Genome instability is a hallmark of aging and cancer cells [1,2,3]. Homologous recombination (HR), an evolutionarily conserved process, plays a crucial role in maintaining genome stability through the repair of DNA double-strand breaks (DSBs) and interstrand crosslinks (ICLs) and the protection and resumption of arrested replication forks [4, 5]. Therefore, HR is generally classified as a tumour suppressor pathway. Consistent with this, driver mutations in many HR genes, such as BRCA1 and BRCA2, are observed in cancer, especially in hereditary breast or ovary cancer [6, 7].

However, a comprehensive assessment of the roles of HR gene products and the consequences of their inactivation, not solely focused on HR itself, can lead to conclusions that challenge the commonly accepted idea that HR is a tumour-suppressing pathway. Indeed, most cancer cells are highly proliferative and therefore actively replicate their genome. However, the progression of replication forks is routinely hampered by both exogenous and endogenous factors. Consequently, cancer cells are subjected to considerable replicative stress. Therefore, mechanisms that allow cancer cells to cope with replication stress should favour tumour progression. Since HR plays a central role in overcoming replication stress, it should thus favour cancer cell survival and proliferation and, consequently, tumour progression. Here, we discuss this provocative point of view.

RAD51 plays a pivotal role in the key step of HR, i.e., the search for homologous sequences and strand exchange of homologous sequences (the step for which the mechanism was named). This role highlights an important concept and topic of this review—the “RAD51 paradox” [7]: the HR genes that are mutated in cancer encode accessory or mediator proteins, but despite the central role of RAD51 in HR, in contrast to that of other HR genes, the inactivation of RAD51 is not associated with cancer predisposition; rather, the overexpression of RAD51 is associated with a poor prognosis in different types of tumours [7,8,9,10,11,12,13,14].

The accessory/mediator proteins, whose genes are mutated in cancer, promote the loading of RAD51 and the stabilisation of the RAD51 filament, which then both induce HR and protect against alternative mutagenic repair processes. The inactivation of the mediator/accessory proteins results in the absence of RAD51 on damaged DNA, thus providing access to alternative nonconservative repair pathways such as single-strand annealing (SSA) or alternative end-joining (A-EJ) [15], which can preserve some cell viability but are exclusively mutagenic. This raises the following question: does cancer predisposition result from an inability to perform HR, from the activation of nonconservative repair processes, or from both mechanisms? Here, we will also discuss a mouse model that allows us to address the above question. Indeed, this mouse model expresses an inducible dominant negative form of RAD51 that specifically down-regulates HR [16] but still prevents the nonconservative repair pathway because it binds to the DNA [15, 16]. Remarkably, decreasing HR without stimulating alternative nonconservative pathways in vivo induced a pronounced premature aging phenotype but did not induce tumorigenesis and even, in contrast, protected against it [16]. This finding does not support the hypothesis that RAD51 and HR are a tumour suppressor pathway but rather that these factors might play a role in facilitating tumour progression in vivo.

The concepts and conclusions of the comment discussed here require that we first recall the main molecular mechanisms of the different repair pathways potentially involved, which have already been largely detailed in the literature, the “RAD51 paradox” and the RAD51 mouse model; then, we discuss them to explain why RAD51-mediated HR can be considered a tumour support pathway. By challenging the common view, the aim of this review is to redefine and determine the actual impact of RAD51 and HR on cancer predisposition/prevention, highlighting the importance of the balance between HR/RAD51 and alternative nonconservative repair pathways.

Main molecular steps of HR

HR-mediated DSB repair

The molecular role of RAD51 in HR can be described in the HR-mediated DSB repair model (Fig. 1), which illustrates the successive HR molecular steps [17]. Briefly, HR is initiated by single-strand resection of the DSB, generating 3’ single-strand DNA (ssDNA), and then RAD51 is loaded on the ssDNA by mediators such as BRCA1-BRCA2-PALB2, leading to the formation of the RAD51-ssDNA filament. Of note, BRCA1 plays an upstream role in the choice of the DSB repair pathway, favouring resection initiation; it is therefore considered a “resection licensing factor”. The other factors such as the RAD51 paralogs are “assisting factor on Rad51 stability and dynamics”. Since HR relies on sequence homology, its pivotal step is the search for DNA sequence homologies and the exchange of homologous sequences. This step is promoted by the RAD51-ssDNA filament, which thus represents the actual “active species” of HR.

