Fourth International Biannual Evolution and Cancer Conference (Resistance, resilience, and robustness: Can we target cancer's evolutionary and ecological nature?)

have reduced proliferative fitness and stably rewired cell cycle control pathways. Mathematical modeling indicates that the tumor’s spatial structure amplifies the fitness penalty of resistant cells and identifies their relative fitness as a critical determinant of the clinical benefit of adaptive therapy. presented a study on the evolutionary dynamics of response to chemotherapies in breast cancer xenografts. This study showed how it is possible to finely resolve evolution in response to multiple chemotherapies by sequencing post- treatment residuals from patient- derived xenografts (PDXs) grown from two triple- negative breast cancer patients combined with exome- sequencing and 1,633 droplet digital PCR (ddPCR) measurements for mutation and CNV quantitation. Using assays of 86 xenografts and 45 derived cell cultures, it was possible to dis-tinguish selection from measurement uncertainty, intraclonal diversity, and spatial drift, with improvements over inferences from exome- sequencing data. Common modes of evolution within these tumors have been observed, including population bottlenecks, spatial diffusion, and stable coexistence between distinct subpopula-tions. Notably, it has been possible to show that a major pre- existing subclone exhibited higher cisplatin sensitivity but was favored when treatment was suspended, indicating an ecology susceptible to re-treatment by adaptive therapy. This demonstrates the importance of intratumoral dynamics in guiding treatment strategy. on the dynamics of and investigated. The evolutionary dynamics of escape from

Carlo Maley then discussed resistance management for cancer, especially drawing on knowledge from pest management that has led to three heuristic achievements in resistance management that could also be related to oncology. Maley explained that the overall aim is to transform cancer from a deadly disease into one that we can live with. This can be summarized as limiting the use of each mode of action (MoA) to the lowest practical level, diversifying the use of MoAs as much as possible, limiting each MoA to no more than two nonconsecutive uses, and partitioning MoAs in space or time so as to segregate their use as much as practically possible. Dr. Noemi Andor (Stanford University, CA, USA) presented her work on the identity of surviving and extinct clones in a longitudinal study of the DNA damage therapy response in gliomas. Overall, she showed that more than half of the clones detected among all patients were found across multiple biopsies of the same patient.
Moreover, mutation profiles and clonal compositions from proximal biopsies were more similar to each other than those from distant biopsies. The study revealed a higher growth rate among clones with more amplifications but only among patients who had received DNA damage therapy.
This first session closed with the keynote talk was given by Dr.
Christina Curtis on the way to quantify the evolutionary dynamics of therapeutic resistance and metastasis. Dr. Michael J. Metzger (Columbia University, NY, USA) next presented a study on the discovery a new kind of contagious cancer (leukemia-like disease) in the soft-shell clam (Mya arenaria), the Pacific blue mussel (Mytilus trossulus), the cockle (Cerastoderma edule), and the carpet shell clam (Polytitapes aureus). Transmission within each of these species is due to the independent horizontal spread of a clonal cancer lineage. However, while the cancer lineages in soft-shell clams, mussels, and cockles are each derived from their respective host species, the cancer cells in P. aureus are derived from Venerupis corrugata, a different species that lives in the same geographic area but which itself is not known to be highly susceptible to disseminated neoplasia. These findings show that transmission of cancer in the marine environment is common in multiple species, that it has originated many times, and that both cross-species transmission and species-specific resistance occur.

ROCHE Et al.
Dr. Chandler Gatenbee (Moffitt Cancer Center, FL, USA) followed with a talk on the characterization of the immunogenic bottleneck. Based on a branching hybrid nonspatial cellular automaton, this study investigated whether the explosive antigenic diversity observed in colorectal cancer can be explained by either a "get lucky" strategy, where clones can have low enough antigenicity to avoid immune detection, or a "get smart" strategy, where clones can acquire active escape mechanisms. Only the "get smart" model is able to recapitulate the observed patterns of antigen burden and change in immune composition, suggesting that an active immune escape mechanism is required for carcinogenesis and implying that the immune system is the first treatment tumors must evolve resistance to.
The parallel session dealt with cancer evolutionary genomics.
The first talk by Dr. Diego Mallo (Arizona State University, AZ, USA) presented the PISCA method, which is a new phylogenetic method for the reconstruction of somatic evolution using somatic chromosomal alteration data. This method, implemented as a plugin in the BEAST phylogeny software, is used to reconstruct the evolution of homogeneous somatic samples (i.e., single cells, single crypts, or deconvoluted clones) using somatic chromosomal alteration data. This method has been used to estimate the acquisition rate of somatic chromosomal alterations in Barrett's esophagus (BE) and its change through time. It has shown that the previously observed slow rate of evolution in this premalignant tissue is due to a low acquisition rate at the crypt level, explaining the low rate of progression from BE to esophageal adenocarcinoma by suggesting that clones with increased mutation rates appear to facilitate this transition. The next talk was given by Dr. Luca Ermini (The Institute of Cancer Research, London, UK) on the evolutionary selection of cancer-risk alleles. Analyzing genomes from five different Caucasian populations available in the 1,000-genome database and using standardized methods to scan for genomic signatures of selection in gene loci associated with cancer risk, the aim of this study was to understand why cancer-risk alleles are so frequent. While no (or neutral) selection was found for most alleles analyzed, a signal for positive selection was found in some variants associated with breast or prostate cancer in all populations analyzed and some populationspecific positive selection was found for some alleles associated with breast cancer. These results highlight new inroads into understanding the biological processes and evolutionary forces shaping cancer risk in humans.
