquely up-regulated components of 14-3-3, IGF-1, PI3K/AKT, and PTEN signaling. Several activated pathways indicated connections from the varied external stimuli in cluster A to the MAP kinase-centered signaling pathways in cluster C: UVB-induced MAPK signaling, RAR activation, NK cell signaling, and EGF, B cell receptor, IL-7, GM-CSF, and Neuregulin signaling. Another example involved down-regulation of four NK cell immunoglobulin-like receptors (KIRs), which link NK cell signaling to dendritic cell crosstalk in Cluster A, representing modulation of activation and repressive immune response signals.

JDM transcript and protein multi-enrichment.

Multi-enrichment analysis in JDM showed consistent transcript and protein pathways in four clusters, despite a smaller number of differentially regulated transcripts and proteins detected in JDM peripheral blood (Figures 2, S5). Overall, the pathways reflected aberrant regulation of immune signaling with elevated inflammatory markers including CXCL16, GDF15, and MCL1, as well as broad suppression of immunoregulatory genes including JAK1, CCL5, CCR3, CCR5, PIK3R6, TNFSF13 transcripts, and TNFRSF17 protein.

Figure 2. Juvenile dermatomyositis multi-enrichment of transcript and protein pathways.

Pathway clusters (large nodes E, F, G, H) are connected to gene loci with altered regulation (small nodes). Node and heatmap colors assigned as in Figure 1, using purple for transcript and blue for protein data. Heatmap cells are colored by direction: blue for down-regulation, red for up-regulation, white for no change, and grey for no protein data available for the corresponding gene.

Abbreviations: JDM, juvenile dermatomyositis; tx, transcript

Th1 and Th2 helper cell activation pathways were enriched in JDM cluster E, although most transcripts and proteins were down-regulated. Galectin-9 (LGALS9) was one of three up-regulated proteins, along with ICOS and IL1RL1. Down-regulated immune components included: transcription factors MAF, NFATC2, and NFATC3; HLA class I, including HLA-B; HLA class II, including HLA-DMA, HLA-DPA1, HLA-DQB1, HLA-DRA, HLA-DRB1, HLA-DRB4, HLA-DR5; and signaling genes JAK1, AKT1, PIK3R6, CXCR3, and CXCR6.

HLA class I and II gene loci were repeated in JDM cluster F, which added down-regulated HLA class I gene loci HLA-C and HLA-G, down-regulated T-cell gamma receptors TRGC2 and TRGV9, T-cell signaling molecules LCK and ZAP70, and NK-cell immunoglobulin-like receptors KIR3DL2 and KIR3DL3. Up-regulated gene loci included BCL2L1, and phagosome/neutrophil extracellular trap (NET) gene loci CTSE, CTSG, and MPO. Additional pathways observed in this cluster reflect antigen presentation and autoimmune mechanisms, such as allograft rejection, graft-versus-host disease, and CTLA4 signaling in cytotoxic T lymphocytes.

Granulocyte adhesion and diapedesis markers were observed in JDM cluster G, consistent with granulocyte expression profiles. Key extracellular signaling markers CXCL16, MMP3, MMP8, FN1, IL1RL1 were also elevated, while megakaryocyte marker PPBP was down-regulated, along with several cytokines, CD99, CCL4, CCL5, XCL1, XCL2, and receptors, CCR3, CCR5, CXCR3, and CXCR6.

Phagosome formation and rheumatoid arthritis pathways were elevated in JDM cluster H, with upregulated transcripts PAK4, ARHGEF12, ACKR1, ITLN1, and proteins FGF8, MCL1, MYL4, GDF15, IL1RL1, and NTN4. Other pathways with decreased expression included PI3K/AKT signaling, phospholipase C signaling, leukocyte extravasation, and semaphorin signaling.

DM and JDM multi-enrichment.

Multi-enrichment analysis was performed across DM and JDM transcript and protein pathways to assess the functional relationship between these two phenotypically similar patient groups. Overall, the analysis represented 411 gene loci across 61 pathways, with 204 genes shared in more than one pathway cluster (Figures 3, S6). Each cluster retained on average 52 unique gene loci, and 123 which were shared with one or more clusters. Forty-four of 70 shared DM-JDM gene loci were included proportionally across the pathway clusters, implicating each functional cluster across both cohorts. Overall, 35 of 61 (57%) pathways were differentially regulated in both DM and JDM, however each patient group also contained enriched pathways that were specific to either DM or JDM.

