Deconvolution of monocyte responses in inflammatory bowel disease reveals an IL-1 cytokine network that regulates IL-23 in genetic and acquired IL-10 resistance

Objective Dysregulated immune responses are the cause of IBDs. Studies in mice and humans suggest a central role of interleukin (IL)-23-producing mononuclear phagocytes in disease pathogenesis. Mechanistic insights into the regulation of IL-23 are prerequisite for selective IL-23 targeting therapies as part of personalised medicine. Design We performed transcriptomic analysis to investigate IL-23 expression in human mononuclear phagocytes and peripheral blood mononuclear cells. We investigated the regulation of IL-23 expression and used single-cell RNA sequencing to derive a transcriptomic signature of hyperinflammatory monocytes. Using gene network correlation analysis, we deconvolved this signature into components associated with homeostasis and inflammation in patient biopsy samples. Results We characterised monocyte subsets of healthy individuals and patients with IBD that express IL-23. We identified autosensing and paracrine sensing of IL-1α/IL-1β and IL-10 as key cytokines that control IL-23-producing monocytes. Whereas Mendelian genetic defects in IL-10 receptor signalling induced IL-23 secretion after lipopolysaccharide stimulation, whole bacteria exposure induced IL-23 production in controls via acquired IL-10 signalling resistance. We found a transcriptional signature of IL-23-producing inflammatory monocytes that predicted both disease and resistance to antitumour necrosis factor (TNF) therapy and differentiated that from an IL-23-associated lymphocyte differentiation signature that was present in homeostasis and in disease. Conclusion Our work identifies IL-10 and IL-1 as critical regulators of monocyte IL-23 production. We differentiate homeostatic IL-23 production from hyperinflammation-associated IL-23 production in patients with severe ulcerating active Crohn’s disease and anti-TNF treatment non-responsiveness. Altogether, we identify subgroups of patients with IBD that might benefit from IL-23p19 and/or IL-1α/IL-1β-targeting therapies upstream of IL-23.


INTRODUCTION
The pathogenesis of inflammatory bowel disease (IBD), which includes Crohn's disease (CD) and

Significance of this study
What is already known on this subject? ► IBD is a complex chronic inflammatory disease caused by the interaction of host and environmental factors in a genetically susceptible setting. Targeting the proinflammatory arm of the immune system such as by blocking tumour necrosis factor (TNF) signalling has proven therapeutically successful in subgroups of patients with IBD. Studies in both mice and humans indicate interleukin (IL)-23 as a key pathogenic proinflammatory factor in IBD. Early-stage clinical trials indicate that blocking IL-23 can induce disease remission in patients with Crohn's disease (CD) and ulcerative colitis (UC). As part of personalised medicine, mechanistic insights into the regulation of IL-23 are prerequisite for selective IL-23 targeting therapies.
What are the new findings? ► We identify autocrine and paracrine sensing of IL-1α/IL-1β and IL-10 as key cytokines that control IL-23-producing monocytes. We determine a transcriptional signature of IL-23-producing monocytes that predicted both disease and resistance to anti-TNF therapy and differentiate this hyperinflammationassociated IL-23 production in patients with severe ulcerating active CD from homeostatic intestinal IL-23 production. We show that exposure of monocytes to whole bacteria induced IL-23 production via acquired IL-10 signalling resistance akin to IL-23 secretion by lipopolysaccharide-stimulated monocytes in Mendelian IL-10 signalling defects.

