SARS-CoV-2 infection induces DNA damage, through CHK1 degradation and impaired 53BP1 recruitment, and cellular senescence

1Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is an RNA virus, responsible for the ongoing coronavirus disease 2019 (COVID-19) pandemic. Its 30 kb genome encodes 26 polypeptides encompassing 16 non-structural proteins (NSPs), 4 structural proteins such as the nucleocapsid (N) protein, and 6 accessory ones.

2Viral infections can impact on several cellular pathways, including the autophagy pathway, the ubiquitin–proteasome system (UPS) and the DNA damage response (DDR). While the interplay between some DNA viruses and DDR has been studied, much less is known about RNA viruses. Although SARS-CoV-2 infection has been suggested to engage components of the DDR machinery–, a thorough characterization and a mechanistic probing of the impact of SARS-CoV-2 on genome integrity and DDR engagement is lacking.

3The DDR is a network of pathways that sense DNA lesions, signal their presence and coordinate their repair. DNA single-strand and double-strand breaks (SSBs and DSBs) are detected by replication protein A (RPA) and by the MRE11–RAD50–NBS1 (MRN) complex, respectively, which guide the recruitment of the apical DDR kinases ataxia telangiectasia and Rad3-related (ATR) or ataxia-telangiectasia mutated (ATM) at SSBs or DSBs, respectively. ATR and ATM undergo autophosphorylation and phosphorylate several DDR factors, including the effector kinases CHK1 and CHK2, which contribute to enforce cell-cycle arrest. DDR activation can cause cellular senescence, and inflammation or cell death.

4We demonstrated that the induction of a DSB results in the recruitment of the RNA polymerase II complex, which transcribes a novel class of RNA molecules named damage-induced long non-coding RNAs (dilncRNAs)–. These RNAs, by interacting with DDR factors such as p53-binding protein 1 (53BP1), are necessary for their condensation into foci at DSBs, by promoting liquid–liquid phase separation (LLPS),. Inhibiting the synthesis or function of dilncRNAs disrupts DDR foci and impairs DNA repair,,. Interestingly, also SARS-CoV-2 N-protein phase-separates in an RNA-dependent manner,.

5In this Article, we demonstrate that SARS-CoV-2 infection causes DNA damage and activation of an altered DDR. DNA damage is the consequence of the degradation of CHK1 by ORF6 and NSP13 viral factors through the proteasome and the autophagy pathways, respectively. Depletion of CHK1 causes loss of RRM2, a component of the ribonucleotide reductase (RNR) complex, which leads to deoxynucleoside triphosphate (dNTP) shortage that causes impaired S-phase progression, DNA damage accumulation, DDR activation, induction of inflammatory pathways and establishment of cellular senescence. The supplementation of deoxynucleosides (dNs) is sufficient to contrast this cascade of events. In addition to that, SARS-CoV-2 N-protein impairs 53BP1 recruitment at DSB by competing with dilncRNAs binding, ultimately hampering DNA repair. These events occur also in vivo in mice infected by SARS-CoV-2 and in patients with COVID-19.

6Overall, our results reveal the impact of SARS-CoV-2 infection on genome integrity and its contribution to the inflammatory response observed in COVID-19 patients and the recently reported virus-induced cellular senescence.

7We studied the engagement of the DDR pathways at different timepoints upon infection by SARS-CoV-2 of Huh7 cells, a human cell line naturally permissive to SARS-CoV-2 (refs.,), by immunoblotting of whole cell lysates. As negative control we used mock-infected cells; as positive control we exposed cells to hydroxyurea (HU), which induces DNA replication stress and activates the ATR–CHK1 axis,, or ionizing radiation (IR) that causes DSBs and activates the ATM–CHK2 pathway (Fig. 1a). We observed that SARS-CoV-2 infection triggered the autophosphorylation, and thus activation, of the master kinases DNA-PK (pDNA-PKS2056, involved in DNA repair) and ATM (pATMS1981) but not ATR (pATRT1989) (Fig. 1a,b). CHK2, the direct downstream target of ATM, was not detectably phosphorylated on its activating site (T68); similarly, CHK1, a target of ATR, was not phosphorylated on S317. Also P53 was not significantly phosphorylated on S15, an ATM/ATR target site (Fig. 1a,b). Differently, KAP1 (also known as TRIM28), a chromatin-bound ATM target, was strongly phosphorylated (pKAP1S824) together with phosphorylated H2AX (γH2AX) and RPA (pRPAS4/8), markers of DSB and SSB, respectively (Fig. 1a,b). Similar results were generated in infected human lung epithelial Calu-3 cells, (Extended Data Fig. 1a,b).

8To confirm and extend at single-cell resolution the impact of SARS-CoV-2 infection on DDR, we performed quantitative immunofluorescence analyses of the conditions aforementioned. We observed increased numbers of pKAP1S824, pRPAS4/8 and γH2AX foci per cell in infected Huh7 compared with mock-infected cells (Fig. 1c,d). In addition, SARS-CoV-2 infection of human nasal epithelial primary cells (HNEpCs) confirmed DDR activation, as detected by pRPAS4/8 and γH2AX foci (Extended Data Fig. 1c,d).

9To directly monitor the impact of the virus on physical DNA integrity, we performed comet assays in Huh7 and Calu-3 cells. We observed DNA fragmentation induction in both SARS-CoV-2-infected cell lines compared with control conditions, as measured by tail moment (Fig. 1e,f and Extended Data Fig. 1e,f).

10Damaged DNA released in the cytoplasm can be sensed by the cGAS–STING pathway triggering an inflammatory response. We therefore investigated cGAS–STING and other inflammatory pathways in cells upon SARS-CoV-2 infection and observed a higher number of micronuclei, which also stained positive for cGAS (Extended Data Fig. 1g,h) in Calu-3 cells, suggestive of the release of damaged nuclear DNA in the cytosol. In infected Huh7 cells, which do not express cGAS and STING, P38 and STAT1, factors involved in the pro-inflammatory response, were activated (Extended Data Fig. 1i).

