The proinammatory response induced by the Cytolethal Distending Toxin depends on cGAS

The Cytolethal Distending Toxin (CDT) is a bacterial genotoxin produced by pathogenic bacteria causing major foodborne diseases worldwide. CDT activates the DNA damage response and induces inammatory signatures in host cells, but the precise relationship between these outcomes has not been addressed so far. Here, we show that the cellular proinammatory response and senescence induced by CDT depend on the cytoplasmic DNA sensor cGAS through the recognition of micronuclei. Indeed, despite active cell cycle checkpoints and in contrast to other DNA damaging agents, cells exposed to CDT reach mitosis where they accumulate massive DNA damage, resulting in chromosome fragmentation and micronucleus formation in daughter cells. These phenotypes are observed with CDT from various origins and in cancer or normal cell lines. Thus, our results establish a direct link between CDT-induced DNA damage, genetic instability and the cellular inammatory response that may be relevant in the context of natural infection associated to chronic inammation or carcinogenesis.


Introduction
Cytolethal Distending Toxin (CDT) is a bacterial genotoxin produced by more than 30 phylogenetically distant Proteobacteria, including pathogenic strains responsible for major in ammatory diseases all over the world, like Escherichia coli (E. col), Campylobacter jejuni (C. jej), Salmonella enterica, Shigella spp., Haemophilus ducreyi (H. duc), Aggregatibacter actinomycetemcomitans (A. act) or Helicobacter spp. [1]. In 2010, CDT-producing bacteria caused 260 million foodborne illnesses worldwide, either diarrhoeal or invasive infectious diseases, resulting in 200,000 deaths [2]. While the current model supports that CDT modulates the host immune microenvironment to promote persistent bacterial colonization, the precise role of CDT in pathogenicity is still to be elucidated.
Several lines of evidence suggest that CDT contributes to the proin ammatory environment in the colonized tissues during bacterial infection. First, more than 80% of H. duc and A. act strains, respectively responsible for chancroid lesions and periodontitis, harbor the CDT operon [3]. In C. jej, the most common reported cause of human bacterial gastroenteritis, the prevalence of CDT genes is higher in strains isolated from diarrhoeal stool samples, reaching 89 to 99% [4][5][6]. More generally, post-infectious irritable bowel syndrome (IBS) from bacterial origins are attributable to four CDT-producing bacteria: Campylobacter spp., Salmonella enterica, Shigella spp. or E. col [7]. Antibodies against the CDT catalytic subunit are frequently produced during in ammatory bowel disease (IBD) with diarrhea and have thus been proposed to serve as biomarkers [8]. CDT proin ammatory effects have also been depicted in rodent models. Partially puri ed CDT from Shigella dysenteriae is diarrheagenic and causes colonic erosions in suckling mice [9]. Moreover, only CDT-positive Helicobacter hepaticus (H. hep) were able to promote typhlocolitis in the IL10 -/mice model for IBD [10]. Of note, mice with genetically altered in ammatory responses are more susceptible to CDT. While an active CDT toxin slightly enhances gastritis in wild-type mice infected by C. jej, this effect is much more pronounced in absence of NF-kB [11].
CDT-mediated proin ammatory responses have been studied through the release of proin ammatory mediators. In mice, production of an active CDT by H. hep promotes the hepatic expression of TNFa, IFNg, Cox2, Interleukin-6 (IL6) and the NF-kB subunits p50 and p65 four months post-infection [12].
Interestingly, these mice latter develop hepatic dysplasic nodules ten months post-infection, questioning on possible link between CDT-mediated chronic in ammatory response and tumorigenicity. Numerous in vitro studies have con rmed that CDT-intoxicated cells produce proin ammatory cytokines [13]. When exposed to A. act CDT, peripheral blood mononuclear cells secrete IL-1b, IL-6, IL-8 and IFN-g [14] and macrophages secrete IL-1b, IL-6 and TNFa [15]. Induction of proin ammatory cytokines after CDT is not restricted to immune cells. Human gingival broblasts treated with A. act CDT produce IL-6 [16]. In the same way, different colorectal cell lines treated with C. jej or H. hep CDT produce IL-8 and other cytokines in a NF-kB-dependent manner [16,17]. Understanding CDT proin ammatory properties is thus necessary to unravel its role during bacterial infection.
CDT is composed by three subunits, CdtA, CdtC and the catalytically active CdtB, except for Salmonella enterica for which CdtB is associated to PltA and PltB [18,19]. Whatever the toxin structure, CdtB catalytic activity is essential to mediate CDT-related cellular defects and pathogenicity [20]. CdtB belongs to the cation-dependent endonuclease-exonuclease-phosphatase (EEP) superfamily and has been structurally and functionally related to DNase I [18,21]. Numerous studies have reported that the genotoxic activity exerted by CDT in a broad range of host cell lineages controls their outcome [22]. Remarkably, CDTexposed cells accumulate DNA strand breaks and therefore activate the DNA Damage Response (DDR) through ATM-CHK2 and ATR-CHK1 signaling pathways to promote cell cycle arrest and apoptotic cell death [23][24][25]. Cells that survive the acute phase of intoxication enter a senescent state, characterized by a permanent cell cycle arrest, persistent DDR activation, enhanced beta-galactosidase (b-Gal) activity and cell distension [26]. Interestingly, another feature of senescent cells, the senescence-associated secretory phenotype consisting in the production of cytokine [27], has also been documented after CDT-induced senescence, notably the proin ammatory mediators IL-1β, IL-6 and IL-8.
It is therefore tempting to speculate on a correlation between DNA damage, senescence and the proin ammatory signatures generally observed in response to CDT. Indeed, unrepaired DNA lesions can result in the formation of micronuclei (MN) after cell division, which induce a proin ammatory response once sensed as cytosolic DNA by the cyclic guanosine monophosphate (GMP)-adenosine monophosphate (AMP) synthase (cGAS) [28,29]. Activated cGAS then synthetizes cyclic GMP-AMP, a second messenger that binds the Stimulation of Interferon Genes (STING) adaptor to activate TBK1, IRF3 and NF-kB, eliciting a type I interferon (IFN) signature and promoting senescence [30].
Here, we identify cGAS as a major regulator of CDT-induced proin ammatory response and senescence. cGAS binds MN in human cells chronically exposed to CDT, promoting a strong proin ammatory response and activation of type I IFN signaling. We show that this response depends on CdtB catalytic activity and DNA damage induction. In contrast to other genotoxic compounds, cells exposed to CDT do not primarily block at G2/M checkpoint of the cell cycle but progress through mitosis, where they suffer massive DNA damage leading to chromosome fragmentation and missegregation. This atypical scenario, observed with CDT from various origins in cancer or normal cell lines, results in MN formation within daughter cells. As MN are drivers of genetic instability and may sustain chronic in ammation through the cGAS-STING axis, these ndings suggest that CDT modulates the host immune response and promotes tumorigenic processes mainly through MN formation.

