Tumor Heterogeneity, Clonal Evolution, and Therapy Resistance: An Opportunity for Multitargeting Therapy

Abstract

Heterogeneity within the cell population is a feature of many tumors. This lack of cellular homogeneity may originate from a number of sources, including differential nutrient status due to the de novo microcirculations of tumors, to infiltration of normal cells into the tumor, and to the hierarchical natures of the cell populations from which cancers arise. Tumors are thought to arise from one or more tumor initiating cells (TIC) within the population and to found hierarchies of progenitors and more differentiated cancer cells. TIC are often derived from tissue stem cells and these cancer stem cells are characterized by resistance to most cytotoxic treatments and by a high metastatic rate. Many of the properties of tumor populations, including the ability to express mutated oncogenes and to evolve new features such as treatment resistance and invasive and metastatic potential appear to depend on the molecular chaperone Hsp90. We discuss the potential of targeting the heterogeneous cell population with Hsp90 inhibitory drugs and its potential ability to inactivate TIC and to block the evolution of new phenotypes in cancer.

Tumor Heterogeneity

Many solid cancers are characterized by heterogeneity within the cell population. In earlier studies, morphologically distinct zones were distinguished in many cancers with, particularly in larger tumors, cell necrosis occurring in central areas remote from the tumor microcirculation (Folkman, 2006; Fowler, 1967). Tumor cells tend to be particularly sensitive to nutrient deprivation due to their characteristically high rates of metabolism and to the utilization of oxidative glycolysis as a primary pathway of energy production (the Warburg effect) (Warburg, 1956; Zhao et al., 2009). Many cancers are thus highly demanding in terms of glucose and oxygen supply (Rohwer et al., 2012). In addition, when cancers arise, they are obliged to assemble an ad hoc microcirculation as they outgrow the tissue capillary network of the surrounding normal tissues (Hanahan and Weinberg, 2011). This newly assembled circulatory system tends to be inefficient in terms of vascular function and is itself asymmetrical, an effect that amplifies tumor heterogeneity (Folkman, 2006; Rohwer et al., 2012). Asymmetric nutrient deprivation in different zones in the tumor, due largely to differential perfusion rates is thus seen as a major driver of heterogeneity. A fraction of the nutrient deprived cells might be destined for necrotic death while some sub-populations such as transiently hypoxic cells might also develop resistance to treatment modalities such as radiotherapy (Brown, 2007).

Another connotation of the term tumor heterogeneity refers to populations of infiltrating cells that have been observed in a number of tumors. These normal cells may include immune effector cells such as immunosuppressive macrophages, myeloid-derived suppressor cells (MDSC), regulatory and cytotoxic T lymphocytes as well as mesenchymal cells, including fibroblasts and mesenchymal stem cells (MSC) (Colotta et al., 2009; Fridman et al., 2011; Mantovani, 2009). This infiltration cascade seen in many tumors is thought to be initiated by bone-marrow-derived MSC that are attracted to cancers in a similar manner to their role in wound healing (Stagg, 2008). MSC have been shown to differentiate into tumor-associated fibroblasts and such cells thus generate the cytokine milieu that may attract macrophages, MDSC, and regulatory T cells (Treg). Infiltrating normal cell populations are thought to provide tumor cells with growth-promoting cytokines and inhibit immune attack on cancer cells through the secretion of immunosuppressive cytokines.

