STING Promotes Homeostasis via Regulation of Cell Proliferation and Chromosomal Stability

Stimulator of IFN genes (STING, also known as TMEM173, MITA, ERIS, and MPYS) is a cytoplasmic pattern recognition receptor shown to be an important transducer of type I interferon (IFN) production triggered by the accumulation of double stranded DNA (dsDNA) in the cytoplasm. STING is localized in the endoplasmic reticulum (ER) and following activation dimerizes, undergoes conformational changes, and translocates to the ER–Golgi intermediate compartments (1–3). STING induces phosphorylation of TANK-binding kinase 1 (TBK1), leading to a signaling cascade involving IRF3 and NFκB and the production of type I IFNs (1, 2, 4, 5). STING directly mediates innate immune responses to cyclic dinucleotides produced by intracellular bacteria and viruses (2, 6) as well as the endogenous cyclic dinucleotide cGAMP (c[G(2′-5′)pA(3′-5′)p]) (7, 8). cGAMP is a second messenger synthesized by the enzyme cGAS, which serves as the primary cytoplasmic dsDNA sensor in metazoans (8–10).

Ionizing radiation (IR) is associated with dsDNA breaks, affecting cellular proliferation and survival (11). The acute DNA damage and changes in growth factors and/or chemokine levels activate intracellular stress sensors that signal cell-cycle progression or inhibition depending on the balance of damage and growth factors (12). Defects in subsequent DNA repair can result in accumulation of chromosomal aberrations (13). Recently, it has been demonstrated that genotoxic stress–induced DNA damage leads to the formation of micronuclei and/or chromosomal fragments that are recognized by cGAS following loss of nuclear compartmentalization, triggering cellular senescence through STING-mediated production of type I IFN and other proinflammatory cytokines (14–18).

STING serves a key role in initiation of immune responses against tumors (19). Several studies have shown that recruitment of tumor-infiltrating host immune cells via STING activation mediates IR-triggered antitumor immune responses (19–21). It has been recently demonstrated that innate immune sensing by dendritic cells following radiotherapy is dominated by the cGAS and STING-dependent cytosolic DNA-sensing pathway, which drives the adaptive immune response to ionizing radiation (21).

Interestingly, cancer cells often display inactivated STING (22, 23). For example, screening 17 different histologic types of tumors from approximately 8,000 patients, Uhlén and colleagues (24) observed significant heterogeneity of STING expression. It remains unclear what selective advantage cancer cells might gain by downregulating STING pathway function. Here we demonstrate that STING also serves homeostatic roles in cell growth and proliferation analogous to those of a tumor suppressor, serving as a brake on cell division and maintaining genomic stability. STING constrains cell proliferation and entry to mitosis via NFκB- and p53-driven activation of the stoichiometric cyclin-dependent kinase inhibitor CDKN1A (p21, WAF1, CIP1). Loss of STING promotes premature entry to S-phase and mitosis, resulting in smaller cell size and increased chromosomal instability (CIN). Radiation potentiates the chromosomal instability. Our findings suggest a function for STING during normal cell division in protecting cells against aneuploidy, serving as a barrier to tumorigenesis.

All animal studies had been approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Chicago (protocol number 72213). Mice were maintained under specific pathogen-free conditions in a barrier facility. Age-matched 6–8 weeks old female C57BL/6 or athymic nude mice were purchased from Harlan. STINGko mice were acquired from Dr. Glen N. Barber (University of Miami, Miami FL).

Wild-type and STINGko primary mouse embryonic fibroblasts (MEF) were derived from 13.5 days postcoitus embryos and cultured in DMEM supplemented with 10% FBS and 1% nonessential amino acids for no more than five passages (25). MEFs were immortalized with a retrovirus expressing SV40-large T antigen (Addgene plasmid 13970). p50^−/− primary MEFs were kindly provided by Dr. Giovanna Bernal (University of Chicago, Chicago, IL). Normal human dermal adult fibroblast primary cell line (HDFA), Human embryonic kidney 293 (HEK293, accession ID: CVCL_0045), human colon adenocarcinoma HCT116 (ID: CVCL_0291), and lung adenocarcinoma A549 (ID: CVCL_0023) were purchased from ATCC. Human glioblastoma D54 (ID: CVCL_7185) was obtained from Dr. Daryll D. Bigner (Duke University Medical Center, Durham, NC), while the murine colon adenocarcinoma MC-38 (ID: CVCL_B288) derived from C57BL/6 mice was obtained from Dr. Xuanming Yang (University of Chicago, Chicago, IL). SCC61 (ID: CVCL_7118) was isolated from human head and neck squamous cell carcinoma (PMID 3458227). Cell lines were submitted to IDEXX BioAnalytics for authentication using short tandem repeat profiling. The cells were cultivated as follows: SCC61 in DMEM/F12 with 20% FBS and 1% HC (hydrocortisone); HEK293, HCT116, and MC-38 cell lines in DMEM-high glucose with 10% FBS; D54 cell line in MEM with 10% FBS, A549 cell line in DMEM/F12 with 10% FBS, and HDFA in fibroblast growth kit with low serum (2%), 5 ng/mL EGF, 50 μg/mL l-glutamine, 1 μg/mL hydrocortisone, and 5 μg/mL insulin. All cell lines used were routinely screened for Mycoplasma contamination using the Lonza MycoAlert Mycoplasma Detection Kit (CN LT07-418). Only cells from early-passage frozen stocks were used for the experiments.

