Aberrant Expression of High Mobility Group Box Protein 1 in the Idiopathic Inflammatory Myopathies

Introduction High Mobility Group Box Protein 1 (HMGB1) is a DNA-binding protein that exerts inflammatory or pro-repair effects upon translocation from the nucleus. We postulate aberrant HMGB1 expression in immune-mediated necrotising myopathy (IMNM). Methods Herein, we compare HMGB1 expression (serological and sarcoplasmic) in patients with IMNM with that of other myositis subtypes using immunohistochemistry and ELISA. Results IMNM (n = 62) and inclusion body myositis (IBM, n = 14) patients had increased sarcoplasmic HMGB1 compared with other myositis patients (n = 46). Sarcoplasmic HMGB1 expression correlated with muscle weakness and histological myonecrosis, inflammation, regeneration and autophagy. Serum HMGB1 levels were elevated in patients with IMNM, dermatomyositis and polymositis, and those myositis patients with extramuscular inflammatory features. Discussion Aberrant HMGB1 expression occurs in myositis patients and correlates with weakness. A unique expression profile of elevated sarcoplasmic and serum HMGB1 was detected in IMNM.


INTRODUCTION
The idiopathic inflammatory myopathies (IIMs) are a group of systemic autoimmune diseases characterized primarily by muscle inflammation, but also potentially accompanied by a range of extra-muscular manifestations. In adults, the term encompasses dermatomyositis (DM), polymyositis (PM), inclusion body myositis (IBM) and immune-mediated necrotising myopathy (IMNM; also called necrotising autoimmune myopathy, NAM). The etiology of these conditions remains obscure and the pathogenic mechanisms likely differ between the subtypes, given their distinct histopathological and immunological features. Immune-mediated necrotising myopathy has been only relatively recently described and the molecular mechanisms underlying the immune attack on muscle are poorly understood (Allenbach and Benveniste, 2013;Allenbach et al., 2018). It is becoming increasingly apparent that a dysregulated innate immune system contributes to the initiation and perpetuation of the IIMs, with roles for type I interferon (IFN), toll-like receptors (TLRs), various cytokines and the alarmin, High Mobility Group Box Protein 1 (HMGB1), now well-established (Day et al., 2017).
HMGB1 is a ubiquitous non-histone nuclear DNA-binding protein that can, under certain physiological and pathological conditions, undergo extra-nuclear translocation where it may act as a signal of tissue damage and a pro-inflammatory mediator (Lotze and Tracey, 2005). In response to injurious stimuli, inflammatory cells actively secrete HMGB1 in a controlled manner, which requires post-translational modification of the protein (Gardella et al., 2002;Lu et al., 2014). HMGB1 is also rapidly passively released from necrotic cells, following nuclear membrane breakdown (Scaffidi et al., 2002). This protein is implicated in a broad range of conditions, including sepsis, malignancy and autoimmune diseases (Harris and Raucci, 2006;Diener et al., 2013;Magna and Pisetsky, 2014). With regards to IIM, extra-nuclear HMGB1 expression has been demonstrated in the muscle of mice with experimental autoimmune myositis (Wang and Qin, 2016) and in muscle from IBM (Muth et al., 2015), PM, and DM (Ulfgren et al., 2004;Grundtman et al., 2010) patients. HMGB1 positive fibers in PM and DM muscle are non-necrotic, suggesting active release of HMGB1 from the muscle cell nucleus (Ulfgren et al., 2004). Of note, sarcoplasmic HMGB1 expression in IIM patients has not been confirmed in all reports (Cseri et al., 2015). Patients with new-onset DM and PM have elevated serum HMGB1, and these levels correlate with survival and the presence of interstitial lung disease (ILD) (Shu et al., 2016). In addition to descriptive research, a pathogenic role for HMGB1 in muscle is suggested by experimental studies. For instance, myocytes or myofibres exposed to recombinant HMGB1 demonstrate intracellular protein aggregation, increased cell death (Muth et al., 2015), aberrant MHC-I expression (Grundtman et al., 2010;Muth et al., 2015) and impaired calcium release during repeated tetanic stimulation, suggesting enhanced muscle fatigue (Grundtman et al., 2010;Zong et al., 2013).
However, HMGB1 is a multifaceted protein that exerts different effects depending on its redox state, the presence of post-translational modifications, the complexes it forms with other stimulatory or inhibitory proteins and the cellular receptor through which it ultimately signals. HMGB1 has been shown to induce muscle regeneration in mouse models of ischaemic myopathy (De Mori et al., 2007) and a non-oxidisable mutant form of exogenous HMGB1 promotes muscle and liver regeneration in mice via interaction with the CXCR4 receptor (Tirone et al., 2018). Cytosolic HMGB1 is a crucial regulator of autophagic responses to cellular stress, where autophagy is a beneficial physiological process enabling cellular proteins to be degraded and recycled (Tang et al., 2010). Within IIM muscle, HMGB1 co-localizes marker of autophagy (Cappelletti et al., 2014). As such, while HMGB1 appears to play a role in IIM pathophysiology and may have direct negative effects on muscle function, it may paradoxically aid in muscle restoration. The role of HMGB1 in IIM appears complex and clearly warrants further investigation. Moreover, research evaluating expression of this protein in IIM has focused on DM, PM and IBM and these discoveries may not be readily extrapolated to the condition of IMNM.
Herein we compare expression of HMGB1 in IMNM patients with that of other IIM patients. We correlate these findings with clinical, histopathological and serological parameters. To our knowledge, this is the largest cohort study evaluating HMGB1 in IIM and the first to describe sarcoplasmic and serum levels of HMGB1 in IMNM.