Fig. 1: HR-mediated DSB repair.
figure 1

First, the MRN complex (MRE11/RAD50/NBS1), in cooperation with the ataxia-telangiectasia mutated (ATM) kinase and the chromatin remodelling machinery, recognises DSBs and induces signalling pathways. The subsequent steps of HR can be summarised as follows: (1) The resection of the DSB generates a 3’ ssDNA stretch, which is coated with the replication protein RPA; (2) and (3) the mediators (BRCA1, BRCA2, PALB2) replace RPA by RAD51 on the ssDNA, forming the RAD51-ssDNA filament, which then promotes the search for homology and the invasion of a duplex DNA molecule harbouring homologous sequences; this corresponds to the central pivotal step of HR; (4) capture of the second DSB; (5) DNA synthesis is then primed using the invading 3’ ssDNA; and (6) the resolution of the HR intermediates leads to gene conversions associated or not with crossover or synthesis-dependent strand annealing (SDSA) or break-induced replication (BIR) [4, 96, 97]. Some of the main factors that are mutated in hereditary breast or ovary cancer are shown in the figure.

Full size image

Recently it has been shown that in the absence of BRCA2, RAD51 may still be able to sustain some suboptimal recombination functions, massively antagonized by FIGNL1. Indeed, in the absence of BRCA2, RAD51 can be loaded onto ssDNA and promote some HR events when FIGNL1 is suppressed [18, 19]. FIGNL1 plays roles in modulating RAD51, by preventing its association to chromatin; loss of FIGNL1 allows thus RAD51 to load at DNA double-stranded breaks in BRCA2-deficient cells, rescuing some HR proficiency and viability upon exposure to Olaparib, a PARP inhibitor, or to cisplatin [18].

HR, a replication fork escort pathway

In addition to DSB repair, HR and RAD51 play essential roles in the protection and resumption of arrested replication forks. Replication fork progression is routinely obstructed by different endogenous factors, such as conflicts with transcription, regions that are difficult to replicate, nicks or gaps in the DNA matrix, molecules that covalently bind to DNA, and damages generated by endogenous reactive oxygen species (ROS) [20,21,22]. Reciprocally replication stress generates the production of ROS by the cell [23,24,25]. Exogenous stresses such as chemicals (notably chemotherapy) or ionizing radiation can also induce DNA damage that blocks the progression of replication forks. Notably, unrepaired DSBs lead to replication fork collapse, and reciprocally, the arrest of replication forks can generate DSBs [26].

In the absence of exogenous stress, defects in HR spontaneously affect replication dynamics [27, 28], notably because of the higher endogenous level of ROS [22]. Moreover, HR proteins are found on normal replicating forks [29].

HR actors are essential for both protecting arrested replication forks and resuming DNA synthesis at these forks (Fig. 2).

Fig. 2: Different examples of the roles of HR/RAD51 in the protection and resumption of arrested replication forks.
figure 2

Different causes can lead to the arrest of replication fork progression, such as a blocking lesion (left panel) or a single-stranded nick or gap (middle panel). When a blocking lesion is reached (left panels), the reversion of the blocked replication fork can form a cruciform tetraplex structure frequently referred to as the “chicken foot” (left panels). RAD51 associated with or not associated with its mediators participates in fork reversion and, in addition, protects the reverted fork from degradation. Then, the resolution of the cruciform intermediate by a structure-specific endonuclease (for example, MUS81) generates a one-ended DSB. In the middle panels, when a progressing replication fork reaches a nick or a gap, this inevitably generates a broken fork with a one-ended DSB (middle panel). RAD51 loaded on the broken forks, arising from both the fork reversion (left panel) or from the nick/gap (middle panel), protects them from degradation and allows replication to resume through strand exchange with a homologous sequence, typically the sister chromatid (middle panel). As shown in the right panel, when the replication fork collapses, uncoupling with replication allows it to continue on the replicating sister chromatid. Then, fork regression generates single-stranded DNA (which corresponds to the molecule on which replication has been arrested). RAD51 loaded on this ssDNA molecule (1) protects it from degradation and (2) allows replication to resume after strand invasion of the downstream sequence, which is homologous.

