Borrelia burgdorferi Engages Mammalian Type I IFN Responses via the cGAS–STING Pathway

Lyme disease results from infection with the spirochetal bacterium Borrelia burgdorferi and presents as a multisystemic inflammatory disease causing debilitating morbidity as a result of fatigue, malaise, severe arthritis, and cardiac and neurologic complications (1–4). It is the most common tick-borne disease in the United States and is a significant public health concern, with the Centers for Disease Control and Infection reporting 39,000 cases per year, but insurance reports indicating a prevalence of greater than 470,000 cases per year (5–7). Antibiotic treatment is highly effective when administered shortly after the tick bite, yet antibiotic efficacy declines as the borrelial infection progresses (8, 9). Lyme disease occurs in stages of localized, disseminated, and chronic infection as the extracellular pathogen spreads from the site of the tick bite to secondary tissues, including the joints, heart, and CNS, causing Lyme arthritis, carditis, and neuroborreliosis, respectively (1–3, 8). The development of arthritis, a characteristic symptom of late Lyme disease in North America, is associated with a robust innate immune response that includes induction of type I IFN (IFN-I) cytokines (10–12).

B. burgdorferi elicits a robust innate immune response, resulting in the secretion of proinflammatory cytokines and chemokines (13–20). In addition, B. burgdorferi triggers IFN-I responses in a wide array of human cells, mouse cells, and infected tissues (11–13, 19, 21–28). Evidence suggests that innate IFN-I signaling plays key roles in several aspects of B. burgdorferi pathology (29–33). Notably, IFN-I and resulting IFN-stimulated gene (ISG) signatures are linked to the development of more severe arthritis in experimental models and also correlate with lingering neurocognitive symptoms of Lyme disease (12, 30, 32, 34). In addition, a recent study from Lochhead and colleagues (35) observed that a robust IFN gene signature correlates with decreased expression of tissue-repair genes in synovial lesion biopsies from patients with postinfectious, B. burgdorferi–induced Lyme arthritis. It is well appreciated that innate immune sensing of the abundant B. burgdorferi lipoproteins via TLR2 triggers production of proinflammatory cytokines and chemokines in vitro and in vivo (14, 36–38). However, lipoprotein binding to TLR2 is not a robust inducer of IFN-I responses in most mouse and human cell types (39, 40). Studies using human cells have implicated nucleic acid–sensing TLRs (TLR7, 8, and 9) as regulators of IFN-I induction in human immune cells challenged with B. burgdorferi in vitro (13, 18, 19). In contrast, other reports have shown that B. burgdorferi can engage IFN-I responses in nonphagocytic fibroblasts and endothelial cells, which do not express a full complement of TLRs (25, 41). Moreover, B. burgdorferi–related ISG induction in murine macrophages is independent of the two primary TLR adaptor proteins MyD88 and TRIF (22, 23). Thus, TLR-mediated sensing of B. burgdorferi pathogen-associated molecular patterns does not appear to be the predominant trigger of IFN-I responses in mammalian cells during infection.

Innate immune pathways that sense cytosolic nucleic acids, such as the RIG-I–like receptor–mitochondrial antiviral signaling (MAVS) or cyclic GMP-AMP synthase (cGAS)-stimulator of IFN genes (STING) pathway, have emerged as key regulators of IFN-I production in both immune and nonimmune cells (42–49). cGAS is an intracellular DNA sensor that localizes to the mammalian cell cytoplasm and nucleus. Upon binding intracellular pathogen DNA, micronuclei, or mitochondrial DNA (mtDNA), cGAS generates the noncanonical cyclic dinucleotide, 2′3′-cyclic guanosine monophosphate-adenosine monophosphate, which binds STING. This results in the recruitment and activation of Tank-binding kinase 1 (TBK1), leading to the phosphorylation of IFN regulatory factor 3 (IRF3) for the induction of IFN-Is (IFNα, β) and ISGs (42–44). Although initially identified as an antiviral host defense pathway, the cGAS–STING pathway is also critical for induction of robust IFN-I responses to intracellular bacteria, such as Mycobacterium tuberculosis and Listeria monocytogenes (45, 50, 51). Moreover, recent work has shown that the cGAS–STING pathway is essential for IFN-I production in response to multiple extracellular pathogens, including Pseudomonas aeruginosa, Klebsiella pneumoniae, and Staphylococcus aureus (52, 53). B. burgdorferi, predominantly an extracellular pathogen, is readily taken up by phagocytic cells and associates with endothelial cells and fibroblasts, which are key sources of IFN-I in the Lyme disease joint (54–62). B. burgdorferi also produces cyclic dinucleotides c-di-GMP and c-di-AMP, which can directly engage STING (63–66). Thus, there are several potential routes by which B. burgdorferi infection could trigger the cGAS–STING–IFN-I pathway.