However over the preceding ten years, the cancer stem cell (CSC) hypothesis has been supported by increasing amounts of evidence and appears to account for a different class of tumor heterogeneity (Reya et al., 2001; Visvader, 2009). It has become clear that, only a minority fraction of cancer cell populations is able to initiate tumors when limiting dilution transplantation studies are carried out in many tumors (Polyak and Weinberg, 2009; Reya et al., 2001; Visvader, 2009). CSC have been envisaged as the shadowy counterparts to the stem cells involved in the renewal of cell populations in normal tissues. In normal tissues, pluripotent stem cells are capable of self-renewal and give rise through cell division to hierarchical cell populations consisting of daughter stem cells, multipotent progenitors, transitamplifying cells, and differentiated cells (Reya et al., 2001; Shibata and Shen, 2012). The stem cell is maintained in its pristine, pluripotent state by signals emanating from the local tissue niche as well as through autoregulation by dedicated pluripotency transcription factors such as Oct4 and nanog (Dontu et al., 2003; Kashyap et al., 2009; Malanchi et al., 2012). In its simplest form one could propose a scheme whereby, when the tissue stem cell undergoes oncogenic mutation, it can give rise to the tumor initiating CSC as well as a hierarchical population of more differentiated cells each bearing the mutation. Such tumor initiating cells (TIC) would thus “breed true” as the single descendants of the mutated tissue stem cells and give rise to a stable hierarchy of daughter cells (Figure 1). Unfortunately, this agreeably simple model does not seem to hold for all tumors. It seems that many tumors contain multiple cell lineages and each of the lineages may be derived both from malignant transformation of the normal pluripotent stem cell as well as from more differentiated progenitors through clonal evolution (Greaves and Maley, 2012; Polyak and Weinberg, 2009; Shibata and Shen, 2012). In addition, TIC may undergo further mutations and epigenetic changes and become the founders of additional lineages within the tumor (Baylin and Jones, 2011; Greaves and Maley, 2012) (Figure 1). As TIC are characterized by resistance to cancer treatment, they also offer a significant challenge for cancer therapy. TIC may resist therapy due to characteristically slow rates of self renewal, to elevated expression of ABC family transporters that evict chemotherapeutic drugs, to the expression of Polycomb factors that protect cells against DNA damage and to the secretion of immunosuppressive cytokines that counteract the effects of immunotherapy (Ginestier et al., 2007; Lagadec et al., 2012; Schatton et al., 2010; Vissers et al., 2012). In addition, TIC/CSC have been implicated as important cells in invasion into the stroma and metastasis to distant sites (Malanchi et al., 2012; Weng et al., 2012).

Figure 1.

Clonal evolution and hierarchical differentiation. In normal tissues, stem cells give rise to descendent cells including pluripotent progenitor, transit amplifying and ultimately cells with differentiated functions within the organ. In Scheme 1, a transforming mutation (X) is shown to give rise to a similar hierarchy of stem (CSC) and descendent cells, each bearing the mutation. In Scheme 2, we show clonal evolution induced by individual second mutations (Y, Z, A, B) in the CSC that already bear mutation X and that can be the founders of individual clones within the polyclonal tumor population. In addition second mutations in the progenitor and other downstream cell types may also give rise to cells with tumor initiating potential (small arrows). We show in the inset box the potential role of multitargeting Hsp90 drugs (Hsp90i) that can suppress the transforming ability of individual mutated oncoproteins (X, Y, Z, A, B), reduce clonal evolution, and potentially kill the cancer cells.

Molecular Chaperones, Mutation, and the Evolution of Tumor Heterogeneity

The major driver of oncogenesis and tumor heterogeneity is thought to be DNA mutation. Activating mutations in the open reading frames and regulatory regions of oncogenes and inactivating DNA damage in tumor suppressor genes are thought to be the primary events, and decrease in the potency of DNA repair is a major contributing factor (Hanahan and Weinberg, 2011). However, many such changes are likely to lead to elevated expression of proteins as well as to polypeptides with point mutations or domain fusions that are likely to be conformationally unstable (Kamal et al., 2003). Such proteins would be likely to be detected by the intracellular protein quality control pathways, polyubiquinated, and targeted to the proteasome for degradation and would thus lose their transforming potential (Calderwood et al., 2009). One way in which these deleterious consequences could be circumvented is through the ability of molecular chaperone complexes to recognize and stabilize client proteins such as aberrant oncoproteins (Trepel et al., 2010). Molecular chaperones tend to be expressed at high levels in cancer and appear to play the role of collaborators fostering the role of oncogenes in tumorigenesis by stabilizing the sensitive conformations of such proteins (Calderwood and Gong, 2012; Ciocca et al., 2013). Many cancers have thus come to be regarded as “addicted to chaperones” due to their dependence on elevated rates of protein folding for growth and survival (Kamal et al., 2003; Neckers and Lee, 2003). One chaperone in particular, heat shock protein 90 (Hsp90) is known to bind and stabilize a wide range of proteins with activity in the cell cycle, in signal transduction and in transcription, many of which are widely implicated in cancer (Kamal et al., 2003; Neckers and Lee, 2003; Neckers and Workman, 2012). The availability of natural product compounds such as geldanomycin that inhibit Hsp90 chaperone function has permitted study of the significance of the “addicted to chaperones” theory and have provided a platform for development of chaper-one-based anti-cancer drugs (Trepel et al., 2010; Whitesell et al., 2012). As mentioned above, tumor heterogeneity and the existence of treatment resistant, highly metastatic TIC suggests the difficulties that may be associated with specific molecularly-targeted cancer therapeutics (O’Hare et al., 2006; Sawyers, 2005). Thus although the development of agents aimed at single oncogenic targets is highly desirable in terms of enhanced treatment specificity and reduced morbidity, many such agents have failed, probably at least partially due to the polyclonal nature of tumor cell populations (O’Hare et al., 2006; Sawyers, 2005). This may offer one explanation for the relative durability of cancer treatment approaches such as traditional chemotherapy and radiation therapy that are likely to have broad spectra of intracellular targets within the cancer cell population. Drugs aimed at inhibiting Hsp90 have been shown to possess multitargeting capacity and could potentially be active against more than one of the potentially multiple tumor cell lineages (Nicolini et al., 2011).

