DNA-induced liquid phase condensation of cGAS activates innate immune signaling

See allHide authors and affiliations

Science  17 Aug 2018:
Vol. 361, Issue 6403, pp. 704-709
DOI: 10.1126/science.aat1022

Liquid droplets step on the cGAS

Spontaneous partitioning of a homogeneous solution of molecules, or liquid-phase separation, underlies the formation of cellular bodies from P granules to nucleoli. Essentially, dense-phase liquid droplets act like cellular compartments. Du and Chen show that DNA binding to its cytoplasmic sensor, cyclic GMP–AMP synthase (cGAS), results in liquid droplets containing activated cGAS (see the Perspective by Ablasser). This phenomenon occurs through multivalent interactions, augmented by zinc, between DNA binding domains on cGAS and DNA in a length-dependent manner. Binding triggers a switchlike reaction that concentrates the enzyme and reactants to enhance STING-dependent interferon responses.

Science, this issue p. 704; see also p. 646


The binding of DNA to cyclic GMP–AMP synthase (cGAS) leads to the production of the secondary messenger cyclic GMP–AMP (cGAMP), which activates innate immune responses. We have shown that DNA binding to cGAS robustly induced the formation of liquidlike droplets in which cGAS was activated. The disordered and positively charged cGAS N terminus enhanced cGAS-DNA phase separation by increasing the valencies of DNA binding. Long DNA was more efficient in promoting cGAS liquid phase separation and cGAS enzyme activity than short DNA. Moreover, free zinc ions enhanced cGAS enzyme activity both in vitro and in cells by promoting cGAS-DNA phase separation. These results demonstrated that the DNA-induced phase transition of cGAS promotes cGAMP production and innate immune signaling.

Cyclic GMP–AMP synthase (cGAS) is a DNA-sensing enzyme that catalyzes the conversion of GTP and ATP to cyclic GMP–AMP (cGAMP), which activates the adaptor protein STING (1, 2). This, in turn, induces type I interferons and other cytokines (35). DNA arising in the cytoplasm activates cGAS and drives the formation of cytoplasmic foci containing cGAS and DNA (1). However, the molecular mechanism and functional effect of such cGAS foci are poorly understood. cGAS contains a disordered and positively charged N terminus and a structured C terminus harboring a nucleotidyltransferase domain (core cGAS). Both the N and C termini of cGAS bind to DNA irrespective of DNA sequence (611). We hypothesized that such multivalent interactions could lead to the formation of large membraneless protein foci through liquid phase separation (Fig. 1A). This physicochemical process has emerged as a key mechanism underlying the formation of cellular bodies such as P granules and nucleoli (1220).

Fig. 1 DNA binding to cGAS induces the formation of liquidlike droplets in which cGAS is activated.

(A) Schematic of hypothetical cGAS-DNA interactions that drive liquid phase condensation. (B) Time-lapse imaging of cGAS-DNA phase separation. Liquid droplets formed after mixing of 10 μM human FL-cGAS (3% was labeled with Alexa 488) with 10 μM 100-bp DNA (2% was labeled with Cy3) and matured over 60 min. The images shown are representative of all fields in the well. (C) Time-lapse micrographs of merging droplets that formed as described in the legend to (B). Data are representative of at least 10 merging droplets. (D) Fluorescence intensities of cGAS-DNA liquid droplets forming over the time course of 120 min. Data were normalized to 100% by maximal fluorescence intensity. Values shown are means ± SD. n = 4 images. AF488, Alexa Fluor 488. (E) EqDiameter (the diameter of a circle with the same area as the measured object) frequency distribution of cGAS-DNA liquid droplets formed at the indicated time points. cGAS, 5 μM; DNA, 5 μM. (F) FRAP of cGAS-DNA liquid droplets. Bleaching was performed at the indicated time points after cGAS (10 μM) and DNA (10 μM) were mixed, and the recovery was allowed to occur at 25° or 37°C. Time 0 indicates the start of recovery after photobleaching. Shown are the means ± SD. n = 3 liquid droplets. n.s., not significant (P > 0.0332); ***P < 0.0002; ****P < 0.0001 [one-way analysis of variance (ANOVA)]. (G) Phase separation diagram of human FL-cGAS and 100-bp DNA at the indicated concentrations. (H) (Left) cGAMP production by the indicated concentrations of cGAS in the presence of ATP, GTP, and HT-DNA. cGAMP production at low cGAS concentrations is shown in the inset. (Right) Normalized cGAMP production divided by cGAS concentrations. Shown are the means ± SD. n = 3 assays. Data are representative of at least three independent experiments unless indicated to be otherwise.