In this study, we tested the hypothesis that B. burgdorferi infection induces IFN-I through the cGAS–STING pathway. We exposed phagocytic and nonphagocytic cells lacking various components of the cytosolic nucleic acid–sensing machinery to viable and sonicated B. burgdorferi and evaluated IFN-I and ISG expression after exposure. We also assessed the degree of association of B. burgdorferi with cultured fibroblasts and examined colocalization of cGAS with intracellular spirochetes. Furthermore, we performed an infectivity study with bioluminescent B. burgdorferi in mice deficient in cGAS or STING to assess borrelial load by in vivo imaging and joint inflammation using histopathology. Our results reveal that B. burgdorferi engages IFN-I responses in a cGAS-STING–dependent manner without significantly altering infection kinetics or borrelial load in tissues.

C57BL/6J (strain 000664), cGAS-deficient (cGAS knockout [cGAS^KO], strain 026554), STING-deficient (STING^KO, strain 017537), IFN-I receptor (IFNAR)-deficient (IFNAR^KO, strain 028288), and MAVS-deficient (MAVS^KO, strain 008634) mice were obtained from the Jackson Laboratory. MAVS^KO mice were backcrossed to C57BL/6J mice for 10 generations before generation of primary cell lines. Mice were group housed in humidity-controlled environments maintained at 22°C on 12-h light–dark cycles (600–1800 h). Food and water were available ad libitum. All animal experiments were conducted in accordance with guidelines established by Department of Health and Human Services Guide for the Care and Use of Laboratory Animals and the Texas A&M University Institutional Animal Care and Use Committee.

Low-passage B. burgdorferi strains B31-A3 and ML23 pBBE22luc were cultured in BSK-II medium with 6% normal rabbit serum (Pel-Freeze Biologicals, Rogers, AR) and grown to midlog phase at 37°C at 5% CO₂ (67–71). ML23 pBBE22luc cultures were supplemented with 300 μg/ml kanamycin.

Primary mouse embryonic fibroblasts (MEFs) were generated from WT, cGAS^KO, STING^KO, IFNAR^KO, and MAVS^KO embryonic day 12.5 to 14.5 embryos. Cells were grown in DMEM (D5756; Millipore Sigma) containing 10% low endotoxin FBS (97068-085; VWR) and cultured for no more than four passages before experiments. SV40 immortalized cGAS^KO MEFs reconstituted with hemagglutinin (HA)-tagged mouse cGAS were previously reported (72). Primary bone marrow–derived macrophages (BMDMs) were generated as described previously (73). In brief, bone marrow cells were collected from the femur and tibia of mice and differentiated into macrophages in DMEM containing 10% low endotoxin FBS and 20% (v/v) conditioned media harvested from L929 cells (CCL-1; ATCC). Cells were plated in petri plates and maintained in L929-conditioned media for 7 d. The day before experiments, macrophages were plated in tissue culture plates and maintained in 5% L929. To generate immortalized mouse macrophages (iBMDMs), we infected BMDMs with J2 recombinant retrovirus (encoding v-myc and v-raf oncogenes) as described previously (74). iBMDMs were passaged for 3–6 mo and were slowly weaned off of L929 conditioned media until they stabilized into cell lines. Human foreskin fibroblasts (HFFs; SCRC-1041; ATCC) were immortalized using a human telomerase-expressing retrovirus (pWZL-Blast-Flag-HA-hTERT, 22396; Addgene).

Bacteria were prepared as previously described (61) with the following exceptions. B. burgdorferi was grown to midexponential phase, centrifuged at 6600 × g for 8 min, washed twice in PBS, and resuspended in DMEM (D5796; Millipore Sigma) with 10% FBS (97068-085; VWR). Spirochetes were enumerated using dark-field microscopy and diluted to the appropriate multiplicity of infection (MOI). Where indicated, plates were spun after the addition of bacteria to mammalian cells for 5 min at 300 × g. Small molecule inhibitors were added to MEFs 1 h before B. burgdorferi infection. cGAS inhibitor RU.521 (HY-114180; MedChemExpress) and STING inhibitor H-151 (HY-112693; MedChemExpress) were added at 10 and 0.5 mM, respectively (75, 76). MEFs or BMDMs were transfected with 2 µg/ml IFN Stimulatory DNA (tlrl-isdn; InvivoGen) complexed with Lipofectamine 2000 (11668019; ThermoFisher) in Opti-MEM media (11058021; Life Technologies) for 5–20 min (73).

