Therapeutic resistance has been shown to result in poorer clinical outcomes in cancer treatment. It has been proposed that evolutionary adaptations of cancer cells to therapy result in the development of resistance with the rate of adaptive change correlating with the heterogeneity of the tumour. These concepts can help overcome therapeutic resistance and have been exploited by Gatenby and others in promising evolutionary double-bind simulations. It was further suggested that tumour vasculature contributes to intra-tumoural heterogeneity through the development of substrate gradients. Increasing analogy between natural ecosystems such as riparian habitats and the tumour environment may allow us to devise novel treatment strategies. This review will briefly examine some of these evolutionary and ecological concepts and how they can be applied to cancer treatment.

Introduction

Carcinogenesis is the process by which normal cells in the body acquire mutations and form tumours. In the 1970s, Peter Nowell characterized this transformation in terms of evolutionary change and this concept has been well accepted by the scientific community. [1] He proposed that genetic instability and mutations form the basis for heritable changes required for natural selection and clonal growth of single cancer cells. Cells are selected for desirable characteristics such as survival and proliferation in response to changes in their immediate environment. [1] Surprisingly, evolutionary principles have seldom been used in the treatment of cancer. Aktipis and colleagues did an analysis of over 6000 papers focusing on therapeutic resistance and cancer relapse and revealed that ‘evolution’ has been used in only 1% of all papers. [2]

As evolution is influenced by changes in the environment, it is possible to view the tumour microenvironment as an ecosystem consisting of heterogeneous populations of cancer cells interacting with one another, and with other cells of the microenvironment. These complex interactions have much in common with ecosystems in nature and consist of analogous abiotic and biotic components which provide novel treatment targets to circumvent therapeutic failure.

Failure of chemotherapy can be attributed to cancer resistance which can be inherent or acquired. Inherent resistance may occur due to over-expression of drug metabolism pathways such as the excision repair cross-complementing 1 gene (ERCC1 — a nucleoside excision repair gene) in resistance against platinum agents while acquired resistance can be caused by altered membrane transport as in the case of the P-glycoprotein transport protein encoded by the multi-drug resistance-1 gene (MDR-1). [3]

Evolutionary game theory

Hypoxia and acidosis within the tumour can exert selective pressures on individual cancer cell populations. These populations may adapt to these conditions through different phenotypic strategies arising from genetic instability and genotypic variations. Gilles and colleagues proposed that these interactions can be understood through the evolutionary game theory. In this theory, the evolutionary rate of a phenotypic strategy is dependent on the amount of phenotypic diversity and the fitness of cancer cell populations. [4] Cancer cell populations will evolve rapidly in the presence of a harsh tumour environment or when cell populations are phenotypically diverse. Selective pressures originating from perturbations outside the tumour microenvironment can also promote further phenotypic diversity. [4,5] Alteration of the tumor environment by chemotherapy can potentially encourage cancer cell populations to diversify and become heterogeneous via de novo mutations arising from therapy or selection of existing chemotherapy-resistant cells in the tumour. [5]

The evolutionary game theory therefore predicts that the probability of the existence and/or emergence of resistant cells correlates with the level of tumour heterogeneity. It also suggests that chemotherapy will inadvertently lead to resistance if chemo-resistant cells (such as cancer stem cells) are already present in the tumour. [4,5] These predictions appear to correlate with clinical findings as advanced cancers which are less responsive to therapies usually exhibit high levels of heterogeneity while the use of high-dose chemotherapy improves survival but seldom cures epithelial cancers. [6]

High-dose chemotherapy regimens were first conceptualized mathematically through the Norton-Simon model. It is hypothesised that administering the maximum tolerated dose (MTD) over a short time period would achieve a high cancer cell kill rate and a low probability of therapy-induced evolution of resistant clones. [7] This model, however, does not account for pre-existing chemo-resistant cells which clonally proliferate and result in cancer relapse after initial treatment. By recognising that resistant cells potentially pre-exist in tumours and that they correlate positively with tumour heterogeneity, certain strategies can be devised. These include controlling the heterogeneity of the tumour to prevent the occurrence of chemo-resistance and, exploiting our ability to predictably alter the adaptive strategies of cancer cells through various treatment modalities.

Controlling tumour heterogeneity: induction of evolutionary bottlenecks and achieving an evolutionary ‘double-bind’

Intra-tumoural heterogeneity is minimal in early neoplasms and the use of low-dose chemotherapy may be sufficient to eliminate early cancers with less risk of resistance. [7] This formed the basis of metronomic chemotherapy where low doses of chemo-drugs were given in frequent intervals. [8] However, intricate strategies involving circumvention of therapeutic resistance would be required as a cancer progresses.

