miR-31-3p Expression and Benefit from Anti-EGFR Inhibitors in Metastatic Colorectal Cancer Patients Enrolled in the Prospective Phase II PROSPECT-C Trial

Colorectal cancer is a leading cause of morbidity and mortality worldwide (1, 2). EGFR mAbs are effective in metastatic colorectal cancer (mCRC) and can be used alone or in combination with chemotherapy (3). Mutations in the RAS pathway are negative predictive biomarkers of response to anti-EGFR antibodies in patients with mCRC, thus RAS testing on tissue is routinely performed in clinical practice for patient selection (3).

We and others have recently shown that implementing RAS genotyping in pretreatment circulating cell-free DNA (cfDNA) can identify patients who are unlikely to benefit from anti-EGFR therapies (4, 5). Furthermore, our mathematical modeling indicated that resistance to anti-EGFR antibodies is often polyclonal, suggesting that multiple genetic and nongenetic drivers might contribute to treatment failure (5).

miRNAs are short noncoding RNAs controlling gene expression at posttranscriptional level (6). miRNAs are involved in developmental and physiologic processes (6), and are often dysregulated in pathologic conditions such as cancer and inflammation (7). miRNA dysregulation is frequently observed in colorectal cancer and multiple lines of evidence suggest that miRNAs affect a number of cancer hallmarks (8), and drive colorectal cancer initiation (9), progression (10), and resistance to treatment (11). Given their relative stability in tissues and other bio-fluids (12, 13), miRNAs have been proposed as potential biomarkers for colorectal cancer early detection (14, 15), diagnosis (16), and prognosis (17).

miRNA upregulation and/or SNPs in miRNA target genes have been postulated as potential determinants of resistance and sensitivity to anti-EGFR mAbs in early and metastatic CRC (18–21). miR-31-3p expression levels have been examined by RT-PCR in retrospective analyses of the FIRE-3, PICCOLO, NEW-EPOC, and PETACC8 trials (22–26); in these studies, low miR-31-3p expression was associated with improved outcome and prolonged benefit from anti-EGFR treatment. Real-time PCR-based assays for the analysis of miR-31-3p on formalin-fixed paraffin embedded (FFPE) tissues are at an advanced stage of validation (23, 27).

The PROSPECT-C trial was a phase II trial of single agent anti-EGFR antibodies in chemo-refractory mCRC. Patients underwent repeated tissue biopsies of metastatic deposits before and after treatment as well as at the time of treatment response in case of partial response (PR; ref. 5).

In this study we aimed to: (i) validate the association between miR-31-3p expression and clinical benefit from anti-EGFR treatment in pre-treatment tissue biopsies; (ii) test changes in miR-31-3p expression in serial tissue biopsies during treatment in order to assess whether miR-31-3p might be a potential biomarker of acquired resistance; (iii) test whether combining miR-31-3p with RAS testing in cfDNA might improve patient selection.

The PROSPECT-C trial (ClinicalTrials.gov identifier: NCT02994888) was a prospective, phase II, open-label, single center, nonrandomized study of biomarkers of response and resistance to anti-EGFR therapies in KRAS/NRAS wild-type (wt) chemo-refractory mCRC. Patients who were at least 18 years old and had a WHO performance status (PS) of 0 to 2 were considered eligible for this study, if they fulfilled all the following criteria: (i) chemo-refractory (at least two lines of chemotherapy) metastatic disease; (ii) KRAS/NRAS wt (on archival material according to clinically accredited molecular testing); (iii) measurable disease; and (iv) metastatic sites amenable to biopsy. Patients received cetuximab/panitumumab through the Cancer Drug Fund. Written informed consent was obtained from all patients. The study was carried out in accordance with the Declaration of Helsinki and was approved by national institutional review boards [National Research Ethics Service (NRES): 12/LO/0914]. The objectives of the study were to validate known mechanisms and identify novel drivers of response/resistance to cetuximab. Treatment consisted of cetuximab 500 mg/m² once every 2 weeks until progression or intolerable side effects. All but one patient received cetuximab and were anti-EGFR naïve at the time of trial entry; the aforementioned patient was switched to panitumumab due to a Common Toxicity Criteria for Adverse Events (CTCAE) 3.0 Grade II allergic reaction after the first dose of cetuximab, and had previously received three cycles of fluorouracil, oxaliplatin, and cetuximab combination with PR as neo-adjuvant chemotherapy for liver resection in the context of the NewEPOC trial 13 months before entering the PROSPECT-C trial.

