Cyclic-G/AMP (cGAMP) synthase (cGAS) triggers host innate immune responses against cytosolic double-stranded (ds)DNA arising from genotoxic stress and pathogen invasion. The canonical activation mechanism of cGAS entails dsDNA-binding and dimerization. Here, we report an unexpected activation mechanism of cGAS in which Mn2+ activates monomeric cGAS without dsDNA. Importantly, the Mn2+-mediated activation positively couples with dsDNA-dependent activation in a concerted manner. Moreover, the positive coupling between Mn2+ and dsDNA length-dependent activation requires the cognate ATP/GTP substrate pair, while negative-cooperativity suppresses Mn2+ utilization by either ATP or GTP alone. Additionally, while Mn2+ accelerates the overall catalytic activity, dsDNA length-dependent dimerization specifically accelerates the cyclization of cGAMP. Together, we demonstrate how the intrinsic allostery of cGAS efficiently yet precisely tunes its activity.
cGAS is crucial for the host defense against various maladies that give rise to cytosolic dsDNA (e.g. pathogen invasion, genotoxic stress, and dysfunctional mitochondria) (1,2). Upon binding and dimerizing on cytosolic dsDNA, cGAS cyclizes ATP and GTP into cGAMP, a second messenger that activates type-I interferon (IFN-I)-mediated innate immune responses via Stimulator of Interferon Genes (STING) (2,3). cGAS is essential for the host defense against numerous pathogens (1,4,5), regulates tumorigenesis (6), and is involved in autoimmunity (7–10).
It is increasingly appreciated that cGAS dynamically tunes the intensity and duration of the IFN-I activity according to the severity of stress, consequently eliciting context-dependent innate immune responses (e.g. autophagy, cellular senescence, and apoptosis) (1,2,11–14). It was recently identified that the length of cytosolic dsDNA is a key measure for cGAS to gauge stress levels (15,16): long stretches of naked dsDNA (≥ 50 base-pairs (bps)) signify major maladies such as the presence of viral genomes or mislocalized mitochondrial dsDNA, while shorter fragments indicate minor DNA damage or degraded viral genomes (15,16). How does dsDNA length regulate cGAS activity? Resting cGAS is an inactive monomer and dimerization is required for activation in addition to dsDNA binding (17,18). dsDNA length dictates the distribution of dsDNA-bound cGAS monomers and dimers, thereby grading the signaling activity (19). Importantly, such an allosteric mechanism allows dsDNA length-dependence to be conditional: substrate binding and elevated cGAS concentrations also induce dimerization, thereby reducing the dependence on dsDNA length (19).
Still, major questions remain regarding how cGAS is regulated. First, it was recently reported that cytosolic Mn2+ originating from damaged organelles potentiates the dsDNA sensing activity of cGAS (20). In fact, Mn2+ was crucial for cGAS-mediated host innate immune responses against DNA viruses (20). Nevertheless, how Mn2+ regulates GAS activity remains unknown. Second, little is known about how cGAS retains substrate specificity and signal fidelity (3,21,22). The catalytic core of cGAS contains a polymerase (pol)-β-type nucleotidyl-transferase (NTase) domain, which is inherently promiscuous (23–25). Although ATP/GTP (1ATP:1GTP) is the cognate substrate pair, cGAS can bind and perform chemistry on non-cognate substrate pairs (e.g. ATP/ATP). Such ‘off-pathway’ interactions can be deleterious by not only competing against cognate substrates, but also generating dinucleotides that do not activate STING (3,21,22,26). Moreover, this issue can also be exacerbated by multiple factors: (a) ATP outnumbers GTP by at least 4-fold in vivo (27) and both NTPs, especially ATP, are present at saturating concentrations for cGAS ([ATP] and [GTP] > KM for cGAS) (8,19,22,28). Thus, noncognate ATP/ATP is likely the most abundant substrate pair for cGAS. How cGAS overcomes this issue and generates cGAMP in high fidelity remains unknown. (b) Although Mn2+ sometimes enhances the catalytic activity of other cGAS-like NTases by substituting for Mg2+, its primary effect is increased substrate promiscuity (23–25). In fact, Mn2+ is considered mutagenic for pol-β-type NTases as it preferentially enhances the affinity for noncognate NTPs (23–25). Currently, how cGAS could compensate for the potentially deleterious effect of Mn2+ remains unknown. (c) Human cGAS is a poor cyclase (3,8,22,28). cGAS first generates 2′-GTP to 5′-ATP-linked linear pppGpA from ATP/GTP, which dissociates and rebinds to be cyclized into 2′-5′ 3′-5 linked cGAMP (3,22) (simply ‘cGAMP’ hereafter). Because of low cyclization efficiency, the pppGpA intermediate accumulates much more rapidly than cGAMP (3,22). This could impinge on promptly mounting host responses, as the intermediate cannot activate STING or even be degraded rapidly. Again, how cGAS might address this issue remains unknown. Here, we unravel an unexpected role of allostery in resolving all these compounding issues.
