DNA crosslink repair safeguards genomic stability during pre-meiotic germ cell development

Abstract

Germline de novo mutations are the basis of evolutionary diversity but also of genetic disease. However, the molecular origin, mechanisms and timing of germline mutagenesis are not fully understood. Here, we define a fundamental role for DNA interstrand crosslink repair in the germline. This repair process is essential for primordial germ cell (PGC) maturation during embryonic development. Inactivation of crosslink repair leads to genetic instability that is restricted to PGCs within the genital ridge during a narrow temporal window. Having successfully activated the PGC transcriptional program, a potent quality control mechanism detects and drives damaged PGCs into apoptosis. These findings therefore define a source of DNA damage and the nature of the subsequent DNA repair response in germ cells, which ensures faithful transmission of the genome between generations.

Introduction

Germ cells are tasked with faithfully transmitting genetic information from one generation to the next¹. Genomic alterations that arise in the germline, known as de novo mutations, can take a variety of forms: from single-nucleotide changes, to insertions and deletions, or large structural rearrangements. The precise mutagenic outcome is determined by the nature of the DNA damage and how it is processed by the repair machinery. Despite considerable knowledge about how the plethora of DNA repair pathways process specific lesions, little is known about the sources of damage or the activity of repair pathways in the mammalian germline.

The earliest mammalian germ cells, known as primordial germ cells (PGCs), emerge during early embryonic development. These cells undergo extensive epigenetic reprogramming before ultimately entering into meiosis². In females, PGCs enter into meiosis during embryonic development but in males the PGCs differentiate into a self-renewing stem cell population that enters meiosis postnatally. Mutations that occur in differentiated germ cells either during spermatogenesis or meiosis are likely confined to an individual offspring. However, mutations that occur in the early PGC population have the potential to be passed to multiple progeny. Therefore, the stage of germ cell development during which de novo mutations arise can play an important role in determining the overall fidelity of genome transmission between generations.

In order to understand the origin of de novo mutations it is also important to understand the molecular mechanisms that give rise to changes in the sequence and structure of the genome. The DNA repair machinery must be tightly regulated because whilst it has the capacity to detect and accurately repair damage to the genome, the DNA repair machinery also has the ability to introduce mutations and structural abnormalities in the genome. One very significant threat to germline genomic stability is meiotic recombination. Failure of meiotic recombination often results in catastrophic karyotypic abnormalities that are incompatible with life. Recently, however, the role of DNA repair proteins in PGCs has become of significant interest as one repair pathway, known as base excision DNA repair, was found to play a key role in epigenetic reprogramming events that occur in PGCs^3–5.

Data from the sequencing of cancer genomes have revealed a surprisingly large spectrum of tissue-specific mutational patterns^6–8. This is likely to represent the interplay between tissue-specific exposure to mutagens and tissue-specific differences in DNA repair capacity. Despite the importance of understanding the origin of germline mutations, little is understood about the sources of DNA damage or repair transactions that occur in the developing germline.

Therefore, significant questions remain about the temporality, source of damage and nature of repair transactions that are active in the germline. These factors ultimately act to shape the evolution of genomes. Here we find that disabling DNA crosslink repair, which is defective in the human disease Fanconi anemia (FA), is critical for the production of viable gametes. We show that crosslink repair is required for embryonic germ cell development prior to entry into meiosis. Loss of crosslink repair leads to genomic instability within the developing PGCs but repair-deficient PGCs are efficiently cleared through apoptosis potentially limiting their ability to pass mutations on to the next generation.

Results

ERCC1 is required for normal fertility

In order to study the role of DNA repair in preventing loss of genetic stability in the germline, we focused on the structure-specific endonuclease XPF-ERCC1. This heterodimeric enzyme cleaves DNA at sites of damage to ensure its accurate repair. XPF-ERCC1 is evolutionary conserved, and plays an important role in sexual reproduction. It is known to regulate the frequency of meiotic crossover in fission yeast, flies and nematode worms, presumably due to its role in the resolution of recombination intermediates^3,9–13. To explore the role of XPF-ERCC1 in mammalian germ cells we generated Ercc1-deficient mice (Ercc1^{tm1a(KOMP)Wtsi}) on a C57BL/6J genetic background. We derived Ercc1^-/- embryonic fibroblasts and found that ERCC1 protein was undetectable and that these cells were hypersensitive to DNA damage (Supplementary Fig. 1a-e). We intercrossed Ercc1^+/- mice and genotyped all living progeny at 14-days old (P14). As previously reported, Ercc1^-/- mice were severely underrepresented (1.9% compared to the expected 25% at P14) and all had died by 21-days old^14,15 (Supplementary Fig. 1f-g).

At P14, we found that there was an obvious reduction in the number of oocytes and spermatogenic cells in the gonads of Ercc1^-/- pups (Fig. 1a-f). Ercc1^-/- females had a 9.5-fold reduction in the number of oocytes compared to congenic wildtype controls (Fig. 1b). Despite this large reduction in the number of oocytes there was only a modest 1.6-fold reduction in the mass of the ovaries (Supplementary Fig. 1h). Strikingly, most seminiferous tubules in Ercc1^-/- males lacked germ cells (Fig. 1c). In agreement with this, Ercc1^-/- males had a significant reduction in the number of cells expressing PLZF, a marker of early spermatogenic cells (6.1-fold, Fig. 1a and d)¹⁶. We noticed that the reduction in the number of PLZF⁺ cells was not uniform among all seminiferous tubules. In the absence of ERCC1, the vast majority of tubules had no PLZF⁺ cells (Fig. 1e), but the few remaining seminiferous tubules contained normal numbers of PLZF⁺ cells compared to wildtype controls (Fig. 1f). This led us to question if the pool of cells that embryonically seeded the seminiferous tubules could have been reduced. We employed mouse vasa homologue (MVH), a marker of germ cells, to measure the numbers of germ cells in the embryonic gonads at day 18.5 of development (E18.5). Both female and male Ercc1^-/- embryos had significant reductions in the numbers of MVH⁺ germ cells compared to wildtype controls (11.1-fold and 9.3-fold, respectively Fig. 1g,h and Supplementary Fig. 2a,b). This was intriguing, as whilst female germ cells have already entered meiosis at E18.5 male germ cells have not, suggesting a common pre-meiotic origin to the germ cell defect in both males and females. We next assessed the E18.5 gonads for evidence of apoptosis. We could not detect a significant increase in the frequency of apoptotic (cleaved-Caspase 3⁺) cells in either Ercc1^-/- male or female gonads (Supplementary Fig. 2c,d). When we looked specifically at germ cells (MVH⁺) we could find no cleaved-Caspase 3⁺ cells (from a total of 149 MVH⁺ germ cells). Taken together these data suggest that the reduction in germ cells observed in both male and female Ercc1^-/- mice precedes meiosis.

Figure 1. Ercc1^-/- mice have reduced numbers of germ cells.

a, Hematoxylin and eosin (H&E)-stained sections of ovaries and H&E- and PLZF(brown)-stained sections of testes at 14 days of age (P14). b, Quantification of follicles per section of P14 ovary (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; n = 6 independent animals per genotype). c, Frequency of Sertoli-cell-only (SCO) tubules per testis at P14 (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; n = 3 independent animals per genotype). d, Quantification and distribution (e) of PLZF⁺ cells per seminiferous tubule at P14 (P value calculated by 2-tailed Mann-Whitney test; data shown represent median and interquartile range; n = 150 tubules per genotype, 50 per mouse). f, Frequency of PLZF⁺ cells per seminiferous tubule excluding tubules with no PLZF⁺ cells in P14 mice (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; n = 150 tubules per genotype, 50 per mouse). g, Representative images of E18.5 ovaries stained for the germ cell marker mouse vasa homologue (MVH) and the frequency of MVH⁺ cells (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.;, each point represents data from one embryo, n = 5 and 3, left to right). h, Representative images of E18.5 testes stained for MVH and the frequency of MVH⁺ cells (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; each point represents data from one embryo, n = 7 and 5, left to right).

