Human cGAS Activates LINE-1 to Trigger MAVS-Dependent Cell Aging

HCA2‐hTERT, HeLa, mouse skin fibroblasts (MSFs), HEK‐293FT cells, and their derived cell lines were maintained in Dulbecco's Modified Eagle's Medium (Corning, Cat. #10‐013‐CVR) supplemented with 10% fetal bovine serum (Gibco, Cat. #10270‐106), 1% nonessential amino acids (Gibco, Cat. #11140‐050), and 1% penicillin–streptomycin (Gibco, Cat. #15140‐122). All cells were cultured at 37°C in a humidified incubator (Thermo Scientific HERAcell 240i) with 5% CO₂ and were routinely tested for mycoplasma contamination using PCR. HCA2‐hTERT and MSFs were transfected using the Lonza 4D Nucleofector system with program DT‐130, while HeLa cells were transfected using program CN‐114. HEK‐293FT cells were transfected via polyethylenimine (PEI)‐mediated method.

The full‐length coding sequences (CDS) of human cGAS, CTCF, and RUNX3 were amplified by PCR using HeLa cDNA as template. The CDS of mouse cGAS, CTCF, and RUNX3 was amplified using MSF cDNA as template. All amplified CDS were cloned into HA‐tag or Flag‐tag vector and validated by Sanger DNA sequencing. shRNA sequences were listed in Table. The following primary antibodies were used: anti‐cGAS (ABclonal, Cat. #A8335), anti‐HA (Cell Signaling Technology, Cat. #3724), anti‐Flag (ABclonal, Cat. #AE005), anti‐p21 (ABclonal, Cat. #A19094), anti‐MAVS (ABclonal, Cat. # A25005), and anti‐GAPDH (Proteintech, Cat. #60004). FITC‐Phalloidin (Cat. #RM02836) was purchased from ABclonal. S1

HeLa cells were co‐transfected with the L1‐5′ UTR‐firefly luciferase reporter (Athanikar et al. 2004; Van Meter et al. 2014) and a vector encoding Renilla luciferase as an internal control. Twenty‐four hours post‐transfection, cells were harvested and lysed. Luciferase activity was measured using a GloMax Luminometer (Promega, Cat. #E5311). Relative L1‐5′ UTR promoter activity was calculated as the ratio of Firefly luciferase luminescence to Renilla luciferase luminescence.

Cells were lysed and total RNA was extracted using the RNASimple Total RNA kit (Tiangen, Cat. #DP419) 24 h post‐transfection. Subsequently, 1 μg of RNA was reverse‐transcribed into cDNA using the TransScript II Reverse Transcriptase kit (Trans, Cat. #AT311‐02). The resulting cDNA was used as the template for quantitative real‐time PCR using SYBR Green qPCR Mix (Roche, Cat. #04913914001) on a ViiA 7 Real‐Time PCR System (Applied Biosystems). RT‐qPCR was performed using three independent biological replicates, yielding consistent findings. Data from one representative replicate are shown in the figures. Primer sequences are listed in Table S2. The primer sequences designed to preferentially amplify elements of the full‐length human‐specific L1HS subfamily were previously reported (De Cecco et al. 2019).

ChIP assays were conducted as previously described (Chen et al. 2023; Jiang et al. 2025). Briefly, to assess cGAS enrichment on the CTCF or RUNX3 promoters, HEK‐293FT cells were transfected with empty vector or HA‐tagged cGAS vector. Cells were harvested 24 h post‐transfection for ChIP analysis using HA‐specific antibodies or control IgG. Precipitated DNA from anti‐HA and IgG samples, along with input samples, was quantified by real‐time PCR using the 2^−∆∆CT method. Relative enrichment was calculated as the ratio of immunoprecipitated DNA to input chromatin, with the value obtained for the IgG sample or empty vector‐transfected sample set to 1. Primer sequences for CTCF and RUNX3 promoters are listed in Table S2.

cGAS ChIP‐seq data were previously reported in Gentili et al. (2019). Filtered human cGAS peaks from the GSE125475↗ dataset were annotated using the annotatePeaks.pl. script from HOMER software (version 4.11). The full‐length L1 annotation file was obtained from the L1base2 database.

cGAS ChIP‐seq data were previously reported in Gentili et al. (2019) and are available under accession number GSE125431↗. Reads were trimmed with fastp 0.23.4, mapped to GRCh38 with Bowtie2 2.5.4 (very‐sensitive, k = 100), and multimappers were reallocated by Allo 1.2.0. Duplicates were marked with samblaster 0.1.26 on name‐sorted BAMs, then coordinate‐sorted and indexed with sambamba 1.0.1. Replicates were merged per condition. Genome‐wide log₂(cGAS/NLS) tracks were generated with deepTools bamCompare using SES scaling, and signal over L1 elements was extracted with computeMatrix and visualized with plotProfile.