RNA was isolated from mammalian cells using the Quick-RNA Micro Prep Kit (R1051; Zymo Research) according to the manufacturer’s instructions. Between 300 and 500 ng of RNA was standardized across samples from each experiment and converted to cDNA with the qScript cDNA Synthesis Kit (95047; QuantaBio). Quantitative PCR was performed on cDNA using the PerfecTa SYBR Green FastMix (95072; QuantaBio) and primers listed in Table I. Each biological sample was assayed in triplicate. Relative expression was determined for each triplicate after normalization against a housekeeping gene (Bactin or Gapdh) using the 2^−ΔΔCT method. DNA contamination of RNA samples was evaluated in a single, no reverse transcriptase reaction for each primer set.

Protein was collected from cells lysed in 1% Nonidet P-40 buffer (50 mM Tris [pH 7.5], 0.15 M NaCl, 1 mM EDTA, 1% Nonidet P-40, and 10% glycerol) supplemented with protease inhibitor (04693159001; Roche) and spun for 10 min at 17,000 × g at 4°C. The supernatant was collected and stored at −80°C. Protein lysates were quantified using the micro-BCA assay (23235; ThermoFisher Scientific, Waltham, MA). Immunoblotting was performed as described by Torres-Odio et al. (77). In brief, between 20 and 30 mg protein was run on 10–20% SDS-PAGE gradient gels and transferred onto 0.22 μM polyvinylidene difluoride membranes (1620177; Bio-Rad). After air-drying to return to a hydrophobic state, membranes were incubated in primary Abs (Table II) at 4°C overnight in 1× PBS containing 1% casein, HRP-conjugated secondary Ab at room temperature for 1 h, and then developed with Luminata Crescendo Western HRP Substrate (WBLUR0500; Millipore).

Cells were seeded on 12- or 18-mm sterile coverslips, allowed to adhere overnight, and infected as described earlier. At the conclusion of infection, cells were washed with DMEM and then 1× PBS, fixed with 4% paraformaldehyde for 15 min at room temperature, and washed twice with 1× PBS for 5 min each. Cells were permeabilized with 0.1% Triton X-100 in PBS for 5 min at room temperature, washed twice with 1× PBS, and blocked for 30 min in PBS containing 5% FBS. Each coverslip was stained with primary (Table II) and secondary Abs for 1 h each. Cells were washed in 1× PBS with 5% FBS three times after each stain for 5 min each. After the last wash, cells were incubated with CellMask Green (H32714; Invitrogen) for 30 min at 1:500 dilution in 1× PBS. Then, cells were washed two times more for 5 min each with 1× PBS. After the last wash, cells were incubated with DAPI (62247; ThermoFisher Scientific) for 2 min at 1:2000 dilution in 1× PBS. Cells were washed two times more for 5 min each with 1× PBS, and coverslips were mounted with ProLong Diamond Antifade Mountant (P36961; Invitrogen) and allowed to dry overnight.

Images in Figs. 2E and 3A were captured with a LSM 780 confocal microscope (Zeiss) with a 63× oil-immersed objective. Z-stack images were processed using Zeiss ZEN 3.3 software. Images in Fig. 3F and Supplemental Fig. 1A were taken with an ECLIPSE Ti2 microscope (Nikon) with a 60× oil-immersed or 40× dry objective, respectively, and NIS-Elements AR 5.21.02 software. Images in Fig. 3F were imported into Huygens Essential software (v21.4.0) and deconvolved using the “aggressive” profile in the Deconvolution Express application. Images in Supplemental Fig. 1E were captured with an Olympus FV3000 confocal laser scanning microscope and a 60× oil-immersed objective.

For quantification of B. burgdorferi associated with cultured MEF and HFF cells (Supplemental Fig. 1A), tiled images from infected cells at 24 h postinfection were taken with 40× objective. From each field, the number of cells with more than one B. burgdorferi present in the cell area, defined by positive CellMask staining, was annotated. B. burgdorferi–positive cells were divided by the total number of cells per field, defined by counting DAPI-positive nuclei, and multiplied by 100 to calculate the percentage of cells associated with B. burgdorferi. Four fields for each cell line 6 and 24 h postinfection were quantified from duplicate biological samples, and the average of each cell line across replicates and fields was plotted.

