INTRODUCTION

Ulcerative colitis (UC) is a chronic, immune-mediated inflammatory bowel disease characterized by mucosal inflammation of the colon and rectum and an unpredictable, relapsing-remitting to chronically active disease course (1,2). Current short- and intermediate-term UC treatment goals focus on clinical response and remission, with improvement and eventual resolution of rectal bleeding and normalization of stool frequency. Long-term UC treatment goals include the achievement of endoscopic and histologic healing, hence preventing colectomy, stoma, and dysplasia and/or cancer (3,4).

The pathogenesis of UC involves a complex interplay between environmental, microbial, and genetic factors, which leads to a dysregulated inflammatory state in the colonic mucosa (2,5,6). Determining differential gene expression in patients with UC before, during, and after treatment can help clarify the underlying molecular mechanisms affected by therapeutic agents.

Interleukin (IL)-23 is a heterodimeric cytokine composed of a unique p19 subunit and a p40 subunit shared with IL-12 (6,7). IL-23 has been shown to be a pivotal component to the pathogenesis of many autoimmune diseases, including UC, and therefore has been identified as a therapeutic target (6,7). Mirikizumab (LY3074828) is a humanized immunoglobulin G4-variant monoclonal antibody that inhibits IL-23 by specifically binding to the p19 subunit of IL-23, but not to the p40 subunit that is common to both IL-12 and IL-23 (8). Mirikizumab and other p19-targeted antibodies have demonstrated clinical efficacy in Crohn's disease (9,10) and UC (8,10).

Understanding the molecular effects of p19-targeted antibodies at the transcriptional and translational level ultimately allows for the identification of profiles associated with response or nonresponse to treatment. To assess treatment efficacy and understand its mechanisms for long-term UC management, it is necessary not only to detect transcriptional changes but also to monitor the stability of those changes over time. We previously reported that transcripts related to UC disease activity and pathways involved in resistance to tumor necrosis factor inhibitors (TNFi) were downregulated from baseline to 12 weeks of induction treatment with mirikizumab (11). In this article, we build on that analysis to identify and measure transcriptional changes in patients with UC who responded to mirikizumab or placebo (PBO) treatment at 12 weeks of induction and continued maintenance mirikizumab or PBO through 52 weeks.

METHODS Study design and patients

We first identified differentially expressed genes (DEGs) during the induction period (weeks 0–12) of the NCT02589665 clinical trial evaluating mirikizumab in patients with UC. Then, we identified which of those DEGs were similarly expressed (DEGSEGs) during the maintenance period (weeks 12–52) of that trial. This generated a transcriptional profile of genes associated with a sustained mirikizumab response vs a spontaneous PBO response.

NCT02589665 was a multicenter, randomized, double-blind, parallel-arm, PBO-controlled, phase 2 trial evaluating mirikizumab in patients with moderate-to-severe active UC. Supplementary Digital Content (see Supplementary Figure 1, https://links.lww.com/CTG/A995) provides the study design. A complete list of inclusion and exclusion criteria and details on endoscopic and histopathology assessments are included in the Supplementary Digital Content (see Supplementary Methods, https://links.lww.com/CTG/B2).

NCT02589665 was compliant with the International Conference on Harmonization guideline on good clinical practice. All informed consent forms and protocols were approved by appropriate ethical review boards before study initiation, and all patients provided written informed consent before receiving study drug.

RNA extraction and gene array methodology

Gene expression was measured in 553 colonic tissue biopsy samples from NCT02589665 using the Affymetrix WT protocol on the GeneChip HTA 2.0 arrays. Biopsies from the same subject for each timepoint were pooled for a total of 277 biopsy RNA samples from 249 patients (8). These went through quality control checkpoints (e.g., RNA sample quality/quantity and amplification) and HTA 2.0 processing. Details on RNA sample preparation, quality control, and gene chip arrays are included in the Supplementary Digital Content (see Supplementary Methods, https://links.lww.com/CTG/B2) and have been published previously (11).

