Search
Close this search box.

p53 regulates diverse tissue-specific outcomes to endogenous DNA damage in mice – Nature Communications

p53 is activated in the absence of Ercc1

To study the relationship between endogenous DNA damage, the p53 checkpoint and tissue-specific outcomes, we employed a previously published Ercc1−/ mouse model11,38. We first confirmed that our model recapitulates the key pathological features of XPF-ERCC1 deficiency in mouse and human. Ercc1−/− mice were generated on a C57BL/6 x 129S4S6/Sv F1 hybrid background using the Ercc1tm1a(KOMP)Wtsi allele. From heterozygous crosses, Ercc1/ pups were born at the expected Mendelian ratio at postnatal day 14 (P14) (24.2%, p = 0.4470). However, they were severely runted compared to littermate controls, with a significant 1.8-fold reduction in body mass 21 days after birth (Supplementary Fig. 1a, b). Consistent with previous reports, Ercc1−/ mice had a significantly reduced lifespan and segmental loss of tissue homoeostasis (Supplementary Fig. 1c–f)4,15. Firstly, these mice had elevated blood serum levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activity, indicative of liver dysfunction (Supplementary Fig. 1d). Secondly, Ercc1/− mice develop renal insufficiency with elevated blood serum levels of urea and proteinuria (Supplementary Fig. 1e, f). Finally, Ercc1−/− mice showed a significant 2.1-fold reduction in the frequency of haematopoietic stem and progenitor cells (HSPCs, lineage c-Kit+ Sca-1+) (Supplementary Fig. 1g)6. Collectively, these data demonstrate that our model recapitulates the major phenotypes of both human XFE progeroid syndrome and previously generated Ercc1−/− models.

To explore tissue differences in response to endogenous DNA damage, we initially focussed our attention on the liver, kidney, and bone marrow of Ercc1/− mice as these are thought to contribute to reduced lifespan4,15,35. We stained tissues of 6-to-8-week-old Ercc1−/ and control mice for the phosphorylated histone variant H2A.X (γ-H2A.X), a marker of DNA breaks39. We observed a significant increase in the frequency of cells with >5 nuclear foci of γ-H2A.X in each tissue (Fig. 1a) and an increase in the average number of foci per cell (Supplementary Fig. 2a). This shows that there is a sufficient burden of endogenous genotoxin, resulting in the accumulation of unrepaired DNA damage in the absence of ERCC1 in all three tissues. Consequently, persistent DNA damage can lead to the activation of the DNA damage response. Whilst p53 is a downstream effector, it is among, if not the most, important regulator of the cellular response to DNA damage20. We observed a significant increase in the frequency of pSer15-p53+ cells – a post-translational modification induced upon DNA damage—in the liver, kidney and bone marrow of Ercc1−/− mice when compared to controls (Fig. 1b). Phosphorylation and stabilisation of p53 result in the activation of a range of p53-transcriptional targets, which ultimately determine the cellular outcome40. One key component of this response is the activation of programmed cell death. We found that whilst all three tissues tested showed persistent DNA damage and accumulation of phosphorylated p53, only the kidney and bone marrow had an increased frequency of cells undergoing apoptosis as determined by cleaved caspase-3 staining (CC3+) and TUNEL assay (Fig. 1c and Supplementary Fig. 2b).

Fig. 1: p53 is activated in the absence of ERCC1.
figure 1

a Representative images of γ-H2A.X foci in liver, kidney and bone marrow sections from wildtype and Ercc1/ mice and quantification of the frequency of cells with >5 nuclear γ-H2A.X foci (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = liver (6 wildtype and 4 Ercc1−/−), kidney (6 wildtype and 4 Ercc1/−), bone marrow (5 wildtype and 3 Ercc1−/−) independent mice). b Representative immunofluorescence images of phosphorylated TP53 (pSer15-TP53) in liver, kidney and bone marrow of wildtype and Ercc1−/ mice and quantification of the frequency of pSer15-TP53+ cells (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = liver (4 wildtype and 4 Ercc1/), kidney (4 wildtype and 4 Ercc1−/−), bone marrow (3 wildtype and 3 Ercc1−/) mice). c Representative immunofluorescence images of cleaved caspase-3 (CC3) in liver, kidney and bone marrow of wildtype and Ercc1−/− mice and quantification of the frequency of CC3+ cells (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 5 Wildtype and 5 Ercc1−/−), kidney (5 wildtype and 5 Ercc1/), bone marrow (5 wildtype and 4 Ercc1/−) mice). Source data are provided as a Source data file.