A critical role for DNA repair in primordial germ cell development

We therefore set out to test if ERCC1 was required in the earlier PGC pool. We intercrossed Ercc1^+/- mice with a PGC-reporter mouse in which the expression of GFP is driven by a fragment of the Oct4 (Pou5f1) promoter (known as GOF18-GFP) on a C57BL/6J background^17,18. At E13.5 the gonads of Ercc1^-/- embryos had a dramatic reduction in the number of PGCs when compared to wildtype littermates (Fig. 2a). We quantified this defect by flow cytometry and found that in the absence of ERCC1 there was there was a 10.0-fold reduction in the number of PGCs (SSEA1⁺GOF18-GFP⁺) (Fig. 2b,c). At E13.5, female PGCs are entering into meiosis, a process that does not occur in males until after birth. We therefore hypothesized that if the loss of PGCs were due to entry into meiosis, then we would find a female-specific reduction in PGCs. However, we found that both Ercc1^-/- males and females had similar reductions in the number of PGCs at E13.5 (Fig. 2d). Female Ercc1^-/- embryos had an 8.7-fold reduction in the number of PGCs when compared to wildtype female embryos at E13.5. Comparably, ERCC1-deficient males had a 9.2-fold reduction in the number of PGCs compared to wildtype males. This pointed to a common defect in both males and females leading to the reduction in PGCs.

Figure 2. ERCC1 is critical for primordial germ cell development.

a, GOF18-GFP fluorescence in E13.5 (male and female) and E12.5 gonads from wildtype and Ercc1^-/- embryos carrying the GOF18-GFP PGC reporter. b-c, Quantification of PGCs (GOF18-GFP⁺SSEA1⁺) by flow cytometry from wildtype and Ercc1^-/- embryos at E13.5 and E12.5 (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; each point represents data from one embryo, E13.5 n = 19 and 14, E12.5 n = 11 and 7, left to right, Gating strategy shown in Supplementary Fig. 18.) d, PGC number separated by sex at E13.5 and E12.5 (P value calculated by 2-tailed Mann-Whitney test; data shown represent and s.e.m.; each point represents data from one embryo, E13.5 n = 5, 10, 15, and 4, E12.5 n = 6, 3, 5, and 4, left to right). e, Quantification of PGCs (GOF18-GFP⁺SSEA1⁺) throughout embryonic development (E9.5-13.5) in wildtype and Ercc1^-/- embryos (data represent mean and s.e.m.; wildtype n =14, 10, 5, 11 and 21, Ercc1^-/- n = 6, 7, 4, 7 and 14 embryos, left to right.). f, H&E-stained histology from 8-12 week old Ercc1^-/f mice carrying Blimp1-Cre (similar results were obtained from 3 independent animals per genotype and gender). g, Quantification of the cumulative number of offspring over time when male or female Ercc1^-/fBlimp1-Cre⁺ mice or controls were mated with wildtype mice and checked for evidence of copulation (data shown represent mean and s.e.m.; n = 3 mice per genotype).

We next examined the frequency of PGCs at E12.5 by which point neither males nor females have yet entered meiosis. We found that Ercc1^-/- embryos had a 10.6-fold reduction in the frequency of PGCs, respectively (Fig. 2c). Both males and females exhibited comparable reductions in the number PGCs (Fig 2d). To ensure that this defect was not an artefact of the GOF18-GFP reporter we crossed Ercc1^+/- mice with an alternative PGC reporter, Stella-GFP¹⁹, and found that at E12.5 Ercc1^-/- embryos had similar reductions in the number of PGCs with both reporters (Supplementary Fig. 3a).

To determine when during PGC development the defect began we analyzed embryos between E9.5 and E13.5. We found that between E11.5-13.5 Ercc1-deficient embryos had significantly reduced numbers of PGCs compared to wildtype littermates (Fig. 2e, Supplementary Fig. 3b). It is noteworthy that the 10.6-fold reduction first observed in PGCs at E12.5 is similar in magnitude to the reduction in MVH⁺ germ cells at E18.5. This strongly suggests that the major reduction of germ cells in Ercc1^-/- mice occurred before E12.5.

Finally, we crossed a conditional allele of Ercc1 with a Cre recombinase expressed in early PGCs (Blimp1-Cre²⁰). We found that loss of ERCC1 in PGCs resulted in an almost complete loss of gamete production in both males and females (Fig. 2f). The histological appearance was consistent with the defects observed in 14-day old constitutive Ercc1^-/- pups. 8- to 12-week old Ercc1^-/fBlimp1-Cre⁺ males showed a significant reduction of PLZF⁺ cells per tubule (Supplementary Fig. 3c and d) similar in magnitude to the reduction observed in Ercc1^-/- constitutive knockout mice at P14 (7.3-fold and 6.1-fold, respectively). When either male or female Ercc1^-/fBlimp1-Cre⁺ mice were crossed with wildtype mates they failed to give rise to offspring despite evidence of copulation (Fig. 2g, Supplementary Fig. 3e and f). These data show that there is a critical requirement for ERCC1 in preserving normal numbers of PGCs prior to entry into meiosis.

ERCC1 acts with the Fanconi pathway to preserve PGC development

XPF-ERCC1 plays a critical role in multiple DNA repair transactions. It is essential for both global genome and transcription-coupled nucleotide excision repair (GG- and TC-NER). XPF-ERCC1 also plays an important role in a subset of homologous recombination transactions²¹. In addition, it is also a critical factor in the Fanconi-mediated repair of DNA interstrand crosslinks (ICLs)^22,23. It is therefore possible that the PGC defect observed in Ercc1^-/- mice is due to loss of one of these individual repair functions or the simultaneous inactivation of multiple routes of DNA repair.

In order to discriminate between these possibilities, we decided to focus our attention on pathway-specific factors. FANCA is a component of the Fanconi-mediated DNA interstrand crosslink repair pathway while XPA interacts with XPF-ERCC1 segregating its function in NER. We first assessed the expression of Ercc1, Fanca and Xpa in E12.5 PGCs by qRT-PCR (Fig. 3a). We found that the levels of Ercc1, Fanca and Xpa expression were higher in PGCs compared to surrounding somatic cells (SSEA1^-GOF18-GFP^-). The expression of both Ercc1 and Fanca increased similarly during PGC development (E8.5-E12.5), peaking at E10.5, whilst the expression of Xpa was relatively constant (Fig. 3b).

Figure 3. ERCC1 acts together with the Fanconi anemia pathway to preserve PGC development.

a, qRT-PCR expression analysis of the DNA repair genes Ercc1, Xpa and Fanca in surrounding somatic cells (SSEA1^-GOF18-GFP^-) and PGCs (SSEA1⁺GOF18-GFP⁺) purified from E12.5 wildtype embryos (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; each point represents data from one embryo, n = 4, 4, 3, ,3 ,3 and 3, left to right). b, qRT-PCR expression analysis of Ercc1, Xpa, Fanca and the germ cell marker Stella in PGCs (SSEA1⁺GOF18-GFP⁺) during development (E8.5-12.5) (data shown represent mean and s.e.m.; n = 3 embryos per genotype per timepoint). c, H&E-stained sections of p14 mice (3 independent animals of each genotype and gender were shown to have similar histology). d, Representative images of GOF18-GFP fluorescence in E12.5 gonads (representative data from 3 independent animals per genotype). e, Quantification of PGCs (SSEA1⁺GOF18-GFP⁺) by flow cytometry from wildtype, Ercc1^-/- and Fanca^-/- embryos at E12.5 (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; each point represents data from one embryo, n = 11, 7, 14 and 10, left to right). f, Quantification of PGCs (SSEA1⁺GOF18-GFP⁺) during embryonic development (E9.5-13.5) by flow cytometry (data shown represent mean and s.e.m.; wildtype n = 25, 19, 15, 25 and 25, Ercc1^-/- n=6, 7, 4, 7, and 14, Fanca^-/- n = 7, 6, 13, 10 and 9 embryos, left to right). g, Generation and quantification of PGCs per gonad of Ercc1^-/-Fanca^-/- E12.5 embryos (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; each point represents data from one embryo, n = 37, 7, 6 and 6, left to right).

We compared the phenotypes of mice deficient in ERCC1 with those deficient in either FANCA or XPA. The ovaries and testes of Xpa^-/- mice were examined at P14 and found to be indistinguishable from wildtype controls (Fig. 3c). In contrast, Fanca^-/- males and females showed little evidence of spermato- or oogenesis, reminiscent of Ercc1^-/- mice (Fig. 3c). We went on to examine the phenotypes of sexually mature mice (8-12 weeks old), but employed conditional Ercc1^-/fBlimp1-Cre⁺ mice as the constitutive knockout mice die before reaching sexual maturity. To fully exclude a role of NER we extended our analysis to Csb^-/- mice (deficient in TC-NER). We found both XPA- and CSB-deficient adults were fertile (Supplementary Fig. 4a and b). In contrast, FANCA-deficient mice were infertile with ovaries and testes that were largely devoid of germ cells (Supplementary Fig. 4a-e). These histological features and fertility defects closely resemble what was observed in Ercc1^-/fBlimp1-Cre⁺ mice (Supplementary Fig. 4c-e).