Human tumor xenografts (1 × 10⁶ cells) were established in the right flank of athymic nude mice, while 1 × 10⁶ MC-38 tumor cells were subcutaneously injected at the back of C57BL/6 mice. For each experiment, up to 5 animals were housed per cage and assigned an ear tag number. Each animal (i.e., technical replicates) in the cage were administered the same treatment. Tumor volumes were measured along three orthogonal axes and calculated using the formula for an hemiellipsoid [(l × w × h)/2]. Tumor growth was measured, recorded, and monitored by a single investigator every 3–4 days for the entire duration of the experiment to ensure consistency of measurement. Tumor growth was quantitatively calculated by extrapolating the linear equation of the growth curve from each measurement to compute for the time for each tumor to reach an arbitrarily assigned volume (300 mm³). Mice were sacrificed when tumor sizes reached a volume of 2,000–3,000 mm³. Cell lines maintained in tissue culture were irradiated using a Gammacell 220 (MDS Nordion) ¹³⁷Cs γ-irradiator.

Cells were seeded at 10,000 cells/well in 24-well plates. At several time points (24, 48, 72, 96, and 120 hours) postseeding, cells were counted using a TC20 automated cell counter (Bio-Rad). Data were plotted as the logarithm of the number of viable cells over time. The growth rate was calculated by extrapolating the slope of the line from the exponential portion of the semilog growth curves. Alternatively, we calculated the doubling time of cells by fitting the linear data to the equation of exponential growth: N(t) = N(0) e⁁(kt) with k as the growth rate constant and N(t) and N(0) the number of cells at time t and time 0, respectively, using the exponential growth analysis function in GraphPad Prism software. Doubling time was calculated from the formula ln(2)/k.

Cells were seeded in 24-well plates at a density of 25,000 cells per well in appropriate growth media and incubated overnight to allow monolayer formation. Thirty minutes prior to IR treatment, media were replaced with fresh growth media containing the nuclear stain Incucyte NucLight Rapid Red Reagent (Essen BioScience) for live cell imaging at 1:3,000 dilution. After irradiation, the plates were placed into either the IncuCyte ZOOM or IncuCyte S3 apparatus and images of the collective cell spreading were recorded every 4 hours for a total duration of 48–72 hours. For each cell line tested and treatment conditions, we performed three identically prepared experimental replicates (n = 3), and experiments were repeated 3–4 times. Basic analyses were performed using the IncuCyte software to plot phase confluence, calculate the number of nuclei-stained cells, and measure the average nuclei area over time. In some experiments, the WEE1 kinase inhibitor MK1775 (Axon Medchem) was added to fresh media at a final concentration of 250 nmol/L together with the NucLight Rapid Red Reagent prior to IR treatment.

Tumor cell lines were transfected with shSTING construct within a TRC2-pLKO-puro vector backbone (Sigma-Aldrich mission shRNA) using Fugene HD transfection reagent at 1:3 plasmid DNA:lipid ratio. Five different shRNA constructs were tested for each human cell line (TRCN0000164628, TRCN0000160895, TRCN0000163296, TRCN161052, and TRCN0000163029), while three shRNA constructs were tested for murine cell line MC-38 (TRCN0000346321, TRCN0000346319, and TRCN0000346264). The TRC2 pLKO.5-puro nonmammalian targeting shRNA (TRCN SHC002 for human cell lines and TRCN SHC202 for murine cells; Sigma-Aldrich) was used as a control. Stable lines from the top two shSTING constructs were selected by growth in culture media containing 5 μg/mL puromycin over multiple passages. Successful knockdown of STING was confirmed by Western blot analysis (Supplementary Fig. S1A–S1C). Stable cell lines from mixed pools following puromycin selection were further assessed for IFNβ production, caspase-3/7 activity, and clonogenic survival as described in ref. 25. For murine tumor models and cell growth studies, we selected the stable cell line from the shSTING construct that yielded the best knockdown for each cell line. The specific product numbers used for each cell line are summarized below:

• D54 shSTING: TRCN0000161052

• HCT116 shSTING: TRCN0000163029

• SCC61 shSTING: V3LHS324286

• A549 shSTING: TRCN0000163029

• MC-38 shSTING: TRCN0000346319

To measure the percent population of actively dividing cells as well as the total DNA content of cells, a dual staining protocol using a fluorescently conjugated anti-EdU/BrdU antibody to measure EdU/BrdU incorporation together with propidium iodide (PI) were performed on tumor cells with or without STING knockdown at different time points following IR treatment. Cells were seeded overnight in 6-well plates at a density of 1 × 10⁵ cells/well. Cells were either mock-irradiated or exposed to 6 Gy IR, and were harvested at different time points posttreatment.

For pulse/chase experiments, the Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Invitrogen) was used together with PI. EdU was added to cells in culture 1 hour prior to harvest time (for pulse experiments) or at a fixed time point (for chase experiments) at a final concentration of 0.03 μmol/L. Cells were harvested at different time points postlabeling and processed for immunostaining following manufacturer's protocol. In the final step, cells were resuspended in PBS containing 25 μg/mL PI (eBioscience) and 100 μg/mL RNase A (Thermo Fisher Scientific). After overnight incubation at 4°C, flow cytometry was performed to measure the number of EdU⁺ cells in the Alexa Fluor 647 channel, while the number of PI-positive cells were detected in the PE channel. Analyses were gated on single cells using the FlowJo software.

For regular cell-cycle experiments, the BrdU/PI staining protocol was performed. Four hours prior to harvest time, the growth media were replaced with BrdU (Invitrogen)-containing fresh media diluted at 1:100. Cells were trypsinized and carefully permeabilized with 70% ethanol for at least 30 minutes on ice. After washing with PBS, cells were incubated in 2 mol/L HCl solution for 30 minutes at room temperature. Cells were washed in PBS, followed by two washes using the blocking buffer (PBS + 1% BSA + 1% Tween-20). Cells were resuspended in blocking buffer and incubated with FITC-conjugated anti-BrdU mAb (eBioscience) overnight at 4°C. The next day, cells were washed 2× with blocking buffer and resuspended in PBS containing 25 μg/mL PI (eBioscience) and 100 μg/mL RNase A (Thermo Fisher Scientific). After 1-hour incubation at 4°C, flow cytometry was performed to measure the number of FITC-conjugated anti-BrdU antibody that bound to actively dividing cells (S phase) in one channel, while the number of PI-positive cells will be used to quantify the percent population of cells in G₁ and G₂–M phase of the cell cycle in another channel. Analysis was gated on single cells using the FlowJo software.

