Abstract
Apart from surgical approaches, the treatment of cancer remains largely underpinned by radiotherapy and pharmacological agents that cause damage to cellular DNA, which ultimately causes cancer cell death. DNA polymerases, which are involved in the repair of cellular DNA damage, are therefore potential targets for inhibitors for improving the efficacy of cancer therapy. They can be divided, according to their main function, into two groups, namely replicative and nonreplicative enzymes. At least 15 different DNA polymerases, including their homologs, have been discovered to date, which vary considerably in processivity and fidelity. Many of the nonreplicative (specialized) DNA polymerases replicate DNA in an error-prone fashion, and they have been shown to participate in multiple DNA damage repair and tolerance pathways, which are often aberrant in cancer cells. Alterations in DNA repair pathways involving DNA polymerases have been linked with cancer survival and with treatment response to radiotherapy or to classes of cytotoxic drugs routinely used for cancer treatment, particularly cisplatin, oxaliplatin, etoposide, and bleomycin. Indeed, there are extensive preclinical data to suggest that DNA polymerase inhibition may prove to be a useful approach for increasing the effectiveness of therapies in patients with cancer. Furthermore, specialized DNA polymerases warrant examination of their potential use as clinical biomarkers to select for particular cancer therapies, to individualize treatment for patients. Antioxid. Redox Signal. 00, 000–000.
I. Introduction
Adapted from (208).
The majority of DNA lesions are amenable to repair by excision of the damaged bases, or nucleotide sequences, and subsequent gap filling against the undamaged template strand, which is carried out by a DNA polymerase (pol). Pols can be divided according to their main function into two groups, namely replicative and nonreplicative enzymes. The doubling of genomic information during the S phase of the cell cycle is carried out by an elaborate replication machinery involving pols that can synthesize the nascent DNA strands with high processivity and fidelity, and enable cells to efficiently replicate their ∼3 billion base pairs of genomic and 17,000 base pairs of mitochondrial DNA (33, 114, 115). Pols α, δ, and ɛ have been identified as the major replicative polymerases in eukaryotic cells, and pols δ and ɛ possess a 3′–5′ exonucleolytic activity that allows the immediate removal of mispaired nucleotides during strand elongation, thus increasing replication fidelity considerably (135, 136) (Fig. 1). Most other pols that have been discovered so far do not assist replication, but carry out a variety of other cellular functions, often associated with the repair of DNA damage. To date, at least 15 different DNA polymerases, and their homologs, have been discovered and characterized, which are grouped into classes according to sequence homology (Table 2). The A family contains pols γ, θ, and ν; the B family comprises pols α, δ, ɛ, and ζ; the X family includes pols β, λ and μ, and the Y polymerase family incorporates pols η, ι, and κ. Additionally, two DNA polymerase homologs have been described, the deoxycytidyltransferase Rev1, belonging to the Y family, and the terminal deoxyribonucleotidyltransferase (TdT), as a part of the X family (40). These pols vary considerably in processivity and fidelity, and many nonreplicative pols lack proofreading activity, therefore replicating DNA in an error-prone fashion.
BER, base excision repair; DSBR, double-strand break repair; NER, nucleotide excision repair; Pol, DNA polymerase; SSBR, single-strand break repair; TLS, translesion synthesis.
Apart from their involvement in DNA replication and repair, certain pols have been linked to the cellular ability to perform strand synthesis past unrepaired DNA lesion sites, thereby temporarily tolerating DNA damage (81, 256) (Table 2). Since cells cannot repair all lesions immediately, some forms of DNA damage may persist into the S phase of the cell cycle, in which they have the potential to inhibit replicative pols and block the replication machinery, resulting in stalling, and ultimately the collapse of replication forks. A disruption of replication threatens cell viability, and therefore cells have developed ways to prevent replication fork stalling (10, 173). Human cells employ a set of specialized pols that are able to perform DNA replication past damage sites, a mechanism called translesion synthesis (TLS) (91, 92, 95, 256). The Y family pols η, ι, κ, and Rev1 have well-characterized TLS activity, but pols of other families, such as pols ζ and β, have also been shown to be able to perform lesion bypass. The TLS pol η is to date the only pol that is causally linked to the development of cancer (35), but deregulations of other pols have been reported in various tumor tissues and have been linked to mutagenesis and drug resistance (33, 134, 135, 140).
In this review article, the links between nonreplicative (specialized) pols and cancer will be discussed in detail. We will focus on the unique properties of nonreplicative pols and discuss the involvement of these enzymes in DNA repair and TLS. The benefits and possibilities of targeting nonreplicative pols will also be reviewed, with a view to improving current treatments for cancer.
II. Biochemistry and Molecular Biology of Nonreplicative DNA Polymerases
A. X family polymerases
The X family of pols contains pols β, λ, μ, and TdT, which share structural similarities and limited sequence homology (264). X family pols have been found across species, and vertebrates express all four pols (242). All X family pols contain an 8-kDa domain, although it only has functional relevance for pols β and λ, where it exhibits 5′-deoxyribose phosphate (dRP) lyase activity. The polymerase domain of all X family pols contains thumb, palm, and finger domains that are required for nucleotide insertions, as has been reported for most pols characterized to date (70, 232, 264). X family pols have been found to participate in the repair of base damage, SSBs, and DSBs, as well as variable, diversity and joining gene segments (VDJ) recombination during antibody diversification (177). Pol β was the first X family pol identified, and it has been extensively studied both mechanistically and in the context of human cancers (231).
1. DNA polymerase beta
Pol β is present in all vertebrates and exhibits a high sequence homology between all mammalian species (20). In yeast, a pol β homologous gene was found encoding a protein with pol β-like enzymatic function, and was named pol IV (207). In humans, pol β is encoded by the POLB gene on chromosome 8 comprising 33 kb and 14 exons, and the protein has a calculated molecular weight of 39 kDa (163). Several different tightly folded protein domains within the pol β protein have been identified that have been attributed to its different activities, namely DNA strand and nucleotide binding, nucleotidyltransferase, and dRP lyase activities (19). The N-terminal region of pol β, with a molecular weight of 8 kDa, comprises the lyase domain, which by itself has a high binding affinity to single-stranded DNA, but only weak affinity to DNA double strands (205). The 31-kDa C-terminal region of pol β holds the pol activity and can be divided into three subdomains: a catalytic, a DNA double-strand binding domain, and a domain binding the nascent base pairs (18). As for most pols, pol β requires divalent cations, Mg2+ or Mn2+, as cofactors for the pol reaction (20). During the insertion of new nucleotides, the pol interacts with the DNA backbone, therefore probing the newly formed base pair for adequate Watson–Crick base pairing and proper formation of hydrogen bonds in the minor groove of DNA. This interaction therefore helps to increase the fidelity of the enzymatic activity of pol β (20, 197).
