Differential effects of tofacitinib on macrophage activation contribute to lack of response in ulcerative colitis patients

Abstract

Background and aims: Tofacitinib, a Janus kinase inhibitor, is approved for the treatment of moderate-to-severe ulcerative colitis. Nonetheless, 40-60% of patients will not respond adequately. The mechanisms underlying responses to tofacitinib remain unknown.

Methods: We applied single-cell and/or bulk RNA analysis to biopsies (n = 23 and 63, respectively) from ulcerative colitis patients (n = 31) before and after tofacitinib treatment. Response was assessed using endoscopic and clinical criteria. In vitro-derived macrophages and primary intestinal fibroblasts were used to validate our findings.

Results: Forty percent of patients responded to tofacitinib. Responders exhibited higher baseline JAK-STAT activity, while non-responders had increased baseline NF-kB pathway activation. Response was associated with significant changes in the abundance and/or activation of immune, epithelial, and stromal cells and the downregulation of S100A9, FCGR3A, MMP12 in resident macrophages. In contrast, non-responders showed a significant increase in the number and activation of macrophages and fibroblasts following tofacitinib treatment, including upregulation of MMP9, IL1B, IL6, CXCL1, CXCL8, and S100A9 compared to baseline. In monocyte-derived macrophages tofacitinib drove the hyperactivation of macrophages in response to lipopolysaccharide, but not TNF or IFNγ. This effect is dependent on the inhibition of IL-10 signaling, which is abundantly induced in response to LPS, but not to TNF or IFNγ. In contrast, cultured fibroblasts, which produced no IL-10 regardless of the stimuli, showed no hyperactivation when pre-treated with tofacitinib.

Conclusions: We conclude that resistance to tofacitinib is mediated by the hyperactivation of myeloid cells and we identify IL-10-dependent macrophages as one cellular subset contributing to this resistance.

Keywords: macrophage; single-cell RNA sequencing; tofacitinib.

Graphical Abstract

Figure 1.

Study design and inferred pathway activity using scRNA-seq. (A) Schematic representation of the study workflow. Biopsies (for single-cell RNA-seq, transcriptional analysis and histological validation) and blood were collected from 31 patients before starting tofacitinib treatment (pre-tx) and during follow-up (post-tx). (B) Nancy histological index in biopsies from responder and non-responder patients before and after tofacitinib treatment. Wilcoxon test for matched data (two-tailed P-value). **P < .01. (C) Uniform manifold approximation and projection (UMAP) of 69,813 intestinal cells color-coded by main cell type. (D) UMAPs showing the PROGENy pathway scores for JAK-STAT and NF-kB separated by condition. (E) Heatmap showing the mean of JAK-STAT and NF-kB pathway scores using PROGENy across all cell types in pre-and post-treatment samples. Wilcoxon signed-rank test adjusted using Bonferroni correction: *P < .05 (R Pre-tx vs NR Pre-tx; * is shown for the group with increased higher mean pathway activation); + P < .05 (Pre-tx vs Post-tx in R and NR).

Figure 2.

Treatment with tofacitinib induces significant changes in the abundance of intestinal cell types in responder and non-responder patients. (A) UMAPs of pre-treatment (pre-tx; 33,941 cells) and post-treatment (post-tx; 35,872 cells) intestinal cells color-coded by response to tofacitinib. Bar plots depict the cellular abundance of the six major cell types (epithelium, stroma, plasma and B cells, myeloid cells, cycling, and T cells) for responder and non-responder patients before (pre-tx) starting tofacitinib and during follow up (post-tx). Comparison was performed using the Chi square test *P < .05. (B) Intestinal cell abundance is shown as the enrichment scores of immune (T cells, plasma, and B cells, cycling and myeloid cells) and non-immune (epithelium and stroma) sub-clusters in responders and non-responders relative to baseline (pre-tx). Comparisons were performed using the Chi square test *P < .05 and **P < .01. (C) Gene expression analysis in whole-biopsy bulk RNA. Selected genes were markers for the main cell lineages. Data is shown separately for responders (blue) pre-tx (n = 13 samples) and post-tx (n = 17 samples); and for non-responders (orange) pre-tx (n = 11 samples) and post-tx (n = 22 samples). Data is expressed as arbitrary units (AU) and presented as boxplots. Dotted lines show the standard error of the mean of biopsies from healthy controls (n = 10). Wilcoxon test. False discovery rate-corrected P-values: *< .05; **< .01, and ***< .001.

Figure 3.