We then assessed the expression of Cdkn1a/p21, Btg2, Mdm2, Gdf15 and Bax which are well-described transcriptional targets of p53 involved in different cellular processes41. We performed RT-qPCR on liver, kidney and HSC samples and, in agreement with the CC3 immunostaining, we observed an induction of the pro-apoptotic factor Bax in both kidney and blood stem cells but not in the liver (Supplementary Fig. 2c). In contrast, Ercc1−/− liver and kidney show a significant induction in the expression of the senescence marker p21/Cdkn1a but this is not observed in HSCs (Supplementary Fig. 2c). Together, these data show that in the absence of ERCC1, endogenous DNA damage accumulates and activates p53 in the liver, kidney, and bone marrow, but results in distinct tissue-specific transcriptional and apoptotic responses.

Differential dependence of tissues on p53 for apoptosis activation

We then set out to genetically determine the contribution of p53 to tissue-specific responses such as apoptosis, cell-cycle arrest and polyploidisation in Ercc1−/− mice. We generated Ercc1/−Trp53−/− mice on a C57BL/6 x 129S4S6/Sv F1 background. Ercc1−/−Trp53−/− mice have previously been generated; however, analysis was restricted to the liver and the consequences on tissue pathology and homoeostasis were not studied37,42. We found that Ercc1−/−Trp53/ mice were present at a significantly reduced frequency at postnatal day P21 (3.0%, 12/395, p = 0.0002; Fig. 2a). This is unexpected as Ercc1−/− single mutants were born at the expected ratio (Supplementary Fig. 1a). Furthermore, the disruption of p53 frequently rescues the lethality of mice deficient in a range of different DNA repair pathways rather than potentiating their phenotypes43,44,45,46,47,48,49,50.

Fig. 2: Differential dependence of tissues on p53 for apoptosis activation.
figure 2

a Ercc1/− mice are observed at expected Mendelian ratios at postnatal day 21 (p value calculated by one-tailed Chi-square test). b Representative photograph of 6-week-old Ercc1/−Trp53/ and littermate control mice. c Weight of Ercc1/Trp53/− and control mice at postnatal day 21 (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 16, 22, 14, 10 independent mice, left to right). d Terminal serum GDF15 levels from 6-to-8-week-old Ercc1/−Trp53/− and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 6, 3, 7 and 5 independent mice, left to right). e Kaplan–Meier survival curve showing the survival of cohorts of Ercc1−/− and wildtype littermate controls (p value calculated by one-tailed Mantel-Cox test, n = 29 wildtype and 23 Ercc1/− mice). f Left—Representative immunofluorescence of cleaved caspase-3 (CC3) staining in the liver, kidney, bone marrow and cerebral cortex of Ercc1−/−Trp53−/ and control mice. Right—Quantification of the frequency of CC3+ cells in the liver, kidney, bone marrow and cerebral cortex of Ercc1/−Trp53−/ and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; each dot represent a mouse, Liver: n = 5, 3, 5, 4; Kidney: n = 4, 3, 5, 4; Bone marrow: n = 4, 3, 4, 3; Cerebral cortex: n = 4, 3, 4, 3 left to right). Source data are provided as a Source data file.

An obvious feature of Ercc1-deficient mice is a profound reduction in mass when compared to littermates, however, the aetiology of this phenotype is unknown (Fig. 2b, c, Supplementary Fig. 1b). We hypothesised that this may in part be due to high blood serum levels of the anorexic hormone growth/differentiation factor 15 (GDF15) that we detected in high levels in the absence of ERCC1 when compared to wildtype controls (Fig. 2d)51. It has previously been shown that the expression of GDF15 is regulated by p53, and consistent with this, we found that the GDF15 response in Ercc1−/ mice was suppressed by p53 loss (Fig. 2d)52. Despite the suppression of GDF15 secretion, Ercc1−/−Trp53−/− mice had a significant reduction in body mass that was comparable to that observed in Ercc1−/− controls (Fig. 2c). Hence, it is unlikely that GDF15 plays an important role in preventing weight gain in patients with XFE progeroid syndrome. The most striking feature of Ercc1 deficiency is the dramatic reduction in lifespan, however, we found that there was no significant difference in the median lifespan of Ercc1−/−Trp53−/− mice when compared to Ercc1−/− controls (55 and 63 days, respectively, p = 0.1224) (Fig. 2e). These data clearly show that loss of p53 does not rescue the reduced weight or longevity of Ercc1-deficient mice.

