Plasma‐induced signatures reveal an extracellular milieu possessing an immunoregulatory bias in treatment‐naive paediatric inflammatory bowel disease

Summary

The inflammatory state associated with Crohn's disease (CD) and ulcerative colitis (UC) remains incompletely defined. To understand more clearly the extracellular milieu associated with inflammatory bowel disease (IBD), we employed a bioassay whereby plasma of treatment naive paediatric IBD patients (n = 22 CD, n = 15 UC) and unrelated healthy controls (uHC, n = 10) were used to induce transcriptional responses in a healthy leucocyte population. After culture, gene expression was measured comprehensively with microarrays and analysed. Relative to uHC, plasma of CD and UC patients induced distinct responses consisting, respectively, of 985 and 895 regulated transcripts [|log₂ ratio| ≥ 0·5 (1·4‐fold); false discovery rates (FDR) ≤ 0·01]. The CD:uHC and UC:uHC signatures shared a non‐random, commonly regulated, intersection of 656 transcripts (χ² = P < 0·001) and were highly correlative [Pearson's correlation coefficient = 0·96, 95% confidence interval (CI) = 0.96, 0.97]. Despite sharing common genetic susceptibility loci, the IBD signature correlated negatively with that driven by plasma of type 1 diabetes (T1D) patients (Pearson's correlation coefficient = –0·51). Ontological analyses revealed the presence of an immunoregulatory plasma milieu in IBD, as transcripts for cytokines/chemokines, receptors and signalling molecules consistent with immune activation were under‐expressed relative to uHC and T1D plasma. Multiplex enzyme‐linked immunosorbent assay (ELISA) and receptor blockade studies confirmed transforming growth factor (TGF)‐β and interleukin (IL)‐10 as contributors to the IBD signature. Analysis of CD patient signatures detected a subset of transcripts associated with responsiveness to 6‐mercaptopurine treatment. Through plasma‐induced signature analysis, we have defined a unique, partially TGF‐β/IL‐10‐dependent immunoregulatory signature associated with IBD that may prove useful in predicting therapeutic responsiveness.

Introduction

Inflammatory bowel disease (IBD) consists of a group of chronic conditions that are classified broadly as Crohn's disease (CD) and ulcerative colitis (UC). They are heterogeneous disorders that possess both distinct and overlapping clinical and pathological characteristics. The pathogenesis of IBD is incompletely understood; however, insight gained over the past decade has revealed the involvement of genetic susceptibility and environmental factors that encompass the intestinal microbiome, that together result in altered immune system function 1, 2, 3, 4, 5. Like many autoimmune diseases, including type 1 diabetes (T1D) 6, the worldwide incidence of IBD is increasing 7, suggesting the presence of new or increasing environmental pressures.

New insights into the pathological mechanisms underlying IBD have been fostered by genomewide association studies, which have so far identified >150 genetic risk loci 1, 8, 9. Most confirmed IBD risk loci confer susceptibility to both CD and UC 1, 10; included among these are components of the interleukin (IL)‐23 pathway (IL23R, IL12B, JAK2, STAT3 and TYK2) 11. Among the identified IBD risk loci, nearly 75% overlap with genetic risk loci implicated in other immune mediated diseases 1, 10, 12, 13. Notably, despite their distinct phenotypes there are >10 susceptibility loci (PTPN2, PTPN22, ORMDL3, IL18RAP, IL27, IL10, IL2/IL21, IL2RA, BACH2 and TYK2) associated with both IBD and T1D 10. An understanding of how these common and distinct genetic loci are manifest in the pathogenic processes associated with IBD, T1D and other autoimmune disorders will lead to better understanding of their pathogenesis and identify new therapeutic targets.

A number of strategies have been used to understand the inflammatory state associated with IBD. Chemokines and cytokines act as communicators for immune cells and are thought to be key elements of IBD pathogenesis that govern disease initiation, progression and, ultimately, its resolution 8, 14. There exists a significant literature describing plasma chemokine/cytokine levels of IBD patients 15, 16, 17, 18, 19, 20. However, the results vary widely among studies, highlighting the limitations that direct measurement of plasma cytokine/chemokine levels can present. While locally in high concentrations at sites of inflammation, cytokines may be dilute in the periphery and difficult to directly measure due to the insufficient sensitivity of conventional immunoassays. Furthermore, measurement of a single or few mediators may be uninformative or even misleading, as combinatorial effects are likely important. Direct gene expression profiling of patient's peripheral blood mononuclear cells (PBMC) has also been used to study the inflammatory state associated with IBD 21, 22, where it has been found sensitive enough to differentiate CD and UC. However, 80% of body's lymphocyte pool resides in the gastrointestinal (GI) tract; therefore studies of peripheral leucocytes may not reflect disease activity completely at the site of inflammation. Lastly, direct transcriptional profiling of intestinal biopsies have also been found informative 23, 24; however, biopsies might not reflect the inflammatory processes of the entire GI tract, especially in CD, which can involve some areas of the gut and spare others.

New approaches are needed to assess the immune state associated with IBD. Towards this goal, we have applied a sensitive and comprehensive genomics‐based bioassay whereby serum‐ or plasma‐borne mediators are used to drive transcription in peripheral blood mononuclear cells (PBMCs) of a well‐controlled healthy blood donor. The response is measured typically with a comprehensive transcriptome‐scale array followed by the application of pathway analyses that enable the interpretation of the data in terms of inflammatory and regulatory immune activities 25. We have used this approach extensively to study T1D 26, 27, 28, 29, 30, 31, where we find that plasma of recent‐onset (RO) T1D patients induce a partially interleukin (IL)‐1‐dependent signature consistent with innate immune activation and pattern recognition receptor (PRR) ligand exposure relative to plasma of unrelated healthy controls (uHC). Importantly, the transcriptional signature appears years before T1D onset and before the emergence of autoantibodies 26, 27, 28, 29, currently the best predictive biomarker in T1D. Furthermore, we have found the signature induced by plasma of recent‐onset type 1 diabetes (RO T1D) patients distinct from those driven by plasma of patients with H1N1 influenza, bacterial pneumonia or cystic fibrosis possessing Pseudomonas colonization 29.