Gene expression analyses

For all DEG and DEGSEG tests, limma (12) fitted linear models with LmFit and calculated probability statistics with eBayes (13). limma identified DEGs and DEGSEGs by fitting the linear model (Lm1): X=Pat+Week where X is the matrix of gene expression values, Pat is a vector of patient IDs, and Week is a vector including 2 timepoints. Kruskal-Wallis tests were used in secondary analyses of differential expression in low numbers of targeted genes already identified with limma.

DEG analyses typically focus on genes expressed at a different level between 2 clinical outcomes or timepoints. In the present analysis, the study design had multiple timepoints at baseline, week 12, and week 52. DEGs from baseline to week 12 needed to be identified as well as those DEGs that did not return to baseline levels at timepoints after week 12 through week 52 (similarly expressed genes [SEGs]). To identify DEGs and SEGs, we created a novel methodology grounded in rigorous statistics called DEGSEG analysis.

In DEGSEG analysis, we performed a test of difference between baseline and week 12 to identify DEGs and then a test of equivalence between weeks 12 and 52 to identify SEGs. First, Lm1 was fit using limma for all genes, all patients in the treatment arms, and the baseline through week-12 timepoints (Weeki∈). Fold changes (FC) greater than 2 and P values less than 0.05 for the Week variable identified DEGs. Lm1 was then fit using limma for all genes, all patients in the treatment arms, and week-12 through week-52 timepoints. (Weeki∈). A 2 one-sided test with Θlog2FC ± 0.26 and 2 one-sided test P value less than 0.05 for the Week variable identified SEGs. The boundary for the equivalence range, 0.26, was selected because this value is roughly equivalent to log21.2, corresponding to a change of approximately 20% in effect size. The intersection of DEGs and SEGs is DEGSEGs. These analyses were conducted for the mirikizumab and PBO treatment arms. Transcriptional changes between timepoints were also analyzed using DEG analyses at weeks 12 and 52.

Correlation statistical analyses

To better understand the identified DEGSEGs' clinical relevance, the transcriptional changes associated with mirikizumab response were correlated with measures of UC disease activity (Robarts Histopathology Index [RHI] and modified Mayo score). Pearson correlation coefficients (PCCs) measured the correlation between changes in each disease activity measure and changes in gene expression at different timepoints. PCCs between the crossed timepoint differential expression of exon groups and the change in clinical metrics between baseline and weeks 12 or 52 were calculated using age, sex, and array chip batch as covariates. Clinical metrics included total Mayo score and RHI. Disease-correlated genes (DCGs) significantly correlated with UC disease activity (PCC > 0.5 and P < 0.05) were identified for different combinations of timepoints, clinical scores, and treatment arms (see Supplementary Methods, https://links.lww.com/CTG/B2). The PCC minimum threshold of 0.5 was selected because this value corresponds to a moderate correlation (14). Clinical scores included modified Mayo score, total Mayo score, Mayo endoscopic subscore, UC Endoscopic Index of Severity total score, Geboes score, and RHI.

Transcriptional response in mirikizumab and placebo treatment arms

Patient samples were clustered to evaluate transcriptional response based on identified DEGSEGs. Clinical response was defined as a decrease in 9-point Mayo subscore (rectal bleeding, stool frequency, and endoscopy) of 2 or more points and 35% or greater from baseline with either a decrease of rectal bleeding subscore of 1 or more or a rectal bleeding subscore of 0 or 1. With mirikizumab responder DEGSEGs, we used uniform manifold approximation and projection to perform dimensionality reduction to 2 dimensions (15). After the dimensionality reduction, we fit a Gaussian mixture model, producing 2 clusters with the ClusterR package (16). The cluster with the majority of baseline timepoints was designated as the transcriptional nonresponse transcriptome; the cluster with the majority of timepoints from weeks 12 to 52 was designated as the transcriptional response transcriptome.