These data suggested that it is XPF-ERCC1’s role in crosslink repair that is critical to ensure normal germ cell development. Fertility defects in mice and humans deficient in the Fanconi-mediated crosslink repair pathway have been documented ^24–27. Previous studies have suggested important roles for several Fanconi proteins in both meiosis and PGC proliferation, but the basis of this infertility remains poorly understood^{24,25,28–32}. Although the data presented here clearly show that in the absence of ERCC1 a PGC defect precedes the onset of meiosis, we did analyze meiosis in mice deficient in either ERCC1 or FANCA (Supplementary Note and Supplementary Figs. 5 and 6). These data reveal that whilst the loss of crosslink repair initially delays progression through meiosis it is not essential for the production of functional gametes^33–37.

In the absence of ERCC1, germ cell defects are observed at E11.5, prior to meiosis. To test if FANCA, like ERCC1, is required for pre-meiotic PGC development we generated mouse embryonic stem cells (mESCs) that lacked ERCC1 or FANCA and carried the GOF18-GFP reporter (Supplementary Fig. 7a and b). In the absence of ERCC1 or FANCA mESCs were defective at differentiating into primordial germ cell-like cells (PGCLCs) in vitro (Supplementary Fig. 7c and d). It is worth noting that day 4 PGCLCs are likely to correspond to early stages of PGC development^38–41. It is therefore interesting that we find a requirement for crosslink repair in PGCLC formation given that we observe only a modest reduction in the number of PGCs at early time points. This highlights the importance of in vivo models to study the role of DNA repair in PGC development.

We therefore set out to test the requirement for FANCA in vivo by crossing Fanca^+/- mice with the GOF18-GFP reporter. We examined the gonads of E12.5 Fanca^-/- embryos and observed a significant 7.9-fold reduction in the numbers of PGCs when compared to wildtype littermates (Fig. 3d and e). The number of PGCs observed in Fanca^-/- embryos was also comparable to the number observed in Ercc1^-/- embryos (Fig. 3e) and no difference could be found in the magnitude of the PGC defect between males and females (Supplementary Fig. 8a and b). Similar to ERCC1-deficiency we observed that the primordial germ cell defect emerged between E10.5 and E12.5 (Fig. 3f).

To test if ERCC1 and FANCA act together to preserve the number of PGCs we generated Ercc1^-/-Fanca^-/- embryos. We found that loss of FANCA did not further exacerbate the PGC defect observed in Ercc1^-/- embryos at day E12.5, revealing that ERCC1 and FANCA work in the same pathway to preserve PGC development (Fig. 3g).

This suggests that there may be a particular requirement for DNA ICL repair during PGC development. To test if the requirement was generalizable to other factors involved in ICL repair we extended our analysis to SLX4 and FAN1. SLX4 directly interacts with XPF-ERCC1 to promote incisions at sites of crosslinked DNA, whilst FAN1 acts in a parallel pathway to XPF-ERCC1^23,42–44. We found that whilst SLX4-deficient mice have a profound fertility defect akin to Fanca^-/- and Ercc1^-/fBlimp1-Cre⁺, Fan1^-/- mice were fertile (Supplementary Fig. 9a and b). At E12.5 Fan1^-/- embryos had an intact germ cell compartment (Supplementary Fig. 9c). This shows that whilst Fanconi-mediated DNA repair is required for germ cell development this requirement is not generalizable to all routes of ICL repair.

Surviving crosslink-repair-deficient PGCs undergo specification and epigenetic reprogramming

In order to determine the fate of FA crosslink repair-deficient PGCs we examined their phenotype in more detail. We measured the expression of PGC markers in early (E10.5) and late (E12.5) genital ridge PGCs. We found that in the absence of crosslink repair PGCs had indistinguishable expression of early (Ap2γ, Prdm14, Blimp-1 and Stella) and late (Mvh and Dazl) PGC markers when compared to wildtype PGCs (Fig. 4a, Supplementary Fig. 9d)^45,46. This suggests that the PGC specification program is intact in the remaining crosslink-repair-deficient PGCs.

Figure 4. Crosslink-repair-deficient PGCs undergo normal PGC transcriptional activation and epigenetic reprogramming.

a, qRT-PCR expression analysis of early (Ap2γ, Prdm14, Blimp-1 and Stella) and late (Mvh and Dazl) germ cell markers in PGCs (SSEA1⁺GOF18-GFP⁺) purified from E10.5 and E12.5 wildtype, Ercc1^-/- and Fanca^-/- embryos (data represent mean and s.e.m.; n = 3 embryos per genotype and timepoint, expression was normalized to Gapdh and made relative to the wildtype PGC sample at E12.5 for each gene). Immunofluorescence (b) staining and quantification (c) of global H3K27me3 and H3K9me2 levels in PGCs (GOF18-GFP⁺) and somatic tissue (GOF18-GFP^-) in E12.5 genital ridges (P value calculated by 2-tailed Mann-Whitney test; data shown represent mean and s.e.m.; each point represents data from one embryo, H3K27me3 n = 4, 5, 2, and 2, H3K9me2 n = 4, 5, 2, and 2, left to right). d, Quantification of DNA methylation of Tex19.1, Mili,, H19 and Line1 transposable elements in fluorescence-activated cell sorted (FACS) PGCs (SSEA1⁺GOF18-GFP⁺) and surrounding somatic cells (SSEA1^-GOF18-GFP^-) (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; each point represents data from one embryo, Tex19.1 n = 9, 3, 3 and 3, Mili n = 9, 3, 4 and 3, H19 n = 10, 3, 3 and 3, Line1 n = 7, 3, 3, and 3, left to right). e, Representative methylation patterns of Tex19.1, Mili, H19 DMR and Line1 in PGCs (SSEA1⁺GOF18-GFP⁺, top) and somatic tissue (SSEA1^-GOF18-GFP^-, bottom) (closed = methylated CpG, open = unmethylated CpG).

We set out to assess if crosslink repair deficiency perturbed PGC migration^47–49. Firstly, we found that when quantified microscopically the PGC defect in ERCC1-deficient embryos was consistent with our findings by flow cytometry (Supplementary Figs. 10, 11 and Fig. 2e). Secondly, at E9.5 and E10.5 Ercc1^-/- PGCs (GOF18-GFP⁺) were similarly distributed compared to wildtype controls, with no evidence of an increase in the frequency of ectopically localized PGCs (Supplementary Fig. 10a-g). Furthermore, at E11.5 the few remaining Ercc1^-/- PGCs were exclusively located within the genital ridge (Supplementary Fig. 11a-c).

In the genital ridge PGCs undergo a process of epigenetic reprogramming, we therefore asked if key events in this process occurred normally in crosslink-repair-deficient PGCs. Early epigenetic reprogramming in PGCs results in altered histone modifications. We assessed PGCs for trimethylation of histone H3 at lysine 27 (H3K27me3 – high levels of which are acquired in PGCs) and dimethylation of histone H3 at lysine 9 (H3K9me2 – which is erased in PGCs)^5,50,51. We found no difference in either H3K27me3 or H3K9me2 at a global level in crosslink-repair-deficient PGCs when compared to wildtype controls (Fig. 4b,c). We went on to determine if DNA demethylation and imprint erasure occurred in the absence of DNA crosslink repair. We assessed the level of DNA methylation at a series of genomic loci. PGCs (SSEA1⁺GOF18-GFP⁺) and surrounding somatic cells (SSEA1^-GOF18-GFP^-) were isolated by flow cytometry from E12.5 embryos and used for sodium bisulphite sequencing. We assessed the methylation of Tex19.1 and Mili (germline factors dependent on promoter demethylation), H19 (an imprinted locus) and Line1 (transposable element) ^46,52,53. In all cases we found that at E12.5 Ercc1^-/- and Fanca^-/- PGCs had erased methylation to the same extent as wildtype PGCs (Fig. 4d,e, Supplementary Figs. 12 and 13). Together these data suggest that in the absence of crosslink repair key events in PGC specification and epigenetic reprogramming occur normally in the remaining PGCs.

