Adaptive Evolution: How Bacteria and Cancer Cells Survive Stressful Conditions and Drug Treatment

Complex multicellular organisms, including mammals, exploit sophisticated and tightly regulated networks and genetic constraints that subject the fitness of individual cells to the fitness of the whole organism. The evolution of multicellularity features underscores selection for cooperation between unicellular entities. Cells must proliferate only when required to fulfill their function, and cell death by apoptosis is sometimes essential to maintain tissue homeostasis (1). However, harmful events, such as somatic mutations or viral infections, may subvert multicellularity constraints, and the hallmarks of cancer (the common traits of human tumors) include many features reminiscent of a “selfish,” unicellular-like life (2).

During tumorigenesis, cancer cells forgo the cellular cooperation and the regulatory mechanisms that preserve homeostasis in multicellular organisms (3). Cancer cells display sustained proliferation, evade signals of growth suppression, resist programmed cell death, and redirect allocation of resources by promoting neoangiogenesis. Moreover, tumor progression is also characterized by progressive cellular dedifferentiation (ref. 2; Fig. 1).

By dysregulating cellular processes associated with the transition from unicellular to multicellular life, cancer cells activate survival strategies, including rapid proliferation and adaptability to stressful environments, that had been perfected by autonomous organisms such as bacteria (4–7). In other words, cancer, a disease of multicellular organisms, shares biological features with prokaryotic cells (Fig. 1).

Keeping in mind the remarkable evolutionary distance between bacteria and mammalian cells, here we discuss the scientific and therapeutic implications of considering cancer cells as selfish forms of life which subvert multicellularity laws and behave as single competing units much like bacteria (7, 8).

Cancer can be considered as an atavistic form of life, characterized by competition among individual cells for access to nutrients, survival, and proliferation, similar to what occurs in bacterial populations (1, 7). As a consequence, several hallmarks and acquired capabilities of cancer cells are strikingly similar to properties displayed by microorganisms (2, 4). For instance, from a metabolic point of view, cancer cells mainly rely on anaerobic glycolysis (the so-called Warburg effect), which is the prevalent energy source among bacteria (9).

In natural environments, microorganisms face a constant battle for nutrients and have evolved several mechanisms to compete with other individuals for resources, including rapid growth, increased matrix production to secure favorable environments, and production of antimicrobial toxins (10, 11). Similarly, human cancers display high levels of molecular heterogeneity, and within a single lesion distinct subclones can compete with each other (12). This can be observed especially under treatment with anticancer agents, when previously unfit drug-resistant subclones outcompete the dominant drug-sensitive cells (13, 14).

Intriguingly, bacteria can also thrive as complex communities regulated by cooperation or competition, either between individuals from the same species or between distinct bacterial strains (4). For example, especially (but not exclusively) when facing hostile environments, bacteria can form biofilms, multicellular aggregates embedded in a proteinaceous extracellular matrix (15). The formation of a biofilm promotes survival of the population, as these complex structures have been shown to protect bacteria from antibiotics, toxins, and solvents (16).

Analogously, the interactions between cancer cells and their stroma can be compared with a special form of biofilm, whereby cancer cells orchestrate neoangiogenesis, recruitment of cancer-associated fibroblasts and immune cells to promote the survival of the entire community (2, 17, 18). Cooperation through paracrine signaling can also promote increased resistance to targeted treatment; for example, we previously showed that in colorectal cancer, treatment with anti-EGFR targeted therapy promotes increased secretion of TGFα and amphiregulin from resistant cells, which in turn protect surrounding sensitive cells in the population from EGFR blockade (19).

Both bacteria and cancer can deploy complex responses to hostile environments to protect the population from eradication. Some of the bacteria inside a biofilm are able to switch to planktonic state and colonize distant sites (3, 20). Moreover, under stressful conditions some individuals can undergo sporulation, generating highly resistant spores that germinate once favorable growth conditions are restored (15, 21). Similarly, some cells within a tumor (possibly in response to challenging circumstances) acquire metastatic competence, i.e., the ability to separate from the cancer of origin and colonize distant organs (22). At the site of colonization, disseminated tumor cells encounter an unfavorable environment, in which they enter a state of dormancy (just like spores) and slowly condition the surrounding stroma until a supportive niche is formed (23).