All participants were required to have mandatory pre-treatment biopsies [baseline (BL), 6 cores], biopsies at 3 months [if PR by Response Evaluation Criteria In Solid Tumors (RECIST) v1.1 criteria (6 cores)] where clinically and technically feasible, and post-treatment at the time of progressive disease (PD) (6 cores from two suitable progressing metastatic sites). Archival material (primary cancer or original diagnostic biopsies) was assessed where available. Plasma for cfDNA analysis was collected every 4 weeks until disease progression.

ISH assays for miR-31-3p expression in baseline tissue was performed using miRCURY LNA miRNA ISH Optimization Kit for FFPE (Qiagen). Archival tumor material at diagnosis was tested if available (n = 12). ISH on tissue sections was performed following the Exiqon protocol with some modifications. Initially paraffin was removed with xylene incubation for 5 minutes followed by ethanol 100% incubation for another 5 minutes. Tissue sections were then dehydrated in ThermoBrite hybridizer (Leica Biosystems) containing 20 μg/mL of Proteinase K (Roche) for 15 minutes at 37°C. The dehydration reaction was stopped by immersing the slides in PBS, and a prehybridization step was then performed by adding 1× ISH buffer (Exiqon) and incubating the sections for 15 minutes at 56°C. Following the removal of the prehybridization solution, previously denatured (90°C) miRCURY LNA miRNA detection probe (hsa-miR-31-3p; Catalog No. YD006116560; Exiqon) was added to the sections at a 200 nmol/L concentration; sections were incubated with the detection probe overnight at 56°C. The following day tissue sections were sequentially immersed in 5×, 1×, and 0.2× SCC solutions at hybridization temperature for 5 minutes each, and finally transferred in PBS solution at room temperature. Blocking was performed at 30°C in hybridizer followed by incubation with anti-digoxygenin-AP fragments (Sigma-Aldrich) for 1 hour. The sections were then washed three times with PBS-Tween 0.2% for 5 minutes each and incubated for 2 hours with BCIP/NBT Liquid Substrate System (Sigma-Aldrich) for developing the reaction. The reaction was stopped with by immersing the slides in KTBT buffer and counterstained in Nuclear Fast Red (Vector Laboratories). The sections were then dehydrated by ethanol 100% and xylene incubations (for 5 minutes each) and covered with a coverslip.

miR-31-3p expression was graded by two independent pathologists as follows: 0 = no staining; 1⁺ = weak staining; 2⁺ = intermediate staining; and 3⁺ = strong staining. Patients with a 0 or 1⁺ expression score were deemed as low expressors whereas those with a score of 2⁺ or 3⁺ were deemed as high expressors.

Prior to RNA extraction, samples were reviewed by the pathologist and cancer areas were marked for subsequent macro-dissection. Total RNA from FFPE slides (4 × 4 μm sections) was extracted using the Ambion Recover All Isolation Kit (Thermo Fisher Scientific), according to manufacturer's instructions. RNA quantity and quality were assessed by NanoDrop2000 (Thermo Fisher Scientific). Ten nanograms of total RNA were retrotranscribed using the TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific), and RT-PCR was performed using the TaqMan assay for miR-31-3p (assay ID 002113). RNU48 was used as housekeeping gene for normalization, and relative expression was calculated using the 2ΔCt method. miR-31-3p expression was scored as high or low based on the median of the distribution.

Progression-free survival (PFS) was calculated from start of treatment with cetuximab to date of progression assessed radiologically, or clinically. Overall survival (OS) was calculated from start of cetuximab to date of death from any cause or last day of follow-up. Differences in PFS and OS between patients with low expression and high miR-31-3p expression pre-treatment were calculated using the Kaplan–Meier method and compared using the log-rank test. Chi-square test was used to assess the effect of miR expression on overall response to cetuximab treatment. Univariate and multivariate analysis using Cox proportional hazards method were performed to assess effects of age, gender, sidedness of tumor in all 42 patients. In the 34 patients for whom baseline ctDNA results were available, a similar approach was used and multivariate analysis performed. A P value of <0.05 was considered significant.

All the authors reviewed and approved the final manuscript. Researcher performing experiments and scoring tissues were blind to clinical outcome information. Analysis was performed by trial statisticians.

The PROSPECT-C trial recruited 45 eligible patients between November 2012 and December 2016 (Fig. 1). Forty-five percent of patients achieved disease control (PR or stable disease by RECIST 1.1. criteria); median PFS and OS were 2.6 months [95% confidential interval (CI), 1.9–4.2] and 8.2 months (95% CI, 4.2–12.0), respectively. These data have been previously reported by our group (5) and are in keeping with available literature for single agent anti-EGFR treatments in chemo-refractory mCRC (28).