Figure 3. Adult and juvenile dermatomyositis multi-enrichment of transcript and protein pathways.

Pathway clusters (large nodes I, J, K, L) are connected to gene loci with altered regulation (small nodes). Node and heatmap colors assigned as in Figures 1 and 2, using red for adult DM transcript, yellow for adult DM protein, purple for JDM transcript, and blue for JDM protein data. Heatmap cells are colored by direction: blue for down-regulation, red for up-regulation, white for no change, and grey for no protein data available for the corresponding gene.

Abbreviations: dermatomyositis, DM; JDM, juvenile dermatomyositis; tx, transcript

Among the shared DM/JDM functional elements, a prominent result in pathway cluster J involved granulocyte adhesion and diapedesis, altered at both the transcript and protein level (Figure S7). The overall pattern of up- and down-regulation was consistent in DM and JDM, involving chemokines, extracellular matrix remodeling, and interleukin signaling molecules. Up-regulated genes CCL7, CXCL10, and CXCL16 represent pleiotropic effects, such as recruitment and stimulation of monocytes, NKT-cells, T-cells, and modulation of cellular adhesion. Down-regulated genes XCL1, CCL5, and CXCL5 are chemotactic for activated T-cells, NK-cells, and neutrophils.

Th1 and Th2 cell activation pathways were also apparent in DM and JDM in cluster I (Figure S8), with shared up-regulation of LGALS9, and down-regulation of CXCR3, IL2RB, RUNX3, and TGFBR3. Granulocyte adhesion and Th1/Th2 activation pathways indicate recruitment of monocytes and granulocytes as a functional core shared by DM and JDM.

The multi-enrichment analysis also identified features unique to either DM or JDM patients. Signaling pathways in cluster K demonstrated effects in DM not seen in JDM, including aryl hydrocarbon, protein ubiquitination, granzyme B, and sirtuin signaling (Figure S9). The pathways in cluster K describe responses to external stimuli, related to the central PI3K/AKT and MAP kinase signaling pathways, which are represented in cluster L. Further, cluster L was characterized by several up-regulated MAP kinases in DM proteins, while JDM enrichment reflected only a subset of the surrounding signaling markers, such as GDF15, LGALS9, MCL1, and PAK4 (Figure S10). Together, the up-regulation of central signaling components indicated multiple, active pro-inflammatory pathways in DM patients.

The most notable JDM-specific signature was broad suppression of gene loci related to antigen-presentation, including numerous HLA class I and class II loci, and associated T-cell and NK-cell marker genes, including CTLA4 signaling (Figure S11). DM did share comparable down-regulation of immunoglobulin heavy chains IGH and IGHG3, as well as T-cell and NK-cell markers.

A functional theme emerged from gene loci up-regulated in both DM and JDM: the regulation of phagocytosis, neutrophil degranulation, and NETs. Overall, 20 of 32 NETosis components were up-regulated, eight of which were seen in both DM and JDM, including AZU1, BPI, CTSG, ELANE, H3C3, LTF, MMP8, and MPO (Figure S12) (31, 32).

DISCUSSION

This study characterized transcriptomic and proteomic changes in the peripheral blood of DM and JDM patients during long-term treatment. These patients with predominantly mild to moderate disease activity still exhibited distinct and consistent immune dysregulation, which may present unique therapeutic targets. The findings describe a persistent inflammatory state in the presence of a limited type I IFN response and may reflect pathogenic features that continue during conventional immunosuppressive therapies when disease remains active.

Although the type I IFN gene signature was not broadly elevated, a subset of type I IFN gene loci exhibited up-regulation, including CXCL10, ISG15, OAS1, CLEC4A, and STAT1. Four of five markers associated with improvement in extramuscular disease activity were also up-regulated in DM patients, including IL1B, STAT3, STAT6, and BCL6, suggesting a positive response to treatment (14). Although previous studies have focused on the elevation of type I IFN in active DM/JDM patients, some reports also showed subgroups of active patients without a strong IFN response, which correlated with lower levels of disease activity, lack of responsive to immunosuppressive therapy, and longer disease duration (5, 11, 12). Therapy for JDM frequently includes high dose pulse corticosteroids, which was previously shown to extinguish the type I IFN signature in SLE patients (33). Together the evidence suggests that the type I IFN signature for patients in this study was diminished in part by their treatment.

