Article Text

Download PDFPDF
Monocyte and bone marrow macrophage transcriptional phenotypes in systemic juvenile idiopathic arthritis reveal TRIM8 as a mediator of IFN-γ hyper-responsiveness and risk for macrophage activation syndrome
  1. Grant S Schulert1,2,
  2. Alex V Pickering3,
  3. Thuy Do1,
  4. Sanjeev Dhakal1,
  5. Ndate Fall1,
  6. Daniel Schnell4,
  7. Mario Medvedovic2,
  8. Nathan Salomonis2,4,
  9. Sherry Thornton1,2,
  10. Alexei A Grom1,2
  1. 1 Rheumatology, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, USA
  2. 2 Pediatrics, University of Cincinnati College of Medicine, Cincinnati, Ohio, USA
  3. 3 Harvard Medical School, Boston, Massachusetts, USA
  4. 4 Biomedical Informatics, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, USA
  1. Correspondence to Dr Grant S Schulert, Rheumatology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH 45229, USA; grant.schulert{at}cchmc.org

Abstract

Objectives Systemic juvenile idiopathic arthritis (SJIA) confers high risk for macrophage activation syndrome (MAS), a life-threatening cytokine storm driven by interferon (IFN)-γ. SJIA monocytes display IFN-γ hyper-responsiveness, but the molecular basis of this remains unclear. The objective of this study is to identify circulating monocyte and bone marrow macrophage (BMM) polarisation phenotypes in SJIA including molecular features contributing to IFN response.

Methods Bulk RNA-seq was performed on peripheral blood monocytes (n=26 SJIA patients) and single cell (sc) RNA-seq was performed on BMM (n=1). Cultured macrophages were used to define consequences of tripartite motif containing 8 (TRIM8) knockdown on IFN-γ signalling.

Results Bulk RNA-seq of SJIA monocytes revealed marked transcriptional changes in patients with elevated ferritin levels. We identified substantial overlap with multiple polarisation states but little evidence of IFN-induced signature. Interestingly, among the most highly upregulated genes was TRIM8, a positive regulator of IFN-γ signalling. In contrast to PBMC from SJIA patients without MAS, scRNA-seq of BMM from a patient with SJIA and MAS identified distinct subpopulations of BMM with altered transcriptomes, including upregulated IFN-γ response pathways. These BMM also showed significantly increased expression of TRIM8. In vitro knockdown of TRIM8 in macrophages significantly reduced IFN-γ responsiveness.

Conclusions Macrophages with an ‘IFN-γ response’ phenotype and TRIM8 overexpression were expanded in the bone marrow from an MAS patient. TRIM8 is also upregulated in SJIA monocytes, and augments macrophage IFN-γ response in vitro, providing both a candidate molecular mechanism and potential therapeutic target for monocyte hyper-responsiveness to IFNγ in cytokine storms including MAS.

  • arthritis
  • juvenile
  • inflammation
  • cytokines

Data availability statement

Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as online supplemental information. Bulk and single-cell RNA-seq datasets have been deposited in gene expression omnibus (GSE147608 and GSE147795, respectively).

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known about this subject?

  • Children with systemic juvenile idiopathic arthritis (SJIA) are at risk for the cytokine storm macrophage activation syndrome (MAS). Hyper-responsiveness of SJIA monocytes to interferon (IFN)-γ is a key driver of MAS, but the molecular mechanisms that promote this are unknown.

What does this study add?

  • SJIA monocytes display both proinflammatory and anti-inflammatory properties, in an attempt to compensate for systemic hyperinflammation.

  • Overexpression of the IFN regulator tripartite motif containing 8 (TRIM8) distinguishes circulating monocytes in SJIA and haemophagocytic bone marrow macrophage subpopulations in one SJIA patient with early MAS.

  • TRIM8 increases macrophage responsiveness to IFN-γ, the pivotal cytokine in MAS, and thus may promote this complication in SJIA.

How might this impact on clinical practice or future developments?

  • TRIM8 represents both a molecular mechanism and novel therapeutic target for monocyte responsiveness to IFN-γ in cytokine storms including MAS.

Introduction

Systemic juvenile idiopathic arthritis (SJIA) is the most severe subtype of JIA, notable for marked systemic immune activation with features of autoinflammation.1 2 The pathophysiology of SJIA is driven by continuous activation of innate immune pathways especially by the monocyte/macrophage lineage, although the precise cellular source of the pivotal IL-1 and IL-6 cytokines in SJIA remains unclear.3 4 Gene expression signatures in peripheral blood mononuclear cells (PBMCs) during active SJIA reveal increased expression of monocyte and macrophage activation markers, genes induced by TLR/interleukin (IL)-1 signalling pathways, and genes involved in negative regulation of innate inflammatory responses.5–7 About 15% of SJIA patients will also develop macrophage activation syndrome (MAS), a life-threatening episode of hyperinflammation driven by excessive activation and expansion of T cells and haemophagocytic macrophages.8–12 This cytokine storm leads to extreme hyperferritinaemia, cytopenias, liver dysfunction and coagulopathy.10 12 For reasons poorly understood, while widespread use of biologics targeting IL-1 and IL-6 has markedly improved overall disease control, children with SJIA remain at risk for MAS.12–14