Using GTEx (V8) data, TPM values were log‐transformed as ln (TPM + 0.00001). Pearson correlation analysis between CGAS and CTCF/RUNX3 expression was performed using R (version 4.2.2), with visualizations generated using the ggplot2 package (version 3.5.1).

Cells were irradiated with 6 Gy X‐ray at 24 h post‐transfection. Six days postirradiation, cells were harvested, washed with PBS, and fixed for 5 min at room temperature using a solution containing 2% formaldehyde and 0.2% glutaraldehyde. After three PBS washes, cells were incubated overnight in staining solution (1 mg/mL X‐gal, 40 mM citric acid/sodium phosphate buffer (pH 6.0), 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 150 mM NaCl, 2 mM MgCl₂) at 37°C. The next day, stained cells were imaged using an inverted microscope and quantified in a blinded manner.

Cells were washed once with PBS, fixed with 4% paraformaldehyde in PBS for 20 min at room temperature, and then washed twice with PBS. Permeabilization was performed using 0.3% Triton X‐100 in PBS for 10 min at room temperature, followed by two PBS washes. Cells were then stained with FITC‐phalloidin working solution (1:200 dilution) for 20 min at room temperature in the dark and washed twice with PBS. Finally, cells were mounted using a DAPI‐containing mounting medium.

No statistical methods were used to predetermine sample size. Sample sizes were determined based on experience with similar experiments. Exact p values are reported in the figures. Analysis was performed using GraphPad Prism (version 10) software.

While our earlier study demonstrated that cGAS promotes degradation of L1 ORF2p at the protein level, its role in L1 transcription was unknown. RT‐qPCR analysis in HeLa cells transfected with human cGAS revealed a dose‐dependent increase in L1 mRNA levels (Figure 1A,B). This effect was confirmed using alternative primer pairs (Figure S1). Conversely, CRISPR/Cas9‐mediated depletion of CGAS significantly reduced L1 mRNA (Figure 1C,D), and shRNA‐mediated gene silencing yielded similar results (Figure 1E,F). We also confirmed the effects of cGAS on L1 mRNA in immortalized human foreskin fibroblasts, HCA2‐hTERT, where cGAS overexpression elevated, and its depletion suppressed, L1 mRNA levels (Figure 1G–J). To determine if cGAS acts through influencing the L1 promoter activity, we employed an L1 5′ UTR‐Luciferase reporter system (Athanikar et al. 2004; Van Meter et al. 2014; Figure 1K). Both CRISPR/Cas9‐mediated depletion (Figure 1L) and shRNA‐mediated knockdown (Figure 1M) of cGAS significantly repressed L1 5′ UTR promoter activity. Crucially, using the enzymatically inactive cGAS mutant D319A, we found that both wild‐type and mutant cGAS promoted L1 mRNA levels and L1 5′ UTR promoter acitivity (Figure 1N–P). These results indicate that cGAS‐mediated transcriptional activation of L1 is independent of its canonical function.

Previous study has reported nuclear cGAS enrichment on LINE DNA repeats (Gentili et al. 2019). Interestingly, although not statistically significant, our reanalysis revealed a trend of cGAS preferentially binding to evolutionarily old, retrotranspositionally inactive L1 subfamilies, such as L1ME3A and L1MB3, over retrotranspositionally active full‐length L1 (Abrusan and Krambeck 2006; Figure 2A). When analyzing the cGAS binding pattern on these L1 sequences, we did not observe specific binding of cGAS to the 5′ UTR promoter region of either retrotranspositionally active L1 or the L1ME3A and L1MB3 subfamilies (Figure 2B, Figure S2). These data suggested that the transcriptional regulation of L1 by cGAS is likely indirect. Therefore, we shifted our focus to searching the literature for known transcriptional regulators of L1. We found that a number of both positive regulators (which promote L1 transcription) and negative regulators (which repress L1 transcription) have been reported (Figure 2C). Analysis of a published dataset (Gluck et al. 2017) identified altered expression of several L1 regulators upon cGAS depletion. Among these, the positive regulators RUNX3 and CTCF (Sun et al. 2018; Yang et al. 2003) (downregulated upon Cgas depletion), and the negative regulator DUSP1 (Briggs et al. 2021) (upregulated upon Cgas depletion), emerged as candidate mediators for cGAS‐induced L1 mRNA changes (Figure 2D). Subsequent RT‐qPCR in cGAS‐depleted HeLa cells confirmed significant reductions in CTCF and RUNX3 mRNA, consistent with RNA‐seq data, while no significant upregulation of DUSP1 was observed (Figure 2E). Consistently, human cGAS overexpression increased the mRNA levels of both CTCF and RUNX3 (Figure 2F–H). In silico analysis of the GTEx dataset further validated positive correlations between cGAS and CTCF or RUNX3 expression across human tissues and sexes (Figure 2I–L, Figure S3), suggesting the impact of cGAS on the expression of CTCF and RUNX3 may be conserved in humans.