WT and STINGko primary MEFs were seeded in 6-well plates at a density of 2 × 10⁵ cells/well. Approximately 15 hours postseeding, cells were either mock-irradiated (unirradiated) or exposed to 6-Gy IR. Total RNA was isolated using TRIzol reagent following the manufacturer's protocol. RNA yield was measured using Qubit RNA broad range kit. One-hundred nanograms of RNA was labeled as per the manufacturer's instructions and profiled in duplicate using the Illumina Mice WG-6 array (Illumina). Background subtraction and quantile normalization was performed across arrays using Illumina Beadstudio software. Analysis was performed as described in ref. 25.

HCT116 and MEFs cell lines were transiently transfected with specific siRNA constructs to CDKN1A (Dharmacon). Nontargeting siRNA was used as a control (Dharmacon D-001210-02-05). Twenty-four hours posttransfection, cells were either mock irradiated or exposed to increasing doses of IR. Cells were then assayed for Western blot and cell-cycle analyses using dual BrdU/PI staining and flow cytometry. The two siRNA constructs that successfully knocked down CDKN1A in HCT116 and MEFs cells, respectively, are:

• siRNA to human CDKN1A: GAUGGAACUUCGACUUUGU (Dharmacon D-003471-01)

• siRNA to murine CDKN1A: CGAGAACGGUGGAACUUUG (Dharmacon D-058636-01)

In some experiments with wild-type MEFs, STING was transiently knocked down, either alone or in combination with p21, using the siGENOME mouse TMEM173 siRNA with sequence CGAAAUAACUGCCGCCUCA (Dharmacon M-024333-00).

Wild-type and STINGko MEFs were plated overnight in two-well chambered slides (at 2 × 10⁴ cells/chamber). Cells were either mock-irradiated or exposed to either 6 Gy or 12 Gy IR. At different time points post-IR treatment (24 or 48 hours), MitoTracker Red (Invitrogen) was added to cells and incubated at 37°C for 1 hour. Cells were washed with PBS and fixed in 3.7% paraformaldehyde for 10 minutes and washed 2× with PBS. Cells were then permeabilized with 70% ethanol for 1 hour at 4°C followed by 2× PBS wash. Cell nuclei were stained with Draq5 solution (Abcam) diluted in PBS to a final concentration of 2 μmol/L, and incubated at room temperature for 30 minutes. After final washing steps (2× PBS), cells were mounted in Prolong Gold (Thermo Fisher Scientific). Cells were visualized using Leica SP5 2-Photon Confocal microscope at 63× magnification. Fifteen images were collected per variable tested. Approximately, n = 150 cells were analyzed per condition.

For the cytokinesis-block micronucleus (CBMN) assay (26), Cytochalasin B (Sigma) dissolved in DMSO was added to each chamber well at a final concentration of 1.5 μg/mL, 24 hours post-IR treatment. After an additional 24-hour incubation at 37°C, MitoTracker Red (Invitrogen) was added to cells and incubated at 37°C for an additional hour. Cells were then processed and imaged as described above. At least n = 50 binucleated cells were analyzed for the CBMN assay.

Primary wild-type and STINGko MEFs at passage 1 were seeded in T-25 flasks at approximately 7.5 × 10⁵ cells in DMEM + 10% FBS media + 1% nonessential amino acids. Fifteen hours postseeding, cells were either mock-irradiated or exposed to 6 Gy IR. Live cells were shipped to Cell Line Genetics and were fixed for karyotype analysis at approximately 48 hours post-IR treatment.

Stable HCT116 cell lines expressing shRNAs targeting STING (shSTING) or scrambled control (shScrambled) were transiently transfected with siRNA targeting CDKN1A or scrambled control. Twenty-four hours posttransfection, selected cells were irradiated at 9 Gy and incubated for an additional 24 hours. Chromatin was then extracted from 2 × 10⁶ cells for chromatin immunoprecipitation (ChIP) using the Chromatin Immunoprecipitation Assay Kit (Sigma Millipore, CN 17-295). Briefly, following chromatin crosslinking with 1% formaldehyde for 10 minutes at 37°C and sonication, the anti-E2F4 antibody (Santa Cruz Biotechnology, sc-398543) was used for immunoprecipitation, and chromatin washes were performed as per the kit. Chromatin was eluted in 100 μL of elution buffer (1% SDS, 100 mmol/L NaHCO₃). Cross-linking was reversed by incubating the samples overnight at 65°C with 200 mmol/L NaCl and 200 mg/mL Proteinase K. DNA was purified with Monarch PCR and DNA Cleanup Kit (New England Biolabs). Quantitative PCR using iTaq Universal SYBR Green Supermix (Bio-Rad) was utilized to determine E2F4 binding at the BUB1, MAD2L1, and GAPDHS promoters using the following primer sequences:

