Identification of Pathway-Specific Serum Biomarkers of Response to Glucocorticoid and Infliximab Treatment in Children with Inflammatory Bowel Disease

The onset, symptoms, and progression of inflammatory bowel disease (IBD) are highly variable and unpredictable. A variety of phenotypes exist; extra-intestinal inflammation may manifest as uveitis, arthritis, or growth failure in children. Therapies for IBD are focused on the induction and maintenance of remission, and the prevention of longer-term complications of chronic inflammation, such as relapsing disease, steroid-dependence, malnutrition, growth-stunting in children, and colorectal cancer. There is a disconnect between patient symptoms and mucosal inflammation, and increasing evidence in adults with IBD shows that long-term clinical outcomes are not improved by treating to symptom remission, but are improved by directly targeting mucosal inflammation.1, 2 Practically, using ileocolonoscopy and imaging to monitor disease response requires waiting 3–6 months for cycles of repair to occur, followed by re-assessment of healing by endoscopy or imaging, adjusting therapy based on these results, and then repeating the evaluation again.3 Though this may currently be the optimal approach available, acceptance of repeated colonoscopy as a widespread clinical practice or as a clinical trial endpoint may be limited by patient discomfort, procedural risk, anesthesia, and high cost. This is a particularly important issue in pediatric clinical care, and a barrier to recruitment in pediatric clinical trials.4 There is a need for biomarkers to predict response to therapy and optimize treatment regimens to improve quality of life. Such biomarkers are particularly important in children.5 Facing a long or lifetime duration of disease encompassing important stages of development, the disease manifestations, side-effects of current therapies, and exposure to repeated invasive procedures may have greater negative impacts on children and their families. In addition to treatment decisions, serum biomarkers may inform earlier dosing, safety, and efficacy decisions in pediatric clinical trials. Currently, a small number of clinically-utilized biomarkers of disease response are available to the IBD clinician, as recently reviewed in Sands et al.6 New candidates have been identified as potential blood-based biomarkers in IBD, but the effect of specific treatments on the prospective change of these biomarkers has not been investigated. These include proteins, as well as micro RNAs (miRNAs), which are emerging as promising treatment-responsive biomarkers. Recently, serum SERPINA1 (α-1-antitrypsin) was shown to differentiate between mild and more severe forms of adult ulcerative colitis (UC), and appears to be superior to C-reactive protein in this regard.7, 8 In two recent studies, circulating miRNAs were measured in patients with Crohn's disease (CD) and/or UC. One study found 24 serum miRNAs differentially expressed in children with CD.9 The other study found three sets of peripheral blood miRNAs that distinguish between active CD, UC, and healthy adult patients.10 Bridging specific biomarkers to drug response (for example, pharmacodynamic biomarkers) will be a necessary step in establishing a biomarker as a treatment monitoring tool or as an outcome measure in clinical trials. In this exploratory study, our goal was to carry out a broad discovery of pharmacodynamic biomarkers that show response to corticosteroid and anti-tumor necrosis factor-α (TNFα) biologic treatment of pediatric IBD. Our hypothesis was that serum inflammatory proteins and miRNAs that lie in the pathway of drug effect, and that change in the same direction after treatment with both classes of drugs, are candidates for further study as potential surrogate outcome measures. Corticosteroid-treated patients received daily oral prednisone at standard 1 mg/kg dosing up to a maximum of 40 mg daily. All but one steroid-treated patient was newly diagnosed, without baseline exposure to other medications. The one previously diagnosed patient with CD had only been exposed to mesalamine at diagnosis. None of the infliximab-treated patients were treated with concurrent steroids or immunomodulators, and received standard induction dosing of infliximab (5 mg/kg/dose) at time 0 and week 2. As our goal was to identify pharmacodynamic effects of therapy, we chose a longitudinal approach. We focused, in this initial discovery study, on obtaining an early read-out of physiologic biomarker change after administration of therapy. We chose not to wait for a longer period to fully assess clinical or endoscopic response or remission (typically 12–52 weeks of treatment), as this would have the potential to introduce more variability and dilute a pharmacological effect. Two blood samples were obtained from each patient, one before treatment (baseline control), and one after treatment with either prednisone or infliximab (paired sample for longitudinal analysis). Duration of steroid treatment ranged from 3 to 18 weeks. For infliximab, post-drug samples were drawn at 