Distinct Domains Contribute to the Subcellular Localization of Human cGAS in Yeast

TL;DR

This study explores how different parts of the cGAS protein influence its location within cells using yeast as a model system.

Contribution

The study reveals how distinct domains of cGAS regulate its subcellular localization using a heterologous yeast expression system.

Findings

cGAS-eGFP localizes to ER-Mitochondria encounter structures and juxtanuclear compartments in yeast.

The N-terminal domain of cGAS prevents nuclear localization and reduces cytoplasmic aggregation.

A DNA-binding motif mutant shows increased nuclear localization of cGAS.

Abstract

Cyclic GMP-AMP synthase (cGAS) functions as a DNA sensor in the cytoplasm, triggering immune responses, but it is also translocated to the nucleus, where it is kept catalytically inactive. It consists of an unstructured N-terminal domain of around 160 amino acids, and a larger C-terminal fold comprising the catalytic and DNA-binding domains. Subcellular localization of cGAS is thought to play a key role in its regulation. Here, we make use of heterologous expression in the eukaryotic model Saccharomyces cerevisiae to study cGAS localization in a neutral cellular environment. cGAS-eGFP was mostly found in aggregates at the endoplasmic reticulum–mitochondria encounter structure (ERMES) and juxtanuclear protein quality compartments (JUNQs), although some cells displayed an association between cGAS-eGFP and the plasma membrane. The N-terminus of cGAS fused to eGFP was unable to associate with the plasma membrane by itself, but its deletion dramatically promoted nuclear localization of cGAS-eGFP and decreased cytoplasmic aggregates. A mutant in the DNA-binding Zn-thumb motif of cGAS also showed a more prominent nuclear localization. Thus, both the N-terminal and C-terminal domains of cGAS seem to cooperate to prevent nuclear localization and to maintain cytoplasmic reservoirs of the protein. Heterologous cGAS expression in yeast is a valuable tool for modeling aspects of its subcellular localization and aggregative features.

Figures

Full text

1. Introduction

Cyclic GMP-AMP synthase (cGAS) detects cytoplasmic DNA, whether exogenous (of pathogenic origin—bacterial or viral DNA) or endogenous, when it is released from the nucleus or mitochondria as a hallmark of cell damage. Its binding to DNA triggers its catalytic activity, producing a cyclic dinucleotide, cGAMP, which acts as a second messenger to activate the STING–TBK1–IRF3 axis for the production of interferons [1,2]. Thus, cGAS is considered a pathogen-associated molecular pattern (PAMP)-recognizing receptor (PRR) essential for the activation of proinflammatory pathways, especially upon viral infection or cell damage [3,4].

Human cGAS is a 522-amino-acid protein that comprises an unstructured N-terminal domain (amino acids 1–159), and a longer C-terminal domain (amino acids 160–522) that contains an NTase core structural domain, an overlapping male abnormal 21 (Mab21) domain, and a dsDNA-binding domain. The NTase fold and the Mab21 domain configure the catalytic domain, which comprises the typical catalytic residues (Glu225, Asp227, and Asp319) [5]. The Mab21 domain includes a canonical Zn-binding motif that is important both for cGAS dimerization and for dsDNA binding [6,7]. The dsDNA-binding domain fold is involved in liquid–liquid phase separation (LLPS) phenomena, mediated by both the disordered N-terminus and stretches rich in basic residues around the catalytic site that favor cGAMP production upon recognition of DNA [8,9]. The same DNA-binding fold has been found to mediate interactions with nucleosomes [10]. Thus, the binding of cGAS to dsDNA, especially long chains, but also to some extent other nucleic acids, promotes both its catalytic activation and LLPS phenomena that contribute to boosting cGAMP production and amplifying the signal.

In spite of the specific function of cGAS in cytoplasmic DNA recognition, activating innate immunity, cGAS is translocated to the nucleus, where it is involved in different functions. Only recently have diverse features related to the molecular mechanisms that regulate cGAS localization, nuclear translocation, and the inhibition of its catalytic activity within the nucleus been unveiled [11,12]. The relevance of cGAS localization in its regulation is a matter of debate. In mammalian cells, it was initially detected at the cytosol [1], but it is mostly found in the nucleus, where it is inactivated by interaction with nucleosomes [13,14,15]. cGAS has been described as being kept in an inactive state in the cytosol by interactions with the plasma membrane (PM) [16], but has also been described as localizing to stress granules [17], lysosomes [18], mitochondria [19], and micronuclei [20], sites where it may recognize nucleic acids and become activated. Moreover, cGAS shows distinct localization in different tissues or cell lines under study [16]. Here, we use the heterologous expression of human cGAS in the single-celled eukaryotic model S. cerevisiae to learn about the intrinsic contribution of the different domains of cGAS to its localization in a simple unbiased cellular system.