Inflammatory bowel disease
ulcerative colitis (UC) and IBD unclassified (IBDu), is caused by dysregulated innate and adaptive immune responses that drive chronic relapsing tissue inflammation. 1 Genetic studies suggest a complex polygenic inheritance driven by defective innate and adaptive immunity. 1 2 Interleukin (IL)-23 signalling and T-helper (Th)1/Th17 immunity are significant mediators of intestinal inflammation as indicated by genetic variants in IL23R encoding the IL-23 receptor 3 as well as RORC, STAT3, IRF5, IL1R1, IL6ST, IL12B, TYK2, IL21, JAK2, IFNG, SMAD3 and CCR6. 4 5 Monocytes, macrophages and dendritic cells (DCs) produce IL-23 in response to bacteria-and fungi-derived microbial stimuli and drive Th1 and Th17 differentiation in the pathogenesis of intestinal inflammation. [6][7][8][9] In addition to a role of pro-inflammatory and anti-inflammatory cytokine networks, 10 mouse models as well as human Mendelian disorders highlight the essential role of IL-10 signalling in controlling inflammatory cytokine responses. Therapeutic approaches that target the proinflammatory arm of the immune system by blocking cytokine signalling or by affecting intestinal cell migration are effective in subgroups of patients with IBD. 8 Mice that lack IL-23p19 are protected from developing colitis in innate as well as lymphocyte replete models of intestinal inflammation induced by IL-10 signalling defects, bacterial colonisation or innate immune stimulation via the anti-CD40. 7 11 These results indicate that IL-10 blockade unleashes an IL-23 mediated immune pathogenesis. Importantly, blockade of IL-23p19 and IL-12p40 showed therapeutic benefit for patients with CD and UC. 12 13 To predict genetic susceptibility or disease course, individual loci and genetic risk scores 14 15 as well as gene expression signatures or transcriptomic scores of peripheral CD8 + T cells, 16 peripheral blood 17 or intestinal tissue 18 are revealing patient stratification strategies in IBD. Membrane-bound tumour necrosis factor (TNF) 19 and IL-6/oncostatin M (OSM) associated cytokines 18 predict anti-TNF non-response. However, transcriptional signatures informed by IL-23 expression have not been described.
Personalised medicine targeting the IL-23 axis requires an understanding of the cellular sources, networks and regulation of IL-23. Here we investigate the regulation of IL-23, describe distinct monocyte subsets that express IL-23 and identify IL-1 signalling as the key cytokine for the differentiation of IL-23producing monocytes. We identify a hyperinflammatory signature of IL-23-producing monocytes in intestinal tissue transcriptomes of patients with IBD and find an additional signature of IL-23 that is associated with lymphocyte cell differentiation in healthy tissue.

METHODS
Detailed methods are outlined in the online supplemental materials.

IL-10 signalling blockade facilitates IL-23 production by a subset of peripheral blood monocytes
We investigated the regulation of IL23A expression by subjecting peripheral blood mononuclear cells (PBMC) from 41 patients with IBD (online supplemental figure 1A and online supplemental table 1) to IBD-relevant stimuli that target different aspects of innate and adaptive immune cell responses. Lipopolysaccharide (LPS), muramyl-dipeptide (MDP), T-cell receptor and costimulation (αCD3/αCD28 coated beads) and IL-10 signalling blockade were used alone or in combination based on the concept that innate pathogen recognition receptor responses in particular, NOD2 14 and TLR4, 20 T-cell responses and IL-10 signalling defects are implicated by multiple genetic IBD susceptibility loci and Mendelian forms of IBD. 21 We performed microarray gene expression analysis of all conditions at 16 hours following stimulation in the presence or absence of IL-10 receptor (IL-10R) blocking antibodies. We identified stimulus-specific and shared gene expression signatures (online supplemental figure 1B,C and online supplemental table 2). LPS, L18MDP and αCD3/αCD28 coated beads induced 319, 187 and 312 genes, respectively, out of which 109, 19 and 175 were condition-specific genes (Benjamini-Hochberg (BH)-adjusted p<0.05, fc≥1.5; online supplemental figure 1B). Changes in IL23A expression were moderate under these conditions (online supplemental figure  1C). We next investigated the role of IL-10 signalling during LPS, L18MDP and αCD3/αCD28 stimulation. In total, 36 genes were upregulated and 29 genes were downregulated by IL-10 blockade, with most of the changes being found during LPS stimulation(fc>1.5,BH-adjustedp<0.05;onlinesupplemental figure 1D,E and online supplemental table 3). IL23A expression was significantly upregulated under conditions of LPS or L18MDP stimulation and IL-10 signalling blockade (online supplemental figure 1C,D).

Significance of this study
How might it impact on clinical practice in the foreseeable future? ► We identify subgroups of patients with IBD that might benefit from IL-23p19 and/or IL-1α/IL-1β-targeting therapies upstream of IL-23.