Full size image

RAD51 plays central roles in both these protection and restarting processes. Notably, RAD51 also plays roles that can be independent of its strand exchange activities and of other HR proteins (for review see) [30]. For example, extensive degradation initiated by MRE11 and EXO1/DNA2 can attack a reversed fork (Fig. 2, left panel). By loading RAD51 on the reversed arm, BRCA2 participates in fork protection against such degradation [31, 32]. Since replication fork reversal can also occur in BRCA2 KO cells [33,34,35], this finding suggests that RAD51 can act even in the absence of BRCA2 at such a step. One hypothesis is that when ssDNA tracts are short, mediators (such as BRCA2) are not required to load RAD51. Moreover, several studies have concluded that strand exchange and fork protection are independent activities (for a review, see) [30]. Indeed, the mutant form of RAD51-II3A is capable of promoting fork reversal, but it is unable to catalyse strand exchange [36]. Consistently, RAD51 cannot catalyse fork reversal alone in vitro [37], whereas it is capable of performing strand exchange, supporting the idea that the two processes are, at least in part, mechanistically different.

Recently, additional roles for RAD51 in the protection of abasic sites preventing replication fork breakage have been described [38].

Therefore, HR, particularly RAD51, plays essential roles in replication fork protection, reversion and the restart of arrested replication. Since replication forks are faced with a series of unavoidable endogenous obstructions, HR components, and RAD51 in particular, are essential for cells to replicate their genome and thus proliferate.

Since tumour cells are highly proliferative and thus actively replicate their genome, they are subjected to strong endogenous replication stress. In addition, they generally exhibit high levels of ROS, which generate replication stress [22], and HR enables cancer cells to cope with replication stress induced by endogenous ROS [22]. Thus, HR helps to manage replication stress, promoting cell survival, genome replication and, ultimately, proliferation. Therefore, according to this view, HR and RAD51 should favour tumour progression.

However, this hypothesis contradicts the fact that mutations in HR genes are found in tumours and even confer a predisposition to cancers, notably hereditary breast and ovarian cancers [7]. To address this contradiction, two points need to be considered first: (i) the mechanisms that compete with HR for DSB repair and (ii) the RAD51 paradox.

Competition between the different DSB repair pathways

Several processes can repair DSBs: on the one hand, canonical nonhomologous end-joining (C-NHEJ), which is dependent on KU80/70-Ligase4, and on the other hand, HR, single-strand annealing (SSA) and alternative end-joining (A-EJ), which are all initiated by ssDNA resection. Importantly, while HR and C-NHEJ are conservative repair processes [4, 39], both SSA and A-EJ are exclusively mutagenic since they systematically lead to deletion at the reseal junctions (Fig. 3); thus, they are nonconservative processes. We have proposed that the selection of the DSB repair process operates in two steps: first, the choice between C-NHEJ and ssDNA resection; second, on resected DNA, the choice between HR and nonconservative repair pathways (SSA, A-EJ) [30, 39,40,41] (Fig. 3):

Fig. 3: Selection of the DSB repair mechanism [30, 39, 40, 41].
figure 3

The selection of the DSB repair process involves two successive steps: 1-C-NHEJ (KU80/70-ligase IV and partners), which seals DSBs and competes with single-strand resection, which generates 3’ ssDNA. 2- On resected single strands, there is competition between HR and nonconservative DSB repair (SSA and A-EJ). Both SSA and A-EJ occur via the annealing of complementary ssDNA exposed by the resection (long annealed sequence for SSA, short annealed sequences for A-EJ). Importantly, both SSA and A-EJ are unavoidably mutagenic events since the intervening sequence is always lost. When RAD51 is loaded on ssDNA (left panel), it promotes HR through its homology search and strand exchange activities and, in parallel, inhibits the annealing of complementary sequences, preventing SSA and A-EJ through DNA occupancy, independent of its HR activities [15]. In the absence of RAD51 (right panel), HR cannot occur, but damaged DNA becomes susceptible to mutagenic SSA and A-EJ.

