Is the tumour microenvironment a critical prognostic factor in early-stage colorectal cancer?

The TNM staging system remains the cornerstone of risk assessment in patients with early-stage colorectal cancer (CRC). However, clinical behaviour is diverse within the same stages, making prognostication an imprecise science. Microsatellite instability (MSI) is the only biomarker routinely considered beyond TNM, although a range of major genomic changes is well established, with contradictory evidence in outcome prediction [1, 2]. So what other markers could improve prognostic precision?

CRC heterogeneity has now been comprehensively characterised at the transcriptomic level as between three and six prognostic and potentially predictive subtypes [3–8]. For clinical application, these competing subtypes were integrated into four consensus molecular subtypes (CMS1–4) by the ColoRectal Cancer Subtyping Consortium [9]. The CMS classification has shown prognostic significance in both the early and advanced settings in multiple cohorts from high-quality clinical trials testing contemporary regimens [10–13]. While CMS2 and CMS3 primarily represent epithelial cancer cell heterogeneity, CMS1 and CMS4 also include cellular components of the tumour microenvironment, mainly immune cells and stromal fibroblasts [9, 10, 14].

The immune microenvironment is known to contribute to CRC patient prognosis [15]. Among tumour microenvironment cells, immune cell infiltrates, specifically cytotoxic T lymphocytes (CytoLym, enriched in CMS1), are associated with better outcomes in early-stage CRC patients [15]. The CMS4 subtype is enriched for cancer-associated fibroblasts (CAFs) and has a poor prognosis [9, 16]. Is it possible that subtype-specific prognosis is driven by the microenvironment?

To address this question and following up on their previous study [1], Dienstmann et al. explored whether tumour microenvironment features such as CytoLym and CAFs are stronger determinants of disease-free survival (DFS) in early-stage CRC patients than the known genomic aberrations (MSI, BRAF and KRAS mutational status) and CMS subtypes [17]. Their large, retrospective study of several public and private clinically annotated stage II/III CRCs (n = 2, 636), both untreated (n = 1, 656) and treated (n = 980), evaluated CMS1/4 subtype scores and microenvironment-based (CytoLym and CAF) continuous scores in silico. In a multivariable model, clinicopathological (including TNM) and microenvironment features were independent prognostic factors. Clinicopathological variables explained the majority (56%–77%) of variation in DFS, followed by immune stromal infiltrating markers (14%–35%), while only <6% of DFS variation was explained by CMS and genomic factors. Moreover, CAF scores were associated with poor prognosis exclusively in stage III cancers while CytoLym scores were associated with good prognosis specifically in stage II and microsatellite stable cancers. Hence, CMS4 and MSI subtypes (enriched for CAFs and CytoLym, respectively [14, 16]) were not prognostic when these specific microenvironment features were included [17].

This study adds to the evidence that the tumour microenvironment plays a crucial prognostic role in CRC. The authors must be appreciated for their efforts in collecting over 2600 patient samples from public and private (including clinical trial) datasets. Their findings are consistent with previous studies, in particular with the immunohistochemistry assay Immunoscore^®, which has been extensively validated in early-stage CRCs [18].

Overall, this represents an excellent summary of the relative contribution of clinicopathological and molecular features in explaining DFS. Validation of these current results in additional, high-quality datasets is recommended, including those in oxaliplatin-treated stage III cancers. However, the study has limitations. Genomic data (MSI and BRAF mutation) were missing for a large number of samples, which were instead imputed computationally, although the prevalence of these aberrations was similar to those previously reported [2]. Conversely, all included samples had CytoLym and CAF scores available. This inconsistent availability of data may have biased the statistical analyses. Hence, the results of these exploratory analyses in this study of multiple retrospective cohorts need to be interpreted appropriately.

Age is usually considered a clinical factor when balancing chemotherapy benefit against side-effects [2]. As highlighted by the authors [17], age accounted for a significant proportion of DFS, especially in the untreated (and older) population, who more often experience non-cancer-related deaths. Hence, it is worth considering only CRC-related relapses for DFS.

Sidedness is a known surrogate biomarker of complex CRC biology: while the poor prognosis and reduced response to anti-epidermal growth factor receptor therapy in the right-sided metastatic setting is increasingly recognised, the prognostic role of sidedness in early-stage CRC is less clear [19]. The current study shows that right-sided early-stage (II and III) CRCs have a better DFS than left-sided CRCs irrespective of MSI status (using partially imputed data) [17], similar to previously published results [20]. Nevertheless, this observation requires further validation using well-annotated data to assess whether these observations are generalisable to early CRCs in the real clinical setting.

The MicroCells approach was limited to detecting only a few immune cell types and did not include, for example, T regulatory cells and specific macrophage subsets, which may be important in governing pro- and anti-inflammatory responses in tumours and, therefore, prognosis. Furthermore, the scores for the microenvironment cell types may vary depending on the cell-type markers and computational methods used. It is worth remembering that computational methods require rigorous validation using established immunohistochemistry or similar experimental methods before clinical application. Nevertheless, the association between CytoLym and MSI status suggests that this study included the most relevant microenvironment scores and important immune populations.

This interesting study now prompts the question of what other factors beyond the microenvironment may affect prognosis in patients with early-stage CRCs. Epigenetics, colonic crypt cell types (similar to previously reported [3]), tumour mutational burden, and neoantigens are promising candidates. Whatever the answer, robust and clinically relevant biomarkers and assays are mandatory for effective clinical translation, which may need to be developed in the future based on the current study. Nevertheless, with further validation, these findings will hopefully facilitate our understanding of the relative contributions of cancer cells and the microenvironment in determining prognosis in early-stage CRC patients and refine personalised medicine approaches in the future.

Funding

None declared.

Disclosure

AS has ownership interest as an inventor for a patent entitled ‘Colorectal cancer classification with differential prognosis and personalised therapeutic responses’ (patent number PCT/IB2013/060416). AS received research funding from Bristol-Myers Squibb, Merck KGaA and Pierre Fabre. All remaining authors have declared no conflicts of interest.