Nonspecific DNA Binding of cGAS N Terminus Promotes cGAS Activation

This article is featured in In This Issue, p.3369

Innate immunity is the first line of host defense against invading pathogens. Different pathogen-associated molecular patterns, such as LPS, flagellins, and nucleic acids, are recognized by specific pattern recognition receptors. In the case of DNA, it is now well known that host- or pathogen-derived cytosolic dsDNA can trigger robust production of type I IFNs and other cytokines (1). Particularly, recognition of aberrant self-dsDNA by innate immune sensors can cause autoimmune diseases such as systemic lupus erythematosus and Aicardi-Goutières syndrome, suggesting that the cytosolic dsDNA signaling pathway is a promising target for treating autoimmune diseases (2).

Many proteins have been reported as DNA sensors over the past few decades, including TLR9, RNA polymerase III, DAI, IFI16, DDX41, DNA-PK, MRE11, STING, and Sox2. However, they all recognize dsDNA in either a cell-type specific or sequence-specific manner (3, 4). It was not until 2013 that the general cytosolic DNA sensor cyclic GMP-AMP synthase (cGAS) was identified by Zhijian Chen’s group (5, 6). The cGAS sensor not only recognizes cytosolic dsDNA but also synthesizes the second messenger 2′3′-cGAMP from ATP and GTP (7–11), which then binds to and activates STING (also known as MITA, ERIS, or MPYS) (12–16). STING undergoes conformational changes (11, 17, 18) and translocation from the endoplasmic reticulum to the Golgi apparatus (19) to encounter TBK1 and IRF3 (20, 21), eventually triggering the production of type I IFNs (22). This newly established cytosolic DNA sensing and signaling pathway has important physiological roles in infection, inflammation, and cancer (4, 23).

Human cGAS (hcGAS) is a 522 aa protein comprising a less-conserved N-terminal domain (residues 1–160) and highly conserved nucleotidyltransferase (NTase) and Mab21 domains (residues 161–522), exhibiting structural and sequence homology to the catalytic domain of oligoadenylate synthase 1 (OAS1). Previous studies showed that hcGAS contains two separate DNA binding domains, 1–160 and 160–212 respectively. In addition, NTase and Mab21 domains are essential and sufficient for cGAS enzymatic activity (5). Moreover, the crystal structures of mouse, porcine, and human cGAS have been determined in both apo form and cGAMP/DNA-bound forms (7, 24, 25). However, in all these structures, cGAS proteins were crystallized as truncated forms, which only include NTase and Mab21 domains without the ∼160 aa residues from the N-terminal domain. Until now, structural and functional studies on the N-terminal domain of cGAS have not been carried out (4).

In this work, we first observed that the N terminus of human cGAS (hcGAS-N160) induces secondary structural changes upon dsDNA binding in solution. Meanwhile, we compared the nonspecific binding property and affinity systematically using recently developed single-molecule technique and traditional biochemical methods for three hcGAS proteins: full length hcGAS (hcGAS-FL), hcGAS-d160 (hcGAS without the 160 N-terminal residues), and hcGAS-N160 (N-terminal 160 residues). The in vitro data demonstrated that hcGAS-N160 helps hcGAS-FL to expand the moving range on dsDNA and facilitates more efficient binding of hcGAS-FL to dsDNA, thereby enduing higher cGAS enzyme activity than hcGAS-d160. More importantly, when stably expressed in cGAS^−/− cells, hcGAS-FL triggers stronger activation of STING/IRF3-mediated signaling upon DNA virus infection or DNA transfection than hcGAS-d160. Together, these results elucidate that the N terminus of cGAS plays an important role in promoting its own activation upon sensing host- or pathogen-derived dsDNA.