Full size image

RAD51 plays a pivotal role in the second step through two separable activities: (1) RAD51 fosters HR through its homology search and strand exchange activities, and in parallel, (2) RAD51 blocks the annealing step required for SSA and A-EJ through DNA occupancy, independent of its strand exchange activity [15]. Indeed, mutant forms of RAD51, such as yeast/mammalian chimeric SMRAD51, are unable to promote strand exchange but are capable of binding ssDNA, protecting against nonconservative SSA and A-EJ showing the separation of HR activity from protection against SSA and A-EJ [15, 16].

Consequently, the absence of RAD51 results in HR defects and, concomitantly, the stimulation of the mutagenic repair processes SSA and A-EJ. Therefore, mutations in genes encoding mediators and accessory proteins of HR (such as BRCA1 and BRCA2), which load RAD51 or stabilise the RAD51-ssDNA filament, result in the absence of the RAD51 protein on damaged DNA and consequently in the stimulation of the mutagenic repair pathways SSA and A-EJ [15, 42,43,44,45,46,47,48,49]. This should (i) compensate, at least in part, for the viability lost due to the HR defect and (ii) increase genetic instability.

This raises the question of whether cancer predisposition actually results from HR defects themselves, from the stimulation of nonconservative pathways, or both.

The “RAD51 paradox”

Many HR genes are mutated in cancer, particularly in hereditary breast or ovarian cancer (Table 1). The majority of the genes listed directly or indirectly control HR and DSB signalling (Table 1) [7]. These findings strongly suggest the importance of this pathway in cancer aetiology and support the concept that HR is a tumour suppressor pathway.

Table 1 List of the genes mutated in familial breast or ovarian cancer [7].
Full size table

However, the fact that HR is mutated in cancer cells is contradictory with the fact that HR is essential for cell viability not only during embryonic and foetal development but also for in vitro cultured cell lines [50,51,52,53,54]. One possibility is that HR is not completely abolished and that residual HR activity would be sufficient to maintain cell viability. For example, several mutations of BRCA1 found in tumours reduce but do not completely abolish HR efficiency [55, 56]. Moreover, one can propose that a compensation mechanism(s) can rescue sufficient viability for cancer cells to survive and proliferate. For example, the re-priming of replication through PRIMPOL, downstream the arrested replication fork might resume replication [57].

HR-deficient cells are sensitive to PARP inhibitors [58, 59], which are now used in the clinic. However, resistance to PARP inhibitors occurs frequently, including in tumours that were not previously treated with anticancer agents [60,61,62,63,64]. Moreover, most of the cell lines generated from HR-deficient tumours have rescued efficient HR [65]. These findings underscore the importance and advantages of maintaining some HR activities or exploiting alternative compensatory processes in cancer cells.

Surprisingly, mutations in the gene encoding RAD51 are absent from the list in Table 1, even though it drives the active nucleoprotein molecule responsible for the HR process (i.e., the search for homology and strand exchange of homologous sequences). Indeed, despite extensive investigation, mutations in RAD51 have not been shown to be tumour driver mutations, in contrast to mediator/accessory HR genes. In contrast, the overexpression of RAD51 is associated with a poor prognosis [7, 9,10,11,12,13,14]. This phenomenon constitutes what we previously named “the RAD51 paradox” [7].