Notably, competition can lead to selection of variants that are better suited to colonize the environment. The evolution of tumors cells and asexual bacteria is governed by the interplay of genetic drift, heritable variation, and Darwinian selection (24). Experimental and computational evidence shows that constitutive mutators are present at low frequency in a bacterial population, because they result in the overproduction of lethal mutations. However, in stressful conditions (such as antibiotic treatment or limited access to nutrients), hypermutators are positively selected and can become fixed in the majority of the population (25–29). Similarly, genome instability is a hallmark of cancer and is favorably selected in tumor cells for its role in fostering the emergence of progressively fitter subclones and supporting tumor evolution (2).

The atavistic hypothesis states that tumorigenesis involves the reactivation of survival programs dating back to the evolution of unicellular organisms, and the concomitant disruption of complex features evolved to support the intercellular communications required for multicellular organisms' development and physiology (7, 30).

Transcriptional analysis of more than 3,000 tumor samples across seven different tumor types unveiled that cancers rely on unicellular processes for survival. Dysregulation of genes and cellular processes unique to multicellularity was detected across multiple tumor types, together with strong upregulation of genes conserved in unicellular organisms, including those involved in proliferative signaling, cell death escape, and genomic instability (5, 31). Expression of highly conserved genes is linked to drug resistance in cancer cells (32), and tumors often activate transcriptional programs associated with dedifferentiation (33, 34).

Analysis of the evolutionary age of genes involved in human cancer revealed increased mutational processes in genes younger than 500 million years, suggesting selection for somatic mutations occurring in younger genes preferentially linked to multicellularity (35).

Goode and colleagues used a computational approach to analyze different types of tumors from more than 9,000 patients, to unveil accumulation of point mutations and copy-number variations (CNV) in genes dating back to metazoan ancestors' genes. Their analysis showed that point mutations disrupt key regulators of multicellularity networks, and that CNVs dysregulate downstream effectors of these pathways (36).

Relatedly, genes with regulatory function in the communication between genes of unicellular and multicellular origin represent vulnerabilities that, when compromised, promote the emergence of tumorigenesis. For example, TP53, NF1, and PI3KCA genes, which are frequently mutated across different tumor types, act as regulatory hubs of multiple processes fundamental for the maintenance of the integrity of genomic networks (31, 36).

Whether atavism plays a role in cancer initiation and progression is still debated, as acquisition of cancer's biological features during tumorigenesis can be explained by stochastic genetic alterations and progressive strong environmental selection rather than reactivation of a conserved survival program. However, as discussed below, the two theories are not necessarily mutually exclusive and can coexist (37). According to basic laws of evolution, an enriched fraction of mutants in a population would increase chances of survival of a few individuals under selective conditions. Indeed, unicellular organisms deploy strategies of adaptive mutability to foster evolution in strongly selective environments (see below). Intriguingly, we and others recently observed that cancer adaptation to stress is at least initially fostered by activation of similar “atavistic” survival programs observed in bacteria under stress, thereby promoting mutagenesis and genetic diversity; selection and expansion of phenotypes capable of replicating under stressful environments then occurs (refs. 38, 39; see details below).

When facing stressful environments, bacteria deploy rapid and reversible survival programs that foster genetic diversity and facilitate adaptation to changes in the environment (5, 40, 41). Effective strategies to evade a stressful condition include relocating to a new environment through swimming and the formation of biofilms, as discussed above (3, 4).

An additional fascinating strategy exploited by unicellular cells to evade stress-induced death relies on temporary phenotypic switches. Indeed, in bacteria a fraction of cells named “persisters” evades antibiotic killing by entering a physiologically dormant state displaying transient phenotypic stress tolerance, without undergoing genetic changes (42, 43). Persister cells, different from permanent genetically resistant cells, reinitiate growth and regenerate a drug-sensitive population upon termination of drug treatment. Importantly, development of resistance eventually occurs upon prolonged drug treatment of persister cells, representing a major cause of antibiotic failure, which leads to recurrent bacterial infections (43).