11To test the consequences of the activation of these pro-inflammatory pathways, we monitored by quantitative reverse transcription polymerase chain reaction (RT–qPCR) the transcriptional induction of IL6, IL8, CXCL9, CXCL10 and TNFα genes in Huh7 and Calu-3 upon SARS-CoV-2 infection. We detected their significant upregulation in both cell types, although generally stronger in Calu-3 (Extended Data Fig. 1j). Since increased expression of pro-inflammatory genes is consistent with the induction of cellular senescence by SARS-CoV-2, as recently reported,, we tested and confirmed the establishment of cellular senescence following SARS-CoV-2 infection in our settings, as demonstrated by increased senescence-associated β-galactosidase (SA-β-gal) activity (Extended Data Fig. 1l,m), augmented P21- and reduced KI67-positive cells (Extended Data Fig. 1n,o), and no significant induction of apoptosis (Extended Data Fig. 1k).

12In sum, our results obtained by different techniques and in three independent cell types indicate that SARS-CoV-2 infection causes DNA damage and an altered DDR; this is associated with the induction of pro-inflammatory pathways and cytokines and cellular senescence.

13While studying activation of individual DDR proteins, we noticed that total CHK1 protein levels progressively decreased in infected Huh7 and Calu-3 cells (Figs. 1a and 2a–d and Extended Data Fig. 2c,d), mainly post-transcriptionally (Extended Data Fig. 2e,f). CHK1 loss is reportedly sufficient to cause DNA replication stress and DNA damage accumulation. CHK1 controls the expression of the ribonucleoside-diphosphate reductase subunit M2 (RRM2), the small subunit of the RNR enzyme that converts ribonucleoside triphosphates (rNTPs) into dNTPs, necessary for DNA synthesis,. By testing RRM2 messenger RNA and protein levels by RT–qPCR, immunoblotting and immunofluorescence, we consistently observed their progressive and significant decrease following SARS-CoV-2 infection (Fig. 2a–d and Extended Data Fig. 2a–f).

14Then, we measured individual dNTP concentrations in SARS-CoV-2 infected Huh7 and Calu-3 cells and observed reduced levels of cellular dNTPs compared with mock-infected conditions (Fig. 2e).

15dNTP shortage can impair DNA synthesis, ultimately hampering S-phase progression,. To monitor cell-cycle progression, we measured DNA content in infected or mock-infected cells by propidium iodide (PI) staining followed by flow cytometry analysis. We observed a significant accumulation of infected cells in S-phase compared with control samples (Fig. 2f and Extended Data Figs. 3d and 8c). This was confirmed by strongly reduced levels of CDT1, a G1-phase marker– (Extended Data Fig. 2g). By pulse labelling with 5-bromo-2′-deoxyuridine (BrdU) for 1 h before flow cytometry, we observed an increased percentage of BrdU-negative cells in S-phase in infected samples (Fig. 2g and Extended Data Figs. 3d and 8c). Altogether, these results indicate reduced dNTP levels and impaired S-phase progression following infection.

16To determine the causal role of reduced dNTP levels, we tested the impact of dN supplementation to culture medium of infected cells. We observed that dN supplementation was sufficient to reduce DDR activation as shown by immunofluorescence and immunoblots of γH2AX, pRPAS4/8 and pKAP1S824 (Fig. 3a–d), DNA damage accumulation detected by comet assays (Fig. 3e,f) and transcription of several pro-inflammatory cytokines (Fig. 3g).

17Overall, these results are consistent with a model in which SARS-CoV-2 gene products cause CHK1 loss, which reduces RRM2 levels and consequently the pool of available dNTPs, causing impaired DNA replication and S-phase progression, DNA damage accumulation and ultimately fuelling an inflammatory response. Supplementation of dNs is sufficient to tame these events.

18To determine whether CHK1 loss is sufficient to recapitulate the events here described following SARS-CoV-2 infection, we studied the impact of CHK1 depletion by RNA interference. Consistent with previous reports, we observed by flow cytometry that cells knocked down for CHK1 accumulate in S-phase (Extended Data Figs. 3a–d and 8c) and pulse labelling with BrdU for 1 h before flow cytometry analysis revealed a higher fraction of BrdU-negative S-phase cells compared with control samples (Extended Data Figs. 3d and 8c).

19We also observed that CHK1 depletion was sufficient to reduce RRM2 levels and cause DNA damage, as shown by increased pRPAS4/8 and γH2AX signals (Extended Data Fig. 3e,f). In addition, CHK1 knockdown led to the activation of P38 and STAT1 (Extended Data Fig. 3e,f) and formation of γH2AX foci and micronuclei, often positive for cGAS (Extended Data Fig. 3g–j), indicating that CHK1 loss in infected cells probably contributes to the activation of pro-inflammatory pathways. Indeed, cells depleted for CHK1 displayed increased expression of most of the cytokine and chemokine genes tested (Extended Data Fig. 3k,l) and increased secretion of IL6, CXCL9 and CXCL10 in Calu-3 cells as monitored by immunoassays (Extended Data Fig. 3m).

20In sum, CHK1 loss is sufficient to recapitulate several of the events observed in SARS-CoV-2 infected cells, namely, RRM2 reduction, S-phase progression impairment, DNA damage and secretion of inflammatory cytokines.

21To identify the viral gene products responsible for CHK1 downregulation, we individually expressed 24 of the 26 annotated SARS-CoV-2 proteins (SARS-CoV-2 reference genome, NC_045512.2) and analysed by immunoblotting their impact on CHK1 levels. Among the gene products tested, ORF6 and NSP13 were the factors with the strongest and most consistent impact on CHK1 protein levels (Fig. 4 and Extended Data Fig. 4a,b). Their sole expression was also sufficient to reduce RRM2 levels and increase γH2AX and RPA phosphorylation (S4/8) (Fig. 4a–d and Extended Data Fig. 4a,b).

22SARS-CoV-2 ORF6 has been shown to associate with the nuclear pore and to interfere with proteins’ nuclear–cytoplasmic trafficking,.

23We observed that SARS-CoV-2-infected cells displayed cytoplasmic CHK1 localization compared with mock-infected cells in which CHK1 is mainly nuclear (Fig. 2c). It has been reported that accumulation of CHK1 in the cytoplasm leads to its degradation through the UPS. To investigate the potential involvement of ORF6 in the cytoplasmic accumulation of CHK1, we took advantage of a point mutant form of ORF6 (ORF6M58R) unable to interact with the nuclear pore complex. Interestingly, CHK1 protein levels did not decrease in Huh7 cells expressing the mutant ORF6, as detected both by immunofluorescence and immunoblots (Fig. 5a–d). Consistent with that, mutant ORF6 expression also had no impact on RRM2 protein levels and DNA damage accumulation as detected by pRPAS4/8 and γH2AX (Fig. 5c,d). This suggests that ORF6, by altering CHK1 nuclear–cytoplasm shuttling, may cause its degradation.