Bioluminescent images of mice were collected as previously described (70, 78). Groups of five WT, five cGAS^KO, and five STING^KO male mice 6–11 wk old were s.c. injected with 100 µl of 10⁵ ML23 pBBE22luc. Before imaging, 5 mg of d-luciferin (Goldbio, St. Louis, MO) was dissolved in PBS and administered to all except one mouse per group via i.p. injection and anesthetized with isoflurane for imaging. Mice were imaged at 1 h and 1, 4, 7, 10, 14, 21, and 28 d postinfection (dpi) using the Perkin Elmer IVIS Spectrum live imaging system. Bioluminescence from treated mice was normalized to the untreated mouse from each group. At 28 dpi, inguinal lymph nodes and skin flanks were collected and transferred to BSKII with 6% normal rabbit serum for outgrowth assays.

Samples of joints and hearts were collected from each mouse after euthanasia at 28 and 35 dpi, fixed by immersion in 10% neutral-buffered formalin at room temperature for 48 h, and stored in 70% ethanol before embedding in paraffin, sectioning at 5–6 µm, and staining with H&E by AML Laboratories (Jacksonville, FL). The tissue sections were examined using bright-field microscopy in a blinded manner by a board-certified anatomic veterinary pathologist and ordinally scored for the degree of mononuclear infiltration. A score of 1 represented minimal infiltration (<5 cells/400× field) and increased to 4 for abundant infiltration (>30 cells/400× field). Tissue sections were also scored for distribution of mononuclear inflammatory cells as focal (1), multifocal (2), or diffuse (3).

Statistical analysis was performed in GraphPad Prism (GraphPad Software, La Jolla, CA). Statistical significance was determined by p ≤ 0.05. Specific tests are detailed in the figure legends. Error bars in figures represent SEM based on the combined triplicate biological samples for all cell culture studies. All cell culture–based results (Figs. 1, 2, 3, Supplemental Figs. 1, 2) are representative of at least three independent experiments. Four biological replicates were used in in vivo bioluminescence imaging (Fig. 4) to determine statistical significance.

We first evaluated the ability of iBMDMs and MEFs derived from C57BL/6J mice to upregulate ISG and inflammatory cytokine transcripts when coincubated with viable or sonicated B. burgdorferi B31-A3 at an MOI of 20 (Fig. 1). Consistent with earlier reports (19, 60), both viable and sonicated B31-A3 were able to induce IFN-β (Ifnb1) and ISGs (Cxcl10 and guanylate-binding protein 2 [Gbp2]), as well as proinflammatory cytokine IL-6 (Il6) expression, in iBMDMs (Fig. 1A, 1B). Viable B. burgdorferi more potently induced ISGs and Il6, inducing expression levels 4- to 10-fold higher than sonicated bacteria (Table I). Likewise, we found that B. burgdorferi elicited ISGs and TNF-α (Tnfa) in primary MEFs, with live spirochetes triggering more robust responses compared with sonicated bacteria (Fig. 1C, 1D and Table II). Collectively, these data indicate that live B. burgdorferi engages robust IFN-I–associated ISG expression in both BMDMs and nonphagocytic MEFs, which lack a full repertoire of TLRs and other pattern recognition receptors (41).

Lipoprotein-rich B. burgdorferi predominately engage TLR2 but can also trigger other TLRs to induce inflammatory cytokines and IFNs (18, 19, 24, 36). However, a role for cytosolic DNA sensing in the innate immune response to B. burgdorferi remains unknown. To examine this directly, we cocultured primary BMDMs from wild-type (WT), cGAS^KO, or STING^KO mice on a C57BL/6J background with B. burgdorferi B31-A3 and assessed ISG and cytokine transcript induction by quantitative RT-PCR (qRT-PCR) (Fig. 2). We first confirmed that cGAS^KO BMDMs were hyporesponsive to immunostimulatory DNA (ISD) delivered into the cytosol by transfection (73). As expected, the induction of ISGs Cxcl10 and Gbp2 in cGAS^KO BMDMs was significantly reduced relative to WT macrophages (Fig. 2A). After 6 h of coculture with B31-A3, the expressions of ISGs Cxcl10 and Gbp2, as well as Ifnb1 transcripts, were significantly reduced in cGAS^KO BMDMs compared with WT (Fig. 2B). In contrast, the levels of Il-6 and Tnfa transcripts were similar in both WT and cGAS^−/− BMDMs (Fig. 2C). These results indicate that the cGAS pathway is not a robust inducer of proinflammatory genes in BMDMs exposed to B. burgdorferi and are consistent with the notion that TLR2 or other TLRs sense B. burgdorferi ligands to engage NF-κB–dependent cytokines. Similar results were obtained from protein analysis of cGAS^KO and STING^KO macrophages 9 h after incubation with B31-A3, with notable reductions in ISGs IFN-induced protein with tetratricopeptide repeats 1 (IFIT1) and Z-DNA binding protein 1 in both cGAS^KO and STING^KO BMDMs (Fig. 2D).