Many of these stress adaptations are tailored to the specific stress. For example, the SOS DNA damage response induces activation of error-prone DNA polymerases that can bypass the damage and terminate replication, while in case of high temperatures, induction of heat-shock chaperones allows the resolution of misfolded or aggregated proteins (44).

The SOS response is triggered by DNA damage caused by cell malfunctioning, stalled DNA replication forks, or conditions such as oxidative stress and antibiotic treatment. In E. coli and related bacteria, the coating of RecA protein to single-strand DNA (ssDNA), generated either as a result of DNA damage itself, following stalled DNA replication forks, or during the DNA-repair process, induces the autoproteolytic cleavage of LexA, a transcriptional repressor, leading to upregulation of SOS regulon genes (45, 46). The SOS regulon includes genes involved in multiple functions, such as DNA recombination and repair, nucleotide excision repair, DNA synthesis past damaged bases, and control of cell division (45, 46). Among these, the activation of specialized low-fidelity DNA polymerases, which temporarily replace canonical replicative polymerases, is responsible for a mutagenic form of DNA replication associated with increased rate of base substitutions and indels (47–49).

However, in addition to stress-specific responses, many bacteria deploy a strong general stress response to react to a variety of growth-limiting conditions and stresses, such as nutrient deprivation, stationary phase growth, DNA damage, and extreme temperatures or pH (44, 50). In E. coli and related bacteria, the master regulator of the general stress response is the sigma factor RpoS. Different mechanisms regulate the increase of RpoS levels and the induction of the multiple RpoS-dependent effectors; indeed, some are dependent only on RpoS, whereas others require certain conditions and additional inputs (44, 50).

Both RpoS and SOS responses seem to play a role in persister formation. In particular the SOS-dependent DNA-repair functions are central for persister formation and recovery. Indeed, emergence of surviving persisters in response to DNA-damaging agents is impaired in SOS-deficient mutants (43).

Cancer cells are constantly exposed to intracellular and extracellular stressful stimuli, such as hypoxia, reduced nutrient availability, glucose deprivation, and DNA damage. In order to foster adaptation and survival, cells deploy a variety of stress responses to deal with adverse environments. Indeed, cancer cells activate multiple survival strategies, alter gene expression, reprogram metabolic pathways, and trigger stress-response mechanisms that can promote mutability and genetic diversity (5).

Cancer cells, like bacteria in response to antibiotics, can enter a reversible drug-tolerant persister state when exposed to targeted therapies (51, 52). Through nongenetic mechanisms of drug tolerance, slowly replicating cancer persister cells escape death and avoid tumor eradication (53). Of clinical relevance, extended treatment of persister cells inevitably leads to the development of resistance and consequent treatment failure (52, 54).

In mammals, the mechanistic target of rapamycin kinase (MTOR) has been proposed as a possible analogue of the bacterial RpoS general stress regulator, which controls the switch to mutagenic break repair in E. coli (5). MTOR has a central role in sensing environmental conditions, including lack of growth factors, nutrient deprivation, low oxygen, and DNA damage, and regulates many fundamental cell processes, such as protein synthesis, metabolism, cell growth, autophagy, and aging (55). Dysregulation of mTOR regulatory signaling is implicated in cancer progression, and mTOR inhibitors have been developed as anticancer agents (55).

Cells are continuously exposed to internal and external sources of damage and have therefore evolved an intricate network of signaling pathways and repair mechanisms to cope with DNA lesions and protect genome stability. Interestingly, several error-prone DNA polymerases have also been described in eukaryotic cells, including yeasts and human cells (49).

Why are error-prone DNA polymerases conserved in highly developed organisms, if programmed cell death allows the elimination of cells with damaged genomes?