To test whether DNA binding induces phase separation of cGAS in vitro, we incubated fluorescently labeled cGAS protein (fig. S1A) with double-stranded DNA (dsDNA) oligonucleotides [100 base pairs (bp)] (table S1 and methods). Upon mixing, cGAS and DNA formed micrometer-sized liquid droplets within 2 min (Fig. 1B and movie S1). Small liquid droplets fused into larger ones (Fig. 1C), accompanied by increased fluorescence intensity and larger equivalent diameter (Fig. 1, D and E, and movie S1). Fluorescence recovery after photobleaching (FRAP) experiments showed that when bleaching was performed 30 min after the initiation of phase separation, the fluorescence of cGAS or DNA was efficiently recovered in a temperature-dependent manner (Fig. 1F and fig. S1, B to D). In contrast, when bleaching occurred 2 hours after mixing of cGAS and DNA, the fluorescence recovery was much slower (Fig. 1F and fig. S1D). Thus, cGAS and DNA molecules within the liquid droplets are mobile and exhibit dynamic internal rearrangement in the early phase but gradually undergo a liquid-to-solid transition and mature into a gel-like state (Fig. 1F) (19, 21).

cGAS and DNA formed liquid droplets when the concentration of each exceeded 30 nM (Fig. 1G) in a buffer mimicking physiological ion concentrations and the composition of the cytoplasm (physiological buffer) (methods). Measurements of cGAMP production by cGAS revealed that the specific activity of cGAS substantially increased at concentrations above 30 nM (Fig. 1H and methods). The in vitro cGAS-DNA phase transition was weakened by increasing salt (NaCl) concentrations (fig. S2A), suggesting that ionic interactions between cGAS and DNA were important for the phase transition. A 45-bp dsDNA commonly used for the stimulation of the cGAS pathway [immune stimulatory DNA (ISD)] also robustly induced cGAS liquid droplet formation (fig. S2, B to D). This effect was abolished by treatment with Benzonase, which degrades DNA (fig. S2E). cGAS also formed liquid droplets with 45-bp double-stranded RNA (fig. S2F), but RNA did not activate cGAS to produce cGAMP (fig. S2G). These results are consistent with previous structural studies indicating that DNA but not RNA binding induces a conformational change that activates cGAS (7). Thus, liquid phase separation is insufficient for cGAS activation in the absence of the correct conformational change induced by DNA.

The cGAS-DNA phase separation was unaffected by the addition of ATP, GTP, or a combination thereof (fig. S3A). Moreover, ATP or GTP could be partitioned into and enriched in the cGAS-DNA liquid droplets (fig. S3, B and C). FRAP experiments showed that ATP was rapidly exchanged into and out of the cGAS-DNA droplets (fig. S3D).

We next examined the formation of cGAS-DNA foci within cells. In the human fibroblast cell line BJ-5ta stably expressing a Halo-tagged cGAS, cGAS formed puncta with fluorescein-labeled ISD in the cytoplasm (Fig. 2A and fig. S4A). To confirm that cGAS formed large granules with DNA in the cytoplasm, we used cGAS-deficient mouse embryonic fibroblast (MEF) cells reconstituted with green fluorescent protein (GFP)–cGAS, transfected the cells with Cy5-ISD, and then permeabilized the cells with saponin (22). cGAS formed puncta with ISD and remained in the cytoplasm after saponin treatment (Fig. 2B). The cGAS-DNA foci exhibited liquidlike properties, as demonstrated by the ability of two foci to fuse with each other (Fig. 2C). Furthermore, upon photobleaching, cGAS in the foci displayed near-complete fluorescence recovery within 120 s (Fig. 2, D and E), indicating that cGAS exhibits dynamic liquidlike behavior within cellular granules.

Fig. 2 DNA-induced liquid phase separation of cGAS in cells.