Finally, confocal imaging revealed that B. burgdorferi are internalized by BMDMs after a 3-h incubation (Fig. 2E). Ab staining against the outer surface protein A (OspA) of B. burgdorferi showed coiled and degraded DAPI-positive spirochetes in the macrophage cytoplasm, as well as punctate, DAPI-negative OspA staining, indicative of bacterial destruction. B. burgdorferi OspA also colocalized with the cytosolic autophagy marker p62/Sequestosome-1, suggesting spirochete degradation via macrophage autophagy and/or LC3-associated phagocytosis pathways (79–81). Collectively, these data indicate that B. burgdorferi are internalized by murine macrophages and trigger the cGAS–STING–IFN-I signaling axis.

To broaden our findings beyond murine macrophages, we exposed MEFs and telomerase immortalized HFFs to B. burgdorferi B31-A3. We observed that OspA-positive B. burgdorferi associated with ∼15% of MEFs (Supplemental Fig. 1A, 1B) and 33% of HFFs (Supplemental Fig. 1C, 1D) after coculture, consistent with a prior study (54). Confocal immunofluorescent imaging with Z-stack reconstitution revealed cell-associated B. burgdorferi in the same focal plane with mitochondria (Fig. 3A), suggestive of spirochete internalization. Additional imaging analysis revealed that coiled and degraded spirochetes strongly colocalized with the intracellular autophagy marker p62, further documenting that B. burgdorferi can access the fibroblast cytoplasm (Supplemental Fig. 1E). Similar to our results in MEFs, exposure of HFFs to live B. burgdorferi increased expression levels of ISGs (IFI44L, IFNB1) and the proinflammatory cytokine gene TNFA (Fig. 3B). Although the synthetic TLR2 ligand PAM3CSK4 (Pam3) induced TNFA levels similar to live B31-A3, we did not observe ISG induction, suggesting that B. burgdorferi lipoprotein engagement of TLR2 is not responsible for IFN-I responses in HFFs. Taken together, these results demonstrate that B. burgdorferi association and internalization within a minority of cultured cells is sufficient to induce robust IFN-I responses.

To next examine whether cGAS contributes to IFN-I responses in fibroblasts exposed to B. burgdorferi, we cocultured primary WT and cGAS^KO MEFs with B31-A3 and subjected them to qRT-PCR analysis for ISG and proinflammatory transcripts. We first confirmed that cGAS^KO MEFs were hyporesponsive to ISD delivered into the cytosol by transfection (Fig. 3C). After coculture with B. burgdorferi, we noted that the induction of ISG (Cxcl10 and Gbp2) and Ifnb1 transcripts was significantly reduced or entirely abrogated in cGAS^KO MEFs relative to WT controls (Fig. 3D). Although Il6 transcripts were similar between B. burgdorferi–exposed WT and cGAS^KO MEFs, Tnfa expression was significantly reduced in the absence of cGAS (Fig. 3E). Immunofluorescence microscopy revealed that B. burgdorferi associated with MEFs after a 6-h incubation, similar to results in BMDMs and HFFs (Fig. 3F). Ab staining against the OspA lipoprotein revealed intact, DAPI-positive spirochetes in the MEF cytoplasm, as well as punctate OspA foci that were DAPI negative. Consistent with a role for cGAS in sensing internalized B. burgdorferi DNA, we observed colocalization of HA-tagged cGAS with coiled spirochetes staining positive for both OspA and DAPI (Fig. 3F). To further document a requirement for the cGAS–STING pathway in the IFN-I response to B. burgdorferi, we employed RU.521, a specific cGAS inhibitor, and H-151, a specific STING inhibitor, to block the pathway during coculture. Both inhibitors were effective at reducing ISG transcripts (Cxcl10 and Gbp2) induced by transfection of ISD (Fig. 3G). MEFs exposed to cGAS and STING inhibitors exhibited reduced ISG expression relative to vehicle-treated MEFs, with no effects on Tnfa induction (Fig. 3H). Taken together, these data indicate that B. burgdorferi associates with fibroblasts in culture, leading to the cGAS-mediated sensing of borrelial DNA from lysed or damaged spirochetes.