In human cells exposed to nonstressful conditions, when the replication fork stalls upon encountering a site of DNA damage, canonical replicative DNA polymerases are replaced by specialized polymerases characterized by poor accuracy in nucleotide incorporation and increased rates of base substitutions and frameshifts, ranging from 10⁻³ to 10⁻¹ errors per base pair (49, 56); propensity to incorporate nucleotides using aberrant DNA primer ends; absence of proofreading 3′→5′ exonuclease activity; and moderate-to-low processivity (49, 57).

The expression of error-prone DNA polymerases is tightly regulated. In physiologic conditions, they are involved in multiple biological functions, such as nonhomologous end joining (NHEJ) and translesion synthesis (TLS) DNA-repair processes (58), somatic hypermutation of immunoglobulin genes (59), and protection against UV-induced genetic instability and skin cancer, through the prevention of replication fork collapse and double-strand breaks (DSB) accumulation (60).

However, according to mathematical models, in growth-limiting hostile environments the evolutionary costs of high-fidelity DNA replication may exceed the costs of error-prone replication, becoming therefore a counterproductive strategy, whereas the proliferation and survival of cells with limited DNA-repair capacity is favored, both in microbes and in cancer cells (61). This suggests that in some instances, also in multicellular organisms, cell survival is prioritized over genome perfection and occurs by downregulation/repression of DNA-repair mechanisms, which is preferred over stalling of DNA replication.

Indeed, DNA damage tolerance (DDT) response, evolved side by side with DNA-repair mechanisms, functions in circumstances where the possibility to make errors has a selective advantage. Through the involvement of specialized polymerases able to bypass DNA damage, DDT allows the replication of a damaged template, and postpones the repair at later time points, thus lowering the overall risk of replication fork collapse, genome instability, genetic rearrangements, and cell death (58).

The association of DDT with cancer is double-faced. On the one hand, defects in DDT are associated with cancer susceptibility and aging (58). On the other hand, uncontrolled overexpression of error-prone polymerases has been associated with accumulation of mutation and tumorigenesis in lung tumors (62), poorer prognosis in patients with glioma (63), and drug resistance (64). Notably, the involvement of DDT and error-prone DNA polymerases in stress responses affects the mutational profile, leading to a temporary increase of genetic diversity, which can favor adaptation to a hostile environment and foster evolution (38, 39, 48).

Mutations represent the primary source of genetic variation and, together with environmental selection, are an important evolutive driving force, in both cancer cells and bacteria (65).

Growing evidence shows that multiple microorganisms, such as bacteria and yeast, adaptively combine stress responses that transiently increase their genomic instability, even during replicative quiescent state, in response to different environmental stresses (5, 41, 66). The resulting stress-induced adaptive mutability promotes stochastic emergence of fitter mutants, leading to higher rates of adaptive evolution (41, 45, 66).

Bacteria enter a dormant state when life is unsustainable and recover when it is beneficial to do so. Acquisition of mutations during the stationary phase indicates that cells exposed to stress respond by promoting mutability; a subset of the newly acquired mutations fosters adaptation to the stressful environment.

Indeed, in a constant environment, well-adapted populations evolve with a constant low mutation rate, while, when organisms are maladapted to their environments (e.g., in conditions of starvation or when exposed to hypoxia or antibiotic treatment), hypermutator phenotypes can emerge (66, 67). These phenotypes transiently provide elevated mutation rates when populations experience bursts of stresses. However, when a new (adapted) population emerges, the mutator phenotype is no longer an advantage and is counterselected, and eventually the genome of the expanding adapted mutants returns to low mutation rates (45, 66).

In E. coli, stress-induced mutagenesis via mutagenic DNA break repair involves three components: (i) repair of DSBs; (ii) activation of the SOS damage response, which upregulates error-prone DNA Pol IV; and (iii) activation of the RpoS-mediated general stress response, which allows the accumulation of Pol IV errors in acts of DSB repair (5, 45, 66). This stress-induced adaptive mutability appears to induce mutagenesis in proximity to DSBs, creating clusters that reduce in density with distance from the DSBs (68). Stress-induced mutagenicity generates distinct forms of genetic changes, including single-nucleotide variants, insertions and deletions (indels), CNVs, and chromosomal rearrangements (45, 66). Induction of error-prone DNA polymerases and other genes of the SOS regulon is required for most point mutagenesis.