We find that Mn2+ activates human cGAS without dsDNA and also positively couples with dsDNA-dependent activation, allowing Mn2+ to dampen dsDNA length-dependence in activating cGAS both in vitro and in macrophages. cGAS preferentially utilizes Mn2+ over Mg2+ for its NTase activity, and Mg2+ facilitates Mn2+ utilization instead of competing against it. Moreover, physiological amounts of Mn2+ can activate cGAS mutants defective in dsDNA/dimerization-dependent activation both in vitro and in cells, indicating that Mn2+ activates monomeric cGAS. The NTase activity with cognate ATP/GTP vs. noncognate ATP/ATP or GTP/GTP is similar when cGAS is bound to short dsDNA; however, longer dsDNA increases the activity only toward ATP/GTP. Likewise, although Mn2+ accelerates the NTase activity of cGAS toward noncognate substrates, ATP/GTP benefits most. Furthermore, ATP/GTP is required for the synergy between Mn2+ and dsDNA binding, while dsDNA binding invokes negative-cooperativity between Mn2+ and noncognate substrates. Finally, Mn2+ accelerates the overall catalytic activity of cGAS; however, dsDNA length-dependent dimerization specifically accelerates the rate-determining cyclization, minimizing the formation of off-pathway products even with excess ATP. Together, we demonstrate how allostery linking Mn2+- and dsDNA-binding maximizes the efficiency and activity of cGAS.
Human cGAS (encoding 2–522, cGASFL or 157–522, cGAScat) was cloned into a pET28b vector (Novagen) in-frame with an N-terminal 6xHis-MBP-tag and a TEV protease cleavage site. Plasmids encoding wild-type and mutant cGAS constructs were transformed in to E. coli BL21-Rosetta 2 DE3 cells and expressed via IPTG as described in ref. (19). Recombinant cGAS was then purified as described in (19), concentrated, and stored in −80°C.
dsDNAs longer than 90-bp were PCR amplified from the human DDX41 gene and agarose gel purified. dsDNAs shorter than 90-bp were purchased from Integrated DNA Technology. ATP and GTP were hydrated separately in 20 mM HEPES buffer and adjusted to pH 7.4 with NaOH.
All experiments were performed at least three times as described in before (19,29). Shown Figures are either representative of these independent experiments or averages with standard deviations (SD). The fits to data (the Hill form of the Michaelis-Menten equation or the binding isotherm) were generated using Kaleidagraph (Synergy Software). All reactions were performed under 25 mM Tris acetate pH 7.4, 125 mM potassium acetate pH 7.4, 0.5 mM TCEP, 5% glycerol and indicated concentrations of MgCl2 and/or MnCl2 at 25 ± 2°C. cGAS concentrations are noted in Figure Legends. To avoid titrating Mn2+ against free NTPs and to mimic the in vivo environment where Mg2+ is present in large excess (20,30), each NTP was pre-mixed with 1 molar equivalent of MgCl2, with the exception of those experiments that were conducted in the complete absence of MgCl2, in which case the NTPs were pre-mixed with equimolar MnCl2 instead. We used as low as 6 μM MnCl2 in our experiments to reflect the physiological amount (≤ 53 μM; (20,31,32)).
40 μl reactions containing cGAS, NTPs, dsDNA (where indicated), MgCl2 and/or MnCl2 were incubated in reaction buffer for the indicated time. Reactions were quenched with 25 mM EDTA, diluted to 80 μl with water and filtered with 3 or 10 KDCO spin columns (Amicon). An aliquot of the flow-through was mixed 1:1 with HPLC buffer A (see below) and fractionated. HPLC fractionation was performed on the Agilent Technologies 120 Infinity II system using a Poroshell EC-C18 column (2.7μm; 4.6 × 100mm). Reaction products were fractionated by the following scheme: 0% B from 0–1min, linear increase to 50% B from 1–27 min, linear increase to 100% B from 27–28 min. Buffer A: 100 mM potassium phosphate monobasic, 5 mM TBA, final pH 6.0; Buffer B: 100 mM potassium phosphate monobasic, 5 mM TBA, final pH 6.0, 30% acetonitrile. Buffers were filtered and degassed. Integrated peak intensities were calculated by the Agilent Technologies ChemStation software. Synthetic and purified compounds were used as standards for peak identification.
THP1-Dual cells (InvivoGen) were passaged and cultured in RPMI 1640 media containing 2 mM L-glutamine, 25 mM HEPES at pH 7.4, 10% heat-inactivated fetal bovine serum (30 min at 56°C), 100 μg/ml Normocin™, Pen-Strep (100 U/ml-100 μg/ml), Blasticidin (10ug/mL) and Zeocin (100ug/ml). HEK293T/17 cells (ATCC) were passaged and cultured in DMEM media containing 2 mM L-glutamine, 1 mM sodium pyruvate and 10% heat-inactivated fetal bovine serum (30 min at 56°C).