Tofacitinib induces significant transcriptional changes in intestinal macrophages and fibroblasts. (A) Uniform manifold approximation and projection (UMAP) of scRNA-seq data for myeloid cells colored by cell subsets (9902 cells) from responders and non-responders before (pre-tx) and after (post-tx) treatment with tofacitinib. (B) Volcano plots showing genes differentially expressed in FOLR2 + macrophages post-tx compared to pre-tx in responder (left panel) and non-responder (right panel) patients. A two-sided Wilcoxon rank sum test was applied. Genes with a false discovery rate adjusted P value < .05, and a fold change (FC) > 1.2 (UPP, dark red) or FC < 0.83 (DWW, dark green) were considered regulated. Additionally, genes with a nominal P-value < .05 are shown in light red (UP) if FC > 1.2 and light green if FC < 0.83. macros, macrophages. (C) Jaccard analyses comparing significantly upregulated (adjusted P.value < .05 & FC > 1.2, UPP) genes in FOLR2 + macrophages post-tx in responder and non-responder patients with canonical markers of M1 or M2 macrophages from Garrido-Trigo et al.(D) UMAP of scRNA-seq data for stromal cells colored by cell subsets (5674 cells) from responders and non-responders before (pre-tx) and after (post-tx) treatment with tofacitinib. (E) Volcano plots showing genes differentially expressed in S1 fibroblasts post-tx compared to pre-tx in responders (left panel) and non-responders (right panel). A two-sided Wilcoxon rank sum test was applied. Genes with a false discovery rate adjusted P value < .05, and a fold change (FC) > 1.2 (UPP, dark red) or FC < 0.83 (DWW, dark green) were considered regulated. Additionally, genes with a nominal P-value < .05 are shown in red (UP) if FC > 1.2 and green if FC < 0.83. (F) Jaccard analyses comparing significantly upregulated (adjusted P.value < .05 & FC > 1.2, UPP) genes in S1 fibroblasts post-tx in responder and non-responder patients with canonical markers of S1 or inflammatory fibroblasts from the study by Garrido-Trigo et al.

Figure 4.

Differential effects of tofacitinib on in vitro-activated macrophages and fibroblasts from UC patients. Experimental design: M-CSF-monocyte-derived macrophages (A) or intestinal fibroblasts (B) from UC patients were exposed to different inflammatory stimuli (LPS, TNF, or IFNγ) in the presence/absence of tofacitinib. Heatmaps showing mean log2 of fold-change (log2FC) in the expression of selected genes. Changes in response to LPS (10 ng/mL), TNF (20 ng/mL), and IFNγ (5 ng/mL) compared to an unstimulated control are shown for (C) M-CSF-monocyte-derived macrophages and (D) fibroblasts. Changes in gene expression induced by tofacitinib (300 nM) on LPS, TNF, or IFNγ-stimulated M-CSF-monocyte-derived macrophages (E) or fibroblasts (F). n = 8 and n = 7 independent experiments, respectively. One-sample t-test false discovery rate-corrected P-values: *<.05, **<.01, ***<.001, ****<.0001. Concentrations of cytokines in supernatants of M-CSF-monocyte-derived macrophages (G) or fibroblasts (H) from UC patients exposed to the different inflammatory stimuli (LPS, IFNγ, or TNF) in the presence of tofacitinib or DMSO as vehicle control. n = 8 and n = 6 independent experiments, respectively. Data are expressed as median ± range. Paired t-test false discovery rate-corrected P-values: *<.05, **<.01, ***<.001, ****<.0001.

Figure 5.

Tofacitinib enhances response to LPS by interfering with IL-10 signaling on macrophages. (A) Heatmap showing the mean log2 fold-change (log2FC) in the expression of selected genes in M-CSF-monocyte-derived macrophages stimulated with LPS, TNF, or IFNγ, in the presence or absence of an anti-IL10 antibody. Data represent n = 3 independent experiments. One-sample t-test false discovery rate-corrected P-values: *<.05. (B) Scatter plot of log2FC of genes regulated by anti-IL-10 mAb in LPS-stimulated macrophages (data from Cuevas et al compared to the log2FC of genes regulated in FOLR2 + macrophages (scRNAseq data) from patients non-responsive to tofacitinib. Pearson test, P = 2.344 × 10^-5. (C) CellPhoneDB cell communication analysis using baseline data in responder (R) and non-responder (NR) patients. Results are presented as a galluvial plot, where lines connect communicating populations, and colors represent the mean intensity of the interactions. ns, not significant; macros, macrophages. (D) Mean expression of mRNA transcripts for each cell type is shown for IL10RB, IL10RA, and IL10 in pre-treatment inflamed samples from R and NR patients. Comparison was performed using Wilcoxon signed-rank test and adjusted using a Bonferroni correction. Significant differences (P < .05) between responders and non-responders are shown in red. (E) IL-10 signaling scores (CytoSig) are shown for pre-treatment inflamed samples. Heatmap shows relative enrichment of IL-10 signaling scores in R and NR patients. Wilcoxon signed-rank test adjusted using Bonferroni correction was applied to compare R versus NR. * P < .05 ** P < .01, *** P < .001.