the evolution and ecology of resistance in cancer therapy

The Evolution and Ecology of Resistancein Cancer Therapy

Robert Gatenby and Joel Brown

Cancer Biology and Evolution Program, Moffitt Cancer Center, Tampa, Florida 33612

Correspondence: [email protected]

Despite continuous deployment of new treatment strategies and agents over many decades,most disseminated cancers remain fatal. Cancer cells, through their access to the vast infor-mation of human genome, have a remarkable capacity to deploy adaptive strategies for eventhe most effective treatments. We note there are two critical steps in the clinical manifestationof treatment resistance. The first, which is widely investigated, requires deployment of amechanism of resistance that usually involves increased expression of molecular machinerynecessary to eliminate the cytotoxic effect of treatment. However, the emergence of a resis-tant phenotype is not in itself clinically significant. That is, resistant cells affect patientoutcomes only when they form a sufficiently large population to allow tumor progressionand treatment failure. Importantly, proliferation of the resistant phenotype is by no meanscertain and, in fact, depends on complex Darwinian dynamics governed by the costs andbenefits of the resistance mechanisms in the context of the local environment and competingpopulations. Attempts to target molecular machinery of resistance have had little clinicalsuccess largely because of the diversity within the human genome—therapeutic interruptionof one mechanism simply results in its replacement by an alternative. We explore an alter-native strategy for overcoming treatment resistance that seeks to understand and exploit thecritical evolutionary dynamics that govern proliferation of the resistant phenotypes. Ingeneral, this approach has shown that, although emergence of resistance mechanisms incancer cells to every current therapy is inevitable, proliferation of the resistant phenotypes isnot and can be delayed and even prevented with sufficient understanding of the underlyingecoevolutionary dynamics.

The dynamic cancer ecosystem, rich in tem-poral and spatial diversity in environmental

conditions and heritable cell phenotypes, isextraordinarily robust to therapeutic perturba-tions. In part, this is caused by cellular diversity,as spatial heterogeneity in the genotypic andphenotypic properties of tumor cells inevitablyalter their response to applied treatments.

In part, this is also caused by temporal and spa-tial heterogeneity in the tumor environment,largely governed by variations in blood flow,which can alter both the delivery of a systemicagent and local concentrations of factors (e.g.,oxygen) that may alter the efficacy of the treat-ing agent. Finally, in part, the response andresistance of cancer cells is governed by their

Editors: Charles Swanton, Alberto Bardelli, Kornelia Polyak, Sohrab Shah, and Trevor A. Graham

Additional Perspectives on Cancer Evolution available at www.perspectivesinmedicine.org

Copyright # 2018 Cold Spring Harbor Laboratory Press; all rights reserved; doi: 10.1101/cshperspect.a033415

Cite this article as Cold Spring Harb Perspect Med 2018;8:a033415

1

ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED

on January 20, 2022 - Published by Cold Spring Harbor Laboratory Presshttp://perspectivesinmedicine.cshlp.org/Downloaded from

mailto:[email protected]



http://www.perspectivesinmedicine.org



http://www.perspectivesinmedicine.org/site/misc/terms.xhtml

http://perspectivesinmedicine.cshlp.org/

complex interactions with adjacent host cells,which can provide local sanctuaries that permittumor cells to survive treatment that wouldordinarily be lethal.

Despite the many advances in cancer therapyover the past few decades, evolution of resistancecontinues to limit their efficacy in reducingoverall survival. Even after a response to therapythat is clinically complete, residual cancer cellsadapt and the tumor almost inevitably returns.Although the evolution of resistance continuesto be the central cause of death in most cancerpatients, recent analysis found that evolutionaryterms were included in fewer than one percentof manuscripts on cancer treatment outcomes;this has remained unchanged for 30 years.

In general, there are two approaches to in-vestigating the dynamics of cancer cell resistanceto therapy. The most common avenue ofresearch focuses on identifying the specific mo-lecular mechanisms of resistance. This is per-haps best illustrated in the extensive literatureon membrane extrusion pumps—P-glycopro-teins (PgP), for example—that use ATP to ac-tively remove cytotoxic agents from the tumorcell cytoplasm and send them back into the en-vironment. This approach has the advantage ofidentifying molecular mechanisms that can betargeted and, thus defeated. However, despitedecades of research in well-recognized adaptivemechanism such as the multidrug resistance(MDR) genes, this approach has not led to clin-ically significant therapy. In large part, this is theresult of the remarkable diversity of adaptivestrategies available through the human genome.Thus, successful therapeutic intervention inone resistance mechanism simply selects for asecond available strategy.

Importantly, however, we note that theexpression of a resistance mechanism does notby any means insure that the resistant popula-tion will rapidly proliferate leading to tumorprogression. Thus, although the mechanismsof adaptation to the toxic effects of therapy aremolecular, the clinical relevance of resistance isentirely dependent on proliferation of resistantpopulations—a process ultimately governed byDarwinian dynamics based on the phenotypiccost and benefit of the molecular mechanisms

of resistance in the context of the local environ-ment and competing cellular populations.