Additional experiments in STING^KO MEFs revealed markedly reduced Gbp2 and Ifnb1 expression, but little change in proinflammatory cytokine transcripts, after challenge with B. burgdorferi (Supplemental Fig. 2A, 2B). MEFs deficient in the IFNAR also exhibited impaired expression of the ISG Gbp2, but not cytokine transcripts, suggesting that ISG induction in MEFs is dependent on autocrine and/or paracrine signaling via IFNAR. In contrast, B. burgdorferi infection of MAVS protein null MEFs (MAVS^KO), which cannot signal in response to intracellular dsRNA, triggered levels of Gbp2 and Ifnb1 mirroring WT cells (Supplemental Fig. 2A). This indicates that MEFs cocultured with B. burgdorferi respond to DNA, not RNA, ligands to induce IFN-I. Expressions of proinflammatory cytokine transcripts Il6 and Tnfa were unchanged or modestly altered in STING^KO, MAVS^KO, and IFNAR^KO MEFs, demonstrating all MEF lines remained responsive to B. burgdorferi lipoprotein engagement of TLR2 during coincubation (Supplemental Fig. 2B). Moreover, qRT-PCR to detect internal levels of borrelial flagellar gene, flaB, within host cells revealed roughly equivalent levels of flaB among WT and mutant MEF lines (Supplemental Fig. 2C). This indicates that altered ISG expression in STING- and IFNAR-deficient cells is not due to changes in the ability of viable B. burgdorferi to associate with these MEFs.

To determine roles for the cGAS–STING pathway in borrelial dissemination, tissue colonization, and inflammation during mammalian infection, we monitored bioluminescent B. burgdorferi, ML23 pBBE22luc, in real time by in vivo imaging in C57BL/6 WT, cGAS^KO, and STING^KO mice (Fig. 4). Mice were injected with d-luciferin before imaging, and one mouse in each group was not injected to serve as a bioluminescence background control (Fig. 4A). The absence of cGAS or STING did not alter the kinetic dissemination of B. burgdorferi or the borrelial load as observed in the images and by quantitative analysis of bioluminescence emission (Fig. 4A, 4B). Outgrowth from infected tissues confirmed that all genotypes were colonized with viable B. burgdorferi 28 d postinfection (Fig. 4C). To investigate whether the cGAS–STING pathway impacts the development of arthritic phenotypes, we collected tibiotarsal joints from WT and cGAS^KO mice for histopathology. Consistent with prior reports (82), B. burgdorferi induced mild arthritis in WT C57BL/6J mice at 28 d postinfection (Supplemental Fig. 3A). Interestingly, H&E staining revealed that the joints of cGAS^KO mice exhibited reduced synovial papillary hyperplasia, immune cell infiltration, and overall joint pathology scores, suggesting a trend toward reduced inflammation at the 28-d time point (Supplemental Fig. 3B). These results suggest that the cGAS–STING pathway may contribute to the development of inflammation during mammalian infection without impacting the ability of B. burgdorferi to readily disseminate and colonize secondary tissues.

B. burgdorferi elicits robust innate and adaptive immune responses that involve induction of both IFN-I (IFN-α and β) and IFN-II (IFN-γ) cytokines that drive expression of an overlapping family of ISGs (12, 31, 35, 83, 84). Synovial tissue from patients with postinfection Lyme arthritis expressed ISGs that are associated with both type I and type II IFNs (25, 35, 85). Ab-mediated IFNAR blockade results in a significant reduction in joint inflammation in mice (12), and the inhibition of IFN-γ delays sustained ISG expression and inflammation during B. burgdorferi infection (12, 86). A number of prior studies have focused on identifying the signaling mechanisms responsible for IFN-I and ISGs elicited by B. burgdorferi. Both in vitro and in vivo studies have demonstrated that TLR adaptors MyD88 and TRIF are not the primary contributors to IFN-I induction (12, 13, 22, 87–90). Specifically, in the absence of TLR2 or TLR5, innate immune receptors that sense lipoproteins or flagella of degraded B. burgdorferi, respectively, the induction of ISGs and joint swelling is similar to WT mice (12, 87–89). Studies using human cells have implicated nucleic acid–sensing, endosomal localized TLRs (TLR7, 8, and 9) as regulators of IFN-I induction in human PBMCs challenged with B. burgdorferi ex vivo (13, 18, 19, 90). However, TLR9 inhibition does not impair ISG induction in BMDMs exposed to B. burgdorferi, in contrast with PBMCs (19, 22, 85). Thus, it is likely that multiple innate immune pathways are responsible for IFN-I induction during B. burgdorferi infection. Differences in IFN-I and inflammatory responses across independent studies might also be explained by the use of distinct borrelial strains that vary in invasion and inflammation (11, 91, 92).