Permanent loss of highly conserved mismatch-repair (MMR) proteins, responsible for the recognition and repair of mispaired bases and 1-base insertion/deletion, generates constitutive mutators (69). Interestingly, the function of the MMR system is temporarily disabled during adaptive mutability in response to stress. Indeed, overexpression of one or more MMR proteins reduces stress-induced mutations (66).

Downregulation of DNA-repair capacity, associated with overexpression of error-prone DNA polymerases, increases the number of errors that persist and become mutations, therefore translating into more mutagenic replication of DNA and genomic instability (70).

Notably, it has been shown that even sublethal concentrations of antibiotics induce downregulation of the MMR system and upregulation of error-prone DNA polymerases which, in turn, promote increased genetic instability (70). This temporary increase in mutation rate promotes the generation of mutants and ultimately the acquisition of resistance to a wide spectrum of drugs, including compounds unrelated to the antibiotic applied (71).

Although it is easy to conceive that adaptive mutability is essential for the survival of unicellular organisms exposed to stress, multicellular organisms evolved multiple constraints to ensure genetic integrity and homeostasis, such as the tightly controlled recruitment of error-prone DNA polymerases in the replication of damaged DNA templates described above (5).

However, it has been shown that, when exposed to stressful conditions, tumor cells exploit survival strategies resembling bacterial stress-induced mutagenesis (38, 39). For example, constitutive activation of the TGFb axis causes downregulation of homologous recombination (HR)–mediated DSB repair, increased genetic instability, and clonal diversity, thus boosting the ability of cancer cells to adapt to drug treatment and the development of chemoresistance (72). Analogously, cancer cells exposed to hypoxia activate a transcriptional regulatory response that downmodulates DNA-repair mechanisms such as MMR and HR (73). Moreover, hypoxia promotes upregulation of error-prone DNA-repair pathways, such as NHEJ, and the use of TLS DNA polymerases (73). This switch from highly accurate to error-prone DNA repair creates a permissive milieu for genomic instability; in fact, hypoxic cancer cells display increased mutability and accumulation of frameshift mutations at repetitive microsatellite sequences in reporter plasmids (74).

The remarkable capacity of tumor cells to adaptively modulate their mutagenicity has also been observed in response to targeted agents, similar to bacteria in response to antibiotics.

Indeed, we recently showed that colorectal cancer cells, under stress imposed by targeted therapy, activate a response recapitulating key element of bacterial stress-induced adaptive mutability. Treatment with the anti-EGFR monoclonal antibody cetuximab, alone or in combination with the BRAF inhibitor dabrafenib, induced transient deficiency of both MMR and HR-mediated DNA-repair proficiency, and a concomitant switch from a high- to a low-fidelity DNA replication (38). Notably, treatment with these non–directly genotoxic drugs caused increased DNA damage and the observed response, in turn, translated into increased genetic instability. The stress-induced adaptive mutability phenotype in cancer cells, as in bacteria, was transitory and confined to maladapted cell populations. When the stress was terminated, either by drug withdrawal or emergence of permanently resistant cells, the MMR and HR levels reverted back (38).

We propose that during tumor development, when breakdown of multicellularity occurs, a stress response aimed at promoting survival over functional and genetic integrity is expediently restored.

Adaptive mutability in response to stress appears to be an evolutionarily conserved process rather than a phenotype restricted to colorectal cancers. Indeed, Cipponi and colleagues observed induction of DNA damage and a switch to mutagenic DNA replication process in melanoma and prostate, breast, and pancreatic cancers exposed to nongenotoxic drugs. This in turn resulted in genetic diversity and development of secondary resistance (39).

Notably, both in melanoma and colorectal cancer cells, the modulation in DNA-repair machinery was not simply a consequence of DNA damage or cell-cycle arrest, but was exquisitely linked to the inhibition of the specific oncogene addiction across tumor types, and was abolished when targeted treatment was applied to resistant cells (38, 64).

