First detailed analysis of the evolution of clear cell renal cell carcinoma (ccRCC)
Loss of 3p, seen in almost all ccRCCs, is commonly caused by chromothripsis, likely decades before a tumour is diagnosed. Initial clonal expansion after 3p loss may start from only a few hundred cells
The genetic paths of ccRCC evolution are highly constrained within seven evolutionary subtypes that can be correlated with clinical phenotypes and outcome
The hallmark genomic drivers of ccRCC metastasis are loss of 9p and 14q
Progression to metastasis requires genomic instability, but not driver mutation load
An estimated 300,000 people worldwide develop kidney cancer every year, with approximately half dying from the disease. The commonest histological sub-type is clear cell renal cell carcinoma (ccRCC), which has a distinctive genome: loss of chromosome 3p is the critical founder genetic event, found in >90% of patients. The deleted region always encompasses four tumour suppressor genes that are frequent targets for subsequent inactivating mutations on the other chromosomal copy: VHL (65%–80%), PBRM1 (40%), BAP1 (10%), and SETD2 (10%). The second most frequent genetic event is gain of chromosome 5q, seen in 65%–70% of patients. ccRCCs have a wide range of clinical behaviour, ranging from highly aggressive to indolent, with differing spatial and temporal patterns of metastasis. Due to this variability of outcome, there is a tendency to over-treatment of small renal lesions, which may never progress to cancer, and a complete lack of predictive biomarkers to manage the more advanced tumours with variable behaviour.
ccRCCs display considerable intra-tumour heterogeneity, which is a critical driver of cancer progression and treatment failure. Tumours, which can be quite large, often contain several geographically localised subclones, but chromosome 3p loss and, when present, VHL mutations or methylation, are always on the trunk of the phylogenetic tree, suggesting that they are key early events in cancer development. Beyond these founder events, the clinical diversity of the disease suggests that subsequently, there are many different paths to progression. Stratifying ccRCCs and understanding their evolution is vital for the success of early diagnosis strategies and personalised management, including risk stratification, decisions regarding surgery, and adjuvant therapies.
To address this need, the TRACERx (TRAcking Cancer Evolution through therapy (Rx)) Renal study was set up in 2014 in concert with the wider TRACERx project (see Case Study 15). In the study, due to complete in late 2023, tumours of ccRCC patients are tracked from diagnosis to post-mortem by multi-region biopsies of both the primary tumour and any metastases. Extensive multiregion sequencing permits inferences to be made about the clonal composition of tumours and the order of mutations. Statistically significant patterns of mutual exclusivity, co-occurrence and ordering are analysed to identify any constraints, and then used for evolutionary classification. From this, the distinguishing features of different evolutionary subtypes are used for prognostication, thus generating both fundamental insights and biomarker opportunity. These data can then be linked to the progression and eventual outcome of the disease.
In 2018, the TracerX Renal consortium published three articles comprising a study of the origin, evolution and routes to metastasis of ccRCCs. (Mitchell et al, 2018; Turajlic et al, 2018a, 2018b). The Swanton lab combined with a host of collaborators to produce a technical and analytical tour de force, providing a molecular dissection of ccRCC at a forensic level never previously achieved for any cancer. Samra Turajlic, the chief clinical investigator on the project, started her own Crick group in 2019, but at the time of the work was a Clinician Scientist with Swanton.
In the first paper (Mitchell et al, 2018), 95 biopsies from 33 patients were whole-genome sequenced, to provide a detailed picture of the early events leading to ccRCC. Looking for potential driver mutations in non-coding DNA, a new point mutation hotspot was identified close to a MYC/MAX-MAD1 repressor binding site in the 5’UTR of TERT. The mutations, which were found in both clonal and subclonal variants, correlated with longer tumour telomere lengths, and likely confer a survival advantage on host cells.
Analysis of the distinctive chromosomal translocations found in ccRCC demonstrated that 3p loss was present in all the tumours, sometimes as a lone event, but generally as a translocation resulting from chromothripsis — a single catastrophic genomic rearrangement. Chromothripsis could be timed as the earliest tumourigenic event in ccRCC, occurring within the first two decades of life, even though the cancer may not be diagnosed for another 30 to 50 years.
This earliest genetic alteration would be predicted to produce a small pool of potentially tumour-initiating cells, which potentially exist in all adults. In individuals carrying a faulty VHL gene, progression to cancer would be all but inevitable within the human lifespan because both copies of VHL are inactivated, but for sporadic cancers, the necessary additional genetic alterations to develop full-blown ccRCC clearly take many years, if they happen at all.