Resistant cells favour tumour progression in a treatment setting but many forms of resistance incur phenotypic costs. If the phenotypic cost is low, for example, due to the ability of the cancer cell to adapt to therapy through up-regulation of xenobiotic mechanisms or usage of a redundant signaling pathway, control of cancer cell proliferation will be less effective. [9] Conversely, if the phenotypic cost is high, for example, due to competition from co-existing cancer cell populations with different proliferative characteristics and biological therapies, robust and long-lasting control may be achieved because cancer cells can only survive by diverting resources away from proliferation. The latter creates an evolutionary double-bind where the only way tumour cells can evade the deleterious effects of treatment is by compromising its fitness attributes, thereby inhibiting its proliferation or ability to develop resistance. [9]

An evolutionary double-bind in a combination therapy setting would require anticipating the adaptation of cancer cell populations to a specific treatment and then targeting the adapted phenotype by a follow-up treatment. [4] In a study by Hunter et al., treatment of glioblastoma multiforme tumours with the alkylating agent temozolomide (TMZ) resulted in hypermutations in the MSH6 mismatch repair gene. [5,10] These mutations were not present in untreated tumours and suggest that chemotherapy selected for MSH6-mutant cells. A clonal selection process was thought to create an evolutionary bottleneck where the majority of the cells were MSH-6 mutants while cancer cells with the wild-type MSH6 gene were eliminated. [5,11]

The transient decrease in genetic heterogeneity following TMZ administration provides a therapeutic window when cancer cells are most susceptible to a secondary treatment. [5] An in vivo study investigating the effects of the oral poly(ADP-ribose) polymerase (PARP) inhibitor ABT-888 on xenograft models of human tumours found that this PARP inhibitor not only synergistically maintains and potentiates the cytotoxic effects of TMZ on different tumours but also overcomes TMZ resistance. [12] ABT-888 and other similar PARP inhibitors may therefore have a role as a secondary treatment in combination therapies as they can eliminate most of the residual chemo-resistant cell populations. A schematic diagram of a two-step evolutionary double-bind is shown in Figure 1.

Figure 1. Evolutionary double-bind. For simplicity, tumour cells can be sensitive (neutral or susceptible) or resistant to a treatment. A two-step setup would involve the first treatment reducing heterogeneity of the tumour by imposing a high phenotypic cost on tumour cells. The second treatment works synergistically with the first treatment, such as in the case of PARP inhibitors and TMZ, to eradicate initially resistant cell populations.

Chemotherapy-based combination therapies

The widespread use of chemotherapy necessitates a scrutinisation of its synergistic and antagonistic effects in cancer treatment. Basanta and colleagues examined the use of an evolutionary double-bind in a combination therapy consisting of the p53 vaccine and chemotherapy. [13] Using a mathematical framework derived from the evolutionary game theory, they found that the p53 vaccine and chemotherapy work synergistically to exert robust anti-tumour effects. Interestingly, depending on whether the p53 vaccine or chemotherapy was used as the first treatment, different effects were observed.

Application of chemotherapy before the p53 vaccine was found to be more effective than using the p53 vaccine initially followed by chemotherapy. [13] This was attributed to a commensalistic relationship between vaccine-resistant cells and other cell populations. Eliminating vaccine-resistant cells in the first instance disrupts the protective effect and results in other cell populations (e.g. chemo-resistant and fully susceptible) being susceptible to immune mechanisms mediated by the p53 vaccine. In other words, ecological interactions between different cell populations of a tumour appear to determine the effectiveness of an evolutionary double-bind.

Although application of the p53 vaccine before chemotherapy had a diminished anti-tumour effect, the effectiveness of this approach can be increased with longer exposure to the p53 vaccine. [13] Indeed, both approaches appeared to be most effective when the first treatment was applied for a longer period. This reflects the importance of the first treatment as a limiting factor in combination therapy. Prolonged exposure to the first treatment widened the therapeutic window and acted as a barrier against therapeutic resistance most likely by reducing tumour heterogeneity through the creation of an evolutionary bottleneck.