New mutations are constantly generated and fixed in cancer cells at a rate that, unless influenced by environmental agents or defects in the mechanisms of DNA repair, in many cases reflects the mutation rate of the tissue of origin (75). However, during cancer development and progression, cancer cells are constantly exposed to multiple stresses and changing environments (such as hypoxia, nutrient deprivation, metastatization to different organs, and ultimately therapy), which can affect their mutagenic ability (74). These punctuated bursts of evolution increase the probability that genetic alterations conferring survival advantage in the environment might be acquired (14). Similarly, targeted therapies represent yet another selective condition where increased generation of variants could facilitate the evolution of fitter subclones and ultimately the survival of cancer cells.

Although the role of stress-induced adaptive mutability in the acquisition of resistance to targeted therapies in cancer requires further elucidation, several indications point in this direction. We observed treatment-induced loss of MMR protein expression in clinical specimens obtained at maximal tumor response (nadir) in patients with MMR-proficient colorectal cancer treated with targeted therapies (38). Moreover, increased mutability at microsatellite regions was detected in preclinical models of acquired resistance (38).

Similarly, treatment of non–small cell lung cancer cell lines with the EGFR inhibitors gefitinib and erlotinib induced increased DNA damage and temporary deficiency in base-excision repair (BER). The resulting increase in genetic instability promoted the acquisition of the EGFR p.T790M mutation, which drives acquired resistance to anti-EGFR treatment (76). In addition, DNA barcoding and mathematical modeling of triple-negative breast cancer cells in response to BET and CDK4/6 inhibitors showed a higher rate of de novo mutagenesis under drug pressure (77).

This growing body of literature suggests that the features of the stress response displayed by microbial organisms have counterparts in cancer cells, and supports the possibility that tumors are a community of individual entities with remarkable survival and evolutionary capabilities (5).

In E. coli, both RpoS-mediated general stress response and SOS DNA damage response need to be activated to initiate stress-induced adaptive mutability (45, 66). The SOS-mediated activation of error-prone DNA polymerases makes the repair of DSBs a source of mutagenesis and increased genetic variations (48). However, RpoS is required to permit the use of error-prone polymerases during stress (47). In addition, although the mechanisms of MMR downregulation during adaptive mutability are not well characterized, RpoS also seems to be implicated in this temporary impairment of DNA repair (78, 79). Importantly, RpoS, and not error-prone polymerases, appears to contribute to genetic amplification, suggesting a central role for RpoS in the induction of genome instability under stress (79).

Recently, MTOR has been proposed as the master regulator umpiring adaptive mutability in response to stress across multiple cancer types (5, 39, 64). Using genome-wide functional screens in response to different types of targeted therapies, Cipponi and colleagues identified two classes of genes involved in the development of secondary resistance to targeted therapy: (i) the “solution” genes, directly responsible for conferring resistance to the specific anticancer agent used (e.g., TP53 alterations in response to the MDM2 inhibitor nutlin); and (ii) “facilitator genes,” capable of indirectly facilitating the process of adaptation to drug-imposed hostile environment. MTOR was identified among the top-ranked facilitator genes, and its pharmacologic inhibition was shown to selectively impair HR and canonical high-fidelity DNA polymerase genes (39).

Moreover, Temprine and colleagues recently showed that treatment of melanoma cell lines with the BRAF inhibitor dabrafenib induced upregulation of error-prone polymerase κ (Pol κ), and this was central to the emergence of drug tolerance (64); furthermore, inhibition of mTOR recapitulated the phenotype (64). Polymerase κ is the mammalian ortholog of E. coli Pol IV, the error-prone DNA polymerase responsible for base substitutions and indels during mutagenic DNA break repair in E. coli (48). Altogether, these results suggest a key role for MTOR in fostering genetic diversity and promoting adaptive evolution in cancer cells during drug treatment.

Overcoming drug resistance remains a major challenge in oncology. Although it is generally accepted that the intrinsic molecular heterogeneity that characterizes most tumors plays a major role in the emergence of drug resistance, it remains unclear whether yet-to-be-discovered biological features of cancer cells play a role in therapeutic failures.