Subjects
Muscle tissue, serum and clinical data were obtained from the South Australian Myositis Database (SAMD), a registry of patients with PM, DM, IBM, non-specific IIM (NSIIM) and 'necrotising myopathy.' Recruitment to the SAMD is based on histological criteria which have previously been described for PM, DM, and IBM (Limaye et al., 2013). All cases of PM, DM and IBM adhered to published classification criteria (Hoogendijk et al., 2004;Rose and Group, 2013). Cases are recorded as 'necrotising myopathy' if there is myofibre necrosis and an absence of histological features consistent with other neuromuscular conditions. This was considered to be immunemediated if the treating clinician documented a diagnosis of IMNM or NAM. Cases of NSIIM have muscle inflammation with insufficient biopsy criteria to allow subclassification (e.g., scattered perimysial or endomysial inflammation that does not surround or invade myofibres). Muscle from 62 patients diagnosed with IMNM between 2001 and 2016, was included. Forty-five patients with a diagnosis of DM, PM, or IBM were included for comparison in addition to 15 patients with NSIIM and 17 controls. Control muscle constituted biopsies obtained from subjects with diffuse myalgia, weakness or unexplained creatine phosphokinase (CK) elevations, but which lacked any myopathic features. Control serum samples were collected from healthy volunteers. Some cases were excluded from this study due to lack of clinical information, alternative diagnosis or technical difficulties (Supplementary Figure 1).

Muscle Biopsy
Muscle samples were obtained via open surgical biopsy or needle biopsy. Specimens were placed into transverse orientation, mounted on cork, frozen in isopentane cooled using liquid nitrogen (Cash and Blumbergs, 1994) and stored in liquid nitrogen until sectioning. Consecutive 9 µm-thick cryostat muscle sections were placed on coated slides. H&E staining was performed on the first and last sections of each series. Sections intended for MHC I, MHCn, CD68, CD45 and LC3 staining were air dried for 30 min then stored at −80 • C until use (1 -68 days). Sections intended for HMGB1 staining were air dried for 30 min, fixed for 10 min in 10% neutral buffered formalin, air dried for 30 min then stored at −80 • C until use (1 -33 days). On the day of staining, antigen retrieval was achieved on slides intended for HMGB1 staining by immersion in sodium citrate buffer (pH 6.0) followed by heat induced epitope retrieval utilizing microwave treatment.
Positive controls were performed for each antibody in every staining procedure. A negative control was performed by omitting primary antibody in every staining procedure, both for an unfixed slide and a formalin-fixed microwave-retrieved slide. For ten cases, a formalin-fixed, antigen-retrieved section underwent staining with a rabbit IgG isotype control (Invitrogen, 31235), at the same concentration as the HMGB1 antibody. Lymphoid tissue was used to confirm HMGB1 staining in nonmuscle tissue.

