Colorectal cancer (CRC) is one of the leading causes of cancer-related death worldwide and the third most common cancer in Korea [1]. Standard palliative treatment for metastatic or unresectable CRC (mCRC) is fluorouracil-based combination chemotherapy (with oxaliplatin or irinotecan), with or without agents targeting angiogenesis (bevacizumab) or epidermal growth factor receptor (cetuximab). The available therapeutic options for later-line chemotherapy are limited; regorafenib and TAS-102 showed only a modest clinical benefit in these patients. The objective response rate (ORR) with regorafenib and TAS-102 was approximately 1%, and median progression-free survival (PFS) was around 2 months for both treatments [2,3]. The long-term outcomes of mCRC are still poor [1], and novel therapeutic approaches are needed.

Growing evidence suggests that patients with mCRC harboring deficient mismatch repair protein (dMMR)/microsatellite instability–high (MSI-H) can obtain clinical benefit from immune checkpoint inhibitors (ICIs) [4-7]. Pembrolizumab and nivolumab, which are anti–programmed death 1 (anti–PD-1) inhibitors, improved ORR and PFS in selected patients with dMMR/MSI-H mCRC [4-7]. Failure to repair DNA replication-associated errors in dMMR/MSI-H mCRC is associated with high mutation loads, tumor neoantigen loads, and dense immune cell infiltration [8]. In fact, the whole-exome sequences revealed higher somatic mutation loads (1,782 mutations per tumor) in patients with dMMR/MSI-H than in patients with proficient MMR (73 mutations per tumor) (p=0.007), and a greater density of CD8-positive lymphocytes and a higher expression of PD-ligand 1 (PD-L1) were observed in patients with dMMR/MSI-H than in patients with proficient MMR [4]. However, the clinical benefit of ICIs is confined to a small proportion of patients, because dMMR/MSI-H is identified in only about 5% in patients with mCRC [9]. This raises the need to expand the number of potential candidates for immunotherapy.

The POLE gene is located in 12q24.33 and encodes the proofreading (exonuclease) subunit of polymerase epsilon (POLE) with 2,286 amino acids [10]. This POLE mutation has been reported in approximately 3% of proficient MMR CRC and represents high somatic mutation loads [10]. According to the Cancer Genome Atlas, up to one-quarter of hypermutated CRC carry somatic mutations in POLE [10]. Because high mutation loads are considered a mechanism of the response of dMMR/MSI-H to ICIs, POLE-mutated cancer may also be susceptible to ICIs. However, to date, clinical data regarding the response to ICIs in POLE-mutated cancer are lacking.

Treatment with avelumab, an anti–PD-L1 inhibitor, achieved an ORR of 33% in patients with previously treated metastatic Merkel cell carcinoma and was approved for the treatment of metastatic Merkel cell carcinoma in early 2017 [11]. Subsequently, avelumab has shown promising antitumor activity in various solid tumors, such as genitourinary tract [12,13], gynecologic [14], and lung [15], but its activity in mCRC harboring dMMR/MSI-H has not been investigated. This study aimed to evaluate the efficacy and safety of avelumab in patients with previously treated CRC with deficient MMR/MSI-H as well as those with POLE mutations.

In this open-label, multicenter, phase II study, avelumab showed promising antitumor activity and manageable toxicity in patients with mCRC harboring dMMR/MSI-H or POLE mutation. The ORR and median PFS were 24.2% and 3.9 months, respectively, and six of eight responders were continuing avelumab treatment with durable response at the end of the analysis. Of note, the ORR and median PFS were 28.6% and 8.1 months, respectively, in patients with MSI-H by PCR or NGS, which were thought to be more reliable methods of determining the MSI status than IHC. Although no patients with POLE mutations had response to avelumab, the limitations of small sample size and variation in mutation sites need to be taken into account. TRAEs of any grade and TRAEs of grade 3 or 4 were observed in 72.7% and 18.2% of patients, respectively, which was consistent with previous studies [17]. There were two discontinuations of treatment, one because of a TRAE and the other because of a serious adverse event, probably related to treatment. There were no treatment-related deaths.

