Gene Expression Signature for Prediction of Golimumab Response in a Phase 2a Open-Label Trial of Patients With Ulcerative Colitis

Golimumab, a tumor necrosis factor antagonist, is an effective treatment for patients with moderate-to-severe ulcerative colitis (UC); however, more than 50% of initial responders lose their response to the drug within the first year of therapy. A gene expression signature identified in colon biopsies collected before treatment was associated with response to infliximab, and was subsequently refined to associate with mucosal healing in response to golimumab. We performed a phase 2a open-label study of 103 golimumab-treated patients with moderate-to-severe UC to test whether the baseline gene expression signature could be used to predict which patients would achieve mucosal healing, clinical response, and clinical remission at weeks 6 and 30 of treatment. The gene expression signature identified patients who went on to achieve mucosal healing at treatment week 6 with an area under the receiver operating characteristic curve (AUCROC) of 0.688 (P = .002) and at week 30 with an AUCROC of 0.671 (P = .006). The signature identified patients with mucosal healing with 87% sensitivity, but only 34% specificity, limiting its clinical utility. The baseline gene expression signature did not identify patients who went on to achieve clinical remission or clinical response with statistical significance. Further studies are needed to identify biomarkers that can be used to predict which patients with UC will respond to treatment with anti–tumor necrosis factor agents. ClinicalTrials.gov no: NCT01988961. Golimumab, a tumor necrosis factor antagonist, is an effective treatment for patients with moderate-to-severe ulcerative colitis (UC); however, more than 50% of initial responders lose their response to the drug within the first year of therapy. A gene expression signature identified in colon biopsies collected before treatment was associated with response to infliximab, and was subsequently refined to associate with mucosal healing in response to golimumab. We performed a phase 2a open-label study of 103 golimumab-treated patients with moderate-to-severe UC to test whether the baseline gene expression signature could be used to predict which patients would achieve mucosal healing, clinical response, and clinical remission at weeks 6 and 30 of treatment. The gene expression signature identified patients who went on to achieve mucosal healing at treatment week 6 with an area under the receiver operating characteristic curve (AUCROC) of 0.688 (P = .002) and at week 30 with an AUCROC of 0.671 (P = .006). The signature identified patients with mucosal healing with 87% sensitivity, but only 34% specificity, limiting its clinical utility. The baseline gene expression signature did not identify patients who went on to achieve clinical remission or clinical response with statistical significance. Further studies are needed to identify biomarkers that can be used to predict which patients with UC will respond to treatment with anti–tumor necrosis factor agents. ClinicalTrials.gov no: NCT01988961. See editorial on page 963. See editorial on page 963. What You Need to KnowBackground and ContextValidated biomarkers are needed to predict which patients with ulcerative colitis will respond to treatment with anti-tumor necrosis factor therapies.New FindingsA colonic gene expression signature predicted mucosal healing response to golimumab with high sensitivity in a prospective phase 2a open-label study of ulcerative colitis patients.LimitationsAlthough the sensitivity of the gene expression signature was high, the specificity was low, reflecting a high false positive rate and limiting its clinical utility.ImpactThis study advances precision medicine in IBD by prospectively evaluating a predictive biomarker and revealing several challenges regarding replication of biomarkers for response prediction. Validated biomarkers are needed to predict which patients with ulcerative colitis will respond to treatment with anti-tumor necrosis factor therapies. A colonic gene expression signature predicted mucosal healing response to golimumab with high sensitivity in a prospective phase 2a open-label study of ulcerative colitis patients. Although the sensitivity of the gene expression signature was high, the specificity was low, reflecting a high false positive rate and limiting its clinical utility. This study advances precision medicine in IBD by prospectively evaluating a predictive biomarker and revealing several challenges regarding replication of biomarkers for response prediction. Golimumab (SIMPONI), an anti–tumor necrosis factor drug, has demonstrated efficacy vs placebo in patients with moderate-to-severe ulcerative colitis (UC); however, many patients are primary anti–tumor necrosis factor nonresponders, defined as a lack of clinical response to induction therapy.1D’Haens G.R. et al.Am J Gastroenterol. 2011; 106: 199-212Crossref PubMed Scopus (386) Google Scholar Additionally, more than 50% of initial responders experience loss of response within the first year of therapy.2Sandborn W.J. et al.Gastroenterology. 