• BUB1 Forward primer: 5′-GAG GGA GGT GGG ACT TGA C-3′

• BUB1 Reverse primer: 5′-CAA ACC TGA ACC GCA AAC TA-3′

• MAD2L1 Forward primer: 5′-CCA CAG CTT TAC AGG GTT CG-3′

• MAD2L1 Reverse primer: 5′-ACC TTA TTC CTG TCC TGC CC-3′

• GAPDH Forward primer: 5′-AGA CCA GCC TGA GCA AAA GA-3′

• GAPDH Reverse primer: 5′-CTA GGC TGG AGT GCA GTG GT-3′

HEK293 or HCT116 (WT or p53^−/−) cells were seeded in a 24-well plate overnight at a density of 7.5 × 10⁴ cells/well. Cells were cotransfected with either an NFκB promoter-driven luciferase reporter or a p65 binding site-mutant CDKN1A promoter–driven reporter construct (27) together with a Renilla luciferase gene driven by a basal promoter (pRL-null) as a transfection control. Transfections were performed using a cationic lipid agent, Fugene-HD (Promega), at a 3:1 lipid:DNA ratio. Twenty-four hours posttransfection, cells were either mock-irradiated or exposed to 6-Gy IR. Forty-eight hours post-IR treatment, cell lysates were collected and 20 μL were transferred to opaque 96-well plates. Following the manufacturer's protocol for the dual luciferase assay (Promega), samples were analyzed for IFNβ luciferase and Renilla luciferase activity using Promega Glomax reader. The transfection efficiency across different wells was normalized by dividing the firefly luciferase activity by the Renilla luciferase control. After correcting for transfection efficiency, all values were normalized to those of nonirradiated cells transfected with the empty pEF-BOS vector.

For confirmation of targeted knockdown experiments as well as transient transfection/reconstitution experiments in both murine and human cell lines, the following primary antibodies were used: anti-STING (clone D2P2F; #13647; Cell Signaling Technology), anti-cGAS (#15102 and #31659; Cell Signaling Technology), anti-p21 (ab109199; Abcam), anti-TBK1 (sc-9910; Santa Cruz Biotechnology), anti-phospho-TBK1 Ser172 (clone D52C2; #5483S; Cell Signaling Technology), anti-IRF3 (clone FL-425; sc-9082; Santa Cruz Biotechnology), anti-phospho IRF3 Ser396 (70R-35220; Fitzgerald Antibodies), anti-STAT1 p84/p91 (clone C-136; sc-464; Santa Cruz Biotechnology), anti- NFκB p65 (clone D14E12; #8242; Cell Signaling Technology), anti-phospho-NF-κB p65 Ser536 (clone 93H1; #3033; Cell Signaling Technology), anti-FLAG (M2 clone; Sigma), anti-CDC2 (#77055S; Cell Signaling Technology), anti-phospho-CDC2 Tyr15 (clone 10A11; #4539; Cell Signaling Technology), anti-Rb (#9313S; Cell Signaling Technology), anti-phospho-Rb Ser807/811 (clone D20B12; #8516; Cell Signaling Technology), and anti-actin-HRP (sc-47778; Santa Cruz Biotechnology). Secondary antibodies conjugated to HRP (Cell Signaling Technology) were used at a 1:5,000 dilution.

All protein bands in Western blot gels were quantified using ImageJ. The pixel units obtained for each band was normalized to the pixel units calculated from their respective β-actin loading control. For phosphorylated proteins, we then quantified the fraction of phosphorylated proteins to the absolute total amount of protein (total protein + phosphorylated protein). For dose-dependent and stimulation assays, the variables tested were normalized to their respective wild-type unstimulated controls.

For animal studies, figures are representative of ≥ 3 biological replicates and each of these experiments contains 5–6 technical replicates. For in vitro assays, data presented are representative of ≥ 3 independent experiments and each of these experiments contains three technical replicates. Data are plotted as SEM of these replicates, with P values calculated using a two‐tailed Student t test. The two‐tailed Student t test is a standard statistical test for measuring the significance of the results from these assays.

All microarray data were submitted to Gene Expression Omnibus (GEO) with accession number GSE113123.

We screened multiple human cancer cell lines as well as normal cells to measure their basal level of STING expression and to examine the role of STING in tumor cells. As shown in Fig. 1A (top), STING is expressed in normal human adult primary dermal fibroblasts (HDFA), HEK293 fibroblasts, glioblastoma cell line D54, the human colorectal carcinoma HCT116, and the murine colon carcinoma MC-38, but was not detected in human A549 lung carcinoma and murine 4T1 mammary epithelial cells. The amount of STING expression from each human and murine tumor cell lines was normalized to primary cell line controls (HDFA for human and MEFs for murine cell lines, respectively). Our data suggest that for human tumor cell lines, HCT116 and SCC61 have lower protein levels of STING compared with the normal HDFA cell lines. D54, on the other hand, exhibited higher STING protein level compared with control (Fig. 1A, bottom). For murine cell lines, large T-antigen immortalization of MEFs led to an increase in STING expression while MC-38 exhibited a slightly lower STING protein level compared with primary MEFs control (Fig. 1A, bottom). To investigate STING function in human tumor cells, we developed stable cell lines expressing shRNAs targeting STING (shSTING) or scrambled control (shScrambled) in D54, HCT116, SCC61, and MC-38, with respectively 54.2%, 82.4%, 99.98%, and 75.63% downregulation of protein expression compared with nontargeting control. (Fig. 1B). Flank tumors formed by the shSTING human tumor cell lines in athymic nude mice or shSTING MC-38 in syngeneic C57BL/6 mice each displayed significantly more rapid tumor growth than controls (Fig. 1C–F; Table 1; P ≤ 0.05). As expected, there was no observed significant difference between A549 shScrambled and shSTING cell lines (Fig. 1G; Table 1; P = 0.659).

To assess the role of STING in cell proliferation, we measured the growth curve of normal and tumor cells in vitro via manual cell counting at different time points postseeding. Growth rate was calculated by extrapolating the slope of the line from the exponential portion of the semi-log growth curves (Table 2). Cell proliferation of shSTING D54, HCT116, and SCC61 human tumor cells as well as MC-38 murine tumor cells was significantly faster and have higher calculated slope, μ, than shScrambled controls (Fig. 1H–K; Table 2; P ≤ 0.05). Similarly, primary and immortalized mouse embryonic fibroblasts (MEF) isolated from STINGko mice exhibited accelerated growth compared with WT control (Fig. 1L and M), suggesting the effects are not limited to transformed cells. Overall, the cell growth data indicate STING depletion confers a shorter cell doubling time compared with controls (Table 2), and as expected, no difference was observed between A549 shScrambled and shSTING cell lines (Fig. 1N; Table 2; P = 0.0.577). These results confirm a previously uncharacterized role of STING in cell proliferation.