As we had shown that Ercc1-deficient tissues had increased frequency of apoptotic cells and that p53 is a regulator of apoptosis, we asked if, in the absence of p53, apoptosis was attenuated. We assessed the frequency of apoptotic cells in the liver, kidney, bone marrow, and cerebral cortex of Ercc1−/−Trp53/ mice (Fig. 2f). In agreement with our earlier data, the liver does not activate apoptosis despite exhibiting DNA damage and increased pSer-p53 (Figs. 1a, b, 2f). We found that whilst the kidney, bone marrow and cerebral cortex of Ercc1−/ mice each had elevated frequency of apoptotic cells, the loss of p53 only led to a significant reduction in the bone marrow (p = 0.0099; Fig. 2f). These results suggest that p53 is the major driver of apoptosis in the bone marrow, but that response is activated by alternative mechanisms in the kidney and cerebral cortex. These data clearly show that p53 has a complex, unpredicted and tissue-specific interaction with Ercc1.

Kidney, neurological and germ cell pathologies persist in the absence of p53

The kidney is a vital organ which becomes dysfunctional in the absence of ERCC1-mediated DNA repair35. Whilst we found that loss of p53 did not rescue the apoptotic response in this organ, we examined the kidney in more detail. Firstly, we asked if loss of p53 exacerbated the burden of DNA damage in Ercc1−/− kidney, as it could explain why the loss of p53 did not affect the frequency of apoptotic cells. We found that the loss of p53 did not alter the frequency of cells with markers of DNA damage when compared to Ercc1−/− littermates (Supplementary Fig. 3a). Next we performed RT-qPCR on kidney samples with a panel of p53-transcriptional targets involved in different responses. We measured the pro-apoptotic factors Bax and Fas, and the cell-cycle/proliferation regulator p21/Cdkn1a and Btg241. We found that they were all upregulated in both Ercc1/− and Ercc1−/Trp53/ kidneys (Fig. 3a). The observation that loss of p53 did not reduce the expression of the pro-apoptotic factors Bax and Fas agrees with the finding that frequency of CC3+ apoptotic kidney cells was not reduced (Supplementary Fig. 3b and Fig. 2f). As we found substantially increased expression of the negative proliferation regulators p21/Cdkn1a and Btg2, we assessed proliferation in the kidney using Ki67 and PCNA (Supplementary Fig. 3c, d)53. We found that, despite the increase in p21/Cdkn1a and Btg2 expression (Supplementary Fig. 3e), there was an increased frequency of proliferating cells. On the surface, the increased expression of negative regulators of proliferation conflicts with the observed increase in the frequency of Ki67+ cells. It is plausible that this discrepancy is due to the induction of proliferation needed for tissue repair to replace the cells lost by apoptosis. As the transcriptional data represents the average across cells, we would be unable to detect a population of undamaged cells undergoing proliferation to maintain tissue homoeostasis. Finally, we assessed the consequence of p53 loss on kidney morphology and function and found that neither of them was altered, with serum creatinine and urea concentrations similar to Ercc1 single mutants (Fig. 3c–e). Together, these data show that p53 does not drive apoptosis or tissue dysfunction in the kidney of Ercc1−/− mice.