Moreover, defects in the HR activity of RAD51 can be dissociated from cancer predisposition. First, germ-line mutations in several HR genes lead to Fanconi anaemia (FA) syndrome. FA is a rare autosomal recessive syndrome associated with bone marrow failure, developmental malformations, and predispositions to acute myeloid leukaemia and cancers. The following HR genes have been found to be mutated in FA: BRCA1 (FA-S), BRCA2 (FA-D1), PALB2 (FA-N), RAD51C (FA-O), BRIP1 (FA-J) and XRCC2 (FA-U) [66]. Mutations in RAD51 have been described in FA-like patients, but while they exhibit many developmental defects, no cancer predisposition has been associated with these RAD51 mutations to date. This contrasts with the other HR genes mutated in FA but shows a similar situation as for breast and ovary cancer and again feeds the “RAD51 paradox” [66,67,68,69,70].

Second, mutations in RAD51 have been described in congenital mirror movement (CMM) syndrome, a hereditary neurodevelopmental disorder affecting pure unimanual and bimanual asymmetric coordinated movements [30, 71]. Remarkably, this syndrome is not associated with cancer predisposition, although the RAD51 mutation affects its capacity to perform strand exchange in vitro [72].

Collectively, these data show that defects in RAD51 do not actually lead to cancer predisposition, challenging the concept that RAD51 is actually a tumour suppressor gene. Note that RAD51 possesses roles that are independent of BRCA1 or BRCA2 and even independent of the strand exchange activity, the non-canonical roles of RAD51 [30], as discussed below. Moreover, these data suggest that alterations in accessory/mediator HR genes can be compensated for viability.

Non canonical roles of RAD51

First, the binding of RAD51 on ssDNA prevents nonconservative mutagenic DNA repair, independently of its strand exchange activity but relies on its ssDNA occupancy [15]. Second, at the arrested replication forks, RAD51 plays several noncanonical roles in the formation, protection, and management of fork reversal, allowing for the resumption of replication [30]. Indeed, replication fork reversal can occur in cells deficient for BRCA2 [33,34,35]. The recruitment of RAD51 on the forks was suggested to be mediated by the direct interaction with DNA polymerase α, RAD54 or RAD51C [32, 37, 73, 74]. Moreover, MCM8 and MCM9, which are involved in the pre-replication complex (pre-RC) formation, DNA replication elongation, have been shown to favour the recruitment of BRCA1 and RAD51 at stalled forks, protecting them from excessive degradation [75]. It is proposed that when the ssDNA tracts are short, no mediator (such as BRCA2) would be required to remove RPA from them. In line with this, fork reversal can occur in cells from a Fanconi patient that express a mutant form of RAD51 RAD51-T131P, which is unable to form RAD51 foci when DNA damage occurs (thus long filament) [35]. Moreover, a mutant form of RAD51, RAD51-II3A, which this unable to catalyze strand exchange, still remains capable of promoting fork reversal [36], exemplifying the dissociation of function between strand exchange (canonical role) and fork reversal activities. Third, overexpression of RAD51-K133R, mutated in the ATP binding/hydrolysis site and defective in HR, in BRCA2-deficient cells restores the resistance of the forks resistant to degradation [31].

Third, RAD51 has also been implicated in post-replicative repair by translesion synthesis in a mechanisms not related to its strand exchange activity [76].

Fourth, RAD51 also exhibits noncanonical roles in RNA-mediated processes. TERRA (telomeric repeat-containing RNA) is a long non-coding RNA that forms R-loops at telomeres [77,78,79]. RAD51 promotes the recruitment of TERRA onto telomeres, but BRCA2 does not play a key role in this process [80]. However, the mutant RAD51-II3A, cannot promote the recruitment of TERRA onto telomeres [80], putting this hypothesis into question. Reactive oxygen species (ROS) can induce the formation of R-loop at the transcription sites. In human cells, the complex RAD51/RAD52/CSB (Cockayne syndrome protein B) detects such R-loops [81], RAD51 playing a role in the R-loops dissolution, independently of BRCA1/2 and the classical HR pathway [81]. More generally, R-loops triggered by transcription conflicts appear to be resolved by RAD51 through mechanisms that are independent from the canonical role of RAD51 in repair [81,82,83].

Finally, the RAD51 pathogenic variants that have been described in the congenital mirror movement syndrome (see above), revealing an unexpected role in brain development, which might be independent of BRCA1 or BRCA2.