Here, we review treatment strategies thatexplicitly incorporate evolutionary principles tolimit proliferation of resistant and prolong timeto progression. Nearly all conventional cancertherapies assume that the maximum patientbenefit is achieved through killing the greatestpossible number of tumor cells. Thus, cancertreatments are typically administered at themaximum tolerated dose (MTD). Indeed, thisprinciple is so universally accepted that phase 1trials—necessary for clinical translation of anycancer drug—seeks to define the MTD, whichis then used in subsequent investigations. Al-though killing the maximum number of tumorcells with the greatest possible drug dose is intu-itively appealing, we propose that it is usuallyevolutionarily unwise. This reflects the principlenoted above—evolution of resistant molecularmachinery is virtually impossible to evade butproliferation of resistant populations is depen-dent on complex Darwinian dynamics. In fact,by applying intense Darwinian selection for re-sistant clones and elimination of all competingpopulation, MTD cancer therapy actually “accel-erates” proliferation of resistant populations(also well-known as an evolutionary phenome-non termed “competitive release”) (Gatenby2009; Renton et al. 2014). Thus, evolutionaryresponses of cancer cells to therapy becomes theproximate cause of death in most cancer patientsand will most likely remain as such in want ofbasic changes in the tumor treatment paradigm.

The majority of conceptual models of tu-mor evolution during chemotherapy stress re-sistance acquired in a step-wise fashion throughsome mutation following the start of therapy(Anderson and Quaranta 2008). If resistancearises stochastically and before treatment thereare no resistant cells present, then maximumdose-density therapy will reduce the numberof cells that can acquire the mutation with theprobability of resistance lessening. However, inmost cases, the resistant cells or some level ofresistance appears before chemotherapy starts(Easwaran et al. 2014).

The assumption that a “resistance muta-tion” is necessary for tumor adaptation under-

R. Gatenby and J. Brown

2 Cite this article as Cold Spring Harb Perspect Med 2018;8:a033415

ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



estimates the vast information content, includ-ing xenobiotic pathways, that is available tocancer cells within the normal human genome(Gatenby 2009). That is, many, and probablymost, resistance strategies only require an in-creased expression of one or more normal genes(PgP; see Schneider et al. 1989; Fletcher et al.2010; Wind and Holen 2011). Therefore, ratherthan an “all or none” phenomenon, resistancecan be graded among cells in a population and,importantly, change in the same cell over timeas it acclimatizes to environmental stresses. Forexample, PgP is a hypoxia inducible factor(HIF)1a client (Thews et al. 2011) and itsexpression often increases in hypoxic and acidicenvironments even without cytotoxins. Inaddition, numerous mechanisms of de novotherapy resistance have been identified. Forexample, in environmentally mediated drugresistance (EMDR), components of tumor mes-enchyma protect cancer cells from what wouldotherwise be lethal concentrations of cytotoxicdrugs (Meads et al. 2009; Moulder 2010). Sim-ilarly, tumor cells in regions of hypoxia may beprotected as a result of increased expression ofPgP, up-regulation survival pathways, increasedmutagenesis, and decreased drug delivery.

EVOLUTIONARY DYNAMICS OF CANCERTHERAPY

Cancers can be described as open complexadaptive systems. “Open” because they freelycommunicate with their surroundings, “com-plex” because they contain multiple compo-nents, and “adaptive” because each elementcan change over time and interact with othercomponents in complicated, often nonlinear,ways. Based on their nonlinear dynamics, acritical attribute is that their dynamics canbe nonintuitive and their response to a pertur-bation can yield unexpected and unintendedconsequences.

Traditional application of systemic therapyto cancer is largely based on an intuitivelyappealing premise because by killing the maxi-mum number of cancer cells, patient benefit isachieved (Norton and Simon 1986; Rodriguesand de Arruda Mancera 2013). This assumption

that more is better is so deeply ingrained, thatthe initial clinical trials (i.e., phase 1 trials), al-ways seeks to establish the maximum tolerateddose for that drug. In other words, there is animplicit assumption that all cancer drugs mustbe administered at the maximum possible doseto find the MTD in the first phase. Generally,only out of concern for fatal toxicity was themaximum dose density of chemotherapy limit-ed. An alternate approach is the “metronomic”strategy (Gnoni et al. 2015), which administerslower more frequent doses of therapy. This hasbeen shown to be beneficial in lowering toxicity,while permitting higher total drug administra-tion, and inhibiting angiogenesis by increasingtumor cell death. However, the intent of mod-ern therapy remains inducing the greatest pos-sible cell death whether administered throughmaximum or metronomic dosing.

Significant flaws in conventional assump-tions emerge when cancer therapy is viewed asan evolutionary process (Axtell and Arends1990; Gatenby 2009; Gatenby et al. 2009a; Silvaet al. 2012; Renton et al. 2014; Jansen et al. 2015;Kam et al. 2015). To be clear, when it is possibleto have curative therapy, then the treatmentstrategy must be designed with that result inmind. However, in a palliative clinical setting(in which metastatic cancer patients nearlyalways die of their disease), treatment for cureis not only futile, it is, in fact, evolutionarilyunsound. In destroying the entire populationof sensitive cells, maximum dose therapy im-poses intense selection for resistant phenotypesand, by eliminating all potential competitors,their proliferation is maximized (a well-knownevolutionary phenomenon termed “competi-tive release”) (Axtell and Arends 1990; Rentonet al. 2014).

Interestingly, insight into these dynamicscan be found in an unlikely source—pest andweed management (Oliveira et al. 2007; Neveet al. 2009). For decades, the application ofhigh-dose pesticides was commonplace, butover time it became clear that this approachvirtually never eradicated the pest and, infact, promoted the rapid emergence of uncon-trollable, resistant strains. Since 1968, the pol-icy toward pest management by the U.S. De-

Evolution of Resistance in Cancer Therapy

Cite this article as Cold Spring Harb Perspect Med 2018;8:a033415 3

ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



partment of Agriculture (USDA) has beenmuch more nuanced placing greater emphasison the controlled application of pesticides. Thegoal is to prevent the emergence of pesticideresistance while maintaining a low and accept-able level of crop damage (Oliveira et al. 2007;Neve et al. 2009). Incorporation of temporaldata sampling and Darwinian dynamics intothe management of invasive species is nowmandated by USDA policy. To assist agricultur-alists in creating the best treatment strategies,computational models guide pest manage-ment, similar to those that we propose foradaptively managing advanced, incurable met-astatic cancers.