MAS bears close clinical resemblance to haemophagocytic lymphohistiocytosis (HLH), a constellation of life-threatening cytokine storm syndromes due to both primary HLH (pHLH) and secondary acquired causes.15 16 pHLH is a group of rare disorders linked to genetic defects affecting the perforin-mediated cytolytic pathway.15 MAS in SJIA is widely viewed as a distinct form of secondary HLH occurring in the setting of inflammatory and rheumatic disorders.12 Interestingly, up to 40% of SJIA patients who develop MAS carry hypomorphic mutations in pHLH genes.17 18 Substantial evidence in pHLH supports interferon (IFN)-γ blockade as novel therapy for this cytokine storm,19 and the anti-IFN-γ antibody emapalumab has been recently approved for this condition.20 Interestingly, while IFN-γ does not play a major role in the pathogenesis of SJIA itself,21 in several studies the development of MAS in SJIA patients paralleled activation of IFN-induced pathways in monocytes, and distinguished acute MAS versus a conventional flare of SJIA.22 23 These observations combined with the fact that neutralisation of IFN-γ reverted MAS in a murine model,24 led to the phase II clinical trial of emapalumab in MAS complicating SJIA (NCT 03311854), the preliminary results of which are promising.25 IL-18 has been identified as another key cytokine in MAS pathophysiology, with elevated IL-18 distinguishing both SJIA and adult-onset Still’s disease and associated with risk for MAS, presumably through augmenting IFN-γ production.4 26 27

Another intriguing observation by us21 and others28 is that monocytes in SJIA exhibit hyper-responsiveness to IFN-γ in vitro that may be further exaggerated by IL-1 and IL-6 inhibiting biologics, which could explain the persistent risk for MAS in SJIA treated with these agents. The mechanistic reasons for such hyper-responsiveness remain unclear, but may be determined by the subtype and polarisation status of monocytes and macrophages.29 While previously considered as a dichotomy between classically activated ‘M1’ and alternatively activated ‘M2’ macrophages, recent work has shown that macrophages are activated towards a diverse spectrum of distinct polarisation phenotypes.30 Previous cell-surface immunophenotyping has demonstrated that monocytes in SJIA do not align with a single polarisation state, but rather exhibit features reflecting multiple activation phenotypes.31 32

The objective of this study was to further characterise the polarisation phenotype of both circulating monocytes and bone marrow macrophages (BMM) from SJIA patients using transcriptional profiling, to identify factors that may influence cellular responsiveness to IFN-γ and risk for development of cytokine storm.

Methods

Patients and peripheral blood samples

Written informed parental consent was obtained for each subject prior to participation, and child assent was obtained where appropriate. Fresh whole blood was collected from SJIA patients (table 1 and online supplemental table S1) with active new onset or established disease; clinically inactive disease (CID) as defined by the Wallace criteria,33 and healthy age-matched controls obtained separately from children undergoing routine phlebotomy. All enrolled patients satisfied the ILAR classification criteria for SJIA.1 MAS patients met the 2016 MAS classification criteria.11 Monocytes were isolated as described.32

Table 1

Summary of clinical and laboratory characteristics for patients in this study

In vitro macrophage polarisation

THP-1 human monocytic cell line (American Type Culture Collection (ATCC)) and primary human monocytes was maintained in RPMI and polarised as described.32

Bulk and single-cell RNA-seq gene expression profiling

Methods regarding bulk RNA-seq analysis of peripheral blood monocytes and single-cell RNA-seq analysis of BM macrophages are in online supplemental methods.

Tripartite motif containing 8 knockdown via siRNAs

THP-1 was transfected with either ON-TARGETplus siRNA against tripartite motif containing 8 (TRIM8) (Dharmacon) or ON-TARGETplus Non-targeting Pool (Dharmacon) using our established protocol.34 TRIM8 primers were: ‘Forward’: 5’-CCTATCTGCCTGCACGTTTT-3’; ‘Reverse’: 5’-GTTGTAGGCCTGGTTGCACT-3’. Primers for GAPDH have been previously reported.32

Monocyte responsiveness to IFN-γ

STAT1 phosphorylation assays were performed as described.21 CXCL9/CXCL10 primers have been previously reported,32 and expression was assessed by RT-PCR.

