DNA-triggered innate immune responses are propagated by gap junction communication

GFP Reporters Sensitive and Specific for IRF3-Activating Stimuli.

dsDNA is known to stimulate the expression of genes with ISRE consensus sequences in their promoters through activation of the IRF3 transcription factor (9, 12). To investigate the spatiotemporal evolution of dsDNA-induced gene expression, we created a stable monoclonal ISRE-GFP reporter cell line in a hepatocyte-derived H35 cell line and selected clones exhibiting low-baseline GFP expression in the absence of stimuli and high dsDNA-induced GFP expression (Fig. 1A). To characterize the reporter, we first assessed its specificity for IRF3-dependent gene expression by exposing confluent reporter monolayers to various immunostimulatory molecules and measuring ISRE reporter fluorescence by using flow cytometry (Fig. 1B). When reporters were exposed to synthetic B-form Poly(dA-dT):Poly(dA-dT) (hereafter referred to as dsDNA) or DNA encoding the red fluorescent protein (dsRED), total population fluorescence increased in a dose-dependent fashion. Similarly, when cells were stimulated with polyinosinic:polycytidylic acid [poly(I:C)], a synthetic double-stranded RNA known to activate IRF3, dose-dependent increases in total fluorescence were also observed. In contrast, IRF3-independent stimuli such as TLR9-dependent CpG DNA, IFNβ, poly(A) ssDNA, and siRNA did not elicit measurable increases in ISRE reporter fluorescence at any of the doses examined. As anticipated, LPS, which classically signals through TLR4 and IRF3, failed to activate the ISRE reporter, consistent with the fact that nonimmune cells do not express significant levels of TLR4 (confirmed by qPCR). To verify that the ISRE reporters retained previously reported dsDNA-induced responses (9), we confirmed the expression of IFNβ and TNFα after 12 h of dsDNA stimulation (Fig. 1C). Taken together, these results demonstrate that the ISRE reporters are sensitive for both dsDNA and Poly(I:C) stimulation, and specific for IRF3-activating stimuli.

When we examined the distribution of ISRE reporter fluorescence after dsDNA stimulation by flow cytometry, we found that increasing doses of dsDNA resulted in increasing numbers of activated cells rather than increasing levels of activation in all cells, indicating that the cellular response was heterogeneous (Fig. 1 D and E). Therefore, we used fluorescence microscopy to examine the distribution of dsDNA-induced GFP expression and discovered a striking spatial pattern. dsDNA stimulation of confluent reporter monolayers exhibited well-delineated clusters of GFP-positive cells or “colonies” in an otherwise dark background of nonactivated cells. Increasing doses of dsDNA led to increasing numbers of colonies with eventual bridging of adjacent colonies (Fig. 1 F and G). To gain further insight into colony formation, we examined the temporal evolution of ISRE-activated colonies by monitoring GFP-reporter induction with time-lapse fluorescence microscopy and quantifying colony area by using custom automated image analysis software (Fig. 1 H and J) (see Movie S1 online). Colony activation began 8–12 h after dsDNA stimulation and grew to a steady-state size with clearly demarcated colony borders characterized by highly induced reporter cells inside and uninduced cells outside. These findings confirm that dsDNA stimulation leads to spatially heterogeneous patterns of ISRE-activated colonies in an otherwise uninduced confluent monolayer of cells. Interestingly, Poly(I:C) stimulation also led to ISRE reporter activation; however, the response was spatially homogeneous across the reporter monolayer and did not result in colony formation.

Because dsDNA requires polyelectrolyte complexing for immunogenicity and high molecular weight complexes have short diffusion distances, we reasoned that dsDNA complexes land in discrete locations within the cultured monolayer, stimulate individual dsDNA-sensing cells, and ISRE colony formation arises by a secondary intercellular communication signal. To clarify which cells were directly and indirectly activated by dsDNA, ISRE reporters were stimulated with dsRED. dsRED stimulation of ISRE-GFP reporters resulted in multicellular ISRE reporter cell colonies surrounding individual dsRED DNA-sensing cells (Fig. 1K and Fig. S1). By using the custom automated image analysis software discussed above, we determined that the average colony was comprised of 23.5 ISRE-activated cells, of which 2.3 cells were dsRED DNA-sensing cells (see Fig. S1D).