APPLICATION OF EVOLUTIONARYPRINCIPLES TO CANCER THERAPYRESISTANCE

Evolutionary therapy usually zeroes in on thecompetition for space and resources amongcancer populations, and in particular, Darwin-ian interactions between resistant and nonresis-tant populations. When a cancer therapy resultsin cell death, this imposes strong selection forcesfor adaptive strategies. The size and complexityof the human genome practically assuresthe presence of numerous potentially adaptivepathways in which to avoid cell death. Thus,although targeting combinations of pathwayscan control human immunodeficiency virus(HIV) with only nine protein-encoding genes,this strategy has not as yet proved successfulin preventing the emergence of resistance inhuman cells.

If resistant populations cannot be prevented,then tumor control requires therapy designed toslow or stop the proliferation of resistance.Here, two basic fundamentals must be empha-sized: first, the growth of the resistant popula-tion is subject to evolutionary forces and,therefore, can be controlled by altering itsfitness or that of competing populations.Second, evolving populations can only adaptto local and current environmental selectionforces; they can never anticipate the future.Importantly, because cancer therapists cananticipate the future and use the knowledge of

these temporal dynamics to a key advantage byusing evolution to inhibit the proliferation ofresistant populations, they can allow treatmentto change over time.

In evolutionary cancer treatment, a majorcomponent of Darwinian dynamics is thecost of resistance. To become resistant, typicallythrough up-regulation of established moleculardefense mechanisms, cancer cells must altertheir observable properties. Expression, mainte-nance, and utilization of these molecular path-ways require resources, which, in an environ-ment of limited substrate, must be divertedfrom proliferation and invasion. These dynam-ics are perhaps most clear in up-regulation ofxenobiotic pathways such as increased expres-sion of PgP, also known as multidrug resistance1 (MDR1) (Schneider et al. 1989; Fletcher et al.2010; Wind and Holen 2011). PgP is a mem-brane transporter that effectively extrudes alarge number of intracellular substrates, includ-ing chemotherapies and, consequently, mini-mizes the efficacy of these compounds. PgPand most other membrane pumps hydrolyzetwo ATP forevery transported molecule. Indeed,in experimental studies this operation cost (aswell as “capital cost” for synthesis and mainte-nance of the pumps) can approach 50% of thecell’s energy budget (Silva et al. 2012). In cellculture conditions with abundant resources,this may have minimum effect on cellular pro-liferation. However, under resource limitation(e.g., in vivo), cancer cells must trade-offresistance costs that permit survival with otheressential functions such as proliferation, move-ment, and cell maintenance. This fitness cost canbe inferred by the simple observation that theMDR phenotype is rare in pretherapy tumorsand becomes common only following treatment(Wind and Holen 2011). Furthermore, the drugresistant phenotype is typically evolved throughregular exposure to a cytotoxic drug and quicklylost when the drug is removed (Kam et al. 2015).

Thus, in the presence of a cytotoxic drug,survival benefits of resistance exceed the fixedand operating costs of the resistance mecha-nisms. But, in the absence of therapy, the costof resistance becomes an ecological burden thatcan be exploited (Silva et al. 2015).



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



The ecological cost of resistance is mostapparent in traditional chemotherapy in whichthe mechanisms of resistance and their associ-ated “construction and operation” costs arefairly clear. This is not as apparent in targetedtherapies. However, Chmielecki et al. (2011)have shown a cost of resistance to tyrosinekinase inhibitors (TKIs).

EXPLOITING THE COST OF RESISTANCE

Theodosius Dobzhansky famously stated “Noth-ing in biology makes sense except in the light ofevolution” (Ayala and Dobzhansky 1977). Yet,the typical cancer therapy is administered with-out regard to the evolutionary consequences.Drugs, doses, and schedules follow a fixed pro-tocol that changes only in the event of unaccept-able toxicity or unequivocal evidence of cancerprogression.

There are three crucial elements that makecancers and their treatment highly dynamic.First, the distribution and abundance of cancerclones. These can be viewed as coexisting cyto-types, with perhaps distinctive niches (Lloydet al. 2016) that can change rapidly. Second,through feedbacks between cancer cell activities,the host and intratumoral properties, the tumorecosystem can rapidly change in size, character,and spatial heterogeneity. Third, the heritabletraits of some or all of the cancer cell typesmay change directly in response to therapy orin response to changes in the tumor environ-ment. The first two dynamics describe theecological dynamics of cancers and the tumorenvironment; the third describes the evolution-ary dynamics of changing frequencies andheritable phenotypes of the cancer cytotypesthemselves. For these reasons, cancers arehighly dynamic systems with tremendous spa-tial and temporal heterogeneity that can rap-idly change with perturbations such as appliedtherapy. Almost certainly, the tumor and itsresident cancer-cell populations treated inthe second cycle of chemotherapy are quitedifferent that those treated in the first cycle.Indeed, any tumor that persists or recurs aftertherapy will differ radically from the tumor atdiagnosis. The persistent or recurrent cancer

cells will possess new phenotypes and geno-typic profiles.

We propose that cancer therapy becomeas dynamic as the tumor system being treated.Ideally, treatment strategies should anticipate,corral, and stay ahead of the intratumoral evo-lution through strategic application of differentdrugs and drug doses that move beyond thetraditional goal of maximal cell death. Maxi-mum control of the tumor becomes the goal,in which the objective can include quality andlongevity of life, cumulative dosing of therapies(radiation, chemo-agents, and/or immuno-therapy), and the patient’s sense of well-beingboth during and in the absence of therapy.