Results

SJIA monocyte transcriptomes reflect multiple polarisation states but lack prominent features of IFNγ-response

Bulk RNA-seq of purified peripheral blood monocytes was performed in 26 patients with SJIA (table 1). This revealed marked transcriptional changes between cells from SJIA patients and healthy controls, regardless of disease activity (figure 1A). Control samples as a group did show some increase in mitochondrial reads and decrease in ribosomal reads, which may reflect some differences in cell quality from patients. However, pathway analysis revealed that the most enriched gene ontology pathways among upregulated genes in SJIA patients included those involved in immune system processes (p=21.9×10-46) and myeloid-mediated immunity (5.08×10–43) (figure 1B). Although not among the top 500 enriched pathways, pathways reflecting response to pro-inflammatory cytokines including IL-1 (p=1.04×10-7), TNF (6.94×10–6), and IFN-γ (4.95×10–5), but not IL-6 or IFN-β, were also significantly enriched. Together these data suggest that SJIA monocytes broadly exhibit altered transcriptional activity reflecting an activated phenotype. Strikingly, no clear separation was observed between monocytes from SJIA patients with active disease versus those with CID (figure 1A).

Figure 1

Differential gene expression in freshly isolated peripheral blood monocytes from patients with systemic JIA (SJIA) versus healthy controls. (A) Distance matrix based on all transcripts where red, yellow and grey colours indicate patients with active SJIA, inactive SJIA, and healthy controls respectively. Colours represent distance matrices calculated by computing the euclidean distance between all sample pairs. (B) Over-representation analyses of gene ontology terms for the upregulated (top) and downregulated (bottom) differentially expressed genes. JIA, juvenile idiopathic arthritis.

Since this comparison considered as the ‘active SJIA’ group a highly heterogeneous collection of all patients with variable disease duration who failed to meet the Wallace criteria for CID,33 we next examined more specific markers of inflammatory activity. Patients with SJIA and particularly those with features of MAS are characterised by hyperferritinaemia. Stratifying SJIA patients with high (≥210 ng/mL) vs normal serum ferritin levels (a cut-off that best paralleled SJIA patient clustering in our previous gene expression study6), showed clear separation of monocytes into two groups, one including 7/8 ‘high ferritin’ samples and the other exclusively ‘normal ferritin’ samples (figure 2A). Interestingly 4/6 ‘normal ferritin’ samples that clustered with the ‘high ferritin’ group had mild elevations in inflammatory markers without hyperferritinaemia. We also noted that there was significant overlap between the ‘high ferritin’ samples and untreated, new-onset SJIA (6/8 ‘high ferritin’ patients; online supplemental table S1). Differential expression analysis revealed 686 upregulated and 418 downregulated genes between ‘high ferritin’ and ‘normal ferritin’ monocytes (figure 2B and online supplemental table S2). The gene set enrichment analysis revealed upregulation of pathways including immune response (4.76×10–44), vesicle-mediated transport (5.06×10–41), myeloid cell activation (3.42×10–40) and secretion (1.26×10–35) (figure 2C). While not among the top 200 enriched pathways, the ‘high ferritin’ signature did show mild enrichment in pathways reflecting response to both IL-1 (3.6×10–3) and IFN-γ (1.94×10–6), supporting the association between elevated ferritin and MAS. There was no significant enrichment in pathways reflecting specific polarisation phenotypes. Among the downregulated genes there were no pathways with adjusted p<0.1. Together, this suggests that monocytes from SJIA patients with elevated ferritin show proinflammatory transcriptional activation.

Figure 2

Analysis of the genes differentially expressed in freshly isolated peripheral blood monocytes from SJIA patients with high versus normal serum ferritin levels. The list of differentially expressed genes were generated using limma moderated t-tests, with FDR 5%. (A) Distance matrix based on all transcripts where red and yellow colours indicate patients with active SJIA and inactive SJIA, respectively, with colours representing distance matrices calculated by computing the euclidean distance between all sample pairs. Dark orange and light orange colours indicate patients with high and normal ferritin levels, respectively. (B) Hierarchical clustering of genes differentially expressed between high and normal ferritin groups. The complete linkage clustering algorithm, in which distance is a measure of similarity, was used to generate the hierarchical clustering tree. In this tree, each row represents a separate gene and each column represents a separate individual. Dark orange, light orange and grey colours indicate patients with high ferritin, normal ferritin and healthy controls, respectively. The scaled normalised expression level for each gene is indicated by colour. (C) Over-representation analysis of gene ontology terms for genes differentially expressed between high and normal ferritin groups (D) Polarisation signature analysis that reflects overlapping patterns of gene expression between SJIA monocytes and in vitro polarised M(LPS+IFN-γ), M(IL-4), M(LPS+IC) and M(IL-10) macrophages. (E) Expression levels of IL18 and IL18BP in peripheral blood monocytes from patients with high ferritin (n=8), normal ferritin (n=18) and controls (n=11). **P<0.01. ***p<0.001 by ANOVA with follow-up Dunnett’s multiple comparisons test. ANOVA, analysis of variance; IFN-γ, interferon-γ; IL-8, interleukin 8; SJIA, systemic juvenile idiopathic arthritis.