Repair-deficient PGCs accumulate unrepaired DNA double strand breaks

We hypothesized that in the absence of crosslink repair PGCs may accumulate unrepaired DNA damage. As the reduction in the number of PGCs occurred between E10.5-E12.5 we chose to assess E11.5 embryos for evidence of DNA damage. We measured 53BP1 (a surrogate marker of DNA double strand breaks, DSBs) and γ-H2A.X focus formation (an alternative, albeit less specific, marker of DSBs). In the absence of either ERCC1 or FANCA a greater proportion of PGCs showed elevated levels of DSB markers (Fig. 5a and b). To directly measure DSBs in PGCs, we employed single-cell gel electrophoresis on FACS-purified PGCs from E11.5 Ercc1^-/-, Fanca^-/- and wildtype embryos. In the absence of either FANCA or ERCC1 there was a significant induction in comet tail moment (Fig. 5c). To assess if these PGCs were undergoing replication, we pulsed pregnant females with EdU and found that both Ercc1^-/- and Fanca^-/- PGCs incorporated EdU to a similar extend as wildtype PGCs (Supplementary Fig. 14a).

Figure 5. PGCs accumulate unrepaired DNA breaks.

a, Representative images of 53BP1 foci in PGCs (GOF18-GFP⁺) from embryos at E11.5 and quantification of the frequency of PGCs with > 5 53BP1 foci per nucleus (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; at least 100 PGCs were scored per embryo, each point represents data from one embryo, n = 5, 3 and 3, left to right). b, Representative images of γ-H2A.X foci in PGCs (GOF18-GFP⁺) at E11.5 and quantification of the frequency of PGCs with > 10 γ-H2A.X foci per nucleus (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; at least 100 PGCs were scored per embryo, each point represents data from one embryo, n = 4, 4 and 5, left to right). c, Single-cell electrophoresis of FACS-sorted PGCs (SSEA1⁺GOF18-GFP⁺) at E11.5 (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; at least 100 PGCs were scored per embryo, n = 3 embryos per genotype). d, Quantification of PGCs (GOF18-GFP⁺) with > 5 γ-H2A.X foci per nucleus in wildtype or Fanca^-/- embryos during embryonic development (E9.5-12.5). (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; each point represents data from one embryo, n = 4, 5, 2, 2, 3, 2, 4 and 7, left to right). e, Quantification of PGCs (SSEA1⁺GOF18-GFP⁺) by flow cytometry in wildtype and Fanca^-/- embryos during embryonic development (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; each point represents data form one embryo, n = 13, 13, 8, 6, 9, 13, 14 and 28, left to right).

Finally, we charted the accumulation of DSB formation during PGC development in Fanca^-/- embryos. We found that E11.5 embryos had the highest frequency of γ-H2A.X⁺ PGCs (Fig. 5d). Interestingly, the maximum induction of DNA damage coincides with the period during which the reduction in PGC number occurred (Fig. 5d,e). DNA damage did not correlate with the global levels of H3K27me3 or H3K9me2 levels in PGCs (Supplementary Fig. 14b-e). These data show that crosslink-repair-deficient PGCs are lost within a narrow temporal window during which they accumulate unrepaired DNA DSBs.

Aldehyde detoxification preserves germ cell genome stability

Although we have shown that the Fanconi pathway and XPF-ERCC1 act together to prevent the loss of genome stability in PGCs, it is unclear what causes the initial DNA damage. It has recently been reported that metabolically produced aldehydes can cause mutagenesis and cell death within Fanconi-deficient hematopoietic stem cells^54–58. Reactive aldehydes (such as formaldehyde and acetaldehyde) can cause DNA interstrand crosslinks but can also cause a plethora of other DNA lesions, including base adducts, intrastrand- and protein-DNA crosslinks.

We therefore set out to test if sufficient acetaldehyde or formaldehyde is generated during fetal development to perturb Fanca^-/- PGC development. In order to do this we employed mice deficient in either ALDH2 (required for acetaldehyde catabolism) or ADH5 (required for formaldehyde catabolism.) This allowed us to generate embryos deficient in both Fanconi-mediated crosslink repair and aldehyde catabolism, Aldh2^-/-Fanca^-/- or Adh5^-/-Fanca^-/- that carried the GOF18-GFP reporter (Fig. 6a). The gonads of E12.5 Aldh2^-/-Fanca^-/- and Adh5^-/-Fanca^-/- embryos had fewer PGCs compared to the Fanca^-/- single mutant (Fig. 6b-d).This further reduction in the number of PGCs is accompanied by a concomitant increase in the proportion of PGCs carrying markers of damaged DNA (Fig. 6e).

Figure 6. Fetal aldehyde catabolism protects developing PGCs.

a, Generation of embryos that lack both aldehyde detoxification (Aldh2^-/- or Adh5^-/-) and Fanconi-mediated crosslink repair (Fanca^-/-) carrying the Gof18-GFP PGC reporter. b, Representative images of GOF18-GFP fluorescence in E12.5 gonads. c, Representative flow cytometry plots of PGCs (SSEA1⁺GOF18-GFP⁺) from E12.5 embryos. d, Quantification of PGCs by flow cytometry from E12.5 embryos (P value calculated by 2-tailed Mann-Whitney test; data represent mean and s.e.m.; each point represents data from one embryo Aldh2^-/-Fanca^-/- n = 60, 30, 8 and 8, Adh5^-/-Fanca^-/- n = 35, 8, 11 and 6, left to right). e, Quantification of PGCs (GOF18-GFP⁺) with > 10 γ-H2A.X foci per nucleus in E11.5 embryos lacking both acetaldehyde catabolism and ICL repair (Aldh2^-/-Fanca^-/-) or formaldehyde catabolism and ICL repair (Adh5^-/-Fanca^-/-) at E11.5 (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; each point represents data from one embryo, Aldh2^-/-Fanca^-/- n = 4, 6, 8 and 3, Adh5^-/-Fanca^-/- n = 4, 3, 8 and 4, left to right).

It is worth noting that ALDH2 and ADH5 cannot compensate for each other in protecting developing germ cells from reactive aldehydes. These data show that sufficient aldehydes (substrates of ALDH2 or ADH5) are generated within the developing fetus to compromise the PGC population at least when the FA pathway is defective, defining endogenous reactive aldehydes as mutagens that can corrupt the genome of the developing germ line. As 540 million people carry inactivating mutations of Aldh2, exposure of developing germ cells to elevated levels of acetaldehyde may be common in the human population⁵⁹.

Damaged PGCs are efficiently cleared through apoptosis

We next questioned if the accumulation of damaged DNA is specific to PGCs or is widespread in FA crosslink-repair-deficient embryos. We assessed γ-H2A.X and 53BP1 focus formation in both the PGCs (GOF18-GFP⁺) and also surrounding somatic cells (GOF18-GFP^-) of the developing gonad at E11.5. As already shown, both Ercc1^-/- and Fanca^-/- embryos had significant increases in the frequency of γ-H2A.X- or 53BP1-positive PGCs (Fig. 5a and b). Surprisingly, in somatic cells there was no difference in the accumulation of these DNA DSB markers between crosslink repair-deficient and wildtype littermates (Fig. 7a and b).

Figure 7. Damaged PGCs are eliminated by apoptosis.

a-b, Quantification of the frequency of PGCs (GOF18-GFP⁺) and surrounding somatic cells (GOF18-GFP^-) with (a) > 5 53BP1 or (b) > 10 γ–H2A.X foci per nucleus at E11.5. (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; at least 100 PGCs or somatic cells were scored per embryo, each point represents data from one embryo, 53BP1 n = 4, 3, 3, 5, 3 and 3, γ–H2A.X n = 3, 3, 3, 4, 4 and 5, left to right). c, Representative images of E11.5 gonads stained for phospho-Ser15-p53 (pSer15-p53) and GFP (GOF18-GFP). Quantification of PGCs (GOF18-GFP⁺) and somatic cells (GOF18-GFP^-) that stain positively for phospho-Ser15-p53 (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m.; at least 100 PGCs or somatic cells were scored per embryo, each point represents data from one embryo, n = 6, 3, 2, 8, 4 and 7, left to right). d, Representative images of E11.5 gonads stained for the marker of apoptosis cleaved-Caspase 3 and GFP (GOF18-GFP) and the quantification of the frequency of PGCs (GOF18-GFP⁺) and somatic cells (GOF18-GFP^-) that stain positively for cleaved-Caspase 3 (P value calculated by unpaired 2-tailed t test; data represent mean and s.e.m., at least 100 PGCs or somatic cells were scored per embryo, each point represents data from one embryo, n = 10, 3, 3, 8, 4 and 4, left to right).