Therefore, we could propose that other unexplored noncanonical roles of RAD51 remain to be discovered and characterised. For example, we could speculate that RAD51 might play some unexpected roles at undamaged chromatin, controlling the homoeostasis equilibriums of the nucleus and more generally of the cell.

Hypotheses accounting for the “RAD51 paradox”

One hypothesis accounting, at least in part, for the “RAD51 paradox”, proposes that mutations of mediator/accessory genes result in the absence of RAD51 on damaged DNA, making it susceptible to further alternative nonconservative processes (see above). These alternative processes allow cancer cells to survive but at the cost of increased genetic instability. This hypothesis implies that inhibition of nonconservative pathways should be toxic in BRCA-deficient cells. The simultaneous association of several mechanisms could be responsible for the survival of BRCA2 deficiency. It is possible that all these mechanisms are necessary in parallel and that the inhibition of only one of them would be sufficient to lead to cell lethality in the context of BRCAness. Indeed, the inhibition of either PARP1 or Pol-θ, which play roles in A-EJ [84,85,86,87], or RAD52 that controls SSA [88], is toxic to HR-deficient cells in a synthetic-lethal manner [58, 59, 87, 89,90,91]. SSA requires homologous sequences, but repeat sequences are highly frequent in the human genome, therefore, SSA can occur in many places. In line with this, the EXO1 resection factor has been recently identified synthetic lethal with BRCA1, compared to wild-type or BRCA2-deficient tumours. Moreover, increased SSA-associated genomic scars have been described in BRCA-tumours [92, 93]. Collectively, these findings support the above hypothesis.

RAD51 mutations that permit its binding to DNA should not stimulate alternative mutagenic repair processes and, consequently, should not confer cancer predisposition. Moreover, RAD51 plays several roles, including roles independent of the mediator/accessory proteins (see above); therefore, the suppression of the RAD51 protein, which affects all RAD51 functions, might be too toxic for cancer cells, and alternative nonconservative repair processes would not be able to compensate for the loss of all these different functions.

Collectively, all the above data raise the question of whether cancer predisposition actually results from an inability to perform HR or from the stimulation of alternative nonconservative repair mechanisms.

In vivo specific inactivation of RAD51 HR activity suppressed tumour development

To address the above questions in vivo, we previously designed a mouse model in which RAD51 HR activity is decreased without stimulation of nonconservative pathways (see below).

In vivo mouse models are powerful models for addressing the above questions. Unfortunately, HR genes, including RAD51, are essential genes, and their knockout leads to early embryonic lethality. Many elaborate strategies have been designed to alter HR in vivo, for example, in specific tissues, and confirm the cancer predisposition of HR defects. However, all these models affect mediator/accessory genes, and none of them target RAD51 itself [50]. Therefore, these models confirm the conclusions from human cancer studies, but they have not addressed the questions related to the “RAD51 paradox”.

To overcome these development issues, we designed a mouse model with doxycycline-inducible expression of a RAD51 dominant-negative form (SMRAD51) [16], which poisons HR but does not stimulate alternative nonconservative pathways (SSA, A-EJ) because of its capacity to bind DNA and to inhibit the annealing of complementary strands [15, 16]. Indeed, SMRAD51 can bind ssDNA and assemble into foci. Importantly, the kinetics of assembly and disassembly of SMRAD51 are similar to those of wild-type RAD51. SMRAD51 inhibits strand exchange and HR but still protects against alternative mutagenic pathways SNA and A-EJ (in contrast with other RAD51 mutant forms); moreover, its ability to bind DNA allows it to protect arrested replication forks from degradation but blocks replication restart by strand exchange [15, 16]. Therefore, this mouse model constitutes a unique tool to separate the decrease in HR from the stimulation of alternative mutagenic pathways.

Decreasing HR through doxycycline supplementation led to rapid premature aging that resulted from exhaustion of the stem cell pool, precluding tissue renewal, and associated with systemic inflammation [16]. Remarkably, a decrease in RAD51 strand exchange activity induced by SMRAD51 expression, without SSA or A-EJ stimulation, did not stimulate oncogenesis, and no tumours were observed [16]. These findings do not support an actual tumour suppressor role for HR in vivo.