Previous work has suggested that monocytes in SJIA display features of multiple polarisation phenotypes. To further characterise the polarisation properties of SJIA monocytes, we first empirically determined transcriptional signatures of primary monocytes from healthy individuals polarised towards well described in vitro phenotypes to generate M(LPS+IFN-γ), M(IL-4), M(LPS+immune complexes (IC)) and M(IL-10) signatures (online supplemental figure S1 and table S3). When comparing the ‘high ferritin’ SJIA monocyte signature to these empirically determined polarisation signatures, we found the highest enrichment in the alternatively activated M(IL-4) and M(IL-10) signatures, with less enrichment with classically activated M(LPS+IFN-γ) signature. Together, these demonstrate that monocytes in SJIA reflecting either a mixed polarisation phenotype, or multiple distinct cell populations (figure 2D).

Active SJIA is also associated with markedly elevated levels of serum IL-18.4 26 27 Since increased free (unbound) serum IL-18 is proposed to promote MAS by enhancing IFN-γ production, we assessed expression of both IL-18 and its natural antagonist IL-18 binding protein (IL-18BP). As shown in figure 2E, compared with healthy controls, SJIA monocytes were expressing significantly higher levels of IL18 and significantly lower levels of IL18BP.

Elevated expression of IFN gamma receptors and TRIM8

Since we and others have shown that SJIA monocytes show increased responsiveness to IFN-γ, we then examined these signatures to identify factors that could modulate IFN signalling. Both ‘high ferritin’ and ‘normal ferritin’ SJIA monocytes expressed significantly higher levels of IFN-γ receptors (IFNGR1 and IFNGR2) compared with monocytes from healthy controls (figure 3A). Increased expression of IFN receptors has been previously shown to contribute to increased IFN responsiveness in human monocytes and macrophages.29 35 To further explore this finding, we examined protein levels of IFNGR (CD119) on the surface of monocytes from patients with active SJIA. As shown in figure 3B,C, we find significantly increased surface expression of CD119 on SJIA patient monocytes, compared with control monocytes.

Figure 3

Differential expression of genes modulating IFNγ signalling pathway. (A) Normalised expression levels of IFNGR1, IFNGR2 and TRIM8 in peripheral monocytes from SJIA patients in the high-ferritin group (n=8), normal-ferritin group (n=18) and healthy controls (n=11). Expression levels of each individual gene for each patient were normalised against the mean expression level in the entire set of samples. ***P<0.001 by ANOVA with follow-up Dunnett’s multiple comparisons test. (B) Representative histograms showing CD119 intensity in monocytes from control (red) and patients with active SJIA (blue) as determined by flow cytometry. (C) Median fluorescence intensity of CD119 as determined by flow cytometry in control and SJIA patient monocytes, pooled from three independent samples. *P<0.05 by t-test. (D) Schematic representation of IFNγ signalling pathway. TRIM8 degrades SOCS1 and PIAS3 levels, both of which negatively regulate STAT1 activity (adapted from reference 37). (E) Normalised expression levels of TRIM8 in whole blood from patients before (day 1) and after (day 3) initiation of canakinumab treatment, stratified by those achieving adapted ACR 50 response. **P<0.01. ***p<0.001 by ANOVA with follow-up Dunnett’s multiple comparisons test. ANOVA, analysis of variance; IFNγ, interferon-γ; SJIA, systemic juvenile idiopathic arthritis; TRIM8, tripartite motif containing 8.

Interestingly, our gene expression data also identified marked overexpression of TRIM8 compared with control monocytes, regardless of SJIA disease activity (figure 3A). As shown in figure 3D, TRIM8 is an E3 ubiquitin-protein ligase and a positive regulator of IFN-signalling,36 37 which participates in the activation of IFN-γ signalling by promoting proteasomal degradation of negative regulators including the suppressor of cytokine signalling 1 (SOCS1).36 TRIM8 has also been reported to positively regulate NF-κB signalling pathways.38 Indeed, we find that in addition to increased TRIM8 expression, SJIA monocytes from patients with high ferritin upregulate more than 10% of genes in the ‘I-kappaB kinase/NF-kappaB signalling’ GO pathway (adjusted p=0.03). To confirm this observation of increased TRIM8 expression, we assessed the whole blood gene expression profiles obtained during the clinical trial of canakinumab in SJIA.7 As shown in figure 3E, TRIM8 expression was upregulated in all SJIA patients compared with controls prior to canakinumab treatment (day 1). By day 3 of treatment, TRIM8 expression significantly decreased in most responders and was comparable to controls, but remained elevated in non-responders. In contrast to SJIA, examination of our previously published gene expression data sets39 40 revealed only subtle trend towards higher expression of TRIM8 in whole blood in active polyarticular JIA (Log FC 0.11, p=0.05) and pHLH (Log FC 0.057, p=0.63). Taken together, these findings demonstrate that monocytes from patients with SJIA demonstrate several gene expression changes that could affect IFN-γ responsiveness, including TRIM8 overexpression that is rather specific to SJIA.