Ecological interactions (e.g. commensalism or competition) between cancer cell populations are important and we can further characterize these interactions by considering the fitness of different cancer cell populations through phenotypic costs. [4] In the absence of treatment, resistant cells are likely to be less fit and have a slower rate of proliferation as compared to sensitive cells since they have to devote more resources to surviving. [4] These cells are most often found in the inner regions of a solid tumour where harsh conditions such as hypoxia and acidosis cause necrosis of tumour cells but favour the selection of resistant clones. Conversely, sensitive cells will be located at the outer rim of the tumour where a close proximity to the vasculature and expression of pro-survival proteins allow them to proliferate easily. [14] We can therefore predict that sensitive cells will be more susceptible to chemotherapy due to their proximity to the blood supply whereas resistant cells are highly affected by metabolic changes.

Silva and Gatenby proposed an evolutionary double-bind strategy consisting of the glucose competitor 2-deoxyglucose (2-DG) and chemotherapy. This was an attempt to reduce the fitness of both sensitive and resistant cell populations as well as stabilize tumour growth through competition via in silico simulations. [15] Different combinations of 2-DG and chemotherapy were modeled mathematically and the combination of 2-DG→chemotherapy was suggested to have the most potent anti-tumour effect. Efficacy was predicted to be lower in chemotherapy→2-DG and lowest in the synchronous administration of 2-DG and chemotherapy. The results become intuitive when we consider tumour cell populations in terms of inner region and outer rim populations. For the 2-DG→chemotherapy approach, the inner region populations are ‘pulverized’ by 2-DG due to their sensitivity to glucose depletion and this increases the surface area for chemotherapy to eliminate the outer rim cells. [15] Furthermore, 2-DG created a ‘pulverized’ morphology where a barrier of cells exists between the outer rim and inner region cells. This potentiates glucose depletion because glucose cannot diffuse effectively from the outer rim to inner region.

Interestingly, 2DG→chemotherapy mirrors the effectiveness of the p53 vaccine→chemotherapy approach. [13] This is probably attributed to the initial targeting of chemo-resistant cells and also the maintenance of a higher proportion of sensitive (and presumably fitter) cells as compared to resistant cells. The latter implies that sensitive cells can impede proliferation of resistant cells via competition for resources. Indeed, the chemotherapy→2-DG approach most likely had a better anti-tumour effect than synchronous administration because, even though the chemo-sensitive outer rim cells were targeted first, the introduction of a break or ‘drug holiday’ between chemotherapy sessions in the study’s protocol allowed the sensitive cells to recover and maintain a sizeable numerical advantage over resistant cells. [16] A similar effect was also noted in previous studies with different treatments. The chemotherapy→2-DG approach fared worse than 2-DG→chemotherapy as glucose can readily diffuse from the outer rim to inner (i.e. allowing chemo-resistant cells to survive) while the synchronous approach was least effective as the outer-rim was readily destroyed by chemotherapy; therefore reducing competition between sensitive cells and resistant cells. [15] Moreover, poor diffusion of chemotherapeutic drugs to areas deeper within the tumour meant that the inner region cells only received sub-lethal doses which favour the development of chemo-resistance.

Out of the three strategies, only the 2DG→chemotherapy approach managed to achieve an almost complete eradication of cancer cells when a bolus of MTD chemotherapy was applied while the other two strategies resulted in chemo-resistance. This result has two implications: firstly, it reflects the point that eradication of tumour cells is possible if tumour heterogeneity is targeted in the first instance and, specifically here, the chemo-resistant population. Secondly, it also implies that delineation of tumour cell populations into sub-groups based on location and proximity to key tumour structures such as the vasculature may be therapeutically significant. In fact, there is evidence that populations of tumour cells often exhibit a convergent phenotype despite genotypic differences between individual cells. [17] Thus, targeting this phenotype may be a more practical option since natural selection acts on phenotypes rather than genotypes.

Riparian ecosystems as an ecological framework for human tumours

Tumour vasculature can contribute to intra-tumoural heterogeneity by creating disparities in substrates such as oxygen and glucose through blood flow gradients, which then select for different populations of cancer cells. [17,18] Alfarouk and colleagues proposed that growth of cancer cell populations can be understood in the context of plant species in a riparian habitat. [18] A riparian habitat is the interface between land and a river stream and two distinct regions of plants can be identified depending on their distance from a river. The mesic region contains lush, tall vegetation which are adjacent to and well nourished by the nutrients from the river. This is followed by an abrupt transition to a xeric region containing sparse, short vegetation which, due to their relatively long distance away from the river, develop adaptations that allow them to conserve water and survive in arid conditions. [19] The rivers and regions of vegetation in a riparian habit are analogous to the vasculature and cancer cells in a tumour respectively.