The analogies between drug-resistance mechanisms arising in cancer and infectious diseases, such as mutations in the drug target or activation/amplification of parallel compensatory pathways, suggest that they can provide insights for the development of novel anticancer treatments (80). For instance, the successful experience of combinatorial strategy at treatment initiation in the case of HIV and tuberculosis infections highlights the relevance of using drug combinations in order to achieve stronger tumor shrinkage, while preventing the outgrowth of resistant clones (80).

Here, we propose that further insights on how tumor cells survive adverse environments, and specifically stress generated by therapeutic treatment, can be obtained by observing the evolutionary strategies used by unicellular microorganisms in response to drug treatments and stressful environments (81), and could be translated into novel concepts for therapeutic interventions.

In detail, if stress response strategies exploited by unicellular organisms represent phylogenetically conserved mechanisms of survival, then conceptual interpretation of cancer as a community of unicellular organisms can be inspirational for developing new anticancer therapies to curb or restrict the emergence of drug resistance (81, 82).

Modern therapeutic regimens have often focused on targeting cell hyperproliferation features (e.g., targeting aberrant and constitutive activation of oncogenes). Though initially effective, activation of parallel pathways and/or compensatory feedback loops usually leads to treatment failure (83, 84), highlighting that cellular proliferation is a redundant, multilayer (and therefore seemingly unbeatable) process in cancer cells. However, by considering at least some features of cancer cells as a phylogenetic throwback, new vulnerabilities emerge. As a matter of fact, drugs targeting cellular processes shared with unicellular organisms are already used in clinical practice, including inhibitors of purine and pyrimidine synthesis, proteasome inhibitors, and inhibitors of the mitotic spindle (7).

Notably, like bacteria in response to antibiotics, cancer cells in quiescent/slowly replicating state may contribute to survival under stress, and represent a reservoir from which mechanisms of drug resistance could eventually emerge (54). Although both tumor progression and response to anticancer drugs are fueled by mutations, the evidence that patients can relapse after a long disease stabilization, or that tumors can be successfully rechallenged with the same drug, cannot be explained by genetic resistance. Dormant drug-tolerant tumor cells might represent a major source of relapse in these instances (53). The importance of targeting persister cells in both cancer and infectious disease in order to prevent relapse has been previously highlighted (80, 85, 86); however, little is known about persistence mechanisms.

We and others recently unveiled that cancer persister cells activate a stress-induced adaptive mutability response as an ancestral survival program that leads to increased mutagenicity, thus fueling selection of newly generated resistant variants (38, 39). Though adaptive mutability increases the chances of survival in stressful environments, this strategy comes at a cost, and stress-induced modulation of DNA-repair pathways might unveil cell vulnerabilities exploitable for innovative therapeutic strategies (81).

We propose here some therapeutic approaches aiming at preventing the development of drug resistance, targeting cancer cell features resembling the unicellular ones.

As discussed above, in bacteria and cancer cells undergoing adaptive mutability in response to stress, the MMR and HR DNA-repair machineries are repressed, and DNA damage is tolerated, thus promoting mutagenesis and improving survival. Interfering with these mechanisms may undermine cellular survival and promote DNA damage–induced cytotoxicity. Indeed, increasing levels of DNA damage to a point that cancer cells cannot tolerate is already exploited in clinical care. Tumors under therapeutic stress display increased endogenous DNA damage, together with delayed repair of radiation-induced and chemically induced DNA lesions, as pointed out by several works in the literature (Fig. 2; refs. 87–94). Indeed, targeted therapies are already successfully combined with radiotherapy or chemotherapy to increase chances of tumor control.