The purpose of this pilot study was to evaluate this method as a tool for advancing our understanding of inflammatory processes in IBD. At the time of diagnosis, the vast majority of patients with IBD show evidence of systemic inflammatory process in the form of fever, weight loss, loss of appetite, raised inflammatory markers and gastrointestinal inflammation, presenting with abdominal pain, diarrhoea, blood in stools and endoscopic lesions. This immunologically active period is probably the optimal period to study the inflammatory process underlying the disease state.

Materials and methods

Subject and subject characterization

Plasma samples were obtained from uHC (n = 10) and treatment‐naive CD (n = 29) and UC (n = 15) patients. These subjects were recruited at Children's Hospital of Wisconsin through protocols approved by its institutional review board (IRB 526542‐5 and 01‐15). Written informed consent was obtained from subjects or their parents/legal guardians. CD and UC was defined and classified according to standard criteria 32, 33. uHC were asymptomatic and lacked a family history for any autoimmune disorder. None of the IBD patients were on immunosuppressive therapy and all subjects were free of any known infection at the time of sample collection. The overall subject characteristics are shown in Table 1. Of the 29 CD patients, plasma from 22 was utilized for initial transcriptional signature studies. For the receptor blocking studies we utilized a total of eight CD patients; only one CD patient overlapped with the 22 subjects studied in the plasma‐induced transcription assays due only to sample limitations. These studies utilized an additional seven CD patients (details are provided in Supporting information, Table S1). Peripheral blood was drawn into acid dextrose solution A or K⁺ ethylenediamine tetraacetic (EDTA) anti‐coagulant. Ficoll Histopaque (Sigma‐Aldrich, St Louis, MO, USA) density gradient centrifugation was used to separate blood components immediately after drawing. Separated plasma was stored at −80ºC until use.

Table 1.

Demographic information of study subjects.

	Crohn's disease	Ulcerative colitis	Controls
Number	29	15	10
Average age, years (range)	12 (5–17)	12 (4–17)	15 (9–17)
Males (%)	55	73	10
Disease extent (%)	L1 ileal ± limited caecal: 10·3 L2 colonic: 31 L3 ileocolonic: 58·7	E1 proctitis: 6·6 E2 left‐sided: 13·3 E3 extensive: 6·6 E4 pancolitis: 73·3	n.a.
Mean CRP* mg/dl(range)	3·3 (range <0·5–24·1)	0·18 (<0·5–1·3)	n.a.
Mean albumin† g/dl	3·7 (3–4·7)	4·1 (3·5–4·8)	n.a.
Disease activity‡ (range)	PCDAI‐25 (5–45)	PUCAI‐46 (35–65)	n.a.

*In Crohn's disease (CD) cohort C‐reactive protein (CRP) data not available in 10 patients and unmeasurable in three patients, in ulcerative colitis (UC) cohort CRP data not available in eight patients and unmeasurable in six patients. ^†Albumin data not available in one CD and two UC patients. ^‡Paediatric Crohn's disease activity index (PCDAI) was not available in one patient and paediatric ulcerative colitis activity index (PUCAI) score was not available in eight patients; n.a. = not applicable.

PBMC cultures, RNA extractions and GeneChip analysis

Plasma‐induced signature analyses were conducted as described previously 25, 27, 29. Briefly, in a 24‐well plate, 5 × 10⁵ cryopreserved PBMC (UPN727; Cellular Technology Ltd, Shaker Heights, OH, USA) were co‐cultured with 40% subject plasma in RPMI‐1640 medium. After 9 h culture, total RNA was extracted using TRIzol reagent (Invitrogen Life Technologies, Carlsbad, CA, USA), and induced transcription was measured using Affymetrix GeneChip Human Genome U133 plus 2·0 arrays (Affymetrix, Santa Clara, CA, USA). This array interrogates >47 000 transcripts/variants and was selected for its overall comprehensive coverage. Image data were collected with the Affymetrix Expression Console Software and normalized with Robust Multichip Analysis (www.bioconductor.org/) prior to determining signal log ratios. Principal components analysis (PCA) was conducted with Partek Genomics Suite 6·5 (Partek, St Louis, MO, USA). This package was also used to determine the statistical significance of differential gene expression, which was assessed through analysis of variance (anova) and false discovery rates (FDR). Hierarchical clustering was conducted with Genesis 34. Ontological analyses were conducted with the Database for Annotation, Visualization and Integrated Discovery version 6·7 (DAVID) 35, and the Ingenuity Pathway Analysis (IPA) package (Ingenuity Systems, Redwood City, CA, USA). Data files are available through the National Center for Biotechnology Information Gene Expression Omnibus, accession number GSE71730.

Measurement of plasma chemokine/cytokine levels

We analysed a panel of mediators associated previously with inflammatory bowel disease, including proinflammatory [tumour necrosis factor (TNF)‐α, IL‐1α, IL‐1β, IL‐2, IL‐6, IL‐8 and IL‐17α], anti‐inflammatory [transforming growth factor (TGF)‐β1, TGF‐β2, TGF‐β3, IL‐10, IL‐1RN] cytokines and soluble receptors (TNF‐R1 and TNF‐R2). This was performed using multiplex enzyme‐linked immunosorbent assay (ELISA) (Luminex xMAP; Ocean Ridge Biosciences, Palm Beach Gardens, FL, USA). Concentrations were calculated with the Bio‐Plex manager 4·1 software; a five‐parameter curve‐fitting algorithm was applied for standard curve calculations.