THP1-Dual and THP1-Dual cGAS-KO reporter monocytes were differentiated into macrophages with 50 nM PMA for 16 hours at 37°C and 5% CO2. The PMA containing media was replaced by fresh media and cells were incubated an additional 48 hours at 37°C 5% CO2. The media was again replaced with fresh media or media plus MgCl2 or MnCl2 and incubated for 2 hrs at 37°C and 5% CO2. Cells were then transfected with various lengths of dsDNA (1.67 μg/ml final) using Lipofectamine LTX containing OptiMEM media. Cells were incubated an additional 24 hours at 37°C and 5% CO2 before supernatants were collected and Lucia production was determined using the Luciferase Reporter Assay System (Promega).
HEK293T cells were pre-incubated for 2 hours at 37°C and 5% CO2 with growth media, or media supplemented with 50 μM of MnCl2 or MgCl2. 10 or 50 ng of pCMV plasmids encoding empty vector or various cGASFL variants were then transfected using Lipofectamine LTX (Invitrogen), along with 5 ng of Renilla Luciferase plasmid, 10 ng of plasmid encoding human STING, and 10 ng of plasmid encoding Firefly Luciferase under an IFN-Iβ promoter. Transfected cells were incubated overnight at 37°C and 5% CO2, washed once with PBS and lysed. Lysates were analyzed for Firefly Luciferase and Renilla Luciferase activities independently using the Dual-Luciferase Reporter Assay System (Promega). Data shown is normalized as the ratio of Firefly output divided by log10 Renilla output of individual wells.
dsDNA length-dependent dimerization is crucial for cGAS activation (15–19). It was reported that cytosolic Mn2+ potentiates the dsDNA-sensing activity of cGAS through an unknown mechanism (20). To explore whether Mn2+ modulates dsDNA length-dependent activation, we preincubated THP1-Dual reporter cells with Mn2+ or Mg2+, and transfected dsDNA of various lengths (Figure 1A). Consistent with previous studies (15,16), we observed step-like dsDNA length-dependent luciferase activity without Mn2+ (i.e. no significant activity for dsDNA < 33 bps; Figure 1A). With Mn2+, dsDNA fragments shorter than 39-bps produced significantly higher IFN-I activity, dampening duplex length-dependence (Figure 1A). By contrast, THP1-Dual cGAS-KO cells did not show increased IFN-1 activity upon transfecting dsDNA (Supplementary Figure S1A), indicating that our observations were specific for cGAS.
While investigating the underlying mechanism, we found that Mn2+ alone does not enhance the dsDNA binding activity of human full-length cGAS (cGASFL; Supplementary Figure S1B–D and Supplementary Table S1), which prompted us to test whether Mn2+ regulates the catalytic activity. Using our PPiase-coupled NTase assay (19), we found that Mn2+ drastically increased the maximal dsDNA-induced NTase activity of cGASFL (kmax) independent of duplex length (Figure 1B), revealing that Mn2+ diminishes dsDNA length-dependence by directly enhancing the catalytic activity. Next, we tested whether Mn2+ enhances the KM or the maximal turnover (kcat) in the presence of saturating amounts of dsDNA with various lengths (Figure 1C). For all dsDNA lengths, KMs improved moderately, but kcats were consistently and significantly higher (Figure 1C, Supplementary Table S1), indicating that the effect of Mn2+ arises from catalysis. Notably, the Hill constants were smaller in the presence of Mn2+ (Hill: ∼2 versus ∼1; Figure 1C, Supplementary Table S2), suggesting positive cooperativity between Mn2+ and dsDNA-binding in activating cGAS; see (33,34) for reference. To explore this unexpected mechanism, we determined whether Mn2+ enhances the apparent affinity of dsDNA (Kact) in the presence of ATP/GTP. Here, we used the catalytic domain (cGAScat) for accurate measurements as it binds dsDNA more weakly than cGASFL (19). Indeed, cGAScat showed enhanced kmaxs and Kacts for both 19- and 72-bp dsDNA with Mn2+ (Figure 1D, Supplementary Table S3). Together, we concluded that Mn2+ not only increases the catalytic activity of cGAS, but also allosterically enhances its dsDNA binding activity in conjunction with ATP/GTP.