Single-cell sequencing identifies inflammatory IL-10regulated monocyte phenotypes
FACS analysis suggested the presence of unknown functional heterogeneity within the population of stimulated CD14 + monocytes (see e.g. figure 1D). To characterise the LPS-induced and IL-10-regulated transcriptional profile of monocytes subsets, and to differentiate population-wide transcriptional changes from subset-specific responses we performed single-cell RNA sequencing ( figure 6A). The majority of unstimulated monocytes were lysozyme + (LYZ), CD52 + (CAMPATH1) and FCN1 + expressing CD14 high monocytes (cluster 0-unstimulated) typical of classical monocytes, while a minority of unstimulated monocytes displayed a CD52 + , FCN1 + , CD16 + (FCGR3A) CD14 low phenotype (cluster 7-unstimulated) characteristic of intermediate or non-classical monocytes in keeping with the flow cytometry data (online supplemental figure 4). 24 Two clusters of cells that emerged following LPS stimulation were composed of CD14 + IL1B + cells. While these two clusters of cells had very similar profiles (online supplemental figure 6B), the first showed lower expression of CD163 and CD300E and was enriched for genes associated with 'T-cell tolerance induction' (cluster 1-LPS-stimulated), while the second displayed higher CD163 and CD300E expression and an enrichment for 'monocyte activation' genes (cluster 2-LPS-stimulated). The three additional monocyte phenotypes that appeared on combined LPS stimulation and IL-10R blockade were demarcated by (1) high expression of type I IFN-responsive genes (eg, IFIs, IFITs, ISGs, OASs and IRFs) and genes linked to antigen processing and presentation (e.g. B2M, CCR7, HLA and CD74) (cluster 3-LPS and anti-IL-10R, called IFN-induced monocytes);(2)highexpressionofITGAX and enrichment for genes associated with 'superoxide generation' and 'positive regulation of neutrophil activation' (cluster 6-LPS and anti-IL-10R, called microbicidal monocytes) and (3) expression of proinflammatory genes including (IL1B, IL6, IL23A, CCL20 and PTGS2) and enrichment of genes associated with 'Th17 cell lineage commitment', 'positive regulation of acute inflammatory response' and 'IFN-γ production' (cluster 4-LPS and anti-IL-10R, called IL-23 + inflammatory monocytes). These three phenotypes in LPS-stimulated anti-IL-10R-treated monocytes were replicated in a second donor (online supplemental figure 6C).
More detailed examination of the expression of key cytokines and their receptors revealed that while IL23A and IL1R1 were detected in a cluster-specific manner under the 'hyperinflammatory' LPS and anti-IL-10R condition, IL10, IL10RA and IL10RB, as well as IL1A/IL1B mRNA showed a broader, cross-condition expression ( figure 2D). Finally, we sought to compare the monocyte phenotypes that we identified following ex vivo stimulation with those reported for colonic mononuclear phagocytes in human IBD. We show that many of the genes associated with inflammatory mononuclear phagocyte phenotypes in CD 9 ('inflammatory macrophages') and UC 25 ('inflammatory monocytes') were upregulated in clusters associated with LPS or LPS+anti-IL-10R stimulation (figure 2E). Clusters mainly composed of cells from the LPS+anti-IL-10R stimulation condition showed the highest induction of the cassettes of diseaseassociated cytokines and chemokines, with the IL23A + cluster 4 being marked by a particularly high expression of IL6, IL1A, IL1B, TNF and OSM, as well as elevated levels of IL10 ( figure 2D).

IL-10-producing monocytes control IL-23-producing monocytes through paracrine signalling
In PBMC, we observed that IL23A and IL10 mRNA expression were strongly correlated (Spearmans' r=0.84, p<0.0001) when stimulated with LPS and IL-10R blockade (online supplemental figure 7A). These data indicate that IL10 mRNA expression itself is tightly regulated by IL-10 signalling as part of a negative feedback mechanism. We assumed that IL23A and IL10 would be produced by the same cells. However, inspection of the singlecell data suggested that this might not be the case. Intracellular flow cytometry demonstrated that IL-10 expression in monocytes at the population level was earlier then the upregulation of IL-23 (online supplemental figure 7B) and that IL-10 and IL-23 largely originate from distinct cells in both monocytes isolated from HD and patients with IBD ( figure 3A,B). While all subsets of IL-23p19 and/or IL-10-producing monocytes were deregulated on IL-10R blockade, monocytes isolated from patients with IBD showed decreased frequencies of IL-10 + single producing monocytes and IL-23p19 + IL-10 + coproducing monocytes but not IL-23p19 single producing monocytes when compared with monocytes isolated from healthy donors (figure 3C). The presence of diverse cytokine profiles in monocytes indicated that IL-10 may regulate IL-23 production through a paracrine mechanism in functionally distinct cells. Alternatively, monocytes might express IL10 and IL23A mRNA sequentially (ie, with early IL-10 producer becoming IL-23 producers). Otherwise, the pattern of segregated IL-10 and IL-23 protein expression might result from oscillatory expression of IL10 and IL23A mRNA. To distinguish between these possibilities, we performed a dualcolour ELISpot assay using MACS-purified monocytes. This approach captured cytokine production in the individual cells over the full duration of their culture in vitro. In line with a paracrine model where IL-10-producing monocytes regulate IL-23 production in others, we found that IL-10-producing monocytes were distinct from IL-23-producing monocytes (figure 3D,E, and online supplemental figure 7C). These results suggest that the IL-10-producing monocytes regulated the IL-23-producing cells through a paracrine mechanism and that the observed monocyte heterogeneity was not the result of sequential or oscillatory cytokine secretion. The absence of aggregates of either IL-10secreting or IL-23-secreting monocytes additionally demonstrated that the development of IL-23-producing monocytes did not result from local cell segregation in tissue culture.