(A) Representative live-cell images of cGAS-DNA puncta formation after BJ cells stably expressing Halo-cGAS, which was covalently labeled with tetramethyl rhodamine, were transfected with fluorescein-ISD. Insets are zoomed images showing cGAS-DNA puncta. Scale bars, 10 μm. These images are representative of at least 10 cells. (B) MEF cells stably expressing GFP-cGAS were transfected with Cy5-ISD for 4 hours, after which they were permeabilized with saponin and analyzed by fluorescence microscopy. Shown in blue is the plasma membrane marker wheat germ agglutinin. Arrows indicate puncta. Scale bars, 50 μm. These images are representative of at least five fields examined. (C) Time-lapse micrographs of cGAS (green) and DNA (magenta) puncta formation and fusion (time 0 represents 30 min after transfection with Cy5–45-bp ISD). Arrows indicate puncta. Scale bar, 15 μm. The fusion events existed in all eight fields examined. (D) Representative micrographs of cGAS-DNA puncta before and after photobleaching (arrow, bleach site). Scale bar, 15 μm. These images are representative of at least three cells in which the cGAS-DNA puncta were photobleached. (E) Quantification of FRAP of cGAS-DNA puncta over a 120-s time course. K, exponential constant; R, normalized plateau after fluorescence recovery. Shown are means ± SD. n = 3 cGAS-DNA puncta. (F) Subcellular fractionation of cGAS activity in cells transfected with DNA. HeLa cells transfected with HT-DNA or untransfected HeLa cells were fractionated by differential centrifugation as depicted in fig. S4B. Fractions were incubated with ATP and GTP, after which cGAMP was measured. Fractions were also analyzed by immunoblotting (IB) with antibodies against histone H2A (nuclear marker), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (cytoplasmic marker), or human cGAS (hcGAS). (G) The P2 fractions from (F) were further separated by OptiPrep gradient ultracentrifugation, and cGAS activity in different fractions was measured as for (F). Fractions from cells not transfected with DNA had no cGAS activity (upper panel). Error bars in (F) and (G) represent the variation range of duplicate assays. Data are representative of at least three independent experiments.

To determine the functional consequence of cGAS liquid droplet formation, we transfected HeLa cells with herring testis DNA (HT-DNA) and measured cGAS activity in subcellular fractions (fig. S4B). Most cGAS activity was present in the pellet obtained after centrifugation at 2,000 × g (designated P2), which contains mainly the nuclei and heavy particles (Fig. 2F). Further separation of the P2 fraction by iodixanol (OptiPrep) density gradient ultracentrifugation revealed two distinct pools of cGAS activity. The first pool was in very heavy fractions (20 to 25% iodixanol) (Fig. 2G), which were separated from the nuclei (27.5 to 30% iodixanol). Thus, cGAS appeared to form heavy particles with transfected DNA that were distinct from cellular organelles and vesicles and that contained active cGAS. The second pool of cGAS activity was in fractions containing the nuclei. However, it remains to be determined whether this activity came from cGAS within the nuclei or some cGAS particles that cosedimented with the nuclei. Similar results were also obtained with human monocytic THP-1 cells (fig. S4, C and D).

Multivalent interactions drive liquid phase separation (16). Long DNA has more binding sites (higher valency) for cGAS than short DNA, and full-length cGAS (FL-cGAS) has higher valency for DNA than core cGAS (Fig. 3A) (6, 7, 11). To test whether cGAS-DNA liquid phase separation is driven by the valency of cGAS and DNA interactions, we incubated FL-cGAS and N-terminally truncated cGAS (ΔN146-cGAS, where the numeral indicates the number of amino acids deleted) with DNA of different lengths in the physiological buffer (15 mM NaCl and 135 mM KCl) or a buffer containing 300 mM NaCl. Human FL-cGAS formed more numerous and larger liquid droplets with longer DNA (Fig. 3B). Both human and mouse FL-cGASs exhibited stronger phase separation than N-terminally truncated cGAS with DNA of the same length either in the physiological buffer (Fig. 3C) or at 300 mM NaCl (fig. S5C). The enzymatic activity of FL-cGAS in the presence of HT-DNA was stronger than that of ΔN146-cGAS in both low-salt buffer (Fig. 3D) and physiological buffer (Fig. 3E).

Fig. 3 Multivalent interactions drive cGAS-DNA condensation and promote cGAS activation.