We noted that the effect of Mn2+ was evident despite excess Mg2+ in our assays (Figure 1B–D), suggesting that cGAS preferentially utilizes Mn2+ over Mg2+. To test this idea, we determined the apparent affinity (EC50) of Mg2+ and Mn2+ in the presence of 339-bp dsDNA (EC50: concentration required for 50% activity). The EC50 of Mn2+ (∼40 μM, Figure 1E) was ∼100-fold tighter compared to Mg2+ (∼4 mM, Supplementary Figure S1E), and this preference was maintained across different dsDNA lengths (Figure 1E; Supplementary Tables S4 and S5). Moreover, in contrast to the Hill constant of 1.2 for Mg2+, the Hill constants were ∼2 for Mn2+ (Figure 1E versus Figure S1E; see also Supplementary Tables S4 and S5), indicating Mn2+ can substitute for Mg2+ in a positively cooperative manner. In the presence of 5 mM Mg2+ (Figure 1F), the EC50 of Mn2+ improved 2- to 3-fold and the Hill constants also decreased to ∼1, albeit the kmax decreased ∼ 2-fold (Supplementary Table S5). Considering that cGAS utilizes two divalent metals at the active site (3), our observations suggest that occupying one metal binding site by Mg2+ increases the affinity of Mn2+ for the second site. The decreased kmax is also consistent with this idea, as the catalytic activity is supported by both Mg2+ and Mn2+. Collectively, our observations suggest that free Mg2+ in the cytoplasm (∼1 mM (35)) would facilitate Mn2+ (≤50 μM (20,31,32)) utilization rather than compete against it (e.g. at 10 μM Mn2+, the activity of cGAS with 72-bp dsDNA would be ∼10-fold higher in the presence of Mg2+).
The NTase activity of cGASFL without dsDNA was notably higher with Mn2+ (Figure 1B, dark blue), raising the possibility that Mn2+ alone activates the enzyme. Indeed, cGASFL showed robust dsDNA-free NTase activity with increasing Mn2+ (Figure 2A). The maximal activation without dsDNA was comparable to the dsDNA-induced activity (Figure 2A, Supplementary Table S6). The steady-state kinetic parameters of cGASFL with 5 mM Mn2+ in the absence of Mg2+ and dsDNA were also similar to those obtained in their presence (Figures 2B versus 1C; Supplementary Table S7). Furthermore, Mn2+-mediated, dsDNA-independent activation still occurred with excess Mg2+ (Figure 2C, Supplementary Table S6); 5 mM Mg2+ decreased the kmax by 2-fold and reduced the Hill constant, similar to when dsDNA was present (Figure 2A versus C, Supplementary Table S6). These observations consistently indicated that Mn2+ activates cGAS even without dsDNA. However, cGAS bound Mn2+ as much as 85-fold less tightly in the absence of dsDNA (EC50: 18 versus 1530 μM; Figure 2A and C; Supplementary Tables S5 and S6). Given that Mn2+ plus ATP/GTP enhances the dsDNA binding affinity of cGAS (Figure 1D), our results indicated that dsDNA-binding couples with Mn2+-utilization in a concerted manner. Collectively, we reveal an unexpected Mn2+-dependent, dsDNA-independent activation mechanism of cGAS. Moreover, our results indicate that dsDNA binding is crucial not only for allowing Mn2+ to be utilized at physiological concentrations, but also for synergizing with Mg2+.
We found previously that ATP/GTP binds cGAS even without dsDNA and modestly induces dimerization (19). Considering the positive relationship between dsDNA and Mn2+, we asked whether Mn2+ also promote substrate-mediated dimerization more efficiently than Mg2+. We thus monitored the changes in Förster resonance energy transfer (FRET) ratios between donor- and acceptor-labeled cGASFL populations with increasing Apcpp/Gpcpp (non-hydrolyzable analogues that prevent turnover (19)). Apcpp/Gpcpp produced higher FRET signals with Mn2+, suggesting that substrate-mediated dimerization is enhanced (Figure 2D). Nevertheless, FRET ratios did not change as robustly as those induced by 72-bp dsDNA (Figure 2D versus Supplementary Figure S2A), indicating that the dimeric fraction was still marginal. Moreover, unlike the NTase activity assays, FRET ratios did not change with increasing Mn2+ when Apcpp/Gpcpp was fixed at a low concentration (Figure 2E; 200 μM, equivalent to [ATP/GTP] used in NTase assays in Figure 2A and C). These observations suggest that dimerization at this substrate concentration is minimal even with Mn2+. Adding Mn2+ and/or Apcpp/Gpcpp did not alter dsDNA-induced FRET ratios, again indicating that dimerization is dictated by dsDNA (Supplementary Figure S2A). Nevertheless, the Mn2+-induced NTase activity of cGAS without dsDNA was as high as those induced by long dsDNA (Figure 2A–C), suggesting that Mn2+ activates monomeric cGAS (each monomer has an intact active-site (3,17,18,36,37)).