Fig. 3: Ercc1−/− kidney and neurological dysfunction are independent of p53.
figure 3

a Quantitative RT-PCR expression analysis of the p53 target genes (Bax, Fas, p21 and Btg2) in the kidney of Ercc1−/Trp53−/− and control mice (p values calculated by two-tailed Mann–Whitney U-test; data were mean ± s.d.; each dot represents an independent mouse; Bax: n = 4, 3, 4, 5; Fas: n = 5, 3, 3, 5; p21: n = 4, 3, 4, 5; Btg2: n = 4, 3, 3, 5 right to left) b H&E-stained sections of kidney from 6-to-8-week-old Ercc1−/−Trp53−/− and control mice. c Pathology assessment of tubular atrophy in the kidney of 6-to-8-week-old Ercc1−/−Trp53−/− and control mice. Scores range from 0 (absent) to 6 (substantial) (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 8, 4, 7 and 5 independent mice, left to right). d Terminal blood serum creatinine levels from Ercc1−/−Trp53/− and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 10, 8, 9 and 5 independent mice, left to right). e Terminal blood serum urea levels from Ercc1/Trp53/− and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 10, 8, 9 and 5 independent mice, left to right). f Representative immunofluorescence of MAC2 staining in the cerebral cortex of Ercc1−/−Trp53−/− and control mice and quantification of the frequency of MAC2+ cells (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4, 3 4, and 3 independent mice, left to right). g Representative immunofluorescence of GFAP staining in the cerebrum of Ercc1−/−Trp53/− and control mice and the frequency of GFAP+ cells (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4 wildtype, 3 Trp53−/, 4 Ercc1−/− and 3 Ercc1−/−Trp53−/). Source data are provided as a Source data file.

It has previously been reported that Ercc1-deficient mice accumulate p53 in the brain and have a range of neurodegenerative changes7,8. Indeed, p53 has been implicated in the neurodegenerative changes associated with DNA damage, aging and to mediate neuronal degeneration49,50,54,55,56,57. We therefore tested if the activation of p53 was responsible for the histophatological changes previously reported in Ercc1−/ mice. We found no difference in mass or gross morphology between Ercc1/ and Ercc1/−Trp53−/− brains (Supplementary Fig. 4a, b). We found an increase in the frequency of senescent (p21+) cells in the brain of both Ercc1−/− and Ercc1−/−Trp53−/− mice (Supplementary Fig. 4c). We then assessed the induction of both microgliosis and astrocytosis by measuring the frequency of MAC2+ and GFAP+ cells, respectively. We found that both were induced in Ercc1−/− mice when compared to wildtype, however, this increase was not suppressed by the loss of p53 (Fig. 3f, g). We went on to measure inflammatory markers previously observed in Ercc1−/ mice by RT-qPCR and found that they remained induced even upon p53 loss (Supplementary Fig. 4d)58. Together these data show that loss of p53 is not sufficient to prevent pathological changes in the brain of Ercc1-deficient mice.

Another hallmark of ERCC1 deficiency is infertility and germ cell failure11,59. We have previously characterised the infertility defect in Ercc1−/ mice and found that Ercc1 acts together with the Fanconi Anaemia DNA repair pathway to safeguard pre-meiotic germ cell development11. We found that in the absence of Ercc1, primordial germ cell (PGC) development is compromised by embryonic day 12.5 of development11. We also found that mutant PGCs accumulate pSer15-p53+ and undergo apoptosis11. Moreover, p53 has been shown to drive germ cell loss in the absence of other DNA repair pathways60. We therefore hypothesised that loss of p53 would also rescue the germ cell defect in Ercc1−/− mice. However, we found that inactivation of p53 did not alter the reduction in either testicular or ovarian mass (Fig. 4a, b). The testis was devoid of germ cells carrying Sertoli cell-only (SCO) tubules (Fig. 4d), and the ovary had a near complete lack of follicles in both Ercc1−/− and Ercc1−/−Trp53−/− mice postnatally (Fig. 4e). However, it was in the embryonic germ cells that we had previously detected pSer15-p53+. We therefore quantified the frequency of PGCs at E12.5 by flow cytometry in embryos carrying the GOF18-GFP reporter and found that there was no difference between Ercc1−/− and Ercc1/−Trp53/− embryos (Fig. 4f)61. Taken together, these data show that, whilst p53 is induced in the kidney, brain and PGCs, it makes a minimal contribution to pathogenesis in those organs.