The last talk of this session was given by Dr. Jeffrey Townsend (Yale University, CT, USA) on ways to quantify the intensity of natural selection on somatic mutations in cancer. Some high profile mutations have lower effect sizes than others whose p values are less significant but that exhibit a high effect size. Examination of the effect size conveys potential new targets for small populations, but also indicates that some high profile somatic nucleotide mutations (e.g., mutations in P53, even PIK3CA) have lower effect sizes than might be expected and may not have a successful therapeutic potential. Thus, a serious problem with using p values or mutation prevalence for ranking genes or mutations emerges from the same source that obviates use of genic mutation prevalence: the effect of mutation rate. Understanding the development of cancer as an evolutionary process permits the adaptation of classical evolutionary theory to use estimates of mutation rate to quantify selection intensity of mutation-cancer effect sizes. These effect sizes are the subject of analyses attempting to quantify the relative importance of mutations to tumorigenesis, cancer progression, and therapeutic resistance. patients treated with intermittent androgen-ablation therapy, this adaptive dynamics model of androgen-ablation therapy was then used to predict PSA dynamics in an independent set of 30 patients from the same clinical study. While predictions were usually reasonably accurate for one cycle, and for some patients up to four cycles, this model had some significant exceptions that can be explained by resistance arising from different mechanisms. Therefore, this modeling approach may provide a noninvasive method to identify emerging resistance mechanisms in nascent hormone-refractory tumors and to plan treatment to delay development of castration resistance.
Dr. Daniel Nichol (Institute of Cancer Research, London, UK) then showed how stochasticity in the genotype-phenotype map can have implications for the robustness and persistence of the "bet-hedging" strategy in cancer cell populations. Drug tolerance mechanisms have been observed without apparent genetic drivers, suggesting that bet-hedging may play a role in driving resistance. Through a simple model involving a molecular switch, it was possible to demonstrate that bet-hedging is resistant to loss from mutations in both the expression of genes and their interactions, suggesting that single-gene knockouts may be insufficient to elucidate the drivers of bet-hedging. The implications for therapy have been investigated, highlighting that the successful attempts to "steer" the evolution of bet-hedging through drug holidays will be dependent on the G-P mapping.
The parallel session was on the evolution of cancer suppression mechanisms and organism robustness. The first speaker, Dr.
Marc Tollis (Arizona State University, Tempe, AZ, USA), discussed a molecular evolutionary approach to understanding cancer suppression, and especially Peto's paradox. While large species should face a higher lifetime risk of cancer due to the greater probability of oncogenic mutations occurring during somatic evolution, zoo necropsy data reveal that elephants have a ~5% probability of death from cancer compared to 11%-25% for humans. This study showed that elephant genomes harbor up to 40 alleles of the tumor suppressor gene TP53. Moreover, functional assays demonstrate that TP53 redundancy in elephants is related to an increased apoptotic response to DNA damage in elephant cells when compared to human cells. Across >50 mammalian genomes, multiple tumor suppressor gene copy-number expansions have been found to co-occur with the evolution of large body size or longevity in elephants, bats, horses, and rhinos, suggesting that convergent evo- have reduced proliferative fitness and stably rewired cell cycle control pathways. Mathematical modeling indicates that the tumor's spatial structure amplifies the fitness penalty of resistant cells and identifies their relative fitness as a critical determinant of the clinical benefit of adaptive therapy.

Dr. Jeffrey Chuang (The Jackson Laboratory for Genomic
Medicine, Farmington, CT, USA) presented a study on the evolutionary dynamics of response to chemotherapies in breast cancer xenografts. This study showed how it is possible to finely resolve evolution in response to multiple chemotherapies by sequencing post-treatment residuals from patient-derived xenografts (PDXs) grown from two triple-negative breast cancer patients combined with exome-sequencing and 1,633 droplet digital PCR (ddPCR) measurements for mutation and CNV quantitation. Using assays of 86 xenografts and 45 derived cell cultures, it was possible to distinguish selection from measurement uncertainty, intraclonal diversity, and spatial drift, with improvements over inferences from exome-sequencing data. Common modes of evolution within these tumors have been observed, including population bottlenecks, spatial diffusion, and stable coexistence between distinct subpopulations. Notably, it has been possible to show that a major pre-existing subclone exhibited higher cisplatin sensitivity but was favored when treatment was suspended, indicating an ecology susceptible to retreatment by adaptive therapy. This demonstrates the importance of intratumoral dynamics in guiding treatment strategy.