24We therefore probed the engagement of UPS in ORF6-dependent CHK1 loss. Treatment of ORF6-expressing Huh7 cells with MG132—a proteasome inhibitor—recovered CHK1 protein levels, detected mostly in the cytoplasm (Fig. 5e–h), consistent with impaired protein trafficking of ORF6-expressing cells.

25To demonstrate that SARS-CoV-2 ORF6 expression causes CHK1 poly-ubiquitination, a prerequisite of proteosome-dependent degradation, we immunoprecipitated endogenous CHK1 in Huh7 cells expressing either ORF6 or ORF6M58R, treated or not with MG132, and probed for ubiquitinated CHK1. Proteasomal inhibition led to a higher accumulation of ubiquitinated CHK1 in ORF6-expressing cells compared with samples that overexpressed ORF6M58R (Fig. 5i,j).

26These results indicate that SARS-CoV-2 ORF6 prevents CHK1 nuclear import, causing its accumulation in the cytoplasm and its consequent degradation through the UPS.

27It has been shown that NSP13 can promote protein degradation in an autophagy-dependent manner. Therefore, to test whether CHK1 loss in NSP13-expressing cells was dependent on the autophagic route, we transiently expressed the viral NSP13 gene in Huh7 cells in the presence of either Bafilomycin A1 (BafA1) or chloroquine (CQ), two specific inhibitors of autophagy; efficacy was confirmed by the observed accumulation of P62 cytoplasmic aggregates (Extended Data Fig. 5a,b). NSP13-mediated reduction of CHK1 protein levels was abolished by both treatments (Fig. 5k,l). A short BafA1 exposure highlighted a clear co-localization of CHK1 with P62 cytoplasmic aggregates, suggesting its accumulation in autophagosomes (Fig. 5m,n). To confirm and extend these results, we individually knocked down Beclin 1 (BECN1) and LC3B—two key regulators of autophagy—and observed a significant restoration of CHK1 protein levels in NSP13-expressing cells (Fig. 5o,p and Extended Data Fig. 5c,d).

28These results indicate that SARS-CoV-2 NSP13 causes the accumulation of CHK1 in the cytoplasm, where it co-localizes with P62, in this way promoting its degradation through autophagy.

29We noticed that γH2AX foci accumulation was not accompanied by co-localizing 53BP1 foci in SARS-CoV-2-infected Huh7, Calu-3 and HNEpC (Fig. 6a–d and Extended Data Fig. 6a,b), despite unaltered 53BP1 protein levels (Extended Data Fig. 6c).

30SARS-CoV-2 N-protein is an RNA-binding protein capable to undergo RNA-dependent LLPS,,–. We previously reported that also 53BP1 phase-separates in an RNA-dependent manner. To test the potential impact of N-protein on 53BP1 foci formation, we expressed the viral N gene in Huh7 cells and exposed them to IR. We observed that irradiated cells expressing N showed increased numbers of γH2AX foci per cell, but fewer 53BP1 foci compared with control cells (Fig. 6e,f). To reduce the possibility of an indirect effect mediated by altered gene expression, we micro-injected purified recombinant N-protein into the nuclei of irradiated cells stably expressing 53BP1-GFP and immediately studied the kinetics of 53BP1 foci by live imaging. We observed 53BP1 foci number decreasing with a faster (~8.5-fold) kinetic in cells injected with the N-protein compared with control cells (Fig. 6g, Extended Data Fig. 6d and Supplementary Video 1).

31Next, we sought to elucidate the molecular mechanisms underlying N-protein impact on 53BP1 functions. In co-immunoprecipitation experiments, 53BP1 did not interact with N-protein (Extended Data Fig. 6e). We previously reported that dilncRNAs generated at DSB drive LLPS of 53BP1 (refs.,). Intriguingly, both viral and cellular RNAs have been reported to associate with N-protein and promote its phase separation, as we confirmed (Extended Data Fig. 6f,g).

32Since N-protein, although mainly cytoplasmic, also localizes in the nucleus– (Extended Data Fig. 6j,k), we tested whether N associates with cellular dilncRNAs by performing RNA immunoprecipitation (RIP) against the N-protein in NIH2/4 cells, which we previously characterized for the expression of dilncRNAs upon DSB induction by I-SceI endonuclease. Therefore, following SARS-CoV-2 N gene expression into NIH2/4 cells and DSB induction by I-SceI, we immunoprecipitated N-protein and analysed the associated RNAs by RT–qPCR. We observed that N-protein was associated with dilncRNA upon DSB generation, but not with H2AX mRNA used as a negative control (Fig. 6h). Next, we immunoprecipitated endogenous 53BP1 in I-SceI-induced NIH2/4 cells expressing or not the viral N-protein, and monitored dilncRNA association with 53BP1. We observed that 53BP1 association with dilncRNAs was reduced in cells expressing N-protein (Fig. 6i), despite unaltered 53BP1 protein levels or immunoprecipitation (IP) efficiency following N-protein overexpression (Extended Data Fig. 6h).

33Since 53BP1 plays important DNA repair functions through non-homologous end-joining (NHEJ), we tested the impact of N-protein on NHEJ. We took advantage of a cell line bearing an integrated GFP construct flanked by two I-SceI recognition sites (EJ5-GFP): following I-SceI expression, DSBs are generated and repair can be quantified by qPCR on genomic DNA (gDNA) with primers flanking the re-joined site. EJ5-GFP U2OS were transfected with a plasmid expressing I-SceI together with N-protein or an EV. Seventy-two hours post-transfection, gDNA was collected and analysed. NHEJ efficiency in cells expressing N-protein was significantly decreased compared with control samples (Fig. 6j) to an extent comparable to that previously observed upon 53BP1 depletion,, while leaving I-SceI levels unchanged (Extended Data Fig. 6i).

34In sum, our evidence indicates that SARS-CoV-2 N competes with 53BP1 for dilncRNAs binding and thus reduces 53BP1 focus formation at DSB, ultimately hampering DNA repair by NHEJ.