Each patient was considered a transcriptional responder or nonresponder using the location of their baseline, week-12, and week-52 timepoints in the transcriptional response and nonresponse clusters. Transcriptional responders had their baseline timepoint in the transcriptional nonresponse cluster and their week-12 and -52 timepoints in the transcriptional response cluster. Patients with any other pattern were considered transcriptional nonresponders. The relative number of transcriptional responders to nonresponders across the PBO and mirikizumab treatment arms was compared to evaluate whether mirikizumab treatment was associated with a greater transcriptional response.

Relationship between coexpressed gene modules and response

Genes from the gene expression matrix were removed if they were in the lower 50% of genes based on expression. The bottom 50% of genes based on variance (i.e., “housekeeping” genes) were subsequently removed, resulting in 25% of genes (N = 5,419) remaining for coexpressed gene (CEG) analysis. The local maximal quasi-clique merger algorithm (17) identified CEG modules in this subset of high-expression and high-variance genes in mirikizumab and PBO responders separately. Gene modules are highly correlated genes that may be associated with a particular pathway; if one gene in a module is expressed, other genes in the module are likely also expressed at the same time. After CEG modules were identified, the Jaccard index (JI) determined the correspondence between mirikizumab- and PBO-derived modules. The JI is a similarity index on a scale of 0–1 where 0 denotes no similarity and 1 denotes complete similarity. Each gene module underwent principal component analysis; the first component was taken as the eigengene. The module eigengenes tracked module expression across timepoints to identify which modules were most related to response.

Ingenuity pathway analysis

Ingenuity pathway analysis (IPA) was performed using the FC and P value from DEGSEG analysis as input. Based on FC and P value, IPA generated enrichment results for pathways and ontologies. This analysis focused on canonical pathways and generated a plot of significantly enriched canonical pathways and a graphical representation of these same ontologies, including their associations.

RESULTS DEGSEG identification from baseline to week 52

In NCT02589665, more than 62% of patients had previously received biologics, and 49% used corticosteroids at baseline; treatment groups were balanced in concomitant therapies at baseline, including corticosteroids, as well as previous exposure to biologics. Baseline disease severity as measured by the Mayo score was balanced across treatment groups (8,11). At week 12, 41.3% (n/N = 26/63), 59.7% (n/N = 37/62), and 49.2% (n/N = 30/61) of patients treated with 50-, 200-, and 600-mg mirikizumab, respectively, had clinical response compared with 20.6% (n/N = 13/63) of patients receiving PBO (P = 0.014, P < 0.001, and P = 0.001, respectively). At week 52, 80.9% (n/N = 38/47) and 76.1% (n/N = 35/46) of patients treated with 200-mg mirikizumab subcutaneously every 4 and 12 weeks, respectively, had clinical response compared with 53.8% (n/N = 7/13) of patients receiving PBO (8).

The initial DEGSEG analysis captured the transcriptional change from baseline to week 12 and from weeks 12 through 52 in responders (Figure 1). Across mirikizumab and PBO groups, 89 DEGSEGs were identified: 63 in mirikizumab responders, 5 in PBO responders, and 21 in both mirikizumab and PBO responder groups (Figure 1a). The 21 intersecting DEGSEGs identified a clear transcriptional response in patients. The identified genes clustered into upregulated and downregulated groups (Figure 1b). Many of the most significant DEGSEGs specific to mirikizumab response were immune-associated (Table 1). When continued transcriptional response was defined as the same direction FC from baseline to week 12 and from weeks 12 to 52, 21 of 35 of the nonimmunoglobulin genes showed a continued response (Table 1).