On a functional level, pol β has been shown to participate in different DNA repair and damage tolerance pathways, of which its involvement in base-excision repair (BER) is the most well understood. BER enables cells to remove damaged DNA bases and to repair DNA SSBs that are created both endogenously and exogenously by DNA-damaging therapies, such as ionizing radiation or various anticancer drugs (66, 78, 221, 257). The initial step of the BER pathway (Fig. 2) is the removal of a damaged base that is recognized and cut from the sugar–phosphate backbone by a damage-specific DNA glycosylase (82, 168). The removal of the damaged base results in the formation of an abasic site (AP site) that is recognized and cleaved by AP endonuclease-1 (APE1), creating an SSB with 5′-dRP and 3′-hydroxyl ends. In the short-patch BER pathway (Fig. 2, left scheme), this 5′-dRP residue is further processed by the dRP lyase activity of pol β, and a 5′-phosphate residue is formed that can be used by the pol activity of pol β, which subsequently adds one nucleotide to the 3′-OH group of the SSB. The DNA ends are then sealed by the DNA ligase 3α–X-ray repair cross-complementing protein 1 (XRCC1) heterodimer (67, 195, 226, 258). As the key pol during short-patch BER, the lyase and pol activities of pol β are both absolutely required, and these enzymatic components are tightly coordinated. Indeed, pol β-deficient cell extracts supplemented with either dRP lyase or pol-deficient pol β mutants could not support short-patch BER (7).
If the 5′-dRP residue is oxidized or otherwise modified, BER requires additional proteins to repair the lesion, since the pol β lyase activity is not able to perform its function on modified 5′-residues (131, 164). During this so-called long-patch BER pathway (Fig. 2, right scheme), the gap-filling step is initiated by pol β, which adds one nucleotide, and pols δ and ɛ then extend from the inserted nucleotide creating a single-stranded flap overhang (13). This overhang is excised by flap endonuclease 1 (FEN1) and the remaining nick is sealed by DNA ligase 1 (long-patch BER) (196, 225). Although long-patch BER pathway is able to replace two to five nucleotides, it likely only contributes to a minority of the total cellular BER capacity (221).
During BER, gap filling by pol β is carried out with relatively high fidelity compared to other nonreplicative pols. The most common errors of pol β are single-nucleotide deletions, and the overall calculated error frequency is about 5–10×10−4 (23, 215). Despite the fact that pol β itself does not have any proofreading activity, the hydrophobic hinge region positions the DNA within the active site of pol β, and has been shown to be critical for the selection of the correct nucleotide (159, 263). There are also suggestions that APE1, as well as WRN and TREX1 or TREX2, may function as an external proofreading enzyme during BER (72, 222, 233).
Apart from its involvement in BER, the activity of pol β has been linked to TLS. Although the majority of TLS is carried out by the low-fidelity pols of the Y family, X family pols, such as pol β, can also replicate past different DNA lesions, particularly AP sites (29, 73). AP sites cannot be bypassed by replicative pols and are therefore usually repaired by BER. However, a minority of these lesions persist to the S phase of the cell cycle, and therefore have the potential to collapse replication forks. Pol β can skip over AP sites by incorporating a nucleotide that is complementary to a template base adjacent to the lesion site, therefore leading to base deletions and substitutions (29). Pol β has also been shown to follow the A-rule when the AP site is opposite a single nucleotide gap (17, 269).
Other DNA lesions may also be amenable to TLS by pol β. Indeed, pol β can incorporate deoxyribonucleotide triphosphates (dNTPs) opposite thymine glycol lesions, and UV-induced cyclobutane dimers and pyrimidine–pyrimidone (6,4) photoproducts (22, 24, 220). Additionally, 1,2-GG and 1,2-AG DNA adducts created by platinum-containing anticancer drugs can be bypassed by pol β in a mutagenic fashion (15). The efficiency of pol β to replicate past these lesions is considerably lower compared to other TLS pols, and the relevance of pol β as a TLS polymerase in vivo remains poorly understood. This is particularly evident considering that pol β wild-type and pol β-deficient cells are equally sensitive to UV irradiation and cisplatin treatment (53, 186, 227). In vitro experiments have shown that fork-like DNA flaps or other higher-order DNA structures can increase the TLS capacity of pol β, but the regulation of this activity in vivo is still largely unclear (63).
2. DNA polymerase lambda
The X family enzyme, pol λ, is encoded by the POLL gene on chromosome 10, and the pol has a molecular weight of 68 kDa (264). It shares 54% sequence homology with pol β and is 34% identical, and pol λ is highly conserved throughout all mammalian species (11). Pol λ comprises domains for DNA binding, nucleotide selection, and 5′-dRP lyase activity, suggesting an involvement in BER (85, 109). While pol λ has been shown to be able to remove 5′-dRP groups from strand-break ends, cell-based studies did not reveal an increased sensitivity of pol λ-deficient cells to alkylating agents (240). Pol λ is thought to act as a backup enzyme in the repair of base damage, since cells depleted of both pol β and λ are incapable of carrying out successful BER (31). Pol λ, like pol β, can carry out gap filling during short-patch and long-patch BER, and while the fidelity of the enzyme is comparable during single-nucleotide insertions, this fidelity decreases significantly with increasing length of the gap. It has therefore been suggested that pol λ may be more suited to short-patch BER (36). Indeed, in vitro assays have confirmed a potential role for pol λ in short-patch BER, and the enzyme has been shown to be able to interact with upstream BER glycosylases (31). Interestingly, pol λ exhibits a BRCA-1 C-terminal (BRCT) domain at its N-terminus, which is known to mediate interactions with other proteins required for the recognition and binding to DNA ends, possibly suggesting a role beyond BER. Indeed, the BRCT domain of pol λ has been shown to enable the protein to interact with nonhomologous end-joining (NHEJ) proteins (Fig. 3) (86, 178).