A strength of the present study was in the complementary results from transcriptomics and proteomics using the multi-enrichment analysis, with the key observation that pathway effects were broadly shared, despite sparse overlap in specific transcripts and proteins. This integrated approach afforded deeper insights into underlying immune mechanisms in DM/JDM patients than from transcriptomics or proteomics alone (5, 6). The results highlighted altered functional regulation of a number of pathways among clinically active patients receiving immunosuppressive therapy, an observation also reported in a small cohort of treated JDM patients (11).

The transcriptomic and proteomic findings in adult DM and JDM, together with review of literature and relevant pathway resources from Ingenuity IPA (QIAGEN Inc.), were integrated into a pathway schematic that shows key components with altered expression that are focused around three intracellular signaling cascades (Figure 4). Connections between genes were derived from literature review of well-studied signaling pathways. However, there may be other mechanisms in DM/JDM by which these components interact which have not yet been described. Our analysis also did not provide the basis for detecting novel interactions between pathways.

Figure 4.

Schematic of gene and protein regulation changes in dermatomyositis (DM) and juvenile dermatomyositis (JDM) peripheral blood, in context of multiple affected pathways. Gene loci are colored red for up-regulation, blue for down-regulation, and white for no change in expression. A solid outline indicates gene loci altered in DM, dashed outline indicates gene loci altered in JDM, and both solid and dashed outlines indicate changes in DM and JDM. Some genes are grouped into labeled boxes by pathway component. An arrow indicates a direct signaling effect, and a red flat cap indicates inhibition.

In adult DM, there was broad up-regulation of PI3K/AKT, ERK/MAPK, and p38/MAPK at both the transcript and protein level. The pathway clustering revealed MAPK genes as core components across 25 canonical pathways, representing a variation on the theme: upstream stimulus, intermediate signaling via MAPK genes, and downstream functional effects. Although JDM findings did not show MAPK gene regulation, aberrant regulation was seen by numerous upstream and downstream components of many of the same pathways. Key immunological markers in DM and JDM, including activation of IL1B, CXCL10, and LGALS9, signify an underlying inflammatory process, while down-regulation of key signaling partners, including KIR3DL2, CXCR3, CXCR6, CCL5, and EGFR, is also indicative of immune suppression. The p38/MAPK activation of IL1B plays an important role in propagating the inflammatory response (34). PI3K/AKT signaling and ubiquitin-conjugating enzymes together were shown to affect disease pathogenesis via ER stress, mis-folded protein response, and autophagy (35, 36).

Among the potential receptor:ligand inputs to PI3K/AKT, ERK, and p38 signaling, supporting evidence from both DM and JDM indicated three patterns of regulation. A subset of these showed concomitant up-regulation of ligand:receptor, including LTF/HSP70:TLR2, NAMPT/S100A8/S100A9:TLR4, LGALS9:HAVCR2, EPO:EPOR, CSF2:CSF2RA, and FGF2/FGF8:FGFR. In contrast, several myositis markers were up-regulated, including CXCL10, CXCL16, and IL1B, yet their corresponding receptors CXCR3, CXCR6, and IL1R1/IL1R2 were down-regulated (37, 38). These cytokines may be targeting muscle, skin, or cell types diminished in peripheral blood. In addition, several growth factors and their receptors associated with T, B, and NK cells were down-regulated, including GH:GHR, INS:INSR, TGFBI:TGFBR3, KITLG:KIT, CD200:CD200R1, CCL5:CCR5, and EGF:EGFR (39).

Activation of PI3K/AKT and MAPK signaling cascades elicits many potential effects, including enhanced degradation by the proteasome and its associated ubiquitination pathways. Both the proteosome and ubiquitination pathways were up-regulated specifically in DM, supported by up-regulation of the immunoproteasome subunit PSMB8 and nine ubiquitin E2 and E3 conjugating genes. Cytokines IL1B and IL18 have been shown to activate the inflammasome and ubiquitin-conjugating enzymes in myositis, although these downstream effects were only observed in adult DM (40–42).

There was a strong shared signature in DM and JDM, consisting of azurophilic (MPO, DEFA1/DEFA4, BPI, AZU1, ELANE, PRTN3) and secondary neutrophilic granules (LTF, MMP8/MMP3, CTSG, LCN2, OLFM4), indicative of degranulation and NET formation. NETosis has previously been shown to correlate with disease activity and muscle damage in DM/JDM (43, 44).