Modified Gomori Trichrome Staining
Consecutive 9 µm-thick cryostat sections of unfixed fresh frozen IBM muscle, normal muscle and a positive "ragged red" control were stained with Harris' haematoxylin for 30 min, washed then stained with modified Gomori trichrome stain (laboratoryprepared, pH 3.4) for 15 min followed by differentiation with 0.2% aqueous acetic acid.

Quantification of Immunohistochemical Staining
Slides were graded using traditional microscopy for HMGB1, MHC I, LC3 and MHCn staining in a semi-quantitative manner by a muscle pathologist (SO) (Supplementary Table 1). The H&E stained sections were graded for degree of necrosis by the same pathologist, where necrosis was defined as muscle cells exhibiting a combination of the following features: swelling, hyalinization, hypereosinophilia, pallor, myophagocytosis. Slides were graded by twice on separate occasions; median grades are reported. A second trained investigator (JD) validated the grading scale by manually counting positive and negative myofibres in 10 randomly selected high power fields (HPFs, magnification × 400) and calculating the percentage of HMGB1+ fibers. Manual cell counts of CD45+ leucocytes and CD68+ macrophages were also performed. Evaluation of each histological parameter was completed for the entire cohort before proceeding to the next parameter. At the time of histological evaluation, investigators were blinded to the clinical details and the grades assigned to other immunoproteins.

Serum Collection and Analysis
A commercial ELISA kit was used to measure serum concentrations of HMGB1 (Cloud-Clone Corporation, Texas, United States). Each sample was diluted 1/100 in phosphate buffered saline and tested in duplicate. One sample was re-tested on each ELISA plate.

Statistics
Statistical analysis was performed using STATA version 14.0. Values were expressed as the median and the interquartile range (IQR). Two group comparisons were performed using the Mann-Whitney U test. When three or more groups were compared, a Kruskal-Wallis H test was conducted to identify whether a statistically significant difference existed, followed by a post hoc Dunn's test. A Bonferroni correction was applied for multiple comparisons. Fisher's exact test was used to analyze categorical data. Spearman correlations were performed to analyze associations between radiological grades and continuous or ordinal parameters. Number of cases analyzed are indicated if a full data set was unavailable. P-values < 0.05 were considered significant.

Immunohistochemistry Grading Reliability
Intra-rater reliability was high (κ > 0.70) for all histological parameters. The average percentage of HMGB1+ myofibres correlated strongly with grades assigned by the muscle pathologist (r s 0.83, p < 0.01).
Sarcoplasmic HMGB1 levels are elevated in muscle from IIM patients but levels differ according to disease subtype HMGB1 sarcoplasmic immunostaining was low grade and stippled in histologically normal control muscle ( Figure 1A) and comparatively strong in IIM patients ( Figure 1B). As expected, muscle nuclei and infiltrating immune cells were strongly HMGB1 positive (e.g., Figure 1B). Sarcoplasmic expression was highest in IMNM followed by IBM and PM (Figure 2). Both IBM and IMNM patients exhibited significantly elevated levels of sarcoplasmic staining compared to those with DM and NSIIM.