Spatiotemporal IRF3-Mediated Gene Expression.

We next sought to determine whether the colonies of GFP⁺ ISRE reporters had functional significance by examining their gene expression profiles and comparing them with their GFP⁻ neighbors. Confluent monolayers of reporters were stimulated with dsDNA for 12 h, sorted into GFP⁺ and GFP⁻ populations by FACS, and examined for expression of IRF3-mediated genes by qPCR (Fig. 2A). GFP⁺ cells represented less than 8% of the total population, yet expressed more than 95% of secreted cytokines and chemokines including IFNβ, TNFα, and IP10 (Fig. 2B). In contrast, expression of IFNβ-inducible genes, with direct antiviral properties such as PKR and OAS1, did not differ significantly between the 2 populations (Fig. 2C). Given the commonly held view that antiviral genes are expressed in response to the paracrine action of IFNβ, these results suggest that dsDNA-sensing in individual cells leads to IRF3 activation in colonies of cells that collectively secrete cytokines, such as IFNβ and TNFα, to establish an antiviral state across the broader population.

IRF3 Activation by Contact-Dependent Intercellular Communication.

We hypothesized that IRF3-activated colony formation required secondary intercellular communication from dsDNA-sensing cells. To test this hypothesis, we developed a transplant coculture system using dsDNA-stimulated nonreporter cells as donors and unstimulated ISRE reporters as recipients. Donor cells were stimulated with dsDNA for 6 h, trypsinized, thoroughly washed, and transplanted onto recipient ISRE reporters that had never been exposed to dsDNA (Fig. 3A). After 18 h of coculture, ISRE reporters were activated in small colonies surrounding donor cells (Fig. 3 B and C). In contrast, no IRF3 activity was observed in the reporters when mock dsDNA stimulated nonreporter cells were transplanted (Fig. 3C). Interestingly, not all dsDNA-stimulated cells were able to activate IRF3 in neighboring reporters. dsDNA-stimulated human cervical cancer cells (HeLa) and mouse neuroblastoma cells (N2A) were unable to activate IRF3 in the reporters (Fig. 3 B and C), demonstrating that the phenomenon is cell-type specific and not an artifact of nonspecific dsDNA carry-over from the donor cells. Donor cells were also stimulated with dsRED DNA, to identify them within the coculture, and to confirm that they make contact and adhere to the reporter cells (see Fig. S4). To investigate whether direct cell contact was necessary for this dsDNA-induced secondary intercellular communication, we cultured dsDNA-stimulated donors and ISRE reporter recipients on opposite surfaces of a microfluidic parallel plate bioreactor (separation gap of ≈50 μm; see Fig. S5). Negligible reporter induction was observed, suggesting that dsDNA-induced intercellular communication is contact dependent. Taken together, these data suggest that dsDNA stimulation induces contact-dependent intercellular communication from dsDNA-sensing cells to their unstimulated neighbors, propagating an IRF3-activating signal and amplifying IRF3-dependent gene expression.

To further clarify the pathways connecting donor dsDNA-sensing and recipient IRF3 activation, we used genetic knockout mouse embryonic fibroblasts (MEFs) to determine the necessity of critical proteins. Wild-type MEF donors stimulated with dsDNA activated ISRE in reporter recipients (Fig. 3 D and E). In addition, MEFs deficient in both TBK1 and IKKε (Tbk1^−/−Ikbke^−/−), kinases necessary for IRF3 activation, also activated ISRE in reporter cells (Fig. 3 D and E). Similarly, dsDNA-stimulated MEFs deficient in both IKKα and IKKβ (IKKα^−/−IKKβ^−/−), kinases essential for NFκB activation, also activated ISRE in reporter cells (Fig. 3 D and E). MEFs deficient in both MyD88 and TRIF (MyD88^−/−TRIF^−/−) were also able to activate ISRE reporter recipients. Together, these data suggest that dsDNA-induced intercellular communication is TLR-independent and occurs upstream of both IRF3 and NFκB activation in dsDNA sensing cells.

