Testing for defective deficiency in DNA mismatch repair (dMMR) (or its surrogate, which is the presence of microsatellite instability, MSI) is now part of the routine diagnostic workup for patients with colorectal cancer (CRC) [1]. In fact, MSI/MMR testing is recommended in all CRC cases for Lynch syndrome screening [2];. MSI/MMR in stage II CRC identifies patients with a lower risk of recurrence and better overall survival (OS) and for whom adjuvant fluoropyrimidine monotherapy may have a questionable benefit [3, 4]. In stage IV patients, MSI/MMR is used to select candidates for immunotherapy with immune checkpoint inhibitors (ICIs) [5, 6].

Cancer is heterogeneous in nature, and this may significantly impact the personalization of patient care [7–9]. MSI/MMR status heterogeneity was recently described in gastric cancer and was associated with a lack of response to pembrolizumab [10]. In CRC, dMMR is considered an early event in the carcinogenetic process, and its heterogeneity has always been considered rather unlikely [11]. In fact, only exceptional reports of heterogeneity in the MMR pathway have been described in CRC [12–14].

Here, we describe a unique case of a metastatic CRC (mCRC) patient showing a heterogeneous MSI/MMR pattern, who was treated with ICIs and underwent an extensive molecular characterization by means of whole genome sequencing (WGS) and whole transcriptome RNA sequencing (RNA-Seq) of two MMR-discordant areas of the tumour DNA. We further discuss the relevance of MMR/MSI heterogeneity in a real world setting by re-evaluating data on MMR obtained at our centre in the last 3-year CRC series.

In December 2013, a 64-year-old man was hospitalized because of worsening asthenia and abdominal pain. Past medical history included a surgically treated pT1 pN0 cM0 left clear cell renal carcinoma in 2005 and a recto-sigmoid resection for colonic low-grade tubular adenoma in 2008. Family history was positive for brain and haematologic tumours not otherwise specified in 2nd degree relatives.

During admission as an inpatient to a Unit of General Medicine of a community hospital, blood tests revealed grade 2 microcytic anaemia, and abdominal radiography showed signs of sub-occlusion. Colonoscopy revealed a right-sided neoplastic lesion. A biopsy documented an adenocarcinoma with a mucinous component. Pre-operative staging total body computed tomography (CT) including the neck, chest, abdomen and pelvis showed thickening of the right bowel wall and diffuse peritoneal nodules with maximum diameter of up to 6 cm.

On December 2013, the patient underwent a palliative right hemicolectomy and diagnostic resection of a peritoneal nodule, the latter only with a diagnostic and confirmatory intent. Figure 1 summarizes the complete clinical course over time. Gross pathology examination described a 9 cm mass completely obstructing the colonic lumen. Pathological reports confirmed the diagnosis of an adenocarcinoma with the presence of heterogeneous phenotypic areas of mucinous (40% of the neoplastic area) and signet ring differentiation (pT4 pN2b [7 metastatic lymph nodes out of 10 examined], pM1c; stage IVC). Routine molecular testing showed a G12D KRAS mutation, whereas NRAS, BRAF, and PIK3CA genes showed a wild-type status (Myriapod Colon status kit; Diatech Pharmacogenetics, Jesi, Italy).Fig1Fig. 1

Clinical course over time, including treatments, diagnostic procedures and timing of disease progression

Despite the recent extensive description and characterization of the molecular heterogeneity of CRC, in everyday practice, it is considered by treating physicians as a rather homogenous disease. Reasons for this reside in a general consistency (between different areas, different metastatic lesions and over time) in the status of markers commonly tested for therapeutic purposes, such as RAS or BRAF mutations. Data on intra-tumour heterogeneity for new emerging biomarkers with therapeutic implications, such as dMMR, are limited.

Here, we describe the case of a mCRC with heterogeneous MMR/MSI status in adjacent tumour areas. The most important novel achievement was the long-lasting response with ICI followed by acquired resistance. The present case offers a unique opportunity to discuss challenges and implications for the diagnostic approach and therapeutic management of this special subgroup of CRCs.

