Intratumour heterogeneity, from hypothesis to the clinic

Abstract

Recent technological advances uncovered intricate biological processes underlying intratumor heterogeneity with clinical implications. These insights led to novel biomarkers for immunotherapies, justified serial tumour biopsies for therapeutic target profiling, inspired new treatment strategies, and ultimately might yield novel therapeutics that target clonal interdependence.

Advances in DNA sequencing technologies, including single-cell sequencing, and in vitro techniques to barcode and lineage-label individual cells have dramatically changed our ability to study clonal heterogeneity within cancers. Since 2012, when the Journal last published a mini-review on cancer heterogeneity [1], In total, 17,500 new articles were published, including “intratumor heterogeneity” in their title or abstract (Google Scholar accessed; August 30, 2022). These studies have shed light on how intratumor heterogeneity influences the tumour immune microenvironment, revealed unexpected functional plasticity in tumour cells, and hinted at clonal cooperation in addition to clonal selection under therapeutic pressure. We also learned that various metrics of heterogeneity predict prognosis and response to chemo- and immunotherapies, and the tumour mutational landscape can change during the disease progression.

Intratumor heterogeneity, tumour microenvironment, prognosis and response to therapy

Intratumor genomic heterogeneity can be quantified through statistical metrics calculated from bulk RNA and DNA sequencing (e.g., RNA expression dispersion distance, distribution of minor allele frequencies, pairwise Hemming distances of DNA sequences, tumour mutation burden [TMB], and clonality estimates), or it can be more directly measured through multi-region tumour sampling or single-cell sequencing [2]. Heterogeneity also exists in epigenetic and metabolic features and extends to the tumour microenvironment. Different heterogeneity metrics capture distinct biological features that differ across cancer types. Melanoma and lung cancer are characterised by a high overall clonal TMB (i.e., similar TMB in the majority of subclones within cancer), while temozolomide-treated low-grade glioblastoma, breast, bladder, and prostate cancers have overall lower TMB with large subclonal variation (i.e., different subpopulations have different TMB) [3]. Tumour heterogeneity metrics correlate with immune cell infiltration, prognosis, and treatment response. In primary triple-negative breast cancer (TNBC), lower clonal heterogeneity and lower neoantigen load are associated with higher immune cell infiltration suggesting an immune pruning effect (i.e., immune editing), and these features also predict for better prognosis and greater chemotherapy sensitivity [4]. In hepatocellular carcinoma, non-small cell lung (NSCLC), acute myeloid leukaemia, and several other cancer types, greater genomic heterogeneity and clonal diversity are associated with worse prognosis and resistance to chemotherapy, which are attributed to greater probability of these tumours spontaneously harbouring drug-resistant clones [5]. On the other hand, high TMB leading to high neoantigen load is generally predictive of response to immunotherapies [6].

Clonal evolution during the course of the disease

Numerous computational biology tools exist that use point mutations and copy number alterations to deduce the evolutionary relationship between cell populations within a tumour and between primary tumours and metastatic lesions [7]. Mutations harboured by every cancer cell are considered clonal, while mutations present only in a subset of cells are called subclonal. The distribution of somatic driver mutations into clonal and subclonal categories varies by cancer types. In breast cancer, PIK3CA mutations are more often clonal, whereas in melanoma, NSCLC, and renal cancers, PIK3CA mutations are often subclonal. Mutations in TP53 are predominantly clonal in NSCLC, TNBC, oesophageal, and ovarian cancers, but are mostly subclonal in clear cell renal cancer and chronic lymphocytic leukaemia. Targeting clonal mutations may be therapeutically more successful than targeting subclonal mutations. During disease evolution, particularly under therapeutic pressure from targeted therapies, the genomic composition of cancers changes. Constitutively active ESR1 mutations can arise during anti-oestrogen therapy for breast cancer. However, emerging data suggest a more complex process than simple clonal dominance [3, 8]. Studies in melanomas treated with RAF inhibition, in colorectal cancer treated with anti-EGFR therapy, in PIK3CA-mutant breast cancer treated with PIK3CA inhibitor, and in BRCA-mutant cancers treated with PARP inhibitors documented simultaneous emergence of distinct clones harbouring different mutations that each confer drug resistance, leading to polyclonal resistance phenotype [9–11].

Analysis of the genomic architecture of primary tumours and metastatic lesions also revealed multiple different paths to metastatic dissemination that co-occur in the same patient. The traditional view of linear evolution with the primary tumour giving rise to all metastatic lesions is rarely the dominant trajectory. In many instances, the metastatic lesions and the primary tumour both arise from a common ancestor and evolve in parallel, while in other instances metastatic lesions give rise to new metastases [12]. Another important observation is that synchronous distant metastases (i.e., de novo Stage IV disease) share greater genomic similarity with the primary tumour than asynchronous metastases (i.e., recurrent metastatic disease) [8].

Functional plasticity and clonal cooperation

Single-cell labelling and genomic barcoding techniques revealed large-scale transcriptional heterogeneity in clonal cell populations indicating that multiple cell states co-exist [13]. The proportion of cells in various cell states changes during stress or exposure to therapy but often returns to baseline [14]. Cell lineage labelling in animal models also demonstrated multiclonality of metastatic lesions suggesting clonal cooperation that facilitates metastatic spread [15, 16]. Cell-cell contact and paracrine secretion of cytokines (IL-6, IL-11, TGFβ, amphiregulin) have been suggested as mediators of these interdependencies [17]. These observations indicate that dynamic epigenetic and stochastic gene expression transitions can be as important to provide population-level fitness as genomic alterations and clonal selection.

Clinical impact

Improved methods to quantify tumour heterogeneity and understanding its relationship to outcome have led to new diagnostic tests (e.g., TMB assessment is approved by the US FDA as a test to select patients for pembrolizumab therapy regardless of histologic type). Recognition that the cancer genome can change during the course of the disease led to the broader use of repeat tumour biopsies to detect therapeutically actionable mutations. Detecting treatment-emergent alterations such as ESR1 mutation in breast cancer will likely play an increasingly important role to select subsequent therapies. The genomic similarity of primary tumours and synchronous distant metastases raise the tantalising possibility that some de novo Stage IV cancers might also be cured with similar combined modality treatment strategies as the ones used as adjuvant therapies that have been highly successful to improve outcomes in early-stage disease (i.e., HER2 amplified breast cancer). In the longer term, a better understanding of the biological processes that sustain tumour heterogeneity and enable clonal cooperation could reveal entirely new therapeutic vulnerabilities and lead to the development of new therapies.

Author contributions

Contributing authors for manuscript preparation: NLS, AK and LP. All authors reviewed the manuscript and approved the final version of the manuscript.

Funding

None.

Data availability

Not applicable.

Competing interests

LP has received consulting fees and honoraria from Pfizer, AstraZeneca, Merck, Novartis, Bristol-Myers Squibb, GlaxoSmithKline, Genentech, Personalis, Daiichi, Natera, Exact Sciences and institutional research funding from Seagen, GlaxoSmithKline, AstraZeneca, Merck, Pfizer and Bristol-Myers Squibb.

Ethics approval and consent to participate

Not applicable.

Consent to publish

Not applicable.