Cytokine receptor blocking studies

To examine the role of IL‐10 and TGF‐β as mediators of the regulatory biased plasma‐induced signatures associated with CD, receptor‐blocking studies were conducted on eight CD patients. Based upon our previously described strategy 27, replicate IBD patient co‐cultures were prepared that possessed 30 µg/ml TGF‐βRII neutralizing antibodies (R&D Systems Inc. catalogue no AF‐241‐NA) or 30 µg/ml IL‐10Rβ neutralizing antibodies (R&D Systems Inc. catalogue no. AF‐874; R&D Systems Ltd, Minneapolis, MN, USA) or both 30 µg/ml TGF‐βRII‐neutralizing antibodies and 30 µg/ml IL‐10Rβ neutralizing antibodies. Experiments utilized non‐specific isotype controls (polyclonal goat IgG; R&D Systems Inc. catalogue no AB‐108‐c). Transcriptional analysis was performed as described above.

Results

Analysis of IBD patients and uHC subjects

A diagnosis and classification of IBD into CD and UC was based on standard criteria 32, 33. We hypothesized that plasma‐borne factors related to inflammatory processes in IBD could be detected through their ability to induce gene expression in a well‐controlled healthy PBMC population. Therefore, co‐cultures were prepared with plasma drawn from newly diagnosed, treatment‐naive CD patients (n = 22), treatment‐naive UC patients (n = 15) and uHC lacking family history of autoimmunity (n = 10). The transcriptional responses of the reporter cell population were measured comprehensively using a high‐density microarray.

First, an unsupervised analysis was conducted that subjected the complete unfiltered data set to PCA. This analysis showed that the uHC samples clustered distinctly from the CD and UC samples. However, the clustering of the CD and UC samples was indistinct (Fig. 1a). To focus the analysis, we identified the differentially expressed probe sets between the CD patients and uHC as well as the UC patients and uHC using thresholds of (|log₂ ratio| ≥ 0·5 (1·4‐fold); FDR ≤ 0·01). The data structure is illustrated in Fig. 1b. Between the two comparisons, a union of 1224 differentially expressed probe sets was identified (910 unique UniGenes). Specifically, 985 regulated probe sets were identified in the CD : uHC comparison and 895 regulated probe sets were identified in the UC : uHC comparison. A significantly non‐random, commonly regulated intersection of 656 (χ² = P < 0·001) transcripts existed between CD : uHC and UC : uHC comparisons. As tabulated (Fig. 1c), the CD : uHC and UC : uHC plasma‐induced signatures exhibited a high correlation (≥0·97). Relative to uHC, the majority of IBD transcripts were regulated in a directionally concordant manner in CD and UC; further, most transcripts were down‐regulated relative to uHC (828 of 985, 84% in the CD patients and 687 of 895, 77% in the UC patients). When comparing the CD and UC groups directly, no probe sets were regulated to the threshold of |log₂ ratio| ≥0·5 (1·4‐fold); FDR ≤ 0·01.

Figure 1.

Identification of a plasma‐induced signature associated with inflammatory bowel disease (IBD). Plasma samples of unrelated healthy controls (uHC) (n = 10) and patients with treatment‐naive Crohn's disease (CD) (n = 22) and ulcerative colitis (UC) (n = 15) were analysed. (a) Unsupervised analysis employed principal component analysis (PCA) and utilized the complete unfiltered data set (54 613 probe sets). uHC are shown as green spheres, CD as orange spheres and UC as purple spheres. (b) Venn diagram illustrating the relationship of the 1224 probe sets regulated to thresholds [|log₂ ratio| ≥ 0·5 (1·4‐fold); false discovery rates (FDR) ≤ 0·01]) in the CD : uHC (n = 985) and UC : uHC (n = 895) comparisons (left panel). One‐way hierarchical clustering (probe sets only) illustrating the mean expression levels of the 1224 probe set union identified when comparing the CD and UC patients to uHC subjects (middle panel). Two‐way hierarchical clustering (probe sets and individual subjects) illustrating the expression levels of the 1224 probe set union identified when comparing the CD and UC patients to uHC subjects (right panel). Expression levels illustrated in the heat map are normalized against the mean expression of the three cohorts. (c) Pearson's correlation coefficients between the CD : uHC and UC : uHC plasma‐induced signatures. (d) Pathway analysis of probe sets regulated differentially in the CD : uHC and UC : uHC signatures. Genes over‐expressed in CD and UC (n = 157 probe sets and n = 208 probe sets, respectively) and those under‐expressed (n = 828 probe sets and n = 687 probe sets, respectively) in CD and UC were evaluated independently for biological pathway enrichment (GO molecular function and GO biological process) using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) annotation tool. Annotations detected through analysis of over‐ or under‐expressed probe sets are denoted by and up or down arrows, respectively. The P‐value defines the significance of the association of a particular biological process or molecular function with the gene list analysed; n.d. = not detected.