How frequent is heterogeneity in MMR/MSI status? To properly address this question, we reviewed our archived CRC samples over the last 3 years. A total of 1855 samples were tested for MMR protein expression by means of immunohistochemistry, and 201 (10.8%; median age 76 years, range 19–91; F/M = 0.93) showed a dMMR phenotype and 1654 a pMMR status (89.2%; median age 71 years, range 33–97; F/M = 0.62). Among the dMMR series, 13 cases showed peculiar patterns of MMR alterations (0.7%; median age 70 years, range 38–85; F/M = 0.44) (Fig. 4):

Eight dMMR/MSI-high cases were characterized by the complete loss of the four MMR proteins in a component of the tumour and presented a second component retaining MHL1/PMS2 (n = 6) or MSH2/MSH6 (n = 2) (Fig. 4a). This finding has already been described in the context of a Lynch syndrome background [1, 17].

These results go beyond the aim of the present report, focused on a specific exceptional clinical case, but they show how heterogeneity - in all its forms - is a rare event in CRC. Nevertheless, it is obvious how in the era of personalized medicine, rarity should not affect the relevance for an individual patient, and the practical consequences are discussed below.

Which could be the main clinical implications?

A) Importance of sampling. Approximately 40% of patients with mCRC may have only a small bioptic sample available for pathological analyses at the time of initial clinical decision making [18]. Divergence between primary tumour and the biopsied tissue could be a hurdle to targeted therapies. Technical issues are extensively discussed in the literature [19]. There is a general agreement that biopsy samples could be a reliable alternative to primary tumours for RAS and BRAF mutational profiling [20]. Instead, regarding MSI status, poor data are available on sampling issues.

B) Therapeutic choice. Predictive markers of response to ICI in MSI-H/dMMR CRC are lacking and nor PD-L1 expression on tumoural cells, abudance of PD-L1 expressing tumour-associated immune cells, BRAF mutation status or Lynch Syndrome were predictive of benefit in the largest clinical study presented so far [21]. Despite MSI/MMR heterogeneity in the primary tumour, our patient responded to ICI treatment. Previously, Kim et al. performed a detailed molecular characterization of 61 patients with metastatic gastric cancer, 7 (11.5%) of whom had dMMR, to explore the determinant of response to pembrolizumab. Only 1 out of 7 cases with dMMR showed a lack of response and rapid progression; that tumour sample was characterized by marked geographic heterogeneity of MLH1 protein on immunohistochemical staining [10].

C) Progression to treatment. Given that mechanisms of clonal selection have already been described as escape strategies for different cancers to different targeted treatments, we initially hypothesized that MSI heterogeneity (and therefore the expansion of MSS sub-clones) could have been the major driver of acquired resistance. Notwithstanding, contrary to our expectations, molecular testing on re-biopsy at the time of progression documented a homogenous dMMR pattern. Unfortunately, the very limited tumoural content of the small biopsy performed at progression did not allow additional analyses.

How does heterogeneity affect emerging markers of ICI efficacy?

TMB is a validated biomarker of ICI response in metastatic melanoma, NSCLC and urothelial bladder cancer [22, 23]. In mCRC, TMB is correlated with MSI status [24] and recent data suggest a role as an independent biomarker of ICI efficacy [25]. WGS and RNA-seq analysis showed consistent findings in our case: MSI areas had high TMB, whereas MSS areas had low TMB.

Similar results were found for tumour infiltrating lymphocytes (TILs). Samples were defined as having high levels of TILs when ≥2.0 per high power field (HPF, 40x) or as having low levels of TILs when < 2.0 [26]. In all analysed specimens, we found high levels of TILs in MSI areas (consistently higher than 3.0) and no TILs in MSS areas (Additional file 1: Figure S1). Similarly, a high number of TILs was observed in the post-treatment MSI-high biopsy sample (i.e., 4.2). Galon et al. previously demonstrated this correlation between MSI-high status and immune infiltration of the tumour [27]. The number of TILs as predictive markers of response to ICI is currently under investigation by our group.

In addition to the main topics discussed above, we also report the relatively uncommon response to regorafenib. In the CORRECT trial, only 5 patients out of 500 treated with regorafenib achieved a partial response (ORR 1.0%) [28]. Therefore, the dimensional decrease and the clear necrotic effects of targeted lesions obtained by our patient were somewhat surprising. Nevertheless, it is interesting to note that expression results are in line with what was previously reported by Teufel et al. [29] regarding a greater efficacy of regorafenib in patients assigned to the consensus molecular subgroup (CMS) 2 (canonical). This evidence is preliminary, and there are no clinical implications at present.

Taken together, the information derived from the present case underlines the importance of critical and rigorous observation of patients in clinical practice, which can be crucial for collecting important data complementary to those derived from large prospective clinical trials.