The cDNA encoding hcGAS was kindly provided by Dr. Z. J. Chen (University of Texas Southwestern Medical Center). hcGAS-FL was subcloned into p3xFlag-CMV-7.1 (Sigma) or pcDNA3.1-HA (Sigma) to generate Flag/HA-tagged proteins. For prokaryotic expression, it was cloned into pET21b (Novagen) between the NdeI and XhoI sites with a C-terminal fusion 6×His-tag. hcGAS-N160 and hcGAS-d160 truncations were generated by the KOD-Plus-Mutagenesis Kit (TOYOBO Bio-Technology). The cDNA encoding mouse cGAS (mcGAS) was amplified from the MEF cDNA library, and mcGAS-FL, mcGAS-d146, or mcGAS-N146 were generated as mentioned above. Other expression plasmids used in this study were described previously (15). All cloned genes were sequence verified.

HEK293T, HeLa, 2fTGH-ISRE cells were cultured in DMEM supplemented with 10% FBS (Life Technologies) and antibiotics. THP-1 cells were cultured in RPMI 1640 medium supplemented with 10% FBS, l-glutamine, and sodium pyruvate. Anti-Flag (F3165; Sigma), anti-HA (H9658; Sigma), anti-GAPDH (sc-25778; Santa Cruz), and anti-p-IRF3 (ab76493; Abcam) Abs were purchased as indicated. Antiserum against STING, cGAS, IRF3, and ISG54 were generated by immunizing mice with the indicated recombinant protein purified from Escherichia coli, at Beijing Biotop Biotechnology, China. Sendai virus (SeV; Congyi Zheng, Wuhan University, China), Vaccinia virus (VACV, Western Reserve strain; Min Fang, Institute of Microbiology, CAS) were gifts from the indicated colleagues, and virus titers were measured by plaque assays using BHK21 cells.

The detailed procedures for SDS-PAGE, Western blotting, luciferase reporter assay, RT-PCR, type I IFNs bioassay, in vitro cGAMP activity assay, isolation of endogenous cGAMP, and chemical synthesis of 2′3′-cGAMP have been described previously (15, 18, 26).

Eight different oligonucleotides were used in this work to test the binding performance of cGAS proteins. All oligonucleotides were synthesized by Sangon (Shanghai, China), and their detailed sequences are listed in Supplemental Table I. Briefly, DNA-16 bp was chosen as the same sequence used in cocrystallization with cGAS Mab21 domain in Mus musculus (7). Poly(dA:dT) and poly(dG:dC) were designed as 20 bp random sequences. IFN-stimulatory DNA (ISD) and HSV-1 DNA were chosen as they were commonly used in the immunological experiments previously reported.

Both forward and reverse synthesized oligonucleotides were dissolved in annealing buffer [10 mM Tris-HCl (pH 7.6), 50 mM NaCl, 1.0 mM EDTA] and annealed by denaturing at 95°C for 5 min, then decreasing 0.5°C every minute to 4°C with equal molar. Annealed nucleotides were further purified by Superdex 75 size-exclusion chromatography (GE Healthcare) in buffer C [20 mM Tris-HCl (pH 7.6), 100 mM NaCl] to get rid of possibly remaining ssDNA.

Fluorescein amidite(FAM)-labeled oligonucleotides share identical sequences and purification steps with unlabeled DNA except for the adding of an FAM fluorescent molecule at the 5′ terminus.

The cGAS proteins were over-expressed in E. coli strain BL21 (DE3) at 18°C for 20 h after induction with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Resuspended cells were lysed by sonication in buffer A [20 mM Tris-HCl (pH 7.6), 1 M NaCl]. The lysates were applied to Ni-chelating columns (5 ml HiTrap HP column; GE Healthcare) equilibrated with buffer A. After washing with 50 mM imidazole, the His₆-tagged proteins were eluted with a binding buffer containing a linear gradient of 50–500 mM imidazole. The eluted proteins were further purified by size-exclusion chromatography (Superdex 75, 24 ml; GE Healthcare) in buffer C [20 mM Tris-HCl (pH 7.6) 100 mM NaCl]. The fractions containing target proteins were concentrated to 1.0 mM, flash frozen in liquid nitrogen, and stored at −80°C for further experiments.