Results
CDT induces a proin ammatory signature related to type I interferon To determine wether chronic exposition to CDT induces an in ammatory response in the host, a model human cancer cell line (HeLa) was chronically exposed to 0.25 ng/ml of E. col CDT, inducing more than 95% cell death after 10 days (Fig. S1). The surviving fraction was cultured for 40 more days in presence of CDT and individual clones were selected as well as a pool of resistant cells (Fig. 1A). Compared to a short-term exposure, cells chronically treated to CDT (50 days total) do not show signi cant increase of gH2AX level, used as a surrogate of DNA damage signaling (Fig. 1B). In addition, these cells were unresponsive to the CDT-mediated G2/M checkpoint (Fig. 1C), suggesting an adaptation to the CDT toxin.
However, chronically exposed-cells exhibit a higher proportion of micronucleated cells, indicative of important chromosomal instability (Fig. 1D). These cells were subjected to transcriptomic analyses and compared to two control groups, i.e. cells without treatment or chronically exposed to the CDT catalytic dead mutant, bearing the H153A substitution on CdtB, which cannot induce DNA damage nor activate DDR [31,32]. As depicted in the heatmap resuming expression pro le of 9703 signi cantly regulated genes between the three conditions, individual clones and the pool of cells chronically exposed to active wild-type (WT) CDT share a common transcriptional adaptation, whereas the two control groups (nontreated and treated with mutant CDT) cannot be distinguished (Fig. 1E). The majority of the most upregulated genes, when comparing the three groups, depends on the catalytic activity of CDT rather than the presence of the toxin solely (Fig. 1F). Strikingly, the most upregulated biological processes in cells chronically exposed to WT CDT mainly rely on proin ammatory responses, more particularly to type I IFN signaling (Fig. 1G). To con rm this result, mRNA expression level of different cytokines (IL1b, IL6, IL8 and IL10) and a panel of type I IFN signaling genes (OAS1, MX1, ISG15, IFIT1, IFIT2, IFI6 and IFI44) were determined after only 2 days of CDT WT or after repeated treatment with CDT WT or H153A during 40 days ( Fig. 1H and 1I). In cells chronically exposed to CDT WT, all tested genes showed a strong mRNA expression enhancement, from 8.8-fold for IL6 to 232-fold for MX1, except the anti-in ammatory mediator IL10 for which the increase is less than 2-fold. This depends on the CdtB catalytic activity, as the H153A mutation abolishes expression pro le modi cation. In the same way, short CDT WT exposure during 2 days does not signi cantly alter the expression level of any tested mRNA. Taken together, these results show that cells chronically exposed to CDT accumulate MN and display a proin ammatory response related to type I IFN signaling. cGAS binds CDT-mediated micronuclei to promotes senescence and type I IFN signaling Recent reports have demonstrated that MN recognition by cGAS triggers innate immune activation related to type I IFN signature [28,29]. We thus questioned whether cGAS could bind to CDT-induced MN ( Fig.  2A). In absence of treatment, 8.6 % of HeLa cells are micronucleated (Fig. 2B). At 0.025 ng/ml of CDT, the proportion of cells with MN reaches 16.4 % after 24 h and remains stable after 72 h. Increasing CDT concentration to 0.25 or 2.5 ng/ml does not clearly impact the percentage of MN-containing cells at 24 h. However, after 72 h, the proportion of micronucleated cells increases to 34.7 % with 0.25 ng/ml of CDT and 37.5 % with 2.5 ng/ml. These data demonstrate that prolonged CDT exposure favorizes MN formation. Then, we examined the proportion of MN recognized by cGAS or stained with a gH2AX antibody. After 24 h of CDT exposure, a dose dependent increase of gH2AX-positive MN can be observed (Fig. 2C). In contrast, the proportion of cGAS-positive MN increases only after 72 h, with MN progressively accumulating gH2AX staining by increasing CDT concentration. Therefore, cGAS recognizes CDT-induced MN, but this binding is delayed in time.