2. Materials and Methods

2.1. Saccharomyces cerevisiae Strains and Growth Conditions

The S. cerevisiae strain used in this study was YPH499 (MATa ade 2-101 trp1-63 leu2-1 ura3-52 his3-200 lys2-801) [21]. YPD (1% yeast extract, 2% peptone, and 2% glucose) agar was the nonselective medium used for growing yeast. The TWY779 strain (MATα LSP1-RFPmars::NAT ura3 trp1 leu2 his3 ade2 can1-100) [22], a gift of T. Walther, was used for visualization of Lsp1. Selective synthetic dextrose (SD) medium (0.17% yeast nitrogen base without amino acids, 0.5% ammonium sulfate, 2% D-glucose, 0.12% drop-out supplements mixture, supplemented with the appropriate amino acids and nucleic acid bases for auxotrophic marker-based plasmid maintenance) was used for plasmid selection and maintenance. Synthetic raffinose medium (SR) and synthetic galactose (SG) medium were SD with 1.5% (w/v) D-raffinose or 2% (w/v) D-galactose, respectively, instead of D-glucose. GAL1-driven protein induction experiments were performed by growing transformants in SR liquid medium for 18 h at 30 °C and then refreshing the cultures to a final concentration of 0.6 × 10^7^ cells/mL in SG liquid medium and incubating under shaking for 5 h at 30 °C. The SR and SG liquid media lacked the appropriate amino acids to maintain plasmids.

2.2. Molecular Techniques and Construction of Yeast Expression Plasmids

Transformation of S. cerevisiae by the lithium acetate protocol and associated molecular techniques were performed by standard procedures. The Escherichia coli strain DH5α was used for molecular biology. The genes CGAS WT, CGAS Nt, CGAS Ct, CGAS^C396^^/^^397A^, CGAS^R71^^/^^75E^, CGAS^R71^^/^^75E C396^^/^^397A^, and CGAS Ct^C396^^/^^397A^ were amplified by standard PCR procedures from the pLenti-CMV-cGAS-HA, pLenti-CMV-cGASΔN-HA, pLenti-CMV-cGAS N-HA, pLenti-CMV-cGAS C396/7A-HA, pLenti-CMV-cGASΔN C396/7A-HA, pLenti-CMV-cGAS R71/75E-HA, and pLenti-CMV-cGAS R71/75E C396/7A-HA plasmids (a gift from J. Kagan, Boston Children’s Hospital, MA, USA) using the primers cGASFL_attB1 (B)_UP, cGASdeltaN_attB1, (B)_UP, and HA_attB2 (L) with the attB flanking sites. Primer sequences are listed in Supplementary Table S1. Invitrogen Gateway Cloning (Addgene kit #1000000011, RRID: SCR_005371) was used to clone the attB-flanked PCR products into a pDONR221 vector by Gateway BP Clonase II reaction to generate entry clones. The inserts from entry clones were subcloned into a pAG425GAL1-ccdB-EGFP, pAG424GAL1-ccdB-EGFP, or pAG424GAL1-ccdB-DsRed destination vectors by LR recombination reaction to obtain the expression clones pAG424-GAL1-cGAS-HA-EGFP, pAG425-GAL1-cGAS-HA-EGFP, pAG425-GAL1-cGAS(Ct)-HA-EGFP, pAG425-GAL1-cGAS(Nt)-HA-EGFP, pAG424-GAL1-cGAS-HA-DsRed, pAG425-GAL1-cGAS^C396/397A^-HA-EGFP, pAG425-GAL1-cGAS^R71/75E^-HA-EGFP, pAG425-GAL1-cGAS^R71/75E C396/397A^-HA-EGFP, and pAG425-GAL1-cGAS(Ct)^C396/397A^-HA-EGFP. The cGAS Ct protein sequence corresponds to amino acids 160 to 522, and the cGAS Nt protein sequence corresponds to amino acids 1 to 159.