Fig. 4: Germ cell failure occurs independently of p53 in Ercc1−/− mice.
figure 4

a Representative images of testes from age-matched Ercc1/−Trp53/ and control mice and b quantification testis mass (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 6, 8, 10 and 4 independent mice, left to right). c Quantification of ovary mass from age-matched Ercc1−/−Trp53/ and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 15, 8, 6 and 5 independent mice, left to right). d H&E-stained sections of testes from 6-to-8-week-old Ercc1−/−Trp53−/ and control mice and the frequency of Sertoli cell-only (SCO) tubules per testis (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 5, 7, 6 and 4 independent mice, left to right). e H&E-stained sections of ovaries from 6-to-8-week-old Ercc1−/ and control mice and quantification of follicles per section (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 7, 7, 5 and 4 independent mice, left to right). f Representative images of GOF18-GFP fluorescence in E12.5 gonads of Ercc1−/−Trp53/− and control embryos and quantification of primordial germ cells (PGCs:SSEA1+ GOF18-GFP+) by flow cytometry (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 18, 16, 4 and 7 independent embryos, left to right). Source data are provided as a Source data file.

p53 drives apoptosis and loss of Ercc1
/ haematopoietic stem cells (HSCs)

We observed a significant increase in the proportion of bone marrow cells entering p53-dependent apoptosis in the absence of ERCC1 (Fig. 2f). The bone marrow dysfunction of both humans and mice deficient in ERCC1 resembles that observed in aging and is thought to be due to reduced haematopoietic reserves. We therefore wanted to assess the role of p53 in regulating the HSC compartment, as these cells are ultimately responsible for maintaining haematopoietic homoeostasis. We assessed the transcriptomes of FACS-purified HSCs (lineage c-Kit+ Sca-1+ CD48 CD150+) from Ercc1−/−Trp53/− and control mice by RNA-seq62. We found the induction of p53 targets, including pro-apoptotic factors (Bax, Fas, among others) in Ercc1−/− HSCs that were suppressed when p53 was ablated (Supplementary Fig. 5a, b). We validated these results by performing RT-qPCR on a panel of p53-transcriptional targets that regulate different cellular responses to DNA damage (Fig. 5a). We found that whilst there was no significant increase in either of the proliferation regulators p21/Cdkn1a or Btg2, there was an induction of both apoptotic factors Bax and Fas in Ercc1/− HSCs as expected. Remarkably, we found that, when p53 was lost, the transcriptional activation of both Bax and Fas was suppressed. These data show, in agreement with the RNA-seq and the total bone marrow data, that p53 drives an apoptotic response in the HSC compartment of Ercc1−/− mice (Figs. 1c, 2f). A key feature of the reduced haematopoietic reserve in the absence of ERCC1 is a reduced frequency of HSCs. Hence, we then set out to ask if the increased p53-dependent apoptosis in Ercc1−/− HSCs led to the observed reduction in frequency. We quantified the frequency of HSCs and HSPCs by flow cytometry and found an ~10-fold reduction in Ercc1−/−. Remarkably, this was completely rescued in Ercc1/−Trp53/− mice (Fig. 5b–d). These data demonstrate that transcriptional activation of apoptosis and HSC loss in Ercc1/ mice is dependent upon p53.

Fig. 5: p53 drives apoptosis and loss of Ercc1−/− HSCs.
figure 5

a Quantitative RT-PCR expression analysis of the p53 target genes (Bax, Fas, p21 and Btg2) in the HSCs of Ercc1/−Trp53−/− and control mice (p values calculated by two-tailed Mann–Whitney U-test; data were mean ± s.d.; each dot represents an independent mouse. p21: n = 7, 3, 7, 3; Btg2: n = 7, 3, 7, 5; Bax: n = 8, 3, 8, 3; Fas: n = 4, 4, 4, 4; right to left b Representative flow cytometry plots of haematopoietic stem and progenitor cells (HSPCs, lineage c-Kit+ Sca-1+) from 6-to-8-week-old adult Ercc1/−Trp53−/ and control mice. c Quantification of HSPCs by flow cytometry from the bone marrow of adult Ercc1−/−Trp53/− and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 9, 6, 8 and 5 independent mice, left to right). d Quantification of HSCs (lineage c-Kit+ Sca-1+ CD41CD48 CD150+) by flow cytometry from the bone marrow of adult Ercc1/Trp53−/− and control mice (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 9, 6, 5 and 4 independent mice, left to right). Source data are provided as a Source data file.

p53 limits liver polyploidisation in the absence of Ercc1

The shortened lifespan of mice lacking Ercc1 is primarily attributed to liver failure4,15,35. At the cellular level, hepatocytes deficient in Ercc1 undergo extensive polyploidization, possibly resulting from endoreduplication or a failure in cytokinesis63,64. p53 is a known regulator of whole genome duplication and liver ploidy; however, conflicting reports exist implicating it in both suppressing and promoting this process32,33,34,31. Moreover, previous studies have indicated the influence of p53 in hepatocyte proliferation and binucleation in Ercc1-deficient mice36,37. Therefore, we aimed to investigate: (i) whether p53 promotes or restricts liver polyploidization in Ercc1/− mice and (ii) the mechanism by which p53 regulates this process.