Deconvolution of intestinal IL-23 gene expression in intestinal tissue
We next sought to understand the context of IL-23 expression during intestinal inflammation. We compared expression of IL-23 in ileal biopsies from 219 patients with CD and noninflamed controls in samples from the Risk Stratification

Inflammatory bowel disease
and Identification of Immunogenetic and Microbial Markers of Rapid Disease Progression in Children with Crohn's Disease(RISK) study 26 (controls without ileal inflammation include 42 healthy donors and 61 patients with UC) (figure 4A). As expected, IL23A and IL12B were expressed in the CD patient samples, along with IL1A, IL1B and IL6. However, we also found IL23A and IL12B expression in non-inflamed control biopsies ( figure 4A,B) where IL23A expression was associated with TNF rather than IL1B expression ( figure 4C,D). We therefore performed weighted gene

Inflammatory bowel disease
correlation network analysis 27 to investigate the possible sources and roles of IL23A in the inflamed and non-inflamed intestines (online supplemental figure 8A). This analysis identified 22 modules of coexpressed genes that were named according to their gene members and enrichments for biological pathways and sets of cell-type marker genes (online supplemental table 4). Clustering of the modules by their expression patterns in the RISK cohort patients identified six major groups that could be broadly distinguished by their correlations with diagnosis of CD, IL23A expression, epithelial gene expression, mitochondrial activity and cell growth genes (online supplemental figure 8A,B).
Most prominently, we found an 'inflammatory cytokine' module that was significantly correlated with both CD (r=0.55) and IL23A expression (r=0.35) (online supplemental figure 8B). This module contained key myeloid and stromal markers genes (CD14 and PDPN), proinflammatory cytokines (including OSM, IL1B and IL6) and fibroblast activation protein (online supplemental figure 8B) in keeping with the emerging concept of a pathogenic myeloid-stromal cell circuit in IBD. 18 24 Intriguingly, however, we found that IL23A expression was most strongly correlated with two modules of 'immune cell differentiation' (r=0.7) and 'lymphocyte differentiation' (r=0.71) genes (online supplemental figure 8A). Of these, the immune cell differentiation module was weakly correlated with CD (r=0.2), associated with expression of myeloid (CD14) and lymphoid (CD79A, CD4) cell markers, and enriched for 'myeloid DC differentiation' (online supplemental figures 8B and 9A). The lymphocyte differentiation module showed an expression signature suggestive of lymphoid follicles, including individual markers of B-cell and T-cell identity (CD79A and CD4) and differentiation (BACH2) 25 26 (online supplemental figure 8B), as well as enrichments for gene ontology categories related to lymphoid cell differentiation, proliferation and selection (online supplemental figure 9A). While the lymphocyte differentiation module was not correlated with CD, it was unique in showing significant enrichments for two sets of IBD Genome-wide association study (GWAS)-associated genes (OR>3), suggesting that this set of genes is relevant for the pathogenesis of IBD (online supplemental figure 9B). We also noted that both of the IL23Aassociated immune cell differentiation and lymphocyte differentiation modules were significantly enriched for genes belonging to the KEGG 'Th17 cell differentiation' pathway (online supplemental figure 9A).
Overall, the network analysis suggested that IL-23 expression was primarily associated with lymphoid cell differentiation in both health and disease. We therefore sought to understand which of the IL-10-regulated monocyte genes were specific to disease-associated inflammation. We correlated the expression of a curated set of IL-10-regulated monocyte genes (35 genes specific to combined LPS and anti-IL-10R stimulation scRNA-Seq;figure 4E,F;onlinesupplementalmethodssection, 'Identification and characterisation of an IL-10-responsive monocyte gene signature') with the eigengenes for the CD-associated inflammatory cytokine and the non-inflammatory lymphocyte differentiation modules in the RISK cohort data. This analysis identified three subsets of IL-10-regulated monocyte genes (figure 4E). These comprised (1) genes correlated only with the CD associated inflammatory cytokine module (black box, figure 4E);(2)asetofgenescorrelatedwithboththe inflammatory cytokine and lymphocyte differentiation modules (grey box, figure 4E); and (3) genes correlated with only the non-inflammatory lymphocyte differentiation module (green box, figure 4E). The IL-10-regulated genes that correlated with the inflammatory cytokine but not the lymphocyte differentiation eigengene (black box, figure 4E) showed a significantly superior ability to predict anti-TNF response after 4-6 weeks of treatment in an independent cohort of patients with UC (UC-cohort GSE12251) 28 and diagnosis of CD in the RISK cohort 26 (figure 4F-H, online supplemental table 5). By contrast, IL-10regulated monocyte genes that correlated with lymphocyte differentiation, but not inflammatory cytokines, had poorer predictive ability ( figure 4F-H). To validate our findings, we analysed intestinal transcriptome data of an additional adult cohort of patients with CD and UC (GSE16879). 28 Similar to our initial results, inflammatory monocytes gene expression predicted disease activity and anti-TNF non-response in adult patients with CD ileal biopsy transcriptomes. In colonic biopsies, these genes were not predictive of disease activity but distinguished patients with CD or UC anti-TNF non-response (online supplemental figure 10A,B).
These data therefore identify a subset of IL-10-regulated monocyte-derived genes, including CXCL1/2, IL1A, IL1B, INHA, IL6, CCL3/4, PTGS2, CSF2/3 and GBP1, which show a specific association for disease-associated intestinal inflammation in the RISK and GSE16879 CD cohorts. By contrast, cytokines such as IL-23 and TNF were not specific for disease suggestive of context-dependent roles in homeostasis and inflammation.