Further evidence for the involvement of the enzyme in NHEJ comes from observations that pol λ was able to perform error-free gap filling that was dependent on XRCC4 and DNA ligase 4, as well as other proteins that are exclusively involved in NHEJ (143, 154). Furthermore, pol λ-deficient mouse embryonic fibroblasts were found to be sensitive to camptothecin and etoposide, which create SSBs and DSBs (48, 250). Recent data also suggest the involvement of pol λ in homologous recombination (HR) (Fig. 4), and a polymorphic splice variant was found to affect cellular HR activity (47). As with other X family pols, pol λ lacks a 3′–5′ exonuclease proofreading ability, and in vitro experiments have shown a low fidelity of pol λ that is comparable to that of pol β (37). Furthermore, it has been shown that a loop structure upstream of the pol domain regulates the fidelity of pol λ by controlling the movements of the template strand induced by the incoming dNTPs, as the enzyme switches from an inactive to an active conformation (21). Phosphorylation of threonine residues on pol λ seems to stabilize the enzyme during cell cycle progression and prevent proteasomal degradation (259).
3. DNA polymerase mu
Pol μ is encoded by the POLM gene on chromosome 7, and pol μ is a 55-kDa single-subunit enzyme (264). Pol μ is primarily found in lymphoid tissue, and mice lacking the pol show defects in B-cell differentiation and impaired rearrangements of IgG light chains (25, 26). Steady-state kinetic analyses have identified pol μ as a lower-fidelity pol with a misinsertion frequency of 10−2–10−4 (211), and one-base deletion frameshifts were reported to be the major polymerization error caused by pol μ (271). The pol μ 8-kDa subunit lacks the 5′-dRP lyase activity reported in pols β and λ, but could interact with the downstream DNA required for the enzyme's end-bridging activity (264). Interestingly, pol μ has been reported to exhibit RNA polymerase activity and could insert ribonucleotides instead of deoxynucleotides on oligonucleotide-based constructs with a similar fidelity (184, 211, 212).
Pol μ has been demonstrated to exhibit TLS activity in vitro, and the pol was found to insert nucleotides opposite various DNA lesion sites with a relatively high efficiency, including AP sites, 8-oxoguanine and 1,N(6)-ethenoadenine lesions, thymidine dimers, and adducts caused by platinum-based chemotherapeutic drugs (62, 101, 270). As described for pol λ, pol μ exhibits an N-terminal BRCT domain that can interact with XRCC4 and DNA ligase 4, which suggests a role in NHEJ during VDJ recombination (64, 153). However, further structural comparisons of the BRCT domains from both pol λ and pol μ are required to provide an insight into their functions. While a lack of pol μ leads to a delay in DSB repair kinetics and an increased amount of unrepaired damage (54), it has yet to be shown if pol μ has a general role in DNA DSB repair outside of VDJ rearrangements. Pol μ has the ability to form foci upon radiation-induced DNA damage that have been described to co-localize with γH2AX foci supporting a possible recruitment of the pol to sites of DNA DSBs (157). However, conflicting data have been reported examining expression patterns of pol μ in the context of radiation-induced DNA damage. Upregulation of pol μ on the protein level has been reported after γ-irradiation, suggesting that pol μ may be involved in the repair of ionizing radiation-induced damage (157). In contrast, POLM mRNA was reportedly downregulated after cellular treatment with various DNA-damaging agents, including alkylating agents, UV light, or γ-irradiation, and it has been suggested that the expression of the pol is regulated on a post-transcriptional level after DNA damage that may include post-transcriptional modifications or stabilization on chromatin containing DNA damage (11).
B. Y family polymerases
The Y family of pols comprises the enzymes, pols η, κ, ι, and Rev1, and all family members have been found to exhibit TLS activity. Generally speaking, the Y family pols are highly conserved throughout species and share structural homology among each other, but also with TLS pols of other pol families (95, 256).
1. TLS pathways
Two models have been proposed for the mechanism of TLS, the polymerase-switch model (144, 165) and the gap-filling model (167, 256). In the polymerase-switch model (Fig. 5, left scheme), TLS activity is required at replication forks when the replication machinery is blocked, and the lesion bypass is crucial to prevent fork collapse and to continue DNA synthesis. During this process, the E3 ubiquitin ligase Rad18 and the E2-conjugating enzyme Rad6 are recruited to the DNA lesion site (83, 255). The Rad18-Rad6 complex then monoubiquitinates the DNA-sliding clamp, proliferating cell nuclear antigen (PCNA), which allows recruitment of TLS pols (9, 103). This is achieved through the binding of monoubiquitinated PCNA to the ubiquitin-binding zinc finger or ubiquitin-binding motif present within most TLS pols (27), although there is evidence that TLS pols can get recruited to sites of DNA damage independent of the presence of Rad18 or PCNA ubiquitination (3, 216). Interestingly, polyubiquitination of PCNA, which also occurs on the same lysine residue (K164) as monoubiquitination, by the MMS2-UBC13 complex and RAD5 has differential effects by recruiting alternative factors required for postreplicative repair (32, 100). Nevertheless, the relevance of PCNA monoubiquitination for the TLS pathway has not yet been completely unraveled, although it has been suggested that the modification may contribute to strengthening the bond between the pol and PCNA, or may facilitate the switch at a stalled replication fork between the blocked replicative pol and the incoming TLS enzyme (55). This pol switch thereby allows the TLS pol to access the DNA lesion site and carry out translesional bypass (144).
It has been suggested that depending on the structure and extent of the DNA damage, TLS may require sequential activity of more than one pol to extend from the lesion site. Insertion of nucleotides opposite certain lesion sites is carried out by specific TLS pols, whereas strand extension from an inserted nucleotide seems to be mainly performed by pol ζ and also by pol κ (99, 118, 254). Upon completion of lesion bypass, a second pol switch may lead to the removal of the TLS pol from the strand and the reinstatement of a replicative pol that can continue high-fidelity nucleotide insertion. Although the exact regulatory mechanism for this second switch is still largely unknown, it has been suggested that deubiquitination of PCNA may be involved in this process (38). Beyond the Rad18–Rad6-mediated regulatory mechanism, additional pathways have also been proposed to lead to the pol switch during TLS. Indeed, the alternative processivity complex 9-1-1 has been shown to be loaded onto damaged DNA strands and interact with the TLS enzymes, pol ζ and pol κ, in yeast models (120, 214).