To test whether Mn2+ activates monomeric cGAS, we monitored the NTase activity of a mutant that cannot dimerize, K394E (Figure 2F, Figure S2B) (19). Both K394E-cGASFL and K394E-cGAScat displayed robust NTase activities with Mn2+ without dsDNA (Figure 2G, Supplementary Figure S2C–F), supporting the mechanism by which Mn2+ activates monomeric cGAS. Of note, cGAS has a Zn2+-finger adjacent to the dimer interface (Figure 2F), and mutating two coordinating Cys residues abrogates the signaling activity of cGAS in cells (17,36,37). Thus, to disrupt the dimerization in a different manner, we removed Zn2+ via EDTA or by mutation (C396A/C397A (36), Figure 2F). Unlike wild-type (WT) cGASFL that behaved a mixture of monomer and dimer in size-exclusion chromatography (SEC), both variants eluted as monomers (Supplementary Figure S2B), corroborating that the Zn2+-finger is important for dimerization. Both EDTA-treated and C396A/C397A-cGASFL variants bound dsDNA (Figure S2G); however, dsDNA failed to activate either variant with Mg2+ (Figure 2G). Nevertheless, both mutants showed robust NTase activities with Mn2+ in the absence of dsDNA or Mg2+ (Figure 2G). Our observations not only support the mechanism by which Mn2+ alone activates monomeric cGAS, but also indicate that Mn2+ does not substitute for Zn2+. Next, to explore biological relevance, we monitored the signaling activity of cGASFL variants with or without Mn2+ in HEK293T cells. A physiological amount of Mn2+ substantially increased the WT cGASFL-STING induced IFN-β activity (Figure 2I). Mn2+ also induced IFN-β activity from K394E- and C396A/C397A-cGAS when they were sufficiently overexpressed (Figure 2I versus J); the dead variant showed no IFN-β activity in all conditions (Figure 2I-J). Together, these results suggest that Mn2+ activates monomeric cGAS.
Dimerization is critical for the dsDNA-mediated activation of cGAS (15–17,19). Because our results suggest that Mn2+ activates monomeric cGAS, we then asked whether Mn2+ could restore dsDNA-mediated activation of cGAS mutants defective in dimerization (Figure 2F and G). Here, we determined the NTase activity of cGASFL mutants using a sub-optimal amount of Mn2+ for dsDNA-free activation of cGASFL (Figure 2H). As expected, all cGASFL variants, except for the catalytically dead, showed moderate Mn2+-induced NTase activity without dsDNA (Figure 2H light purple). Strikingly, dsDNA further stimulated the Mn2+-induced NTase activities of K394E, EDTA-treated, and C396A/397A variants (Figure 2H dark purple), which suggested that the positive coupling between Mn2+ and dsDNA occurs within a monomer (i.e. in cis). Next, to further test the cis coupling mechanism, we mutated two side-chains at the hinge of the ‘spine helix’ that changes conformation upon dsDNA binding (K173E/R176E (36), Figure 2F, Supplementary Figure S2H); the dsDNA-induced conformation change of the spine helix restructures the ‘activation loop’ in cis (3,18,36) (Supplementary Figure S2H). K173E/R176E-cGASFL migrated mostly as a dimer in SEC and bound dsDNA (Supplementary Figures S2B and G). 5 mM Mn2+ also activated K173E/R176E-cGASFL without dsDNA and Mg2+ (Figure 2G), but dsDNA failed to further activate this cGASFL mutant even with sub-saturating Mn2+ (Figure 2H). Surprisingly, however, K173E/R176E-cGAS showed robust IFN-β activity when overexpressed in HEK293T cells (Figure 2J). Considering that dsDNA does not activate this mutant (Figure 2H), our observation suggests that even the marginal catalytic activity of cGAS induced by Mn2+ alone could be sufficient for robust IFN-β signaling in vivo. Overall, our observations consistently indicate that the positive cooperativity between Mn2+ and dsDNA occurs within the cGAS monomer.
Although ATP/GTP is the cognate substrate pair, the active-site of cGAS is inherently promiscuous; cGAS can create linear dinucleotides using either ATP or GTP alone (pppApA and pppGpG, respectively) (3,21,22). This could be problematic in vivo, as ATP outnumbers GTP and neither pppApA nor pppGpG activates STING (3,21,22,26). Moreover, Mn2+ could exacerbate this issue, as it increases the substrate promiscuity of cGAS-like NTases (23–25). Nonetheless, cGAS can sustain the fidelity and efficiency of its signaling pathway. To understand the underlying mechanism, we first investigated how efficiently cGAS utilizes either ATP/ATP or GTP/GTP with Mg2+ and varying lengths of dsDNA using our NTase assay. With 19-bp dsDNA, both kcats and KMs for noncognate substrates were ∼2-fold worse than those from ATP/GTP, resulting in ∼3-fold lower kcat/KMs (Figure 3A, B and Supplementary Table S8; see Supplementary Figure S3A for GTP/GTP). Our observations suggest that cGAS produces cGAMP and off-pathway dinucleotides with similar efficiency when bound to short dsDNA. Strikingly, and in contrast to ATP/GTP, longer dsDNA failed to enhance either kcat or KM against ATP/ATP or GTP/GTP, resulting in an up to 12-fold difference in kcat/KM between cognate and noncognate substrates (Figure 3A and B, Supplementary Table S8). These results suggest an unexpected role of dsDNA length in governing the substrate specific catalytic activity of cGAS.