To better understand the role of cGAS in response to CDT injury, cGAS knockout HeLa cells (cGAS -/-) were generated (Fig. 2D). cGAS -/cells are more resistant to low CDT concentrations (0.025 and 0.25 ng/ml) than their WT cGAS counterpart (Fig. 2E). Moreover, despite similar viability loss between cGAS +/+ and cGAS -/cells at 2.5 ng/ml of CDT, cell distention is not observed in cGAS mutant cells (Fig. 2F). In the same way, the increase of β-Gal staining, a marker of cellular senescence, is abolished in cGAS-depleted cells ( Fig. 2G and 2H). Finally, mRNA expression of three type I IFN signaling target genes (MX1, ISG15 and IFI44) was compared between cGAS +/+ and cGAS -/cells after 10 days of CDT intoxication. An approximately 3-fold increased expression level of each tested gene after chronic CDT exposure was observed in cGAS +/+ but absent in cGAS -/cells (Fig. 2I). Altogether, these results unravel the essential role of cGAS during CDT intoxication through MN recognition, eliciting cell distention, senescence and type I IFN signature. CDT-exposed cells reach mitosis despite active G2 cell cycle checkpoint As cGAS-mediated proin ammatory response depends on MN recognition, we next asked whether MN formation is the direct consequence of CDT intoxication. Previous studies from our lab and others showed that CDT-induced DNA damage activate the G2/M checkpoint [22]. However, G2/M checkpoint arrest is inconsistent with MN formation that requires mitosis completion. To explain the accumulation of MN following CDT treatment, we rst monitored DNA damage markers. As DNA damage checkpoints relies on DDR activation, phosphorylation of H2AX at S139 (gH2AX), CHK1 at S345 (pCHK1) and CHK2 at T68 (pCHK2) were measured after a 24 h treatment with CDT (Fig. 3A). Strong DDR activation is only observed at high concentration of CDT (2.5 and 25 ng/ml). This result supports that at low concentrations, the proliferation defects induced by CDT (Fig. S1) is unlikely the consequence of a rapid DDR activation and a checkpoint-induced cell cycle arrest. Indeed, the CDT-mediated cell cycle arrest signi cantly increases from 24 to 72 h (Fig. 3B), implying that at least a part of CDT-exposed cells reach mitosis before to block their cell cycle during the next rounds of cell division. In contrast, exposure to etoposide (etop), camptothecin (campto) or mitomycin C (MMC), three other genotoxic compounds, induces a rapid and stable cell cycle block over time (Fig. S2). To test whether DDR does not effectively abrogate cell proliferation during early phase of CDT intoxication, cells were co-exposed during 24 h to CDT and ATR inhibitor VE-821 (ATRi), given that the G2/M checkpoint mostly depends on ATR rather than ATM under these conditions (Fig. S3). Contrary to other genotoxic treatments, ATR inactivation does not sensitize HeLa cells to CDT during the rst 24 h of exposure (Fig. 3C), further supporting that cells do not activate checkpoint during early phase of CDT intoxication. Moreover, while exposure to control genotoxicants or high CDT concentration (25 ng/ml) block mitotic entry in an ATR-dependent manner, indicative of active G2 checkpoint, lowest CDT concentrations signi cantly increase the mitotic index, con rming that cells do progress through mitosis (Fig. 3D). This result demonstrates that ATR is only crucial after high treatment with CDT to protect cells from mitotic catastrophe by inducing a G2/M arrest, at least during the rst 24 h of exposure. Finally, in contrast to high CDT concentrations or DNA damaging agents, low CDT concentrations (0.025 and 0.25 ng/ml) induce MN formation that is not aggravated by the presence of ATRi (Fig. 3E). Altogether, these data demonstrate that except for high concentrations, CDT exposure allows mitotic entry and MN generation, despite the presence of active cell cycle checkpoints.