We first assessed the impact of p53 loss on the liver function of Ercc1-deficient mice. We found that there were elevated serum concentrations of ALT and AST—markers of liver damage—in the absence of Ercc1 and that this was not affected by the loss of p53 (Supplementary Fig. 6a, b). Furthermore, when we assessed liver synthetic function, there was a reduction in the serum concentration of albumin in the absence of Ercc1, which remained reduced in the Ercc1−/−Trp53−/− mice (Supplementary Fig. 6c). These data show that loss of p53 does not subvert liver dysfunction of Ercc1−/− mice, which is consistent with their dramatically reduced lifespan (Fig. 2e).

When we microscopically examined Ercc1−/ livers, we observed a significant increase in the frequency of hepatocytes with enlarged and deformed nuclei consistent with previous reports (Supplementary Fig. 6d). The loss of p53 greatly exacerbated the morphological changes observed in Ercc1-deficient mice (Supplementary Fig. 6d). We isolated nuclei from Ercc1−/−Trp53−/− livers and isogenic controls and assessed their DNA content by flow cytometry65. Firstly, we found that there was an increase in the frequency of polyploid cells in the absence of Ercc1 (Fig. 6a, b and Supplementary Fig. 6e). Not only did we observe an increase in the frequency of polyploid hepatocytes in Ercc1−/Trp53−/ mice, but we found they had undergone more extensive polyploidization, with a significant 81.6-fold increase in the frequency of cells reaching 64n (Fig. 6a, b). These data show that the hepatocyte nuclear enlargement is due to polyploidisation and that p53 acts to limit polyploidisation in the absence of Ercc1.

Fig. 6: p16 is a failsafe proliferation brake in the absence of the p53-p21 axis.
figure 6

a Top—Representative immunofluorescence of DAPI staining in the liver and quantification of hepatocyte nuclear area (p values calculated by two-tailed Mann–Whitney U-test, data were median and interquartile range; n = 150 hepatocytes per genotype, 50 per mouse). Bottom— Representative flow cytometry plots of DNA content in the nuclei of liver cells of 6-to-8-week-old mice. b Quantification of the frequency of 64n cells (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4, 3, 6 and 4 independent mice, left to right). c Representative immunofluorescence of γ-H2A.X staining in the liver. d Quantification of γ-H2A.X+ cells (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; each dot corresponds to individual mice; n = 8, 10, 16, 14 left to right). e Representative immunofluorescence of Ki67 staining in the liver. f Quantification of Ki67+ cells (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4, 4, 5 and 5 independent mice, left to right). g Representative immunofluorescence of p21 staining in the liver. h Quantification of p21+ cells (p value calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; each point represents one independent mouse, n = 4, 3, 4, 4 left to right). i, j Quantitative RT-PCR expression analysis of Cdkn1a (top) and Cdkn2a (bottom) in the liver (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = Cdkn1a (11, 5 and 4) and Cdkn2a (9, 8 and 6), mice, left to right). k, l Representative immunofluorescence of Ki67 staining in the liver of young (3-to-5-week-old) and older (6-to-9-week-old) mice and quantification of the frequency of Ki67+ hepatocytes (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4 independent mice per genotype and timepoint). m Quantitative RT-PCR expression analysis of p16/Cdkn2a in the liver of young (3-to-5-week-old) and older (6-to-9-week-old) mice (p values calculated by two-tailed Mann–Whitney U-test, data were mean ± s.d.; n = 4 independent mice per genotype and timepoint). Source data are provided as a Source data file.