ADAPTIVE THERAPY

One such approach is adaptive therapy (AT)(Gatenby 2009; Gatenby et al. 2009a). Thepremise of AT is similar to current pest manage-ment in that the time to progression of extantanticancer drugs can be revised if their admin-istration is guided by Darwinian principles.This approach has a number of features thatdiffer from conventional chemotherapy strate-gies. First, the goal of AT is maximizing progres-sion-free survival and not abatement in cancerburden. Second, the amount of drug adminis-tered is the minimum necessary to maintainpatient quality of life and tumor stability. Third,the drugs and dose schedules are not fixedbut instead constantly adjusted to steer andexploit evolutionary dynamics and maintain astable tumor.

Although there are a number of Darwiniandynamics that may be exploited during cancertherapy, the theory behind AT focuses onthe phenotypic costs of the molecular mecha-nism(s) of resistance. Extant or potential cancerphenotypes within the ecological context of thetumor reside on an adaptive landscape describ-ing fitness as a function of a cancer cell’sphenotype. It is dependent on the fitness ofthat phenotype compared with that of otherexistent populations. Note that the fitness ofany phenotype depends on context. A pheno-type that, for example, expressing the PgPmembrane pump is fitter than nonexpressing



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



cells in the presence of chemotherapy. However,in the absence of therapy, owing to the cost ofthe resistance mechanism, the fitness of the re-sistant cells will be lower than that of sensitivecells (Fig. 1). In this case, the success of a cancercell depends on its ability to compete success-fully with other cancer cells within the contextof the presence or absence of therapy. To exploitthis, AT administers limited therapy with theexplicit goal of maintaining a stable populationof treatment-sensitive cells (Fig. 2). Therapyis reduced or withheld once the tumor size isstabilized. Although this may allow for sometumor cell proliferation, at the expense of theresistant cells, the fitness advantage of sensitivecells will allow their population to grow. Repeat-ed administrations of small doses of drugs arethen used to reduce the tumor volume. The goalis to administer “much less than maximum”dose while shrinking or maintaining at anacceptable level the overall tumor burden andtotal population size of cancer cells.

The success of AT is dependent on the costof resistance and the competitive interactionbetween sensitive and resistant cancer cells. ATcan be expected to be superior to maximumdose if three conditions are met. First, the ther-apy is highly effective against sensitive cancer

cells. Second, sensitive cells outcompete resis-tant cells in the absence of therapy. Third,sensitive cells in the absence of therapy canproliferate faster than resistant cells can in thepresence of therapy. This leads to either persis-tent or declining cycles of total tumor burden.Initial therapy rapidly drops the population sizeof sensitive cells. This results in the evolutionand increase of resistant populations of cancercells. Therapy is then stopped before the resis-tant cells completely replacing the sensitivecells. At this point, the resistant cells by virtueof their faster proliferation rates and greatercompetitive ability will both increase and sup-press the resistant cells. Before the sensitive cellsreturning to threatening population sizes, ther-apy must be resumed. This sequence representsa complete cycle. Such cycling works because ofthe therapy can decisively crash the sensitivecells, and the recovery of the sensitive cellsprovide the therapy for the resistant cells. Inpractice, effective AT will require tailoring thetherapy protocol to the nature of the interven-tions, the characteristics of the particularcancer, the nature and cost of resistance, usefulmetrics of tumor burden and composition, andassociated mathematical model to integrate tu-mor metrics to determine and manage optimaltiming of therapy dosing.

The AT hypothesis was initially framed us-ing catastrophe theory (Gatenby and Frieden2008). For any drug or drug combination, itcan be shown that the probability of eradicatingall of the tumor cells is maximized when phe-notypes and environments are homogeneous.In this mathematical model, intratumoral het-erogeneity results in phase differences betweentherapy and tumor that permits survival. Thebiological interpretation of “phase difference” istherapy that is unsuccessful because of the pre-treatment presence of resistant phenotypes orenvironments that are relatively sheltered fromthe toxic effects of therapy and permit rapidevolution and proliferation of cancer cells withresistant strategies. It is not unsuccessful be-cause of the evolution of resistance during ther-apy. Showing the feasibility of this theoreticalmodel, a preclinical model of ovarian cancer(OVCAR-3 cells) was developed and treated

1.0

0.8

0.6

0.4

0.2

0.00 10 30 40 5020

Time (days)

Pop

ulat

ion

frac

tion

SensitiveResistant

Figure 1. Illustrating the evolutionary consequence ofthe cost of resistance. To mimic growth dynamics ofresistant and sensitive cell populations, labeled MCF7(sensitive) and MCF7/Dox (resistant) cells were co-cultured in the absence or doxorubicin with physio-logical levels of glucose. The phenotypic cost of resis-tance decreased fitness of the resistant cells andallowed the sensitive population to proliferate at theexpense of the resistant phenotype.



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



with carboplatin. Control, standard therapy(60 mg/kg q4 days x 3), and AT groups wereestablished. The AT algorithm was based solelyon tumor volume. Following the initial carbo-platin dose of 50 mg/kg, tumor volumes weremeasured every three days. Subsequent doses ofcarboplatin were then adjusted to maintainstable tumor volume. For example, if the tumorvolume increased in size for two consecutivemeasurements, the dose of carboplatin wouldbe increased. Similarly, doses would be de-creased when tumor size declined. Comparedwith control and standard therapy, this ATresulted in prolonged tumor control and im-proved survival of mice (Gatenby et al. 2009a).