CDT induces mitotic delay and cell death
The increased mitotic index observed in CDT-exposed cells (Fig. 3D) is accompanied by a dose-dependent diminution of the anaphase population (Fig. 3F). In order to gain insight into the mitotic phenotype of CDT-treated cells, live-cell imaging has been performed on HeLa cells stably expressing the chromatibody fused to GFP, enabling real-time chromatin visualization [33]. When measuring the timing needed to complete metaphase, we found that unperturbed mitosis takes an average of 64 min that signi cantly increases to 109 and 164 min after treatment with 0.25 ng/ml and 2.5 ng/ml of CDT respectively (Fig.  3G). Moreover, monitoring cell death during the course of live imaging revealed that an important fraction of CDT-exposed cells preferentially dies at metaphase (Fig. 3H). In conclusion, mitotic cells are particularly affected during CDT intoxication, as evidenced by a prolonged metaphase duration that eventually result in cell death.

CDT-exposed cells experience DNA damage at mitosis
To better understand the relationship between CDT-mediated DNA damage and cell cycle defects, cell cycle analyses were conducted after immunostaining with antibodies directed against gH2AX and H3 histone phosphorylated at S10 (pH3) to identify mitotic cells. Cells treated with CDT for 24h present a dose-dependent augmentation of pH3 and gH2AX positive cells, representing a 12-fold increase at 2.5 ng/ml of CDT compared to control cells. In contrast, after exposure with moderate concentration of control genotoxic compounds, only cells without gH2AX staining do progress to mitosis (Fig. 4A). Thus, CDT-exposed cells progress through mitosis with damaged DNA, representing a unique feature over other genotoxic insult. Strikingly, asynchronous cells exposed to CDT display an intense gH2AX signal in mitosis compared to interphase (Fig. 4B). This staining is clearly distinguishable from the basal DNA damage-independent gH2AX signal described in unchallenged mitotic cells [34], that is diffuse all along the condensed chromosomes from prometaphase to anaphase, or from few gH2AX foci observed in mitotic cells exposed to other DNA damaging agents (Fig. S4). The huge gH2AX increase at mitosis is observed with CDT from other bacterial origins or with other cell lines (Fig. S5), thus representing a general cellular response to CDT. Moreover, the fraction of gH2AX-positive cells is more important in mitosis compared to interphase, after 24 h or even a shorter incubation of 8 h with CDT (Fig. 4C), demonstrating that mitotic cells represent the rst population to be damaged during the course of CDT treatment. The strong gH2AX signal after CDT can be observed all along the mitotic phases (Fig. 4D). Finally, CDT exposure induces a dose-dependent increase of chromosome fragments that does not properly align during metaphase or segregate at anaphase (Fig. 4E), therefore explaining the high level of MN observed after CDT treatment.