Figure 1.:

Identified DEGSEGs. (a) A Venn diagram shows the DEGSEGs identified in the miri and PBO treatment arms. (b) The 21 miri-PBO intersecting DEGSEGs are plotted on a heatmap, showing their expression. In Figure 1b, rows are clustered based on expression, and columns are sorted by timepoint (week) and treatment arm (miri or PBO). Brackets indicate whether DEGSEGs were upregulated or downregulated. The black box indicates PBO responders at week 12, and the brown box indicates nonresponders (miri and PBO) at week 52. DEG, differentially expressed gene; DEGSEG, differentially expressed gene that is a similarly expressed gene; Mayo, Mayo score; miri, mirikizumab; PBO, placebo.

Table 1. - Mirikizumab-specific DEGSEGs with immunoglobulins removed DEG logFC DEG FDR SEG logFC SEG FDR SEG TOST P value Response REG1A −3.05E+00 3.10E-06 −6.38E-02 9.73E-01 4.60E-06 Continued MMP3 −2.33E+00 4.23E-05 6.12E-02 9.60E-01 5.44E-03 Reversed TMIGD1 1.73E+00 2.02E-05 3.26E-01 6.78E-01 6.03E-09 Continued CXCL1 −1.72E+00 8.36E-06 6.12E-02 9.54E-01 3.41E-03 Reversed REG3A −1.39E+00 8.46E-05 −1.31E-01 9.03E-01 1.05E-09 Continued TNIP3 −1.69E+00 8.36E-06 2.93E-01 7.17E-01 8.48E-04 Reversed CA1 1.14E+00 2.67E-04 2.50E-01 5.47E-01 1.30E-07 Continued TRPM6 1.18E+00 3.37E-05 1.83E-01 7.85E-01 1.36E-04 Continued MNDA −1.34E+00 1.34E-04 −2.98E-02 9.69E-01 5.98E-08 Continued LOC101928405 1.15E+00 1.33E-04 1.95E-01 8.01E-01 4.99E-08 Continued CD180 −1.30E+00 2.51E-05 −2.49E-02 9.80E-01 5.06E-05 Continued ABCG2 1.29E+00 1.14E-05 1.66E-02 9.82E-01 3.50E-08 Continued GBP5 −1.26E+00 4.01E-05 −8.52E-03 9.87E-01 2.74E-08 Continued LYPD8 1.00E+00 2.49E-04 2.43E-01 6.15E-01 2.68E-06 Continued CHI3L1 −1.27E+00 6.24E-05 2.63E-02 9.69E-01 5.56E-05 Reversed PECAM1 −1.03E+00 1.94E-04 −1.92E-01 7.56E-01 2.15E-03 Continued CXCL9 −1.49E+00 4.23E-05 2.74E-01 6.53E-01 6.72E-06 Reversed CXCL10 −1.20E+00 9.45E-04 −8.98E-03 9.93E-01 1.11E-06 Continued TNC −1.11E+00 7.90E-05 −8.67E-02 8.42E-01 1.10E-03 Continued PLEK −1.24E+00 3.22E-05 4.75E-02 9.43E-01 2.83E-04 Reversed MS4A12 1.03E+00 5.46E-04 1.46E-01 8.04E-01 6.63E-05 Continued SRGN −1.05E+00 4.23E-05 −1.11E-01 8.43E-01 8.55E-07 Continued HMGCS2 1.21E+00 5.13E-06 −7.24E-02 9.24E-01 2.30E-02 Reversed CCL18 −1.25E+00 7.36E-05 1.29E-01 7.89E-01 1.08E-09 Reversed CLDN1 −1.06E+00 2.54E-06 −6.04E-02 9.44E-01 4.63E-02 Continued SLC26A3 1.02E+00 3.69E-04 9.27E-02 8.40E-01 2.49E-04 Continued SLAMF7 −1.03E+00 1.94E-05 −6.48E-02 9.11E-01 2.01E-02 Continued DEFB4A −1.01E+00 5.19E-05 −7.72E-02 9.26E-01 3.11E-09 Continued DEFB4B −1.01E+00 5.19E-05 −7.72E-02 9.26E-01 5.56E-05 Continued CXCL2 −1.18E+00 2.02E-05 1.08E-01 8.75E-01 8.58E-08 Reversed PADI2 1.13E+00 3.54E-06 −6.76E-02 9.03E-01 4.00E-09 Reversed KYNU −1.04E+00 2.02E-05 3.57E-02 9.48E-01 6.38E-09 Reversed CD38 −1.00E+00 1.87E-05 1.62E-02 9.73E-01 8.23E-03 Reversed CFI −1.04E+00 5.83E-07 5.93E-02 9.25E-01 4.76E-03 Reversed CXCL13 −1.10E+00 1.15E-04 2.28E-01 6.50E-01 4.52E-03 Reversed CD274 −1.01E+00 5.19E-05 1.58E-01 7.17E-01 8.49E-10 Reversed