Gap Junctions Are Necessary for Amplified IRF3 Activation.

Gap junction communication enables the rapid, localized exchange of information between cells linked through connexin protein channels (16). To determine the necessity of gap junctions in the dsDNA-induced intercellular communication, we pretreated ISRE reporters with 18β-glycyrrhetinic acid (18βGA), a molecular inhibitor of gap junctions (20), before dsDNA stimulation. Compared with vehicle controls, pretreatment with 18βGA dramatically reduced colony size, resulting in mostly single cell reporter activation (Fig. 4A). Images were quantified by creating isointensity contour maps and calculating the average size per colony for each colony present in the images (Fig. 4B). The average size of dsDNA-induced ISRE-activated colonies was reduced by more than 10-fold with 18βGA treatment compared with no 18βGA treatment. More than 75% of all colonies formed in the presence of 18βGA contained fewer than 3 cells, whereas more than 90% of all colonies formed in the absence 18βGA contained more than 3 cells (Fig. 4C). In addition, the overall number of ISRE-activated GFP-positive cells was decreased from 38% to 5%, as a result of gap junction blockage (Fig. 4D). These results were further validated by Cx32 knockdown analysis using Cx32-targeted siRNA (see Fig. S3). Compared with control siRNA, Cx32 siRNA knockdown resulted in a significant decrease in both the size of dsDNA-induced ISRE active colonies and in the overall number of dsDNA-induced GFP-positive reporters. More than 70% of all colonies formed in the presence of Cx32 knockdown contained fewer than 3 cells, whereas more than 85% of all colonies formed in the presence of control knockdown contained more than 3 cells (see Fig. S3).

The utility of chemical gap junction inhibitors such as 18βGA is limited because of their nonspecific side effects and unknown mechanism of action (21). Therefore, to definitively demonstrate the necessity of gap junction communication for the propagation of IRF3 activity, we used the transplant coculture system with genetically modified cell lines. HeLa and N2A cell lines have been historically shown to be gap junction- and connexin-deficient (21, 22). We obtained modified monoclonal HeLa and N2A cells stably transfected with individual connexin transgenes (Cx26, Cx32, or Cx43), thereby reconstituting connexin expression and gap junction communication (21, 22). PCR analysis was performed to verify appropriate connexin expression in all HeLa and N2A cells. PCR analysis showed that ISRE reporters express only Cx32 of the β-subset and therefore only form functional gap junction channels with other cells expressing β connexins 26 or 32. When WT and Cx43-expressing HeLa and N2A cells were stimulated with dsDNA, minimal ISRE reporter activation was measured (HeLa WT, 0.5%; N2A WT, 1.0%; HeLa Cx43, 2%; N2A Cx43, 1.5%), suggesting a lack of intercellular communication (Fig. 4 E–H). However, when Cx26- and Cx32-expressing cells were stimulated with dsDNA, significant ISRE reporter activation was observed (Fig. 4 E and F), with 63% of reporters activated by Cx32-expressing HeLa cells and 23% by Cx32-expressing N2A cells (Fig. 4 G and H). These data suggest that functional gap junction channels are necessary for dsDNA-induced intercellular communication and for amplifying IRF3 activation. To further generalize the utility of gap junctions for amplifying IRF3 activation in other cell types, we constructed another ISRE-GFP reporter in a stromal hepatic stellate cell line (HSC). When the HSC ISRE-GFP reporters were stimulated with dsDNA, they were able to form multicellular colonies of IRF3-activated reporters in a gap junction-dependent manner (see Fig. S2).