In these and in more recent studies usingbreast cancer cell lines, an unexpected observa-tion is a biphasic pattern in tumor response toATwas noted. That is, when treatment is initiallyapplied, the tumors are typically growing expo-nentially. Forcing the growth curve to plateaurequired the full treatment dose. However, oncethe tumor volume stabilized the amount ofdrug necessary to maintain stability diminished

rapid. In the above experiment, prolongedtumor control was often maintained using5 mg/kg carboplatin. Of note, the computa-tional models did not predict this phenomenonand it remains under investigation, but may becaused by tumor vascularity. Over time, itappeared that the tumor’s blood system equili-brated on a more normal vascularity than thegreatly dysregulated angiogenesis seen in the tu-mors of the control and standard of care mice.This observation points to the need to modeland appreciate all three dynamics: cancer cellpopulation size and evolution, tumor size andheterogeneity, and cancer cell–tumor ecosys-tem feedbacks.

INCREASING THE COST OF RESISTANCE—ERSATZDROGES

As mentioned previously, membrane extrusionpumps are a common mechanism of cancercell resistance to chemotherapy. Targeting thesepumps with treatments that reduces their effec-tiveness has been explored extensively but has

Maximum dosedensity

treatmentRepeat

treatment

TreatmentTreatment

Sensitive

Sensitive cell

Resistant cell

Repeattreatment

Figure 2. Conventional, high dose-density treatment compared to an evolution-based strategy. (Top row)Conventional high dose-density therapy explicitly aims to eliminate all cancer cells that are sensitive to therapy.However, this maximally selects for resistant phenotypes and eliminates competitors permitting rapid progres-sion—an evolutionary dynamic termed “competitive release.” (Bottom row) Adaptive therapy (AT) explicitlyaims to maintain a small population of cells that are sensitive to treatment. Although the resistant cells survive,the metabolic cost of the molecular machinery of resistance (Fig. 1) renders them less fit in the absence oftherapy. Thus, when therapy is withdrawn, the tumor will regrow but the fitness advantage of the sensitive cellsallows them to proliferate at the expense of the resistant population. At the end of each cycle, the tumor remainssensitive to therapy.



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



not resulted in anyclinicallyeffective strategiestoprolong treatment response (Gatenby et al.2009b). As noted above, an alternative approachexploits the evolutionary consequences of themetabolic capital and operation costs for main-tenance of the pumps. For example, Broxtermanet al. (1988) showed that the extrusion activity ofPgP pumps activated by verapamil could con-sume cell’s ATP budget by 50%. Importantly, inthe intratumoral environment of limited re-sources, this increased ATP demand requiresdiversion of substrate from other activities in-cluding proliferation and invasion. Thus, analternative treatment strategy to blocking pumpfunction, seeks to exhaust the cancer cell’s re-sources by maximizing pump activity throughadministration of nontoxic (or minimally toxic)substrate (Kam et al. 2015). Thus, by forcing cellsto expend energy to extrude a “fake drug” (hencethe name ersatzdroges) in the absence of chemo-therapy, this strategy increases the phenotypiccost of the cells’ resistance strategy.

Currently available ersatzdroges, includingantibiotics and verapamil (a calcium channelblocker prescribed to treat arrhythmia), areused in the treatment of other diseases. Thus,when the tumor cell detects the presenceof these “fake” drugs, it uses its resistant mech-anisms such as the MDR1 system to pump thedrug from the cytosol.

Exposure to verapamil in a preclinicalmodel of breast cancer dramatically altered theenergy dynamics with cancer cells and reducedboth proliferation and invasion. In vivo and invitro, administration of various ersatzdrogessignificantly increased glucose flux in tumorsand reduced tumor growth (Kam et al. 2015).The use of the esatzdroges as chaff exploits eco-logical and evolutionary dynamics. The resis-tant cells waste energy bailing the ersatzdroges.The cell becomes less competitive relative toother cells with fewer pumps. And, evolutionshould now disfavor the membrane pumpsor select for fewer pumps thus maintainingoriginal drug efficacy. This represents a formof evolutionary therapy known as a “doublebind therapy” (Gatenby et al. 2009b). A doublebind therapy uses the cancer’s adaptive responseto one therapy (in this case a cytotoxin) to make

another more effective (forced bailing of theersatzdroges) and vice-versa.

TURNING THE TABLES: TARGETINGTHE ADAPTIVE STRATEGIES

In general, the application of evolutionarystrategies to optimize tumor therapy requires“temporal” thinking. As shown in AT, the ther-apist must look beyond the immediate effectsof treatment (i.e., tumor cell reduction). It re-quires anticipating and exploiting the longer-term ecological and evolutionary dynamics asnew phenotypic properties arise and increaseamong the cancer cells. Thinking ahead mayidentify the opportunities for double bindtherapies (Fig. 1) (Gatenby et al. 2009b) or“sucker’s gambits” (Maley et al. 2004). Howoften can one find a first line treatment to in-duce a phenotypic adaptation that is then morehighly susceptible to an especially chosen sec-ond line therapy?

A double-bind strategy was successfullyapplied to antibiotic treatment of Heliobacterpylori (Fuentes-Hernandez et al. 2015). In can-cer treatment, Antonia et al. (2006) examinedthe efficacy of a p53 vaccine in patients (n ¼ 29)with small cell lung cancer. Most patientselicited immune responses but only one partialresponse was observed. However, when the pa-tients subsequently underwent chemotherapy, aresponse rate of 67% was observed (comparedwith the historic response rate of ,5%); effica-cy increased in those patients that had the great-est immune activation (Chiappori et al. 2010).We interpret this as an example of an evolution-ary double bind in which the tumor cells’ adap-tive strategy to the immunotherapy renderedthem more vulnerable to cytotoxic drugs.