Elevated TRIM8 and IFN-γ-induced signature in haemophagocytic BMM in MAS

Circulating monocytes are recruited to inflammatory sites, where in the context of a specific cytokine milieu, they mature into resident macrophages. As such, blood monocytes may not reflect the phenotype of myeloid cells during SJIA and emergence of MAS. We utilised single cell (sc) RNA-seq to better understand the specific gene expression signatures of BMM in SJIA (figure 4A). Three independent control samples yielded 180 single BMM. While there was substantial interindividual variability, a core set of approximately 1400 genes were identified that contributed to the heterogeneity of normal BMM population (online supplemental figure S2A). Control macrophages formed three primary cellular clusters, which were distinguished based on expression of genes associated with inflammatory responses including IFNGR2 (cluster 1), granulocyte-monocyte colony stimulating factor (GM-CSF) signalling (cluster 2) and aurora B signalling (cluster 3) (figure 4B). We also noted that these profiles represent a dominant signature (spanning all three clusters and >10 000 genes), obscuring identification of other subclusters. We, thus, performed an unsupervised analysis, excluding this dominant signature, in ICGS2 (see online supplemental methods),41 which identified 11 additional sub-clusters, including a small IFN- response enriched cell population (online supplemental figure S2B).

Figure 4

Single-cell RNA sequencing of BMM identifies distinct subpopulations in MAS with features of interferon response. (A) Isolation of BMM by flow cytometry. Cells were gated for 7AAD- (live)/CD15−, and then for CD14+CD163+ macrophages. (B) Identification of macrophage populations in normal BM samples (d1–3), using HOPACH clustering of the most highly variable genes. This clustering shows significant variability in expression in BMM from three independent normal clinical biopsy samples. At least three distinct cluster of macrophages and at least three groups of genes could be discriminated. Along left side of the plot are enriched functional pathways within the gene cluster (PathwayCommons), and representative genes listed along right edge. (C) BM biopsy from patient with new-onset SJIA. Immunohistochemical staining with CD163 shows increased macrophages/histiocytes with rare haemophagocytosis. (D) Proportion of CD14 +CD163+CD15- macrophages isolated from SJIA patient compared with those from normal BM samples. (E) Distinct macrophage population with altered transcriptional profile in MAS. Unsupervised clustering of scRNA-seq (ICGS version 2) from three normal samples (d1–3) and one patient with SJIA and early MAS. The black bars at the bottom of the plot denote SJIA specific or highly enriched clusters. The top marker gene is shown to the right of the plot and TRIM8 is denoted in red. (F) Network representation of statically enriched transcription factor regulated targets from the software GO-Elite (TFTarget database) for upregulated genes in the TRIM8 MAS BM expanded cell cluster. Red circles denote upregulated genes and yellow boxes denote the predicted regulatory transcription factor (G) Gene-set enrichment of differentially expressed genes in systemic JIA BM macrophage subpopulation compared with all control BMM using the ToppFun website.49 d1, d2, d3 represent control BMM donor samples. 7AAD, 7-aminoactinomycin D; BMM, bone marrow macrophage; ICGS, iterative clustering and guide-gene selection; MAS, macrophage activation syndrome; scRNA, single cell RNA; SJIA, systemic juvenile idiopathic arthritis; TRIM8, tripartite motif containing 8.