Monocyte exposure to whole bacteria causes aquired IL-10 resistance and IL-23 secretion
The analysis of gene expression data in the RISK cohort suggests IL-23 expression despite the presence of IL-10 in a subset of patients with macroscopic inflammation and deep ulcerating IBD in the absence of Mendelian IL-10 or IL-10R loss of function DNA variants, that is, functional IL-10 resistance. In light of the epithelial barrier defect in ulcerating disease that predisposes to bacterial translocation, we tested whether exposure of monocytes to bacteria can induce functional IL-10 resistance. We treated monocytes for 16 hours with heat killed Salmonella typhimurium or Escherichia coli strain Nissle. Interestingly, monocytes produced IL-23 and IL-10 on stimulation (figure 5A,B) reminiscent of LPS stimulation in presence of IL-10R blockade. The cellular dichotomy of IL-23 and IL-10 production was maintained under these conditions (figure 5B). To assess IL-10 responsiveness in previously stimulated monocytes, we extensively washed cell cultures, subsequently exposed those to recombinant human IL-10 and evaluated phosphorylation of signal transducer and activator of transcription (STAT)3 ( figure 5C). Indeed, bacterial stimulation was associated with an increased proportion of IL-10 non-responsive cells, demonstrating that monocyte exposure to whole bacteria induced a state of acquired IL-10 resistance ( figure 5D,E). Importantly, the reduced phosphorylation of STAT3 in monocytes that were exposed to whole heat killed bacteria was not due to increased cell death under these culture conditions (online supplemental figure 11A,B).