Finally, the ideal double bind therapycreates an evolutionarily futile cycle in whichcyclical application of the treatments matchesprecisely the pattern of evolution of resistancein the underlying cancer populations. In ourclinical example, it is possible that this evolu-tionary cycle could have been completed byrevaccinating the patients after chemotherapy,which may have selected for increased p53expression.



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



EXPLOITING PROPERTIES OF COMPLEXDYNAMIC SYSTEMS FOR CANCERCONTROL

All current cancer therapies act as biocides withvarious mechanisms for killing cancer cells.This is certainly reasonable but also inevitablyproduces Darwinian forces that promote resis-tance. An alternative approach targets environ-mental selection forces to alter the underlyingevolutionary dynamics with the explicit goal ofpromoting a less proliferative or invasive tumorphenotype. In general, cancers can be viewedas open complex dynamical systems: “complex”because the tumor has many components, “dy-namic” because the components interact andchange over time (often nonlinearly), and“open” because it must freely interact with thehost (for instance, all cancer cells derive theirnutrients from the host). Traditional analysisof such systems emphasizes the challenge inpredicting outcomes because of their nonlineardynamics and sensitivity to initial conditionsand/or slight perturbations of current condi-tions (weather is probably the most familiarcomplex dynamic system). For example, thefamous “butterfly effect” posits an insect flap-ping its wings in Asia can cause a tornado inNorth America.

Importantly, however, it has been notedthat this established pattern of complex systemsto magnify some small perturbations can beexploited to steer the system along a desiredcourse with minimal application of force(Wang et al. 2012). This, of course, requires amodest understanding of the underlying intra-tumoral dynamics, but it does not require afull understanding. One has a sufficient levelof understanding when interventions suggestthemselves that perturb the tumor environmentin ways that dissipate the cancer system or atleast coax it along a less clinically aggressivepath. What small, yet decisive, biological forcesare available? Perturbing pH may prove useful.Most tumors are net-producers of acid andintratumoral acidosis has been shown to selectfor phenotypes that are highly motile and inva-sive. However, small perturbations of the extra-cellular pH (an increase of �0.2 pH units) can

alter these Darwinian dynamics and select forless aggressive, more indolent phenotypes(Ibrahim-Hashim et al. 2012).

CLINICAL APPLICATIONS

Mastering complex systems (as in weather fore-casting) (Wang et al. 2012) require three compo-nents: (1) defining and mathematically framingfirst principles; (2) necessary and sufficient datato parameterize mathematical models; and (3)sophisticated computational methods.

We propose that evolutionary and ecologi-cal dynamics serve as first principles in cancertherapy and that computational models canbe readily constructed (e.g., computer modelsfor pest management are widely available). Theabsence of usable data is the greatest stumblingblock to the clinical application of evolutionaryprinciples to cancer therapy. Although concernsfor dealing with “big data” from molecularanalysis are often expressed in oncology, thesedata lack spatial and temporal resolution andtherefore have limited value in evolutionarymodels. In fact, good time-series data in oncol-ogy is largely limited to serum markers andrepeated cross-sectional imaging. There is anurgent need to develop methods that can usesparse data for computational models andextract more information from available clinicaldata (e.g., radiomic analysis of computed to-mography [CT] and magnetic resonance imag-ing [MRI] images [Gatenby et al. 2013]).

Limitations withstanding, some clinicaltrials that illustrate successful application ofevolutionary dynamics have been and are beingperformed (Schweizer et al. 2015). We suggestthat the applications of evolutionary conceptsand therapies can allow us make quantumstrides by simply changing our use of existingtherapies, patient and tumor metrics, and dataanalysis and modeling. Such results may bepossible at a fraction of the cost of novel drugdiscovery and development.

SUMMARY AND FUTURE DIRECTIONS

Cancers are complex, dynamic systems that be-gin to evolve resistance strategies immediately



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



on the application of any therapy. Currentcancer treatment is typically applied in a staticfashion: the same drugs, doses, and schedulesare administered until the protocol ends or thetumor progresses. When the cancer cells of atumor respond to treatment, the typical strategyis to simply give more of the same. In contrast,evolution-based therapy seeks to become asadaptive and flexible as the tumor populationsbeing treated. This will require a new para-digm in cancer therapy in which Darwiniandynamics are explicitly incorporated into trialdesign and execution (Fig. 3). In AT, for ex-ample, once the tumor response to a specifictreatment is observed, the best course might beto switch to a new strategy or withdraw therapybecause ongoing treatment with the successfuldrug will only result in greater selection forresistance.

In preclinical experiments (Gatenby et al.2009a; Silva et al. 2012; Kam et al. 2015), wehave shown that applying evolutionary princi-ples to conventional chemotherapy agents can

substantially prolong progression-free survivalin both breast and ovarian cancer. Schweizeret al. (2015) recently showed that evolutionaryprinciples could be used to prolong response toanti-androgen therapy in a cohort of men withcastrate-resistant prostate cancer. A clinical trialusing an adaptive-therapy algorithm for abira-terone therapy in men with castrate-resistantprostate cancer has recently opened and appearssuperior to continuous therapy.