Propagation of Antiviral and Inflammatory Responses by Gap Junctions.

To investigate the physiological significance of gap junction-mediated amplification of IRF3 activity, we disrupted gap junctions and examined the ability of cells to mount innate immune responses. When ISRE reporters were pretreated with 18βGA and stimulated with dsDNA, they expressed significantly lower levels of critical antiviral cytokines in comparison to vehicle pretreatment. After 8 h of dsDNA stimulation, 18βGA pretreated cells expressed 6-fold less IFNβ than vehicle-treated cells (Fig. 5A). Expression of the IFNβ-stimulated antiviral protein PKR was also significantly reduced by 2-fold (P < 0.05) in the 18βGA-treated cells, compared with vehicle treatment (Fig. 5A). Similarly, the expression of the proinflammatory cytokine, TNFα, was reduced by 3.5-fold with gap junction inhibition (Fig. 5A).

We further verified the role of gap junctions in innate immune responses by using genetically modified cells and the transplant coculture system described above. Rat ISRE reporter cells, which express only Cx32, were stimulated with dsDNA and then cocultured with unstimulated human HeLa cells that either lacked gap junctions (HeLa WT) or expressed Cx32 gap junctions (HeLa Cx32). Recipient HeLa cells were then analyzed for cytokine expression and secretion using human-specific PCR primers and ELISAs. When dsDNA-stimulated reporter cells were cocultured with HeLa Cx32 cells, HeLa IFNβ and TNFα expression was increased by 6- and 4-fold respectively, compared to coculture with gap junction-deficient HeLa WT cells (Fig. 5B). Additionally, coculture of dsDNA-stimulated reporters with HeLa Cx32 cells triggered a 5-fold increase in production of human IFNβ compared to coculture with gap junction-deficient HeLa WT cells (Fig. 5C). Taken together, these data suggest that gap junctions amplify dsDNA-induced expression and secretion of IFNβ and TNFα, and that without functional gap junctions, cytokine secretion is impaired, consequently reducing the antiviral state in the overall population.

Cells and Reagents.

Hepatocyte-derived H35 cells were maintained as previously described (29). Construction of the ISRE-GFP reporter was performed as described in SI Materials and Methods. MEFs from WT and knockout mice and Cx26-, Cx32-, and Cx43-expressing HeLa and N2A cells were gifts (see Acknowledgments). Synthetic polydeoxynucleotides poly(dA-dT):poly(dA-dT) dsDNA and poly(dA-dT) ssDNA, synthetic poly(I:C), CpG ODN, LPS, 18β-glycyrrhetinic acid and pdsRED was purchased from commercial sources (see SI Materials and Methods). DNA transfections were performed by using Lipofectamine LTX (Invitrogen) at a ratio of 1.5:1 (volume/weight) with DNA as per manufacturer's protocol. For experiments involving the use of 18βGA, cells were pretreated with 25 μM 18βGA in DMEM for 1 h before dsDNA stimulation and during stimulation.

Quantitative RT-PCR and ELISA.

Analysis of gene expression by quantitative RT-PCR, and protein secretion by ELISA is detailed in SI Materials and Methods. RT-PCR primer sequences are provided in Table S1.

Transplant Coculture Assay.

Donor cells were transfected with 10 μg/mL of poly(A:T) dsDNA. Six hours after transfection, donor cells were trypsinized, washed three times in PBS, and counted. dsDNA-stimulated donor cells (1 × 10⁵ cells/mL) were transplanted onto a subconfluent layer of recipient ISRE reporters, maintaining a 1:10 ratio between donor and recipient cells. After 18 h of coculture, recipient reporter cells were analyzed by fluorescent microscopy and FACS for ISRE activity. In some cases, IFNβ and TNFα secretion by recipient cells was measured by ELISA. Gene expression analysis on recipient cells was performed by quantitative RT-PCR using recipient cell species-specific primers (provided in Table S2).