The second model proposed involves gap filling by the TLS pathway postreplication and is independent of replication fork stalling. TLS-mediated gap filling (Fig. 5, right scheme) may contribute to the sealing of single-stranded gaps that have been left unresolved by the replication machinery during the S phase of the cell cycle (167, 256). The gap-filling model was established in yeast models and has not yet been sufficiently studied in human cells. The recruitment of TLS pols to DNA lesion sites outside of the S phase may involve the same mechanisms as reported for the pol switch model, but since the replication machinery has already passed the damage site, a pol switch is not required for the TLS pol to access the lesion-containing strand. However, after sealing of the remaining nick by a DNA ligase, a pol switch to a replicative pol may still occur.
2. DNA polymerase eta
Pol η was the first TLS pol to be discovered and characterized, and still to date, pol η is the most well-studied TLS enzyme, and is the only pol causally linked to the development of cancer (35). Pol η was first described as the enzyme lacking in the variant form of xeroderma pigmentosum (XPV), a rare autosomal-recessive disorder that leads to a massive increase in UV sensitivity in patients (117, 125, 162). The symptoms of XPV patients resemble those suffering from other forms of xeroderma pigmentosum, despite a functional nucleotide excision repair (NER) pathway (Fig. 6) (147).
Pol η is encoded by the POLH gene on chromosome 6, and the protein contains 713 amino acids (272). The active site is located in the N-terminal region of the enzyme and is highly conserved throughout the Y pol family. It exhibits a more open configuration to enable the enzyme to accommodate even bulky DNA adducts, such as thymidine dimers caused by UV irradiation (8, 203). The C-terminus of pol η contains sequences that allow the enzyme to be recruited to DNA lesion sites (35, 123). Pol η has been cocrystallized with DNA, and it has been found that it only shares minimal sequence homology with other pols outside the Y family (8, 28, 224). However, it shares basic structural similarities, such as a palm domain containing the glutamate and aspartate residues to coordinate the divalent cations that facilitate the nucleophilic attack on the alpha-phosphate and stabilize the pyrophosphate-leaving group on the incoming dNTP (14). Pol η also contains thumb and finger domains that hold the DNA and make connections to the primer and template strands, respectively (203). Furthermore, the pol contains a polymerase-associated domain, which is a characteristic of all Y family pol members, which makes additional contacts with the DNA (150). Moreover, as with all pols of the Y family, pol η lacks intrinsic 3′–5′-exonuclease proofreading activity, and despite this, after insertion of an incorrect nucleotide, pol η was found not to be able to continue strand elongation and dissociates from the DNA lesion site (253). It is believed that the misinsertion rate of pol η is determined by hydrogen bonds formed between the catalytic domain of the enzyme, the incoming dNTPs, and the nascent DNA strand (8). Therefore, the fidelity of the TLS pol depends on its ability to insert the correct nucleotide and to only continue strand elongation from the correctly paired primer termini.
Pol η can bypass thymidine–thymidine dimers (cyclobutane pyrimidine dimers [CPDs]) and pyrimidine–6/4-pyrimidone photoadducts caused by UV radiation in an almost error-free manner (253). Apart from UV-induced damage, pol η has been reported to be able to bypass a variety of other DNA lesions, including 8-oxoguanine, O6-methylguanine, benzo[a]pyrene-N2-deoxyguanine DNA adducts, as well as DNA damage created by treatment with the anticancer drugs, cisplatin, gemcitabine, and oxaliplatin (57, 58, 156, 210, 244). Processivity of pol η largely depends on the DNA structure, and while it is low on undamaged DNA, pol η was reported to insert nucleotides opposite CPDs with significantly higher processivity (166). Since overexpression of pol η in human cells does not increase cellular mutagenesis, and vector-based multicopy expression proved nonviable in cell lines, it has been concluded that pol η activity is highly regulated, and the enzyme is prevented from accessing undamaged DNA in vivo (130, 239).
Pol η requires Rad18 to monoubiquitinate PCNA to be loaded onto damaged DNA (255), although it has been suggested that Rad18-dependent PCNA ubiquitination may not be essential for Pol η recruitment (3, 216). Pol η itself then gets monoubiquitinated, although the relevance of this modification for its TLS activity has not been revealed (27). Translesional bypass by pol η only spans the lesion site, and a few adjacent bases before the binding of the pol to the primer terminus get destabilized, and a pol switch removes the enzyme, allowing replicative pols α or δ to continue strand elongation (137, 161). Beyond its involvement in TLS, pol η has been suggested to be involved in the repair of DSBs by HR, and in vitro experiments have shown that the pol is able to perform the D-loop extension step during HR (126, 170).
3. DNA polymerase iota
Despite its sequence homology with pol η, pol ι has been suggested to play a distinct role in TLS. Saccharomyces cerevisiae, often used as a model organism for the study of the translesional pathways, lacks the enzyme, and the function and involvement of pol ι in the TLS system are therefore relatively unexplored (169). Human pol ι is encoded by the POLI gene on chromosome 18, and it is believed that the pol ι protein is used as a backup pol employed for the bypass of UV-induced lesions in the absence of pol η, where it has been shown to contribute to a unique spectrum and high rate of mutations, as reported in XPV (94, 252). Lack of pol ι in turn increases mutation rates and has been shown to result in a higher incidence of different cancers in both humans and in mouse models (71, 142, 193). However, unlike in the absence of pol η, there is no distinct phenotype associated with pol ι deficiency, and no immune defects or deficiencies of UV lesion bypass have been reported in the absence of pol ι in mice (94, 223). Similarly, deficiency of pol ι did not increase cellular sensitivity to alkylating agents (201). A deficiency of pol ι in human cell lines has been demonstrated to lead to an increased sensitivity to oxidative damage, and due to the established interaction of pol ι with the BER enzyme XRCC1, this evidence suggests a role for pol ι in the cellular response to DNA damage induced by oxidative stress (198).
Pol ι has a considerably lower fidelity than pol η, but in vitro experiments have shown that the fidelity of pol ι could be increased by substituting the divalent cation, Mg2+ with Mn2+ (79). The fidelity of pol ι was found to vary between different template nucleotides, and the enzyme incorporated dNTPs opposite purines with higher fidelity than opposite pyrimidines, a finding that may be explained by the preference of pol ι for Hogsteen base pairing (59, 116, 181). The presence of the auxiliary proteins, PCNA, replication protein A (RPA), and replication factor C (RFC) has been shown to increase the enzymatic activity of the pol in vitro, and structural analysis of pol ι has revealed a binding motif for ubiquitinated PCNA and dependence on Rad18 activity (41, 98, 121, 156). Pol ι was also found to be able to physically interact with pol η, although the impact and mechanistic relevance of this interaction are not currently understood (124).