Ontological analysis of the CD : uHC and UC : uHC data sets was conducted with DAVID. For each data set the up‐regulated and down‐regulated probe sets were analysed independently (Fig. 1d). Representative down‐regulated Gene Ontology (GO) Biological Processes and Molecular Functions were related to cytokine activity, inflammation and immune cell activation, intracellular signalling. Transcripts annotated under these terms revealed that, relative to uHC plasma, IBD patient plasma induced lower transcription of numerous immune signalling molecules and receptors. This included IL‐1 cytokine family members and signal transducers (IL1A, IL1B, IL1RN, IL36G, IRAK2, IRAK3), TNF family members (TNF, TRAF1, TNFRSF4 and TNFSF9) and the acute‐phase reactant IL6. Relative to uHC plasma, the plasma of IBD patients induced lower levels of chemokine transcripts. This included CXCL1, CXCL2, CXCL3, CCL3, CXCL5, CXCL6 and IL8 involved in neutrophil chemotaxis; CXCL3, CCL4, CCL20 and CCL24 involved in monocyte and lymphocyte chemotaxis as well as CCL22, which has broad chemotactic activity. IBD patient plasma also down‐regulated many other transcripts related to immune function, including TLR2 (Toll‐like receptor 2), TREM1 (triggering receptor on myeloid cells 1) and NLRP3 [NACHT, LRR and PYD domains‐containing protein 3], a member of the NALP3 inflammasome complex and activator of nuclear factor kappaB (NF‐κB) signalling. CD and UC plasma also down‐regulated transcription of genes involved in cell adhesion and motility; these included ICAM1, TJP2 (tight junction protein 2), GJB2 (gap junction beta 2) and GJA2 (gap junction alpha 2).

Both CD and UC plasma induced transcription of a number of transcripts consistent with the presence of interferons (IFNs), including SAMHD1 (SAM domain and HD domain‐containing protein 1), an enzyme that blocks HIV replication in innate immune cells by depleting nucleotide triphosphates; ITGAL (integrin, alpha L), involved in cellular adhesion and co‐stimulatory signaling and SP100 nuclear antigen, as well as IFI44 and IFI44L. Notably, CD plasma induced to the statistical thresholds a greater number of IFN‐induced transcripts related to anti‐viral responses. These included (OAS1) (2′‐5′‐oligoadenylate synthetase 1) and OAS2, IFIT1 (IFN‐induced protein with tetratricopeptide repeat 1), IFIT2, IFIT3, STAT2, IRF7 (IFN‐regulatory factor 7), IRF9 and MX1 (myxoma‐virus resistance 1).

We have used plasma‐induced signature analysis to study extensively the inflammatory state associated with T1D 25, 26, 27, 29, which is more subtle than that observed in IBD. When comparing RO T1D patients to uHC, we previously defined a signature consisting of 762 significantly regulated probe sets (|log₂ ratio| ≥ 0·263, FDR ≤ 20%, anova P ≤ 0·036) 27, 29. RO T1D plasma induces a partially IL‐1‐dependent transcriptional response that includes up‐regulation of transcripts for chemokines, immune receptors and signalling molecules, and down‐regulates genes involved with regulatory processes. Given that both T1D and IBD are autoimmune diseases with common predisposing genetic susceptibility loci, we examined the overall identity of the IBD and T1D signatures (Fig. 2a,b). Unexpectedly, the RO T1D : uHC signature was related inversely to the CD : uHC (Pearson's correlation coefficient: −0·51) and UC : uHC (Pearson's correlation coefficient: −0·64) signatures. This is reflected further by the opposite direction of induction of the aforementioned transcripts (Fig. 2c).

Figure 2.

Comparison of the inflammatory bowel disease (IBD) and recent‐onset type 1 diabetes (RO T1D) plasma‐induced signatures. The 1224 probe set IBD : unrelated healthy controls (uHC) signature defined in Fig. 1 was aligned with the signature defined previously when comparing RO T1D patients (n = 47) to uHC (n = 44). The RO T1D : uHC signature consisted of 762 probe sets regulated to thresholds [|log₂ ratio| ≥ 0·263 (1·2‐fold); analysis of variance (anova) P ≤ 0·036 and false discovery rates (FDR) ≤ 0·2] 29. (a) Venn diagram illustrating the relationship between the IBD : uHC and RO T1D signatures (left panel). One‐way hierarchical clustering (probe sets only) illustrating the mean expression levels of the 1890 probe set union identified when comparing the Crohn's disease (CD) : uHC, ulcerative colitis (UC) : uHC and T1D : uHC signatures (right panel). (b) Pearson's correlation coefficients between the CD : uHC versus T1D : uHC and UC : uHC versus T1D : uHC signatures. (c) Representative regulated transcripts annotated by the ontological analysis (Figure 1D); (a–c) indicate regulated to the defined thresholds in the CD : uHC, UC : uHC, and T1D : uHC analyses, respectively.

Overall, plasma of CD and UC patients induced a transcriptional programme in healthy PBMCs that was highly distinct from the partially IL‐1‐dependent response induced by new onset T1D patients. Consistent with the extent of inflamed tissue in IBD compared to T1D, CD and UC plasma induced greater folds of change relative to uHC plasma than did RO T1D plasma. Paradoxically, relative to uHC and T1D plasma, the combinatorial response induced by IBD plasma was consistent with an extracellular milieu possessing an abundance of immunoregulatory molecules.

Mediators underlying the IBD signatures

We utilized the IPA upstream regulator analysis to identify candidate chemokines, cytokines and signalling intermediaries responsible for the observed plasma‐induced signatures. Overall, the analysis showed a significant inhibition of proinflammatory mediators that included TNF, IFN‐γ, IL‐6, IL‐12 and IL‐1 (Fig. 3a). Consistent with the failure of CD and UC plasma to induce many of the transcripts associated typically with proinflammatory processes, upstream regulator analysis identified IL‐10, long recognized as a hallmark cytokine of regulatory T cells, as a candidate mediator in both the CD : uHC (P = 1·2E‐24, activation Z‐score = 2·3) and the UC : uHC (P = 7·8E‐19, activation Z‐score = 2·4) comparisons. Interestingly, these analyses also identified activation of SMAD7 (mothers against decapentaplegic homologue 7), which is induced transcriptionally by TGF‐β 36, and inhibits downstream signalling of TGF‐β and thus its regulatory functions.