CDT induces DNA double-strand breaks during mitosis
To exclude the possibility that mitotic gH2AX signal originates from DNA damage induced before mitotic entry, HeLa cells were enriched in mitosis by a 22 h nocodazole block and then co-exposed during the last 6 h to CDT or genotoxic control agents. Similar to observations made on asynchronous cells, cells treated with CDT during mitosis exhibit a strong gH2AX level compared to etop, campto or MMC (Fig. 5A). To con rm that the gH2AX level increase in mitosis depends on DSB induction, cells arrested in mitosis were exposed to CDT before to be subjected to neutral comet assay (Fig. 5B). Mitotic cells treated with 0.25 or 2.5 ng/ml of CDT show a signi cant increase in comet tail moment (Fig. 5C). Taken together, these data demonstrate that CDT induces DSB during mitosis leading to chromosome fragmentation and missegregation.
CDT promotes proin ammatory response and mitotic defects in normal colonic epithelial cells We next assessed whether non-transformed cells exhibit similar in ammatory responses to CDT genotoxic activity. Immortalized normal human colonic epithelial cells (HCECs), previously shown to be susceptible to CDT intoxication [35], were exposed to CDT for 96 h and STAT1 phosphorylation at Y701 (pSTAT1) was measured as a surrogate for in ammatory pathway activation after DNA damage [28]. HCECs cells show a dose dependent increase of pSTAT1 (Fig. 6A), con rming that CDT activates in ammatory pathway in non-transformed cells. CDT also induces senescence in HCEC cells, as evidenced by β-Gal staining ( Fig. 6B and C). Importantly, the proportion of β-Gal positive cells decreases when adding CDK1 inhibitor RO-3306 (CDK1i), which blocks cells at G2/M boundary, supporting that CDTtreated cells must go through mitosis before entering a senescent state. Compared to control displaying only 4% of micronucleated cells, a 24 h exposure to CDT enables MN formation in approximately 16% of HCEC cells (Fig. 6D). This result indicates that normal cells do not stop at G2/M checkpoint during early course of CDT intoxication, but divide and form MN in daughter cells. Moreover, CDT-induced pSTAT1 increase is prevented by STING inhibitor H-151 (STINGi) (Fig. 6E), suggesting that similarly to HeLa cells, MN formed in HCEC cells in response to CDT activate the cGAS-STING axis to enhance proin ammatory responses. We further explored the mitotic phenotype of CDT-exposed HCEC cells and observed a decrease of the anaphase population in cells exposed to 0.25 ng/ml of CDT (Fig. 6F), as shown above (Fig. 3F). Moreover, mitotic HCEC cells displayed the same intense gH2AX signal as in cancer cell lines (Fig. 6G). Finally, the proportion of gH2AX positive cells after an 8 h treatment to 0.025 ng/ml of CDT is signi cantly higher in mitosis compared to interphase (Fig. 6H). Taken together, these data highlight the crucial role of cGAS-STING pathways in proin ammatory activation in normal HCEC cells, and demonstrate that this requires passage through mitosis where chromosomes accumulate DNA damage and form MN in daughter cells.