4. DNA polymerase kappa
Pol κ, encoded by the POLK gene on chromosome 5, is different from the other Y family TLS pols in several respects. Pol κ is the most highly expressed and most-conserved TLS pol characterized to date (256) and has only weak translesional activity and limited ability in DNA lesion bypass. Indeed, the highest TLS efficiency and accuracy of pol κ have been demonstrated across N2 guanine adducts or interstrand crosslinks (60, 174). Unlike most other Y family pols, pol κ is capable of replicating undamaged DNA with relatively high processivity. However, while the Escherichia coli pol κ homolog, pol IV, has been shown to perform high-fidelity replication on undamaged DNA, human pol κ has misinsertion frequencies comparable with other TLS pols (132, 190). Apart from its translesional activity, pol κ performs extension from nucleotides inserted opposite DNA lesions and acts as the extender pol during TLS, which makes it dependent on other inserter pols to bypass the site of DNA damage (152, 254). Beyond its involvement in TLS, a role for pol κ has been implied during the polymerization steps of NER (Fig. 6), most likely in cooperation with pol δ (189).
Pol κ deficiency leads to an increase in cellular sensitivity to UV radiation and alkylating agents, and the enzyme has been found to be involved in high-fidelity bypass past UV-induced CPDs, in collaboration with pol η in Xenopus laevis oocytes (188, 235, 262). Pol κ has been found to interact with PCNA and the 9-1-1 complex, but unlike pol η and ι, pol κ foci were found not to co-localize with PCNA upon DNA damage, suggesting that the pol may not be a vital part of the TLS machinery during the S phase of the cell cycle (187, 191). Overexpression of pol κ has been shown to result in genomic instability and an increase in DNA damage, and mice deficient in pol κ were found to exhibit a mutator phenotype (16, 230). Furthermore, pol κ overexpression resulted in slowing down of replication fork progression without activation of the replication checkpoint (199).
C. A family polymerases
1. DNA polymerase theta
The A family pol, pol θ, is encoded by the POLQ gene located on chromosome 3. The enzyme is composed of three domains, an N-terminal ATPase–helicase-like domain, a central spacer domain, and a pol domain located at the C-terminus (217). On undamaged DNA, pol θ was found to insert nucleotides with an ∼100-fold increased misinsertion rate (2.4×10−3 nucleotides) compared to other A family pols (12). On the other hand, pol θ was observed to be very efficient at bypassing AP sites (107, 218). Pol θ has also been shown to contain a 5′-dRP lyase domain, and therefore a dual role for pol θ as a pol involved in both TLS and short-patch BER of AP sites has been suggested, indicating an important role for pol θ in the repair of these DNA lesions (206). These findings are supported by data showing that pol β and pol θ have overlapping functions and cooperate during BER in vitro (268).
Studies have also suggested that pol θ acts as an extender pol that can extend from mismatches and base pairs containing CPDs or pyrimidine–6/4-pyrimidone photoproducts in vitro (219). Investigations of pol θ in Drosophila melanogaster have uncovered an important role for the enzyme in a DNA ligase 4-independent alternative end-joining pathway (52), and C. elegans pol θ has been implied in interstrand crosslink repair (179), although the relevance of these findings for higher eukaryotes remains unclear. The lack of pol θ in human cells was found to increase radiosensitivity in tumor and bone marrow cell lines (89, 105), and mice deficient in pol θ are viable, but cells from these mice produce an increased level of ionizing radiation-induced micronuclei (89). Furthermore, pol θ has previously been implicated in somatic hypermutation and immunoglobulin diversification (133), although this is still under debate (160).
D. B family polymerases
1. DNA polymerase zeta
The B family pol, pol ζ, is comprised of two subunits, the catalytic subunit Rev3, encoded by the REV3L gene on chromosome 6, and the accessory subunit Rev7, which has been shown to considerably increase the catalytic activity of the pol (256). Mice lacking pol ζ are not viable, and furthermore, loss of the catalytic subunit of pol ζ leads to early embryonic lethality (75, 260). In contrast to the other pols of the B family, the replicative enzymes pol δ and ɛ, pol ζ is devoid of a 3′–5′-proofreading exonuclease activity, and it has been suggested that pol ζ contributes considerably to mutagenesis in the majority of eukaryotic organisms (1, 68). However, a deficiency of this pol has been found to result in increased chromosomal instability, and it has been suggested that pol ζ mediates cellular tolerance to spontaneously occurring DNA damage (261). Although pol ζ was shown to be able to insert nucleotides opposite certain DNA lesions, its main role is believed to be in extension from nucleotides inserted opposite DNA lesions, thereby cooperating with another specialized pol during TLS (2, 273). Indeed, several interactions between pol ζ and other TLS proteins have been described, and both the catalytic Rev3 subunit and the accessory Rev7 subunit can bind to the Y family pol-like protein Rev1 (97). Additionally, the enzymatic activity of pol ζ has been found to increase in the presence of the replication factors PCNA, RPA, and RFC, as well as the 9-1-1 complex (214, 273). Pol ζ protein expression levels are reportedly low in most cell types, and studies performed in S. cerevisiae showed increased mutagenesis and resistance upon UV irradiation when pol ζ was upregulated (209).
III. Biological Relevance of Nonreplicative DNA Polymerases to Carcinogenesis
A. X family polymerases
While no causal connection between X family pols and cancer has been established to date, various associations and deregulation patterns have been described for certain enzymes.
Pol β has been the most thoroughly studied X pol member in the context of cancer. Because of the significant amount of DNA damage occurring spontaneously under normal conditions, all cells depend on BER to repair the majority of DNA lesions (108). While the DNA glycosylases involved in this pathway have redundant functions, pol β has a unique function in gap filling, and has been shown to be essential for cellular viability. Mice deficient in pol β die very early during embryogenesis, and embryonic fibroblasts derived from these animals are hypersensitive to methylating agents that induce DNA damage repaired by BER (93, 202, 234). Due to its essential role during BER, pol β is constitutively expressed in all tissues, and pol β expression has been shown not to change during the cell cycle (274). Pol β levels have been reported to be low in most tissue types, with the exception of normal testicular tissue, and low pol expression may help to prevent pol β from interfering with replication by high-fidelity pols (106, 221). Based on data from artificial model systems, it has also proposed that overexpressed pol β has the ability to replace the replicative pols, pol δ, and ɛ, during the gap-filling step of NER (46).