6 weeks after treatment initiation, just prior to the third induction dose. Pediatric Crohn’s Disease Activity Index and Pediatric Ulcerative Colitis Activity Index were recorded at both pre- and post-treatment time points, as were C-reactive protein and erythrocyte sedimentation rate. The specific SOMAscan panels utilized for the two groups were slightly different, in that the prednisone-treated group was tested using a 1,120 protein panel, whereas the infliximab-treated group was tested with a newer 1,300 protein panel. Thus, ~180 additional proteins were assayed in the infliximab group. Eleven of these significantly changed with infliximab, but were not able to be cross-referenced in the prednisone-treated data set (Supplementary Table S4). These include CD177 antigen (CD177), chitinase-3-like protein 1 (CHI3L1), protein S100A12 (S100A12), and C-X-C motif chemokine 9 (CXCL9), which are inflammatory proteins that decreased with infliximab. To identify potential steroid-specific responses, we used bioinformatics to screen differentially expressed proteins for established interactions with either the GR or TNFα. We found that the genes encoding three of these proteins were direct targets of the GR (Figure 2). These include apolipoprotein E, complement subcomponent C1r, and reticulon 4. An additional two, apolipoprotein B and afamin, were direct targets of apolipoprotein E. Other markers were more indirectly connected to the GR or were regulated by TNFα as well. Because they changed exclusively in response to prednisone, and are known to be regulated by the GR, these five proteins may represent steroid-specific biomarkers. We did not aim to correlate biomarkers with established clinical or biochemical response following this rather short-treatment period. However, we did note that a majority of patients in both groups did show expected improvement in clinical and biochemical measures following treatment. Prednisone treatment (n=12) resulted in a significant improvement in C-related protein (P<0.005) and pediatric Crohn’s disease index (P<0.001, n=10 Crohn’s patients), along with a trend of improvement in erythrocyte sedimentation rate (P=0.06). Infliximab treatment (n=7) led to a trend of improvement in each of these three measures (C-related protein, P=0.2; Pediatric Crohn’s Disease Activity Index, P=0.09; erythrocyte sedimentation rate, P=0.07). Non-statistically significant trends also likely reflect our small sample size. We found that 11 of the identified proteins are established to have connections to each other through interactions with NF-κB, TNFα, and/or the GR (Figure 5). This further establishes these molecular changes as part of a systemic network response to effective anti-inflammatory treatment. We have identified 18 proteins and 3 miRNAs that are responsive to both prednisone and infliximab treatment. Both of these drugs target inflammatory TNFα and NF-κB signaling, but they do so via distinctly different mechanisms. Infliximab is an antibody that specifically binds to and inactivates the TNFα cytokine (ligand), blocking its NF-κB-mediated antiapoptotic effects.27 In contrast, prednisone is a corticosteroid that binds to and activates the GR. Activated GR then binds to NF-κB complexes to inhibit downstream TNFα-mediated inflammation.28 However, prednisone-activated GR shows many additional activities outside of NF-κB transrepression, as it also binds directly to hundreds of DNA elements that affect additional gene expression pathways.29 By overlaying molecular changes in serum from patients treated with these partially overlapping drug mechanisms, we have identified ~20 serum-based pharmacodynamic markers of anti-inflammatory therapies. All five of the NF-κB-regulated proteins that decrease with treatment have roles in inflammation. The first is α-1-antitrypsin, or SERPINA1, a serine protease inhibitor that induces T-regulatory cell expansion.30, 31, 32 Notably, serum SERPINA1 already shows potential as a diagnostic biomarker in IBD.7, 8 A second protein, CCL23, is a chemokine known to be elevated in both IBD and rheumatoid arthritis.33, 34 NF-κB regulates CCL23, which then increases expression of pro-inflammatory cytokines, chemokines and adhesion molecules.35 Another protein marker, resistin, or RETN, is an adipocytokine that shows species-specific tissue expression and is induced by TNFα in human macrophages.25, 36 Corticosteroids are known to affect RETN, though these effects may be time-specific.37, 38 Finally, both IGFBP1 and IGFBP2 decrease with prednisone and infliximab. The IGF-binding proteins regulate endocrine actions of insulin-like growth factors, and IGFBP1 is primarily produced by the liver during states of inflammation. IGFBP2 and IGFBP3 have been shown to decrease after high-dose corticosteroids in Crohn’s disease and UC.39 Our results suggest these effects on IGFBP proteins may not be steroid-specific. Twelve proteins increased with treatment of IBD by both drugs. Many of these have a role in resolving tissue damage. CNTF is a polypeptide hormone, which acts as a growth and survival factor to reduce