Turajlic and co-workers (Turajlic et al, 2018a) next generated a comprehensive picture of the genetic and clonal evolution of ccRCCs, by multi-regional panel sequencing of 1206 regions within 101 primary ccRCC tumours. They were able to show that there were up to 30 driver events per tumour, challenging the assumption that tumours rely on no more than 5-10 drivers. Crucially, the data showed that the trajectory of a tumour is constrained, with every step narrowing the subsequent path it can take. It is this repeatability that gives hope that prediction of the “next move” may be possible.
As we start to consider the conditions under which different evolutionary paths are selected, we are leveraging the Crick’s mathematical modelling, AI expertise and sophisticated platforms for the study of tumour microenvironment.
From this analysis, the group was able to stratify ccRCC into seven evolutionary subtypes, which could be linked to clinical outcome. Two subtypes, defined by either multiple clonal drivers or BAP1 mutation, had lower levels of intratumour heterogeneity (ITH) and higher levels of genomic instability, and this correlated with rapid dissemination and poor clinical outcomes. Poorest overall survival was linked to a rare subtype with no driver alterations, including no VHL functional loss, that displayed increased ITH and genomic instability, and higher stage and grade of tumours.
The other four subtypes were less aggressive: early PBRM1 mutation followed by any of SETD2, PI3K mutation or somatic copy number alterations characterised slow-growing primary tumours with high ITH, and tumours in which VHL was the monodriver were indolent, with decreased ITH, genome instability (assessed using a weighted genome instability index; wGII) and size. Of note, the majority of small renal masses in the cohort had low ITH and low genome instability, but some fell into the less benign subtypes, and could progress in the absence of surgery. Evolutionary classification could therefore be used as an aid in active surveillance of small renal masses.
The evolutionary subtype group sizes were too small for formal survival analysis, which awaits assessment in the full TracerX Renal study cohort, but being able to stratify ccRCCs by genomic profiling is a valuable new tool. Key to this is the critical question of how many regions in a tumour need to be sampled to provide an adequate measure of subclonal alterations. The group found that four biopsies would capture ≥75% of variants in tumours with low ITH, but in cases with high ITH, eight biopsies were necessary to detect a similar proportion. The question of sufficient sampling and inference of evolutionary subtype is challenging for routine practice, so the group has recently developed a novel method that aims to overcome this (Litchfield et al, 2020).
In the final paper of the trilogy (Turajlic et al, 2018b), Turajlic and her colleagues provided a comprehensive picture of the genetic underpinning and evolutionary patterns of metastasis. To distinguish metastasis-competent from incompetent clones in the primary ccRCC tumour, and to examine the routes and timing of metastases across multiple anatomic sites, they analysed 575 primary and 335 metastatic biopsies from 100 patients.
Metastases exhibited signs of evolutionary bottlenecking: the overall number of driver events was lower compared to primary tumours, metastases were more homogeneous, and very few new drivers had arisen, suggesting that most driver diversity accumulates at the primary tumour site, and then metastatic-competent populations are selected.
Populations able to progress to metastases proliferated more quickly, and there was some evidence of immune evasion, as indicated by loss of heterozygosity at the HLA locus. Most notably, genomic instability, with a concomitant increase in ploidy, was significantly elevated. Instability was correlated with loss of chromosomes 9p and 14q, which together were hallmarks for metastatic potential. Critically for their potential use as biomarkers, both 9p and 14q loss were predominantly subclonal in the primary tumours, and may therefore be missed by single biopsy approaches.
In the context of the evolutionary subgroups of the primary tumours, rapid progression to multiple metastatic sites was associated with the multiple clonal drivers and BAP1 subtypes, which had low heterogeneity and higher genomic instability. The VHL wild-type tumours, which also had high levels of genomic instability, also fell into this category. In these subtypes, metastatic competence is acquired early in the primary tumour’s evolution, which drives rapid dissemination, leading to surgical failure, poor response to systemic therapy and early death from disease.
Conversely, in the four primary subtypes with high heterogeneity and lower levels of genome instability, metastatic competence was acquired gradually in subclonal populations, with only one or a few metastases observed. Metastatic capacity increased over time, and could eventually result in widespread disease, suggesting that for patients in these groups, surgical removal of the primary tumour before systemic therapy might remove the “evolutionary sink” from which metastatic mutations arise, thus minimising the risk of future metastatic seeding.
These papers have provided oncologists with important and much-needed new biomarkers-in-principle for stratification of renal cancer patients. Evolutionary classification of tumours could determine which patients would benefit from surgery in the presence of metastatic disease, how to manage patients following surgery with curative intent, including decisions on surveillance schedule and adjuvant therapy, and in the context of active surveillance of small renal masses.