Cellular upregulation of pol β has been reported after treatment with cytotoxic agents or exposure to oxidative stress, and increased pol β levels correlate with resistance to oxidative agents, such as hydrogen peroxide and the alkylating agents, methyl methanesulfonate, melphalan, and cisplatin (44, 56). Upregulation of pol β was found to lead to an increase in frameshift mutations during BER (51); conversely, downregulated levels of pol β have also been linked to an increased rate of spontaneous chromosomal aberrations and increased cancer risk, as observed in pol β happloinsufficient mice (42, 43). Interestingly, a reduction of pol β levels in some tumor tissues was reported to correlate with a downregulation of other pols, suggesting a connection between BER and other pol-dependent DNA repair pathways (6).
Pol β levels have been studied in clinical samples of various tumor types. Analysis of mRNA expression arrays has shown a reduction in pol β levels in ∼20% of all screened cancer samples, and downregulation of pol β has been found mainly in breast and colorectal cancers (6). In general, pol β overexpression seems to be considerably more prevalent than underexpression in tumor samples, and it has also been shown for various cancer cell lines (45, 228). It has been reported that increased pol β mRNA and protein levels are observed in ∼35% of tumor tissues in comparison to the corresponding normal tissues, and overexpression was most marked in samples from ovarian, uterine, prostate, and gastric cancers (6).
The short arm of chromosome 8 that contains the POLB gene coding for pol β has been found to be lost in various tumor types, including hepatocellular, colorectal, prostate, and lung cancers (74, 155, 247). In a meta-analysis, pol β mutants or splice variants have been reported in almost half of all cancers, and about 10% of the analyzed cancer samples express a truncated pol β protein (231). Several pol β variants that are expressed in tumor cells have been found to impede BER activity or pol fidelity (138, 139), and one deletion variant has been reported to act in a dominant-negative fashion by impeding gap filling during BER (61, 182, 183, 249).
Pol λ has also been found to be overexpressed in different tumor tissues, and more than a twofold upregulation was reported for almost 25% of examined samples (6). In a study examining pol λ expression in bronchial tissue from smokers, expression correlated with the amount of smoking in normal tissues, but not cancer tissues, and lack of pol λ expression in heavy smokers was linked to an advanced tumor stage (192). Interestingly, a naturally occurring pol λ splice variant has been characterized that results in a mutator phenotype by impeding the NHEJ pathway (238).
The third X family pol, pol μ, has been shown to be preferentially expressed in secondary lymphoid tissue where it seems to contribute to somatic hypermutation (213). However, little is known about expression of pol μ in cancer tissues, and potential deregulations in expression have not yet been characterized.
B. Y family polymerases
Y family pols have been implicated in mutagenesis and tumorigenesis, and to date, pol η is the only pol that appears to exhibit a causal connection with the development of skin cancers (35). Various mutations in the POLH gene leading to nonfunctional pol η protein have been described in XPV patients from Europe, Asia, and North America (110, 236). Additional variants of the POLH gene have been suggested as low-penetrance alleles predisposing for malignant melanoma (65). In contrast, no mutations or variations in the POLH gene were found in a panel of basal and squamous skin cancers (77, 87).
Expression patterns of the Y family TLS pols have been studied in a range of tumor samples, although with conflicting results. An upregulation of pol η and ι in a panel of tumor samples, whereas pol κ generally showed lower levels, compared to corresponding normal tissues has been reported (6). In this study, pol ι was found to be increased more than twofold in almost 30% of all analyzed samples, whereas a downregulation was observed in about one-quarter of the samples. In a panel of breast cancer biopsies, pol ι was generally found overexpressed, and it has been suggested that deregulation of the pol leads to a decrease in fidelity, thereby contributing to increased genomic instability in these tumors (265). In glioma samples, pol ι upregulation was observed in 27% of all examined cases, and detection of increased expression by immunohistochemistry correlated with shorter patient survival (251). It has also been suggested that hypoxia is able to upregulate pol ι expression in human tumor cell lines (112).
In contrast, a general downregulation of the TLS pols, pol ι, η, κ, and ζ, in tumor specimens derived from lung, gastric, and colorectal cancers has been reported (194). Conversely, two further studies reported no alterations in pol η expression in lung and gastric cancer compared to normal tissues, but a significant inverse correlation between expression levels of pol η and patient survival was found (50, 237). Another study assessed gene expression profiles in nonsmall-cell lung cancer biopsies and reported an upregulation of pol η (246). Furthermore, in a small cohort of skin cancer patients, pol η was found downregulated in 4 of the 10 examined basal-cell carcinoma biopsies and upregulated in 3 of 10 specimens, whereas only 1 out of 7 squamous cell cancer samples exhibited a significant increase in pol η (77).
In a panel of lung cancer samples, increased pol κ levels were found in 21 of 29 examined samples (185), and pol κ upregulation has been linked to genetic alterations and tumorigenesis in a mouse model (16). In a glioma study, pol κ was reported to be upregulated in 72 out of 104 tumor samples, and immunohistochemical detection of pol κ in these samples correlated with a poor prognosis (251).
C. A and B family polymerases
Only a few studies to date have been published on the expression of A and B family DNA pols in cancer tissues. Expression of the Rev3 subunit of the B family pol, pol ζ, has been found to be significantly reduced in a panel of 74 colorectal cancer samples when compared to normal colon tissue (34). Several single-nucleotide polymorphisms in the POLZ gene have also been found to correlate with either breast cancer development or prognosis (248).
The A family pol, pol θ, is generally observed at very low levels in most tissues, although significant normal tissue expression has been reported for the lymphatic system, and has also been observed in certain tumors (127). Two studies found a correlation between overexpression of pol θ in tissues of breast cancer patients and an adverse prognosis for these patients (104, 145). Furthermore, overexpression of the pol in colorectal cancer samples has been linked with poor patient survival (200). In cell lines, pol θ overexpression led to disturbances in replication, and eventually genomic instability, which may contribute to carcinogenesis (145).
IV. The Potential Role of Nonreplicative DNA Polymerases in Cancer Treatment
An understanding of the complexity of the involvement of multiple nonreplicative pols in DNA damage repair pathways (summarized in Table 2), and the implication that nonreplicative pols are involved in carcinogenesis (Section III above) has led to studies on the role of these enzymes in determining cellular responses to cancer treatment and the search for potential correlations with patient response and survival.
