All patients were recruited to the Juvenile Dermatomyositis National Cohort Biomarker Study and Repository for Idiopathic Inflammatory Myopathies (hereafter called the JDM Cohort study), a UK-wide multicentre cohort study and biobank for JDM through the UK JDM Research Group (JDRG) 15 . The JDM cohort study had ethical approval from the North Yorkshire Multi-Centre Research Ethics Committee and was also approved by the Steering Committee of the UK JDM Cohort Study. A total of 56 patients who fulfilled the Bohan and Peter criteria for definite or probable JDM 16 were included in this study. Eighteen children (6 boys, 12 girls, mean age 5.3 years), who had a muscle biopsy taken as part of the investigation of diverse symptoms, such as clumsiness and falls, were used as controls. As previously reported 17 , none of the controls received a diagnosis of myositis or myopathy and the biopsies were all reviewed by two expert histopathologists and confirmed to show normal muscle histology for age and, therefore, considered to represent suitable control samples. Ethical approval was obtained from the National Hospital for Neurology and Neurosurgery and the Institute Of Neurology Joint research ethics committee to use control muscle biopsies and serum from healthy controls. All patients (or their carers) provided consent to participate in the study.

Serum levels of the muscle enzymes lactate dehydrogenase (LDH) and creatine kinase (CK) were analysed by the Great Ormond Street Hospital clinical biochemistry department. Erythrocyte sedimentation rate (ESR) was measured by Great Ormond Street Hospital haematology department. Standardised outcome variables for the assessment of JDM, physician determined disease activity visual analogue score (VAS), childhood health assessment questionnaire (CHAQ), childhood myositis assessment score (CMAS) and parental global disease activity VAS 18 19 were collected prospectively as part of the UK JDM Cohort Study, as described 15 .

Serum was extracted within three hours of collection of peripheral blood and stored at −80°C. Serum concentrations of MRP8/14 were determined by sandwich enzyme-linked immunosorbent assay (ELISA) as previously described 13 . Monocyte chemoattractant protein-1 (MCP-1) levels were measured in serum using multiplex immunoassay (Meso Scale Discovery MSD technology, Meso Scale Diagnostics, USA). MCP-1 was measured in cell culture supernatants by ELISA (eBioscience, Hatfield, UK). IL-6, and in some experiments IL-1β, IL-10, IL-17, TNF-α and IFN-γ, measurements were performed using multiplex immunoassay 20 .

Immunohistochemistry and immunofluorescence staining were performed on biopsies from quadriceps (vastus lateralis) which were snap frozen within one hour and stored at −80°C. Staining was performed on acetone fixed 7 μm cryostat sections. Immunohistochemistry was performed as described 21 using anti-human monoclonal antibodies to CD68 (KP1), CD3 (UCHT1), MHC class l heavy chain (W6/32; Novacastra, Milton Keynes, UK) and MRP8/14 (27E10 22 ). Degree of CD3 and CD68 infiltration was scored as 0, 1 or 2 as described 21 : briefly <4 cells in ×20 field scored 0, >4 cells in ×20 and/or 1 cluster scored 1, >20 cells in ×20 and/or >2 clusters in whole biopsy scored 2. For double immunofluorescence staining the following anti-human monoclonal antibodies were used: CD68 (KP1; Novocastra,), CD14 (TUK4), CD15 (C3D-1; DakoCytomation, Cambridgeshire, UK), CD163 (RM3/1 23 ) and anti-human polyclonal MRP14 10 . Sections were incubated with primary antibodies at 4°C overnight followed by 30 minutes at room temperature with secondary antibodies sheep anti-mouse fluorescein isothiocyanate (FITC) (Sigma, Dorset, UK) and goat anti-rabbit alexa 568 (Life Technologies, Paisley, UK). Sections were mounted with Vectashield mounting medium with diamidino-2-phenylindole (DAPI) (Vector, Peterborough, UK) and analysed by fluorescent microscopy. For each patient, cells positive for lineage markers and MRP14 were counted in ten 10× fields. The mean percentage MRP14 cells expressing specific lineage markers were determined.