We next asked if the role of p53 in limiting polyploidisation was specific to the liver or generalisable to other tissues from Ercc1−/− mice. We, therefore, measured the nuclear area, as a surrogate marker of polyploidisation, in cells from the liver, kidney, bone marrow and cerebral cortex of Ercc1−/− mice (Supplementary Fig. 7a). We found that only the liver exhibited a significant increase in nuclear area in the absence of Ercc1, whilst it led to a negligible change in the kidney, bone marrow and brain when compared to controls (Supplementary Fig. 7a, b). Furthermore, the loss of p53 only led to an increase in the nuclear area of Ercc1−/− hepatocytes (Supplementary Fig. 7a, b). Together, these data show that the liver uniquely undergoes polyploidisation in the absence of Ercc1 and that this is restrained by p53.

p16 is a failsafe proliferation brake in the absence of the p53-p21 axis

We next set out to investigate why the loss of p53 should lead to increased polyploidisation in the liver (Fig. 6a, b). A canonical function of p53 is the activation of apoptosis; however, we did not detect an increased frequency of apoptotic hepatocytes (CC3+) nor transcriptional activation of apoptotic factors in Ercc1−/− livers (Figs. 1c, 2f and Supplementary Fig. 2a). However, despite not detecting evidence of apoptosis in hepatocytes (CC3+ or a transcriptional signal) we cannot exclude a role for apoptosis (e.g. at an earlier developmental stage). We also considered that the polyploidisation could be driven by an increased burden of DNA damage in Ercc1−/−Trp53−/− compared to Ercc1−/− mice. We found that whilst Ercc1−/− hepatocytes had a greater burden of DNA damage than wildtype, this was no different in Ercc1−/−Trp53−/− mice, suggesting that the absence of p53 does not lead to the accumulation of more DNA breaks (Fig. 6c, d and Supplementary Fig. 7c).

Finally, we examined whether the involvement of p53 in proliferation regulation provides insight into its role in controlling ploidy. In order for polyploidisation to occur, hepatocytes must either undergo iterative genome duplications or nuclear fusion events64. We hypothesised that a p53-dependent cell-cycle arrest mechanism was lost in Ercc1−/−Trp53−/− hepatocytes, leading to uncontrolled cell-cycle progression. To test this, we stained the livers of young 3-to-6-week-old mice for the marker of proliferation Ki67 and observed a significant increase in the frequency of Ki67+ hepatocytes in Ercc1−/−Trp53−/− compared to Ercc1−/− mice (Fig. 6e, f). Furthermore, we found that, in the absence of Ercc1, there was a dramatic induction of the negative cell-cycle regulator Cdkn1a/p21; however, this is entirely suppressed when p53 is ablated (Fig. 6g–i)66,67,68. It is likely that loss of the p53-p21 axis allows further hepatocyte proliferation enabling polyploidisation. Therefore, the massive polyploidisation in Ercc1-deficiency is likely the result of endoreduplication, the process of genome replication without mitosis.

Polyploidisation is often observed upon liver aging; indeed, Ercc1 deficiency has been proposed as a progeroid model. However, liver polyploidisation can occur either by failed cytokinesis, often associated with age-dependent polyploidisation or by the process of endoreduplication more associated with damage and regeneration69. We performed bulk RNA sequencing on purified hepatocytes from Ercc1−/− and control mice (Supplementary Fig. 7d). We first confirmed that, in the absence of Ercc1, p21/Cdkn1a was induced and that this was suppressed when p53 was ablated (Supplementary Fig. 7e). We next assessed the expression of a ‘liver aging signature’ defined in the Mouse Aging Cell Atlas (Tabula Muris Senis)70. Interestingly, we observed a strong concordance between the hepatocytes of young 6-to-8-week-old Ercc1−/− mice and the aging signature (Supplementary Fig. 8a). We then assessed the transcriptional changes when p53 was ablated and found that the aging signature persisted (Supplementary Fig. 8a). One hallmark of aging is dysregulated nutrient signalling as the growth hormone (GH)/insulin-like growth factor (IGF) somatotrophic axis is downregulated both in natural aging and in models of premature aging71,72,73,74. It has previously been shown that Ercc1-deficient mice suppress the GH/IGF somatotrophic axis and it is hypothesised that this attenuation diverts resources from growth towards a pro-survival stress response to cope with systemic damage4. However, it remains unclear how the loss of ERCC1 leads to attenuated endocrine signalling. We asked if p53 signalling contributes to the dysregulation of GH/IGF signalling. Consistent with previous studies, our analysis of the differentially expressed genes in Ercc1−/− hepatocytes compared to wildtype controls revealed altered expression of the REACTOME pathway “MMU-381426 – regulation of IGF transport and uptake by IGF binding proteins (IGFBPs)” (p = 4.68 × 10−15, rank 3rd). Strikingly, when we compared the transcriptome of Ercc1−/−Trp53/− hepatocytes with Ercc1/− controls the most enriched REACTOME pathway was MMU-381426 (p = 3.67 × 10−7, rank 1st), demonstrating that dysregulation of the IGF signalling pathway persists even in the absence of p53 (Supplementary Fig. 8b, c). We also observed transcriptional inflammatory and immune responses in Ercc1/−Trp53−/− livers which are also characteristic of aged tissues (Supplementary Fig. 8b)71. Together, these led us to conclude that Ercc1/Trp53−/− liver had multiple transcriptional changes reminiscent of premature aging.