Multiple pieces of evidence indicate modulation in DNA-repair systems upon treatment with targeted therapies across different tissues (Fig. 2; refs. 38, 39, 64, 76, 89, 91, 93–102). For example, blockade of the RAF/MEK/ERK pathway induces a BRCAness phenotype that sensitizes cells to PARP inhibitors (PARPi), both in vitro and in vivo (Fig. 2). Sun and colleagues showed that targeted therapy inhibiting the MAPK pathway restores sensitivity to PARPi in otherwise PARPi-resistant ovarian cancer cell lines (102). Relatedly, Maertens and colleagues found that cotreatment of melanoma cells with vemurafenib and histone deacetylases (HDAC) inhibitors leads to concomitant downregulation of both HR and NHEJ, an ineffective repair of DSBs, and ultimately cell catastrophe (ref. 101; Fig. 2). Based on this, synthetic lethality approaches capitalizing on the transient deficiency of DNA-repair systems in stressed cancer cells might be exploited to prevent cancer adaptation and evolution.

Analogously, inhibition of the transcriptional networks controlling induction of adaptive mutability in cancer cells might be successful in preventing development of resistance through de novo mutagenesis. Although MTOR has emerged as a possible master regulator of adaptive stress response in multiple tumors, it has been shown that mTOR inhibition recapitulates the adaptive mutability phenotype rather than preventing it (39). Therefore, the applicability of clinically available mTOR inhibitors for this purpose remains uncertain.

On the contrary, inhibition of hyperactivated error-prone DNA polymerases might eventually interfere with enhanced mutagenic capacity. For example, overexpression of Pol κ, which plays a central role in the adaptive response by regulating genes associated with drug resistance and immune surveillance through noncanonical mechanisms, further increases resistance to treatment, suggesting it as a putative target to curb drug tolerance (64).

Following the lesson learned from infectious disease, in order to optimize the potential success of these strategies in preventing cancer adaptation and curbing the tumor evolution, targeted therapies and drugs targeting the stress-response phenotype should be administered in combination from the initiation of the treatment (i.e., before development of resistance).

We and others previously reported the preclinical and clinical potential of concomitant inhibition of multiple nodes along the same pathway in terms of prevention of secondary resistance (83, 103) and overall survival in patients with colorectal cancer (104, 105).

Notably, as we previously demonstrated, interfering with oncogenic addiction causes increased DNA damage and shifts the dependency of cancer cells toward DDT pathways to cope with increased instability (38). The concomitant inhibition of multiple pathways with agents simultaneously targeting oncogene addiction and the adaptive response activated in cancer cells exposed to targeted therapies might therefore restrain the emergence of resistant variants by denying activation of the salvation mechanisms in cancer cells.

The concept of a paradigmatic shift from a reactive to a proactive strategy is indeed central to the approach of targeting mechanisms of adaptive mutability and resistance. Currently, in most instances medical oncologists wait for resistance to emerge; next, the resistant tumor is molecularly profiled, and then therapy is modified accordingly. Instead, one can envision acting preemptively by interfering with the cellular factors involved in adaptive mutability to restrain the emergence of resistance, rather than passively waiting for the tumor to evolve under therapy.

Although it remains unclear whether the stress-induced adaptive response also occurs in normal cells, inhibition of DNA damage response has been previously tested in patients with cancer with manageable toxicity (106). Moreover, combinatorial regimens might allow the administration of lower doses of single agents by exploiting synergism exerted by parallel inhibition of multiple survival pathways in cancer cells.

In conclusion, further exploration of the mechanisms through which bacteria and cancer cells survive and evolve under stress could provide new therapeutic options aimed at curbing adaptive mutability unleashed by drug pressures in cancer cells with altered genome functions, while limiting the side effects on normal cells.

A. Bardelli reports grants from AIRC during the conduct of the study; grants from Neophore and AstraZeneca, and personal fees from Inivata outside the submitted work; SAB at Horizon Discovery, Inivata, and Roche; and shareholding at Neophore. No disclosures were reported by the other authors.

This work was supported by FONDAZIONE AIRC under 5 per Mille 2018—ID. 21091 program—P.I. Alberto Bardelli; AIRC under IG 2018—ID. 21923 project—P.I. Alberto Bardelli; AIRC-CRUK-FC AECC Accelerator Award contract 22795 to A. Bardelli; IMI contract n. 101007937 PERSIST-SEQ to A. Bardelli; Ministero Salute, RC 2020 to A. Bardelli.