The efficacy of avelumab (an anti–PD-L1 inhibitor) for mCRC with dMMR/MSI-H, specifically in patients with MSI-H by PCR or NGS, is comparable to that of pembrolizumab and nivolumab (anti–PD-1 inhibitors) in this setting. In the KEYNOTE-016 trial, only patients with MSI-H CRC had the objective response to pembrolizumab (ORR 40%), whereas none of those with MSS CRC had the objective response [4]. A subsequent multicenter trial of pembrolizumab for dMMR/MSI-H CRC, the KEYNOTE-164 phase II study, resulted in an ORR of 33% and a median PFS of 4.1 months; the 12-month PFS and OS rates were 34% to 41% and 72% to 76%, respectively, according to prior line of treatment [6,7]. Nivolumab also showed promising antitumor activity in terms of ORR (31.1%) in dMMR/MSI-H CRC, and the 12-month PFS and OS rates were 50% and 73%, respectively [5]. All these favorable results in dMMR/MSI-H CRC contrast sharply with those of later-line conventional chemotherapy for overall mCRC with treatments such as regorafenib or TAS-102, which resulted in ORR of only 1% and median PFS of around 2 months [2,3].

Several phase I results from the JAVELIN Solid Tumor Trials have shown promising ORRs and disease stabilization with avelumab in various types of advanced tumors. Specifically, among 53 patients with metastatic or locally advanced previously treated solid tumors, four (8%) achieved responses and 30 (57%) had SD [18]. The ORR with avelumab ranged from 6.7% to 18.2%, depending on tumor type such as metastatic or unresectable previously treated renal cell carcinoma [12], urothelial carcinoma [13], non-small cell lung cancer [15], and ovarian cancer [14]. In a phase II study of 88 patients with chemotherapy-refractory metastatic Merkel cell carcinoma [11], the ORR with avelumab was 33%, including a CR rate of 11.4%. To date, avelumab has been approved for treatment of previously treated metastatic Merkel cell carcinoma, urothelial carcinoma, and renal cell carcinoma in combination with axitinib. Several explorations to expand its use in various clinical settings are ongoing. The present study adds to the evidence of clinical activity of avelumab by showing durable objective responses in mCRC with dMMR/MSI-H.

dMMR/MSI-H is an established biomarker for the efficacy of ICIs in mCRC, and its predictive value has also been confirmed in other various tumors. MSI-H metastatic gastric cancer (mGC) had a higher ORR with pembrolizumab than did MSS mGC (57% vs. 9%) in the KEYNOTE-059 study [19]. Recently, the clinical benefit of pembrolizumab was demonstrated among patients with dMMR/MSI-H non-CRC, including endometrial cancer, gastric cancer, cholangiocarcinoma, and pancreatic cancer in terms of ORR (34.3%) and median PFS (4.1 months) [20]. Although the exact mechanism is unknown, there are several proposed explanations for T-cell checkpoint blockade [21]. Since dMMR/MSI-H results in diverse neoantigens, T-cell epitopes that are newly formed as a consequence of tumor-specific mutations, which can increase neoantigen-driven T-cell response. Another possible explanation is that dMMR/MSI-H is associated with the activation of signaling pathways through altered cytokines or chemokines, resulting in the tumor microenvironment becoming more inflamed. Cellular stress induced by dMMR/MSI-H promotes innate immune cells, such as T cells and natural killer cells, or tumor recognition.

However, not all patients with mCRC harboring dMMR/MSI-H respond to ICIs, and the TMB varies even within dMMR/MSI-H mCRC. The updated analysis from the initial single-center study showed an ORR of 50% [4], but subsequent multicenter phase II trials revealed ORR of around 30% [5-7]. In the present study, the ORR was 28.6% in patients with MSI-H by PCR or NGS. Approximately 37 to 41 mutations per megabase may be a cutoff value, and low TMB was significantly associated with poor response to ICIs and worse PFS within the dMMR/MSI-H population, which may be an explanation for the heterogeneity in response in recent clinical trials of dMMR/MSI-H CRC [22]. Moreover, substantial genomic variation is observed within dMMR/MSI-H tumors. In particular, the genome-wide intensity of MSI and the accumulation of insertion–deletion mutational loads are responsible for a wide diversity of clinical benefits with ICIs [23], and the activated WNT/β-catenin signaling pathway is associated with immune escape, despite a high TMB and high numbers of tumor-infiltrating lymphocytes [24].