In addition to surgical approaches to cancer, chemotherapeutic treatment with cytotoxic drugs and radiotherapy using ionizing radiation remain the mainstays of current cancer therapy. Both of these nonsurgical approaches can use the DNA of cancer cells as a target, thereby inhibiting further proliferation and eventually leading to cell death. However, cancer cells have evolved pathways that can repair treatment-induced DNA lesions and prevent the accumulation of lethal damage. Several of these pathways require specialized pols for intermediate steps, for example, for gap filling or lesion bypass. Based on this knowledge, certain specialized pols may prove to be valuable targets to improve tumor therapy, or to inhibit the development of resistance in various cancers.
A. Involvement of DNA polymerases in cellular responses to chemotherapy
A significant proportion of solid tumors are either inherently resistant to different chemotherapeutic drugs or develop resistance during the course of treatment. Several mechanisms have been proposed in the development of resistance to cytotoxic treatment, among them decreased cellular uptake (96, 151) or cytosolic inactivation (171, 172) (Fig. 7). Alterations in several DNA repair pathways have also been linked to the development of drug resistance in cellular assays and have been shown to correlate with treatment response or outcome in patient-based studies. As specialized pols are integral parts of many DNA repair pathways, their role in the development of resistance to chemotherapy has been widely examined.
In vitro assays have shown that the X family pol, pol β, can efficiently bypass or repair DNA adducts created by cisplatin and oxaliplatin, and it has therefore been hypothesized that pol β may mediate tolerance of cancer cells to platinum-based treatments (15, 53, 243, 245) (Fig. 8). Furthermore, an increased resistance of cultured cells overexpressing pol β to cisplatin, melphalan, and mechlorethamine has been observed (44), and cellular depletion of pol β by RNA interference has revealed an increased sensitivity to monofunctional alkylating agents, such as methyl methanesulfonate (186). Cells deficient in pol β were also found to be hypersensitive to oxaliplatin (266). Similarly, inhibition of pol β, but not pol λ, sensitized cancer cells to the alkylating agent, temozolomide, an effect that was augmented in cells with mutations in the BRCA2 gene (229, 240, 241). The cellular response to the topoisomerase II inhibitor, etoposide, was also reported to be affected by pol β status, since inhibition of the pol resulted in an increased amount of DNA damage and an increased sensitivity to the chemotherapeutic drug (141). In addition, in vitro data have shown that pol β was able to incorporate antimetabolite drugs, gemcitabine and cytarabine, into DNA, which would suggest that these drugs may be more beneficial in tumors overexpressing pol β (204).
Similar to pol β, the Y family pol, pol η, has also been shown to be able to bypass DNA adducts created by platinum-based cytotoxic drugs in vitro, and the structural basis for this bypass mechanism has been studied in detail (8, 15, 244). Moreover, a deficiency of pol η has been shown to increase sensitivity to cisplatin and oxaliplatin, as well as a cisplatin/gemcitabine combination treatment in cell lines (5, 57). It has been suggested that expression levels of POLH mRNA may predict response to treatment with cisplatin in nonsmall-cell lung cancer patients and to treatment with oxaliplatin in gastric cancer patients (50, 237). Pol η was furthermore found to be capable of bypassing DNA adducts induced by the cancer drug tamoxifen in vitro (267). In contrast, the enzymatic activity of the Y family pols, pol ι or pol κ, has not been linked to the response of cells to chemotherapeutic agents, and in particular, a deficiency in pol ι was shown in several studies not to increase cellular sensitivity to alkylating agents (198, 201, 240). Conversely, in vitro studies in S. cerevisiae showed that both pol ι and pol κ could bypass DNA damage caused by the alkylating agent methyl methanesulfonate, suggesting that they may have a nonexclusive role in the cellular response to alkylating agents (119). Similar results have been published for pol θ, which did not mediate a cellular response to the alkylating and methylating agent, temozolomide (105). However, cells derived from pol θ-deficient mice have been demonstrated to be hypersensitive to the chemotherapeutic drug bleomycin (89).
The B family pol, pol ζ, has been suggested to play a role in the cellular bypass of platinum-induced DNA adducts, since RNA interference-based depletion of pol ζ in head-and-neck squamous cell carcinoma cell lines resulted in an increased cellular response to cisplatin treatment (4). Similar results have been reported for tumors with an acquired resistance to platinum-based chemotherapy, and furthermore, a knockdown of pol ζ in colorectal cancer cell lines deficient in mismatch repair was reported to resensitize these cells to cisplatin (148). In a mouse model of cisplatin-resistant lung cancer, suppression of the catalytic subunit of pol ζ, Rev3, led to resensitization of these tumors to cisplatin treatment, and resulted in an increased survival of the treated mice (69).
B. The role of nonreplicative polymerases in cellular responses to radiotherapy
Similar to cancer treatment with cytotoxic drugs, radiotherapy targets cellular DNA causing DNA damage that ultimately leads to cell death. Unlike most chemotherapeutic agents that create specific DNA lesions, ionizing radiation results in a plethora of different types of DNA damage, including oxidative base damage, SSBs, and DSBs, as well as complex DNA damage containing two or more different lesions within one helical turn of the DNA (90, 158) (Fig. 9). Although DSBs are a minority of all DNA damage induced by ionising radiation, they are believed to be the major DNA lesions by which radiation causes cell death. Because of the complexity of radiation-induced DNA damage, several DNA repair pathways are involved in the repair of these lesions, and specialized pols have been implicated in the cellular repair of ionizing radiation-induced damage (Fig. 9).
Recently, a link between deficiency in the A family pol member, pol θ, and the cellular response to ionizing radiation has been found by two independent research teams. Pol θ has been identified as a mediator of the cellular radiation response in a siRNA-based screen, and subsequently, a knockdown of pol θ increased radiosensitivity in tumor cell lines (105). Similarly, a deficiency in pol θ was found to radiosensitize bone marrow stromal cells, causing an increase in cellular micronuclei, a marker of DNA damage (89).