Figure 3.

Identification of ligands underlying plasma‐induced signatures. (a) Upstream regulator analysis using Ingenuity Pathway Analysis (IPA). Analysis was performed on both the Crohn's disease (CD) : unrelated healthy controls (uHC) and ulcerative colitis (UC) : uHC comparisons. Indicated are upstream regulator and its predicted activation status. A Z‐score possessing an absolute value >2 is considered significant. The P‐value (determined with Fisher's exact test) reflects the significance of the overlap between the regulated probe sets within the data set and genes regulated by the transcriptional regulator; n.d. = not detected. (b) A total of 985 significantly regulated probe sets were identified when comparing the CD and uHC cohorts directly. Adding 30 μg/ml α‐transforming growth factor (TGF)‐βRII and 30 μg/ml α‐interleukin (IL)‐10Rβ^– neutralizing antibodies to CD plasma (n = 8 subjects) modulated transcription to mimic uHC sera, directionally altering expression of 735 of 985 probe sets (74·6, χ² P = 2·7E‐8; yellow box). (c–d) Independent addition of α‐TGF‐βRII^– or α‐IL‐10Rβ^–neutralizing antibodies to CD plasma (n = 5 and 7 subjects, respectively) modulated fewer genes (695 of 985 and 633 of 985). (e) Selected well‐annotated proinflammatory transcripts that are down‐regulated by CD patient plasma but reversed when co‐cultures are supplemented with antibodies to TGF‐βRII and IL‐10Rβ. Tabulated are the fold of change and FDR associated with selected probe sets; (a) and (b) denote regulation by TGF‐β and IL‐10, respectively, per Ingenuity Pathway Analysis (IPA) upstream regulator analysis. Non‐specific isotypic control antibodies did not modulate expression significantly (data not shown).

To define further the extracellular milieu, multiplex ELISA was used to measure directly proinflammatory and regulatory mediators implicated in IBD (Table 2). Most of the 15 analytes tested were either not present at detectable levels in any of the cohorts (TGF‐β3, IL‐1α, IL‐6, IL‐17A, TNF‐α, TNF‐β) or not present at significantly different levels when comparing CD or UC patients to uHC (TGF‐β1, TGF‐β2, TNF‐R2, IL‐1β, IL‐2). Among the proinflammatory mediators, IL‐8, a chemotactic factor for neutrophils and other granulocytes, was detected significantly in UC plasma compared to CD plasma. The regulatory cytokines IL‐10 and IL‐1RN (IL‐1 receptor antagonist) were detected at significantly higher levels in UC patient plasma compared to uHC plasma. Soluble TNF receptors function as TNF‐binding proteins and modulate TNF activity. Notably, compared to uHC plasma, tumour necrosis factor receptor‐1 (TNFR1) was found to be significantly lower in both UC and CD plasma.

Table 2.

Multiplex enzyme‐linked immunosorbent assay (ELISA) analysis of inflammatory bowel disease (IBD) patients and healthy controls.

Cytokine/receptor	CD (pg/ml)	UC (pg/ml)	uHC (pg/ml)	Fold CD : uHC	Fold UC : uHC	Lower detection limit (pg/ml)
TGF‐β1	36 691·1 ± 13 903·2	33 860 ± 14 247·5	31 697·9 ± 17 000·1	1·2	1·1	1187·5
TGF‐β2	4565·7 ± 1878·1	4260·5 ± 1974·1	3837·4 ± 2053·8	1·2	1·1	296·9
TGF‐β3	128·6 ± 65·9	140·6 ± 82·8	142·3 ± 75·2	−1·1	−1·0	296·9
IL‐10	0·8 ± 0·8	2·3 ± 2·3 *	0·8 ± 0·4	−1·1	2·8	1·8
IL‐1RN	144·4 ± 148·6	231·3 ± 183·1 *	112·9 ± 76·2	1·3	2·0	37·6
TNF‐R1	339·5 ± 377·8 *	392·9 ± 421·8 *	680·5 ± 253·2	−2	−1·7	158·7
TNF‐R2	383·4 ± 191·7	508·2 ± 228·8	459·2 ± 151·4	−1·2	1·1	232·0
IL‐1α	0·0 ± 0·0	2·0 ± 6·2	0·2 ± 0·9	1·0	8·0	3·2
IL‐1β	0·9 ± 1·7	1·5 ± 1·9	1·0 ± 0·9	−1·1	1·6	0·5
IL‐2	3·6 ± 5·9	4·9 ± 6·1	1·9 ± 3·1	1·8	2·5	4·4
IL‐6	1·4 ± 4·2	1·5 ± 3·3	0·1 ± 0·3	20·6	21·9	10·3
IL‐8	2·4 ± 4·0	10·6 ± 20·8	1·0 ± 1·4	2·6	11·1	1·7
IL‐17A	0·4 ± 0·8	1·1 ± 1·0	0·3 ± 0·4	1·3	3·2	2·2
TNF‐α	1·9 ± 3·7	3·0 ± 3·5	1·4 ± 1·5	1·3	2·1	4·5
TNF‐β	0·0 ± 0·0	0·0 ± 0·0	0·0 ± 0·0	1·0	1·0	4·1

* P < 0·05 two‐sample t‐test. A total of 10 Crohn's disease (CD), 10 ulcerative colitis (UC) and 20 unrelated healthy controls (uHC) subjects were assayed in duplicate, respectively. All the CD and UC patients included in cytokine assays were analysed in the expression studies. Tabulated data are expressed as cohort mean values ± standard error in pg/ml. IL = interleukin; TNF = tumour necrosis factor; TGF = transforming growth factor.