Placebo responder DEGSEGs are not included in this list. The logFC and FDR are listed for both DEGs and SEGs. SEG TOST P value indicates the P value of TOST to determine whether a gene is a SEG. The Response column indicates whether the direction of FC from baseline to week 12 and weeks 12 to 52 are the same (continued) or different (reversed).

DEG, differentially expressed gene; DEGSEG, differentially expressed gene that is a similarly expressed gene; FC, fold change; FDR, Benjamini-Hochberg false discovery rate; SEG, similarly expressed gene; TOST, two 1-sided test.

We also identified 2 mirikizumab-associated DEGSEGs that significantly changed in the 5 week-12 responders who subsequently lost response between weeks 12 and 52. VNN1 is a downregulated DEGSEG that significantly increased from weeks 12 to 52 in the 5 patients who lost response between weeks 12 and 52 (log2FC = −0.31, Kruskal-Wallis P value = 4.72E-2). HMGCS2 is an upregulated DEGSEG that significantly decreased from weeks 12 to 52 in this same patient population of 5 week-12 responders who later lost response (log2FC = 0.14, P value = 4.72E-2). VNN1 was downregulated in responders vs nonresponders at week 52. HMGCS2 was upregulated in responders vs nonresponders at week 52.

Mirikizumab responder DEGSEGs were evaluated for differential expression between responders and nonresponders at week 52. The pattern between responders and nonresponders mirrored patterns in nonresponders as measured by the Mayo score from weeks 12 to 52. In order of significance through the Kruskal-Wallis test, VNN1, ABCA12, TCN1, HMGCS2, OTP2, DUOX2, CFI, AQP8, C4BPA, GUCA2B, SLC6A14, CLDN1, PADI2, and AQP9 expression were all significantly different in responders vs nonresponders at week 52 (Table 2).

Table 2. - Mirikizumab-specific DEGSEGs demonstrating FC in Mayo score responders vs nonresponders at week 52 logFC P value VNN1 −0.50 4.08E-03 ABCA12 −0.53 1.86E-02 TCN1 −0.25 1.86E-02 HMGCS2 0.16 2.62E-2 OTOP2 0.19 2.62E-2 DUOX2 −0.33 3.62E-2 CFI −0.25 3.62E-2 AQP8 0.26 3.62E-2 C4BPA −0.28 4.02E-2 GUCA2B 0.18 4.02E-2 SLC6A14 −0.38 4.46E-2 CLDN1 −0.24 4.46E-2 PADI2 0.14 4.94E-2 AQP9 0.16 4.94E-2

The Kruskal-Wallis test determined P values.

DEGSEG, differentially expressed gene that is a similarly expressed gene; FC, fold change.