2014; 146: 96-109Abstract Full Text Full Text PDF PubMed Scopus (581) Google Scholar Predicting whether a patient will respond to a particular treatment would allow for a patient-specific treatment plan, with expectations of improved safety and efficacy.3Bek S. et al.Aliment Pharmacol Ther. 2016; 44: 554-567Crossref PubMed Scopus (73) Google Scholar, 4Tew G.W. et al.Gastroenterology. 2016; 150: 477-487Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar The development of biomarkers that predict individual patient responses to specific inflammatory bowel disease (IBD) therapies has been challenging, partially due to disease complexity, cohort variability, and heterogeneity of molecular, endoscopic, and clinical endpoints. Here we describe the identification and prospective replication of a colonic gene expression signature to predict golimumab mucosal healing response in patients with moderate-to-severe UC. The predictive gene expression signature was initially identified from colon biopsies collected in the ACT1 infliximab study (Supplementary Methods).5Arijs I. Li K. Toedter G. et al.Gut. 2009; 58: 1612-1619Crossref PubMed Scopus (173) Google Scholar Baseline expression levels of a panel of 109 probe sets distinguished responders (n = 12) from nonresponders (n = 10) at week 8 with >90% sensitivity and specificity. Response was defined as complete mucosal healing and histologic normalization (a Mayo endoscopic subscore of 0 or 1 and a grade of 0 or 1 on the Geboes histological scale).6Geboes K. et al.Gut. 2000; 47: 404-409Crossref PubMed Scopus (351) Google Scholar The predictive panel of 109 probe sets, representing 81 unique genes, was refined using baseline gene expression from colon biopsies (Supplementary Methods) collected in the PURSUIT golimumab study.7Sandborn W.J. et al.Gastroenterology. 2014; 146: 85-95Abstract Full Text Full Text PDF PubMed Scopus (369) Google Scholar A 13-gene signature (Table 1) achieved the maximum area under the receiver operating characteristic curve (AUCROC) value for predicting week 6 mucosal healing response in PURSUIT (AUCROC of 0.768). Hereafter, the length-13 gene signature is referred to as the molecular prediction signature (MPS).Table 1Genes Included in the MPS PanelGene symbolGene nameCMTM2CKLF-like MARVEL transmembrane domain containing 2C5AR1complement C5a receptor 1FGF2fibroblast growth factor 2GKglycerol kinaseHGFhepatocyte growth factorIL1RNinterleukin 1 receptor antagonistLILRA2leukocyte immunoglobulin like receptor A2NAMPTnicotinamide phosphoribosyltransferasePAPPApappalysin 1SNCAsynuclein alphaSOD2superoxide dismutase 2; mitochondrialSTEAP4STEAP4 metalloreductaseZBED3zinc finger BED-type containing 3MPS, molecular prediction signature. Open table in a new tab MPS, molecular prediction signature. The MPS was then prospectively evaluated for prediction of mucosal healing, clinical response, and clinical remission in an open-label study (PROgECT; NCT01988961) of 103 patients treated with golimumab (Supplementary Methods). The primary objective of the PROgECT study was to evaluate the accuracy of the MPS derived from gene expression in colon biopsies collected at screening to predict mucosal healing at week 6. Secondary objectives to evaluate the accuracy of the MPS to predict clinical response and clinical remission at weeks 6 and 30 and mucosal healing at week the primary a receiver operating characteristic curve for MPS was for mucosal healing on the of and false at week 6. The AUCROC was 0.688 (P = a accuracy of the MPS to predict week 6 mucosal to patients mucosal healing or for of on A with a of of and a specificity of on with a of of and a specificity of of and Secondary of 6 = 30 = the 103 treated patients in the primary patients from 1 due to and 6 patients from due to lack of biomarker area under the receiver operating characteristic MPS, molecular prediction of patients. Open table in a new tab the 103 treated patients in the primary patients from 1 due to and 6 patients from due to lack of biomarker area under the receiver operating characteristic MPS, molecular prediction of patients. Additionally, the MPS predicted mucosal healing at week 30 = of the for clinical response at weeks 6 and 30 and for clinical remission at week 6 that the accuracy of prediction was than (Table of clinical remission at week 30 a positive = The PROgECT study prospectively the of a gene panel in colon biopsies to predict golimumab mucosal healing response in patients with moderate-to-severe UC. The predictive of the MPS was by the and that the MPS was than at predicting mucosal healing at weeks 6 and The of the MPS was the high sensitivity of the the specificity of the MPS was in PROgECT than in reflecting a high or of mucosal healing the MPS is not as a response prediction of the specificity of the MPS evaluated and are in the in the PROgECT and PURSUIT patient baseline of and a rate of in PURSUIT with the of the MPS from to Additionally, in the was The week 6 the and in of the PROgECT in a in a patient a mucosal healing or in of the patients (Supplementary the MPS is to predict an defined the MPS will be in the the MPS This is by the that of the MPS to predict mucosal healing or mucosal healing defined by an of 0 improved the of the MPS (Supplementary golimumab levels in patients who predicted to but did the to response prediction. was a of patients in the with the at week the in regarding the