IL-1α and IL-1β are essential for monocyte IL-23 production
The regulation of monocyte-derived IL-23 and IL-10 suggested coregulation via additional factors (likely cytokines). We therefore investigated the functional effects of 11 coregulated cytokines on monocyte IL-23 production (figure 6A,B and online supplemental figure 12A,B). Strikingly, when blocking IL-1R1, IL-23 production was near completely abolished (figure 6A,B and online supplemental figure 12A-C). Blocking IL-1β or IL-1α alone had only partial or no effect on IL-23 production, indicating that either cytokine can compensate for the absence of the other in driving monocyte IL-23 production (figure 6A and online supplemental figure 12A). The effect of IL-1β on IL-23 expression was context specific, since addition of IL-1β in the presence of anti-IL-10R blockade alone did not induce IL-23 expression (online supplemental figure 12D). None of the other tested cytokines (GM-CSF, IL-6, IL-11, IL-17A, IL-18, IL-19, IL-23, IL-24, IL-36γ, TNF and type I IFN) demonstrated a significant impact on IL-23 production. Although not essential, addition of IFN-γ or IL-12 significantly increased monocyte IL-23 production (figure 6A,B and online supplemental figure  12A,B). To investigate the effects of IL-1R1 blockade on functional monocyte clusters, we stimulated PBMC from HDs with combinations of LPS and anti-IL-10R and analysed monocyte metaclusters-associated protein expression. In line with the single-cell mRNA sequencing experiments, the predicted clusters associated with LPS and combined LPS and anti-IL-10R stimulation ( figure 6C). Blockade of IL-1R1 specifically inhibited the development of IL-23-producing monocytes, while other monocyte clusters remained largely unaffected ( figure 6C,D). Importantly, monocyte IL-23 production induced by whole bacteria exposure was similarly dependent on IL-1R1 signalling (online supplemental figure 13A,B), suggesting that IL-23 is downstream of IL-1 signalling also in the context of bacteria-induced acquired IL-10 signalling defects.
We next used a mathematical model of ordinary differential equations as an objective analysis tool to investigate the relative direct and indirect impacts of the addition or blockade of various cytokines or blockade of cytokine receptors (figure 6E). The model tests whether cytokine production is modulated by other key cytokines that are produced by monocytes (TNF, IL-1α, IL-1β, IL-6, IL-10 and IL-23; false discovery rate (FDR)-corrected paired Wilcoxon test). Modelling all feasible network configurations led to the conclusion that most of the n=2 31-1 configurations could be excluded a priori. We identified an optimal network describing the core network dynamics by ranking the models based on their fit to the data using the Akaike information criterion ( figure 6F). This model supported IL-10 signalling as negative feedback mechanism and IL-1β as a positive regulator of IL-23 production, with IFN-γ acting as an amplifier. In addition, the model predicted the positive effect of IL-1β on IL-23 production to be independent of regulating IL-1α, IL-6, IL-10 or TNF and the negative regulation of IL-23 by IL-10 to be independent of the effect of IL-10 on the other cytokines present in the model (IL-1α, IL-1β, IL-6 and TNF). Interestingly, the model suggests that IFN-γ might amplify IL-1α through the suppression of IL-10, but amplify IL-23 independently of its inhibitory effect on IL-10 (the upregulatory edge from IFN-γ to IL-1α present in figure 6E was removed in the final model since this can be accounted for by a decrease in IL-10; figure 6F). Finally, negative regulation of IL-23 by IL-10 is predicted to be independent of the effect of IL-10 on the other modelled cytokines.

IL-1 regulates IL-23 production in genetic IL-10 signalling defects
To confirm these predictions, we investigated the effect of IL-1R1 blockade on IL-23 expression in the context of genetic deficiency of IL-10 signalling. We stimulated PBMC obtained from patients with infantile-onset IBD due to an IL10RA or IL10RB gene defect. Interestingly, patient-derived monocytes produced IL-23 in response to LPS stimulation alone, indicating the intrinsic defect in IL-10R-dependent regulation, confirmed by the inability of exogenous IL-10 to suppress monocyte IL-1β production. Strikingly, IL-1R1 blockade inhibited monocyte IL-23 production in LPS-stimulated PBMC from IL-10R deficient patients ( figure 7A,B). Together, these analyses confirmed IL-10 as the major negative regulator of IL-23 production, while IL-1 signalling (and in particular IL-1β) is essential for monocyte IL-23 synthesis.