Changes in DEGs from weeks 12 to 52

DEG analysis evaluated transcriptional changes between timepoints and between mirikizumab and PBO treatment arms at weeks 12 and 52. The pattern of transcriptional changes in the mirikizumab group markedly differed from PBO group, most notable in the FC size between baseline and later timepoints (Figure 2). The DEGs' significance level was also much higher in the mirikizumab responder group vs PBO. In the mirikizumab group, 4 DEGs at week 12 and 6 DEGs at week 52 had an absolute log2FC greater than 2 (Figure 2a, see Supplementary Figures 3A and 3B, https://links.lww.com/CTG/A997). By contrast, the PBO group had 0 DEGs at week 12 and 2 DEGs at week 52 with absolute log2FC greater than 2 (Figure 2C, see Supplementary Figures 3C and 3D, https://links.lww.com/CTG/A997). At week 12, most DEGs in PBO responders were below an absolute log2FC of 1. These results indicate mirikizumab responders' transcriptional response is much greater and more immediate than in PBO.

Figure 2.:

DEGs at weeks 12 and 52. (a) Volcano plot of baseline to week 12 and baseline to week 52 DEGs in mirikizumab responders, overlaid for all genes. (b) Top DEGs from baseline to week-12 DEGs (|FC|>2.4, P < 0.05) at weeks 12 and 52 in mirikizumab responders. (c) Volcano plot of baseline to week-12 and baseline to week-52 DEGs in placebo responders, overlaid for all genes. (d) Top DEGs identified in (b) from baseline to week-12 DEGs at weeks 12 and 52 in placebo responders. In (a) and (c), the same gene is connected by a solid black line between the week-12 and -52 timepoints. In (b) and (d), the same genes are plotted (i.e., the top baseline to week-12 DEGs in mirikizumab responders). DEG, differentially expressed gene; FC, fold change; NS, nonsignificant.

The mirikizumab group's large effect size at week 12 was also sustained through week 52 (Figure 2b). P values calculated for the mirikizumab group at week 12 vs baseline decreased from week 52 vs baseline, whereas effect sizes were mostly unchanged; this was expected because the transcriptome profiles became more diverse from weeks 12 to 52. The variance of genes expressed at week 52 was greater than that at week 12, which created comparable effect sizes with a decrease in significance. Although many PBO responder DEGs continued to increase in FC magnitude between weeks 12 and 52, the response did not reach the effect sizes observed in mirikizumab responders, even at week 52 (Figure 2d). Transcriptionally, the mirikizumab response was both greater in magnitude and more sustained than the PBO.

Transcriptional response patterns in mirikizumab and placebo treatment arms

Uniform manifold approximation and projection performed dimensionality reduction down to 2 dimensions using the 63 DEGSEGs identified in mirikizumab responders and 21 DEGSEGs in mirikizumab and PBO responders (Figure 3a). After dimensionality reduction, a Gaussian mixture model was fitted, resulting in 2 clearly differentiated clusters. No subclustering appeared within these larger clusters. Each patient represented a set of 3 points. The top cluster contained a greater proportion of points from baseline and was considered the transcriptional nonresponse cluster. The lower cluster contained a greater proportion of points from weeks 12 to 52 and was considered the transcriptional response cluster (Figure 3b). When a patient's baseline point was in the transcriptional nonresponse cluster and their week-12 and -52 points were in the transcriptional response cluster, that patient was considered a transcriptional responder. None of the 7 PBO responders had this transcriptional response pattern (Figure 3c), whereas 17 of the 31 mirikizumab responders were transcriptional responders (Figure 3d; P = 0.02 from the Fisher exact test).

Figure 3.:

Identification of transcriptional “response.” (a) UMAP visualization of all mirikizumab and placebo treatment arm samples by timepoint and GMM cluster. (b) Transcriptional response and nonresponse clusters. (c) UMAP visualization of placebo treatment arm samples by timepoint. (d) Mirikizumab treatment arm samples by timepoint. In (b), (c), and (d), the dashed red line stratifies the transcriptional response and nonresponse clusters. In (c) and (d), each patient's samples are connected by a solid gray line. If the gray line crosses the dashed red line from baseline to week 12 and not from weeks 12 to 52, that patient is a transcriptional responder. GMM, Gaussian mixture model; UMAP, uniform manifold approximation and projection.