The plasmids pAG424-Mdm34-DsRed [23], pESC-LEU-chFP-Ubc9ts [24], and pRS426-GFP2XPH(PLCδ) [25] have been described previously.

2.3. Cell Microscopy Techniques and Image Processing

For in vivo bright-field differential interference contrast (DIC) or fluorescence microscopy to visualize green or red fluorescent proteins, transformants were cultured and induced in galactose as described above, and then harvested by centrifugation at 2500 rpm for 3 min and observed directly under either a Leica DMi8 microscope (Leica Microsystems; RRID: SCR_026672) or an Eclipse TE2000U inverted fluorescence microscope (Nikon, RRID: SCR_023161). In the former, images were acquired with a Leica K8 charge-coupled-device camera using LAS X software v5.3.1 (Leica, Wetzlar, Germany; RRID: SCR_013673), and further processed using Fiji v1.54p (ImageJ; Bethesda, MD, USA; RRID: SCR_002285). In the latter, digital images were acquired with an Orca C4742-95-12ER camera (Hamamatsu Photonics, Hamamatsu, Japan; RRID: SCR_017105) and HCImage software v2.0.1.16 (Hamamatsu Photonics, Hamamatsu, Japan; RRID: SCR_015041), and further processed by Adobe Photoshop (RRID: SCR_014199). For the visualization of nuclei, cells were resuspended in 100 µL of culture medium and incubated with DAPI (4′,6-diamidino-2-phenylindole) (ThermoFisher Scientific, Waltham, MA, USA; RRID: SCR_008452) at a final concentration of 10 µg/mL for 5 min at room temperature and in the dark. After incubation, the cells were harvested again, washed three times with PBS, and observed.

2.4. Statistical Analyses

Data were analyzed using Microsoft Excel (RRID: SCR_016137) Microsoft Corp., Edmond, WA, USA, and IBM SPSS v28.0.1.1 statistics (RRID: SCR_016479) SPSS Inc., Chicago, IL, USA. All data sets were tested for normality using the Shapiro–Wilk test. When a normal distribution was confirmed, Student’s t-test was used for the statistical comparison of two samples. In all cases a significance level of 0.05 was selected. The asterisks in the figures correspond to the following p-values: * p < 0.05, ** p < 0.01, and *** p < 0.001. Experiments were performed as biological triplicates on different clones, and data with error bars are represented as mean ± standard deviation.

3. Results

3.1. cGAS-eGFP Forms Cytoplasmic Aggregates in Yeast Cells, with a Minority of Cells Showing Plasma Membrane Localization

In order to study the performance of the human cGAS protein in the heterologous S. cerevisiae model, we cloned the cDNA encoding cGAS C-terminally fused to either eGFP or DsRed in its C-terminus under the control of the inducible GAL1 promoter. Expression of cGAS in yeast from this strong galactose-inducible promoter was verified by immunoblotting (Supplementary Figure S1a,c). Human cGAS was well tolerated, as its expression did not have any effect on yeast growth (Supplementary Figure S1b,d). Since both the eGFP and DsRed fusions behaved similarly, unless otherwise noted, all experiments henceforth refer to eGFP fusions.

Mammalian cGAS comprises two domains: an N-terminal intrinsic disordered region (IDR) and a C-terminal DNA-binding catalytic fold (Figure 1a). Since many aspects related to its subcellular localization remain controversial, we proceeded to study the localization of cGAS in yeast, as well as the contribution of both domains. In yeast cells, cGAS had a nucleocytoplasmic distribution with two features: most cells (96.11% ± 3.23) presented at least 2–3 distinct cytoplasmic spots, but in a minority of cells (16.56% ± 3.58), usually those with a higher fluorescent intensity, spots associated with the cell periphery were observed (Figure 1b–e and Figure 2b,c; for confocal planes of a representative cell, see Supplementary Figure S2a). To verify that the latter cGAS spots were associated with the PM, we counterstained with a protein known to associate with the PM in yeast, the eisosome core component Lsp1. Although cGAS-eGFP covered wider areas than Lsp1 at the cell periphery, cGAS and Lsp1 signals were spatially concordant, confirming PM association (Figure 1b). Co-expression of cGAS-eGFP with markers of cytoplasmic structures and organelles tagged with red fluorescent proteins showed that some spots co-localized with Mdm34-DsRed, suggesting that cGAS was accumulating at the ER–mitochondria encounter structure (ERMES) (Figure 1c), while at least one spot per cell co-localized with mCherry-Ubc9ts, a marker of juxtanuclear protein quality compartments (JUNQs) (Figure 1d).