To assess changes in BMM populations in SJIA, we profiled a BMM sample from a patient with newly diagnosed SJIA with histologic findings on BM biopsy of mild histiocytic hyperplasia with rare haemophagocytosis (figure 4C), consistent with early or subclinical MAS (further patient description in online supplemental methods). Indeed, the sorted BM aspirate demonstrated a twofold increase in BMM than control aspirates (figure 4D). BMM expression profiles from this SJIA/MAS patient were largely distributed among control donor BMM clusters identified above (figure 4E). However, two clusters were identified with distinct subpopulations of BMM from the SJIA/MAS patient that exhibited markedly altered transcriptional profiles (figure 4E), and TRIM8 was among the top marker genes of the smaller of these clusters. To understand the broader molecular impact of this subtype, we identified the cells with this TRIM8-associated signature to identify differentially expressed genes vs control macrophages (see online supplemental methods). In total this signature was present in 20% (12/61) of SJIA patient cells. In addition to upregulation of TRIM8 (3.8-fold change), this SJIA/MAS macrophage population showed a strong IFNγ-induced signature (‘cellular response to IFN gamma’, adjusted p=4.9×10−3). In addition, this signature also demonstrated significant upregulation of gene pathways including response to cytokines (p=3.7×10−4) and innate immune response (p=1.5×10−3), and a large activated transcription-factor network (figure 4F,G). Notably this macrophage population signature included significant upregulation of pathways involved in intracellular granule movement (p=2.7×10−4) including the MAS-associated gene STXBP2 (p=3.1×10−5), suggesting that we correctly identified the population of haemophagocytic macrophages (figure 4G, online supplemental table S4 and S5). There were no other specific cytokine response pathways that were significantly enriched. Crayne et al 42 recently suggested that haemophagocytic macrophages may have anti-inflammatory properties including heme-oxygenase (HO-1) production and secretion of IL-10 and IL-4. Although HO-1 was expressed in this population of macrophages, IL4 and IL10 were not on the list of DEGs. In contrast, overexpression of several DAMPs capable of serving as endogenous TLR agonists (HMGB1, HMBG2, S100A12 and S100 A8/9) was prominent. The presence of a strong IFN-γ-induced signature on the other hand was consistent with work demonstrating that IFN-γ alone can act directly on macrophages to induce haemophagocytosis leading to consumptive anaemia of inflammation.43 Together, this suggests that there exists distinct and activated proinflammatory BMM subpopulations during early/subclinical MAS, with the potential for exaggerated responses to IFN-γ.

Previously, Cepika et al 44 identified decreased aryl hydrocarbon receptor (AHR) expression in SJIA monocytes as a factor promoting differentiation of monocytes into macrophages in SJIA patients. The authors felt that this was an important factor contributing to the risk for MAS. Although in our data set AHR was not on the list of the DEGs in the population of haemophagocytic BMM, there were two other genes from the same signalling pathway: AIP and AHRR, both overexpressed. AHRR encodes AHR repressor that functions as a feedback modulator by repressing AHR-dependent gene expression.45 The AIP gene (AHR interacting protein) has also been implicated in negative regulation of AhR signalling.46 Overexpression of these two genes in proinflammatory likely haemophagocytic macrophages would lead to downregulation of the AHR pathway in transition to MAS. Therefore, our data does support the observation made by Cepika et al.

TRIM8 knockdown via siRNAs in THP-1 macrophages led to decreased production of CXCL9 and CXCL11 in response to stimulation with IFN-γ in vitro

Based on prior work, we hypothesised that TRIM8 overexpression will decrease repression of IFN-induced signalling in SJIA monocytes and macrophages, leading to exaggerated responsiveness to IFN-γ and contributing to the development of MAS. We assessed the effects of TRIM8 knockdown in THP-1 macrophages on expression of the IFN-induced genes CXCL9/CXCL11 on IFN-γ stimulation in vitro. TRIM8 siRNA was used to reduce expression of TRIM8 at both mRNA and protein levels (figure 5A,B). As shown in figure 5C, on IFN-γ treatment, macrophages with reduced levels of TRIM8 demonstrated significantly reduced upregulation of CXCL9 and CXCL10. Notably, TRIM8 knockdown did not simply change the kinetics of IFN-γ response, as diminished IFN-γ-induced expression of CXCL9/CXCXL11 was observed at 4, 16 and 24 hours post-treatment (figure 5D).

Figure 5

Effects of TRIM8 knockdown on CXCL9 and CXCXL11 production in THP-1-derived macrophages stimulated with IFN-γ in vitro. Macrophages were incubated with either negative control (NC) or TRIM8 siRNAs. (A) TRIM8 mRNA levels in macrophages treated with either NC or TRIM8 siRNAs assessed by RT-PCR. ***P<0.001 by t-test. (B) TRIM8 protein levels in triplicate samples of macrophages treated with either NC or TRIM8 siRNAs assessed by Western Blot. (C) Fold increase in CXCL9 and CXCL11 mRNA levels as determined by RT-PCR in macrophages pretreated with either NC or TRIM8 siRNAs at 4 hours after stimulation with IFN-γ in vitro. **P<0.01, ***p<0.001 by ANOVA with follow-up Dunnett’s multiple comparisons test. (D) Expression of CXCL9 as determined by RT-PCR relative to GAPDH in macrophages pretreated with either NC or TRIM8 siRNAs at 1,12,16 and 24 hours after stimulation with IFN-γ in vitro. All experiments were performed in triplicates. *P<0.05. **p<0.01 by ANOVA with follow-up Dunnett’s multiple comparisons test. ANOVA, analysis of variance; IFN-γ, interferon-γ; TRIM8, tripartite motif containing 8.