35We next extended our analyses in in vivo settings of SARS-CoV-2 infection. Lung sections of mice expressing human ACE2 (hACE2) were stained for DDR markers following intranasal administration of SARS-CoV-2 (Fig. 7a,b and Extended Data Fig. 7a,b); mice exposed to IR were used as positive control (Extended Data Fig. 7c,d). Immunostaining of γH2AX and pRPAS4/8 together with N-protein demonstrated strong γH2AX and pRPAS4/8 signal induction in infected samples compared with mock-infected ones. Instead, 53BP1, CHK1 and RRM2 signals were reduced in infected murine lungs (Fig. 7a,b). Thus, consistent with our observations in cultured cells, SARS-CoV-2 infection in mice causes DNA damage accumulation, impairs 53BP1 activation and reduces CHK1 and RRM2 protein levels.

36As SARS-CoV-2 infection has been recently shown to induce cellular senescence and contribute to inflammation in vivo,,,, we probed the same tissues for p16 and p21 and observed an increase of both in infected lungs (Extended Data Fig. 7e,f), with p21 detected almost exclusively in pneumocytes and in bronchial epithelium of the infected lungs, while p16 also associated with inflammatory cells populating the lung parenchyma of infected mice (Extended Data Fig. 7g).

37We next analysed lungs and nasal mucosa sections from patients who died with a clinical diagnosis of COVID-19, determined by a positive naso-pharyngeal swab or bronchoalveolar lavage during their last hospitalization. In situ hybridization (ISH) and immunohistochemistry (IHC) were used to confirm the presence of SARS-CoV-2 RNA or N-protein, respectively. As negative control, we studied the lungs of patients not diagnosed for the pathology (non-COVID) but affected by viral pneumonia of different aetiologies. Since for ethical reasons we were not able to collect nasal mucosa from non-COVID subjects, we compared samples scored for the presence or absence of SARS-CoV-2, as detected by fluorescence in situ hybridization (FISH).

38We probed lung and nasal mucosa for γH2AX, SARS-CoV-2 genome and cytokeratin-18 (CK18), a marker of epithelial cells, and observed increase γH2AX foci in epithelial cells of infected specimens, compared with controls (Fig. 8a,c). Similarly, pRPAS4/8 levels were strongly induced in the lungs of patients with COVID-19 (COVID) compared with non-COVID subjects (Fig. 8e,f).

39We also analysed 53BP1 in both lung and nasal mucosa. In lung samples of patients with COVID, γH2AX foci were significantly less associated with 53BP1 than ex vivo bleomycin-treated sections of human non-COVID lungs (non-COVID bleo+) used as positive control, indicating impaired 53BP1 recruitment in infected lungs (Fig. 8b,d). Similarly, in nasal mucosa, COVID FISH+ cells displayed a reduced recruitment of 53BP1 compared with COVID FISH− cells (Fig. 8b,d). Both CHK1 and RRM2 expression was invariably and significantly lower in infected tissues compared with not infected ones (Fig. 8e–h).

40Staining of lung tissues for P16 and P21 together with SARS-CoV-2 markers revealed that both were increased in patients with COVID compared with non-COVID individuals (Extended Data Fig. 8a,b).

41Overall, these results indicate that SARS-CoV-2 infection in vivo causes an altered DDR activation that is associated with increased levels of DNA damage and cellular senescence in mouse and patients with COVID-19.

42Viruses are known to hijack cellular activities, including DDR, as a strategy to promote their replication,. This may have deleterious effects on the cell, potentially leading to genome instability. SARS-CoV-2 infection has been reported to alter different host pathways,, and to correlate with the activation of some DDR markers and senescence–,,–. The observation that some DDR inhibitors reduce SARS-CoV-2 replication further hints at a mutual interplay. However, when studied, DNA damage was correlated mainly with reactive oxygen species,.

43Here, we showed that SARS-CoV-2 infection causes DNA damage (Fig. 1e,f and Extended Data Fig. 1e,f), as observed in two immortal cell lines, in primary human cells and in vivo in mice and humans. We identified at least two mechanisms responsible for DNA damage accumulation: one impacting on cellular dNTP metabolism leading to DNA replication impairment; another impeding 53BP1 activation and reducing DNA repair (Fig. 8i). The DNA damage accumulated triggers DDR activation, but in an altered way (Fig. 1a,d and Extended Data Fig. 1a,d). For instance, CHK1, together with P53, decreases following SARS-CoV-2 infection (Figs. 1a and 2a–d and Extended Data Fig. 2c,d). Degradation of DDR factors is a strategy shared by different viruses to override host defences,–.

44CHK1 is known to control the expression of E2F transcription factors, important regulators of cell-cycle progression, and consequently of the RRM2 gene, allowing DNA synthesis in S-phase,,. We demonstrated that SARS-CoV-2 infection leads to CHK1 loss and consequent RRM2 decrease (Figs. 1a and 2a–d), causing dNTP shortage and prolonged S-phase (Fig. 2e–g), consistent with the generation of DNA replication stress and DNA damage. This cascade of events leads to the establishment of cellular senescence and activation of pro-inflammatory pathways (Extended Data Figs. 1g–o, 7e–g and 8).

45CHK1 depletion is sufficient to recapitulate RRM2 reduction, DNA damage accumulation and cytokine expression (Extended Data Fig. 3). Importantly, the administration of dNs to SARS-CoV-2-infected cells reduced virally induced DNA damage, DDR activation and cytokine expression (Fig. 3), thus demonstrating the causative role of dNTP depletion in these events.

46We propose that this is probably the unmeant consequence of the dire need for rNTPs of SARS-CoV-2. Staggering two-thirds of total RNA in SARS-CoV-2-infected cells is of viral origin: thus infected cells need to triple their normal RNA synthesis capacity. Therefore, the virus has been under evolutionary pressure to boost rNTP levels. One way is to reduce CHK1 levels, causing decreased RRM2 activity and consequent accumulation of rNTPs at the expense of dNTPs. Interestingly, a similar yet opposite mechanism was observed in the DNA virus HPV31, which boosts RRM2 to increase dNTPs and favour its genome replication.