Moreover, we crossed our mouse model with a breast cancer predisposition model (PyMT, which expresses the mouse mammary tumour virus-polyomavirus middle T antigen). SMRAD51 expression in vivo does not favour tumorigenesis but rather decreases the number and size of breast tumours [16]. These findings indicate that not only HR and RAD51 are not tumour-suppressive factors but in fact favour tumour development in vivo. Notably, SMRAD51 expression led to increased levels of the DNA damage response markers γH2AX (which mainly detects DSBs) and pCHK1 (which are activated upon replication stress), which is consistent with HR defects, in hyperplastic as well as in carcinoma mammary cells [16]. Therefore, RAD51 likely functions at the early step of cancer progression, at pretumorigenic steps (hyperplasia), and should persist at tumorigenic steps (carcinoma). These conclusions are consistent with the replication fork escort role of HR since hyperplastic cells are highly proliferative and thus actively replicate their genome.

Noteworthy, the expression of SMRAD51 in MEFs leads to telomere shortening (data to be published). This can account for the premature ageing phenotype in mice, in addition to the attrition to the stem cell pools. Moreover, this could account for the anti-tumour phenotype in mice, since cancer cells need to maintain telomeres. Therefore, according to this function of telomere maintenance, this supports again a protumour role of functional RAD51.

However, the overexpression of wild-type RAD51 in a similar doxycycline-inducible mouse model did not promote tumorigenesis [16], suggesting that if RAD51 favours oncogenesis, it is not a true oncogene that promotes cancer initiation. Nevertheless, the effects of wild-type RAD51 overexpression on carcinogenesis remain to be explored in the PyMt model of breast cancer predisposition. Additionally, the impact of wild-type RAD51 overexpression following mutagenic genotoxic stresses or treatments that promote cancer initiation should also be addressed.

Collectively, these data suggest that RAD51 promotes progression in the early pretumorigenic state (hyperplasia) and persists at the carcinoma stages.

Discussion

The above data are summarised in Fig. 4A and discussed below.

Fig. 4: The dual roles of HR/RAD51.
figure 4

A RAD51 plays a central role in HR, repairing damaged DNA and coping with replication stress. These findings support cell viability and proliferation, preventing premature aging. Indeed, the division of progenitor stem cells is required to maintain the stem cell pools, enabling tissue renewal. RAD51 and HR escort the replication of progenitor stem cells, helping maintain stem cell pools. In parallel, RAD51 on damaged DNA protects against mutagenic nonconservative repair processes such as SSA or A-EJ. The absence of RAD51 on damaged DNA impairs HR and concomitantly makes damaged DNA accessible for nonconservative repair. This allows partial rescue of cell viability but induces genetic instability. B During cancer progression, cells are highly proliferative, thus replicating their genome in the hyperplastic (nontumor) and carcinogenic steps (tumour). Therefore, they are subjected to high replication stress. HR, through its role as a replication fork escort, allows cells to cope with this high replication stress and thus should facilitate cancer progression as soon as the early nontumorigenic step (hyperplasia).

Full size image

HR/RAD51, a pro-tumour pathway

The loading of Rad51 onto DNA is a prerequisite step for the protection and resumption of arrested replication forks. Therefore, HR/RAD51 is essential for the survival and proliferation of normal dividing cells as well as cancer dividing cells, which are confronted with a series of unavoidable endogenous obstacles that hinder the progression of replication forks during each cell division cycle.

Therefore, HR is essential during embryonic development, and the KO of HR genes leads to early embryonic lethality [50]. Moreover, HR/RAD51 plays a key role in somatic cells as a replication fork escort. In this context, HR plays a prime role in protection against premature aging by maintaining the pool of stem cells that need to divide [16].