of drug levels on the MPS The PROgECT study advances precision medicine in IBD and disease in which treatment could the MPS in several challenges regarding replication of biomarkers for response prediction in an disease like challenges to identify response for was by and of to study and and of of the and of the for to study and of and of of the and of the for to study and and of of the of the for and to study and and of and of the for to study and and of and of the for to of of the of the for and study was in study and and of and of the for to of and of of the for and statistical to study and and of and of the for to and of of the and of the for to study and of and of the for to study and and of and of the for to and of and of the for to study and and of of the and of the for to study and and of and of the for to study and of and of of the and of the for to study and and of and of the for to study and and of the for to study and and of and of the for to study and of and of of the and of the for to study and and and of statistical and of the for The predictive gene expression signature was initially identified in the ACT1 infliximab study from a of patients who to in the I. Li K. Toedter G. et al.Gut. 2009; 58: 1612-1619Crossref PubMed Scopus Google Scholar mucosal biopsies collected before infliximab treatment was and then with Baseline gene expression was evaluated for the to week 8 responders (n = 12) from nonresponders (n = Response was defined as complete mucosal healing and histologic a Mayo endoscopic subscore of 0 or 1 and a grade of 0 or 1 on the Geboes histological for ulcerative colitis was defined as an endoscopic subscore of or and a grade of or on the histological K. et al.Gut. 2000; 47: 404-409Crossref PubMed Scopus Google Scholar A of 109 probe sets was at baseline responders and nonresponders the 109 probe sets, of infliximab for patient was with the S. et 2000; Scopus Google Scholar using The panel of 109 probe sets was to predict week 8 response with >90% sensitivity and specificity. The predictive panel of 109 probe sets, which to 81 unique genes, was then in an the PURSUIT golimumab W.J. et al.Gastroenterology. 2014; 146: 85-95Abstract Full Text Full Text PDF PubMed Scopus Google Scholar using gene expression from patient collected at The 109 probe panel was to predict mucosal healing response at week 6 in PURSUIT (n = with an area under the receiver operating characteristic curve (AUCROC) of The ACT1 and PURSUIT for by evaluating the gene expression of cohort of with UC in to that of colon biopsies collected from (n = which the UC A of the that in ACT1 patients with was in PURSUIT patients with associated with and The predictive panel of 109 probe sets was subsequently refined in the PURSUIT study to a of that could predict mucosal healing at week 6. healing was from the subscore of the Mayo and defined as an subscore of 0 or collected at baseline from a of patients in PURSUIT initially by using on the with the objective of predictive and the before development on the A of representing 81 unique to the 109 probe and and and a of the gene expression to and to with and and which used the gene expression of to with and to that associated with response in patients used within the S. Scopus Google et Scopus Google Scholar was evaluated using receiver operating characteristic with many for on the of The of the and of the of the gene of for gene signature in the PURSUIT study using on the was used for and and normalization of the PURSUIT study that using on the and as genes, which achieved an of with achieved the maximum area under the curve (AUCROC) value for prediction of week 6 mucosal healing response, a gene signature unique length-13 gene signature was to predict week 6 mucosal healing responders (n = and nonresponders (n = at baseline with a sensitivity and specificity of and and an AUCROC of The length-13 gene signature was to predict week 8 clinical response, defined as complete mucosal healing and histologic in the ACT1 cohort (n = with an AUCROC of A of was to the length-13 gene signature to patients or for mucosal This was the maximum of sensitivity and specificity of prediction. of which the positive predictive was The associated with inflammatory response, and with baseline expression of in mucosal healing nonresponders with mucosal healing Hereafter, the length-13 gene signature is referred to as the molecular prediction signature The PROgECT study was a open-label study and The and by an and patients to the study and and the This study was in with the in the of and was with and with at a was identified and and patients at the from the efficacy patients an of UC at and moderate-to-severe disease defined as a Mayo of 6 to with an endoscopic subscore on the subscore by an response or to 1 or more of the or to of UC K. et al.Gut. 2000; 47: 404-409Crossref PubMed Scopus Google Scholar patients in the study the induction of at week 0 and at week week 6 and week patients the of golimumab that was for UC in the weeks or the treatment was in which golimumab was not for patients with a of weeks was an the treatment phase of the study was by an safety with a safety at week (Supplementary who or and at the of the study their the study or was due to or a maximum of the week which the could be at the of