Purified proteins used in single-molecular dynamic imaging were labeled by reaction with Cy3B-maleimide (GE Healthcare) at surface cysteines of different hcGAS proteins as previously described (27). Briefly, ∼5.0 μM target proteins were mixed with Cy3B-maleimide (C_final = 60 μM) in 50 mM Tris-HCl (pH 7.8), 200 mM NaCl, 15% glycerol, and 500 μM Tris-(2-carboxyethyl) phosphine hydrochloride (Sigma). The mixture was degassed with nitrogen for 10 min and incubated at 4°C overnight. A desalting column (PD MinTrap G-25; GE Healthcare) separation was performed to separate target proteins and unreacted free dye molecules with the above buffer except Tris-(2-carboxyethyl) phosphine hydrochloride.

Size-exclusion chromatography with multiangle light scattering (SEC-MALS) experiments were performed using a fast protein liquid chromatography system (GE Healthcare) connected to a Wyatt Mini-DAWN TREOS MALS instrument and a Wyatt Optilab rEX differential refractometer (28, 29). Indicated proteins at 2.0 mg/ml with or without incubation with DNA for 30 min were injected into a Superdex 75-column equilibrated in buffer C [20 mM Tris-HCl (pH 7.6) and 100 mM NaCl]. BSA (68 kDa; Sangon Biotech, China) was used as a standard to calibrate the system, and data were collected every 0.5 s at a flow rate of 0.5 ml/min. Data analysis was carried out using the program ASTRA, yielding the molar mass and mass distribution (polydispersity) of the sample.

The single-molecule measurements were carried out inside a home-made flow-cell, designed according to the protocol previously described (27, 30). The assay buffer contained 20 mM Tris (pH 7), 5.0 mM NaCl, 5.0 mM MgAc₂, 0.5 mg/ml BSA, 5.0 mM DTT, 10% (v/v) glycerol, 0.8% (w/v) glucose, 1.0 mg/ml glucose oxidase, 0.1 mg/ml catalase, and 2.0 mM trolox (all reagents were purchased from Sigma). Degassed assay buffer was carefully filtered with a 0.22 μm syringe filter (Millipore). The λDNA (Thermo Fisher Fermentas) was biotinylated and immobilized on a cover slip as previously described (31, 32) with incubation for 1 h at a concentration of 50 pM in 50 mM Tris-HCl (pH 8), 10 mM NaCl, 0.1 mM Na₂EDTA, and 1.0 mg/ml BSA. During acquisition of the assay images, a buffer stream was kept flowing at ∼4 ml/h utilizing a syringe pump (Model LSP02-1B; Longer Pump, China).

The single-molecule fluorescence imaging performed in this study followed the protocol from previous publications (27, 31, 33, 34). Briefly, an inverted microscope (Nikon Ti; Nikon, Japan) was used for total internal reflection illumination. An area consisting of ∼10⁻⁴ cm² was illuminated with 5–10 W/cm² by focusing the beam of a solid-state 532 nm green laser (MGL-III-532/100 mM; Changchun New Industries Optoelectronics Tech., Changchun, China) into the back aperture of a 100× objective lens (numerical aperture 1.49; PlanApo, Nikon). Fluorescence measurements were collected using the same objective lens, through a dichroic beam splitter (model Di01-R532-25*36; Semrock) and long pass filter (model BLP01-532R-25; Semrock) to be imaged on a back-illuminated electron multiplying charge coupled device camera (iXon3 888; Andor Technology). Fluorescent images were recorded every 0.05 s for a duration of 3 min.

Fluorescent images were processed using algorithms written in Matlab (The Mathworks, Natick, MA) as previously described (33, 35). The one-dimensional diffusion coefficient was calculated with the formula:

Where x is the move range of the protein and t is the time of moving.

Circular dichroism (CD) spectra were collected at MOS-500 Circular Dichroism Spectrometer (BioLogic, France) using 1.0 mm path length cells. The spectra from 185 to 250 nm were acquired every 1 nm with 2 s of averaging time per point and a 2 nm band pass at 25°C. The protein concentration used for each sample was 0.2 mg/ml, the final spectra were normalized by the protein concentration or the total concentration of protein and DNA for the complex. The complex samples of hcGAS-N160 and DNA ligands (equal molar ratio) were preincubated on ice for 15 min. All samples were in 20 mM HEPES buffer (pH 7.1) and 100 mM KCl. A buffer baseline was collected in the same 1.0 mm path length cuvette for background correcting.