Given the inflammatory phenotype of IBD, the down‐regulation of proinflammatory molecules in the reporter PBMCs and the identification of IL‐10 and SMAD7 in the upstream regulator analysis were unexpected. While the ELISA analyses, which measured total (latent and active) TGF‐β, showed increased levels of TGF‐β1 and TGF‐β2 in IBD patient plasma relative to uHC, the differences did not reach statistical significance. We reasoned that if IBD patient plasma possessed higher levels of IL‐10 and/or active TGF‐β, preparing co‐cultures with CD plasma and supplemented with receptor neutralizing antibodies to IL‐10R and TGF‐βR should reverse significant portions of the plasma‐induced signature (Fig. 3b–e). This analysis utilized an independent cohort of eight CD patients (Supporting information, Table 1). The partial dependence of the CD : uHC signature on TGF‐β was confirmed as introduction of receptor neutralizing anti‐TGF‐βR antibodies directionally reversed induction of 695 of 985 (70·6%; χ², P = 1·8E‐6) of the probe sets comprising the CD : uHC signature. The partial dependence of the CD : uHC signature on IL‐10 was confirmed as introduction of receptor neutralizing IL‐10R directionally reversed induction of 633 of 985 (64·3%; χ², P = 4·5E‐4) of the probe sets comprising the CD : uHC signature. Consistent with the synergistic action of TGF‐β and IL‐10, blocking both TGF‐βR and IL‐10R reversed the induction of a greater proportion of the CD : uHC signature (n = 735 of 985, 74·6%; χ² P = 2·7E‐8; Fig. 3). Overall, these analyses suggest the existence of an overall immunoregulatory extracellular milieu associated with IBD that is TGF‐β‐ and IL‐10‐dependent.

Relationship between the CD plasma‐induced signature and therapeutic response

Limiting drug‐induced toxicity and sustaining corticosteroid‐free remission are the primary goals of medication management in IBD. In paediatric CD, this currently relies upon the use of immunomodulators such as azathioprine or 6‐mercaptopurine (6‐MP), methotrexate and anti‐TNF antibodies such as infliximab and adalimumab, as well as anti‐integrin antibodies such as natalizumab. Among the initial CD cohort of 22 patients analysed by plasma‐induced signature analysis, 15 patients were treated initially with 6‐MP, seven of which had a sustained remission ranging from 7 months to 6 years (mean 2·64 years). The remaining eight subjects failed to respond to 6‐MP (five responded to anti‐TNF antibodies and three failed all available therapies). Approaches that can predict therapeutic response are needed. Therefore, we investigated the relationship between the plasma‐induced signature and therapeutic outcome by comparing directly the 6‐MP responders and non‐responders. A total of 349 significantly regulated probe sets were identified (|log₂ ratio| ≥ 0·5 (1·4‐fold); FDR ≤ 0·01); among these, 11·2% (39 of 349, χ², P = 4·5E‐4) were identified in the initial analyses comparing the IBD patients to uHC (Fig. 1). Significantly identified GO terms associated with this data set were related to inflammation, bacterial ligand exposure, immune activation and chemotaxis. Compared to 6‐MP non‐responders, treatment‐naive plasma of 6‐MP responders induced higher expression of a number of receptors and signalling molecules, including IL13RA1, which binds tyrosine kinase TYK2 37, TLR2, ARHGAP18, which encodes a Rho GTPase activating protein and NCALD, a regulator of G protein‐coupled receptor signal transduction. The plasma of 6‐MP responders also induced greater expression of ACER3, which encodes a ceramidase that promotes cell proliferation 38, the antigen‐presenting co‐stimulatory molecule, CD86, CD1D, as well as CD93 and THBD, members of the group XIV C‐type lectin family. Compared to 6‐MP non‐responders, plasma of 6‐MP responders induced lower expression of IL36G, a proinflammatory member of the IL‐1 family (also known as IL1F9), ANXA1, which encodes the anti‐inflammatory molecule annexin 1, as well as IKZF3, a transcription factor important in leucocyte proliferation and differentiation. Through this analysis of a very limited number of subjects, the data suggest that a continuum of immune states is present among newly diagnosed, treatment‐naive CD patients. The data also suggest that it may be possible to identify at least a subset of 6‐MP responders from non‐responders. Notably, the three CD patients that were non‐responsive to all therapy (indicated in Fig. 4a) also gave rise to a distinct signature with this set of transcripts.

Figure 4.

Plasma‐induced signatures associated with responsiveness to 6‐mercaptopurine (6‐MP) therapy in Crohn's disease (CD). (a) Heat‐maps illustrating 349 significantly regulated probe sets [|log₂ ratio| > 0·5 (1·4‐fold); false discovery rates (FDR) <0·01] identified when comparing seven treatment naive CD patients that were 6MP responders to eight treatment‐naive CD patients that were 6MP non‐responders. Left: mean of 6MP‐responder, 6MP‐non‐responder and unrelated healthy control (uHC) groups; Middle: individual 6MP‐responder and 6MP‐non‐responder; right: expression levels of well‐annotated transcripts, among these 94·6% (330 of 349) were identified in the initial analyses comparing the IBD patients to uHC (Fig. 1). (b) Regulated probe sets associated 6‐MP therapeutic response was evaluated for biological pathway enrichment using Database for Annotation, Visualization, and Integrated Discovery (DAVID) to identify regulated Gene Ontology biological processes and molecular functions. Representative pathway terms, the number of identified genes and significance of enrichment are tabulated. (c) Regulated probe sets associated with 6‐MP therapeutic response were evaluated using the Ingenuity Pathway Analysis (IPA) upstream regulator analysis function. Indicated are upstream regulator and its predicted activation status. A Z‐score possessing an absolute value > 2 is considered significant. The P‐value (determined with Fisher's exact test) reflects the significance of the overlap between the regulated probe sets within the data set and genes regulated by the transcriptional regulator.