Mn2+-induced dsDNA-free NTase activity of cGASFL toward ATP/ATP or GTP/GTP was ∼3-fold higher compared those induced by dsDNA and Mg2+ (Figure 3C versus Figure 3A and Supplementary Figure S3A; see also Supplementary Table S8). Nevertheless, ATP/GTP still produced the highest Mn2+-induced kcat (Figure 3C), which suggested that Mn2+ preserves the substrate preference of cGAS unlike several other related NTases (23–25). We then investigated whether ATP/ATP or GTP/GTP allows Mn2+ to synergize with dimerization/dsDNA-mediated activation. Here, we found that increasing Apcpp or Gpcpp alone boosted the FRET ratios between two labeled cGASFL populations (Figure 3D), suggesting that binding of any NTP at the active-site promotes dimerization. However, Mn2+ did not further increase the FRET ratio for either NTP alone (Figure 3D versus 2D), suggesting that noncognate substrates do not allow Mn2+ to promote dimerization. Next, we asked how well cGAS utilizes Mn2+ in its NTase activity against noncognate pairs. Without dsDNA or excess Mg2+, the EC50 of Mn2+ against ATP/ATP or GTP/GTP was 2-fold weaker compared to ATP/GTP, but the Hill constants were still near two (Figures 2A versus 3E, F, Supplementary Figure S3B). Adding 5 mM Mg2+ in the absence of dsDNA moderately enhanced the EC50 of Mn2+ and reduced the Hill constant to near one (Supplementary Figure S3B-C and Table S9). However, with 339-bp dsDNA, the EC50 of Mn2+ against noncognate NTPs could not be determined due to apparent negative-cooperativity (Hill < 1; Figure 3E, F and Supplementary Figure S3B-C and Supplementary Table S9). These results suggest that dsDNA binding allosterically suppresses the utilization of Mn2+ in catalyzing noncognate substrates. Fixing the Hill constant to one for data fitting estimates the EC50s around 350 μM for both NTPs, which is 4-fold tighter than in the absence of dsDNA (Supplementary Table S9). Such changes pale in comparison to the 100-fold enhancement for ATP/GTP (EC50: 18 μM, Supplementary Table S5). Moreover, the low affinity would make the utilization of Mn2+ against noncognate substrates unlikely in vivo (20,30). Together, we concluded that positive coupling between dsDNA and Mn2+ requires cognate ATP/GTP.
cGAS first generates pppGpA from ATP/GTP then cyclizes the intermediate into cGAMP (3,22). It remained unclear which catalytic step was affected by Mn2+ and whether the same products were being produced. We thus analyzed various cGASFL reactions using High Pressured Liquid Chromatography (HPLC). We first established the retention times for ATP, GTP, and cGAMP using synthetic materials (Figure 4A, gray). With 5 mM Mg2+, we then confirmed that the peaks corresponding to cGAMP or pppGpA appeared only if dsDNA was present (Figure 4A, black versus orange). With saturating Mn2+, but without dsDNA or Mg2+, the same two peaks emerged (Figure 4A, blue), indicating that Mn2+ does not change the intrinsic cGAMP synthesis activity. dsDNA/Mg2+-induced cGAS reactions with or without 50 μM Mn2+ also produced the same peaks (Figure 4B). Notably, however, both cGAMP and pppGpA peaks were almost equally higher when Mn2+ was present (Figure 4B and C). These results suggest that either the first catalytic step, or both catalytic steps were accelerated. We thus concluded that Mn2+ potentiates the NTase activity of cGAS without altering product identity.
Next, we asked whether dsDNA also accelerates both the first and second linkage formation in a length-dependent manner. Here, cGAS reaction products were identical across different dsDNA lengths (Figure 4D). However, while the peak corresponding to pppGpA was increased by 1.4-fold from 19- to 339-bp dsDNA, the cGAMP peak was 4.1-fold higher (Figure 4D and E), which are in stark contrast to when Mn2+ accelerated both steps equally (Figure 4C, b). Importantly, our results also suggest that dsDNA length allosterically dictates the rate-limiting cyclization. To further test this, we monitored reaction products resulting from different cGAS amounts, as promoting dimerization by raising protein concentrations decreases the dependence on dsDNA length in activating cGAS (19). pppGpA production increased linearly with increasing cGAS concentrations and dsDNA length-dependence was consistently marginal (Figure 4F and Supplementary Figure S4A–C). On the other hand, while cGAMP production depended on dsDNA length at lower cGAS concentrations, it became essentially length-independent at the highest concentration (Figure 4F, b, Supplementary Figure S4D-F), Our observations corroborate that dsDNA length-dependent dimerization is crucial for cyclization. Mn2+ did not alter this trend, but increased the amount of pppGpA and cGAMP for all dsDNA lengths even at the highest cGAS concentration (Supplementary Figure S4). Our observations again indicate that Mn2+ alone activates cGAS, and also synergizes with the allosteric activation. Overall, these results consistently suggest that dsDNA length-dependent dimerization regulates the cyclization of cGAMP from the intermediate.