TRIM8 knockdown decreases STAT1 phosphorylation in response to IFN-γ in vitro

Finally, to assess whether decreased production of CXCL9/11 was associated with decreased IFN-induced signalling, phosphoflow was used to measure STAT1 phosphorylation in THP-1 cells in response to stimulation with IFN-γ in vitro. As shown in figure 6, pretreatment with TRIM8 siRNA decreased the intensity of pSTAT1 signal assessed 30 min after IFN-γ stimulation. Taken together, we show that TRIM8 expression is required for full macrophage responsiveness to IFN-γ, and could represent a key and targetable pathway in MAS pathogenesis.

Figure 6

TRIM8 knockdown reduces IFNγ-mediated STAT1 phosphorylation. THP-1-derived macrophages were incubated with either negative control (NC) or TRIM8 siRNA. Intracellular STAT1 phosphorylation was assessed by flow cytometry at 30 min after in vitro stimulation with IFNγ. (A) Representative histograms showing pSTAT1 intensity in macrophages treated with NC (left) and TRIM8 (right) siRNA at baseline and 30 min after stimulation with IFN-γ in vitro. (B) Median fluorescence intensity of pSTAT1 indicating the amount of pSTAT1 produced in cells, pooled from three independent experiments. ***P<0.001 by ANOVA with follow-up Dunnett’s multiple comparisons test. ANOVA, analysis of variance; IFN-γ, interferon-γ; TRIM8, tripartite motif containing 8.

Discussion

Children with SJIA demonstrate continuous activation of monocytes and macrophages.3 However, the precise function of these cells in systemic hyperinflammation remains poorly understood. Several gene expression, immunophenotyping and microRNA analysis studies suggest that monocytes in SJIA display a mixed polarisation state, with markers reflecting both classical activation and multiple alternatively activated phenotypes.5 6 28 31 32 Here, we report extensive transcriptional profiling of purified monocytes from patients with SJIA. First, we find that SJIA monocytes show dramatically distinct transcriptomes from control monocytes, regardless of disease activity. This is in agreement with our previous work showing persistently altered microRNA profiles in monocytes during inactive disease, and may suggest persistent and more durable epigenetic changes in these cells. Second, we identify a robust transcriptional signature of myeloid cell activation present in monocytes from SJIA patients with elevated serum ferritin levels. Third, we show that this ‘high ferritin’ signature was enriched for genes representing multiple polarisation phenotypes, but most enriched for alternatively activated conditions such as M(IL-4) and M(IL-10). Together, these data suggest that SJIA monocytes are functioning in an attempt to compensate for systemic hyperinflammation, and display both proinflammatory and anti-inflammatory properties.

MAS remains a critical complication of 10%–15% of SJIA patients despite introduction of IL1- and IL6-inhibiting biologics. IFN-γ, a cytokine not considered a major player in SJIA itself, is increasingly recognised as a pivotal driver of MAS.12 23 24 Consistent with this concept, preliminary results of the ongoing Phase II clinical trial of the anti-IFN-γ antibody emapalumab in MAS/SJIA (NCT03311854) are very promising.25 Interestingly, and consistent with prior work,5 6 we found little evidence of IFN-γ-mediated activation in circulating SJIA monocytes. However, during MAS circulating monocytes are recruited to inflammatory sites where they mature into activated tissue macrophages. To explore that, we report the first transcriptional profile of haemophagocytic BMM during MAS at the sc level. These cells showed upregulated gene pathways that would be predicted for haemophagocytes, including cytokine response, granule secretion, and MAS-associated genes. They also exhibited a strong IFN-γ-induced signature, which is among the most significantly enriched gene ontology pathways. These findings are consistent with a model where in MAS, inflammatory monocytes rapidly traffic to tissue on IFN-γ activation. It also highlights the importance of studying key effector cells in tissue in conjunction with the periphery.

Overall our data support the concept that increased IFN-γ activity observed during MAS could be facilitated by two factors. The first is strikingly high levels of free IL-18, a cytokine that augments production of IFN-γ in response to various stimuli.4 26 27 Our data here confirm that monocytes in SJIA show a progressive increase in IL18 expression, and decrease in IL18BP expression, when stratified by disease activity and degree of hyperferritinaemia. Notably, the primary cellular source of IL-18 in SJIA remains uncertain and may include epithelial cells.4 However, a second key factor is the exaggerated responsiveness of monocytes and macrophages to IFN-γ which we have previously noted.21 Indeed, markedly increased expression of the IFN gamma receptors (IFNGR1 and IFNGR2) both transcriptionally and on the surface SJIA monocytes and macrophages may serve as one mechanism to exagerate these cell’s responsiveness to IFN-γ. Similar IFN hyper-responsiveness has been recently reported in lupus, where increased expression of IFNAR1 was found in monocytes from both mouse models and human patients, and linked to higher IFN-α-stimulated gene expression.29 35

