Molecular chess? Hallmarks of anti-cancer drug resistance

Alteration of Drug Targets: While it is common to separate drugs used in chemotherapy from newer agents targeting molecular pathways, it is of course a truism that all drugs have targets. These targets can be altered by cells in a number of ways. Rapid down-regulation of a target gene expression is an obvious ploy, exemplified by the effect of doxorubicin on topoisomerase IIα [ 12], but more subtle alteration of drug targets by mutation is also common particularly in response to targeted agents such as receptor tyrosine kinase inhibitors [ 21, 22, 25, 30, 32]. If the target is part of a pathway activated by other molecules, then the cell may activate an alternative molecular mechanism – mutation of EGFR in ALK fusion gene positive lung cancer is a good example [ 28, 35, 36].

Expression of drug efflux pumps: The ATP-binding cassette (ABC) superfamily of proteins includes a number of membrane proteins able to transport a wide diversity of substrates. Besides an ability to transport of toxins out of cells, other substrates include amino acids, peptides, sugars, lipids, steroids, bile salts, nucleotides and endogenous metabolites [ 10]. These pumps act to protect cells by ejecting a wide variety of toxins. Although in bacteria this toxin might be an antibiotic, in human cancer it is often an anticancer drug. Classical drug resistance is mediated by the MDR1 ( ABCB1) gene, which encodes a membrane-based xenobiotic pump molecule, known as phenolic glycoprotein (PgP). This pump is relatively promiscuous and ejects drugs from the cell at a rate that may exceed their entry, rendering the cell resistant. One of the more important molecules of the blood-brain barrier, it has been much studied. This in turn led to the discovery of numerous other pumps, and the human genome contains 49 ABC transporter molecules [ 10], many of which can pump drugs. Besides MDR1 the best known are multidrug resistance related protein ( MRP1, ABCC1) and breast cancer related protein ( BCRP, ABCG2). Pharmaceutical chemists now design drugs with this mind, so that pump mechanisms are less problematic than they were, though even some TKIs, including gefitinib and erlotinib [ 37, 38], are pumped. Metabolite and nucleotide pumps have also been found to be of importance, and genes such as hENT1 have been reported to be important mediators of chemosensitivity in gene expression studies [ 13– 15]. Rapid up-regulation of drugs pumps can occur in cancer cells and lead to resistance [ 12].

Expression of detoxification mechanisms: Drug metabolism occurs at the host level, where it underlies the pharmacokinetics of many drugs, and within cancer cells themselves, where there may be considerable heterogeneity. Molecules such as gluthathione S-transferase (GSTπ) are well known to be up-regulated in some cancers and a potential cause of resistance [ 12, 39]. It is possible that conjugation and excretion of drugs at the luminal surface of some well-differentiated adenocarcinomas may explain the relationship between differentiation and drug sensitivity to some drugs, but this remains uncertain [ 40– 42]. Altered local drug metabolism and detoxification are key resistance mechanisms across many cancers. As an example, these processes have been investigated in the plasma cell cancer, Multiple Myeloma (MM), where a majority of patients repeatedly relapse and finally succumb to the disease [ 43]. The expression of 350 genes encoding for uptake carriers, xenobiotic receptors, phase I and II drug metabolising enzymes and efflux transporters was assessed in MM cells of newly-diagnosed patients. There was a global downregulation of genes encoding for xenobiotic receptors and downstream detoxification genes in patients with an unfavourable outcome. However, there was a higher expression of genes encoding for the aryl hydrocarbon receptor nuclear translocator and Nrf2 pathways as well as the ABC transporters in these patients [ 43].