The in vivo concentration of ATP is well above the KM for cGAS and also present in ∼5-fold excess to GTP (27). Because Mn2+ boosts the NTase activity of cGAS against ATP/ATP (Figure 3), we envisioned that cytosolic Mn2+ might instigate futile catalytic cycles, accumulating an off-pathway product (pppApA). To test this possibility, we thus analyzed cGAS reaction products using physiologically relevant amounts of ATP:GTP (5ATP/1GTP). With 5 mM Mn2+, and without dsDNA or any Mg2+, we noted an additional peak indicating the production of pppApA along with cGAMP and pppGpA (Figure 5A, blue, see also Supplementary Figure S5A–C for ATP/ATP and GTP/GTP reactions). Adding 339-bp dsDNA did not change the pppGpA peak, but increased the cGAMP peak (Figure 5A, B, Supplementary Figure S5D). Remarkably, dsDNA also decreased pppApA production (Figure 5A, B, Supplementary Figure S5D), suggesting that the acceleration of cyclization by long dsDNA in turn suppresses off-pathway product formation. We also analyzed dsDNA/Mg2+-stimulated cGAS reaction with 5ATP/1GTP in the absence or presence of 50 μM Mn2+ (Figure 5C). Here, Mn2+ specifically increased the production of pppGpA and cGAMP without altering pppApA formation (Figure 5C, D, Supplementary Figure S5E), indicating that Mn2+ and longer dsDNA synergistically ensure the efficiency and fidelity of cGAS.
Finally, to further test the role of dsDNA length in suppressing off-pathway product formation, we monitored the kinetics of cGASFL reaction with 5ATP:1GTP in the presence of 19- or 339-bp dsDNA. In the presence of 19-bp dsDNA, both pppApA and pppGpA accumulated linearly, with pppGpA accumulating 5-fold faster than pppApA (Figure 5E, Supplementary Table S10). cGAMP production occurred with a prominent lag then reached a steady-steady (linear) rate (Figure 5E), corroborating the stepwise synthesis model (3,22). With 339-bp dsDNA, both pppGpA and pppApA accumulated about 1.3-fold faster than 19-bp dsDNA-induced production (Figure 5E versus F, Supplementary Table S10). However, the steady-state production of cGAMP was reached 4.3-fold faster than that induced by 19-bp dsDNA, quickly outpacing pppApA production (Figure 5E versus 5F, Supplementary Table S10). The negligible boost in pppApA formation by longer dsDNA also agrees with the dsDNA length-independent NTase activity of cGAS against ATP/ATP (Figure 3A). These results further support the mechanism by which dsDNA length-dependent allostery controls the signal fidelity of cGAS.
cGAS is central to the host defense against various maladies that give rise to cytosolic dsDNA (1). Despite extensive cellular and structural studies (3,15,18,19,28,36,37), how human cGAS operates remain poorly understood at the molecular level. For instance, how cytosolic Mn2+ released from damaged organelles potentiates cGAS activity remains unknown (20). Moreover, it is unclear how cGAS manages the inherent substrate promiscuity at its NTase active-site in vivo when its NTP substrates exist well above their respective KMs (Figure 3) with ATP present in large excess over GTP (27). Here, we identify an unexpected activation mechanism in which Mn2+ activates monomeric cGAS without dsDNA. More importantly, this noncanonical activation synergizes with the canonical dsDNA-dependent activation. Furthermore, we show that dsDNA length-dependent dimerization regulates the substrate specificity and cyclase activity of cGAS, with Mn2+ further polarizing these activities toward cognate ATP/GTP. Together, we propose that this uniquely synergetic coupling mechanism underpins the signaling efficiency, substrate specificity, and product fidelity of cGAS (Figure 6).
Cytosolic Mn2+ is crucial for the dsDNA sensing activity of cGAS (20). Unlike abundant Mg2+, Mn2+ is present in trace amounts and also sequestered in mitochondria and the Golgi (31,32,35). Pathogen invasion or exposure to toxic reagents damages these vital organelles, releasing Mn2+ to the cytosolic space (20). That is, the co-presence of Mn2+ and long dsDNA in the cytosol would signal one of the most critical conditions such as viral replication and/or massive mitochondrion deaths. Our work demonstrates the mechanisms by which cGAS senses and integrates multiple danger-associated molecular patterns within the environment that gives rise to cytosolic dsDNA, thus dialing its signaling outputs accordingly. The concentration of Mn2+ for dsDNA-free activation (EC50 ∼ 1 mM; Figure 2) is much higher than physiological levels (e.g. <10 μM in THP1 cells (20) and ∼50 μM in brain cells (30)); however, the affinity for Mn2+ drastically improves by dsDNA (Figure 1E and F). Such a highly cooperative mechanism would then ensure the host to properly initiate IFN-I responses.