inflammatory tissue destruction in a variety of injury types.40, 41 Cadherin 3, or p-cadherin, is involved with cell adhesion in epithelial tissues and may be involved in wound repair.42 CCL25 and CD36 are particularly interesting as potential efficacy markers in IBD. Both are expressed by M2 macrophages associated with resolving inflammation. CCL25 is required for protection against chronic colitis in animal models.43 CD36 is required, together with lipolysis, for activation of M2 pro-resolution macrophages.44 Together, these markers may reflect the repair of intestinal tissue damage in IBD. We identified three miRNAs that are responsive to both drugs and are known to be induced by inflammatory signaling. Notable among these is miR-146a, which is elevated in several inflammatory diseases and is induced by NF-κB in immune and muscle cells.13, 45, 46, 47 Interestingly, miR-146a (on chromosome 5) has both acute anti-inflammatory affects45 and chronic pro-inflammatory effects,48 and it is proposed that these contrasting effects may help in the staging of the inflammation/resolution process.49 The miR-146 family also includes miR-146b (on chromosome 10), which we find is affected by both treatments as well. These two miRNAs have shared gene targets, although timing of their expression may differ. A treatment response of this miR-146 family is consistent with studies in muscular dystrophy, where we find both prednisone and a novel dissociative steroid (VBP15) reduce TNFα-mediated induction of miR-146b and miR-146a.13, 50 Though less studied, miR-320a is linked to inflammatory disease as well. In colon biopsies from adults with UC, reduced miR-320a is associated with reduced inflammation.51 Together, measuring levels of these miRNAs could help assess inflammatory disease, as well as therapeutic response. The pharmacodynamic biomarkers identified here hold potential for improving clinical care, and for streamlining development of new treatments. However, there are several limitations to our study. This is an exploratory, proof-of-concept study with limited patient numbers, variable sample timing, and heterogeneous patient populations. Because our statistical approach did not adjust for multiple comparisons, there is potential for type II error and false discovery. Further studies are necessary to replicate these findings, and to develop these markers as surrogate outcome measures. Summary data have been included in the Supplementary Data section. Moving forward, a prospective study with more acute and less-variable blood sampling, hours or days after treatment, will define which biomarkers show the earliest responses to treatment, as well as validate biomarker discovery. Acutely responsive biomarkers should then be bridged to objective measures of intestinal inflammation and downstream clinical outcomes, with larger numbers of pediatric IBD patients followed out over a longer-term outcome period. This subset of biomarkers could then be used as outcome measures in drug development for dose finding studies (Phase 2a), and as surrogate biochemical outcome measures to support efficacy (Phase 2b or Phase 3 studies). In addition to biologics and small molecule inhibitors for IBD therapy, novel anti-inflammatory drugs with improved safety profiles are in development for IBD and other chronic inflammatory disorders.50, 52 Based on the successes of this pilot and feasibility study, drug-specific responses to treatment can be measured using specific serum biomarkers. Unanswered questions about the most clinically appropriate biomarkers to track various components of a complex disease like IBD (for example, the type and degree of the inflammatory process, and the state and activity of the healing process) remain. The time-line of the responses, the sensitivity and specificity of response over time compared to established (invasive) clinical measures, cost, and other issues must be addressed. Guarantor of the article: Laurie S. Conklin, MD. Specific author contributions: Heier, Conklin, Hathout, Fiorillo, Damsker, and Hoffman contributed to the study concept and design. Heier, Chaisson, Hathout, and Conklin acquired the data. Gordish-Dressman performed statistical analysis. Heier and Conklin drafted the manuscript. Hoffman and Fiorillo offered critical revision for intellectual content. Funding was obtained by Conklin, Damsker, Heier, and Hoffman. Financial support: This work was supported by the following National Institution of Health grants: National Center for Rehabilitation Medicine pilot grant R24HD050846, Research Program Projects and Centers P50AR060836, U54HD071601 Small Business Technology Transfer (STTR) Grant 1R41DK102235, Clinical and Translational Science Institute at Children’s National UL1TR000075, NIH Award Number UL1RR031988, and the NIH Pathway to Independence Award K99HL130035. Work was also supported by a Sheikh Zayed Institute for Surgical Innovation pilot grant, and The Clark Charitable Foundation. Potential competing interests: None. Supplementary Information accompanies this paper on the Clinical and Translational Gastroenterology website