To examine the molecular mechanisms underlying deregulated growth in the absence of STING, we asked whether the absence of STING confers changes in cell-cycle distribution. MEFs were treated with EdU (5-ethynyl-2′-deooxyuridine) to identify S-phase cells, and fixed at the indicated time points. The EdU was labeled with azido-Alexa Fluor 647 by copper(I)-catalyzed azide-alkyne cycloaddition Click chemistry and costained with propidium iodide (PI), and cells were examined by flow cytometry to distinguish G₁ (2N), S, and G₂–M (4N) cells by DNA content. Cells were exposed to a single dose of ionizing radiation (6 Gy) to induce genotoxic stress–mediated cell-cycle arrest. Using the gating strategy shown in Fig. 2A for data analysis, we demonstrate that, strikingly, STINGko MEFs displayed almost 2-fold higher proportion of EdU⁺ S-phase cells versus WT controls at all time-points, independent of ionizing radiation (Fig. 2B; Supplementary Fig. S2A). The greater S-phase content was associated with fewer interphase cells overall, but the distributions in G₁ and G₂–M displayed a reciprocal pattern. STINGko MEFs displayed a decreased fraction with G₂–M DNA content (Fig. 2B; P ≤ 0.05). STINGko and WT MEFs displayed similar accumulation of cells in G₂–M delay after irradiation. The fractions of unirradiated STINGko and WT MEFs in G₁ were more similar across the time course, but with a slightly increased G₁ content in STINGko MEFs (Fig. 2B). Both STINGko and WT MEFs displayed similar depletion of G₁ content after irradiation.

To distinguish cells that were already in S-phase at the time of irradiation from those entering S-phase afterwards, we “chased” the EdU and measured the percent of EdU-labeled MEFs at different time points (Fig. 2C). Analyzing these cells as above and applying the gating strategy in Fig. 2D, we were able to detect a distinct population of EdU-positive cells that accumulate in the 2N G₁ DNA content peak at later time points, corresponding to cells that completed mitosis during the time course (Supplementary Fig. S2B; ref. 28). As shown in Fig. 2E, the fraction of STINGko MEFs that are at the onset of S-phase (EdU⁺ cells) increases at a more rapid rate than WT over time. Gating for cells newly entering S-phase during the time course (Fig. 2E, S-phase, second cycle) demonstrated that in the absence of STING, cells reenter the cell cycle earlier than WT, with or without prior irradiation.

A simple interpretation of this pattern is that STING is required to properly delay onset of cell division, allowing completion of S-phase DNA replication, G₂ cell growth, and early events in mitosis. Deregulated mitotic entry promotes aneuploidy and polyploidy, leading to CIN (13) and mitotic catastrophe. Accordingly, loss of STING in MEFs or depletion in HCT116 tumor cells yielded a higher fraction of polyploid cells compared with their respective controls, as indicated by EdU⁺ cells with greater than 4N DNA content (Fig. 2F and G). Taken together, our data suggest that STING has critical roles in maintaining proper timing of the cell cycle, with a particular defect at onset of mitosis.

All EdU labeling experiments shown on Fig. 2 were performed at early time points post-radiation treatment (2–10 hours post-IR). Our data suggest that within this time frame, both WT and STINGko cell lines exhibited similar response to IR. However, STING has previously been reported to have a role in mediating type I IFN signaling in response to IR (4, 5, 21). We measured type I IFN signaling in WT and STINGko MEFs at 48 hours post-IR exposure. As expected, STINGko MEFs fail to induce TBK1 and IRF3 phosphorylation in response to IR exposure (Fig. 3A; Supplementary Fig. S3A). Consequently, IR-induced IFNβ production is abrogated in STINGko MEFs as well as shSTING MC-38 cells (Fig. 3B and C) while overexpression of FLAG-tagged STING in HEK293 cells led to a radiation dose-dependent increase in IFNβ promoter–driven luciferase activity (Fig. 3D).

To investigate transcriptional programs associated with STING-mediated regulation of cell cycle and response to IR, we used transcriptional profiling of STINGko and WT primary MEFs that were either mock-irradiated or exposed to 6 Gy IR. We analyzed the number of differentially expressed genes (DEG) and found that the absence of STING led to 2,946 DEGs. Of the 367 genes differentially expressed by WT cells 48 hours after irradiation, 265 were also identified as STING-dependent genes (Fig. 3E).

As expected, functional analysis of the DEGs using Ingenuity Pathway Analysis (IPA; Qiagen) revealed that the absence of STING led to significant differential expression of genes involved in activation of IFN regulatory factors (IRF) by cytosolic pattern recognition receptors as well as IFN signaling (Fig. 3F). But prominent in the IPA analysis were canonical pathways involved in cell-cycle control of chromosomal replication, chromosomal stability and segregation, and DNA damage response (Fig. 3F). To validate the cell-autonomous role of STING in DNA damage response, we performed colony-forming assays in STING-depleted human and murine tumor cell lines that were exposed to IR. The shSTING cells displayed less growth suppression after a single 6 Gy radiation dose compared with shScrambled control (Supplementary Fig. S3B–S3F). Consistent with low endogenous STING expression, shSTING A549 cells displayed similar growth and radiation response to shScrambled control (Supplementary Fig. S3G). In turn, both primary and immortalized MEFs from STINGko mice exhibited higher viability compared with WT MEFs in response to IR (Supplementary Fig. S3H and S3I).