To test the association between miR-31-3p expression and benefit from anti-EGFR treatment, we initially scored miR-31-3p by ISH in 91 baseline tissue core biopsies from 45 patients. Forty-two of those patients (88 cores) could be tested further in this study, as, following extensive previous analyses (5, 29), no cancer was left in three cases.

miR-31-3p was marked as low if scored negative or 1⁺, and as high if scored 2⁺ or 3⁺ in cancer (Fig. 2), whereas stromal miR-31-3p staining due to inflammatory or immune infiltrate (Supplementary Fig. S1) was not taken into account. Positive cells in the stroma were represented by plasma cells, macrophages, and endothelial cells, and most of them showed a faint miR-31-3p expression, with only a limited number of cells characterized by a moderate expression. No significant difference in the proportion of positive stromal cells was observed between anti-EGFR responder and resistant tumors.

At least two different slides for each core were tested and concordance in miR-31-3p scoring among different sections from the same core was 100%. In 32 patients, two different cores from the same metastasis were tested, although, in four cases one of the two cores showed only necrosis and/or inflammation and thus comparison with its parental core was not possible. The concordance in miR-31-3p scoring among different cores in the remaining 28 patients was 89% (three cases scored in different categories and were attributed to the high miR-31-3p category as the average score was above 1⁺). Overall, 24 patients were scored as miR-31-3p low and 18 as miR-31-3p high; patients' demographics based on miR-31-3p expression are presented in Table 1.

To validate the results obtained by ISH, we performed miR-31-3p expression analysis using RT-PCR on 46 cores from 18 patients for whom material was available for RNA extraction. Two different core biopsies were available in 19 cases and the average expression (based on RT-PCR) between the two cores was used to determine the miR-31-3p scoring; for the remaining 8 cases only one core biopsy was tested. The 46 cores included four primary tumors, 29 pretreatment (baseline), four on-treatment (at the time of PR), and nine posttreatment (progression) biopsies. A statistically significant correlation (chi-squared exact test P: 0.003) with 77% concordance between the two miR-31-3p expression tests was observed (Supplementary Table S1).

Next we tested the association between miR-31-3p score based on ISH and clinical benefit from anti-EGFR mAbs. Low miR-31-3p expression was associated with better overall response rate (ORR) defined as PR or stable disease, with 58.3% (14/24) patients showing response in the miR-31-3p low group versus 22.2% (4/18) in the miR-31-3p high group (Supplementary Table S2; chi-squared exact test P: 0.029). A significant depth and duration of response was observed in patients with low miR-31-3p expression (Fig. 3A and B). Median PFS was 4.21 months (95% CI, 1.91–5.56) and 2.27 months (95% CI, 1.55–2.53) in patients with low and high miR-31-3p respectively [HR for miR-31-3p high: 2.03 (95% CI, 1.06–3.91); P = 0.03] (Fig. 3C). Similarly, median OS was 8.88 months (95% CI, 5.53–18.36) and 4.14 months (95% CI, 2.96–8.68) in patients with low and high miR-31-3p respectively [HR for miR-31-3p high: 2.20 (95% CI, 1.09–4.43); P = 0.03] (Fig. 3D). Multivariable Cox regression analysis including miR-31-3p expression, age at diagnosis, gender, and sidedness (30) in the trial cohort (n = 42) confirmed that miR-31-3p was an independent predictor of PFS (Supplementary Table S3) and OS (Supplementary Table S4).

Changes in miR-31-3p expression during single-agent cetuximab treatment have never been investigated before. Here we took advantage of repeated serial tissue sampling in our trial and we tested whether miR-31-3p scoring is altered during or after EGFR inhibition. Analysis of nine patients with long-term response (PFS ≥6 months) and patients with primary progression (PFS ≤ 3 months) revealed no changes in miR-31-3p scoring in either of the two groups (Supplementary Table S5; Fig. 4A). Indeed, analysis of liver, nodal, abdominal wall, and pelvic metastases showed consistent miR-31-3p expression even when different metastases were tested (Fig. 4A and B).

Comparison of miR-31-3p score between archival, treatment naïve tissue (primary colorectal cancer in most of the cases) and pre-cetuximab tissue biopsies (Supplementary Table S6) was concordant in 11 of 12 cases (Fisher exact test P = 0.01) (Supplementary Table S7).