the The in the PROgECT and and and and and and and and and and The primary objective of the study was to evaluate the accuracy of the MPS to predict mucosal healing at week 6. The objectives to evaluate the accuracy of the MPS to predict clinical response and clinical remission at weeks 6 and 30 and mucosal healing at week objectives to evaluate the accuracy of the MPS to predict mucosal healing, clinical response, and clinical remission defined as the response at weeks 6 and disease Mayo at week and week et J PubMed Scopus Google Scholar at baseline and the of the treatment at week 6 and week 30 on the subscore by a from a panel of who to patient and The endoscopic on the identified in the bowel the at high of colon by was for patients. the of and the subscore from the before the study was W.J. et al.Gastroenterology. 2014; 146: 85-95Abstract Full Text Full Text PDF PubMed Scopus Google Scholar healing was defined as a Mayo subscore of 0 or G. W.J. et al.Gastroenterology. Full Text Full Text PDF PubMed Scopus Google Scholar, et Ther. Full Text Full Text PDF PubMed Scopus Google Scholar response was defined as a from baseline in the Mayo and with a subscore of 0 or 1 or a from baseline in the subscore and clinical remission was defined as a Mayo with individual subscore W.J. et J PubMed Scopus Google Scholar, G. W.J. et al.Gastroenterology. Full Text Full Text PDF PubMed Scopus Google Scholar to from the at screening used to and the expression levels of the MPS using the A signature on the MPS was for collected at baseline and weeks and for of collected at baseline and weeks and for and collected at baseline and weeks and for of golimumab golimumab was using a with a of et Ther. Full Text Full Text PDF PubMed Scopus Google Scholar for the of a collected at baseline and weeks and as positive at in their and clinical and the and baseline disease for treated patients. on treated and biomarker performed for treated patients who biomarker at for treated patients. The primary was that the AUCROC of the MPS to predict mucosal healing subscore of 0 or 1) at week 6 would be than accuracy than AUCROC value predictive A curve was by the positive the on with the MPS, using of MPS K. J Google Scholar The AUCROC was using a to the accuracy of the MPS to predict the efficacy of healing, clinical or clinical PubMed Scopus Google Scholar the AUCROC is than the accuracy is than and a AUCROC value predictive The with its and is AUCROC of of the primary sensitivity and and specificity using sensitivity and and positive predictive to performed for the primary for which are the accuracy of the MPS in predicting clinical response at weeks 6 and clinical remission at weeks 6 and and mucosal healing at week not for as and for and and for used to The was used to the its and the associated week 6 of the golimumab and the MPS, in in MPS the 1 and and and The for study was on from the PURSUIT W.J. et al.Gastroenterology. 2014; 146: 85-95Abstract Full Text Full Text PDF PubMed Scopus Google Scholar in which mucosal healing with an AUCROC of an AUCROC of patients in at for an AUCROC that is than = to that the MPS sensitivity with specificity than in predicting mucosal healing, patients for the patients for which a statistical of for the primary The PROgECT study was at in (n = (n = and (n = the 103 patients patients golimumab patient the PROgECT PURSUIT and PURSUIT that in the used for development of the MPS (Supplementary The of patients at baseline was the study the of patients was in PROgECT than in Additionally, baseline levels in PROgECT with 103 patients golimumab at baseline and at week patients study before week 6 and patients weeks and patients weeks treatment was by of patients before week 30 and of patients before week of patients study week The of treatment was with a of The of was and Baseline et W.J. et al.Gastroenterology. 2014; 146: 85-95Abstract Full Text Full Text PDF PubMed Scopus Google patient patient was not or in and disease in UC of to of Mayo of UC Mayo to Mayo in in in of of patients with ulcerative patient patient was not Open table in a new tab of of patients with ulcerative the 103 patients in the efficacy patients from 1 from efficacy due to week of the induction of patients achieved mucosal healing, clinical remission was in of patients. of the patients achieved clinical response at week 6. week of patients achieved mucosal healing and clinical response as for week clinical remission was in as many patients mucosal healing was achieved in of clinical response and clinical remission achieved in and of The defined as the response at weeks 6 and The Mayo at 6 the the was the and and did not the week 6 and week the and and the was at week 6 and at week was at week 6 and at week week patients in the efficacy for and patients positive for to golimumab. efficacy was in patients with to golimumab and of patients achieved mucosal healing, clinical response, and clinical than in patients to golimumab and of patients achieved mucosal healing, clinical response, and clinical golimumab at week with a of week the golimumab was The of and and week 6 and the safety are in the and the was UC. 1 was The only was patients who an of the safety 1 of the was A of the safety and only within the first 6 weeks of the associated with of in the