Tumour cell populations can be broadly separated into ‘mesic’ and ‘xeric’ cells depending if they are adjacent or distal to a blood vessel. [18] Mesic tumour cells and their proximity to blood vessels would render them highly susceptible to angiogenesis inhibitors by systemic administration. Since the ‘lush’ mesic region is expected to contain many tumor cells, a drastic reduction in tumour volume can be achieved. [18] However, the elimination of mesic tumour cells favours unprohibited proliferation of xeric tumour cells and an early treatment directed against the xeric region would be necessary. Phase I and II trials have shown that pro-drug carriers (containing chemotherapeutic drugs) based on 2-nitromidazoles can target hypoxic regions of a tumor and have shown strong anti-tumour effects. [20,21] Combining pro-drug carriers with an intra-tumoural route of administration may improve the accuracy of this approach. Considering the scarcity of xeric tumour cells, prolonged early treatment may be extremely effective. A summary of the different strategies described above is shown in Figure 2.

Figure 2. Best predicted outcomes in evolutionarily and ecologically enlightened strategies. (i) In silico studies suggest that p53-resistant cells and p53-sensitive cells exist in a state of commensalism. The initial introduction of p53 eliminates vaccine-resistant cells and predisposes all remaining cells to destruction by chemotherapy. A greater effect is seen with prolonged p53 administration. (ii) 2DG targets and ‘pulverizes’ resistant cells, creating physical barrier between resistant cells but retains an outer-rim chemo-sensitive cells that inhibits cancer spread. (iii) Riparian-based therapy may achieve maximal tumour cell death through a localized targeting of mesic cells by hypoxia-based strategies followed by targeting of xeric cells by angiogenesis inhibitors.

Discussion and conclusion

Tumours are resilient in nature because they consist of a heterogeneous system of cells locked in a constant state of feedback. [22] Any perturbations in the environment of these cells may simply reinforce tumourigenic processes which restore overall tumour fitness. Although all therapies inherently disturb this fragile equilibrium, in silico studies have demonstrated proof of principle that a well-designed strategy such as an evolutionary double-bind can control and potentially eradicate most tumour cells. While modelling methods may not translate to immediate clinical benefits, they are an inexpensive way of exploring theoretical concepts in a controlled situation and provide a sound framework for further in vivo studies and clinical trials. The models described here can also readily be modified to study other forms of combination therapy, illustrating their flexibility and broad applicability to the clinical environment. One limitation, though, is that the parameters used in models have to be as realistic as possible and this can only occur through close cooperation between experimentalists and clinicians.

Key features highlighted here such as the need for prolonged initial treatment to reduce intra-tumoural heterogeneity, enhancing competition between resistant and sensitive cells and combining systemic and localized approaches are intuitive and feasible options that can be readily applied to existing treatment protocols. High-dose chemotherapy is no longer considered as a first-line approach except occasionally as salvage treatment for relapsed disease. This is not surprising in light of possible selection for chemo-resistance and increasing preference for low-dose maintenance and adaptive regimens. [6] The examples discussed in this review focused primarily on solid tumours due to easy visualisation and amenability to mathematical modelling. However, treatment of haematological malignancies would also benefit from a double-bind approach as evident from the restoration of drug sensitivity by second-generation tyrosine kinase inhibitors in treatment-resistant chronic myelogenous leukemia. [23]

There are several potential areas for further research. Firstly, we need to understand why natural selection appears to control cancers but does not eliminate them. In fact, a parallel exists with infectious diseases and high fitness costs and the tendency for organisms to evolve tolerance mechanisms may account for this phenomenon. Secondly, we should consider maximising the potential of new treatment modalities such as immunotherapy in evolutionary double-binds. [24] The limited efficacy of immunotherapy appears to contradict observations in natural ecosystems which indicate that biological control incurs higher phenotypic costs and achieves robust control of pests. This implies that inappropriate immune targets are being selected and, therefore, a true double-bind cannot be achieved. [4,23]

In conclusion, therapeutic resistance is a major obstacle to the optimisation of cancer treatment. Evolutionary and ecological principles may appear far-fetched concepts with little direct relevance to oncology but a closer inspection of the evolutionary origins and the spatial organisation of cancer cells reveal strategies that can improve clinical outcomes. Under-utilisation of these concepts is most likely a reflection of an inability to change our mindset rather than an issue of practicality. These encouraging modelling results provide a sound foundation for further translational research.

Conflict of interest

None declared.

Acknowledgments

None

References

[1] Nowell PC. The clonal evolution of tumour cell populations. Science. 1976; 194: 23-28.