In fact, a considerable portion of the primary resistance of mCRC treated with ICIs may be due to misdiagnosis of dMMR/MSI-H. One study showed that three of five patients (60%) who had PD at the first evaluation were reassessed as MSS by central review with PCR, contrary to the diagnosis of dMMR/MSI-H by local assessment [25]. In the Check-Mate-142 study, there were discrepancies between local and central assessments in 14 of 74 patients (19%) [5]. Most of the patients with assessments (10/14) were initially determined as dMMR by IHC at the local laboratory, but central review with PCR reclassified them as MSS [5]. In this study, there were some discrepancies between IHC and PCR results, and six of nine patients with dMMR by IHC only showed MSS by PCR or NGS, leading to different ORRs depending on how dMMR/MSI-H CRC was defined. After review of the available IHC results in eight of these patients, five patients were revised to proficient MMR, while their PCR or NGS showed all MSS. The reliability and reproducibility of IHC results have always been a concern, because IHC results are largely affected by intra- and interobserver variation and tissue preservation status [26], and local assessment without central review contributed to these results. Likewise, in the recent phase II study with a similar design, differences in ORRs with avelumab in patients with previously treated endometrial cancer were observed between dMMR determined by IHC only and MSI-H determined by NGS (26.7% vs. 30%) [27]. dMMR/MSI-H could be misdiagnosed if IHC results are not supported by PCR or NGS, so the IHC test alone should be carefully interpreted by experienced pathologists. Further, IHC tests on tumor samples achieved after chemoradiotherapy, old samples, or poorly preserved samples are associated with a high risk of unreliable results.

POLE-mutated CRC has been characterized by young age, male predominance, right-sided CRC, earlier stage of disease, and excellent prognosis [10]. POLE has a crucial role in chromosomal DNA replication by its proofreading capacity and is known to be mutually exclusive with dMMR/MSI-H [10]. Because of high immunogenicity and enrichment of mutation-associated neoantigens, POLE-mutated cancer has been considered a candidate for treatment with ICIs. However, there are limited data, and there is only one case report [28] of a patient with mCRC harboring a POLE mutation (V411L) and MSS. This patient had a response after three cycles of pembrolizumab, and CD8 infiltrating lymphocytes with PD-1 expression were observed in the primary colon tumor. All responders in previous reports had mutations at the P286R or V411L sites, which are considered hotspots of POLE mutation [10]. Unfortunately, three patients with POLE mutations did not respond to ICIs in the present study. One patient with a P286C mutation was associated with a hypermutation phenotype by NGS, but the other two (with R559W and G1086S mutations) did not show high TMB, although they were tested with different panels (S1 Table), and the identified sites of POLE mutation might not have been hotspots, which could have led to negative results. Further clinical studies with larger sample sizes are necessary to evaluate the activity of ICIs and its association with sites in POLE-mutated CRC.

One of the responders initially showed dMMR by IHC and MSS by PCR at the time of enrollment. However, the IHC result was revised to proficient MMR in the post hoc review. Although NGS could not be performed due to an inadequate amount of tumor tissue, leaving the POLE mutation status of this patient unknown, he achieved CR and has continued avelumab treatment for approximately 18 months (Fig. 3). In this regard, there may be unknown factors to explain the mechanism of response to ICIs other than MMR/MSI status or POLE mutation, such as PD-L1 or PD-L2 amplification [29,30], although we could not explore the cause of responsiveness in this patient due to the lack of adequate tumor tissue.

This study has several limitations. The consistency of eligibility and response evaluation could not be fully ensured, because central adjudication of the dMMR/MSI status of tumor tissue and central review of the radiologic response were not performed. Tumor samples were not collected prospectively for research purposes, and therefore translational studies, such as investigation of TMB, tumor-infiltrating lymphocytes, PD-L1 expression, or transcriptome, have not yet been performed. Further post hoc studies are planned with available tissue samples to elucidate their association with the response to ICIs.

In conclusion, avelumab is a promising anti–PD-L1 inhibitor in patients with metastatic or unresectable CRC harboring dMMR/MSI-H or POLE mutations. For the determination of dMMR/MSI-H, the conventional IHC method alone appears to be insufficient to select patients who would benefit from immunotherapy. Further studies to identify accurate strategies to select optimal candidates for immunotherapy are needed.