Reduced susceptibility to apoptosis and cell death: Apoptosis was recognised as a unique form of cell death by Currie and others in the 1970s [ 44, 45]. It attracted the attention from pathologists, but it was not until experiments by Gerard Evan et al. [ 46– 48] that it became clear that avoidance of apoptosis underpinned the development of cancer and was an important resistance mechanism for cancer cells to both chemotherapy [ 8, 47, 48] and agents targeting signaling pathways [ 49– 51]. Other forms of cell death may also be triggered by anti-cancer drugs, including necrosis, necroptosis, and autophagy [ 52]. In all cases, the key feature in resistance seems to be survival signalling which prevents cell death. Not all forms of cell death are the same, and the level of damage required to achieve cell death is variable. This particularly true of autophagy, which can either promote chemosensitisation or chemoresistance [ 53]. In some cases, its inhibition can chemosensitise tumours [ 54]. Necroptosis is a caspase-independent form of cell death induced by receptor-interacting protein kinases (RIP1 and RIP3) or mixed lineage kinase domain-like protein (MLKL). Its importance in cancer treatment is controversial, but its induction may circumvent anti-apoptotic mechanisms [ 55].

Increased ability to repair DNA damage: As cancers must acquire permanent genomic mutations, cancer can be viewed as a disease of DNA repair as changes in these genes produce the mutator phenotype essential the for the acquisition of further mutations. Once a mutation is acquired cancers often become addicted to a different DNA repair pathway. A good example of this is exemplified by BRCA1/2. As BRACA1/2 are key components of a DNA double strand repair pathway these cancers become dependent on another DNA repair component, PARP1, for replication fork progression [ 56, 57]. Inhibition of PARP1 in these cancer cells is catastrophic and results in their death. This is the concept of synthetic lethality [ 58] and has been proposed as a potential Achilles heel in a cancer cell’s defense. Although this concept may enable the clinician to increase the therapeutic index between cancer and normal cells, it is expected that these approaches will also have the potential to develop resistance. DNA damage is recognised by cells, and if they cannot repair the damage, this leads to apoptosis [ 12, 59]. If apoptotic potential is reduced, then cells can survive considerable DNA damage, but an alternative is to up-regulate DNA repair [ 59]. Many cells of course do both.

Altered proliferation: The normal response to DNA damage that cannot be repaired is apoptosis, but as Gerard Evan showed in diploid fibroblasts, the threshold for death is much higher in cells that are not growing [ 46]. Transient reduction in growth is mediated in part by P53 [ 60]. Levels of P53 rise and at first simply reduce cell cycle, only tipping over to stimulate apoptosis at a certain threshold [ 60].

Topoisomerase II inhibitors remain a mainstay of both haematological and solid tumour therapy, but their clinical efficacy is often limited by resistance. Many mechanisms may contribute to this resistance including reduced drug accumulation and/or increased efflux, site specific mutations affecting drug-induced topo II mediated DNA damage, post-translational modifications resulting in altered DNA damage and downstream cytotoxic responses [ 12, 15].

Anti-HER2 antibodies such as Herceptin develop acquired resistance through a variety of mechanisms, including activation of tyrosine kinase in CSCs, upregulation of HER3, activating mutations in the p110a subunit of PIKK (PIK3CA), enhanced HER-ligand autocrine signaling and alterations in apoptotic pathways [ 69]. HER3 has now also been proposed as potentially driving survival of HER2+ cells once they have developed resistance to HER2 inhibitors such as lapatinib and trastuzumab [ 69].

Bortezomib was the first proteasome inhibitor to enter practice. Again a wide range of mechanisms have been reported in acquired resistance to this drug, which has an important role in the treatment of several haematological cancers. These include mutations in proteasome subunits, unfolded protein response, XBP1 and MARCKS proteins, aggresomes, the role of constitutive and immunoproteasomes, alterations in pro-survival signalling pathways, changes in bone marrow microenvironment and autophagy as well as other multidrug resistance mechanisms [ 70, 71].

Antibody-drug conjugates can also be limited by acquired resistance [ 72]. As with small molecules, this resistance is multifactorial and can include altered interaction with the target, altered apoptosis pathways and altered survival pathways. In addition, the payload for any antibody drug conjugate approach is likely to be sensitive to the same range of resistance mechanisms described for small molecule drugs.