DISCUSSION
We identify key regulatory circuits of IL-23 production by inflammatory monocytes. These include failure of paracrine IL-10mediated control, as well as autocrine and paracrine signalling of IL-1α and IL-1β in response to inflammatory stimuli. Indeed, increased IL-23 production can be observed in monocytes from patients with infantile-onset Mendelian IL-10 signalling deficiency or experimental blockade of the IL-10 receptor. In addition, we newly show that monocytes that respond to whole bacteria express IL-23 as a consequence of acquired IL-10 non-responsiveness. Most importantly, we define a transcriptional signature of IL23A + inflammatory monocytes that indicates a state of acquired IL-10 non-responsiveness in a subgroup of patients with deep ulcers and compromised epithelial function where monocytes may directly respond to whole bacteria. This signature predicts both diagnosis of CD and resistance to anti-TNF treatment similar to previously described OSM expression. 18 Although derived from analysis of ileal inflammation in CD, this signature was also able to predict anti-TNF failure from the transcriptional profiles of colonic samples in UC, suggesting a common role for hyperinflammatory monocytes in treatment-resistant disease in both IBD subtypes. Interestingly, loss of IL-10 responsiveness induces IL23A mRNA and IL-23 protein expression only in a small fraction of monocytes that express the IL-1 receptor IL1R1. This selective IL-23 expression contrasted with pervasive induction of IL-1α and IL-1β expression, as well as a broad induction of IL-10 expression in monocytes under the same hyperinflammatory condition.
Our data support context-specific roles of IL-23 in driving the differentiation of pathogenic Th17/Th1 cells in the presence of IL-1 29 that are enriched in inflamed intestinal tissue from patients with IBD 30 31 while supporting the development of nonpathogenic Th17 cells under homeostatic conditions such as those found in mucosa-associated lymphoid tissue (MALT) -in healthy individuals.
Together, these results suggest that subgroups of patients with active disease may benefit from anti-IL-23p19 therapy and that the targeting of upstream cytokines that regulate IL-23 expression in inflammation, such as IL-1 might provide a selective means to block pathogenic IL-23 expression. The critical element that differentiates homeostatic IL-23 and TNF expression from hyperinflammation is the coexpression of a multitude of IL-10-regulated factors, including IL-1α and IL-1β in patients with severe active disease and anti-TNF treatment non-responsiveness.
The strong predictive ability of an IL-10-sensitive inflammatory monocyte signature (CD14, IL1A, IL1B, OSM, PTGS2, IL6, CCL2/3 and CXCL1/2) emphasises a surprising and underestimated extent of IL-10 non-responsiveness in IBD. In those patients, inflammation is present despite transcription of IL-10 (patients express a large amount of IL-10) and absence of pathogenic variants within the IL-10 receptor (exome sequencing did not reveal Mendelian forms of IL-10 receptor signalling defects in this cohort). 32 The high expression of PTGS2 in this subgroup of patients with IBD represents an additional indicator for deregulated IL-10 responses and intestinal antimicrobial immunity. 33 Our experiments suggest that direct contact of monocytes with whole bacteria can induce IL-10 resistance that allows coexpression of IL-23 and IL-1 explaining the signature of IL-10 non-responsiveness in a subgroup of patients with intestinal ulceration. IgG receptor IIa signalling in myeloid cells in response to immunoglobulin coated microbes may additionally amplify inflammasome-dependent induction of IL-1 synthesis. 34 An additional mechanism of reduced IL-10 responsiveness may be differential expression of IL-10R1. 35 36 Our ELISpot experiments exclude differential spatial clustering of IL-10-secreting or IL-23-secreting monocytes, as well as oscillatory or sequential IL-10 and IL-23 transcription.