FAM-labeled oligonucleotides were prepared as above described and used for EMSA. Oligonucleotides of 18.9 pmol were mixed with 1.5, 3, 7.5, 15, 30, 60, 120, 300, and 600 pmol target proteins respectively with a total volume of 15 μl.

The in vitro DNA-protein binding buffer contained 20 mM Tris-HCl (pH 7.6), 100 mM NaCl, and 5 mM MgCl₂. After incubation on ice for 30 min, the DNA-protein mixture was loaded to 8% native PAGE and electrophoresed for 50 min at 120 V in 0.5× Tris-borate–EDTA buffer. Equal amount of oligonucleotides were handled alone in the same way as the free DNA control.

The polyacrylamide gels were pictured with ImageLab (Bio-Rad) in ChemiDocXRS (Bio-Rad), and the intensity of each lane detected by Gel-Pro Analyzer (Media Cybernetics). The fraction of bound DNA can be determined by measuring the counts present in the bound species and dividing by the total counts present in the lane. Combined with the concentration of DNA and protein, a fit to the Hill equation was then performed (36) to calculate the dissociation constant K_d by OriginPro 7.0 (OriginLab).

FAM-labeled oligonucleotides were prepared as described above and used for microscale thermophoresis (MST) experiments, which have been described in detail elsewhere (37). Briefly, we first chose a suitable concentration for the FAM-labeled oligonucleotides, which ensured the fluorescence value was between 800 and 1000. The different cGAS proteins were serially diluted to 16 different concentrations with a total volume of 20 μl, in the presence of FAM-labeled oligonucleotides. Then 10 μl of the samples was loaded into standard treated capillaries after incubation for 10 min at room temperature. Measurements were performed on a Monolith NT.115 instrument (NanoTemper, Germany).

Samples were fractionated by Mono Q 5/50 GL column (GE Healthcare) at 1 ml/min flow rate, with a gradient of buffer B [20 mM Tris-HCl, (pH 8.5), 1 M NaCl] in buffer A [20 mM Tris-HCl (pH 8.5)].

To measure cGAS activity, purified recombinant cGAS proteins from E. coli were incubated in 20 μl reaction system mainly containing 2 μM cGAS protein, 2.5 mM ATP, 2.5 mM GTP, 5 mM MgCl₂, and 0.1 mg/ml ISD. After incubation at 37°C for 20 min, the mixture was heated at 95°C for 5 min, centrifuged at 14,000 × g for 10 min, then the heat-resistant supernatant was followed by Mono Q analysis to measure the content of synthesized 2′3′-cGAMP or mixed with perfringolysin-permeabilized THP1 to measure phosphorylation of IRF3.

We performed statistical analysis with a Student t test, and considered p < 0.05 to be statistically significant.

hcGAS is a 522 aa cytosolic protein comprising undetermined N-terminal residues (1–160), conserved NTase, and Mab21 domains (161–522) (Fig. 1A), whereas mcGAS has 507 aa containing ∼146 residues in its N-terminal domain. Recent bioinformatic and functional studies revealed that the cGAS-cGAMP-STING pathway is highly conserved in vertebrates, and the readily identifiable cGAS homologs all contain the conserved residues involved in cGAMP catalysis and DNA binding (3, 38). To further characterize the N terminus of cGAS, we first selected 25 animal cGAS homologs, which ranged from fish to human, and aligned their corresponding sequences. Different from the highly conserved C terminus, the cGAS N-terminal shows vast diversity both in length and sequence among 25 cGAS homologs (Supplemental Fig. 1). We further aligned six mammalian cGAS homologs including mouse, rat, monkey, and human. Intriguingly, all these mammalian cGAS homologs contain the ∼160 residues in the N terminus and show moderate sequence identity, such as highly conserved P20/53/108/144, R55/60/80/111/127, and K7/62/63 in humans (Fig. 1B). Moreover, consistent with previously reported DNA-binding capability (5), the N terminus of mammalian cGAS homologs shares a rather high rate of positively charged residues that can contribute to DNA binding: R (10.6% in humans, 17.8% in mice) and K (6.2% in humans, 4.8% in mice). Collectively, these results indicate that the DNA binding role of the cGAS N terminus is relatively conserved among mammals, although less conserved in their primary sequences.