The development of new synthetic analogues of existing drugs has been the usual response to try to circumvent resistance. It is possibly best exemplified in the vinca alkaloid derived drugs, where greater potency has been achieved by chemical alteration of molecules [ 74, 75]. In some cases however, this approach has been less than successful, as it tends to increase toxicity.

Combinations have been used in oncology since multiple drugs became available. Most combinations have been developed empirically, on the basis that if two drugs are active, then the combination should be more active still. This has been a successful approach, but as the number of possible combinations has risen, the number of expensive clinical trials required to fine-tune such combinations has made this approach less attractive. Cell lines have been used to design combinations, with some success, but the reality is that highly passaged cell lines are poor models of cancer cell behavior [ 76, 77]. We have previously used primary cell culture to develop new combinations, with considerable success [ 78].

It is clearly important to stratify patients based on whether they are likely to respond to a particular therapy or combination. Although cell lines can provide a useful first step they are unable to effectively model the complex tumour–stroma interactions that contribute to the development of drug resistance. It is now suggested that combining therapies that target two or more orthogonal, ‘independent’ pathways, will be preferable to attempting to hit two or more targets on the same pathway. It is hoped that this approach will reduce the tumour’s ability to mount an effective resistance campaign.

Sequential strategies have much to recommend them, both to increase efficacy and reduce toxicity. Despite some success, relatively few sequential combinations have entered clinical practice, as until recently the molecular understanding of their efficacy has been lacking [ 79]. DNA and RNA sequencing technologies are now at a point where they can be used as companion diagnostic technologies, and the effects of sequential drug administration can be predicted [ 80].

Synthetic lethality is used to describe a mechanistic approach to combination and sequence design. Tumour-specific genetic changes can make cancer cells more vulnerable to synthetic-lethality strategies and so enable the clinician to target tumour cells while sparing normal cells. These mutations in cancer genes may be either loss or gain of function and the concept can be extended to contextual synthetic lethality to include defects in metabolic processes and rewiring signaling networks and tumour-associated hypoxia [ 81]. Nevertheless, even with a new generation of novel targeted cancer therapies based on the concept of synthetic lethality, the potential for secondary acquired resistance remains. Mutation or inactivation of P53 is usually thought to be anti-apoptotic, allowing cells to avoid the induction of apoptosis. However, chemosensitivity experiments in ovarian cancer showed that this was not always the case [ 82], and subsequent studies have shown that under certain conditions, mutation of P53 can confer susceptibility to apoptosis [ 60]. It is increasingly clear that such approaches to synthetic lethality are achievable with sufficient knowledge of the molecular makeup of individual cancers [ 60]. In high grade serous ovarian cancer, characterised by P53 mutation, 20% of patients have BRCA1 and BRCA2 mutations rendering them susceptible to PARP inhibitors, and methylation of the BRCA1 promoter has a similar effect [ 83]. Drug development and companion diagnostic development strategies need to be aligned, and tested in a variety of pre-clinical settings before use in man.

Immunotherapy has long been suggested as a solution to many of the problems of anti-cancer drug resistance. The advent of anti-CTLA antibody treatment with ipilimumab in melanoma [ 84] showed that the promise was likely to be fulfilled, and the success of anti-PD1 and anti-PDL1 antibodies alone and in combination with anti-CTLA4 antibody treatment is nothing short of a revolution in melanoma and lung cancer, to mention but two target cancers of the many that are likely to benefit from these agents. Understanding of resistance to these agents is at an early stage [ 85], and the benefits of combination or sequential use of immunotherapeutics with chemotherapy or targeted agents has yet to be established. However, it is already clear that PDL1 expression by the neoplastic cells is useful as a companion diagnostic, despite difficulties of implementation [ 86]. Neo-antigen load is related to mutational load [ 87], and cancers with high mutational load seem to respond well to immunotherapy [ 88, 89]. It is entirely possible that accurate immune profiling of tumours will require multiple methods.