Inflammatory bowel disease
This suggests that the inflammatory microenvironment drives in parallel several monocyte effector programmes and monocyte to macrophage and DC differentiation processes, reminiscent of plasmacytoid DC stimulation with a single stimulus (R848 or CpG) that induces diverse transcriptional states and cellular functions. 37 This is likely a biological mechanism to ensure functional heterogeneity under hyperinflammatory conditions, when diversity in the host defence towards a range of potential pathogens is essential.
Single-cell RNA transcriptomic approaches identified transcriptional signatures in circulating and tissue human monocytes and macrophages, 9 25 38-42 demonstrating phenotypical and functional diversity in monocyte, macrophage and DC populations. We have focused on the cellular heterogeneity and inflammatory response of peripheral monocytes because those cells are recruited into the gut and differentiate into proinflammatory major histocompatibility complex (MHC)-II-high monocytes and macrophages that outnumber resident, yolk sac-derived tissue macrophages during inflammation. [43][44][45] Such proinflammatory macrophages have been characterised following in vitro differentiation using LPS stimulation and express IL-12 and IL-23 in response to STAT1-dependent IFN-γ signalling. 46 Recently IL1Β + 'inflammatory monocyte' 25 and IL1B + IL23A + 'inflammatory macrophage' 9 populations associated with inflammation and resistance to anti-TNF treatment have been described in single-cell studies of tissue from patients with UC 25 and CD. 9 We found that, following stimulation with LPS in the presence of anti-IL-10R blockade, a fraction of blood monocytes acquired a IL1B + IL23A + transcriptional signature that is very similar to those described for mononuclear phagocytes isolated from the inflamed intestinal biopsies of patients with CD and UC. This observation demonstrates the relevance of functional studies of peripheral monocytes for understanding the behaviour of these cells following their recruitment into the inflamed intestinal mucosa. Together with the previous findings, our observations suggest that inflammatory monocytes and CX3CR1 + IL-1β + macrophages 47 48 are likely to be major sources of IL-23 in the context of inflammasome activation and IL-1 production in the inflamed intestine and support a critical role for IL-1 in intestinal anti-TNF-treatment refractory inflammation.
Importantly, several cell types including regulatory T cells, B cells, macrophages and DCs produce IL-10 49 and may therefore contribute to the negative regulation of proinflammatory IL-23producing monocytes in tissue in vivo. While we have unravelled the regulation of peripheral blood monocytes in IL-23 responses ex vivo and related those to processes of inflammation in intestinal tissue from patients with IBD, in the tissue context complex and niche-specific cellular communication networks are expected to further modulate the regulation of monocyte inflammatory effector functions and IL-23 expression.
In individuals without intestinal inflammation, our network analysis suggested that IL23A expression was likely derived from terminal ileal MALT because the lymphocyte differentiation to which it was assigned also showed strong correlations with orthologues of genes associated with murine small intestine lymphoid tissue (SILT), including IL22RA2, ITGAX 27 as well as with VCAM1, which is a marker of lymphoid-associated villi in humans (online supplemental figure 5B). In such MALT tissue, which is hyperplastic in paediatric patients, 50 the source of IL23A is most likely DCs, tolerogenic CD103 + DC 51 or macrophages as is known to be the case in mouse SILT. 47 This work demonstrates the utility of network-based approaches for deconvolving variance in gene expression that arises from the analysis of biopsy samples from tissues with complex microanatomical heterogeneity such as the small intestine.
Our mathematical model pinpoints IL-1 as a key positive regulator of IL-23 expression in monocytes. Studies in monogenic forms of IBD suggest that IL-1R1 blockade can resolve intestinal inflammation in patients with IL-10 receptor defects, 52 mevalonate kinase defects and potentially gain-of-function NLRC4 53 defects, which are all characterised by increased inflammasome activation and IL-1 secretion. It is therefore an attractive hypothesis that the hyperinflammatory IL1 + PTGS + IL6 + signature might define an anti-IL-1R1 responsive group of patients with IBD. Interventional studies are required to identify differential effects between IL-23 and IL-1 blockade since both have complex effects on licensing for effector cytokine production and differentiation of Th1 and Th17 cells. 54 Altogether, our findings may inform transcriptional diagnostic efforts to guide use of IL-23p19 targeting therapies and suggest a role for therapeutics capable of selectively blocking IL-23 in disease by targeting context-specific upstream factors such as IL-1.

Patient and public involvement
Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.

Data availability statement
The data supporting the findings described in this study are available from the corresponding author upon request. Single-cell RNAsequencing analysis of the IL-10 dependend response of human CD14+ monocytes to LPS have been deposited at NCBI's Gene Expression Omnibus and are accessible under GEO Series accession number GSE130070. Microarray gene expression analysis of PBMC from patients with IBD stimulated by combinations of LPS, MDP, anti-CD3/anti-CD28 antibodies and anti-IL-10R blocking antibodies have been deposited at NCBI's Gene Expression Omnibus and are accessible under GEO Series accession number GSE137680. Both data sets are available via a GEO SuperSeries that represents the publication as a whole (GSE138009). The links to the deposited data are link to the GEO Superseries (https://www. ncbi. nlm. nih. gov/ geo/ query/ acc. cgi? acc= GSE138009), link to the microarray data (https://www. ncbi. nlm. nih. gov/ geo/ query/ acc. cgi? acc= GSE137680), link to the scRNA-seq data (https://www. ncbi. nlm. nih. gov/ geo/ query/ acc. cgi? acc= GSE130070).

Open access
This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https:// creativecommons. org/ licenses/ by/ 4. 0/.