3.2. The cGAS N-Terminal Disordered Extension Is Dispensable for Plasma Membrane Attachment

The IDR of cGAS, comprising the first 160 amino acids that precede the structurally defined C-terminus, has been shown to contribute to DNA binding, DNA-induced assembly, and the proper fluidity of LLPS droplets [1,9,26], but it has also been found necessary and sufficient for binding to the PM upon recognition of PtdIns(4,5)P_2_ [16]. Indeed, the pattern of PM localization observed in a minor population of cGAS-eGFP-expressing yeast cells is reminiscent of our previous observations with TIRAP, another PtdIns(4,5)P_2_-binding protein [27]. To assess whether the observed subcellular localization pattern in the heterologous model relied on the N-terminal IDR, we followed the rationale of Barnett et al. [16] and split cGAS to generate two truncated versions, cGAS(Nt) (aa 1–159) and cGAS(Ct) (aa 160–522), fused to the fluorescent proteins (Figure 1a). Like the full-length protein, the fusions produced were not toxic for the yeast cell (Supplementary Figure S1b,d), and their expression was detected by immunoblotting (Supplementary Figure S1a,c). Peculiarly, as previously reported in a macrophage cell line [16], cGAS(Ct) was produced in yeast at much lower levels than either full-length cGAS or cGAS(Nt) (Supplementary Figure S1a,c), suggesting that the IDR is important for the intrinsic stability of the protein. However, although PM localization was significantly reduced by 1/3, 5.19% (±0.97) of the yeast cells expressing cGAS(Ct)-eGFP still presented this pattern (Figure 1e and Figure 2c). Furthermore, cGAS(Nt) displayed a diffused cytosolic pattern, failing either to produce any sort of cytoplasmic spots or to associate with the PM, as determined by the lack of overlap with a PH-PLC (Figure 1f). Thus, in yeast, the IDR of cGAS partially contributes to localization clues at the PM, but it is not sufficient to drive PM localization by itself.

3.3. Deletion of the N-Terminus of cGAS Promotes Nuclear Localization to the Detriment of Cytoplasmic Aggregates

While analyzing the localization of cGAS(Ct)-eGFP in yeast, we noticed that, in contrast to full-length cGAS-eGFP, most cells showed a bright compartment consistent with the nucleus (see Figure 1e). We stained yeast cells with DAPI, thus confirming that deletion of the N-terminal IDR of cGAS caused the nuclear accumulation of this protein (Figure 2a). The percentage of cells displaying a clear nuclear enrichment as a major localization increased from 3.39% (±2.79 SD) in cGAS-eGFP to 74.32% (±5.21 SD) in cGAS(Ct)-eGFP (Figure 2b). Conversely, the overall count of cells with visible cytoplasmic aggregates was reduced from 96.11% (±3.23 SD) in cGAS-eGFP to 22.89% (±6.34 SD) in cGAS(Ct)-eGFP (Figure 2c). These results indicate that the disordered N-terminal extension of cGAS prevents the localization of cGAS in the nucleus while favoring the formation of cytoplasmic aggregates.

3.4. The DNA-Binding Zn Finger Motif Partially Contributes to All Localization Patterns

To gain insight into the contribution of the DNA-binding motifs to cGAS subcellular localization in yeast, we constructed a mutant cGAS allele defective in two consecutive Cys residues at the zinc-binding site of the thumb, reported as essential for DNA binding (C396A/C397A) [6], both in full-length cGAS and in its Δ1–159 version [cGAS(Ct)]. The mutant cGAS C396A/C397A fusions were not deleterious for yeast growth, and were expressed at the same levels as their WT and Ct counterparts, with the truncated version being less abundant in yeast lysates (Supplementary Figure S1a,b). Peculiarly, the cGAS C396A/C397A mutant behaved with a trend similar to the cGAS(Ct) truncation: accumulation at the nucleus was significantly enhanced (96.54% ± 0.92 cells with nuclear signal higher than cytoplasmic background), although with a lower percentage of cells showing exclusive nuclear localization (39.08% ± 4.58 cells) than when cGAS(Ct) was expressed (Figure 2b). The Zn thumb mutation also had a significant impact on the appearance of cGAS cytoplasmic spots, decreasing the population of cells showing this as a predominant pattern to 1.71% (±1.5% SD) (Figure 2c). Relative to cGAS(Ct), the truncated cGAS(Ct) C396A/C397A mutant displayed a significantly decreased proportion of cells with exclusive nuclear localization (42.41% ± 1.59), and a higher number of cells with a combination of nuclear localization and cytoplasmic spots (55.87% ± 2.14) (Figure 2b,c). However, unlike the truncated cGAS(Ct) mutant, the C396A/C397A mutant did not seem to significantly affect the minor percentage of cells with cGAS localized to the PM (9.41% ± 4.66), which was otherwise clearly diminished in cGAS(Ct) C396A/C397A (0.53% ± 0.46) (Figure 2d). This PM association pattern fully overlapped with that of co-expressed TIRAP-mCherry (Supplementary Figure S2b), suggesting PtdIns(4,5)P_2_ binding, as previously reported [16,27]. Thus, the integrity of the DNA-binding Zn thumb motif of cGAS seems to contribute to form cytoplasmic aggregates, to prevent nuclear translocation, and to interact with the PM, but to a lesser extent than its N-terminal IDR.