The failure to accurately repair DNA DSBs leads to the loss of genome stability and mutation. We therefore wanted to ask how PGCs are prevented from passing on mutagenized genetic material to the next generation. We found that in the absence of either FANCA or ERCC1, PGCs accumulated a phosphorylated form of p53 (pSer15-p53) (Fig. 7c). Given the dramatic reduction in PGC number we hypothesized that they may be lost through p53-dependent apoptosis. Ercc1^-/- and Fanca^-/- embryos had increased frequency of apoptotic PGCs compared to littermates (cleaved-Caspase 3⁺, 7.7-fold and 7.4-fold, respectively) (Fig. 7d). Of the PGCs that carried evidence of unrepaired DNA DSBs (persistent γ-H2A.X) we found that a large proportion of these cells were also positive for pSer15-p53 (Supplementary Fig. 15). Together these data show that crosslink repair-deficient PGCs accumulate DNA damage, pSer15-p53, and enter into apoptosis.

Discussion

Germ cells are single-handedly tasked with passing genetic information from one generation to the next, it is therefore of paramount importance that these cells maintain the integrity of their genomes. We use constitutive knockout mice to show that in the absence of ERCC1 or FANCA germ cell defects precede meiosis. The loss of PGCs occurs between E10.5 and E12.5, which is before either male or female germ cells enter into meiosis. The results presented here reveal that in the absence of FA crosslink repair PGCs accumulate DSBs, undergo apoptosis and are lost prior to entry into meiosis. This sheds new light on the nature of DNA damage, repair processes and quality control mechanisms that prevent de novo germline mutagenesis.

Whilst XPF-ERCC1 is required for many different DNA repair transactions, FANCA is restricted to a role in the Fanconi ICL repair pathway. We find that the magnitude and temporality of PGC loss is similar between Ercc1^-/- and Fanca^-/- embryos. Most strikingly, we find that Ercc1^-/-Fanca^-/- embryos have a PGC defect that is indistinguishable from either of the single mutants. These data reveal that XPF-ERCC1 acts together with the Fanconi DNA repair pathway in order to ensure normal PGC development. In contrast, whilst FAN1 is required to maintain cellular resistance to crosslinking agents, we find that it is dispensable for PGC development. Hence, PGCs have a specific requirement for Fanconi-mediated DNA interstrand crosslink repair to ensure normal development.

Crosslink-repair-deficient PGCs accumulate unresolved DNA breaks that can lead to cell death or, if illegitimately repaired, insertions, deletions or structural changes. Therefore, the role of DNA crosslink repair in suppressing germline DNA DSBs is critical to maintain germline genome stability. It is worth noting that whilst crosslink-repair-deficient PGCs accumulate unresolved DSBs, the surrounding somatic cells do not exhibit elevated levels of DNA damage. It has been suggested that PGCs have different responses to DNA damage compared to somatic cells⁶⁰, however, the fact that PGCs specifically accumulate damaged DNA strongly suggests the PGCs face unique threats to their genetic integrity, necessitating crosslink repair. The observation that crosslink-repair-deficient PGCs accumulate stabilized p53 and enter into apoptosis during the period in which they exhibit elevated levels of DNA damage may provide a simple mechanism to prevent the transmission of corrupted genetic information to the next generation (Supplementary Fig. 16).

Crosslink repair is crucial during a narrow temporal window (E10.15-E12.5) coinciding with PGC epigenetic reprogramming and expansion. It is plausible that the epigenetic reprogramming could in-itself pose a threat to genome integrity when crosslink repair is inactivated. It is worth noting that in the absence of FA crosslink repair there is a significant defect in the ability to generate induced pluripotent stem cells (iPSCs)^61,62. Although by no means the same process, iPSC reprogramming does share some similarities with PGC development (activation of pluripotency markers, DNA demethylation and alterations to histone modifications)⁶³. DNA demethylation, histone demethylation or retrotransposon reactivation may pose threats to genetic integrity during PGC reprogramming necessitating the FA pathway^3,64. However, we find that in the absence of crosslink repair, the remaining PGCs have undergone the major landmarks of epigenetic reprogramming. Alternatively, PGC epigenetic reprogramming may induce a unique genomic state which when compounded by crosslink repair deficiency results in cellular toxicity. Going forward, it will be critical to uncover if there are other sources of DNA damage that pose unique threats to the genetic integrity of PGCs.

There are human patients with mutations that lead to defective FA-mediated DNA crosslink repair. In agreement with our mouse models, both male and female FA patients (including those with mutations in FANCA and FANCQ/XPF) have fertility defects^{15,22,26,27,65–68}. Despite limited studies of fertility in human patients both males and females exhibit hypogonadism with hormonal dysregulation. Females have been shown to have primary ovarian insufficiency with premature menopause. The male patients reveal Sertoli cell-only tubules, the same pathological feature that we find in our mouse models. The phenotypes reported in FA patients are consistent with germ cell loss due to failure of PGC development. This study provides a plausible explanation for the infertility observed in human patients with FA. More generally, this work suggests that the process of embryonic germ cell development may pose unique challenges to the DNA damage repair machinery.

Methods

Mice

All animal experiments undertaken in this study were with approval of the MRC Laboratory of Molecular Biology animal welfare and ethical review body and the UK Home Office under the Animal [Scientific Procedures] Act 1986 license 70/8325. All mice were maintained under specific pathogen-free conditions in individually ventilated cages (Techniplast GM500, Techniplast) on Ligno-cel FS14 spruce bedding (IPS, Ltd.) with environmental enrichment (fun tunnel, chew stick, and Enviro-Dri nesting material (LBS)) at 19-23 °C with light from 7:00 a.m. to 7:00 p.m. and fed Dietex CRM pellets (Special Diet Services) ad libitum. No animals were wild and no field collected samples were used. Unless otherwise stated, mice were maintained on a C57BL/6J background. Embryos were used at various stages from E8.5-E18.5 as indicated in the text. Samples were collected from animals at 14 days postpartum or at 8-12 weeks as specified in the text. Females used in timed mating experiments were aged between 6 and 18 weeks. The investigators were blinded to the genotypes of animals throughout the study and data were acquired by relying purely on identification numbers. Fanca^{tm1a(EUCOMM)Wtsi}; MGI ID: 4434431, C57BL/6N, the conditional Fanca^{tm1c(EUCOMM)Wtsi} allele, and the Aldh2^{tm1a(EUCOMM)Wtsi}; MGI:4431566 allele have been described previously^54,58. Fan1^{tm1a(KOMP)Wtsi}; MGI ID: 4940765, C57BL/6N/129S4 were a kind gift from F Hildebrandt⁴⁴. Xpa^tm1Gvh; MGI ID: 5312129, C57BL/6 and Ercc6^tm1Gvh (Csb); MGI ID: 1932102, C57BL/6 mice were a kind gift from G.T. van der Horst^69,70. Stella-GFP (Tg(Dppa3/EGFP)6-25Masu); MGI:5519126 mice were a kind gift from Azim Surani¹⁹. GOF18-GFP (Tg(Pou5f1-EGFP)2Mnn); MGI:3057158; JAX:004654, and Stra8-iCre (Tg(Stra8-icre)1Reb/J); MGI:3779079; Jax:008208 mice were purchased from the Jackson Laboratory.

Generation of Ercc1-deficient mice

Ercc1-mutant mice were generated from C57BL/6 ES cells (Ercc1^{tm1a(KOMP)Wtsi};MGI:4362172; Clone G12), carrying a targeted disruption of the Ercc1 locus, obtained from the EUCOMM consortium. These mice were backcrossed onto a C57BL/6J background. Targeting was confirmed by long-range PCR at both the 5’ and 3’ using oligonucleotide pairs LAR3+Ercc1GF3, and RAF5+Ercc1GR3 respectively. Germline transmission was achieved and progeny were genotyped using ER1F2, ER1R1 and En2a oligonucleotides. This allele was made conditional by crossing with mice transgenic for the Flp-e recombinase⁷¹ to generate the Ercc1^{tm1c(KOMP)Wtsi} allele. Cre recombinase led to the excision of exon 5 resulting in the Ercc1^{tm1d(KOMP)Wtsi} allele. The Ercc1^{tm1d(KOMP)Wtsi} allele was determined by genotyping using PCR with ER1NF and ER1NR oligonucleotides. Sequence of oligonucleotides are provided in Supplementary Table 1.