Sarcoplasmic HMGB1 expression correlates with multifactorial processes in IIM
Sarcoplasmic HMGB1 grades correlated strongly (r s 0.62 -0.77, p < 0.01) with the degree of muscle cell necrosis for all IIM subtypes except NSIIM, suggesting that necrosis is an important driver of sarcoplasmic HMGB1 staining even in those subtypes where this is not the dominant histological feature. Both macrophage and CD45+ leucocyte infiltration were strongly associated with sarcoplasmic HMGB1 expression (r s 0.63, 0.61, p < 0.001), although these correlations were less robust and not significant for IBM patients. In fact, for IBM patients, the degree of MHCn+ (regenerating) fibers (r s 0.77, p < 0.001) and LC3+ staining (r s 0.75, p < 0.001) correlated most strongly with HMGB1 expression. On a cellular level, these processes colocalized within individual muscle fibers. Many necrotic cells exhibited strong positive staining for HMGB1 (Figures 3A,B). Regenerating fibers were frequently present in IIM muscle and always stained positively for HMGB1 (Figures 3C,D). In IBM, abnormal HMGB1 staining was visualized in fibers exhibiting features of abnormal cytoplasmic protein inclusions, vacuolar change and autophagic protein accumulation (Figures 3E-H). Isotype control staining was negative (Supplementary Figure 2). Only occasionally was HMGB1 positivity noted in relatively normal appearing mature myofibres. This typically occurred in areas of inflammation and might reflect active secretion of HMGB1 by activated myofibres. Analysis of sarcoplasmic HMGB1 levels by individual autoantibodies was limited by small numbers. Sarcoplasmic HMGB1 levels were high in all anti-SRP+ IMNM patients (Grade ≥ 2, n = 5) whereas a range of staining patterns were observed in anti-Ro+ IIM, antisynthetase+ IIM and anti-HMGCR+ IIM.

Reduced sarcoplasmic HMGB1 staining in some IIM patients may reflect corticosteroid exposure
We found a modest negative correlation between cumulative corticosteroid dose and sarcoplasmic HMGB1 expression (r s −0.30, p < 0.01). Patients with inflammatory arthritis had lower muscular HMGB1 staining than those without arthritis (p < 0.05). These patients also had higher cumulative prednisolone exposure (325 mg vs. 0 mg, p = 0.04), which may explain the comparatively reduced sarcoplasmic HMGB1 expression.

Sarcoplasmic HMGB1 correlates with muscle weakness
Twenty-four IIM patients had MMT8 assessments at the time of muscle biopsy and there was a strong negative correlation between strength and sarcoplasmic HMGB1 levels. This was true for IMNM (r s −0.57, p = 0.03, n = 14) and non-IMNM IIM patients (r s −0.75, p = 0.01, n = 19). Sarcoplasmic HMGB1 expression consistently correlated with every bedside clinical disease activity index tested (Figure 4). There was a modest correlation with serum CK level (r s 0.31, p < 0.01, n = 95). There were no correlations between HMGB1 grades and indices of  (  disease-related damage (MDI scores) or patient disability (HAQ scores). Sarcoplasmic HMGB1 expression modestly negatively correlated with symptom duration in IMNM patients (r s −0.38, p < 0.01, n = 50) but trended toward a positive correlation in IBM patients (r s 0.62, p = 0.07, n = 9), likely reflecting that different processes are responsible for HMGB1 expression in these diseases.

Serum HMGB1 levels are elevated in patients with PM, DM, and IMNM
Serum HMGB1 was elevated in IIM compared with healthy controls (p < 0.001), however, levels varied markedly by IIM subtype ( Figure 5A). As such, patterns of serum HMGB1 expression differed from that observed within muscle. For instance, despite exhibiting notable sarcoplasmic staining, circulating levels of HMGB1 in IBM patients did not differ from controls. Conversely, high serum HMGB1 was detected in DM patients despite these patients exhibiting low sarcoplasmic levels. Patients with IMNM had notably high intramuscular and high serum HMGB1 levels. There were no significant differences in serum HMGB1 levels between IIM patients who had serum collected early in the disease process (within 6 months of diagnosis) and those whose serum collection was more delayed (Figure 5B).  There was no correlation between serum HMGB1 levels in IIM patients and symptom duration. Sarcoplasmic and serum levels of HMGB1 did not correlate in the 58 patients with both samples available, possibly owing to the multifactorial nature of intramuscular HMGB1 expression and because these samples were collected at different time-points of the disease trajectory.