In addition to its role in chaperoning oncogenic proteins, Hsp90 appears to play an unusual role in evolution at the species and cellular levels (Queitsch et al., 2002; Rutherford and Lindquist, 1998). Under normal circumstances, Hsp90 has been shown to be involved in canalization, buffering against phenotypic changes despite the background development of sporadic mutations and the occurrence of environmental fluctuations, thus permitting the acquisition of developmental robustness (Sangster et al., 2008). In addition to canalization, Hsp90 is thought to permit the accumulation within cells of genetic variability, presumably by chaperoning the products of sporadic polymorphisms and mutations that occur continuously within cell populations (Sangster et al., 2008). The chaperone has thus been described a capacitor for phenotypic diversity (Queitsch et al., 2002). A mutator phenotype has been observed in many cancers and would be predicted to accelerate the generation of genetic diversity and the rate of evolution within tumors (Fishel and Kolodner, 1995; Hanahan and Weinberg, 2011). In addition to increased mismatch repair, other important pathways such as non-homologous end joining repair (NHEJ) also become inhibited during tumorigenesis (Lord and Ashworth, 2012). One would thus predict a priori the evolution of multiple tumor cell clones that have in fact been observed experimentally (Greaves and Maley, 2012; Polyak and Weinberg, 2009; Shibata and Shen, 2012). Cancer treatment with cytotoxic agents could select the fittest, most treatment resistant clones from the diverse malignant cell population maintained by Hsp90 and thus promote the evolution of resistance. The capacitor for evolution hypothesis was tested using Hsp90 inhibitors or heat shock to prevent the buffering of phenotypes and permit expression of altered proteins and new phenotypes (Queitsch et al., 2002; Rutherford and Lindquist, 1998). It is however not clear to what extent Hsp90 would inhibit treatment-induced evolution by cytotoxic therapy due to its role in canalization and prevention of expression of the diverse phenotype that it can buffer. One interesting property of Hsp90 that might relate to the cancer microenvironment and tumor heterogeneity is that the chaperone is exquisitely sensitive to energy deprivation. Hsp90 molecular chaperone function requires a relatively high intracellular ATP concentration and low ADP levels (Peng et al., 2005). Therefore in the energy-deprived cells remote from the microcirculation one might see the emergence of cells with novel phenotypes due to loss of Hsp90 activity. This effect might also be exaggerated by the increases in genetic instability and new phenotypes such as increased metastatic ability that have been observed in hypoxic cells in these areas remote from the blood supply (Bindra and Glazer, 2005). Indeed increases in cells with a surface CSC phenotype have also been observed in the hypoxic zones of tumors (Conley et al., 2012; Currie et al., 2012; Xing et al., 2011). The mechanisms underlying these effects of hypoxia are not entirely clear but may be at least in part due to upregulation of the transcription factor hypoxiainducible factor 1a (HIF1a) (Rohwer et al., 2012). HIF1a can influence DNA repair through multiple interactions including association with BRCA1 (breast cancer susceptibility gene 1) and DNA dependent protein kinase, a key regulator of NHEJ (Rohwer et al., 2012). Therefore, nutrient-generated tumor heterogeneity as well as cell hierarchy-generated heterogeneity appear to be mechanistically linked. In addition, hypoxia, through the induction of the hypoxia inducible transcription factors (HIF), can increase the responsiveness of breast cancer cells to cytokines secreted by infiltrating MSC (Chaturvedi et al., 2013).