Isolation and characterization of Ercc1^-/- mouse embryonic fibroblasts

Timed matings were performed between Ercc1 heterozygous mice and pregnant females were culled by cervical dislocation at E12.5. Primary MEF cultures were obtained and immortalized using the SV40 large T antigen as described previously²³. Briefly, pBABE-SV40-Puro virus was produced using Plat-E cells (Cell Biolabs). 48 hours following transfection, the culture medium containing retroviruses was collected and passed through a 0.22 μm filter. The filtered retrovirus was mixed 1:1 with complete medium and supplemented with 4 μg.ml^-1 polybrene, the resulting infective medium was added to primary MEF cultures. Transformed clones were selected for 10 days using 3.5 μg.ml^-1 puromycin. Sensitivity to DNA-damaging agents was determined by seeding 500 transformed MEFs per well of a 96-well flat-bottom plate and exposed to mitomycin c (MMC) or ultraviolet (UV) irradiation. After 10 days of culture post-exposure the MTS cell viability reagent (Promega) was added and plates incubated at 37 °C for 4 hours, and absorbance at 492 nm was measured.

Histological analysis

Tissues were fixed in 10% neutral-buffered formalin for 24-36 hours and transferred to 70% ethanol. Fixed samples were embedded in paraffin and 4 μm sections cut, deparaffinized, rehydrated and stained with hematoxylin and eosin following standard methods. Images were captured using a Zeiss AxioPlan 2 microscope (Zeiss) and tissue architecture was scored blindly.

Immunohistochemistry

Sections of formalin-fixed paraffin-embedded samples were deparaffinized and rehydrated following standard methods. Slides were placed into antigen retrieval buffer (10 mM sodium citrate, pH 6.0) and boiled for 10 minutes. Slides were allowed to cool to room temperature before being washed three times in water for 5 minutes and then once in TBS, 0.1% w/v Tween-20 for 5 minutes. A large hydrophobic ring was drawn around the tissue and the samples incubated in blocking buffer (TBS, 5% v/v goat serum, 0.1% w/v Tween-20) for 1 hour at room temperature. Samples were incubated at 4 °C overnight with the following primary antibodies diluted in blocking buffer; anti-MVH 1:100 (Abcam, ab27591), anti-PLZF 1:200 (Santa Cruz Biotechnology, sc-28319), anti-ERCC1 1:50 (Cell Signaling Technology, D6G6) and anti-cleaved-Caspase 3 1:400 (Cell Signaling Technology, 9661). Slides were washed three times with TBS, 0.1% w/v Tween-20 and then incubated with the following secondary antibodies diluted in blocking buffer for 1 hour at room temperature; goat anti-mouse Alexa Fluor 488 1:500 (A11029, Life Technologies), goat anti-rabbit Alexa Fluor 594 1:1,000 (A11037, Life Technologies), swine anti-rabbit or anti-mouse HRP-conjugated immunoglobulins 1:200 (Dako). Afterwards, slides were washed three times in TBS, 0.1% w/v Tween-20 for 5 minutes. For fluorescence-based immunohistochemistry, slides were incubated with 0.5 μg.ml^-1 4’-6-diamidino-2-phenylindole (DAPI) diluted in PBS for 10 minutes. Slides were then washed once in water and mounted with ProLong Gold antifade reagent (P36934, Molecular Probes) and coverslips placed on top of the slides and allowed to cure for 48 hours. Images were captured using a Zeiss LSM 780 confocal microscope (Zeiss). The frequency of MVH-positive cells was scored blindly. For HRP-based immunohistochemistry, slides were incubated for 3 minutes with diaminobenzidine (Cell Signaling Technology, 8059P) and then washed once in water. Samples were counterstained with 50% hematoxylin diluted in water for 5 minutes. Afterwards slides were washed 2 times in water for 5 minutes and dehydrated in an ethanol gradient and finally xylene before being mounted with D.P.X (Sigma, 317616) and coverslips placed onto slides. Images were captured using a Zeiss AxioPlan 2 microscope (Zeiss). The frequency of PLZF-positive cells and presence of ERCC1 were scored blindly.

Preparation and spreading of meiotic cells

The spreading and staining of meiotic cells was performed as described previously with some modifications^33–37,72. Briefly, the seminiferous tubules of 21-day old (P21) or 50-day old (P50) male mice or the ovaries of embryonic day 16.5 (E16.5) or 18.5 (E18.5) female fetuses were incubated in hypotonic extraction buffer (30 mM Tris, 50 mM sucrose, 17 mM trisodium citrate dehydrate, 5 mM ethylenediaminetetraacetic acid (EDTA), 0.5 mM Dithiothreitol (DTT) 0.5 mM phenylmethane sulfonyl fluoride, pH 8.2) for 1 hour at room temperature with gentle agitation. Samples were allowed to settle under gravity and the supernatant discarded. Samples were placed onto a glass slide and excess buffer removed. Samples were then disaggregated in 100 μl of 100 mM sucrose solution, pH 8.2, using a fine pair of forceps and a scalpel. The cell suspension was made up to a final volume of 200 μl (100 mM sucrose, pH 8.2), and dropped from a height of 0.5 m onto glass slides coated in 10 mM sodium borate, 1% w/v PFA, 0.15% w/v Triton X-100, pH 9.2. Slides were left to air dry for 2 hours before being washed three times in PBS for 5 minutes. Samples were blocked in PBS, 3% w/v BSA and 0.1% w/v Triton X-100 for 30 minutes at room temperature. The following primary antibodies were diluted in blocking buffer and incubated overnight at room temperature; anti-γ-H2A.X 1:1,000 (Millipore, JBW301) and anti-SCP3 1:300 (Santa Cruz Biotechnology, sc-33195). Slides are were washed three times in PBS, 0.1% w/v Triton X-100 for 5 minutes and incubated in the following secondary antibodies diluted in blocking buffer for 1 hour at room temperature; goat anti-mouse Alexa Fluor 488 (A11029, Life Technologies) and goat anti-rabbit Alexa Fluor 594 (A11037, Life Technologies). The slides were then washed three times in PBS, 0.1% w/v Triton X-100 for 5 minutes and stained with 0.5 μg.ml^-1 DAPI diluted in PBS for 10 minutes. Slides were washed once in water and mounted with ProLong Gold antifade reagent (P36934, Molecular Probes) and coverslips placed on top of the slides. Images were captured using a Zeiss LSM 780 confocal microscope (Zeiss). The frequencies of meiotic intermediates were scored blindly following standard criteria.

Assessing the fertility of DNA-repair-deficient mice

To determine if mutant mice were capable of giving rise to offspring, test mice were paired with wildtype C57BL/6J mates of the opposite gender. Females were examined daily for the presence of copulation plugs. The numbers of pups born over three successive months were recorded (only mating pairs in which at least 3 copulation plugs were found were included in the final analysis). Individuals performing copulation plug checks were blinded to the genotype of mice.

Embryo isolation and fetal gonad dissection

Timed matings were performed overnight and female mice were assessed for the presence of copulation plugs the following day and separated from males. Halfway through the light cycle on the day a copulation plug was observed was designated E0.5. Pregnant mice were culled by cervical dislocation and exsanguination at noon of the appropriate day during gestation (E9.5-18.5) and the embryos harvested. Embryos were dissected and individual where appropriate fetal gonads placed into ice-cold PBS and further analysis performed immediately.

Whole-mount fluorescence imaging of fetal gonads

Gonads from developing embryos (E11.5-13.5) carrying the GOF18-GFP PGC reporter were dissected as previously described. Briefly, embryonic gonads were dissected and placed in ice-cold PBS and washed three times in ice-cold PBS. Whole-mount fluorescence images were acquired on a Leica DM IL inverted microscope (Leica) using a Lumen200 light source (Scitech).

Analysis of PGC distribution in whole-mount embryos

Embryos were harvested at E9.5, E10.5 and E11.5 following timed matings between Ercc1 heterozygous mice carrying the GOF18-GFP reporter, as described previously. Embryos were fixed in 10% neutral-buffered formalin overnight at 4 °C and washed three times in PBS for five minutes each. In the case of E10.5 embryos the right body wall was dissected to reveal the genital ridge and placed on their left side and embedded in 1% w/v low melting point agarose. In the case of E11.5 embryos the front abdominal wall and visceral organs were dissected away to reveal the genital ridge⁴⁷ and embryos were placed ventral side down and embedded in 1% w/v low melting point agarose to facilitate visualization of GOF18-GFP⁺ PGCs. Samples were incubated at 4 °C for 30 minutes, until the agarose had fully set. Whole-mount fluorescence images were acquired using an ECLIPSE TE2000 inverted epifluorescence microscope (NIKON). In the case of E10.5 embryos somites were used as physiological landmarks to assess the distribution of PGCs along the hindgut.