47At least two SARS-CoV-2 products cause CHK1 degradation. ORF6, by associating with the nuclear pore complex, interferes with CHK1 nuclear import, leading to CHK1 cytoplasmic mis-localization and consequent proteasomal degradation. Notably, a point mutation that disrupts ORF6 binding to the nuclear pore complex prevented CHK1 poly-ubiquitination, degradation and DNA damage accumulation (Figs. 5a–d,i,j). In addition, proteasome inhibition with MG132 in ORF6-expressing cells was sufficient to rescue CHK1 levels (Fig. 5e–h). Differently, NSP13 leads to CHK1 depletion through the autophagic route, as indicated by the recovery of CHK1 levels upon treatment with autophagy inhibitors or with RNAi against key autophagy factors (Fig. 5k–p).

48In addition to induce DNA damage, SARS-CoV-2 inhibits its repair. We observed a strikingly reduced ability of 53BP1 to form DDR foci, despite unaltered protein levels, in infected cells (Fig. 6a–d and Extended Data Fig. 6a–c). We propose that SARS-CoV-2 N, an avid RNA-binding protein, impairs 53BP1 condensation at DSB by competing for dilncRNA binding. Indeed, both 53BP1 and N-protein undergo LLPS in an RNA-dependent manner, and we demonstrate that N-protein, just like 53BP1 (ref. ), binds to dilncRNA (Fig. 6g–i and Supplementary Video 1).

49These data suggest a nuclear role of SARS-CoV-2 N-protein. Although both SARS-CoV and SARS-CoV-2 N-proteins bear functional nuclear localization signals, they are only partly nuclear– (for example, Extended Data Fig. 6j,k), but phylogenetic studies have correlated the enhancement of motifs that promote nuclear localization of viral N-proteins with coronavirus pathogenicity and virulence.

50Intriguingly, enoxacin, a molecule that we reported to boost RNA-mediated 53BP1 foci assembly and DNA repair, has been predicted in silico to reduce SARS-CoV-2 replication. It is tempting to speculate that enoxacin, by enhancing 53BP1 activities, counteracts SARS-CoV-2 N-protein competition with 53BP1 for dilncRNAs.

51Hyperactivation of inflammatory pathways is responsible for fatal COVID-19 cases. DNA damage accumulation and chronic DDR activation are potent inducers of inflammation. Consistent with previous studies,, we observed that SARS-CoV-2 infection of cultured cells activates multiple pro-inflammatory signalling pathways, including cGAS/STING, STAT1 and p38/MAPK (Extended Data Fig. 1g–j), similar to CHK1 depletion (Extended Data Fig. 3e–j). Supported by reports that disruption of the CHK1–RRM2 pathway triggers cellular senescence and our own evidence (Extended Data Figs. 1l–o, 7e–g and 8), we propose that SARS-CoV-2-mediated CHK1 loss promotes a pro-inflammatory programme akin to the senescence-associated secretory phenotype (Extended Data Fig. 3k–m).

52A role for SARS-CoV-2-induced senescence in promoting macrophage infiltration and inflammation in vivo has been proposed. We observed that SARS-CoV-2 infection causes DNA damage accumulation that correlates with markers of cellular senescence, in primary cells and in vivo (Extended Data Figs. 1l–o, 7e–g and 8). In particular, infected pneumocytes express high p21 levels, while polymorphonuclear and monocytoid inflammatory elements have elevated p16 (Extended Data Fig. 7g), reminiscent of a two-wave model of inflammatory response: an initial cell-intrinsic one and a second one triggered by the immune system,.

53Altogether, our results indicate that SARS-CoV-2-induced DNA damage triggers a cell-intrinsic pro-inflammatory programme that, in concert with the immune response, fuels the strong inflammatory response observed in patients with COVID-19. The observed ageing phenotypes recently reported in patients with severe COVID-19 (refs.,) are consistent with our observations.

54Finally, by proposing a mechanism for the generation of DNA damage and the activation of DDR pathways and of a pro-inflammatory programme, we provide a model to improve our understanding of SARS-CoV-2-induced cellular senescence,. In this regard, it will also be interesting to determine if persistent DNA damage and DDR activation, features of cellular senescence,, following SARS-CoV-2 infection, contribute to the chronic manifestations of the pathology known as long COVID.

55Experiments involving animals have been carried out in accordance with the Italian Laws (D.lgs. 26/2014), which enforce Directive 2010/63/EU (Directive 2010/63/EU of the European Parliament and of the Council of 22 September 2010 on the protection of animals used for scientific purposes). Accordingly, the project has been authorized by the Italian Competent Authority (Ministry of Health). B6.Cg-Tg(K18-ACE2)2Prlmn/J mice were purchased from The Jackson Laboratory. Mice were housed under specific-pathogen-free conditions as already described. SARS-CoV-2 isolation and propagation for infection in mice has been carried out as shown before. Virus infection of K18-hACE2 mice was performed via intranasal administration of 1 × 105 tissue culture infectious dose 50 per mouse under isoflurane 2% (# IsoVet250) anaesthesia, as described. Mice were monitored to record body weight, clinical and respiratory parameters. Mice were killed by cervical dislocation. At the time of autopsy, mice were perfused through the right ventricle with phosphate-buffered saline (PBS). Lung tissues were collected in liquid nitrogen or in Zn-formalin and transferred into 70% ethanol 24 h later. For irradiation experiments, an 11-month-old C57BL/6J mouse was irradiated with 2 Gy TBI using GADGIL X-Ray irradiator. An age- and sex-matched mouse was used as control. Mice were killed by CO2 inhalation at 1 h post-IR, and lungs were collected for fixation in 10% neutral buffered formalin overnight, washed in water and paraffin-embedded for histological analysis. RNA from mouse lungs was extracted and analysed as described.

56Histological analysis of lungs of patients with COVID-19 and subjects not diagnosed with the pathology (non-COVID)—but affected by viral pneumonia (influenza virus infection)—was performed by expert technicians and pathologists at the Pathology Unit of Trieste University Hospital. The same pathologists analysed all samples considered in this study, excluding operator-dependent biases. Viral presence in nasal mucosa of patients with COVID-19 was assessed by FISH against SARS-CoV-2 RNA. This study was approved by the Joint Ethical Committee of the Regione Friuli Venezia Giulia, Italy (re. 0019072/P/GEN/ARCS). All patients provided written informed consent to the use of their samples for research purposes at the time of hospital admission.