(A) Schematic of hypothetical cGAS and DNA valencies. (B) Representative images of phase separation by mixing of cGAS (10 μM) with dsDNA of different lengths (10 μM) in physiological buffer. Scale bar, 10 μm. (C) Bright-field images of phase separation by mixing of DNA of different lengths with full-length or N-terminally truncated human or mouse cGAS as indicated. Scale bar, 20 μm. The images shown in (B) and (C) are representative of all fields in the wells. (D and E) cGAMP production by different concentrations of recombinant human FL-cGAS or N-terminally truncated cGAS in low-salt buffer (D) or physiological buffer (E). Shown are the means ± SD. n = 3 assays. (F) Quantification of cGAS-DNA puncta by imaging of MEF cells expressing GFP-tagged human FL-cGAS or ΔN160-cGAS after transfection with Cy5-ISD. Representative images are shown in fig. S6. Values shown are means ± SD. n = 5 images. (G) cGAMP production in the MEF cells expressing human FL-cGAS or ΔN160-cGAS after transfection with ISD or HT-DNA. Values are means ± SD. n = 3. (F) and (G): ****P < 0.0001 (multiple t tests). cGAS expression levels are shown in fig. S5F. Data are representative of at least three independent experiments.

To investigate the role of the cGAS N terminus in cells, we reconstituted cGAS-deficient MEF cells with human FL-cGAS or ΔN160-cGAS (fig. S5F). After transfection with Cy5-ISD, FL-cGAS formed puncta with Cy5-ISD in the MEF cells, whereas ΔN160-cGAS formed fewer puncta (Fig. 3F and fig. S6). cGAMP production was also higher in cells stably expressing FL-cGAS than in those expressing ΔN160-cGAS upon transfection with 45-bp ISD or HT-DNA (Fig. 3G; cGAS expression levels are shown in fig. S5F).

cGAS enzyme activity was much weaker in the assay with physiological buffer than in the assay with low-salt buffer (fig. S5, D and E). This raises the question of how cGAS is activated in cells. We found that zinc ions substantially promoted the activity of recombinant cGAS in the physiological buffer and that this enhancement could be partially replaced by other ions, such as Mn2+ or Co2+ (Fig. 4A and fig. S7A), which is a general characteristic of enzymes that require zinc (23). Similarly, Zn2+ was the most efficient in activating mouse cGAS (fig. S7, B and C). The optimal concentrations of Zn2+ were 160 to 625 μM (Fig. 4A and fig. S7C), which is within the physiological concentration range of zinc ions in cells (24, 25). Zn2+ at ~200 μM markedly facilitated DNA-induced cGAS phase separation at low concentrations of cGAS and DNA (Fig. 4B and fig. S7D). Moreover, cGAS activity at low concentrations was markedly enhanced in the presence of Zn2+ (Fig. 4C). At a concentration as low as 2.5 nM, which is below the concentration of cGAS in the cytoplasm of HeLa cells (approximately 8 to 12 nM) (methods), cGAS underwent phase transition and catalyzed cGAMP synthesis (Fig. 4, B and C). At 10 nM cGAS, 100-bp DNA activated cGAS with a median effective concentration (EC50) of ~1.4 nM, whereas 45-bp DNA had much weaker activity. DNA at 25 bp or shorter had no detectable activity in the in vitro assay (Fig. 4D). This DNA length–dependent activation of cGAS was largely mirrored in the cellular assay in which a THP-1 reporter cell line was transfected with DNA (Fig. 4E).

Fig. 4 Zinc ions promote DNA-induced phase separation and activation of cGAS.

(A) Zn2+ enhances cGAS activation in vitro. Recombinant human FL-cGAS (15 nM) was incubated with ATP, GTP, and DNA in a physiological buffer containing the indicated concentrations of Zn2+, and cGAMP production was measured. (B) Quantification of cGAS-DNA condensates in the presence or absence of zinc. Liquid-phase condensates formed after mixing of Alexa Fluor 488–labeled human FL-cGAS with 45-bp Cy3-labeled ISD at the indicated concentrations of each in physiological buffer with or without Zn2+ (200 μM). Images were then captured by confocal microscopy, and representative images are shown in fig. S7D. Values are means ± SD. n = 5 images. P values are from multiple t tests. (C) cGAMP production in physiological buffer containing HT-DNA and different concentrations of cGAS in the presence or absence of Zn2+ (200 μM). The activity of cGAS at low concentrations is shown in the inset. P values are from multiple t tests. (D) cGAMP production by 10 nM cGAS in physiological buffer containing 200 μM Zn2+ and different concentrations of DNA of the indicated lengths. (E) THP-1–Lucia ISG cells, which harbor a luciferase gene under the ISG54 promoter, were transfected with the indicated DNA for 24 hours, after which the secreted luciferase activity was measured. RLU, relative luciferase units. (F) Thermal shift assay to measure the stability of cGAS or the cGAS-DNA complex in the presence or absence of Zn2+ (200 μM). Tm, protein melting temperature. Values are means ± SD. n = 3. P values are from an unpaired t test. (G) Measurement of cGAS binding to zinc. Zinc ions (10 μM) were incubated with various concentrations of DNA, cGAS, or both, and the solution was passed through a centrifugal filter, after which the zinc ion concentration in the filtrate was measured. The dissociation constant (Kd) values for zinc binding to cGAS and the cGAS-DNA complex were 3.9 ± 1.3 μM and 3.0 ± 0.4 μM, respectively. (H) Depletion of intracellular zinc inhibits cGAS activation by DNA. L929 cells were incubated with the indicated concentrations of the Zn2+ chelator TPEN for 2 hours before transfection with HT-DNA. cGAMP production was measured by a bioassay. Images depicting intracellular zinc depletion are shown in fig. S8B. (B), (C), and (F): n.s., P > 0.0332; *P < 0.0332; **P < 0.0021; ***P < 0.0002; ****P < 0.0001. Error bars represent the variation range of duplicate assays unless otherwise indicated. Data are representative of at least three independent experiments.