Quantification of PGCs in vivo

For PGC quantification, urogenital ridges of developing embryos (E11.5-13.5) or embryos (E9.5-10.5) carrying either the GOF18-GFP or Stella-GFP reporter were isolated and placed into 150 μl (or 500 μl in the case of E9.5-10.5 embryos) of trypsin solution (2.5 μg.ml^-1 trypsin (Gibco), 25 mM Tris, 120 mM NaCl, 25 mM KCl, 25 mM KH₂PO₄, 25 mM Glucose, 25 mM EDTA, pH 7.6) pre-warmed to 37 °C and incubated for 10 minutes at 37 °C. Subsequently 1 μl of Benzonase (EMD Millipore) was added and the sample disaggregated by gentle pipetting and incubated for a further 5 minutes at 37 °C. The trypsin was inactivated by the addition of 1 ml of PBS, 5% v/v fetal calf serum (FCS). Following 10 minutes of centrifugation at 3,300 r.p.m., the cell pellet was resuspended in 100 μl of Alexa Fluor 647-conjugated anti-human/mouse CD15 (SSEA1) (BioLegend, MC-480) diluted 1:100 in PBS, 2.5% v/v FCS and incubated at room temperature for 10 minutes. 300 μl of PBS, 2.5% v/v FCS was added to the cell suspension and the samples immediately run on an Eclipse analyzer (Sony Biotechnology Inc.) and the data analyzed using FlowJo v10.1r5.

Assessment of PGCs for DNA damage and DNA damage response markers

E11.5 fetal gonads were dissected as previously described and fixed in PBS, 4% w/v paraformaldehyde (PFA) for 30 minutes at 4 °C. Fixed gonads were then washed three times in PBS, 1% w/v Triton X-100 for 15 minutes at room temperature before being pressed onto glass slides and a hydrophobic circle drawn around the tissue. Slides were blocked in PBS, 1% w/v BSA, 1% w/v Triton X-100 for 30 minutes at room temperature before being incubated overnight at 4 °C with the following primary antibodies diluted in blocking buffer; anti-GFP 1:500 (Nacalai, GF090R), anti-γ-H2A.X 1:1,000 (Millipore, JBW301), anti-53BP1 1:1,000 (Novus, NB100-304), anti-cleaved Caspase-3 1:400 (Cell Signaling Technology, 9661), anti-phospho-p53 (Ser15) 1:500 (Cell Signaling Technology, D4S1H). Slides were then washed three times in PBS, 1% w/v Triton X-100 for 5 minutes and incubated with the following secondary antibodies diluted in blocking buffer, for 1 hour at room temperature; goat anti-rat Alexa Fluor 488 1:1,000 (A11006, Life Technologies), goat anti-mouse Alexa Fluor 594 1:1,000 (A11032, Life Technologies) and goat anti-rabbit Alexa Fluor 594 1:1,000 (A11037, Life Technologies). The slides were then washed three times in PBS, 1% w/v Triton X-100 for 5 minutes and stained with 0.5 μg.ml^-1 DAPI diluted in PBS for 10 minutes. Slides were washed once in water and mounted with ProLong Gold anti-fade reagent (P36934, Molecular Probes) and coverslips placed onto slides. Images were captured using a Zeiss LSM 780 confocal microscope (Zeiss). DNA damage foci per nucleus or staining of DNA damage response markers were scored blindly.

Isolation of primordial germ cells

For FACS sorting of PGCs (GOF18-GFP⁺) and surrounding somatic cells (GOF18-GFP^-) individual gonads of developing embryos (E11.5-13.5) or whole embryos (E8.5-10.5) carrying the GOF18-GFP reporter were isolated as previously described and placed into pre-warmed trypsin solution (2.5 μg.ml^-1 trypsin (Gibco), 25 mM Tris, 120 mM NaCl, 25 mM KCl, 25 mM KH₂PO₄, 25 mM Glucose, 25 mM EDTA, pH 7.6) and incubated for 10 minutes at 37 °C. Subsequently 1 μl of Benzonase (EMD Millipore) was added and the sample disaggregated by pipetting and incubated for a further 5 minutes at 37 °C. The trypsin was inactivated by the addition of 1 ml of PBS, 5% v/v fetal calf serum (FCS). Following 10 minutes of centrifugation the cell pellet was resuspended in 100 μl of Alexa Fluor 647-conjugated anti-human/mouse CD15 (SSEA1) (BioLegend, MC-480) diluted 1:100 in PBS, 2.5% v/v FCS and incubated at room temperature for 10 minutes. 300 μl of PBS, 2.5% v/v FCS was added to the cell suspension and the samples passed through a 70 μm cell strainer. Samples were immediately run on a Synergy sorter (Sony Biotechnology Inc.) and the cells sorted into 10 μL PBS. Sorted PGCs were centrifuged at 13,000 r.p.m. in a centrifuge chilled to 4°C and the cell pellets stored at -80 °C until further analysis.

Quantitative RT-PCR and gene expression analysis

Primordial germ cells (GOF18-GFP⁺) and surrounding somatic cells (GOF18-GFP^-) were isolated from E8.5-12.5 embryos carrying the GOF18-GFP reporter by FACS as described above. Total RNA was extracted using the PicoPure™ RNA Isolation kit (Life Technologies) and first strand cDNA was synthesized using the Superscript III reverse transcriptase (Life Technologies) following the manufacturer’s instructions. Quantitative real-time PCR analysis for the expression of Ap2γ, Prdm14, Blimp-1, Stella, Mvh and Dazl was performed using oligonucleotides described previously^45,46. qPCR was performed using Brilliant II SYBR^® Green QPCR Master Mix (Invitrogen) in a Viia7 (ThermoFisher) cycler at 95 °C for 10 minutes and 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute. Mean threshold cycles were determined from 3 technical repeats using the comparative CT methodology. All expression levels were normalised to Gapdh. Taqman probes were used for qRT-PCR expression analysis of Ercc1 (m00468337_m1), Xpa (mm00457111_m1), Fanca (mm01243361_g1), Plzf (mm01176868_m1) and Stella (mm01184198_g1) PCR amplification was performed using the TaqMan Fast Advanced Master Mix (ThermoFisher Scientific). PCR amplification was performed on a Viaa7 cycler for 95 °C for 15 seconds and 60°C for 1 minute. Mean threshold cycles were determined from 3 technical repeats using the comparative CT methodology. All expression levels were normalized to Gapdh (mm99999915_g1).

Bisulfite Sequencing

Primordial germ cells (GOF18-GFP⁺) and surrounding somatic cells (GOF18-GFP^-) were purified from E12.5 embryos carrying the GOF18-GFP reporter by FACS as previously described. Genomic DNA extraction and bisulfite conversion was performing using the EZ DNA Methylation-Direct™ Kit (Zymo Research) following the manufacturer’s instructions. Nested primer pairs were used to amplify the following genomic regions using the ZymoTaq™ polymerase (Zymo Research). The oligonucleotides used to amplify Mili⁴⁶, Tex19.1⁴⁶, H19DMR⁷³, and Line1⁵² were described previously. PCR products were ran out on an agarose gel and a single-band excised, gel-extracted using the QIAquick^® Gel Extraction Kit (QIAGEN) following manufacturer’s instructions and cloned into the pGEM-T Easy Vector System I (Promega) for sequencing. Sequencing reads were analyzed using the free online software quantification tool for methylation analysis (QUMA) using standard quality control settings.

Single-cell gel electrophoresis (Comet assay)

Primordial germ cells (GOF18-GFP⁺) were purified from E11.5 embryos carrying the GOF18-GFP reporter by FACS as previously described. Purified PGCs were sorted into PBS at a concentration of 100,000 cells per ml and were mixed with molten low melting point agarose (LMAgarose), which was boiled and then equilibrated to 37 °C at 1:10 v/v: LMAgarose). 50 μl of cell suspension (1 part cell solution 10 parts molten LMAgarose) was added to each well of a 2-well CometSlide™ (Trevigen^®) to form a thin, uniform layer and allowed to set for 30 minutes at 4°C in the dark. Samples were lysed overnight at in pre-chilled lysis solution (Trevigen^®) at 4 °C in the dark. Samples were drained and incubated in pre-chilled alkaline unwinding solution (200 mM NaOH, 1 mM EDTA, pH > 13) for 1 hour at 4 °C. Slides were transferred to an electrophoresis tank containing alkaline electrophoresis solution (200 mM NaOH, 1 mM EDTA, pH > 13) and 1 volt/cm was applied for 30 minutes, as described in the manufacturer’s instructions. Slides were drained and washed twice in water for 5 minutes each before being placed into 70% ethanol for a further 5 minutes. Slides were then dried at 37 °C in the dark for 10 minutes. Comets were stained with 100 μl of 1:10,000 (v/v) SYBR™ gold nuclei acid stain (Invitrogen) for 30 minutes at RT in the dark. Comets were imaged by epifluorescence microscopy using a Zeiss AxioPlan 2 microscope (Zeiss) and comet tail moment was calculated using the CometScore software (TriTek).