Although great efforts have been made in the structural studies of different cGAS proteins, no crystal structure of full-length cGAS has been reported, most likely due to the structural flexibility of its N-terminal domain. In line with this, the secondary structures of the N terminus of three mammalian cGAS homologs (including human, mouse, and porcine) were predicted to have great tendency to be disordered, whereas the C terminus showed a low disorder tendency (Fig. 1C). We further used CD spectra to study the secondary structures of purified hcGAS-N160 (Fig. 2C) either alone or in a complex with dsDNA. The results indicated that hcGAS-N160 is not a well-structured protein as predicted by exhibiting a disordered pattern around 198 nm (Fig. 2A). Strikingly, consistent with the dramatic conformational changes observed in the C terminus of hcGAS upon dsDNA binding (39), the CD spectrum of the incubated hcGAS-N160/DNA complex showed significant changes compared to the protein itself or DNA alone, as there was an obvious shift of α-helix pattern at 192 nm, indicating the increase in α-helical structures (Fig. 2A). This observed difference was most likely caused by changes in the secondary protein structure, because the CD curve of hcGAS-N160/DNA complex was different from the expected value calculated from free protein and dsDNA (40). Meanwhile, we used another DNA ligand, poly(ATCG) (Supplemental Table I), and the same condition was applied to the incubated hcGAS-N160/poly(ATCG) complex. Similarly, the CD spectrum of this mixture showed an obvious shift of α-helix pattern at 192 nm, different from the calculated curve (Fig. 2B). These data collectively suggest that DNA binding by hcGAS-N160 induces secondary structural changes.

To further investigate the exact composition of hcGAS-N160 and the DNA ligand, we used SEC-MALS to determine the exact molecular mass of our proteins and complex, as SEC-MALS has previously been shown to be useful for this purpose (28, 29). The results showed that hcGAS-N160 is a monomer with molecular mass 17.78 ± 0.456 kDa (Supplemental Fig. 2A, 2C). Subsequently, when incubated hcGAS-N160 with DNA-14 bp, peak 1 (molecular mass 27.02 ± 0.746 kDa) suggested this complex is composed of monomeric hcGAS-N160 bound to one molecule of DNA-14 bp. Consistently, peak 2 (molecular mass 18.53 ± 0.332 kDa) represented the free monomeric hcGAS-N160 protein (Supplemental Fig. 2B, 2C). Taken together, we concluded that hcGAS-N160 forms a 1:1 complex induced by DNA binding in solution.

To investigate if hcGAS-N160 affects the DNA binding affinity of hcGAS-FL, proteins of hcGAS-FL, hcGAS-d160, and hcGAS-N160 were expressed, purified (Fig. 3D), and tested for the interactions with dsDNA by biochemical methods. EMSAs of three hcGAS proteins with five different kinds of DNA ligands were tested first (Fig. 3A). These DNA ligands included poly(dA:dT), poly(dG:dC), DNA-16 bp, HSV-1 DNA, and ISD (Supplemental Table I). Consistent with previous work, hcGAS-N160 bound to DNA with relatively high affinity (Fig. 3A). We also calculated the dissociation constant (K_d) of three hcGAS proteins with different DNA according to the remaining nucleic acids and relative ratio of DNA versus protein in each lane of native PAGE gels. The results clearly showed that with respect to all DNA ligands, hcGAS-d160 presented the weakest affinity (corresponding to highest K_d), whereas hcGAS-N160 and hcGAS-FL had a K_d of around 10⁻⁶ M, suggesting that hcGAS-N160 and hcGAS-FL bind to dsDNA with higher affinity (Fig. 3B). Furthermore, we carried out an MST assay, a recently developed technique for analysis of biomolecular interactions, to confirm the results above. Mechanistically, any change of the hydration shell of proteins due to changes in their structure results in a relative change of the movement along the temperature gradient, thus MST is well suited to determine binding affinities. As expected, we obtained comparable results. That is, hcGAS-d160 showed the weakest binding affinity with both DNA ligands, whereas the binding affinity of hcGAS-N160 and hcGAS-FL were much higher; especially hcGAS-FL, which was almost 30–90 fold stronger than hcGAS-d160 (Fig. 3C, Supplemental Fig. 2D). Consistently, EMSA (Fig. 4E, 4F) and MST assays (Fig. 4G, Supplemental Fig. 2E) of three mcGAS proteins (mcGAS-FL, mcGAS-d146, and mcGAS-N146; Fig. 4A, 4B) also demonstrated that the affinity of full-length mcGAS with dsDNA is largely enhanced by its N-terminal domain. In sum, these results unveiled a hitherto unappreciated role of N terminus of cGAS, which is enhancing its DNA binding ability.