57Vero E6 (ATCC-1586), Huh7 (kindly provided by Ralf Bartenschlager, University of Heidelberg, Germany) and Calu-3 cell lines (ATCC HTB-55) were cultured in Dulbecco’s modified Eagle medium (DMEM, ThermoFisher) supplemented with 10% foetal bovine serum (FBS, ThermoFisher) and 50 µg ml−1 gentamicin. 53BP1-GFP U2OS, EJ5-GFP U2OS and NIH2/4 cell lines were cultured as already described,,. Cell cultures were maintained at 37 °C under 5% CO2. Cells were routinely tested for mycoplasma contamination. DNA damage was induced in cultured cells by 6 mM HU treatment for 4 h, or by exposure to IR (2 Gy) and analysed 15 or 60 min post-IR. Proteasome inhibition was conducted by incubating Huh7 cells for 6 h in growth medium containing 10 μM MG132 (M7449, Sigma). To inhibit autophagy, Huh7 cells were incubated with medium supplemented with either 100 nM Bafilomycin A1 (S1413, Selleckchem) or 50 μM CQ (C6628, Sigma) for 6 h, unless indicated otherwise.

58Working stocks of SARS-CoV-2 ICGEB-FVG_5 isolated and sequenced in Trieste, Italy, were propagated on semiconfluent Vero E6 cells. Cultured Huh7 and Calu-3 cells were mock-infected or infected at a multiplicity of infection (MOI) of 0.1, in DMEM without supplements for 1 h at 37 °C and 5% CO2. Then, the non-bound virus was rinsed off with PBS 1× and fresh DMEM containing 2% of heat-inactivated FBS was added to the cells. Uniformity of viral infection in all experiments was confirmed by viral titration, RNA detection and immunofluorescence.

59HNEpCs were collected from healthy adult volunteers. Nasal cavities were anaesthetized using lidocaine, and nasal epithelial cells were collected by repeatedly scraping turbinates with a disposable bronchial cytology brush (CONMED). The tissue was resuspended in DMEM supplemented with 10% FBS and centrifuged at 300g for 5 min. The pellet was digested in a solution containing 1,000 U of Accutase (Sigma, A6964), 5,000 U of Dispase (Corning, 354235) and 1 mg ml−1 DNAse II (Sigma, D8764) for 8 min at 37 °C. The digestion was stopped by adding an equal volume of Pneumacult medium (Stem Cell Technologies, 5050) supplemented with 10% FBS and filtered through a 100 µm cell strainer. The cell suspension was centrifuged at 300g for 5 min, and the pellet was resuspended in red blood cell lysis solution (150 mM NH4Cl and 10 mM KHCO3) for 2 min. Cells were centrifuged at 300g for 5 min, resuspended in medium and seeded on eight-well chambers (Ibidi) at a concentration of 5 × 105 cells per well. After 48 h, cells were infected with SARS-CoV-2 (ref. ) at 1 MOI as described above. Cells were cultured for 48 or 72 h and fixed in 4% paraformaldehyde for 15 min at room temperature for staining.

60Fixed HNEpCs were permeabilized with 0.5% Triton X-100 (Sigma, T8787) in PBS for 15 min at room temperature. FISH to detect SARS-CoV-2 genome was performed using a kit by Molecular Instruments, following the manufacturer’s instructions. After FISH, samples were incubated with a blocking solution (0.5% Triton X-100, 10% horse serum in PBS) for 30 min and incubated for 16 h at 4 °C with primary antibodies (Supplementary Table 1) diluted in the blocking solution. Samples were stained with the appropriate secondary antibody (Supplementary Table 1). Nuclei were counter-stained using Hoechst 33342 trihydrochloride (Invitrogen, h3570) and samples mounted using Mowiol (Sigma, 81381).

61Whole cell extracts were obtained by lysing Huh7 and Calu-3 in 1× Laemmli buffer (2% SDS, 10% glycerol and 60 mM Tris pH 6.8). Before fractionation on 4–12% gradient SDS–PAGE (ThermoFisher), whole extracts were boiled and sonicated for 15 s at low intensity using Bioruptor Next Gen (Diagenode) in a water bath at 4 °C. Proteins were then transferred onto nitrocellulose membrane and analysed as described before with the antibodies listed in Supplementary Table 1. Bio-Rad Image Lab 6.1 was used for immunoblot data collection.

62Quantitative immunofluorescence assays were carried out in cultured cells as described, with minor modifications. Specifically, Calu-3 were fixed with 4% paraformaldehyde and permeabilized first with methanol/acetone (1:1) for 2 min at room temperature and then with Triton 0.2% in PBS 1× for 10 min at room temperature. The antibodies used are listed in Supplementary Table 1. Images at widefield microscope were acquired with MetaVue software or with Zen 2.0 Software (Zeiss); confocal microscope images were collected with Leica Application Suite X; data analyses were carried out by using CellProfiler 3.1.9 or ImageJ 1.53a.

63SARS-CoV-2-infected or mock-infected Huh7 and Calu-3 were collected in TriFast (EMR507100, Euroclone), and total RNA was extracted using RNeasy Kit (Qiagen). RNA from short interfering RNA (siRNA)-transfected Calu-3 and Huh7 was purified with Maxwell RSC simplyRNA Tissue Kit (Promega). DNase I was added during RNA purification, following the manufacturers’ protocols. Purified total RNA was reverse transcribed with SuperScript VILO cDNA Synthesis Kit (ThermoFisher), and complementary DNA was analysed by qPCR using SYBR Green I Master Mix (Roche) with the primers listed in Supplementary Table 2.

64Alkaline comet assay was performed on 48 h SARS-CoV-2-infected or mock-infected Huh7, or on 24 h infected or mock-infected Calu-3 using CometAssay Reagent Kit for Single Cell Gel Electrophoresis Assay (Trevigen, 4250-050-K), following the manufacturer’s instructions. Tail moment was measured using CometScore 2.0 software.

65SA-β-gal activity was measured in 48 h SARS-CoV-2-infected or mock-infected Huh7 or in 72 h infected or mock-infected HNEpC as described. Cells treated with 6 mM HU treatment for 4 h were used as positive controls of senescence induction.

66dNTP pools were extracted from 48 h SARS-CoV-2-infected or mock-infected Huh7 or from 24 h infected or mock-infected Calu-3. Cell plates were carefully washed free of medium with cold PBS and extracted with ice-cold 60% methanol. Methanolic extract was centrifuged, boiled for 3 min, brought to dryness by centrifugal evaporation (Savant, SC100 SpeedVac Concentrator and RT100A Refrigerated Condensation Trap) and stored at −80 °C until use. Cells left on the plate were dried and dissolved on 0.3 M NaOH, and the absorbance at 260 nm of the lysates was used as an index of cell mass, in turn an approximation for cell number. The dry residue was dissolved in water and used to determine the size of dNTP pools by the DNA polymerase-based assay as described. Two different aliquots of each pool extract were analysed, and pool sizes were normalized by the A260nm of the NaOH lysates.