Extramuscular features are associated with elevated serum HMGB1 levels
High serum HMGB1 levels were observed in IIM patients with RP, ILD and inflammatory joint disease ( Table 2). Myalgia was common and was also associated with significantly elevated serum HMGB1. Myalgic patients also had more polyarthralgia (10/33, 30% versus 4/36, 11%, p = 0.046) compared with nonmyalgic IIM patients, but did not exhibit higher serum CK, myonecrosis or muscle inflammation. Together this suggests that myalgia may be systemically driven or reflect joint inflammation, rather than intramuscular pathology per se.
Patients with anti-Ro52+ antibodies had significantly elevated serum HMGB1 levels ( Table 2). We did not observe any association between circulating HMGB1 levels and muscle disease activity measures, antisynthetase antibodies or the presence of cancer.

DISCUSSION
Understanding the molecular events underpinning IIM pathophysiology and how this differs between subtypes is critical in the pursuit of developing targeted therapies, which is increasingly the goal in rheumatological practice. Herein we evaluated expression of HMGB1 in all forms of IIM, including subtypes in which levels of expression of this protein have been hitherto unknown, such as IMNM and NSIIM. Structures demonstrating HMGB1 expression included infiltrating immune cells, necrotic myofibres and those exhibiting regeneration, autophagy and mitochondrial dysfunction. We additionally observed strong HMGB1 staining in all muscle nuclei and low-grade staining in our histologically normal controls. Given small amounts of HMGB1 are present cytosol under normal cellular conditions (Kuehl et al., 1984;Tang et al., 2010), the sarcoplasmic HMGB1 observed in our control tissue is likely physiological. As such, we have shown HMGB1 to be associated with multiple processes in the muscle microenvironment ranging from physiological to pathological, damaging to restorative. The relative balance and temporal evolution of these processes may explain the varying degrees of HMGB1 expression we observed across IIM subtypes. While it seems paradoxical that one protein could be associated with multiple processes, it is in keeping with the complex functional properties of HMGB1. The biological actions of HMGB1 are strikingly diverse and evolve over time owing to its unique biochemistry, its ability to undergo post-translation modifications, complex with other proteins and signal through a multitude of receptors. As others have emphasized (Magna and Pisetsky, 2014), HMGB1 should be conceptualized as an ensemble of proteins rather than a single species and with a fixed structure or function.
Overall, our data imply a deleterious role for HMGB1 in IIM, as sarcoplasmic expression to correlated with clinical disease activity and histological inflammation and necrosis. This is consistent with evidence demonstrating HMGB1 to have pro-inflammatory properties and correlate with inflammatory disease activity. In addition to cytokine and chemokine properties, HMGB1 can activate the classical pathway of complement (Kim et al., 2018), where complement deposition is a key mediator of myonecrosis (Engel and Biesecker, 1982). It is conceivable that HMGB1 released from necrotic myofibres could trigger further local muscle cytolysis and perpetuate further damage in a self-sustaining process. Therapeutic blockade of HMGB1 in inflammatory disease states in is under consideration and been evaluated in experimental models of drug-induced liver injury (Lundback et al., 2016) and inflammatory arthritis (Kokkola et al., 2003), with promising results.
However, data herein and elsewhere links HMGB1 with ostensibly beneficial processes such as myofibre regeneration (De Mori et al., 2007;Tirone et al., 2018) and metabolic functions such as autophagy (Muth et al., 2015;Tang et al., 2010). Accelerated tissue regeneration is observed following administration of recombinant HMGB1 in mouse models of muscle injury (De Mori et al., 2007). This myoregenerative effect is particularly pronounced when HMGB1 is administered in a fully reduced isoform (Tirone et al., 2018). As such, promoting certain HMGB1 pathways therapeutically might accelerate tissue repair in various clinical scenarios. Development of therapeutics that specifically inhibit isoforms of HMGB1 contributing to inflammatory pathology or promote those isoforms involved in reparative processes would clearly be the most desirable strategy. Of note, regenerating myofibres have been implicated in the elaborate pathophysiological mechanisms that underpin IIM (Tournadre and Miossec, 2013), and attempts to