The striking number of STING-dependent genes involved in cell proliferation pointed to STING functions that have not been fully explored to date. Several STING-dependent genes were components of the CDKN1A/RB/MDM2 network (Fig. 3G), including the stoichiometric cyclin-dependent kinase inhibitor CDKN1A (p21, CIP1, WAF1), yielding a P value of 2.2 × 10⁻³⁵ and an activation z-score of −3.16, indicating strong repression in the absence of STING. Western blot analysis of wild-type and STINGko primary and immortalized cell lines revealed that basal p21 expression was greater in WT than STINGko cells and p21 induction by IR was attenuated in the STINGko mutants (Fig. 3H and I; Supplementary Fig. S4A and S4B). Similarly, STING-depleted HCT116, and SCC61 tumor cells also expressed lower p21 compared with their shScrambled controls (Fig. 3J and K; Supplementary Fig. S4C and S4D). HEK293 cells overexpressing FLAG-tagged STING displayed modestly increased p21 expression and lower Rb phosphorylation compared with vector controls (Supplementary Fig. S4E). Confirming order of function between STING and p21, transient knockdown of p21 in HCT116 tumor cells using siRNA increased phosphorylated Rb, but did not affect STING expression or phospho-TBK1 levels (Supplementary Fig. S4F), while transient siRNA-mediated knockdown of STING in wild-type MEFs attenuated p21 expression and worked as well as transient knockdown of p21 (Supplementary Fig. S4G), indicating that STING acts upstream of p21. Time-lapse imaging revealed enhanced proliferation in both untreated and irradiated p21 knockdown HCT116 cells (Fig. 3L), partly recapitulating the phenotype of shSTING cell lines. Live cell imaging of nuclei-stained tumor cells also revealed that transient knockdown of p21 in stable STING–depleted tumor cell lines led to greater cell proliferation compared with their respective nontargeting controls (Supplementary Fig. S4H–S4J), suggesting that even though p21 acts as a downstream effector, STING only has partial control of p21.

Activated STING triggers the classical NFκB (p50/p65 heterodimer) transcriptional response, leading to induction of inflammatory mediators (1, 29). NFκB has also been implicated in transcriptional regulation of p21 (30, 31). We recently showed that, in response to DNA damage, NFκB binds to a consensus element in the p21 promoter, and induces p21 expression in p53-deficient acute myeloid leukemia cells (27). In HEK293 cells transiently transfected with an NFκB-dependent luciferase reporter along with a STING overexpression vector, treatment with cGAMP to induce STING led to a 6-fold increase in luciferase activity at 24 hours compared with vector control (Fig. 4A). Similarly, treating WT primary MEFs with 2′3′-cGAMP to induce STING led to upregulation of the NFκB subunit p65 and phospho-p65 along with the expected increase in phospho-TBK1 and higher expression of STAT1 (Fig. 4B; Supplementary Fig. S5A). Expression of p21 displayed a similar pattern to p65. As expected, STINGko primary MEFs stimulated with cGAMP failed to induce upregulation of phospho-TBK1, STAT1, phospho-p65, and p21 (Fig. 4C; Supplementary Fig. S5B). Likewise, STINGko MEFs displayed decreased phospho-p65 and lower p21 expression after IR (Fig. 4D; Supplementary Fig. S5C). Diminished p21 expression after IR was also observed in the NFκB DNA-binding domain–deficient p50^−/− MEFs (Fig. 4E; Supplementary Fig. S5D). Taken together, our data demonstrate that STING-mediated regulation of p21 involves the canonical NFκB pathway.

In response to IR-induced DNA damage, p21 is known to be predominantly regulated by p53 (32). To test the relative involvement of both NFκB and p53 in the transcriptional regulation of the p21 gene CDKN1A, we used luciferase reporter constructs driven by the CDKN1A promoter (27). When HCT116 cells were transfected with reporter constructs encoding a wild-type (GGGactccCC) or mutated (AATactccCC) p65 RE and stimulated with cGAMP, the mutant p65 RE displayed decreased activation at 24 hours (Fig. 4F). When the p21 reporters were similarly transfected in p53^−/− HCT116 cells, the wild-type p21 reporter was no longer induced by cGAMP and the mutant p65 RE reporter displayed reduced basal expression and no cGAMP induction. Taken together, our data indicate that STING regulates p21 activation through NFκB and p53 to modulate cellular proliferation.

Micronuclei accumulate in cells that enter mitosis with unrepaired chromosomal breaks following irradiation (33) and can be detected using the cytokinesis-block micronucleus (CBMN) assay (26). CBMN analysis detected micronuclei in both STINGko and WT MEFs after IR (Supplementary Fig. S6). Karyotyping metaphase cells from primary STINGko and WT MEFs at early passage detected greater aneuploidy in STINGko compared with WT MEFs, which was increased by IR (Fig. 5A). We have karyotyped a total of 40 metaphase cells each for unirradiated WT and STINGko primary MEFs, and 50 metaphase cells each for irradiated cells. Unirradiated primary WT MEFs displayed 19% aneuploidy that increased to 52% after 6 Gy while STINGko MEFS displayed 46% without IR and 64% after 6 Gy. Both Fisher exact test and χ² test for a 2 × 2 contingency table were performed to determine statistical significance of data. For unirradiated WT and STINGko cells, the two-tailed P value were 0.0307 using Fisher test and 0.017 using χ² test, suggesting that the association between the groups and outcomes is statistically significant. After radiation treatment (6 Gy), on the other hand, the outcomes were not statistically significant, as the calculated two-tailed P value were 0.3111 and 0.2241 for Fisher test and χ² test, respectively. The response to IR was more drastic in WT compared with STINGko MEFs, and is consistent with our previous observation that STINGko MEFs exhibited higher viability compared with WT cells. On the basis of detecting cells that appeared to have failed to complete mitosis and increased ploidy, we used flow cytometry and dual BrdU/PI staining to monitor the degree of polyploidy in STINGko and WT MEFs and compared these cells to HCT116 depleted of either p53 or p21 (Fig. 5B and C; Supplementary Fig. S7A–S7C). Consistent with prior studies (34, 35), p53-deficient HCT116 cells displayed increased polyploidy while p21 deficiency conferred a similar effect that was further enhanced after IR (Fig. 5C; Supplementary Fig. S7C). Western blot analysis revealed higher levels of BUB1 (Budding Uninhibited by Benzimidazoles 1) and MAD2L1 (Mitotic Arrest Deficient 2-like 1), key mitotic checkpoint proteins involved in controlling the alignment and segregation of sister chromatids during mitosis (36–38), in both STINGko and WT MEFs and HCT116 cells lacking either p21 or p53 (Supplementary Fig. S7D–S7F). Both BUB1 and MAD2L1 have been demonstrated to be negatively regulated by the p53-p21-DREAM (DP, Rb-like, E2F4m and MuvB) pathway, which function through the cell-cycle–dependent element and promoter sites of several late cell-cycle target genes (39). ChIP data revealed that the E2F4 transcription factor failed to bind to the promoter region of BUB1 and MAD2L1 in either STING- or p21-depleted HCT116 cells, both at basal level and in response to IR (Supplementary Fig. S7G). This helps explain why BUB1 and MAD2L1 protein levels are upregulated in the absence of STING and provides a key mechanism for p53-p21–mediated G₂–M cell-cycle arrest in response to genotoxic stress.