Discussion
The discovery of CDT more than 30 years ago relied on the distinctive phenotype of progressive cytoplasmic and nuclear enlargement this toxin caused in cultured mammalian cells [36]. This cell distention and other features of CDT exposure, namely proin ammatory cytokines expression or permanent DDR activation with cell cycle arrest, have then been associated to cellular senescence [26].
Our study now establishes a direct link between all these CDT-mediated cellular outcomes and the cytosolic DNA sensor cGAS recognizing MN that, at least in part, may arise from DNA damage induced during mitosis.
Mitosis coordinates the proper segregation of sister chromatids to ensure faithful transmission of equal genetic material to the next cell generation. On the other side, DSB repair is inhibited during mitosis, through the phosphorylation of several DDR factors by CDK1 and PLK1 kinases, noticeably to prevent telomere fusion [37]. Therefore, controlling chromosome integrity prior separation is performed at the previous G2 phase and implies cell cycle arrest and DNA repair to impede transition to mitosis with damaged DNA [38]. As duration of M phase is short and constant [39], only a very minor part of cells from an asynchronous population may suffer DNA damage at mitosis. The data presented here depict an atypical scenario with CDT-treated cells being able to reach mitosis despite continuous DNA damage induction. Contrary to ionizing radiations that instantaneously induce massive DNA damage, or to chemicals and metabolites from pathogens with limited stability or capacity to induce repeated lesions, the CDT catalytic subunit CdtB may exert unceasing genotoxic attacks, given that its nuclease activity remains active for at least 48 h after cellular internalization [25,32]. CDT is the only characterized encoded toxin from mammalian pathogens that has evolved to create DSBs in host genomic DNA [40]. This continuous nuclease activity is opposed to the host DNA repair machinery, implying that under a certain threshold, CDT-induced DNA damage do not effectively activate the DDR and therefore allows G2/M transition. In this context, previous studies used most of the time CDT concentration around 1 mg/ml and beyond, resulting in substantial cell cycle arrest at the G2 checkpoint. Similar effects have been observed here in HeLa cells with 25 ng/ml of CDT within the rst 24 h of exposure, whereas a 1000fold lower concentration does not strongly activate the DDR nor alter cell cycle, but still induces DNA damage, chromosome fragmentation and a delay at mitosis followed by increased MN formation and eventually viability loss. Currently, the physiological concentration of CDT to which cells are exposed during a natural infection is unknown, but one may expect that the cellular effects caused by moderate concentrations of toxin used in this study should be at least as relevant than analyzing highest doses.
We demonstrate here that once entered in mitosis, CDT-treated cells experience massive DNA damage, as evidenced by the intense mitotic gH2AX staining observed in several human cell lines, with toxins from various bacterial origins. This indicates that in icting DNA lesion during mitosis is a general mode of action of CDT in proliferative cells. The gH2AX signal in mitosis is correlated to a dose-dependent increase of mitotic cells with chromosome fragments, as revealed by microscopic observation or neutral comet assay. These DNA damage impose a dose-dependent mitotic delay at metaphase, similar to cells depleted for the mitotic resolvase GEN1 [41,42]. More generally, mitotic duration is governed by the capacity of cells to dissolve or resolve Hollyday junctions [43], questioning on CdtB abilities to target homologous recombination intermediates that are processed later in mitosis.
We speculate that the progressive accumulation of DNA damage, during at least 48 h, in interphasic CDTexposed cells [31] may partly be a consequence of unrepaired DNA lesions transmitted from the previous cell division. In comparison, control genotoxic compounds induce only few mitotic DSBs at selected concentrations that only partially block cells in G2, in an ATR-dependent manner. Etoposide at 10 mM has been shown to induce important DNA damage on mitotic HeLa cells, however synchronization in mitosis is required to ensure that an important proportion of cells is affected [44]. Alternatively, HT-29 cells arrested in G2 after a 25 nM treatment of camptothecin, a concentration that is comparable to the present study, nally reach mitosis 48 h post-treatment with highly damaged chromatin [45]. Such recovery from G2 arrest in the presence of DNA damage, referred to as checkpoint adaptation, has been observed with many DNA damaging agents and give rise to important MN formation in post-mitotic cells [46]. In the same way, a substantial increase of micronucleated cells treated with etop, campto or MMC was also observed here at 72 h (data not shown). Therefore, CDT exposure globally recapitulates the phenotype of cells subjected to checkpoint adaptation, but with more rapid kinetics due to a weak ATR-dependent G2 checkpoint activation.
The cGAS-STING axis activates IRF3 and NF-κB, two transcriptional inducers of type I IFNs and other cytokines [30]. cGAS has also been shown to play a crucial role during cellular senescence induced by DNA damage [47]. Therefore, our ndings that HeLa cells chronically exposed to CDT display high MN frequency and type I IFN signature prompted us to examine the role of cGAS in response to CDT genotoxin. Our data reveal that cGAS -/cells are resistant to low CDT concentrations and cannot induce the expression of type I IFN signaling target genes. Moreover, cGAS de ciency prevents CDT-mediated cell distention and senescence. These ndings support that cGAS impedes the proliferation of cells experiencing genetic instability after CDT exposure by inducing senescence. However, we cannot exclude that in CDT-exposed cells, cGAS inhibits DSB repair and/or promotes mitotic cell death upon retardation of metaphase-anaphase transition [48,49], therefore participating in the sensitivity to CDT. Finally, we bring evidence that non-transformed cells display similar behavior in response to CDT-mediated MN formation, with mitosis-dependent senescence induction and STING-dependent proin ammatory pathway activation. Thus, our data indicate that normal proliferative cells may in uence the modulation of the immune environment in tissues infected by CDT-producing bacteria.
Persistent infection causes chronic in ammation, a driving force of tumor development [50]. CDT production during bacterial infection promotes in ammation and persistent colonization, and may eventually contribute to tumorigenesis [10][11][12]51]. In a broad range of cellular lineages, CDT intoxination induces DNA damage and a proin ammatory signature [13], but how these two events are linked was never addressed so far. Our ndings identify cGAS as the missing piece between CDT genotoxic and immunomodulatory activities. cGAS is a key modulator of the innate immune response that can in uence the tumor microenvironment in ways that may be detrimental or bene cial [52]. Further investigations will be necessary to unravel the role of the cGAS-dependent surveillance of MN during natural infection with CDT producing bacteria.
E.col CDT (WT or the catalytic dead mutant H153A) were produced and puri ed as previously described . ATM inhibitor KU-55933 (ATMi), ATR inhibitor VE-821 (ATRi), CDK1 inhibitor RO-3306 (CDK1i) and STING inhibitor H-151 (STINGi) were purchased from Selleckchem. ATMi and ATRi were used at a nal concentration of 10 mM, CDKi at 9 mM and STINGi at 0.5 mM. All other chemicals and reagents were purchased from Sigma-Aldrich.