More importantly, our study identified TRIM8 as a likely contributor to the exaggerated responsiveness of monocytes and macrophages to IFN-γ. This observation was confirmed using the publically available blood gene expression profiles obtained during the clinical trial of canakinumab in SJIA.7 Separately, our scRNA-seq of BMM identified TRIM8 overexpression as one of the features distinguishing multiple populations of proinflammatory macrophages with IFN-γ response signatures from other macrophages in the BM during SJIA. Of note, this patient had features of early/subclinical MAS without clinically overt disease; whether more extensive transcriptional changes are seen during progression to ‘full-blown’ MAS remains to be seen. TRIM8 is an E3 ubiquitin-protein ligase that plays important roles in innate immune pathways.36–38 Thus, TRIM8 plays a positive role in the TNF- and IL-1β signalling pathways. Little is known regarding the transcriptional regulation of TRIM8. The main TRIM8 regulatory region contains ChIP-seq peaks from multiple transcription factors including the vital hematopoietic transcription factor GATA2. It likely has complex cytokine regulation, as it contains both a STAT1 peak suggesting induction by IFN-γ, and TRIM22, an IFN-γ-induced epigenetic repressor.47 48 Mechanistically, TRIM8 induces the lys-63 polyubiquitination of MAP3K7/TAK1 component leading to the activation of NF- κB,38 and was associated here with upregulation of NF-κB-induced genes in SJIA monocytes. TRIM8 also activates IFN-γ signalling by promoting proteasomal degradation of the IFN-γ repressors SOCS1 and PIAS3.36 37 SOCS1 is induced by various proinflammatory cytokines including IFN-γ and negatively regulates IFN-signalling by inhibiting IFN-induced JAK-STAT activation.35 Indeed, we show that TRIM8 knockdown resulted in decreased STAT1 phosphorylation and decreased expression of CXCL9 and CXCL11 in response to stimulation with IFN-γ in vitro. The observed effects of TRIM8 knockdown on STAT1-phosphorylation may suggest that TRIM8 functions primarily through degradation of SOCS1 (figure 3B), but the mechanisms by which TRIM8 overexpression potentiates IFN responsiveness remains to be investigated. In addition, the effect of TRIM8 knockdown on IFN-induced responses in primary SJIA monocytes still needs to be assessed

We note that this study is limited by the BMM experiments being derived from a single patient. However, these experiments address the question of localisation of observed IFN-γ pathway associated gene expression responses at the single-cell level. A broader single-cell study would be needed to assess single-cell variability in a larger cohort. In addition, the specificity of TRIM8 elevation to SJIA monocytes remains to be tested in similarly isolated cells from other inflammatory disorders, as well as a critical need to test the impact of TRIM8 knockdown on primary SJIA monocytes and defining the mechanisms by which TRIM8 mediates IFN hyper-responsiveness. We also note that our SJIA cohort has a female predominance that is not observed in the disease more broadly. Finally, we note that the collection and processing of control samples as separate batches may introduce unmeasured variance in the gene expression profiles.

Augmented production of IFN-γ facilitated by IL-18, combined with exaggerated responsiveness to IFN-γ, that is, at least partially, caused by increased TRIM8, may be two pathophysiological features that explain strikingly high rates of MAS in SJIA. The in vitro experiments demonstrating the effects of TRIM8 inhibition on macrophage responsiveness to IFN-γ provide a rationale for using TRIM8 as a biomarker for risk for MAS. Indeed, future work will explore the potential of TRIM8 as a possible therapeutic target both in acute episodes of cytokine storm including MAS, as well as a long-term prophylactic for SJIA patients at risk for recurrent MAS.

DNA sequencing datasets

Bulk and single-cell RNA-seq datasets have been deposited in gene expression omnibus (GSE147608 and GSE147795, respectively).

Data availability statement

Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as online supplemental information. Bulk and single-cell RNA-seq datasets have been deposited in gene expression omnibus (GSE147608 and GSE147795, respectively).

Ethics statements

Ethics approval

This study was approved by the institutional review board of CCHMC (IRB# 2012–0160).

References

Footnotes

  • Handling editor Josef S Smolen

  • Twitter @GrantSchulert

  • Contributors GS and AAG designed the study. GS, TD, SD, NF and ST performed the experiments. GS, AP, DS, MM, NS and AAG performed gene expression analysis. GS and AAG wrote the first draft of the manuscript. All authors contributed to the final manuscript and approved its submission.

  • Funding This work was supported by the Systemic Juvenile Idiopathic Arthritis Foundation; National Institutes of Health K08-AR072075 (GS), R01-AR059049 (AAG) and P30-AR070549; Cincinnati Children’s Research Foundation ARC Grant (GS&AAG); and an unrestricted gift from the Jellen Family Foundation.

  • Competing interests GS has served as a consultant for Novartis and Sobi. AAG has served as a consultant for Juno and Novartis, and has received research support from Sobi and AB2Bio. All other authors declare no conflicts of interest.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.