Discussion

Cytokines and chemokines play a crucial role in pathogenesis of IBD. This understanding has been developed largely through genetic/genomic analyses of human patients, as well as more mechanistic studies conducted with animal models of experimental colitis. To date, characterization of the complex cytokine/chemokine milieu in peripheral blood of IBD patients has yielded inconsistent and sometimes conflicting results. Both biological and technical causes contribute to this variability, including disease heterogeneity, the use of different reagents and the limited sensitivity of currently available methods such as ELISA.

In understanding the immune state associated with IBD, recognition of the combinatorial effect of the multitude of molecules, inclusive of chemokines, cytokines, inflammatory lipids and microRNAs, is probably more important than studying the levels of individual mediators. To address this gap, in the current study we obtained a well‐controlled PBMC population drawn from a single healthy donor to use as transcriptional reporters to gain insight into the nature of the overall extracellular milieu associated with IBD. We studied newly diagnosed paediatric IBD patients that were naive to treatment and demonstrated that, relative to uHC plasma, CD and UC patient plasma possess a chemokine/cytokine milieu that has an overarching immunoregulatory bias. Specifically, we observed a more than five‐fold reduced expression of genes related to proinflammatory cytokines, neutrophil and monocyte chemotactic factors and other molecules involved in cell movement and signalling. These observations were in stark contrast with what was expected in diseases with significant inflammatory phenotypes.

The results raised several questions. First, what mediators accounted, at least in part, for the overarching immunoregulatory bias? Secondly, if IBD patient plasma reflects an immunoregulatory environment, why do they have an inflammatory phenotype? Finally, is this phenotype specific to IBD or common with other autoinflammatory disorders?

To identify candidate upstream transcriptional regulators underlying the transcriptional signatures we utilized Ingenuity Pathway Analysis (IPA) upstream regulatory analysis. We identified activation of regulatory molecules [IL‐10, IL‐37 and TNF‐α‐induced protein 3 (TNFAIP3)] and inhibition of proinflammatory molecules (TNF, IL‐1, IL‐6, and IL‐17a). Receptor neutralization studies were conducted to characterize further the roles of IL‐10 and TGF‐β in the observed signatures of treatment‐naive CD patients where blocking of TGF‐βIIR and IL‐10R directionally reversed 74·6% of the transcripts regulated by CD plasma. Associated with this reversal was induction of proinflammatory transcripts, including numerous chemokines (CXCL1, CXCL2, CXCL3, CCL3, CCl4 and CCL20) and cytokines (IL‐1a and IL‐1b, IL‐2RA, IL‐6, IL‐8, IL‐19, IL‐24, IL‐36G, TNF and TNFAIP). Our observations are consistent with previous reports that describe a role for TGF‐β in IBD. Babyatsky et al. demonstrated increased TGF‐β transcript levels in the affected mucosa and lamina propria mononuclear cells (LPMC) of patients with active CD and UC 39. Stadnicki et al. reported elevated serum levels of TGF‐β1 in UC patients irrespective of disease activity compared to unrelated healthy controls 40. This report also showed a significantly increased expression of TGF‐β1 mRNA in colonic tissue of adults with UC 40. TGF‐β is known to modulate functional characteristics of the entire spectrum of cells present in the inflammatory infiltrate in CD, including B and T lymphocytes and macrophages 39. TGF‐β has also been found to be a potent chemotactic factor for neutrophils. Finally, Graham et al. showed that TGF‐β enhances collagen deposition by smooth muscle cells isolated from CD tissue, a process that may contribute to fibrosis and stricture formation 41.

The identification of an immunoregulatory bias in the extracellular milieu of IBD patients is enigmatic. Notably, SMAD7 was also identified in the upstream regulatory analysis. SMAD7 is an inhibitory SMAD molecule 42 that is induced by TGF‐β 36, NF‐κB and signal transducer and activator of transcription fator‐1 (STAT‐1) 43, 44, and is known to block downstream signalling of TGF‐β by blocking the phosphorylation of SMAD2 and SMAD3 45. Monteleone et al. has reported over‐expression of SMAD7 in whole mucosal and LPMC samples from IBD patients 46. Moreover, in‐vitro studies with LPMC isolated from the colon of IBD patients revealed that these cells do not phosphorylate SMAD3 and maintain high levels of inflammatory cytokines, following exposure to exogenous TGF‐β1 46. Overall, our analyses and these reports suggest the possibility that IBD is a state of TGF‐β resistance rather than one of decreased TGF‐β production. This concept is consistent with the lack of efficacy that was observed in trials of IL‐10 therapy in the management of CD 47. Recent clinical trials of an oral SMAD7 anti‐sense oligonucleotide, which functions by decreasing the production of SMAD7 and restoring TGF‐β signalling, have shown remarkable efficacy in the treatment of CD 48. While the basis for activation of SMAD7 remains incompletely understood, through in‐vitro studies Ulloa et al. has demonstrated that IFN‐γ signalling through the Janus kinase 1 (JAK1)/STAT‐1 pathway increases SMAD7 rapidly, causing inhibition of TGF‐β‐mediated SMAD3 phosphorylation and a loss of TGF‐β signalling to the nucleus 44. Notably, relative to uHC plasma, CD plasma induced transcription significantly consistent with the presence of IFN, including IFN‐induced protein 44‐like (IFI44L), IFIT1, IFIT2, IFIT3, STAT‐2, IFN regulatory factor 7 (IRF7), IRF9 and MX1.