Furthermore, one of the best-characterised hallmarks of aging is cellular senescence, which is not only a marker of increased age but also of tissue dysfunction13,70,71,75,76. We therefore sought to determine if the p53 response regulates senescence in the liver of Ercc1−/− mice. To do this, we quantified the expression of the well-characterised marker of senescence and cell-cycle regulator, p16/Cdkn2a. Surprisingly, we found that the liver of Ercc1/−Trp53−/− mice had a significant induction in the expression of p16/Cdkn2a compared to Ercc1−/ controls (Fig. 6j). This appeared contradictory to our previous results showing that the loss of the p53-p21 led to increased proliferation of Ercc1−/Trp53/− hepatocytes (Fig. 6e, f). To reconcile this, we compared p16/Cdkn2a expression and proliferation in the liver at different time points. When we compared the frequency of proliferating hepatocytes in young (3-to-5-week-old) and older (6-to-9-week-old) Ercc1/−Trp53/− mice, we observed a significant reduction in the frequency of Ki67+ cells with age (Fig. 6k, l). We confirmed this result by staining the liver of old (6-to-9-week-old) Ercc1−/−Trp53/− mice with the alternative proliferation marker PCNA (Supplementary Fig. 8d). Concomitantly, we observed a significant increase in the relative expression of p16/Cdkn2a in the liver of Ercc1/−Trp53−/− mice (Fig. 6m). These data show that, in the absence of the p53-p21 axis, p16/Cdkn2a is induced, and proliferation is suppressed, a defining feature of cellular senescence. In the absence of the p53-p21 axis, we see engagement of p16 and arrested proliferation. Given the documented role of p16 in cell-cycle regulation, this may represent a failsafe mechanism to prevent proliferation and endoreduplication.

The p53-axis protects the liver from chemotherapy-induced polyploidisation

The XPF-ERCC1 endonuclease acts to counteract several forms of DNA damage, including helix-distorting adducts, intra- and inter-strand crosslinks3,77,78,79. However, it is unclear which of these lesions could drive the liver phenotype. We hypothesised that inter-strand crosslinks could likely contribute to liver polyploidisation as mice deficient in the ability to repair ICLs (Fan1−/−, Adh5/Fancd2−/−) also exhibit liver polyploidisation65,80. We turned our attention to platinum-based DNA inter-strand crosslinking agents, as these are also frequently used in chemotherapy, including the treatment of hepatocellular carcinoma81,82. We exposed wildtype mice to cisplatin and observed a significant increase in the frequency of cells with >5 nuclear γ-H2A.X foci in the liver (Supplementary Fig. 9a, b). Moreover, the treatment led to the accumulation of pSer15-TP53 in the liver (Supplementary Fig. 9a, b). These responses are similar in magnitude to what was observed spontaneously in Ercc1−/− mice (Fig. 1a, b). Therefore, we next wanted to ask if this induction of p53 acted to restrain liver polyploidisation in the face of exogenous challenge with DNA inter-strand crosslinking agents. We found that chronic exposure to cisplatin-induced a significant increase in the frequency of polyploid (>16n) cells in the liver of Trp53−/− mice, which was not observed in p53-proficient mice (Supplementary Fig. 9c, d). These data demonstrate that DNA crosslinking agents can induce liver polyploidisation and that the same p53-dependent mechanism acts to restrain polyploidisation following exposure to chemotherapy.