Targeting Molecular Chaperones

Hsp90 has emerged as a major pharmaceutical target in cancer therapy. This molecular chaperone plays a broad role in cancer as described above, and drug targeting this molecule could offer the possibility of inhibiting multiple intracellular targets simultaneously; in addition the existence of natural product prototype drugs such as geldanomycin that effectively inhibit its function have accelerated progress in this area (Neckers and Workman, 2012; Whitesell and Lindquist, 2005). Thus multiple Hsp90 drugs have been synthesized in recent years, many of which are undergoing clinical trial for cancer treatment (Neckers, 2002; Neckers and Workman, 2012). The heterogeneous, polyclonal nature of many cancers mentioned above indicates the desirability of drugs such as Hsp90 inhibitors with multiple targets. However, CSC subpopulations, important TIC with powerful metastatic functions, are intrinsically resistant to many types of therapy and could be similarly resistant to Hsp90 inhibitory drugs. It has however been shown that the Hsp90 inhibitor IPI-504 can suppress leukemic stem cells expressing a mutant form of BCR-ABL resistant to kinase inhibitory drugs and sensitize the cells to imatinib (Peng et al., 2007). In addition, the Hsp90 inhibitor 17-AAG was shown to reduce the growth of glioma derived stem cells and to synergize with radiation therapy in the inhibition of growth of intracranial tumors in vivo (Sauvageot et al., 2009). Likewise Hsp90 inhibitors sensitized cancer side population cells with CSC function to chemotherapy and triggered apoptosis (Sobhan et al., 2012). Indeed when used at low concentrations 17-AAG was shown to selectively eliminate lymphoma and leukemic CSC compared to the population of rapidly growing progenitor cells by disrupting the activity of HIF1a (Newman et al., 2012). In addition, Hsp90 is required for the activity of the pluripotency transcription factors Oct4 and Nanog essential for maintaining stemness, indicating a fundamental requirement for the chaperone in the stem cell phenotype (Bradley et al., 2012). Thus CSC appear to be, at least in some circumstances selectively susceptible to Hsp90 inhibitors.

However, despite the versatility of Hsp90 inhibitors in terms of inactivating multiple molecular targets in tumor cells, it is still not clear that, at systemically tol erable doses, the drugs will be cytotoxic to commonly observed cancers in human patients. It has therefore been suggested by Whitesell et al. (2012) that Hsp90 inhibitors might be most effectively used in combination with other, more cytotoxic agents. The Hsp90 inhibitors would be envisaged as inhibiting the accumulation of variant phenotypes and clonal evolution within the tumor population and thus reducing the development of treatment resistant tumor sub-populations that are a complication in many protocols (Whitesell et al., 2012). In this context, the Hsp90 inhibitors could also potentially reduce tumor stem cell reprogramming, a treatment complication that has been shown to be induced by cytotoxic agents (Lagadec et al., 2012). Although the extent of treatment-induced CSC reprogramming is currently not clear, such an effect could be potentially limiting to cancer therapy by rapidly increasing levels of treatment resistant cells. In addition, CSC are often highly metastatic, compounding the potential problems of stem cell reprogramming (Weng et al., 2012). CSC reprogramming by radiation appears to involve similar processes to those involved in the generation of inducible pluripotent stem cells, with a prominent role for the transcription factor STAT3 in both processes (Ho et al., 2010; Lagadec et al., 2012; Tang et al., 2012). Interestingly, activation of STAT3 in multiple myeloma was associated with enhanced sensitivity to Hsp90 inhibitory drugs and the drugs effectively inhibited STAT3 activity (Lin et al., 2013).

Another area for novel drug development involving molecular chaperones is the potential targeting of cochaperones, co-factor proteins known to be required for protein folding by the primary chaperone (Cox and Johnson, 2011). Hsp90 can interact with a wide range of such co-chaperones and many of these proteins are essential for effective activity of Hsp90 in cells and for intracellular regulation. Expression of the Hsp90 cochaperones Hop, p23, Cdc37, and the immunophilin molecule FKBP is associated with enhanced tumorigenesis and these molecules may offer novel targets to be investigated in inhibiting tumor growth, clonal evolution, and treatment resistance (Calderwood, 2013; Gray et al., 2008). Cdc37 in particular is expressed to high levels in prostate cancer and could provide an important alternate target in this type of malignancy (Gray et al., 2007). In addition other molecular chaperones such as heat shock protein 70 may also be important in tumorigenesis and clonal evolution and are potential new targets for novel drug development (Powers et al., 2011).

Conclusion

In conclusion therefore, many tumors contain heterogeneous cell populations composed of multiple cell lineages that each might respond quite differently to therapies. In addition such populations may generate a cache of novel mutated proteins that can be utilized to evolve resistance against cytotoxic treatments. It appears, however, that such capacity to evolve is at least partially dependent on molecular chaperones to maintain these potential new traits. Utilization of drugs targeting molecular chaperones may thus be currently indicated in the face of such evolutionary potential. The development of future understanding of the mechanisms of evolution and the generation of phenotypic diversity could indicate whole new areas for study in cancer.