Western blot analyses
Cells were incubated on ice for 30 min in lysis buffer (50 mM Tris-HCl pH7.5, 500 mM NaCl and 0.5% NP40) containing the HaltTM Protease & Phosphatase inhibitor cocktail (Thermo Scienti c) and sonicated on a VibraCell 72434 (Bioblock Scienti c). Cell lysates were centrifugated and the supernatant containing total soluble proteins was kept. Proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane (Amersham). Membranes were incubated with the primary antibody over-night at 4°C. γH2AX antibody was purchased from Merck/Millipore (05-636), pChk1 (133D3), pChk2 (C13C1), pH3 (D2C8), cGAS (D1D3G) and pSTAT1 (58D6) antibodies from Cell Signaling and GAPDH (GTX100118) antibody from GeneTex. The secondary anti-mouse or anti-rabbit HRP-conjugated antibodies (Jackson Immunoresearch laboratories) were incubated for 1 h at room temperature. Proteins were visualized with the enhanced chemiluminescence substrate ECL (Biorad) and imaged using the ChemiDoc XRS Biorad Imager and Image Lab Software.
Immuno uorescence analyses and micronucleus assay Cells were grown on glass coverslips. After at least 24 h of culture, cells were xed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, blocked with 3% BSA and 0.05% IGEPAL, and stained with primary antibodies for 2 h at room temperature in blocking solution (all solutions were prepared in PBS). Cells were washed three times with PBS 0.05% IGEPAL and incubated with the secondary antibodies for 1 h (Alexa Fluor 488 Goat anti-mouse (A32723) or Alexa Fluor 594 anti-rabbit (A32740), purchased from Invitrogen). DNA was stained with 4.6-diamino-2-phenyl indole (DAPI) 30 nM.
Coverslips were mounted onto slides with PBS-glycerol (90%) containing 1 mg/ml paraphenylenediamine and observed at 40x magni cation with a Nikon 50i uorescence microscope equipped with a Luca S camera. Cells were counted positive for foci formation when > 10 foci/interphase nuclei or when > 5 foci/mitotic nuclei were detected.

Cell viability assays
For clonogenic assay, cells were plated in triplicate at a density of 300 to 3000 cells per well in 6 wells plate. One day after seeding, cells were treated and grown for 10 days. Formed colonies were xed and stained with a 0.25% methylene blue and 100% methanol solution. Colonies containing more than 50 cells were counted and the surviving rate calculated.
For rapid viability testing at 24 h, cells were plated in triplicate at a density of 10000 per well in 96 wells plate. One day after seeding, cells were treated and grown for 24 h. Viability was assessed using the CellTiter-Glo ® Luminescent Cell Viability Assay (Promega) according to the manufacturer's instructions.

Senescence assay
The senescence-associated b-galactosidase staining was performed using the Senescence b-Galactosidase Staining Kit (Cell Signaling) according to the manufacturer's instructions. Images were captured by light microscopy.

Flow cytometry analyses
Cells were xed with 4% paraformaldehyde for 15 min at room temperature in PBS. For immunostaining, xed cells were permeabilized with 0.2% triton X-100 and 1% BSA in PBS for 20 min at room temperature.
Antibodies against γH2AX and pH3 were incubated for 2 h in 1% BSA at room temperature. After three washes, cells were incubated with the secondary antibodies Alexa Fluor 488 anti-mouse and Alexa Fluor 594 anti-rabbit for 30 min at room temperature. Then, cells were washed three times and incubated in PBS containing DAPI (1 μg/mL) for 15 min before samples were processed using ow cytometry (MACSQuant, Miltenyi Biotec). At least 10000 events were analyzed per sample using FlowLogic software (Miltenyi Biotec).
Comet assay on mitotic cells Cells were treated for 18 h with 2 mM thymidine, washed twice with PBS and released in fresh media for 9 h. Cells were treated again with 2 mM thymidine for 17 h to be synchronized in early S, washed twice with PBS and released in fresh media for 6 h before adding 50 ng/ml nocodazole. After 4 h, mitotic cells were collected by shake-off and exposed for 12 h to CDT or the genotoxic controls in presence of nocodazole. Comet assay was performed in neutral conditions using the Comet SCGE assay kit (Enzo) according to the manufacturer's instructions. At least 60 cells were analyzed per sample using OpenComet software.