The topic of the next talk, given by Dr. Benjamin Werner (The Institute of Cancer Research, London, UK), was about forecasting resistance evolution in cancer from liquid biopsies. After discussing some approaches on how this heterogeneity might be better classified from multiregion sequencing data and how this might improve the selection for potential targets of treatment, this study has shown how sequential sampling of circulating tumor DNA (ctDNA) in patients during treatment can be used to forecast the evolution of treatment resistance. Interestingly, a combination of sequential sampling and evolutionary modeling does not only detect resistance but also allows quantifying some properties of the evolutionary process.
Dr. Nara Yoon (Cleveland Clinic Foundation, OH, USA) followed with a talk on optimal chemotherapy scheduling based on a pair of collaterally sensitive drugs. To avoid drug resistance, researchers have proposed sequential drug therapies so that the resistance developed by a previous drug can be relieved by the next one, a concept called collateral sensitivity. In this study, dynamic models were developed and revealed that the optimal treatment strategy consists of two stages: (Stage 1) the initial stage in which a chosen "better" drug is utilized until a specific time point, T; and then (Stage 2) a combination of the two drugs with a relative intensity (f) for Drug A and (1-f) for Drug B. Importantly, the initial period during which the first drug is administered, T, has to be shorter than the period in which it remains effective, contrary to clinical intuition. Finally, the last talk of this session was given by Dr. Kimberly J. Bussey (NantOmics, Phoenix, AZ, USA), who showed that a noninherited mutation is constrained by the genomic evolutionary history in nonintuitive ways. This study identified the noninherited (de novo) single nucleotide variants (SNVs) in 129 individuals using the methodology of somatic variant calling. Through different data treatments, it was observed that the SNVs filtered out had different evolutionary properties depending on the filter applied. SNVs that were filtered out at data quality control stages were enriched for regions of the genome that pose difficulties for unique alignment, such as segmental duplication and inversion regions (SDRs) and nonallelic homologous recombination (NAHR) substrates, but not LTRs. In contrast, data filtered out by allele frequency were enriched in LTRs and homologous synteny blocks and excluded from SDRs, NAHRs, and evolutionarily re-used breakpoints. Additionally, SNVs filtered at the data quality steps were slightly enriched to be in clusters of variants while those filtered by allele frequency were excluded from clusters, suggesting that clustering is dominated by somatic/early germline events. In general, SNVs preferentially affected genes younger than 1,500 MY, but strong filtering tends to create a bias against recovering this pattern. reports of cancer, these species must be particularly resilient to mutations or rely on highly effective molecular mechanisms of damage prevention, DNA repair, or tissue-level cancer control that are worth investigating. This has been tested on three invertebrates for which there have been no reports of cancer: Trichoplax adhaerens (Placozoa), Tethya wilhelma (sponge), and Macrostomum lignano (flatworm). Dr. Fortunato observed that T. adhaerens have an elevated resistance (~160 Gy) to X-rays, where cellular aggregates with a different morphology were observed after several weeks of high-dose exposure (even though it is unsure that these were a form of cancer).
T. wilhelma, which adapts well to being cultured in a laboratory setting, is even more resistant to X-rays (~700 Gy). Finally, the flatworm M. lignano has an elevated regenerative ability conferred by its high percentage of stem cells (the highest recorded in an animal) and is much less resistant to X-rays (~60 Gy).
This session ended with a talk by Dr. Pierre Martinez (Cancer Research Center of Lyon, Lyon, France) on the evolution of Barrett's esophagus (BE) through space and time at single-crypt and whole-biopsy levels. In this study, researchers noted copynumber alterations (CNA) from SNP arrays in 6-11 biopsies over two time points in each of eight individuals with Barrett's esophagus, including four cancer progressors. Eight individual crypts and the remaining epithelium were assayed for each biopsy, yielding 358 valid samples. This allowed the characterization of genetic diversity at an unprecedented resolution and the reconstruction of corresponding phylogenies. In six patients, CNAs could be detected in all crypts and biopsies, suggesting lesions derived from a single ancestor. While crypts contained private mutations, mutational load and rates in crypts were similar to those in whole biopsies; thus, biopsies were adequate for evolutionary studies.
Moreover, Dr. Martinez observed that "macrodiversity" between biopsies reflected the "microdiversity" between crypts of a biopsy, that genetic distances between crypts were unrelated to physical distances, and that rare clonal expansions indicated that BE lesions are mostly evolving neutrally. These results shed new light on the evolutionary dynamics underlying BE genetic evolution and reveal they are adequately described by biopsy-level macroscopic heterogeneity.
These two parallel sessions were followed by a plenary talk by