There are difficulties in clinical application.These include the requirement to collect reliabletime-series data allowing the internal evolu-tionary dynamics of the cancer to be observed,measured, and estimated. In fact, repeated mea-sures of changes in tumors over time and spaceare typically quite sparse, largely limited toserum markers and clinical imaging. Future di-rections must focus on converting clinical datainto a dynamical understanding of the complexenvironmental and phenotypic changes thatdrives tumor evolution in response to therapy.Ultimately, this understanding will likely re-

Molecular, Cellular,and ImagingDiagnostics

Targetedtherapy Chemotherapy Immunotherapy

EcoevolutionaryPrinciples

Outcomes

Figure 3. Current “personalized medicine” paradigms almost exclusive focus on defining predictive biomarkersthat can identify effective treatments. This approach, however, fails to recognize that even highly effectivetherapies are almost always defeated by evolution of resistance. Here, we propose that “precision medicine”in cancer care requires both identification of optimal treatment modality and understanding of the Darwiniandynamics that govern response and resistance to therapy and, thus, ultimate patient outcomes.



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



quire sophisticated patient-specific computa-tional models to provide treating physicianswith decision support tools to optimize cancertherapy.

ACKNOWLEDGMENTS

This work is supported by National Institutesof Health (NIH) Grants U54CA143970-1 andRO1CA170595 and a grant from the JamesS. McDonnell Foundation.

REFERENCES

Anderson AR, Quaranta V. 2008. Integrative mathematicaloncology. Nat Rev Cancer 8: 227–234.

Antonia SJ, Mirza N, Fricke I, Chiappori A, Thompson P,Williams N, Bepler G, Simon G, Janssen W, et al. 2006.Combination of p53 cancer vaccine with chemotherapyin patients with extensive stage small cell lung cancer.Clin Cancer Res 12: 878–887.

Axtell RC, Arends JJ. 1990. Ecology and managementof arthropod pests of poultry. Annu Rev Entomol 35:101–126.

Ayala FJ, Dobzhansky T. 1977. “Nothing in biology makessense except in the light of evolution”: Theodosius Dob-zhansky: 1900–1975. J Hered 68: 3–10.

Broxterman HJ, Pinedo HM, Kuiper CM, Kaptein LC,Schuurhuis GJ, Lankelma J. 1988. Induction by verapamilof a rapid increase in ATP consumption in multidrug-resistant tumor cells. FASEB J 2: 2278–2282.

Chiappori AA, Soliman H, Janssen WE, Antonia SJ,Gabrilovich DI. 2010. INGN-225: A dendritic cell-basedp53 vaccine (Ad.p53-DC) in small cell lung cancer:Observed association between immune response and en-hanced chemotherapy effect. Expert Opin Biol Ther 10:983–991.

Chmielecki J, Foo J, Oxnard GR, Hutchinson K, Ohashi K,Somwar R, Wang L, Amato KR, Arcila M, Sos ML, et al.2011. Optimization of dosing for EGFR-mutant non-small cell lung cancer with evolutionary cancer model-ing. Sci Transl Med 3: 90ra59.

Easwaran H, Tsai HC, Baylin SB. 2014. Cancer epigenetics:Tumor heterogeneity, plasticity of stem-like states, anddrug resistance. Mol Cell 54: 716–727.

Fletcher JI, Haber M, Henderson MJ, Norris MD. 2010. ABCtransporters in cancer: More than just drug efflux pumps.Nat Rev Cancer 10: 147–156.

Fuentes-Hernandez A, Plucain J, Gori F, Pena-Miller R, Red-ing C, Jansen G, Schulenburg H, Gudelj I, Beardmore R.2015. Using a sequential regimen to eliminate bacteria atsublethal antibiotic dosages. PLoS Biol 13: e1002104.

Gatenby RA. 2009. A change of strategy in the war on cancer.Nature 459: 508–509.

Gatenby RA, Frieden BR. 2008. Inducing catastrophe inmalignant growth. Math Med Biol 25: 267–283.

Gatenby RA, Silva AS, Gillies RJ, Frieden BR. 2009a. Adap-tive therapy. Cancer Res 69: 4894–4903.

Gatenby RA, Brown J, Vincent T. 2009b. Lessons fromapplied ecology: Cancer control using an evolutionarydouble bind. Cancer Res 69: 7499–7502.

Gatenby RA, Grove O, Gillies RJ. 2013. Quantitative imagingin cancer evolution and ecology. Radiology 269: 8–15.

Gnoni A, Silvestris N, Licchetta A, Santini D, Scartozzi M, RiaR, Pisconti S, Petrelli F, Vacca A, Lorusso V. 2015.Metronomic chemotherapy from rationale to clinicalstudies: A dream or reality? Crit Rev Oncol Hematol 95:46–61.

Ibrahim-Hashim A, Cornnell HH, Abrahams D, Lloyd M,Bui M, Gillies RJ, Gatenby RA. 2012. Systemic buffersinhibit carcinogenesis in TRAMP mice. J Urol 188:624–631.

Jansen G, Gatenby R, Aktipis CA. 2015. Opinion: Controlvs. eradication: Applying infectious disease treatmentstrategies to cancer. Proc Natl Acad Sci 112: 937–938.

KamY,DasT,TianH,ForoutanP,RuizE,MartinezG,MintonS, Gillies RJ, Gatenby RA. 2015. Sweat but no gain: Inhib-iting proliferation of multidrug resistant cancer cells with“ersatzdroges.” Int J Cancer 136: E188–E196.

Lloyd MC, Cunningham JJ, Bui MM, Gillies RJ, Brown JS,Gatenby RA. 2016. Darwinian dynamics of intratumoralheterogeneity: Not solely random mutations but alsovariable environmental selection forces. Cancer Res 76:3136–3144.

Maley CC, Reid BJ, Forrest S. 2004. Cancer prevention strat-egies that address the evolutionary dynamics of neoplas-tic cells: Simulating benign cell boosters and selection forchemosensitivity. Cancer Epidemiol Biomarkers Prev 13:1375–1384.

Meads MB, Gatenby RA, Dalton WS. 2009. Environment-mediated drug resistance: A major contributor to mini-mal residual disease. Nat Rev Cancer 9: 665–674.