The active-sites of cGAS-like NTases show varying degrees of promiscuity (23–25,28), and cGAS accepts different NTP pairs and generate off-pathway products (3,18,21,22,28). By contrast, human STING is hyper-selective towards cGAMP (3,21,22). cGAS is an unusually slow enzyme, as one round of cGAMP synthesis takes ∼20 s even under the best-case scenario (Figure 1C). Thus, if cGAS excessively generates off-pathway dinucleotides or fails to cyclize the on-pathway intermediate in a timely fashion would impinge on promptly mounting stress-responses. Further compounding issues include the imbalance between ATP and GTP (27), rate-limiting cyclization efficiency (3,8,22,28) and Mn2+ that often found to be mutagenic in other NTases (23–25). We find that the intrinsic allostery of cGAS solves these potentially graving problems. For instance, all biochemical activities required for efficient signaling are concertedly and selectively maximized with the correct signal, cognate substrates, and a specific metal cofactor (Figure 6). On the flip side, the same allosteric mechanism also permits greater promiscuity under normal, non-stressful conditions to minimize spurious activation (Figure 6). For example, unlike other NTases where Mn2+ favors noncognate NTPs (23–25), Mn2+ preferentially accelerates the kcat of ATP/GTP for cGAS (Figure 3). Second, ATP/GTP is essential for the synergy between Mn2+ and dsDNA length-dependent activation (Figure 3). Third, dsDNA binding invokes anti-cooperativity in utilizing Mn2+ against either ATP or GTP alone, thereby further suppressing misincorporation of Mn2+ with noncognate substrates. Fourth, dsDNA length is irrelevant for the catalysis against ATP/ATP or GTP/GTP (Figure 3). Finally, longer dsDNA specifically accelerates the rate-limiting cyclization of cGAMP (Figures 3–5), which also suppresses off-pathway product formation even with Mn2+ and excess ATP (Figure 5). Considering that longer dsDNA promotes oligomerization (19) and subsequent liquid-phase condensation (LPC) of cGAS (38), it is tempting to speculate that phase-separated cGAS•dsDNA complexes trap the pppGpA intermediate to facilitate cyclization. Together, all these factors would synergistically maximize host innate immune responses against major intracellular assaults. On the other hand, short (or no) dsDNA decreases the affinity for Mn2+ (Figures 1 and 2), reduces the cyclization efficiency (Figures 4 and 5), and allows off-pathway reaction with noncognate NTPs to be almost as favorable as the on-pathway reaction with ATP/GTP (Figure 3B). All these actions would then synergistically minimize the signaling activity of cGAS (i.e. viral dsDNA has been degraded by host nucleases). Combined, such a tightly coupled allosteric mechanism would control the host innate immune system in a precise yet efficient manner.
Mg2+ is thought to be the canonical cofactor for cGAS-like NTases (23–25,39). However, a recent study revealed that Mn2+ could activate the catalytic efficiency of several bacterial cGAS-like NTases (39); the ligand requirements for these NTases remain largely unknown (39). Considering the highly concerted relationship between the dsDNA ligand and the Mn2+ cofactor we observe here, future mechanistic studies will reveal whether other NTases with signaling functions might entail similar ligand-cofactor selection mechanisms as cGAS. On the other hand, it is noteworthy that ‘short’ dsDNA (<40 bp) is relatively abundant in the cytoplasm (1,40,41). We found that a physiologically relevant amount of Mn2+ (50 μM) activated cGAS with minimal (or no) dsDNA when both wild-type and various mutant enzymes were overexpressed in HEK293T cells (Figure 2J). For instance, K173E/R176E-cGASFL did not respond to dsDNA in vitro, but showed robust activity in HEK293T cells with Mn2+ (Figure 2G versus J). Also importantly, the catalytic activity of these cGAS mutants were nonetheless marginal compared to wild-type (Figure 2H), indicating that cGAS does not have to operate at the maximal catalytic capacity to induce robust IFN-I signaling. Considering that IFN-I drives the overexpression of cGAS via a positive feedback loop (42), it is tempting to speculate that the Mn2+-induced activity of cGAS might underpin many type-I interferonopathies that do not involve pathogen infection (i.e. sterile inflammation), which include manganism (32), Parkinson's disease (43,44), and a wide range of autoimmune disorders such as Aicardi-Goutieres syndrome and systemic lupus erythematosus (7–10).
We thank Dmitriy Bobrovnikov, David Bartee, Caren Freel-Meyers, Taekjip Ha for the help with HPLC.
Author contributions: R.M.H. conceptualized the project, designed and performed experiments, analyzed data, and wrote the paper. G.M. and K.R. designed and performed experiments, analyzed data, and edited the manuscript. M.A.C. analyzed the data, supervised cell-based experiments, and edited the manuscript. J.S. supervised the overall project, analyzed data, and wrote the paper.
Supplementary Data are available at NAR Online.
American Cancer Society [DMC-RG-15-224]; National Institutes of Health (NIH) [GM129342A1 to J..S.]; NIH [AI102696 to M.A.C.]; Johns Hopkins School of Medicine Synergy Award (to J.S. and M.A.C.). Funding for open access charge: NIH.
Conflict of interest statement. None declared.
Present address: Richard M. Hooy, Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720, USA.