While p21 is effective at restraining progression at G₁–S, it has limited impact at G₂–M, where the STING-deficient cells displayed the most dramatic cell-cycle defect. In the normal cell cycle, premature activation of the mitotic cyclin-dependent CDK1 (CDC2) is blocked during S-phase and G₂ by inhibitory Tyr15 phosphorylation by WEE1 (40). DNA damage in S or G₂ activates the checkpoint and inhibits Tyr15 dephosphorylation, delaying mitosis to allow time for repair. Western blot analysis showed that the Tyr15 phosphorylated CDK1 was lower in STINGko MEFs compared with WT both at the basal level and after IR (Fig. 5D; Supplementary Fig. S8A). Staining for mitochondria and DNA in STINGko MEFs suggested that both the cytoplasmic and nuclear volumes were markedly reduced compared with WT (Fig. 5E). Total nuclear area in both STINGko MEFs and shSTING D54 cells were significantly smaller, based on time-lapse imaging (Fig. 5F and G), suggesting that STINGko MEFs have a similar phenotype to wee mutants that are defective in protein regulators that normally prevent cells from dividing when they are too small (41). Although STINGko MEFs appeared to increase in size slightly after IR, the effect was markedly less than for WT cells (Fig. 5E and F).

Inhibition of the kinase WEE1 in combination with DNA-damaging agents is a promising strategy in cancer therapy (42). The resulting activation of CDK can induce replication stress (43) or premature mitotic entry (42), leading to mitotic catastrophe and cell death. We hypothesized that further impairing G₂–M regulation in STING-deficient cells by blocking CDK1 Tyr15 phosphorylation by WEE1 might be differentially toxic and/or further sensitize the cells to IR. Time-lapse analysis of cell proliferation revealed that WEE1 inhibitor MK1775 (42, 44) conferred greater growth inhibition and enhanced the effects of IR on STINGko MEFs (Fig. 5H and I). A similar pattern was observed for shSTING D54, HCT116, and SCC61 (Fig. 5J–L). As expected, WEE1 inhibitor in combination with IR also inhibited cell proliferation in WT MEFs and shScrambled D54, HCT116, and SCC61 tumor cells (Supplementary Fig. S8B–S8E).

We demonstrated that STING has cell-intrinsic functions in the regulation of the cell cycle and chromosomal stability. Depletion of STING conferred a shorter doubling time in all cell types compared with wild-type, implicating STING in cell-intrinsic control of proliferation and response to IR exposure. Our finding is consistent with recent data correlating STING suppression and poor prognosis in patients with hepatocellular carcinoma (45). RNA expressional profiling demonstrated that STING is involved in cell-cycle control through CDKN1A. We further hypothesized a mechanistic link between STING and CDKN1A can be provided by NFκB based on published observations that NFκB transcriptionally activates CDKN1A (30, 31). Here we provide the first evidence that both p65 and p53 are required for the STING-mediated activation of p21, implicating a previously unrecognized STING/NFκB/p53/CDKN1A pathway in cell-intrinsic regulation of cell proliferation. This finding is consistent with a previous report demonstrating a high concentration of p65 transcription factor bound to the chromatin of senescent cells, and that senescence was bypassed using a combination of NFκB inhibition and p53 inactivation (46). These results provide additional support to recent reports implicating the cGAS/STING pathway in regulation of senescence phenotype (14, 17), as well as a previous study demonstrating that wild-type, but not STINGko mice, exhibited accelerated aging and massive hair graying three months after exposure to IR (18). However, our data indicate that STING regulates the G₁–S transition independently from cGAS, because at least one of the cell lines tested (HCT116) does not express cGAS (see ref. 22 and Fig. 3G).

Progression from G₂ to M phase is driven by the activation of the CDK1/cyclin B1 complex (40). The phosphorylation of CDK1 on Thr14 and Tyr15 by Myt1 and Wee1 kinase, respectively, inhibits CDK1 activity during G₂, while dephosphorylation by CDC25 phosphatase at these tyrosine sites enable progression through M-phase (42). The CDK1/cyclin B1 heterodimer induces mitosis by activating enzymes involved in spindle assembly and mitosis-specific microtubule reorganization. Karyotypic analyses on primary MEFs revealed that the absence of STING led to a higher degree of chromosomal aberrations and polyploidy and these aberrations are enhanced by IR. This result suggests that STING modulates entry to mitosis and consequently, the assembly of the mitotic spindle. We observed that STINGko MEFs exhibited lower levels of inactive CDK1 and higher levels of the mitotic checkpoint markers BUB1 and MAD2L1 compared with WT both at basal level and in response to IR. Upregulation of BUB1 and MAD2L1 in tumor cells are associated with tumor cell proliferation (37, 47, 48). Analysis of normal and tumor samples from The Cancer Genome Atlas project demonstrated a positive correlation between high copy number variations and high expression of genes implicated in cell-cycle regulation, G₂–M checkpoints, mitosis, and chromosome maintenance, suggesting that there is increased proliferation in high aneuploid tumors (49). Taken together, our findings suggest a putative tumor-suppressive role for STING.