Moulder S. 2010. Intrinsic resistance to chemotherapy inbreast cancer. Womens Health 6: 821–830.

Neve P, Vila-Aiub M, Roux F. 2009. Evolutionary-thinkingin agricultural weed management. New Phytologist 184:783–793.

Norton L, Simon R. 1986. The Norton-Simon hypothesisrevisited. Cancer Treat Rep 70: 163–169.

Oliveira EE, Guedes RNC, Totola MR, De Marco P Jr. 2007.Competition between insecticide-susceptible and -resis-tant populations of the maize weevil, Sitophilus zeamais.Chemosphere 69: 17–24.

Renton M, Busi R, Neve P, Thornby D, Vila-Aiub M. 2014.Herbicide resistance modelling: Past, present and future.Pest Manag Sci 70: 1394–1404.

Rodrigues DS, de Arruda Mancera PF. 2013. Mathematicalanalysis and simulations involving chemotherapy andsurgery on large human tumours under a suitable cell-kill functional response. Math Biosci Eng 10: 221–234.

Schneider J, Bak M, Efferth T, Kaufmann M, Mattern J, VolmM. 1989. P-glycoprotein expression in treated and un-treated human breast cancer. Br J Cancer 60: 815–818.

Schweizer MT, Antonarakis ES, Wang H, Ajiboye AS, SpitzA, Cao H, Luo J, Haffner MC, Yegnasubramanian S, Car-ducci MA, et al. 2015. Effect of bipolar androgen therapyfor asymptomatic men with castration-resistant prostate



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



cancer: Results from a pilot clinical study. Sci Transl Med7: 269ra2.

Silva AS, Kam Y, Khin ZP, Minton SE, Gillies RJ, GatenbyRA. 2012. Evolutionary approaches to prolong progres-sion-free survival in breast cancer. Cancer Res 72: 6362–6370.

Silva R, Vilas-Boas V, Carmo H, Dinis-Oliveira RJ, CarvalhoF, de Lourdes Bastos M, Remiao F. 2015. Modulation of P-glycoprotein efflux pump: Induction and activation as atherapeutic strategy. Pharmacol Ther 149: 1–123.

Thews O, Nowak M, Sauvant C, Gekle M. 2011. Hypoxia-induced extracellular acidosis increases p-glycoproteinactivity and chemoresistance in tumors in vivo via p38signaling pathway. Adv Exp Med Biol 701: 115–122.

Wang WX, Ni X, Lai YC, Grebogi C. 2012. Optimizing con-trollability of complex networks by minimum structuralperturbations. Phys Rev E 85: 026115.

Wind NS, Holen I. 2011. Multidrug resistance in breastcancer: From in vitro models to clinical studies. Int JBreast Cancer 2011: 967419.



ww

w.p

ersp

ecti

vesi

nm

edic

ine.

org

RETRACTED



14, 20172018; doi: 10.1101/cshperspect.a033415 originally published online JulyCold Spring Harb Perspect Med

Robert Gatenby and Joel Brown The Evolution and Ecology of Resistance in Cancer Therapy

Subject Collection Cancer Evolution

Cancer TherapyThe Evolution and Ecology of Resistance in

Robert A. Gatenby and Joel S. BrownCancer TherapyThe Evolution and Ecology of Resistance in

Robert Gatenby and Joel BrownBig Bang Tumor Growth and Clonal Evolution

Ruping Sun, Zheng Hu and Christina Curtis HeterogeneityPhylogenetic Quantification of Intratumor

Thomas B.K. Watkins and Roland F. Schwarz

Fitness Models for CancerSpatiotemporal Axes: A Prelude to Quantitative Observing Clonal Dynamics across

Sohrab P. ShahAndrew W. McPherson, Fong Chun Chan and

Strategiesfor the Development of Novel Immunotherapeutic The ''Achilles' Heel'' of Cancer and Its Implications

al.Kroopa Joshi, Benjamin M. Chain, Karl S. Peggs, et

Evolution of Premalignant Disease

GrahamKit Curtius, Nicholas A. Wright and Trevor A. Perspective of Cancer

Homeostasis Back and Forth: An Ecoevolutionary

David Basanta and Alexander R.A. AndersonThe Role of Aneuploidy in Cancer Evolution

Laurent Sansregret and Charles Swanton Architecture of CancersPrinciples of Reconstructing the Subclonal

LooStefan C. Dentro, David C. Wedge and Peter Van

Pressures Sculpt Cancer EvolutionTreatment-Induced Mutagenesis and Selective

S. Taylor, et al.Subramanian Venkatesan, Charles Swanton, Barry

Responses to TherapyTumor Microenvironment and Differential

Eishu Hirata and Erik Sahai

Heterogeneity and EvolutionChromosomal Instability as a Driver of Tumor

Samuel F. Bakhoum and Dan Avi LandauInfluences Cancer EvolutionOrder Matters: The Order of Somatic Mutations

David G. Kent and Anthony R. Green

in Chronic Lymphocytic LeukemiaCoevolution of Leukemia and Host Immune Cells

Noelia Purroy and Catherine J. WuCancerThe Cellular Origin and Evolution of Breast

Mei Zhang, Adrian V. Lee and Jeffrey M. Rosen

http://perspectivesinmedicine.cshlp.org/cgi/collection/ For additional articles in this collection, see

Copyright © 2018 Cold Spring Harbor Laboratory Press; all rights reserved


http://perspectivesinmedicine.cshlp.org/cgi/collection/

http://perspectivesinmedicine.cshlp.org/cgi/collection/


the evolution and ecology of resistance in cancer therapy

Documents