In fact, hcGAS-N160 has a more basic isoelectric point of 11.07, compared with those of hcGAS-FL (9.54) and hcGAS-d160 (9.12), which are consistent with mcGAS (Fig. 4C, 4D). This difference is mainly due to the rich content of arginine (10.6% in humans, 17.8% in mice) and lysine (6.2% in humans, 4.8% in mice) of N-terminal of cGAS as mentioned previously. Therefore, it is reasonable to suggest that with such a positive charge and basic isoelectric point, the N-terminal domain helps the hcGAS-FL bind to dsDNA with higher affinity, acting as a DNA-binding enhancer.

To explore whether hcGAS-N160 could affect other behaviors in hcGAS-FL besides DNA binding affinity, we used single-molecule flow-cell method to visualize the direct interaction between fluorescently labeled hcGAS proteins and λDNA. Because nonspecific interaction is difficult to detect by traditional biochemical methods, a single-molecule experiment provides us an excellent assay to observe this interaction visually and sensitively (30). The fluorescence images of representative hcGAS-FL, hcGAS-d160, and hcGAS-N160 moving along a stretched λDNA (dsDNA) were captured (Fig. 5B) by a home-made flow-cell system (27) (Fig. 5A). We observed that hcGAS-FL and hcGAS-N160 were distributed quite broadly along the λDNA, whereas hcGAS-d160 was only detected at a few positions (Fig. 5B, Supplemental Videos 1–3). These results indicate that hcGAS-FL and hcGAS-N160 could nonspecifically bind to λDNA with more sites and stay longer when compared with hcGAS-d160. In other words, hcGAS-d160 binds to λDNA more weakly than the other two proteins.

We also showed the trajectories of these three proteins in vertical and horizontal dimensions (Fig. 5C) from time-series images as shown in Fig. 5B. Of note, moving range along the vertical dimension was checked as a reference to show whether the λDNA is fully stretched. According to the trajectories of these three proteins, hcGAS-FL bound and slid rapidly on a broad range along the λDNA, with the majority of protein molecules moving up to 0.6 μm (1.8 kbp), and some reaching almost 1.0 μm (3 kbp) (Fig. 5C). In contrast, hcGAS-d160 was only detected at a few points and moved within a very parochial range of 0.2 μm (600 bp), which is probably due to the DNA thermal motion (see y-axis), suggesting that this protein can hardly slide along dsDNA (Fig. 5C). Intriguingly, it is noteworthy that hcGAS-N160 seemed to move in an even broader range than the full-length protein (Fig. 5C). Aside from the moving range, hcGAS-FL and hcGAS-N160 were also shown to slide faster than hcGAS-d160 based on the one-dimensional diffusion coefficient of these three proteins, in which the slopes of hcGAS-FL and hcGAS-N160 are much steeper than that of hcGAS-d160 (Fig. 5D, 5E). Taken together, single-molecule imaging experiments revealed that the hcGAS-N160 helps hcGAS expand its nonspecificity and moving range on dsDNA.