LHCNM2, a human skeletal myoblast cell line, was derived from the pectoralis major muscle of a 41-year-old male Caucasian heart-transplant donor, in accordance with local ethical legislation 24 . Myoblasts were grown in (Dulbecco’s) modified Eagle’s medium ((D)MEM) containing 4.5 mg/ml glucose + MEM 199 (at a ratio of 4:1), supplemented with 20% foetal calf serum (FCS), 100 IU penicillin, 100 μg streptomycin (GIBCO/Invitrogen, Paisley, UK). Myoblasts were cultured at 5 × 10⁴/ml in six-well plates at 37°C in 5% C0₂ and the medium was refreshed when cells reached 60% confluence. To test the effects of TLR4 agonists, myoblasts were stimulated with lipopolysaccharide (LPS) or recombinant human MRP8, MRP14 or MRP8/14 25 or left in culture medium alone. In some experiments myoblasts were pre-incubated for four hours with thapsigargin, an ER calcium-ATPase inhibitor which induces ER stress 26 . Thapsigargin was used at 0.05 μM, LPS at 100 ng/ml and MRP proteins at 5 μg/ml. Supernatants and cells were harvested as a time course over a 24-hour period. Supernatants were stored at −80°C for cytokine and chemokine measurements and cells were lysed in TRizol reagent (Invitrogen) for RNA extraction.

RNA from both myoblast cells and muscle tissue was isolated with TRIzol according to the manufacturer’s instructions. cDNA was synthesised using superscript II RT and random hexamers (Invitrogen). TLR4 expression on muscle tissue was analysed using Quantitect TLR4 primers (Qiagen, Crawley, UK). As a read out of ER stress, expression of x-box binding protein (XBP)-1 splice variants were analysed by PCR using the following primers: forward CGGAAGCCAAGGGGAATGAA, reverse CCCAACAGGATATCAGACTCTGA, at an annealing temperature of 62°C for 35 cycles. Primers to detect HPRT mRNA (Qiagen) were used as a housekeeping control for all PCR experiments. PCR products were visualised by 1.5% agarose gel electrophoresis. MCP-1 mRNA levels were quantified using real time PCR, 27 and HPRT gene expression was used for normalisation. cDNA was amplified using SYBR green (Bio-Rad Hertfordshire, UK) and thermo cycler rotor-gene 6000 Corbert (Qiagen). Data were analysed using Rotor-gene 6000 software (Qiagen).

Serum MRP levels, muscle enzymes and ESR were compared with clinical measures of disease activity using Spearman’s Rank. Where parametrically distributed, data were compared using 1 way or 2 way analysis of variance (ANOVA) and if appropriate paired t-tests. For non-parametric results, unpaired data were compared using the Mann–Whitney test. Differences in CD3/CD68 frequency were analysed by the Chi-Square test for trend. Results of P <0.05 were considered statistically significant. Analyses were carried out in SPSS version 18 (IBM, New York, USA).

The JDM patients in this study were representative of the whole JDM Cohort Study 15 and had evidence of moderately severe disease activity at the time of sampling with a median CMAS 33, physician’s global assessment 4.2, and CHAQ 1.3 and were analysed early in the course of their disease (median disease duration six months, Table 1 and Additional file 1: Table S1).

CK, creatine kinase; CHAQ, childhood health assessment questionnaire; CMAS, childhood myositis assessment score; ESR, erythrocyte sedimentation rate; LDH, lactate dehydrogenase.

Serum levels of the heterodimer MRP8/14 were significantly higher in JDM patients compared to age-matched controls but did not vary according to drug treatment (Figure 1A, B). Importantly, MRP8/14 levels strongly correlated with validated clinical measures of disease activity (Figure 1B-E), including physician global assessment of disease activity (PGA, R = 0.65, P = 0.0003) and strength/stamina (CMAS, R = −0.55, P = 0.004), and only to a limited extent with disability assessment (CHAQ, R = 0.56, P = 0.062) and the muscle enzyme, CK (R = 0.4, P = 0.046).

To test if infiltrating myeloid cells in myositis secrete MRP proteins, biopsies from 46 JDM patients were stained for MRP14 or the MRP8/14 heterodimer and compared to muscle biopsies from 14 age-matched controls. In JDM muscle the staining patterns observed suggested that these cells secrete MRP8/14 (Figure 2A, top right panel). The staining pattern of MRP8/14 in control samples in contrast was quite distinct (Figure 2A, bottom panel). Thus, control biopsies showed only very occasional myeloid cells, which although they stained positive for MRP8/14, showed no MRP8/14 protein secretion around the cells. To characterise MRP8/14-secreting cells in JDM biopsies, cells were analysed by dual colour immunofluorescence using polyclonal rabbit antibodies against MRP14 co-stained together with markers specific for cells of the myeloid lineage CD14, CD15 and CD68 as well as the scavenger receptor CD163 (Figure 2B). By immunohistochemistry, antibody to MRP14 reliably identified the same cells as those identified by antibody to MRP8/14 heterodimer (data not shown). The majority of MRP14 was secreted by CD68 positive cells (Figure 2B, left hand bar graph). The expression of the scavenger receptor CD163 on tissue macrophages has been described as a marker of anti-inflammatory or ‘alternately activated’ macrophages 28 and the proportion of CD163+ macrophages in muscle has been shown to increase after exercise 29 . Double staining in the JDM biopsies with MRP14 and CD163 showed that the majority of MRP8/14 secreting myeloid cells within the inflamed muscle were CD163- (Figure 2B right hand bar graph).