67For cell-cycle analysis, SARS-CoV-2-infected or mock-infected Huh7 were pulsed with 10 µM BrdU (Sigma-Aldrich) for 1 h. Then, cells were collected and fixed first in formaldehyde 2% and then in 75% ethanol. Cells were probed with anti-BrdU primary antibody (Supplementary Table 1) diluted in PBS 1% bovine serum albumin (BSA) at room temperature for 1 h. After washing, cells were incubated with the secondary antibody, diluted 1:400, at room temperature for 1 h in the same buffer. Finally, cells were stained with PI (Sigma-Aldrich, 50 µg ml−1) in PBS 1% BSA and RNase A (Sigma-Aldrich, 250 µg ml−1). Samples were acquired with Attune NxT (ThermoFisher) using a 561 nm laser and 695/40 filter for PI; 488 nm laser and 530/30 filter for BrdU. Analysis was carried out using FlowJo 10.7.1 (BD Biosciences). At least 104 events were analysed for each sample.

68pLVX-EF1α-2xStrep-IRES-Puro vectors encoding for SARS-CoV-2 proteins are a kind gift from Professor Nevan J. Krogan. pCAGGS-ORF6 and -ORF6M58R vectors were gently provided by Lisa Miorin. pcDNA3.1 (+) Mammalian Expression Vector (ThermoFisher) was used as a control EV where indicated. Plasmids were transfected with Lipofectamine 2000 Transfection Reagent (ThermoFisher) in Opti-MEM (ThermoFisher). siRNAs were purchased from Dharmacon (Horizon) and transfected into Huh7 or Calu-3 with Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher) in Opti-MEM. Cells were collected for analyses at 48 or 72 h post-transfection.

69Bio-Plex Pro-Human Cytokine Immunoassay kits (Bio-Rad) were used to measure the concentration of pro-inflammatory cytokines and chemokines in cell supernatants, according to the manufacturer’s guidelines. The magnetic-bead-based antibody detection kit allows simultaneous quantification of the analytes of interest. Each sample was tested undiluted and in duplicate. Positive and negative controls were included in the plate. The plate was read on the Bio-Plex 200 system (Bio-Rad), powered by the Luminex xMAP technology. The concentration of analyte bound to each bead was proportional to the median fluorescence intensity of the reporter signal, and was corrected by the standards provided in the kit (Bio-Rad). Data were expressed in pg ml−1.

70Following exposure to SARS-CoV-2 (0.1 MOI) for 1 h, Huh7 were washed once with PBS 1× and then incubated with fresh DMEM containing 2% of heat-inactivated FBS supplemented with 5 μM deoxy-adenosine (dA), deoxy-cytidine (dC) and deoxy-guanidine (dG) (D8668, D0776 and D0901, Sigma) for 48 h before being collected for downstream analyses. dN supplementation in Calu-3 was carried out by incubating cells with medium containing 50 μM dA and dC, 5 μM dG and deoxy-thymidine (dT, T1895, Sigma) for 24 h.

71Forty-eight hours post plasmid transfection, Huh7 were lysed with IP buffer (150 mM KCl, 25 mM Tris–HCl pH 7.4, 5% glycerol, 0.5% NP40, 10 mM MgCl2 and 1 mM CaCl2), supplemented with 1× protease inhibitors (Roche) and 250 U ml−1 Benzonase (E1014-25KU, Sigma) for 45 min at 4 °C. Lysates were cleared by centrifugation and supernatants incubated overnight at 4 °C with 1:200 of anti-CHK1 (Supplementary Table 1), previously coupled with Protein G Dynabeads (10004D, ThermoFisher). After five washes with IP buffer, proteins were eluted with 1× LDS buffer (B0007, ThermoFisher), separated by 4–12% SDS–PAGE (ThermoFisher) and transferred onto nitrocellulose membrane (Amersham). Membrane was autoclaved before overnight blocking at 4 °C in 5% BSA TBS-T. After blocking, membrane was incubated with anti-Ubiquitin antibody (Supplementary Table 1) 1:1,000 in 5% BSA TBS-T for 1 h at room temperature.

72pET28 plasmid bearing SARS-CoV-2 N-protein fused to a C-terminal His6-tag (a kind gift from S. Pasqualato, Human Technopole, Milan, Italy) was expressed in Escherichia coli BL21-CodonPlus (DE3)-RP (Agilent) upon induction with 500 µM isopropyl-β-d-1-thiogalactopyranoside for 16 h at 18 °C. Cells were resuspended in resuspension buffer (25 mM Tris pH 8, 500 mM NaCl, 5% glycerol, 2 mM β-mercaptoethanol and 10 mM imidazole), supplemented with protease inhibitors and TurboNuclease and sonicated. After polyethyleneimine addition and centrifugation, the supernatant was applied onto Ni-NTA agarose beads (Qiagen) and the His6-nucleocapsid protein was eluted in elution buffers containing 250 mM and 500 mM imidazole. The eluted protein was applied on a HiTrap Heparin HP column (Cytiva) and loaded on a Superdex 200 Increase 10/300 GL (Cytiva) pre-equilibrated in SEC buffer (25 mM Tris pH 8, 500 mM NaCl and 2 mM β-mercaptoethanol). Protein purity was assessed by Comassie blue SDS–PAGE as >90% pure.