[2] Aktipis CA, Kwan VSY, Johnson KA, Neuberg SL, Maley CC. Overlooking evolution: a systemic analysis of cancer relapse and therapeutic resistance research. PLoS One. 2011;6(11): e26100. doi: 10.1371/journal.pone.0026100.

[3] Luqmani YA. Mechanisms of drug reistance in cancer chemotherapy. Med Princ Pract. 2005;14:35-48.

[4] Gillies RJ, Verduzco D, Gatenby RA. Evolutionary dynamics of carcinogenesis and why targeted therapy does not work. Nat Rev Cancer. 2012;12(7):487-93.

[5] Gerlinger M, Swanton C. How Darwinian models inform therapeutic failure initiated by clonal heterogeneity in cancer medicine. BJC. 2010;103:1139-43.

[6] Gatenby RA, Silva AS, Gillies RJ, Frieden BR. Adaptive therapy. Cancer Res. 2009; 69(11):4894-903.

[7] Norton L, Simon R. Tumour size, sensitivity to therapy, and design of treatment schedules. Cancer Treat Rep. 1977;61(7):1307-17.

[8] Gately S, Kerbel R. Antiangiogenic scheduling of lower dose cancer chemotherapy. Cancer J. 2001;7(5):427-36.

[9] Gatenby RA, Brown J, Vincent T. Lessons from applied ecology: cancer control using an evolutionary double bind. Cancer Res. 2009;69:7499-502.

[10] Hunter C, Smith R, Cahill DP, Stephens P, Stephens C, Teague J, et al. A hypermutation phenotype and somatic MSH6 mutations in recent human malignant gliomas after alkylator chemotherapy. Cancer Res. 2006;66(8):3987-91.

[11] Merlo LMF, Pepper JW, Reid BJ, Maley CC. Cancer as an evolutionary and ecological process. Nat Rev. 2006;6:924-35.

[12] Palma JP, Wang YC, Rodriguez LE, Montgomery D, Ellis PA, Bukofzer G, et al. ABT-888 confers broad in vivo activity in combination with temozolomide in diverse tumours. Clin Cancer Res. 2009;15(23):7277-90.

[13] Basanta D, Gatenby RA, Anderson ARA. Exploiting evolution to treat drug resistance: combination therapy and the double bind. Mol Pharm. 2012;9(4):914-21.

[14] Tannock I, Hill RP, Bristow TG, Harrington L. The Basic Science of Oncology. MacGraw-Hill Professional; 2005.

[15] Silva AS, Gatenby RA. A theoretical quantitative model for evolution of cancer chemotherapy resistance. Biol Direct. 2010;5(25):1-17.

[16] Labianca R, Sobrero A, Isa L, Cortesi E, Barn S, Nicolella D, et al. Intermittent versus continuous chemotherapy in advanced colorectal cancer: a randomized ‘GISCAD’ trial. Ann Oncol. 2011;22(5):1236-42.

[17] Yap TA, Gerlinger M, Futreal PA, Pusztai L, Swanton C. Intratumour heterogeneity: seeing the wood for the trees. Sci Transl Med. 2012;4(127): 127ps10. doi: 10.1126/scitranslmed.3003854.

[18] Alfarouk KO, Ibrahim ME, Gatenby RA, Brown JS. Riparian ecosystems in human cancers. Evol Appl. 2013;6(1):46-53.

[19] Helmlinger G, Yuan F, Dellian M, Jain RK. Interstitial pH and p02 gradients in solid tumours in vivo: high-resolution measurements reveal a lack of correlation. Nat Med. 1997; 3:177-82.

[20] Wilson WR, Hay MP. Targeting hypoxia in cancer therapy. Nat Rev Cancer. 2011;11(6): 393-410.

[21] Duan JX, Jiao H, Kaizerman J, Stanton T, Evans JW, Lan L, et al. Potent and highly selective hypoxia-activated achiral phosphoramidate mustards as anticancer drugs. J Med Chem. 2008;51(8):2412-20.

[22] Cunningham JJ, Gatenby RA, Brown JS. Evolutionary dynamics in cancer therapy. Mol Pharm. 2011;8(6):2094-100.

[23] Thomas, F, Fisher D, Fort P, Marie JP, Daoust S, Roche BJ, et al. Applying ecological and evolutionary theory to cancer: a long and winding road. Evol Appl. 2013;6(1):1-10.

[24] Hochberg ME, Thomas F, Assenat E, Hibner U. Preventive evolutionary medicine of cancers. Evol Appl. 2013;6(1):134-43.