Examining muscle biopsies for myeloid sub-populations, we frequently observed a heavy infiltration of macrophages, often without a T cell aggregate (Figure 2C, top panels). In contrast biopsies with large numbers of T cells invariably had a significant macrophage infiltrate (Figure 2C, bottom panels). This observation raised the possibility that T cell recruitment could depend on the early macrophage infiltrate, which establishes the appropriate chemokine milieu to attract lymphocytes from the circulation. To quantify the relative abundance of macrophages and T cells in JDM muscle, we used our biopsy score tool 21 which includes a semi quantitative assessment of cell infiltration (macrophage infiltration and T cell infiltration, each scored as 0, 1, or 2). Analysis of these biopsies (n = 28) showed that muscle sections scored significantly higher for macrophage infiltration, than for T cells (P <0.001, Chi square test for trend, Figure 2D); indeed, 25% of these early biopsies scored 0 for T cell infiltrate. Interestingly, none of the biopsies assessed had T cell infiltration in the absence of a macrophage population, which suggests that myeloid cells may be more important in the early tissue inflammation of JDM than previously recognised.

Myoblasts are known to respond to pathogen associated molecular patterns (PAMPs), including bacterial LPS, by secreting inflammatory cytokines and chemokines 30 . Since MRP8 stimulates TLR4, similar to LPS, we hypothesised that tissue macrophages secreting MRP8/14 may propagate leukocyte recruitment in JDM by inducing skeletal muscle-derived cytokines. To test this hypothesis we first confirmed expression of TLR4 on human skeletal muscle from patients, controls and cultured human myoblasts (Figure 3A). The effects of MRP homo- and heterodimers on skeletal muscle were tested by culturing LHCNM2 cells in vitro in the presence of LPS or MRP proteins. LPS induced high levels of IL-6 and the chemokine MCP-1 from muscle cells, whilst MRP8 led to more modest but still significant increases compared to unstimulated controls (Figure 3B). IL-1β, IL-10, IL-17, TNF-α and IFN-γ were not detected in muscle supernatants after stimulation. We have previously shown that MRP8 is the active component in TLR4 ligation 11 : again in this muscle system, MRP14 and MRP8/14 did not significantly upregulate MCP-1 or IL-6 from myoblasts. As MRP8 and MRP14 are present at increased levels in the muscles of JDM patients (Figure 2A), we asked if this would lead to a corresponding increase in tissue levels of MCP-1. Examining JDM muscle, we found significantly higher mRNA expression of MCP-1 when compared to control muscle biopsies (0.3 versus 0.02 relative units, P = 0.0095, Figure 3C). Analysis of JDM serum also showed significantly higher levels of MCP-1, as well as IL-6 (Figure 3D and E, respectively), than samples taken from healthy controls.

Class I MHC over-expression in skeletal muscle (Figure 4A), which is an early pathological change in JDM, is known to be associated with ER stress, an established mechanism in the pathogenesis of myositis 7 9 31 . Recent reports have suggested that XBP-1, a key transcription factor involved in ER stress 32 , may act via an alternative pathway in monocytes and augment the secretion of TLR induced cytokines 33 . It is, therefore, important to understand how the interplay between ER stress and TLR stimulation, in the form of MRP8, would influence the secretion of inflammatory mediators from skeletal muscle. LHCNM2 cells were treated with thapsigargin, to induce a stress response which was confirmed by detecting expression of the active XBP-1 splice product (26 bp smaller), seen maximally between four to eight hours (Figure 4B). As previously shown, XBP-1 splicing is regulated by a negative feedback loop 34 , and 16 hours after exposure to thapsigargin, only the unspliced variant was detected. To test whether TLR ligation in muscle cells synergised with ER stress, LHCNM2 cells were cultured in vitro in the presence of the TLR4 agonist LPS, either alone or after pre-incubation with thapsigargin. ER stress clearly synergised with LPS to significantly increase IL-6 secretion from myoblasts when compared to stimulation with LPS alone (Figure 4C). MRP8, although acting in a TLR4 dependent manner, is less potent than LPS in activating downstream signalling cascades, including NF-κB 11 . Consequently, stimulation with MRP8 resulted in lower IL-6 levels when compared to LPS (Figure 4D), but there was still a trend for ER stress to augment the action of MRP8 (mean IL-6 114.2 pg/ml) when compared to MRP8 alone (72.8 pg/ml, P = 0.06). ER stress did not upregulate MCP-1 production following stimulation with either LPS or MRP8 (data not shown).