We and others have recently demonstrated the predictive value of RAS testing in pretreatment cfDNA as a valuable and more specific alternative to tissue analysis in the selection of patients eligible for anti-EGFR treatments (4, 5). When we included miR-31-3p in a multivariable Cox regression analysis including age at diagnosis, gender, sidedness, and RAS genotyping in pretreatment cfDNA in patients for whom all the information were available (n = 34), miR-31-3p showed no independent value in predicting PFS (Supplementary Table S8) or OS (Supplementary Table S9). In keeping with these data, when we generated a statistical model combining miR-31-3p status in tissues and RAS genotyping in pretreatment cfDNA (presence/absence of mutations) (Supplementary Fig. S2A–S2C; Supplementary Tables S10 and S11), the interaction tests for ORR, PFS, and OS were nonsignificant (P = 0.213, 0.178, and 0.067, respectively). Among patients who tested as RAS wt in cfDNA from baseline bloods, ORR was 78%, median PFS was 5.10 months (95% CI, 1.91–16.88) and median OS was 15.23 months (95% CI, 1.91–34.08) in patients with low miR-31-3p expression (n = 9). On the contrary, ORR was 25%, median PFS was 2.27 months (95% CI, 1.91–16.88) and median OS was 4.67 months (95% CI, 1.51–12.04) in patients with high miR-31-3p expression (n = 8).

miR-31-3p expression has been tested in a number of retrospective series and retrospective analyses of prospective trials (18, 22–24, 31). Low miR-31-3p expression has been associated with sustained PFS and OS as well as improved ORR in response to EGFR inhibitors. Although these findings have been validated in several studies, the interpretation of these data remains challenging due to the fact that in most of these series anti-EGFR mAbs (cetuximab or panitumumab) were used in combination with different chemotherapy backbones and in different lines of treatment. In the PROSPECT-C trial (5), cetuximab was used as a single agent in a prospective and homogeneous cohort of chemo-refractory mCRC patients; furthermore, miR-31-3p was tested in ad hoc pre- and posttreatment tissues biopsies. Thus, despite a relatively small sample size, the trial provided an excellent opportunity to validate the predictive role of miR-31-3p in a prospective cohort and allowed to test dynamic changes in miR-31-3p expression overtreatment. The results presented here are largely consistent with available literature (22–24, 27) and suggest that low miR-31-3p might be an indicator of response and better prognosis in patients treated with anti-EGFR mAbs.

Even though our data align well with available literature, several questions remain open. First, the biology underpinning a potential role for miR-31-3p in driving resistance to anti-EGFR agents is not clear. Preclinical in vitro and in vivo data in colon and lung cancer respectively suggest that miR-31-3p targets a number of negative regulators of the RAS-MAPK cascade (32, 33). However, despite the link between miR-31-3p overexpression and RAS signaling pathway activation appears solid, no experimental evidence has, as yet, confirmed whether these mechanisms are responsible for resistance to cetuximab.

A second question relates to the source of material and the technology to be used for miR-31-3p testing. Mir-31-3p is overexpressed in early stages of sporadic and inflammation-related colorectal cancer carcinogenesis but expression does not appear to change in more advanced or metastatic stages of disease (34–36). In keeping with these data, no significant changes in miR-31-3p scoring were observed in our trial when comparing primary cancers and metastatic sites. Similarly, in our series, no changes in miR-31-3p scoring were observed in sequential tissues biopsies collected before, during, and after cetuximab treatment. On the contrary, in the NEW-EPOC trial (22), a nonsignificant correlation for miR-31-3p expression was observed between paired primary colorectal cancer specimens and liver metastases in patients receiving preoperative cetuximab, whereas a positive and significant correlation was observed in patients treated with chemotherapy alone, suggesting that cetuximab treatment might induce changes in miR-31-3p expression. One of the potential explanations for the discrepancy between the NEW-EPOC (22) and the PROSPECT-C trial (5) is that, in the former trial, miR-31-3p levels were tested by ISH whereas in the latter, the analysis was performed by RT-PCR. Cetuximab is known to trigger intratumor inflammatory infiltration in liver metastases, with an enrichment of CD3⁻, CD8⁻, and CD56⁺ cells (37). Given miR-31-3p is also involved in immune and inflammatory cells homeostasis (38–41), its overexpression might sometimes be due to intratumor infiltration from lymphocites or inflammatory cells. Under these circumastances, despite tissue microdissection, contamination by inflammatory cells might potentially lead to a bias in miR-31-3p scoring when using high sensitivity RT-PCR-based assays. In line with this hypothesis, our ISH did detect areas of miR-31-3p overexpression in the stromal compartment of tumors otherwise scored as miR-31-3p low. Furthermore, even though the comparison between ISH-based and RT-PCR-based miR-31-3p scoring in our cohort showed a good concordance, several cases were classified in different miR-31-3p expression categories by the two assays, thus highlighting some hurdles in selecting the best approach for evaluating miR-31-3p expression as a biomarker for anti-EGFR mAbs. Given RT-PCR-based assays have been recently validated for miR-31-3p clinical testing (23, 27), caution in the analysis and interpretation of data should be exerted in cases with intense inflammatory and immune infiltrate as these might affect miR-31-3p classification.