Correlation of DEGSEGs with UC disease activity indices

Genes identified as significantly correlated with UC disease activity (PCC >0.5 and P < 0.05) were designated DCGs. Figure 4 presents DCGs significantly correlated with RHI and/or Mayo score at both weeks 12 and 52. Transcriptional changes in multiple immune-related genes were correlated with changes in disease activity, including IL20RA, one of the top 20 downregulated DCGs in the RHI baseline to week-12 and RHI baseline to week-52 comparisons. Supplementary Digital Content (see Supplementary Table 1, https://links.lww.com/CTG/B2) details the top 20 most highly correlated DCGs present in RHI baseline to week 12, RHI baseline to week 52, Mayo score baseline to week 12, Mayo score baseline to week 52, and RHI and Mayo score at week 52. These correlations were not as apparent nor strong in the PBO responders vs mirikizumab responders (see Supplementary Figure 4, https://links.lww.com/CTG/A998), although the very low number of PBO responders (N = 7) limits interpretation.

Figure 4.:

Correlation of mirikizumab responder DEGSEGs with UC disease activity indices. (a) PCC between DEGSEG changes from baseline to week 12 and RHI. (b) PCC between DEGSEG changes from baseline to week 52 and RHI. (c) PCC between DEGSEG changes from baseline to week 12 and Mayo score. (d) PCC between DEGSEG changes from baseline to week 52 and Mayo score. The 20 most positively correlated and 20 most negatively correlated genes are labeled in each panel. DEGSEG, differentially expressed gene that is a similarly expressed gene; NS, nonsignificant; PCC, Pearson correlation coefficient; RHI, Robarts Histopathology Index; UC, ulcerative colitis.

Relationship between coexpressed gene modules and response

CEGs, the coexpression patterns of identified DEGSEGs, were also analyzed across different timepoints. We identified 19 CEG modules in mirikizumab responders (see Supplementary Figure 5A, https://links.lww.com/CTG/A999) and 18 modules in PBO responders (see Supplementary Figure 5B, https://links.lww.com/CTG/A990). There was low correspondence between gene modules as measured by JI. The median JI in mirikizumab CEG modules corresponding to their closest match in PBO CEG modules was 0.22. The median JI in PBO CEG modules corresponding to their closest match in mirikizumab CEG modules was 0.17.

Several mirikizumab responder CEG modules (e.g., 3, 5, and 7) had a very large transcriptional change between baseline and week 12 sustained through week 52, correlating with UC disease activity indices (see Supplementary Figure 5A, https://links.lww.com/CTG/A999). The eigengene from CEG module 3 was significantly correlated with both RHI and Mayo score across timepoints and was associated with immune cell activation and immune response (see Supplementary Figure 6A, https://links.lww.com/CTG/A1000). Similarly, in PBO responders, many CEG module eigengenes progress from baseline through week 52 in either an increasing or decreasing trend. The PBO responder CEG modules 3 (decreasing), 4 (increasing), and 5 (increasing) follow a trend across timepoints, and the CEG module eigengene values also correlated with both RHI and Mayo score. The PBO responder CEG module 4 eigengene had PCC values of 0.56 for RHI and 0.45 for Mayo score and was enriched for other immune processes (e.g., cytokine activity; see Supplementary Figure 5B, https://links.lww.com/CTG/A999).

Pathway enrichment

Pathway enrichment results were generated for the 84 mirikizumab and mirikizumab/PBO DEGSEGs with IPA. Multiple immune-related concepts and pathways were identified, including the IL17 A/F pathway, which is directly downstream of IL-23 (Figure 5).

Figure 5.:

Pathway enrichment in the 84 mirikizumab and mirikizumab/placebo DEGSEGs. (a) Canonical IPA results for the top 10 most significant pathways. (b) Graphical summary of IPA results. DEGSEG, differentially expressed gene that is a similarly expressed gene; IL, interleukin;