To evaluate the disease specificity of the IBD : uHC signature, we compared plasma‐induced signatures of CD and UC patients to those of patients with T1D. Plasma of CD and UC patients induced a more robust signature, in terms of fold‐of‐change, that correlated negatively with the partially IL‐1‐dependent T1D signature. This observation was unexpected, given the existence of common predisposing genetic susceptibilities between IBD and T1D. Importantly, the results highlight how similar genetic dispositions can be diversely manifest in different disease settings.

We have used plasma‐induced transcription to study T1D families, specifically healthy siblings of T1D probands who lacked titres for autoantibodies towards islet antigens. Siblings of T1D patients have a ∼6% probability of progressing to T1D, and most of those who progress to disease possess a (DR3) and/or DR4 HLA haplotype 49, 50. We analysed healthy autoantibody (AA)‐negative high HLA risk siblings (DR3 and/or DR4, termed HRS) and healthy autoantibody‐negative low HLA risk siblings (non‐DR3/non‐DR4, termed LRS) and compared them to RO T1D patients and uHC with no family history of T1D 27. Notably, relative to uHC, T1D family members exhibited an elevated, partially‐IL‐1‐dependent, inflammatory state that included significantly elevated plasma IL‐1α, IL‐12p40, CCL2, CCL3 and CCL4 levels 27. Among AA‐HRS, this familial inflammatory state was found more regulated and possessed elevated IL‐10 and TGF‐β levels compared to LRS. In longitudinal studies of HRS, the familial inflammatory state was found to be supplanted temporally by an IL‐10/TGF‐β‐mediated regulatory state. Through flow cytometry studies, emergence of this regulated state was found to parallel peripheral increases in activated CD4⁺/CD45RA^–/forkhead box protein 3 (FoxP3)^high regulatory T cell (T_reg) frequencies 27. Taken together, these findings offer a mechanistic basis to the juvenile nature of T1D. Furthermore, the studies suggest that failures in endogenous regulatory mechanisms that normally manage inherited T1D risk may underlie disease progression. TGF‐β elevation was not identified in T1D as it may be a specific response to dampen inflammation in the GI tract, which is exposed constantly to bacterial and food antigens 42, 51, 52. We hypothesize that activation of SMAD7 by the complex cytokine milieu of IBD creates a TGF‐β refractory state and perpetuation of intestinal inflammation.

Being a cross‐sectional observational study that examined a modest number of IBD patients, this work has inherent limitations. Our aim was to characterize the inflammatory state of treatment‐naive IBD patients. Typically, immunosuppressive therapy is initiated upon establishment of a diagnosis, thus there is generally a very small window for conducting temporal studies of inflammatory processes in IBD. Clearly, longitudinal studies are needed to understand the dynamics of IBD progression. As IBD is a complex disease, healthy siblings of IBD probands will share varying degrees of genetic risk; importantly, it remains to be determined if unaffected family members possess signatures intermediate of those of IBD patients and uHC or whether they possess signatures with even greater regulatory bias. Studying larger populations of children with IBD may also associate unique signatures with specific phenotypes (e.g. penetrating versus non‐penetrating CD).

Clinical trials published during the last 10 years have demonstrated that the absolute rates of remission maintenance are in the range of 25–30% with thiopurines 53 and 40–70% with anti‐TNF drugs 54. Currently there is a need for reliable methods that could predict patient responsiveness to different IBD medications. We explored this possibility by comparing the plasma‐induced signatures of seven CD patients who responded to 6‐MP therapy with a sustained remission to eight patients who did not. This analysis identified a signature of 349 significantly regulated probe sets that included higher expression of number of immune receptors and signalling molecules, and co‐stimulatory molecules. This limited analysis suggests that it may be possible to correlate positive clinical responses to the plasma‐induced signatures of samples collected prior to treatment. This would be an important step towards personalized therapies. However, more extensive studies are needed in order to make such prognostic predictions.

Despite being largely undetectable by direct analysis, we have determined that plasma of CD and UC patients possess factors that induce a unique, partially IL‐10‐ and TGF‐β‐dependent, immunoregulatory signature in PBMCs relative to that of uHC and T1D patients. The fact that IBD patients possess an inflammatory phenotype, despite having elevated levels of TGF‐β and IL‐10, might implicate a cellular unresponsiveness to TGF‐β, possibly stemming from molecules such as SMAD7. Subanalyses of the CD cohort suggest that this approach may lead to the development of biomarkers of therapeutic response in IBD. Although the overall sample size is modest, these studies lay the foundation for further work that aim to characterize more fully the immune state associated with IBD and to define the utility of this approach to identify predictive biomarkers of therapeutic response.

Disclosure

The authors have no disclosures.

Author contributions

B. G. and M. J. H. designed the study, analysed the data and wrote and revised the manuscript; M. L. K. and S. J. conducted the experiments and data analysis; N. H. S., B. L., M. S. and M. R. S. recruited patients, reviewed and revised the manuscript.

Supporting information

Additional Supporting information may be found in the online version of this article at the publisher's web‐site:

Table S1. Additional Subject Characteristics.