Micronuclei formation following genotoxic stress has been used as a marker for CIN (33). Double-strand breaks in chromosomes give rise to chromatin fragments that do not segregate with the rest of the genome following mitosis, which eventually form into micronuclei (33). Micronuclei have been observed to spontaneously undergo nuclear envelope collapse in normal and tumor cells during interphase, leading to the release of cytoplasmic DNA fragments that trigger cGAS/STING-mediated innate immune activation of cytokines and chemokines (15, 16). Both WT and STINGko MEFs exposed to IR were observed to form micronuclei (Supplementary Fig. S6), but STINGko cells exhibited higher degree of CIN than WT (Fig. 5A), presumably as a result of its inability to sense local danger signals. Consequently, STING-depleted cells will continue to exhibit a deregulated cell cycle, accumulate errors in their genome, and fail to activate host immune proinflammatory responses. This is consistent with previous reports demonstrating that the cGAS–STING pathway is critical for the induction of inflammatory cytokines and chemokines that modulate the host immune responses (14, 17, 18). Our findings help shed light on the connection between accumulation of nuclear structure abnormalities in tumor cells and the mechanisms that they employ to evade host-mediated immune responses. However, in contrast to our results, Bakhoum and colleagues (50) reported that continuous chromosome segregation errors promoted tumor metastasis through cGAS-STING–mediated activation of noncanonical NFκB following recognition of accumulated cytosolic DNA. This correlated with a shorter disease-free survival in patients with breast and lung cancer (50). The discrepancy between our findings and their study suggest that STING may be a tumor suppressor in certain tumor types but could have tumor-promoting effects in others. These paradoxical roles of STING warrant further investigation.

Quantifying STING expression in tumor samples might be useful in precision oncology in that alternative modes of killing or greater radiation doses might be required in STINGko tumors due to increased perturbations in cell-cycle regulation that favors tumor survival. In addition, because most normal cells express STING and many tumor cells do not, this difference might be exploited in cancer therapy. It is postulated that induction of mitotic catastrophe through WEE1 inhibition may be used as radiosensitizers in the clinical setting (42). We demonstrated that STING-depleted cells treated with the WEE1 inhibitor MK1775 in combination with IR exhibited cell growth delay compared with WT control. WEE1 inhibitors have been shown to trigger mitotic entry prior to completion of DNA synthesis and DNA damage repair, which can lead to mitotic catastrophe and cell death (42). Downregulation of WEE1 also leads to elevated CDK activity during S-phase, resulting in replication stress (43). Considering current findings regarding frequent deletion/suppression of cGAS/STING pathway in tumor cells, our data pose important questions about tumor-suppressive functions of STING as a potential “gatekeeper” or homeostatic regulator, the absence of which may be associated with early stages of oncogenesis (12). Further investigations on the regulation and function of STING in normal and cancer cells are warranted by these insights and may provide additional therapeutic targets to help improve cancer therapy.

R.R. Weichselbaum has ownership interest (including stock, patents, etc.) in OncoSenescence and Boost Therapeutics Inc. R.C. Widau is a clinical research scientist and has ownership interest (including stock, patents, etc.) in Eli Lilly and Company. R. Parry has ownership interest (including stock, patents, etc.) in Varian Medical Systems. S.J. Kron reports receiving commercial research grant from AbbVie and has ownership interest (including stock, patents, etc.) in OncoSenescence. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D.R.E. Ranoa, M.J. Bolt, R. Parry, N.N. Khodarev, R.R. Weichselbaum

Development of methodology: D.R.E. Ranoa, A.D. Parekh, C.M. Nicolae, G.-L. Moldovan, N.N. Khodarev

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.R.E. Ranoa, R.C. Widau, A.D. Parekh, R.R. Weichselbaum

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.R.E. Ranoa, R.C. Widau, S. Mallon, M.J. Bolt, A. Arina, R.R. Weichselbaum

Writing, review, and/or revision of the manuscript: D.R.E. Ranoa, R.C. Widau, S. Mallon, A. Arina, R. Parry, S.J. Kron, N.N. Khodarev, R.R. Weichselbaum

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.R.E. Ranoa, X. Huang, R.R. Weichselbaum

Study supervision: D.R.E. Ranoa, S.J. Kron, N.N. Khodarev, R.R. Weichselbaum

The authors would like to thank Rolando Torres, Michael Beckett, Dr. Giovanna Bernal, Dr. Longtao Wu, and Dr. David Hosfield for technical assistance; Amy K. Huser for assistance in manuscript preparation; Dr. David J. Grdina for intellectual discussions; Drs. Michael Spiotto and Geoffrey Greene for sharing reagents and equipment; Yue Liu for critical reading of the manuscript; Dr. Pieter Farber and the Functional Genomics Facility staff; Dr. Vytas Bindokas, Shirley Bond, and Dr. Christine Labno of the Integrated Microscopy Core Facility; and Dr. Siquan Chen of the Cellular Screening Center. This work was supported by the Ludwig Center for Metastasis Research, the Virginia and Ludwig Foundation for Cancer Research, Varian Medical Systems, the Lung Cancer Research Foundation (all to R.R. Weichselbaum), NIH NCI RO1 grant (CA199663 to S.J. Kron), and a generous gift from Mr. and Mrs. Vincent Foglia. The Integrated Light Microscopy Facility is funded by the University of Chicago Institute for Translational Medicine and the Cancer Center Support Grant (P30 CA014599) at the University of Chicago.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.