Another downstream effect of inflammatory signaling involves elevated glycolysis, where up-regulated PKM was shown to induce skeletal muscle pyroptosis via the NLRP3 inflammasome, which also correlated with high IL1B (16). PKM was up-regulated among nine glycolytic enzymes: eight seen exclusively in DM, and one, PGAM1, shared between DM and JDM. Glycolysis was also the strongest up-regulated pathway among JDM treatment responders in a study comparing them to non-responders (12). Glycolysis is elevated by activated PI3K/AKT and MAP signaling, which stimulates cell proliferation and survival (45). Its activation is a sign of M1 macrophage and dendritic cell proliferation, and is required for neutrophil degranulation and NETosis. Although the specific signaling and glycolytic components varied between DM and JDM, they shared a common signature of cell growth, highlighted by up-regulation of anti-apoptotic marker MCL1 at both the transcript and protein level. Together, the enhanced glycolytic pathway, PI3K/MAPK signaling, and Bcl-2 components suggest increased cell proliferation and survival are findings shared across DM and JDM.

Several core signaling pathways including type I IFN, MAPK and PI3K/AKT signaling, NETosis, glycolysis, and Bcl-2 apoptotic signaling have been activated in other autoimmune and inflammatory diseases (46). However, preceding studies suggest the constellation of pathways, along with the direction of change in specific genes, may define a unique signature of DM and JDM. Although type I IFN signature has been observed in the peripheral blood from patients with both DM and SLE, the specific genes involved and magnitude of up-regulation differ between these conditions (47, 48). The DM/JDM changes seen in this study need to be compared to those seen in other autoimmune diseases to determine if they are distinctive.

An important benefit of proteomics integration was the specific detection of certain protein classes not found at the transcript level. The transcriptome array provided genome-wide coverage, yet proteomics data identified significant expression changes in 253 additional proteins, 88 of which were secreted proteins. The proteomics protocol also included fluid phase proteins, an advantage for peripheral blood studies where important immune cytokines may not be detectable by transcriptomics, even with single cell resolution (49). Together the results represent an expansion of transcriptomic pathways with proteomics leading to insights only possible with a study designed for this analytical approach.

The current study has several limitations. Relatively small numbers of DM and JDM patients were examined, and at only one time point, which prevents the assessment of long-term changes in gene regulation. Most patients had only low to moderate disease activity and also received several years of immunosuppressive therapy, which may have dampened the signal of other pathways, including type I IFNs. However, this patient group represents the state of patients undergoing standard-of-care therapy, and the consistent gene regulation signatures across patients in each group suggests these patients reached a relatively steady state. Significant gene changes in JDM patients overall were substantially lower than adult DM. The JDM patients were receiving higher corticosteroid dosages, which may have reduced the ability to detect some alterations in gene regulation.

The study analyzed whole peripheral blood samples in which there may be variations in cellular composition; however, similar expression profiles have been described when comparing whole peripheral blood and muscle biopsies, including in treatment-refractory patients with active disease (8, 11). Prior myositis gene expression studies have shown broadly concordant changes between blood and affected tissues, however the degree of regulation and number of affected molecules has been diminished in blood compared to tissues (8)(48)(11). The use of peripheral blood may have reduced the ability to detect expression changes in affected muscle or skin, and may not represent all pathways regulated in DM/JDM patients. Although this study was not powered for comparisons across myositis autoantibody groups, the consistency of gene and protein regulation observed within each clinical subgroup is also supported by recent studies reporting relatively homogeneous regulation patterns across autoantibody groups (7, 50). These findings need confirmation in an independent patient cohort with a similar profile of treatment and disease activity.

The study design of combining transcriptomics and proteomics data and using multi-enrichment pathway analysis was quite powerful, as the results showed consistent gene and protein expression despite the limited sample size. The key altered pathways featured increased expression of PI3K/AKT, ERK, and p38 MAPK signaling in adult DM. Both DM and JDM shared elevation of upstream components of these pathways, as well as downstream effects on neutrophil degranulation, NETosis, and glycolysis. Upstream components of these signaling pathways may be potential therapeutic targets, that may include membrane receptors TLR2, TLR4, HAVCR2, EPOR, CSF2RA, and FGFR. Attenuation of these key signaling pathways may aid in improving control of active disease, and therefore should be a focus of future investigation.