73Before micro-injection experiments, both purified N-protein and acetylated BSA (ThermoFisher) were dialysed overnight against micro-injection buffer (25 mM Tris pH 8 and 150 mM NaCl). Micro-injection experiments were performed as already described, with minor changes. An UltraVIEW VoX spinning-disk confocal system (PerkinElmer) with a motorized Luigs & Neumann SM7 micromanipulator was used. Glass borosilicate capillaries were pulled to a diameter of 0.7 μm using a Flaming/Brown micropipette puller Model P-1000 (Sutter Instrument) and loaded with 6 mg ml−1 N-protein or BSA solution as control, in presence of Rhodamine B Dextran (ThermoFisher) as marker for micro-injected cells. Samples were injected into cell nuclei 30 min after irradiation (2 Gy) using a FemtoJet pump (Eppendorf) at a constant pressure of 20 hPa. A z-stack was acquired every minute for 100 min total using a 60× oil-immersion objective (Nikon Plan Apo VC, 1.4 numerical aperture). To excite GFP and Rhodamine B, 488 nm and 561 nm lasers were used, respectively. Images in micro-injection experiments were acquired with Volocity 6.4.0. Quantification of 53BP1-GFP foci per nucleus per single frame was performed using the software CellProfiler 3.1.9. Foci number per nucleus was plotted along time, and the corresponding curve was fitted to an exponential function to determine the decay rate (k) using Prism 9.3.0 software.

74Purified recombinant N-protein (5 mM) in micro-injection buffer (25 mM Tris pH 8, 150 mM NaCl) was mixed with increasing concentrations of RNA extracted from SARS-CoV-2-infected (48 h) or mock-infected Huh7. Turbidity of the solutions was measured at 350 nm using a NanoVue Plus Spectrophotometer (Cytiva) after 30 min, 2 h and 24 h of incubation at room temperature.

75Huh7 were transfected with the plasmid encoding for N-protein or with an EV as control. Forty-eight hours post-transfection, cells were irradiated or not. One hour post-IR, cells were collected by trypsinization and washed twice in ice-cold 1× PBS. Cell pellets were lysed in IP buffer supplemented with 1× protease inhibitors (Roche), 0.5 mM dithiothreitol, 40 U ml−1 RNaseOUT and 1,000 U ml−1 DNase I (Roche), and incubated at room temperature for 15 min and at 4 °C for an extra 15 min with gentle rotation. Lysates were cleared by centrifugation at maximum speed for 20 min at 4 °C. After addition of 5 mM EDTA (pH 8), lysates were incubated overnight at 4 °C with 5 μg of anti-N-protein (40588-T62 Sino Biological) or with normal rabbit IgGs (Cell Signaling), which were previously bound to Dynabeads Protein G (ThermoFisher). After five washes with 1 ml of IP buffer, immunoprecipitated proteins were analysed by immunoblotting.

76RIP and dilncRNA analysis was carried out in NIH2/4 as previously described, with minor modifications. Briefly, I-SceI-GR-expressing NIH2/4 cells were transfected with the plasmid encoding for N-protein or with an EV as a control at 24 h before triamcinolone acetonide 0.1 µM (Sigma-Aldrich) administration. IP was performed using 5 μg of anti-N-protein, or 10 μg of anti-53BP1 (Supplementary Table 1), or with normal rabbit IgGs (Cell Signaling) as a mock IP.

77Evaluation of repair efficiency by NHEJ following viral N-protein expression was conducted as shown previously, with minor modification: HA-I-SceI expressing plasmid was transfected along with the plasmid encoding for SARS-CoV-2 N-protein.

78IHC staining was performed as previously reported. Double IHC staining was conducted by applying SignalStainBoost IHC Detection (#18653; #31926, Cell Signaling) and Vulcan Fast Red as substrate chromogen.

79For combined ISH and IHC staining, SARS-CoV-2 RNA probe hybridization (RNAscope Probe - V-nCoV2019-S; 848561; ACD) was performed using RNAscope 2.5 HD Detection Reagent-BROWN (ACD), following the manufacturer’s protocol. After blocking, samples were incubated with primary antibodies (Supplementary Table 1) following by SignalStainBoost IHC Detection and Vulcan Fast Red. Nuclei were counter-stained with Gill’s Hematoxylin n.1 (Bio-Optica). Quantitative analyses of IHC experiments were performed by calculating the average percentage of positive signals in five non-overlapping fields at medium-power magnification (×200) using the Nuclear v9 or Positive Pixel Count v9 ImageScope software (v12.3.2.8013, Leica Biosystems). Combined ISH/IHC images were analysed through the use of a segmentation-based software (HALO v3.5.3577.140, Indica Labs), quantifying the percentage of positive cells.

80Double-marker immunofluorescence was conducted as previously described.

81All the data are shown as mean ± standard error of the mean (s.e.m.) of three independent experiments, unless stated otherwise in the figure legends. We did not use any criteria to determine the sample size. As much data as possible was collected depending on the nature of the experiments or to allow statistical analysis. Where single-cell analysis was performed, we scored at least 30 cells per condition in each biological replicate. Throughout the manuscript, no data were excluded. Only in rare occasions were individual data points removed following unbiased criteria of outlier identification using Prism 9.3.0 software. In addition, in the case of IHC staining in infected mouse lungs, we excluded the tissue sections that show low rates of SARS-CoV-2 infection. The number of animals and replicates are indicated in each figure legend. Animals were randomized to the experimental groups. For in vitro experiments, wells were randomly assigned into each group and all cells were analysed equally. No blinding method was applied. No statistical method was used to pre-determine sample sizes. Ordinary one-way analysis of variance (ANOVA) with Dunnet’s post hoc test was applied in Figs. 1b and 2b and Extended Data Figs. 1b, 2d–f, 4d and 5c. Unpaired two-tailed t-test was applied in Figs. 1d,f, 2d, 6b,d,g, 7b and 8b,d,f,h and Extended Data Figs. 1d,f,h,o, 2b,n,o,p,r, 5b, 6d,f and 7b. Multiple paired two-tailed t-test was applied in Extended Data Fig. 1j. Ordinary one-way ANOVA with Fisher’s least significant difference test was applied in Figs. 2j,l,n,o, 3d, 4h and 5b,f and Extended Data Figs. 1m and 2q. Ordinary one-way ANOVA with Šidák’s post hoc test was applied in Figs. 2e, 4j and 6f,i,j. Ordinary two-way ANOVA with Šidák’s post hoc test was applied in Fig. 2g and Extended Data Fig. 5d. Ratio paired two-tailed t-test was applied in Figs. 2h and 6h. Ordinary two-way ANOVA with Bonferroni’s post hoc test was applied in Extended Data Fig. 2k. Repeated measures one-way ANOVA with Fisher’s least significant difference test was applied in Extended Data Fig. 2s. Ordinary one-way ANOVA with Tukey’s post hoc test was applied in Figs. 3b, 4b,d,f and 5d and Extended Data Figs. 4b and 5k.

82Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

83Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41556-023-01096-x.