3.5. Basic Residues Within the N-Terminus of cGAS Cooperate with the C-Terminus in Determining Its Subcellular Localization in Yeast

Barnett et al. [16] defined the basic residues R71 and R75 in human cGAS as essential for interaction with PtdIns(4,5)P_2_ in mammalian cell lines. We reproduced the R71E R75E double mutation to study the contribution of these residues to the phenomena observed in yeast. Except for a mild, although significant, enhancement of cells with an exclusive nuclear pattern (16.27% ± 5.36) (Figure 2b), the cGAS R71E/R75E mutant did not significantly differ to the WT cGAS concerning PM localization and aggregate formation (Figure 2c,d). Thus, the contribution of these residues to the role of the N-terminal extension of cGAS in preventing nuclear localization is partial. However, when we combined the R71E/R75E mutations with the C396A/C397A mutations, the cGAS R71E/R75E C396A/C397A mutant did phenocopy the cGAS(Ct) C396A/C397A version, showing a high proportion of cells with clear nuclear cGAS localization (60.55% ± 4.18), as well as a reduction in cytoplasmic aggregates (39.09% ± 3.96) that were intermediate between cGAS(Ct) and cGAS(Ct) C396A/C397A (Figure 2b,c). Also, like cGAS(Ct) and cGAS(Ct) C396A/C397A, the proportion of cells with PM localization (2.63% ± 0.90) was significantly lower than that of wild-type cGAS or the R71E/R75E mutant (Figure 2d). Thus, the mutation of the Arg residues at the N-terminus of cGAS seemed to have an additive effect to that of the Zn thumb motif in terms of nuclear vs. cytoplasmic aggregate localization, which was even more effective than the deletion of the whole N-terminus. None of the mutants generated led significantly to a diffuse localization pattern like that of cGAS(Nt)-eGFP (Figure 2e).

4. Discussion

The subcellular localization of cGAS in mammalian cells and its relevance for detecting both pathogenic and intracellular nucleic acids is a matter of debate. We made use of the heterologous S. cerevisiae model in search of universal spatial clues that may shed light on this topic. Current knowledge advocates for a preeminent role of cGAS as a DNA-sensing PRR at the cytoplasm, whereas its translocation to the nucleus would obstruct this function [2]. Cytoplasmic cGAS aggregates are purportedly related to its recognition of DNA and its activation [3,9]. Developing yeast models to study cGAS could provide useful platforms to test potential inhibitors that may eventually be used as immunomodulators or anti-inflammatory or antitumor drugs [28]. Targeting the allosteric site of cGAS with compounds considered to be protein condensation inhibitors reduces its aggregative properties, eliminating cGAS LLPS in vivo [29]. So far, we have been unable to detect cGAMP by mass spectroscopy in yeast lysates overproducing different versions of cGAS, hindering the design of functional analyses in the yeast heterologous model. Nevertheless, the presence of cytoplasmic condensates of cGAS in yeast may indicate that the protein maintains its aggregative properties in the system. Specifically, we found that cGAS accumulates both at the yeast ERMES and JUNQs. The latter localization in a proteostatic stress compartment may reveal a misfolded protein fraction [30], and may be artefactual. In contrast, we have previously observed localization to the ERMES in other heterologous self-aggregating proteins involved in innate immune signaling, like MyD88 [31], which could be related to the LLPS properties of these signaling molecules [32]. Further studies will be required to elucidate whether mitochondria-ER contact sites are able to seed LLPS condensates of proteins that are prone to this behavior.