Fibroblast and endothelial cells, which do not express a full complement of TLRs, are necessary for B. burgdorferi–induced IFN-I responses in joints (25, 41). Thus, TLR detection of B. burgdorferi nucleic acids is likely not the predominant pathway governing IFN-I responses during B. burgdorferi infection in the mammalian host. Therefore, we hypothesized that B. burgdorferi might engage an intracellular innate immune pathway leading to IFN-I. Using primary fibroblasts and macrophages from a panel of C57BL/6J mice lacking intracellular nucleic acid sensors or downstream adaptors, we find that the DNA-sensing, cGAS–STING pathway is critically required for robust induction of ISGs after exposure to live B. burgdorferi. This finding was confirmed through small molecule inhibitor studies, because we observed reduced ISG responses to B. burgdorferi when cGAS or STING was inhibited in MEFs. In contrast, we show that deficiency of the main RIG-I–like receptor adaptor MAVS does not impact expression of ISGs or proinflammatory cytokine transcripts in fibroblasts, indicating that intracellular sensing of B. burgdorferi RNA is not a major contributor to IFN-I responses in cultured fibroblasts.

The phenomenon of host cell internalization by B. burgdorferi has not been observed during experimental infection in vivo; however, prior studies have shown that B. burgdorferi can become internalized into both phagocytic (macrophages, monocytes, and dendritic cells) and nonphagocytic (endothelial, fibroblast, and neuroglial cells) cells in culture (54, 57–62, 93–95). Our immunofluorescence microscopy analyses revealed borrelial OspA staining associated with a proportion of cultured mouse and human fibroblasts after short incubations with B. burgdorferi B31-A3. In both macrophages and fibroblasts, we observed OspA colocalization with the intracellular autophagy adaptor p62, which substantiates the notion that spirochetes interact with mammalian cells and can become internalized into the cytoplasm. Recent work has revealed that the autophagy machinery regulates the cGAS–STING pathway, because p62 is found in close proximity to TBK1, a key kinase necessary for IFN-I induction by STING (96). In addition, phosphorylation of p62 by TBK1 is important for the negative regulation and turnover of STING (97). Activation of cGAS by the obligate intracellular pathogen Mycobacterium tuberculosis can drive selective autophagy for clearance of intracellular bacteria (45). Interestingly, we observed coiled and degraded B. burgdorferi in BMDMs and MEFs colocalizing with p62, suggesting that the autophagy could be a mechanism to turn over internalized spirochetes. Autophagy has also been linked to inflammatory cytokine responses during B. burgdorferi infection in vitro (79). Therefore, internalization of B. burgdorferi and autophagic targeting may serve to fully engage intracellular signaling leading to IFN-I or other innate immune responses.

The cGAS–STING pathway is a major inducer of IFN-I in response to both exogenous, pathogen-derived DNA and host mtDNA and nuclear DNA (45, 51, 52, 72, 98–100). cGAS nondiscriminately binds dsDNA and produces 2′3′-cyclic guanosine monophosphate-adenosine monophosphate that binds and activates STING on the endoplasmic reticulum. This leads to recruitment and phosphorylation of the kinase TBK1, which phosphorylates and induces nuclear translocation of transcription factors IRF3 or IRF7 (101). Miller et al. (12) demonstrated that B. burgdorferi engages IFN-I response genes in an IRF3-dependent manner, further supporting that the signaling machinery downstream of cGAS–STING and other intracellular nucleic acid sensors is required for ISG induction. Multiple intracellular bacterial pathogens, including Mycobacterium tuberculosis, Listeria monocytogenes, and Chlamydia trachomatis, engage the cGAS–STING axis to induce IFN-I and influence pathogenesis (45, 50, 102–105). More recently, extracellular bacterial pathogens Pseudomonas aeruginosa and group B Streptococcus have been shown to trigger IFN-I responses through cGAS–STING in macrophages and dendritic cells (52, 106). The mechanisms by which B. burgdorferi and other extracellular bacteria engage cGAS remain unclear. However, we have observed recruitment and colocalization of cGAS with intracellular borrelia in MEFs, suggesting that internalized spirochetes may shed DNA or undergo host-induced membrane damage that liberates bacterial genomic material for detection by cGAS. Bacterial pathogens can also cause host cell stress, resulting in the release and accumulation of nuclear or mtDNA (107, 108). Mitochondrial stress and mtDNA release, in particular, is linked to innate immune responses through the cGAS–STING axis and is associated with autoimmune disorders (i.e., lupus and rheumatoid arthritis) and other conditions characterized by elevated IFN-I (98). Thus, it is possible that the IFN-I response induced by B. burgdorferi is at least partially dependent on host mtDNA or nuclear DNA sensing. It is likely that B. burgdorferi engages distinct mechanisms in phagocytic and nonphagocytic cells to elicit IFN-I responses via the cGAS–STING pathway, and future work is required to reveal the molecular mechanisms involved.