To determine the difference in enzymatic activity between hcGAS-FL and hcGAS-d160, we used these two proteins purified from E. coli to test their corresponding enzyme activity in vitro. After incubation with ATP and GTP in the presence of DNA, the reaction product was eluted with a Mono Q column. In agreement with previous reports, hcGAS-d160 was sufficient for cGAS activity and produced a certain amount of 2′3′-cGAMP (5, 7). On the other hand, hcGAS-FL catalyzed a relatively higher amount of 2′3′-cGAMP at the same reaction time (Fig. 6A). Meanwhile, the cGAMP activity assay of the same products was performed and confirmed that hcGAS-FL produces relatively more cGAMP than hcGAS-d160 (Fig. 6B). Furthermore, we also purified these two proteins from HEK293T cells and performed the cGAMP activity assay to further confirm the result above (Fig. 6C). To mimic more physiological conditions, we transfected a low dose of the two cGAS expression constructs into HEK293T cells, in which the transfected plasmid itself acts as a DNA ligand. The IFN-β luciferase assay showed that upon low doses of cGAS plasmid transfection, hcGAS-d160 showed a much weaker activity compared with hcGAS-FL (Fig. 6D). Nevertheless, upon a high-dose transfection, thus over-expression, the two proteins showed similar activity (Supplemental Fig. 3A) as reported previously (5). Together, we demonstrate that the loss of the N-terminal domain relatively reduces the enzyme activity of cGAS.

By conducting in vitro biochemical assays, single-molecule imaging experiments, and enzyme activity assays, we have concluded that hcGAS-N160 would facilitate hcGAS-FL to bind to dsDNA more efficiently and enhance a certain level of its enzyme activity. We then reasoned that N-terminal 160 residues of hcGAS have an important physiological function in STING-dependent antiviral responses. To test this, we used a HeLa cell line harboring an intact DNA signaling pathway. When we knocked out cGAS by CRISPR/Cas9 in HeLa cells, the production of type I IFNs as well as the phosphorylation of STING and IRF3 was completely blocked upon DNA transfection or DNA virus infection (VACV) (Fig. 7A, 7B). In contrast, the loss of cGAS in HeLa cells did not affect type I IFNs induction by RNA virus (SeV), which is known to activate the RIG-I-MAVS pathway (Fig. 7A, 7B). We subsequently stably expressed Flag-tagged hcGAS-FL and hcGAS-d160 proteins in HeLa cGAS-knockout (KO) cells, and tested the antiviral responses of these two stable cell lines. Immunoblot results showed that upon DNA virus infection, hcGAS-FL stably expressed cells activate stronger STING-IRF3 dependent downstream signaling than hcGAS-d160 cells (Fig. 7C). In contrast, these two stable cell lines showed no obvious difference in RNA signaling pathway (Fig. 7C), suggesting the DNA signaling pathway is specifically enhanced in hcGAS-FL stably expressed cells. Meanwhile, bioassay and RT-PCR analysis further confirmed above results that upon SeV infection, there was no observable difference between these two stable cell lines (Fig. 7E, 7F). Furthermore, we found that only plasmid DNA transfection (activates cGAS), but not 2′3′-cGAMP treatment (activates STING directly), showed stronger activation in hcGAS-FL stably expressed cells, suggesting the difference is specifically at the level of cGAS (Fig. 7D). Collectively, these results showed clearly that hcGAS-FL triggers stronger STING/IRF3-dependent downstream signaling than hcGAS-d160 by DNA virus infection or DNA transfection, demonstrating that the N terminus of cGAS has a physiological function in promoting its activation.