It has been proposed that cGAS may remain attached to the PM in an inactive state in lymphoid cells, thus preventing aberrant activation by endogenous DNA species that may leak from nuclei or mitochondria in physiological conditions. DNA binding might release cGAS from this location, facilitating the formation of LLPS droplets in the cytoplasm [16], which is important for signal establishment and amplification. Indeed, it has been shown that the transition from solid aggregates to LLPS droplets upon DNA recognition is essential for cGAS’s catalytic activity [9]. The N-terminal IDR of cGAS has been described as crucial for both PM recognition via PtdIns(4,5)P_2_ binding and, recently, for the transition to LLPS [16,26]. In yeast, we found that cGAS indeed decorated the PM, although this is restricted to a subset of cells. This localization was preferentially observed in cells with a high cGAS-eGFP signal, suggesting a threshold effect, but further studies will be necessary to understand whether this particular localization may require molecular clues additional to the sheer presence of PtdIns(4,5)P_2_. In any event, the N-terminus of cGAS was not sufficient to recognize the yeast PM, in contrast to that of TIRAP, for instance [27,31]. Furthermore, in yeast, the IDR was only partially necessary for the PM localization of cGAS, and removal of the Arg residues described as involved in PM recognition exhibited a very minor contribution, suggesting that molecular determinants in the Ct are also contributing to this localization. Actually, the combination of either the deletion of the N-terminus or the R71E/R75E mutation with that of Cys residues at the DNA-binding Zn thumb essential for catalytic activity did have a dramatic impact on PM association in yeast.

Both the removal of the N-terminal extension of cGAS and the mutation of C396A/C397A led to a decrease in the formation of cytoplasmic puncta in yeast, supporting the view that both regions are important for the formation of aggregates. The N-terminal extension on its own was not sufficient to assemble aggregates. However, again, the combination of either an N-terminal truncation or Arg-to-Glu mutations with the substitution of the Cys residues at the Zn thumb led to different levels of aggregate decrease, suggesting cooperative mechanisms between both domains for their formation. Interestingly, the ability of the studied cGAS mutants to concentrate at the nucleus seemed to negatively correlate with their ability to form cytoplasmic aggregates, reinforcing the idea that the formation of cytoplasmic aggregates prevents nuclear translocation. Thus, yeast seems to recapitulate the nuclear translocation of cGAS(Ct) observed by Gentili et al. in mammalian cells [33], underlining the role of the IDR in the regulation of cGAS localization, although cGAS(Nt) failed to display this predominantly nuclear localization in yeast, as reported by these authors in mammalian cells. Regulation of the nuclear translocation of cGAS may be crucial to understanding its function [11]. The fact that the C396A/C397A mutation also promoted cGAS nuclear localization indicates a putative role for dimerization and DNA-binding in cytosolic retention. This mutation should be expected to affect Zn coordination for DNA recognition and the subsequent conformational changes that lead to catalytic activation [6], but the basic residues involved in interactions with histones in the nucleosome [34] should be intact, allowing for nuclear retention. It is remarkable that nuclear localization of these mutant versions of cGAS is preserved in yeast, despite its evolutionary distance, which suggests that universal cues of nucleosome structure and/or histone modification are involved in this behavior. In sum, we show that the heterologous yeast model can be exploited to understand diverse aspects of cGAS’s intrinsic molecular properties, like PM association, the formation of cytoplasmic aggregates, and nuclear retention.

5. Conclusions

In this work we provide what are, to our knowledge, the first studies on human cGAS expression in a heterologous yeast model and perform mutagenesis to study the determinants of its aggregative, plasma membrane binding and nuclear translocation. We conclude that, in an unbiased cellular environment, cGAS mainly forms cytoplasmic aggregates but is occasionally found at the plasma membrane. Removal of its unstructured N-terminal extension reduced aggregates and shifted cGAS to the nucleus, but did not abrogate its minor cGAS plasma membrane localization. Likewise, mutation of residues at the Zn finger involved in DNA binding also favored nuclear cGAS localization.

Thus, our experiments on cGAS mutants in yeast revealed that both its N-terminal extension and its DNA-binding domain contribute to the cytoplasmic localization of the protein, and that the presence of cytoplasmic aggregates and the nuclear localization of cGAS correlate negatively.