enhance myoregeneration with exogenous HMGB1 may not be prudent in these diseases. Further research evaluating the role of specific HMGB1 isoforms in IIM and the role of regenerating myofibres in disease perpetuation is clearly required before therapeutic intervention exploiting HMGB1 pathways can be considered. We observed elevated serum HMGB1 levels in IMNM, PM and DM patients, however, the source of circulating HMGB1 may differ between subtypes. In IMNM, this likely reflects rapid, passive release of HMGB1 into the extracellular space due to myonecrosis. Indeed, HMGB1 has been used as a marker of necrosis in experimental studies of tumor pathophysiology (Jeon et al., 2013;Kang et al., 2014). However, HMGB1 can also be released from activated immune cells present in inflamed tissues; this may explain the association between HMGB1 levels in serum and the presence of extra-muscular autoimmune manifestations, such as RP, joint disease and ILD. Previous studies have demonstrated elevated HMGB1 in the serum and/or broncho-alveolar fluid of IIM-related ILD and other inflammatory fibrotic lung conditions (Ebina et al., 2011;Shu et al., 2016;Ying et al., 2017;Shimizu et al., 2018). These results support a clinical role for measuring serum HMGB1 levels; this could supplement muscle biopsy in the subtyping of IIM and, potentially, screening for IIM-related ILD (Shu et al., 2016). Our findings suggest that this assessment will have discriminatory utility even in patients with well-established disease who are receiving immunomodulatory therapy, although further studies evaluating the sensitivity and validity of this minimally invasive test are required. Importantly, HMGB1 that is actively secreted by inflammatory cells undergoes critical post-translational modifications (acetylation) of 'nuclear localization sites' (NLSs) contained within the protein in order to exit the nucleus (Lotze and Tracey, 2005). Conversely, passive release does not involve NLS modification and thus necrotic cells do not generate hyperacetylated HMGB1 (Yang et al., 2015). An assay that differentiates these HMGB1 isoforms could allow clinicians to quantify the degrees of necrosis versus inflammation in individual patients, and could conceivably aid in IIM subtyping. The oxidation state of HMGB1 also differs according to the mechanism of cellular release but, considering this can rapidly alter in the extracellular milieu , assays determining the degree of NLS acetylation may have more discriminatory value. Unfortunately, HMGB1 isoform analysis is challenging, requires high-end mass spectrometry instrumentation and is currently successfully performed by only one research group worldwide on a collaborative basis (Yang et al., 2012). Development of further reliable isoform assays would advance scientific understanding regarding the role of HMGB1 isoforms in disease pathogenesis, information vital for clinical translation of therapeutics targeting HMGB1.
This study has several limitations. It is descriptive in nature and the complex mechanisms underpinning the associations we have observed cannot be determined. We did not perform HMGB1 receptor staining or have access to HMGB1 isoform analysis, which would have provided added insights. Our sample size was small, owing to the rare nature of these disorders. However, this is a large study evaluating HMGB1 expression in IIM and the first to describe elevated levels in the muscle and serum of IMNM patients. These important findings may inform future critical mechanistic studies regarding the role of HMGB1 in autoimmune muscle disorders.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
This study was approved by the Central Adelaide Local Health Network Ethics Committee, Adelaide, Australia. We confirm that we have read Muscle and Nerve's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

AUTHOR CONTRIBUTIONS
JD: hypothesis revision, conducting experiments, acquiring and analyzing data, writing manuscript, and acquiring funding. VL: SAMD custodian, hypothesis initiation and revision, acquiring data, manuscript revision, acquiring funding, and providing materials. JH: hypothesis initiation and revision, providing funding, manuscript revision, and providing materials. SP: hypothesis revision, manuscript revision, acquiring data, and providing funding. PE: hypothesis revision, experimental design, and laboratory supervision of JD. PH and SO: hypothesis revision, acquiring data, and providing materials. KC: hypothesis revision, experimental design, conducting experiments, and laboratory supervision of JD. All authors: approval of the final manuscript.