To sum up, we have demonstrated that hcGAS-N160, as a highly basic protein with many positive charges at the physiological condition, binds to negatively charged dsDNA in a nonspecific manner, and moves broadly along the λDNA. Compared to the structurally flexible hcGAS-N160, hcGAS-d160 could only bind to dsDNA within the same interface as shown in all the reported crystal structures (7, 24, 41). Nevertheless, when hcGAS-N160 and hcGAS-d160 were combined, the resulting hcGAS-FL could not move along the dsDNA as freely and quickly as hcGAS-N160, likely due to the stereo hindrance of hcGAS-d160 (Supplemental Fig. 3B). This model explains why the one-dimensional diffusion coefficient of hcGAS-FL lies in between the other two proteins (Fig. 5D). Our experiments, therefore, revealed that it is hcGAS-N160 that facilitates hcGAS-FL to expand the binding and moving range on dsDNA. Moreover, hcGAS-N160 also helps to enhance the enzyme activity of cGAS and trigger stronger activation of innate immune signaling upon cytosolic DNA recognition.

In this work, we have combined recently developed single-molecule technique with other biochemical and cellular assays to demonstrate the role of N-terminal domain of cGAS in enhancing the function of cGAS. Previously, hcGAS-N160 was only shown to have a DNA binding ability as one of the two DNA binding domains. However, because hcGAS-d160 retained cGAMP synthesis activity, the N-terminal 160 aa residues were believed to be largely dispensable for DNA binding and catalysis by cGAS. To our knowledge, we systemically investigated the cGAS N terminus for the first time, and revealed its important function of enhancing the activation of cGAS. From an evolutionary perspective, we have good reason to believe that the N-terminal of cGAS plays a role in physiological conditions, especially in mammals, because all mammalian cGASs contain the ∼160 residues in N terminus and show a certain degree of sequence identity. Moreover, the sensitive single-molecule technique provides us with an unconventional but effective way to observe visually and directly the kinetics of nonspecific interaction.

Most likely due to the structural flexibility, the reported crystal structures of mouse, porcine, and human cGAS were all crystallized as truncated forms without the N-terminal domains. Further structural work on the N-terminal of cGAS (residues 1–160 in humans and 1–146 in mice) or the full length cGAS complexed with DNA should help us to understand the function of native cGAS under physiological conditions. Besides dsDNA, recently, ssDNA, Y-form DNA (containing unpaired guanosines), and RNA:DNA hybrids, either from pathogens or host, have all shown immunostimulatory contents detected by cGAS (42, 43). Considering that cGAS has two DNA-binding domains and the C terminus of cGAS could bind to dsDNA within the same interface as shown in all reported crystal structures, it is possible that the cGAS N-terminal domain plays a role in recognizing these newly found, unconventional nucleic acid ligands.

It is now well recognized that the cGAS-mediated DNA signaling pathway plays an essential role in infection, inflammation, and tumor therapy. Consequently, the N-terminal domain is an important part of full-length cGAS and enhances its role in anti-viral and anti-tumor immune responses. Some recent work has shown that the function of cGAS could also be regulated by other processes (4, 44–46). Thus, regulating cGAS via the N-terminal domain may represent one effective way. Recently our colleagues have found that pyroptotic caspases regulate innate immunity through cGAS cleavage during canonical or noncanonical inflammasome activation. Particularly, caspase-1 is activated by various inflammasome agonists to recruit and cleave human cGAS at D^140/157 (in the N terminus), resulting in reduced cGAMP production and cytokine expression. Consequently, inflammasome defects enhance DNA virus resistance both in vitro and in vivo (47). It is also possible that the N terminus of cGAS has been involved in other reported or undiscovered interactions such as Beclin-1 (48), Akt (49), cyclin kinases (50), PQBP1 (51), TTLL4/TTLL6 and CCP5/CCP6 (52), TRIM38, and Senp2 (53). Interestingly, K⁷ in hcGAS was shown to be acetylated (54) and S¹⁴³ to be phosphorylated (55), indicating the importance of cGAS N terminus. Further work on this will be necessary to understand the detailed function of full-length cGAS.

We thank Yong-Ping Xu, Zhao-Yang Ye, Juyi Gao, Yukun Guan, Pengfei Gao, and Rui Zhang for providing technical assistance, Prof. Yi Li for assistance with MST experiments, and Dr. Gui-lan Li and the Core Facilities (School of Life Sciences, Peking University) for assistance with protein purification and SEC-MALS.

The authors have no financial conflicts of interest.