Using a thermal shift assay, we found that DNA binding destabilized cGAS but that Zn2+ stabilized the cGAS-DNA complex (Fig. 4F and fig. S8A). Measurements of free Zn2+ concentrations revealed that Zn2+ bound to cGAS but not to DNA (Fig. 4G). To determine whether zinc plays a role in cGAS activation within cells, we depleted L929 cells of zinc with the zinc-specific chelator N,N,N′,N′-tetrakis(2-pyridinylmethyl)-1,2-ethanediamine (TPEN). Cellular cGAMP production upon transfection with HT-DNA was decreased in a TPEN concentration–dependent manner (Fig. 4H and fig. S8B). Under these conditions, TPEN did not affect the viability of L929 cells (fig. S8C). Live-cell imaging with a zinc-specific fluorescent probe revealed that cGAS-DNA puncta contained zinc (fig. S8, D and E). These results showed that zinc facilitated cGAS activation in cells by promoting cGAS phase transition in the presence of cytosolic DNA.

Thus, DNA binding to cGAS induces a robust phase transition to liquidlike droplets, which function as microreactors in which the enzyme and reactants are concentrated to greatly enhance the production of cGAMP. This mechanism allows cGAS to detect the presence of DNA in the cytoplasm above a certain threshold to trigger a switchlike response. Such a switchlike response is made possible by the multivalent interactions between the DNA binding domains of cGAS and DNA in a manner that depends on the DNA length. This also provides an explanation for why long DNA activates cGAS more efficiently. The binding between cGAS and DNA involves extensive ionic interactions between the positively charged surfaces of cGAS and negatively charged DNA. Such interactions are vulnerable to cytoplasmic salt concentrations, which may be a mechanism to prevent spurious activation of cGAS by self-DNA below a certain threshold. However, we found that zinc ions could substantially enhance cGAS phase separation and its enzymatic activation at physiological salt concentrations. Free zinc ions are stored mainly in organelles such as the mitochondria and the endoplasmic reticulum (26), and their delivery to the cytosol may be another avenue by which cGAS activity is regulated in cells. DNA binding to cGAS induced formation of cGAS-DNA condensates, which were observed as cytoplasmic foci within cells. Further characterization of the dynamics and composition of the cGAS condensates should provide deeper insights into the mechanism by which cGAS activity is tightly regulated to trigger an appropriate immune response to pathogens while simultaneously avoiding autoimmune reactions to self-tissues.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Table S1

References (2730)

Movie S1

References and Notes

Acknowledgments: We thank H. Yang for generating the MEFcGAS KO-GFP-cGAS and MEFcGAS KO-GFP-ΔN160cGAS cell lines, L. Sun for helping with cGAMP bioassays, C. Zhang for helping with live-cell imaging, and S. Banani for discussions regarding liquid-liquid phase separation of proteins. Funding: This work was supported by grants from the Lupus Research Alliance, the Cancer Prevention and Research Institute of Texas (RP120718 and RP150498), and the Welch Foundation (I-1389). Z.J.C. is an investigator of the Howard Hughes Medical Institute. Author contributions: M.D. designed and performed experiments. M.D. and Z.J.C. wrote and revised the manuscript. Competing interests: All authors declare no conflicts of interest. Data and materials availability: All data needed to evaluate the conclusions in this paper are present either in the main text or in the supplementary materials.
View Abstract

Navigate This Article