Notably, cGAS depletion reduced L1 mRNA levels, and this reduction was rescued by restoring CTCF or RUNX3 expression (Figure 2M–P). Similarly, CTCF or RUNX3 overexpression rescued the diminished L1 5′ UTR promoter activity in cGAS‐depleted cells (Figure 2Q,R). Furthermore, DUSP1 knockdown in cGAS‐depleted cells did not rescue the reductions in L1 mRNA levels or L1 5′ UTR promoter activity (Figure S4A,B). These data demonstrated that DUSP1 is not responsible for cGAS‐mediated activation of L1 transcription. Taken together, our results revealed that cGAS transcriptionally regulates L1 through CTCF and RUNX3. A recent study has reported that cGAS could be recruited to the promoter region to regulate gene expression (Zhang et al. 2024). Supporting this mechanism, our ChIP experiments revealed that cGAS enriched at the promoters of the CTCF and RUNX3 genes, independent of its enzymatic activity (Figure 2S,T, Figure S4C,D), suggesting potential direct regulation of these targets. Due to the lack of commercial antibodies against cGAS suitable for ChIP, we could not confirm endogenous cGAS binding to these promoters. This validation should be addressed in future research. Previous studies (Sun et al. 2018; Yang et al. 2003) and analysis of publicly accessible data further confirmed the enrichment of both CTCF and RUNX3 at the 5′ UTR of the L1 sequence (Figure S4E,F). Moreover, while cGAS significantly enhanced the activity of the full‐length L1 5′ UTR promoter, this effect was markedly diminished upon deletion of either the +83 to +101 region (Yang et al. 2003) or the +390 to +526 region (Sun et al. 2018), which correspond to the binding sites for RUNX3 and CTCF respectively (Figure 2U,V). Additionally, knocking down CTCF or RUNX3 abolished the increase in L1 mRNA induced by cGAS overexpression (Figure 2W), confirming these factors are essential mediators of cGAS‐dependent L1 regulation.

When analyzing the effect of mouse cGAS overexpression on L1 expression in mouse skin fibroblasts, we unexpectedly observed no induction of L1 mRNA; instead, mouse cGAS induced a mild but significant inhibitory effect on L1 transcription. Similar results were obtained using three additional primer pairs (Figure 3A). This discrepancy prompted us to test whether human cGAS could promote L1 expression in mouse cells. However, human cGAS also failed to induce L1 mRNA (Figure 3B). In contrast, mouse cGAS increased L1 mRNA levels in human cells (Figure 3C). These data suggest that the distinct effects of cGAS on L1 expression are independent of functional divergences in cGAS itself, but rather dependent on cellular context (Figure 3D).

We therefore hypothesized that intracellular signaling differences between species are critical determinants. We found that human cGAS overexpression in mouse cells retained the capacity to induce mouse CTCF and RUNX3 expression, mirroring the effect of mouse cGAS (Figure 3E,F). These findings indicate that the critical divergence point likely resides downstream of CTCF/RUNX3 induction, within the CTCF/RUNX3–L1 regulatory axis.

To directly compare the roles of CTCF and RUNX3 across species, we overexpressed human or mouse versions of these factors in both human and mouse cells. Consistent with previous reports, human CTCF and RUNX3 enhanced L1 mRNA levels in human cells (Figure 3G, Figure S5A). Notably, they also promoted L1 transcription in mouse cells (Figure 3H, Figure S5A,B). Conversely, overexpression of mouse CTCF and RUNX3 failed to induce L1 transcription in either human or mouse cells (Figure 3I,J, Figure S5C–E), demonstrating that the regulatory axis linking CTCF/RUNX3 to L1 expression is not conserved.