Selection of mCRC patients' candidate to anti-EGFR treatment relies on primary tumor location (sidedness; ref. 30) and RAS testing (3). As we and others have suggested (4, 5), moving RAS testing to plasma cfDNA might represent a more sensitive and cost/effective option than tissue analysis. In our study, we combined RAS genotyping in cfDNA with miR-31-3p expression in order to test whether this would result in a more accurate prediction of response to cetuximab. The test for interaction between the two categorical variables was not significant possibly due to the very small sample size; however, in patients with no cfDNA RAS abnormalities, ORR, PFS, and OS appeared better for patients with miR-31-3p low tumors. Although larger studies will need to confirm these findings, a key question remains open: do we need another test to select mCRC patients for cetuximab treatment, or are we at risk of ultra-selecting patients? Our data, in line with the analyses of FIRE-3 (26) trial, suggest that miR-31-3p expression may be an indicator of depth of response to anti-EGFR inhibition; this, in our opinion, might represent the ideal scenario where a more accurate identification of patients likely to achieve resectability and/or symptom control may justify a more thorough selection of patients (Fig. 5).

In conclusion, our results confirm the potential predictive role of miR-31-3p for the selection of patients undergoing anti-EGFR treatment. Further studies are needed to test if miR-31-3p might be combined with RAS testing in cfDNA to further identify best responders in specific clinical niches.

D. Cunningham reports receiving commercial research support from Amgen, Sanofi, Merrimack, AstraZeneca, Celgene, MedImmune, Bayer, Emerge, Clovis, Eli Lilly, and Janssen. K.H. Khan is a consultant/advisory board member for Bayer Oncology Group. S. Rao is a consultant/advisory board member for Roche and Celgene. N. Starling reports receiving speakers bureau honoraria from AstraZeneca, Eli Lilly, and Merck, and reports receiving commercial research grants from AstraZeneca, Bristol-Myers Squibb, and Merck. C. Braconi reports receiving speakers bureau honoraria from Merck Serono, Bayer, and Eli Lilly, and did a consultation with Lumien. I. Chau is a consultant/advisory board member for Eli-Lilly, Bristol-Myers Squibb, MSD, Bayer, Roche, Merck Serono, Five Prime Therapeutics, AstraZeneca, and Oncologie International, and reports receiving commercial research grants from Eli-Lilly, Janssen-Cilag, and Sanofi Oncology. N. Valeri reports receiving speakers bureau honoraria from Bayer, Pfizer, Eli Lilly, and Merck-Serono. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Anandappa, D. Cunningham, K.H. Khan, C. Braconi, C. Peckitt, N. Valeri

Development of methodology: G. Anandappa, A. Lampis, D. Cunningham, K.H. Khan, N. Starling, M. Darvish-Damavandi, M. Fassan, N. Valeri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Anandappa, A. Lampis, D. Cunningham, K.H. Khan, S. Rao, D. Watkins, H. Lote, J. Thomas, N. Khan, N. Fotiadis, R. Begum, I. Rana, J.C. Hahne, I. Chau, N. Valeri

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Anandappa, D. Cunningham, K.H. Khan, K. Kouvelakis, G. Vlachogiannis, N. Tunariu, N. Starling, R. Kalaitzaki, J.C. Hahne, I. Chau, M. Fassan, N. Valeri

Writing, review, and/or revision of the manuscript: G. Anandappa, A. Lampis, D. Cunningham, K.H. Khan, K. Kouvelakis, N. Tunariu, S. Rao, D. Watkins, N. Starling, C. Braconi, M. Darvish-Damavandi, C. Peckitt, R. Kalaitzaki, J.C. Hahne, I. Chau, M. Fassan, N. Valeri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.H. Khan, S. Hedayat, D. Watkins, N. Khan, R. Begum, I. Rana, A. Bryant, N. Valeri

Study supervision: D. Cunningham, K.H. Khan, D. Watkins, M. Rugge, N. Valeri

This work was supported by Cancer Research UK (grant number CEA A18052), the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) at The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research (grant numbers A62, A100, A101, A159), and the European Union FP7 (grant number CIG 334261) to N. Valeri. The authors acknowledge support from the National Institute for Health Research Biomedical Research Centre at The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.