Derivation, maintenance and gene targeting of mouse embryonic stem cells

Wildtype and Fanca^-/- mouse embryonic stem cells (mESCs) were derived from mice carrying the GOF18-GFP reporter. Pluripotent mESCs were maintained in N2B27 media supplemented with the Glycogen synthase kinase-3 inhibitor, CHIR99021 (Axon Medchem, 1386), the MAPK/ERK inhibitor PD0325901 (Axon Medchem, 1408), and mouse leukemia inhibitory factor (LIF) (Cambridge Stem Cell Institute) referred to as to 2i + LIF media. Ercc1^-/- GOF18-GFP mESCs were generated by transfecting wildtype mESCs with px461 containing the Cas9 (D10A) and the guides mmErcc1Nicka and mmErcc1Nickb. Sequence of oligonucleotides are provided in Supplementary Table 1. Whole mESC lysates were probed using the anti-ERCC1 (Santa Cruz Biotechnology, sc-1708) primary antibody as described previously and successfully targeted clones expanded. This resulted in a homozygous deletion of 59bp in the Ercc1 locus resulting in transcripts containing a frameshift.

Generation and quantification of PGCLCs in vitro

Wildtype, Ercc1^-/- and Fanca^-/- mESCs carrying the GOF18-GFP reporter were maintained as previously described and differentiated into primordial germ cell-like cells (PGCLCs) as described previously³⁸. To determine the frequency of PGCLCs generated from mESCs, PGCLC colonies were harvested by centrifugation at 3,300 r.p.m. for 10 minutes and cell pellets resuspended in 500 μl of pre-warmed trypsin solution and incubated for 10 minutes at 37 °C. Afterwards, 1 μl of Benzonase (EMD Millipore) was added and the samples disaggregated by pipetting and incubated for a further 5 minutes at 37 °C. The trypsin was inactivated by the addition of 1 ml of PBS, 5% v/v FCS. Following 10 minutes of centrifugation at 3,300 r.p.m., the cell pellet was resuspended in 100 μl of Alexa Fluor 647-conjugated anti-human/mouse CD15 (SSEA1) (BioLegend, MC-480) diluted 1:100 in PBS, 2.5% v/v FCS and incubated at room temperature for 10 minutes. The cell suspension was diluted by added 300 μl of PBS, 2.5% v/v FCS and the frequency of PGCLCs (SSEA1⁺GFP⁺) determined by flow cytometric analysis on a BD LSRFortessa analyser (BD Biosciences) and the data analyzed using FlowJo v10.1r5.

Western Immunoblotting

Unless otherwise stated the following primary antibodies were used for western immunoblotting, diluted in 5% w/v BSA, 0.1% Tween-20 TBS and incubated at 4 °C overnight with gentle agitation; Anti-ERCC1 antibody 1:100 (Santa Cruz Biotechnology, sc-1708), anti-FANCA 1:1,000 (Cell Signaling Technology, D1L2Z), anti-PLZF 1:100 (Santa Cruz Biotechnology, sc-28319), anti-Histone H3 1:5,000 (Abcam, 1791), anti-β-ACTIN 1:3,000 (Abcam, 8227), anti-VINCULLIN 1:2,000 (Abcam, 129002). The following secondary antibodies were diluted in 5% w/v BSA, 0.1% Tween-20 TBS and incubated for 1 hour at room temperature with gentle agitation; Swine anti-rabbit or anti-mouse HRP-conjugated immunoglobulins 1:2,000 (Dako). To demonstrate the absence of ERCC1 in Ercc1-mutant MEFs derived from Ercc1^{tm1a(KOMP)Wtsi} embryos anti-ERCC1 1:1,000 (Cell Signaling Technology, D6G6) was employed, following the same western immunoblotting protocol described previously.

Cell cycle analysis following EdU incorporation

To assess the incorporation of EdU into the DNA of PGCs in vivo, pregnant mice were given a single dose of 50 mg/kg EdU by intraperitoneal (IP) injection at 10 ml/kg. Following a 1 hour incubation period the female was sacrificed by cervical dislocation and the fetal gonads of E11.5 embryos were harvested. Following a 1 hour incubation period the fetal gonads of E11.5 embryos were harvested as previously described and placed into 150 μl pre-warmed trypsin and incubated for 10 minutes at 37 °C. Subsequently 1 μl of Benzonase (EMD Millipore) was added and the colonies disaggregated by pipetting and incubated for a further 5 minutes at 37 °C. Trypsin was inactivated by the addition of 1 ml of PBS, 5% v/v FCS. Following 10 minutes of centrifugation at 3,300 r.p.m., the cell pellet was re-suspended in 100 μl of Alexa Fluor 647-conjugated anti-human/mouse CD15 (SSEA-1) (BioLegend, MC-480) diluted 1:100 in PBS, 2.5% v/v FCS and incubated at room temperature for 10 minutes. 1 ml of PBS, 2.5% v/v FCS was added and the samples centrifuged at 3,300 r.p.m. After centrifugation the samples were re-suspended in 100 μl of IntraPrep fixative reagent (A07803, Beckman Coulter) and incubated at room temperature for 15 minutes. 1 ml of PBS, 5% FCS was added the samples and centrifuged at 3,300 r.p.m. Cell pellets were then re-suspended in 100 μl of IntraPrep permeabilisation buffer (A07803, Beckman Coulter) and incubated at room temperature whilst the Click-iT^® azide solution (C10646, Invitrogen) is prepared (following the manufacturer’s instructions). 500 μl of Alexa Fluor 594 picolyl azide solution is added to each sample and incubated at room temperature for 30 minutes. 1 ml of saponin solution (C10646, Invitrogen) is added before the samples are centrifuged at 3,300 r.p.m. Samples are resuspended in 500 μl of PBS, 2.5% v/v FCS, 0.5 μg.ml^-1 DAPI and incubated at room temperature for 10 minutes. The samples were analyzed immediately on a BD LSRFortessa (BD Biosciences) analyzer and the data analyzed using FlowJo v10.1r5.

Immunoblotting of spermatogonial stem cells

To assess the success of Stra8-iCre-mediated disruption of the conditional Ercc1 and Fanca alleles, spermatogonial stem cells were sorted from adult testes for western blot analysis, essentially as described previously⁷⁴. Briefly, 8- to 12-week old mice were culled by cervical dislocation, the testes dissected and tunica albuginea removed. Decapsulated testes were incubated in 3.5 ml Gey’s balanced salt solution (GBSS) (Sigma, G9779) containing 120 units.ml^-1 collagenase type I (Sigma, SCR103) and 10 µl of DNase I (Sigma, DN25) stock solution (1 mg.ml^-1) and incubated at 55 °C for 15 minutes with gentle agitation. The testicular tubules were allowed to settle under gravity and the supernatant discarded, and the process repeated once more. Following the aspiration of the second supernatant, samples were incubated in 2.5 ml GBSS containing 120 units.ml^-1 collagenase type I, 1 mg.ml^-1 trypsin and 10 µl DNase I for 15 minutes at 55 °C with gentle agitation. After gentle disaggregation using a Pasteur pipette samples were stained with Hoechst 33342 (Life Technology, H3570) added to a final concentration 350 ug.ml^-1. The samples were run on a Synergy sorter (Sony Biotechnology Inc.) with 5 ug.ml^-1 propidium iodide. Viable spermatogonia, defined as PI-negative, Hoechst red low and Hoechst blue low, were sorted into 500 μl PBS. Sorted spermatogonia were centrifuged at 13,000 r.p.m. in a centrifuge chilled to 4 °C and the cell pellets lysed in SDS-PAGE running buffer (Thermo Fisher Scientific, NP0008). Whole cell lysates of 100,000 spermatogonial stem cells were immunoblotted for ERCC1 or FANCA using the anti-ERCC1 antibody 1:100 (Santa Cruz Biotechnology, sc-1708) and the anti-FANCA antibody 1:1,000 (Cell Signaling Technology, D1L2Z) as previously described.

Statistical analysis

The number of independent biological samples and technical repeats (n) are indicated in figure legends. Unless otherwise stated data are shown as mean ± standard error of the mean (s.e.m.) and the nonparametric Mann-Whitney test was employed to determine statistical significant. Analysis was performed in GraphPad Prism version 8.