Likewise, the X family pol member, pol β, has been linked to the cellular response to radiation. While a deficiency in pol β alone was reported not to result in an increase in radiosensitivity, a dominant-negative variant has been described that could radiosensitize cells by inhibiting the later stages of DNA repair (182, 183, 249). In addition, deregulation of pol β, at levels corresponding to those reported in some tumors, has been suggested to contribute to nucleotidic perturbances and chromosomal instabilities after ionizing radiation treatment (80). Similar observations have been reported for another X family pol member, pol μ. Pol μ is thought to be involved in DSB repair through the NHEJ pathway, and overexpression of pol μ has been shown to lead to an increase in cellular sensitivity to ionizing radiation (49, 157).
At present, the role of other specialized pols in the cellular response to ionizing radiation treatment remains unclear. There is certainly scope for further research in this area, since, using in vitro or cell-based assays, several other pols have been shown to contribute to the various DNA repair pathways employed during the repair of ionizing radiation-induced DNA damage (Fig. 9). This research is required to further establish whether nonreplicative pols may be viable targets for inhibitors to improve the efficacy of radiotherapy in cancer treatment.
C. Targeting specialized DNA polymerases to improve cancer treatment
Taking into account the effects of specialized pols in the cellular responses to cytotoxic cancer treatments and radiotherapy, there may be scope for targeting these enzymes with pharmacological agents (Fig. 10) as a means of improving cancer treatment. Indeed, inhibitors for several specialized pols have been identified and characterized already, and several studies have tested the potential benefits of pol inhibition in combination with cytotoxic drugs in cell culture.
Inhibition of the X family pol member, pol β, has been shown to sensitize cells to the alkylating agent, temozolomide, and the DNA cross-linker, cisplatin, highlighting the important involvement of pol β in repairing the types of DNA damage generated by these agents that are processed by BER and TLS pathways (30, 113). Therefore, the potential importance of targeting pol β in cancer treatment is evident (88). Furthermore, an additional benefit of BER inhibition by blocking pol β activity was reported in BRCA2-deficient tumors, which showed a further increase in the cellular response to temozolomide treatment (229). This concept of synthetic lethality previously observed by inhibition of poly(ADP-ribose) polymerase-1 (PARP-1) in BRCA2-deficient tumors (39, 76) suggests that targeting pol β activity, in combination with a deficiency in DSB repair through HR, may serve as an important approach to cancer treatment. Several molecules have also been described that have differential inhibitory effects on pol β, and while the triterpenoid inhibitors oleanolic acid and betulinic acid were shown to inhibit both the 5′-dRP lyase and pol activities of pol β, the plant sterol inhibitor stigmasterol was found to selectively block lyase activity (84) (Fig. 10). Stigmasterol-mediated inhibition of pol β has been shown to lead to a potentiation of the cytotoxic effect of the radiomimetic cancer agent, bleomycin, which induces DNA strand breaks that may require the SSB repair activity of pol β (84, 146).
Tocotrienol is an agent that can inhibit the activity of pol λ and may have additional effects as an angiogenesis inhibitor (Fig. 10), although no benefit as a combinatorial treatment with any anticancer drug has yet been demonstrated (176). Several other inhibitors for X family pols have been identified and characterized in vitro, but have not yet been tested for potential increases in tumor cell sensitivity to chemotherapeutic treatment (111, 122, 128, 180). Furthermore, Penicillols A and B were the first published inhibitors shown to specifically inhibit Y family pols in vitro (129). Since then, the Y family pol inhibitor 3-O-methylfunicone has been shown to inhibit tumor cell growth and decrease clonogenic survival of cells treated with UV irradiation, linking pol inhibition with the cellular response to DNA damage (175).
The potential advantage of inhibiting nonreplicative pols has been shown in vitro, and agents that may be used for this purpose are currently at preclinical stages of development for cancer treatment. While the potential use of novel inhibitors targeting nonreplicative pols as potential cancer therapy targets is at an early stage, one caveat is the redundancy among the pols, and therefore inhibition of one particular pol may not have a significant effect on tumor cell killing due to its replacement with another pol that is able to undertake the same function. Therefore, multiple pols may need to be targeted, although the effect of inhibiting two or more of these enzymes in normal cells, in comparison to tumor cells, would need to be studied in detail to ensure preferential toxicity to the tumor cells. It should also be noted that due to the differential expression of nonreplicative pols in various normal as well as cancer tissues, but also due to the interindividual variation of expression of these enzymes, these considerations must be taken into effect when using inhibitors against these pols as potential cancer therapy agents. Therefore, more, detailed studies on tissue-specific expression of nonreplicative pols are required as well as further information on the expression of these enzymes in various different tumor types. The molecular mechanisms that control the expression of nonreplicative pols also warrant future research, and both the regulation of expression at a transcriptional level, but also protein stability via post-translational modifications, should be investigated in more detail and subsequently analyzed in various tumor cells to establish whether these mechanisms are defective.
V. Conclusions
Cancer is routinely treated by agents that act via cellular DNA damage and ultimately cause tumor cell death. Nonreplicative (specialized) pols have been shown to participate in DNA damage repair and tolerance pathways relevant to carcinogenesis and the treatment of cancer. An alteration in these DNA repair pathways, specifically in tumor cells, has been linked with resistance of these cells to certain classes of cytotoxic drugs used during cancer treatment. There is also evidence that specialized pols can contribute to cellular tolerance to certain types of DNA damage. In vitro inhibition of specialized pols by RNA interference or small-molecule inhibitors has been reported to increase sensitivity of tumor cells to cancer treatment, and pol inhibition may ultimately prove to be a useful approach for increasing the effectiveness of the current therapies for patients with cancer. However, more detailed information is required on the expression, as well as the mechanisms controlling this expression, of nonreplicative pols in various normal and tumor tissues to be able to predict the normal-versus-tumor-tissue response to pol inhibition in individual patients. Furthermore, as drugs that target nonreplicative pols become available, full mechanistic details on their mode of action will be required to fully ensure that pol inhibition is responsible for mediating the toxic effects to tumor cells. Subsequently, when these drugs are available for clinical trials, clinical studies will be required to evaluate the potential of pol inhibition as a way of optimizing noninvasive cancer treatment and a means of targeting treatment-resistant tumors. In the meantime, specialized pols warrant examination of their potential use as clinical biomarkers to select for certain chemotherapies and radiotherapy, with the ultimate goal of individualizing cancer treatment.
Footnotes
Acknowledgments
R.A.S. is funded by the NIHR Biomedical Research Centre Oxford, the Experimental Cancer Medicine Centre, Cancer Research UK, the UK Medical Research Council, and the Higher Education Funding Council for England.