Why is this regulatory axis not conserved? One possibility is evolutionary divergence, where sequence variability—particularly in the CTCF/RUNX3 amino acid residues—affects their binding affinity to L1. Protein sequence alignment revealed that while the DNA‐binding domains of CTCF and RUNX3 are highly conserved between human and mouse, sequence variability exists in their C‐terminal regions (Figure S6). Whether these divergences contribute to different L1 regulation remains to be investigated. Variations in the L1 5′ UTR region could also contribute. However, since mouse CTCF/RUNX3 failed to promote L1 transcription even in human cells (Figure 3I), this factor likely plays a minor role in the observed divergence. It is also noteworthy that while mouse cGAS inhibits L1 mRNA, mouse CTCF/RUNX3 overexpression has no significant effect (Figure 3A,J), suggesting the involvement of additional regulators in cGAS‐mediated L1 repression in mouse cells. These aspects remain unresolved and warrant further investigation.

Cytosolic RNA activates RNA‐sensing pathways, triggering the expression of interferons and inflammatory factors, many of which are SASP factors (Li, Zhu, et al. 2024). To assess whether cGAS‐mediated elevation of L1 mRNA similarly exacerbates the SASP and potentially cellular senescence, we depleted MAVS, a critical adaptor protein in the RNA‐sensing pathway, in HCA2‐hTERT cells (Figure 4A, Figure S7A). Consistent with the established role of cGAS in facilitating senescence, we found that human cGAS overexpression induced a remarkable increase in senescent cells (Yang et al. 2017), assayed by β‐Galactosidase staining. However, depletion of MAVS significantly abolished this effect in human cells (Figure 4B,C), indicating that cGAS‐triggered cellular senescence is partially dependent on MAVS. As expected, Mavs depletion did not affect cGAS‐overexpression‐induced cellular senescence in mouse cells (Figure S7B–D). Analysis of additional reported senescence markers, such as cell enlargement and upregulation of antiapoptotic factor BCL2 (Suryadevara et al. 2024), yielded consistent results (Figure 4D–F).

SASP factors comprise a panel of secretory interferons, cytokines, chemokines, and proteases. While cGAS has been identified as a key mediator of the SASP (Gluck et al. 2017; Victorelli et al. 2023; Yang et al. 2017), we sought to determine whether the RNA‐sensing pathway also contributes partially to this cGAS‐mediated SASP induction. To assess this, we compared the effect of cGAS overexpression on the mRNA levels of key SASP factors in control cells versus MAVS‐depleted cells. cGAS overexpression significantly increased the expression of IFNB and several critical SASP factors, including IL1A, IL8, CCL2, CXCL1, and MMP3. MAVS depletion significantly attenuated this induction in human cells but not in mouse cells (Figure 4G–L, Figure S7E–H). Western blot analysis revealed that cGAS overexpression also significantly upregulated the classical senescence marker p21. Again, this effect was abolished by MAVS depletion in human cells but not in mouse cells (Figure 4M,N, Figure S7I,J). Moreover, simultaneous depletion of both cGAS and MAVS did not further inhibit cellular senescence beyond single cGAS knockdown (Figure S8A–J), and L1 depletion abolished the pro‐senescence effect induced by cGAS overexpression (Figure S8K–Q). Overall, these data suggest that the L1–MAVS pathway mediates cGAS‐dependent senescence promotion.

Since DNA fragments generated by L1 reverse transcription also contribute to cellular senescence, we treated cells with the reverse transcriptase inhibitor 3TC. Although 3TC treatment significantly reduced senescence and inflammation as previously reported, cGAS overexpression still induced robust increases in the percentage of senescent cells and in p21 and SASP factor expression, even in the presence of 3TC (Figure). Moreover, inhibition of the DNA sensing pathway using the STING inhibitor H‐151 attenuated senescence markers, while cGAS overexpression continued to promote cellular senescence under these conditions (Figure). These data indicate that L1 promotes senescence through both DNA‐dependent and ‐independent mechanisms. To further confirm the role of L1 mRNA in activating RNA‐sensing pathways, we knocked down L1 in control and MAVS‐depleted cells. Our data showed that L1 depletion significantly attenuated IFNβ transcriptional activation and cellular senescence in control cells, but had minimal effect on these phenotypes in MAVS‐depleted cells (Figure). S9A–G S9H–O S10

Collectively, these data demonstrated that cGAS potentiates human cellular senescence, an effect that partially requires the L1‐mediated activation of the RNA‐sensing pathway (Figure 4O).