The was and patients study due to of the 103 patients the study The for before week 30 was an of UC by an than of UC and patient was to A of patients study their safety of the safety of due to with 1 or of of patients with Open table in a new tab of of patients with The and in week 6 was of the patients in PROgECT = we the week 6 by the to the by the The week 6 the and in of the PROgECT patients (Supplementary The in in a in a patient a mucosal healing or at week 6 in of the patients. The primary of the was in patients who an of by the vs an of 1 by the at week 6 of PROgECT (Supplementary The of the MPS to predict mucosal healing as by the was The AUCROC of the MPS to predict week 6 mucosal healing as by the was of = The AUCROC of the MPS to predict week 30 mucosal healing as by the was of = of = at week Open table in a new tab The accuracy of the MPS to predict mucosal healing was than of of and the to predict clinical response of of and = or clinical remission of of and = was not in we that a of patients who an of by the a of 1 by the the predictive to only patients at the of the = 0 or the mucosal healing be predicted by the a was performed to whether the patients with of 1 and would the accuracy of the MPS in predicting mucosal healing A of patients the for (n = patients with a week 6 of = patients with a week 6 of The MPS was to predict mucosal healing in of patients with an AUCROC of was performed to whether golimumab associated with the predictive of the on their drug at week 6. AUCROC value on the MPS was derived for was to that drug to the specificity of the MPS Additionally, the of patients who false at week 6 the MPS predicted to be mucosal healing responders but did not in of their week 6 drug was a of patients in the with the was patients who before the week 30 from the MPS prediction of week 30 mucosal A of patients the for The MPS was to predict mucosal healing in of patients with an AUCROC of = of 6 area under the receiver operating characteristic MPS, molecular prediction of not a 1 and and and Open table in a new tab area under the receiver operating characteristic MPS, molecular prediction of not a 1 and and and of the specificity of the MPS been the in the PROgECT and PURSUIT patient disease than the could be in the baseline gene expression the of the Baseline disease Mayo disease and and in PROgECT and in the of patients in PURSUIT who in the predictive in baseline levels PROgECT and with in PROgECT in PROgECT with in Additionally, of patients at baseline in PURSUIT and a of patients in PURSUIT of patients in PROgECT in of in study than in be due to a more that was used for is to on the of many on the accuracy of the the that the cohort used to the MPS from the cohort used to the MPS in that the of the MPS from to for the specificity or the high of in the in the week 6 and mucosal healing was in a patient a mucosal healing vs a could the predictive of the MPS, as the MPS was to predict mucosal healing We that the week 6 the and in of the PROgECT patients. The in in a in a patient a mucosal healing or in of the patients. The predictive accuracy of the MPS did not using endoscopic whether an was by a or a not the specificity of the MPS in PROgECT in with the in the of and the of the mucosal healing which is high to prediction of mucosal the MPS is an defined the MPS will be This is by the that of the MPS to predict mucosal healing or mucosal healing defined by an of 0 improved the predictive of the is that of the in the mucosal healing is and to on the of the response are needed to was that patients who predicted to but did to achieve drug levels at week 6 and week 30 and to their response prediction. the MPS was to of patients was to that drug levels to the specificity of the was a of patients patients who the MPS predicted to respond but did not in the with the The in patients regarding the drug levels and the predictive of the We the week 6 is a to achieve mucosal healing in a of patients with a placebo week 6 be to allow for of mucosal healing response in an individual we the of the MPS to predict week 30 mucosal The of the gene panel to predict week 30 mucosal healing (AUCROC = was to the at week 6 (AUCROC = We the that the MPS patients who are to the and response and that week 6 responders who response by week 30 would the predictive we the of the MPS to predict mucosal of patients mucosal healing, was a an in the predictive (AUCROC that in the the predictive Further we that the MPS only to patients at the of the patients with of 0 or in an in the predictive accuracy (AUCROC = Although the specificity of the MPS in PROgECT with the sensitivity Although the MPS did not identify mucosal healing nonresponders, of predicted to be nonresponders the the MPS, which associated with inflammatory response, and baseline expression levels in mucosal healing nonresponders with This of gene expression that patients with high of to anti–tumor necrosis factor inflammatory at the molecular that is more to with treatment. The prediction of will be an to as be a more with will of as prediction of has for or patients in drug on individual biomarker in in by therapies of the of an of in effective for inflammatory bowel disease to in of more than to a and The of an biomarker in of in area is and to the that the for PDF