The replication back-up role of HR should protect non cancer cells, exhibiting an apparent dual role for HR. Nonetheless, this protective role should also act on tumour cells. This should be even more essential for tumour cells than for primary cells since cancer cells are generally highly proliferative cells, actively replicating their genome and thus are subjected to intense replication stress (in contrast with non-transformed cells). Therefore, through this role on replication stress, HR should help tumour cells cope with this high replication stress. We propose that this is the main reason the pro-tumour role of HR. Thus, there are not two actual opposite faces of HR, but the same face that is an advantage both in non-cancer and in cancer cells. Accounting for its replication escort role, HR should facilitate cancer progression and thus act as a pro-tumour pathway.

The impact of Tp53 on HR is consistent with the above hypothesis. Indeed, the main role of Tp53, which is mutated at a very high frequency in tumours, is protection against cancer through the control of numerous processes, including cell cycle checkpoints, apoptosis and senescence. Therefore, Tp53 favours tumour suppressor pathways and prevents tumour drivers. The tumour suppressor Tp53 does not favour HR but rather restricts HR (for a review), see [94]. These findings suggest that HR is recognised by Tp53 as a tumour support pathway and thus should be restricted.

Dual roles of RAD51

On the one hand, RAD51 (and HR) protects genome stability through its roles in DNA repair and replication fork escort and against mutagenic pathways such as A-EJ and SSA. If we admit that genome stability maintenance prevents oncogenesis, RAD51 (and HR) plays a tumour-suppressive role. Nonetheless, this function of replication back-up is also beneficial for tumour cells. as discussed above. In addition, maintaining genome stability in cancer cells helps them to survive (high genetic instability is toxic) and to maintain the genetic formula that confers them their proliferative advantage. Finally, the protective role of HR may be counterbalanced by the fact that HR can also induce genome instability [95], which could favour cancer initiation.

On the other hand, RAD51 might favour cancer progression through its replication fork protection functions, making it a tumour driver (as discussed above). Therefore, RAD51 has opposite dual faces. As a double-edged sword, HR should thus be tightly controlled. However, HR inherently helps cancer cells proliferate, and consistent with its role in replication escort, HR should act as soon as the replication program is activated, even at the early step of the precancerous step (hyperplasia) of cancer progression.

Concluding remarks

Collectively, these data highlight the dark side of HR and, more specifically, RAD51 as a tumour supports at early pretumorigenic steps; the cancer predisposition associated with HR mutations might not result from HR defects per se but, at least in part, from the concomitant stimulation of alternative nonconservative pathways that compensate for defects in HR.

Figure 4B summarises the dual effects of HR: 1- Prior to cancer initiation, HR potentially functions as a tumour suppressor through its role in maintaining genome stability; however, this function could be counterbalanced by the capacity of HR to induce genetic instability [95]. 2- After initiation of the hyperproliferation program, HR promotes cell proliferation, though its replication fork functions, both at the precancerous stage (hyperplasia) and cancer stage (carcinoma).

Additional work might complete this comprehension, notably how cell-cycle and checkpoint control might impact the processes described above. Moreover, the impact, in vivo, of the genetic alteration of nonconservative repair pathways in different HR-deficient mouse models or to analyse the consequences of the stimulation of nonconservative repair pathways in the absence of a decrease in HR. In order to have a precise view, it is important to identify and characterise all noncanonical roles of RAD51, particularly those that are independent of BRCA1 and BRCA2. Elucidating the defect that causes the CMM syndrome is crucial and should be informative for both the pathology and for academic knowledge about brain development. Moreover, it remains to be explained how HR-deficient cancer cells can survive and proliferate. This might identify novel targets for synthetic lethality strategies in cancer therapy.

More specifically, these findings might have important implications for the design of anticancer strategies. Indeed, they suggest that targeting HR itself could be a promising strategy provided that alternative nonconservative pathways are not induced. Moreover, they identify nonconservative repair pathways that can compensate for HR defects as potential targets for synthetic lethality strategies alternative to PARP or Pol-θ inhibition.

By challenging common views, this review allows us to precisely redefine the actual impact of RAD51 and HR and the importance of balancing alternative nonconservative repair pathways on cancer predisposition/prevention.