Time-lapse imaging
HeLa cells stably expressing the chromatibody-GFP to visualize chromatin in living cells  were generated by TALEN insertion at AAVS1 site with co-transfection of SHDP-CMV-VHH-HA-GFP donor plasmid and AAVS1 right and left Talen plasmids (genome TALER AAVS1 safe harbor cloning kit, Genecopoeia) using TransIT-LT1 (MirusBio) according to manufacturer's instructions. Cells were plated in 4-compartments Petri dish (CellView Greiner Bio-One). Immediately after treatment, time-lapse uorescent microscopy was performed with a Zeiss Axio-observer inverted videomicroscope with controlled temperature (37 °C), CO2 level (5 %) and humidity. Images were acquired for 90 h every 9 minutes with an AxioCam 506 camera (Zeiss) in two focal planes 5µM apart from each other. Illumination has been set to 75 ms with 20 % laser intensity to reduce phototoxicity. The two focal planes were attened and 100 to 200 mitosis per sample were analyzed with FIJI software from 4 to 90 h post-exposure to measure the duration of mitotic phases and cell death.

Generation of cGAS knock-out HeLa cells
The CRISPR plasmid was a generous gift from J. Dupuy (Toxalim, France). The donor plasmid (pCMV-Cas9-GFP, Sigma), containing the sgRNA, Cas9 and GFP, was modi ed to substitute the sgRNA sequence for cGAS sgRNA 1 (5'-GCTTCCGCACGGAATGCCAGG) or cGAS sgRNA 2 (5'-CGATGGATCCCACCGAGTCT) targeting the rst exon of the cGAS gene, using the Q5 site-directed mutagenesis kit (New England Biolabs) according to manufacturer's instructions. The two plasmids bearing cGAS sgRNA1 and 2 were co-transfected in HeLa cells using TransIT-LT1 (MirusBio) according to manufacturer's instructions. Individual clones were selected by cGAS immuno uorescence and con rmed by Western blot analyses. Two con rmed cGAS -/clones were used in this study.

Gene expression
Total RNA was extracted with TRIzol reagent (Invitrogen, Carlsbad, CA, USA) mRNA were extracted and the quality of these samples was assessed (Agilent RNA 6000 Nano Kit Quick, Agilent Bioanalyzer 2100); RNA Integrity Number (RIN) of these mRNA was superior to 9.8.
For transcriptomic data, gene expression pro les were obtained at the GeT-TRiX facility (GénoToul, Génopole Toulouse Midi-Pyrénées, France) using Agilent SurePrint G3 Human GE v2 microarrays (8x60K, design 039494) following the manufacturer's instructions. For each sample, Cyanine-3 (Cy3) labeled cRNA was prepared from 200 ng of total RNA using the One-Color Quick Amp Labeling kit (Agilent Technologies) according to the manufacturer's instructions, followed by Agencourt RNAClean XP (Agencourt Bioscience Corporation, Beverly, Massachusetts). Dye incorporation and cRNA yield were checked using Dropsense™ 96 UV/VIS droplet reader (Trinean, Belgium). 600 ng of Cy3-labelled cRNA were hybridized on the microarray slides following the manufacturer's instructions. Immediately after washing, the slides were scanned on Agilent G2505C Microarray Scanner using Agilent Scan Control A.8.5.1 software and uorescence signal extracted using Agilent Feature Extraction software v10. 10

Statistical analysis
Microarray data were analyzed using R and Bioconductor packages [53]. Raw data (median signal intensity) were ltered, log2 transformed, and normalized using quantile method [54]. A model was tted using the limma lmFit function [55]. Pair-wise comparisons between biological conditions were applied using speci c contrasts. A correction for multiple testing was applied using Benjamini-Hochberg procedure to control the False Discovery Rate (FDR). Probes with FDR≤0.05 were considered to be differentially expressed between conditions. Hierarchical clustering was applied to the samples and the differentially expressed probes using 1-Pearson correlation coe cient as distance and Ward's criterion for agglomeration.