An IFN-I response occurs early in mammalian infection, resulting in the development of inflammation and arthritis that persist in part because of IFN-γ during later stages of disease (32, 35, 83, 84). We therefore investigated how absence of cGAS and STING impacted the kinetics of borrelial infection in mice using in vivo imaging to track bioluminescent B. burgdorferi in real time through the progression of disease (70, 109). Bioluminescent B. burgdorferi was able to successfully disseminate and colonize secondary tissues independently of cGAS or STING. Joint inflammation also showed a moderate, although not statistically significant, reduction at 28 dpi in cGAS^KO mice, suggesting that cGAS–STING signaling to IFN-I or other inflammatory response pathways might contribute to arthritic phenotypes during B. burgdorferi infection. It is important to note that the absence of cGAS did not eliminate inflammation entirely, indicating that B. burgdorferi–associated arthritis is multifactorial and involves additional innate and adaptive immune pathways as have been reported by others (13, 17, 19, 20, 90).

Although this study is impactful in that it is, to our knowledge, the first to link the cGAS–STING pathway to B. burgdorferi–induced IFN-I responses, it is not without limitations. C3H is the preferred model for borrelial infection because this background develops inflammation and pathologic disease at a significantly higher level than C57BL/6 or BALB/c mice (34, 82). An elegant study identified the Bbaa1 locus in C3H mice is responsible for IFN-I responses observed during borrelial infection and showed that introduction of this locus into C57BL/6 mice yielded similar inflammatory responses to C3H (29). Our study employed WT, cGAS, and STING KO strains on a pure C57BL/6 background because they are commercially available and well characterized. We employed highly invasive and inflammatory RST1 B. burgdorferi B31-A3 for our studies and were able to observe IFN-I induction even in cells from the C57BL/6 background. These responses were markedly attenuated or lost when using sonicated B. burgdorferi B31-A3. Other studies have revealed weak or absent ISG expression during in vitro challenge with B. burgdorferi, although most used sonicated bacteria or less inflammatory RST3 strains (12, 22, 82, 87). This indicates that the C57BL/6 background is not devoid of IFN-I responsiveness to borrelial infection. Our work and previous studies show C57BL/6 mice and cell lines do produce measurable IFN-I and inflammatory responses to borrelial infection, and therefore we propose it is an appropriate model for this initial characterization of cGAS–STING in the innate immune response to B. burgdorferi (12). However, the development of cGAS or STING KO lines on the C3H background will be necessary to fully characterize the role of this innate immune pathway in B. burgdorferi infection dynamics, tissue-specific inflammation, and arthritic phenotypes in vivo.

In conclusion, to our knowledge, our study is the first to show that B. burgdorferi triggers the intracellular cGAS–STING DNA-sensing pathway to shape IFN-I responses in cultured cells. Future studies are needed to investigate the how borrelial cells initiate IFN-I responses through this pathway by characterizing the source of DNA that binds to cGAS (i.e., bacterial or host) and determining whether borrelial internalization is required. Loss of cGAS or STING does not appear to alter B. burgdorferi dissemination or its ability to colonize secondary tissues, but studies in C3H mice lacking this innate immune pathway may reveal differential bacterial kinetics and/or inflammatory responses that do impact the course of infection. Additional work to clarify these open questions may support the development cGAS- and/or STING-based immunotherapeutics that may be effective against active infection or persistent symptoms of B. burgdorferi, such as Lyme arthritis.

The authors have no financial conflicts of interest.

We thank members of the Hyde and West laboratories for helpful discussions. We thank Dr. Jon Skare for providing the OspA Ab. We thank Yuanjiu Lei for assistance preparing the visual abstract.