SARS-CoV-2 spike triggers TLR7-dependent endolysosome dysfunction and senescence in human astrocytes

Abstract

SARS-CoV-2 infection is associated with long-lasting neuropsychiatric and cognitive symptoms, collectively referred to as neuro-PASC. Emerging studies indicates that accelerate brain aging and cellular senescence in COVID brain could lead to altered neuroimmune responses and neurodegenerative outcomes. However, little is known about how cellular senescence is development in neuro-PASC. Here, we examined the role of spike protein subunit S1, a persistent viral antigen, in driving the development of cellular senescence in primary human astrocytes. We have demonstrated that S1 enters endolysosomes and induces endolysosome dysfunction and cellular senescence. Moreover, the multibasic motif is critical for such S1-induced damaging effects. Importantly, we identified Toll-like receptor 7 (TLR7), an endolysosome-resident pattern recognition receptor, as a critical mediator of S1-induced damaging effects. Mechanistically, S1 interacts with TLR7 at the site of the endolysosome lumen and activates p38 MAPK signaling of downstream of TLR7, which drive the development of cellular senescence. Together, these findings suggest that TLR7 mediates S1-induced endolysosome dysfunction and cellular senescence, and that TLR7 represents a therapeutic target for mitigating neuro-PASC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03586-1.

Introduction

COVID-19 patients continue to experience neurological and mental health-related symptoms ranging from cognitive impairments, such as brain fog, to mental health disorders, including anxiety and depression, beyond the acute phase of SARS-CoV-2 infection [1–10]. This is known as neurological post-acute sequelae of COVID-19 (neuro-PASC) [11]. The incidence of neuro-PASC is high [12–14], and even mild COVID can lead to such lasting neurological complications [6, 15], which are more common in younger and middle-aged individuals [16]. A recent large meta-analysis examining the worldwide prevalence of long COVID found that the prevenance of neurological subtype of long COVID is about 16% among confirmed COVID cases (> 777 million) [17], and thus over 100 million people may experience neuro-PASC. Although the underlying mechanisms remain unclear, accumulating evidence implicates that persistent viral antigens [18–23] and dysregulated neuroimmune responses [24–26] may drive the development of neuro-PASC [3, 27, 28]. Emerging evidence indicates that the COVID-19 brain exhibits molecular signatures of brain aging [29, 30] and cellular senescence [31, 32]. The accumulation of senescent cells [32], which secrete pro-inflammatory factors called senescence-associated secretory phenotype (SASP) that elicit deleterious paracrine-like effects, could lead to neuroinflammation and neurodegeneration [33–37] and contribute to the development of neuro-PASC. However, it remains unclear how cellular senescence is developed in neuro-PASC.

Spike (S) protein persistence has emerged as a key viral factor in neuro-PASC [18, 19, 22, 23], with multiple studies detecting S protein in postmortem brain tissue [38–40], even in the absence of detectable SARS-CoV-2 RNA [41]. S protein on the outer surface of SARS-CoV-2 is responsible for its cell attachment and internalization [42, 43] via cell surface receptors [44, 45]. S protein can be cleaved into S1 and S2 proteins by various host cellular proteases, a process that is critical for viral infection [46, 47]. S1 proteins can dissociate from the viral membrane, resulting the release of S1 into bodily fluids [48, 49]. Each SARS-CoV-2 virion can release about 300 monomeric S1 subunits [50], and serum levels of S1 [51] persist for up to 15 months [52] following the diagnosis of COVID-19 [49], and are detectable in post-COVID-19 mRNA vaccination [23]. Furthermore, S1 can cross the blood–brain barrier [53] and S1 is detected in the postmortem brain of COVID-19 victims [22, 38, 40, 41]. Importantly, S1 contains the binding domain for receptors, such as ACE2 [44], neuropilin-1 [45] and others [54–58], all of which are expressed in CNS cells [59–65]. As such, S1 present in the brain, could directly interact with CNS cells contributing to neuroinflammation [22, 66–68] and neuronal injury [69]. Moreover, S1 induces senescence-like phenotype in endothelial cells [70, 71], raising the possibility that S1 in the brain may similarly trigger cellular senescence.

As a released viral protein, present in the brain, S1 is predominantly present on astrocytes [40]. S1 enters endolysosomes via receptor-mediated endocytosis [42, 55, 72]. Based on our findings that S1 enters endolysosomes and induces endolysosome dysfunction [69, 73], and others’ findings that endolysosome dysfunction is strongly linked to cellular senescence [74–77], we investigate the role of endolysosomes in S1-induced senescence-like phenotype in human astrocytes.

Results

Multibasic motif is critical for S1-induced endolysosome dysfunction in human astrocytes

The cleavage of S proteins by furin results in the exposure of a multibasic motif with multiple arginine residues (RRAR) on the C-terminus of S1 (Fig. 1A) [46, 78], which is critical for SARS-CoV-2 infection [45, 46, 79, 80]. Our earlier studies have shown that this multibasic motif is critical for S1-induced endolysosome dysfunction in U87MG cells [73]. Using recombinant S1 and mutant S1 lacking the multibasic motif, we determined the role of the multibasic motif in S1-induced endolysosome dysfunction in primary human astrocytes. First, we examined the uptake of S1 into endolysosomes. Using immunoblotting assay, we demonstrated the internalization of both wild-type and mutant S1 into astrocytes (Fig. 1B). Furthermore, both iFluo-488 labeled S1 and Alexa-488 labeled mutant S1 lacking the multibasic motif entered endolysosomes identified with LysoTracker (Fig. 1C), and the colocalization of lysoTracker with S1 or mS1 has similar Pearson’s coefficient value, indicating that the multibasic motif is not essential for S1 uptake. Such a finding is not surprising given that both S1 and mutant S1 contain the receptor binding domain (RBD), which could interact with cell surface receptors, such as ACE2 [42, 44], expressed on astrocytes (Fig S1). However, only the S1, but not mutant S1, increased endolysosome pH (Fig. 1D) and increased endolysosome mass (Fig. 1E). Additionally, we found that S1, but not mutant S1, induced endolysosome membrane leakage, as evidenced by the formation of galectin-3 puncta within LAMP1-positive endolysosomes (Fig. 1F). Galectin-3 is a cytosolic protein, and endolysosome membrane damage enables galectin-3 to access galactose molecules in the lumen of endolysosomes, forming galectin-3 puncta [81, 82]. Endolysosome-localized galectin-3 is involved in the repair of damaged endolysosomes [83] and can be secreted into the extracellular space possibly via lysosome exocytosis [84]. Consistent with this notion, we demonstrated that S1, but not mutant S1, increased the release of galectin-3 (Fig. 1G) and cathepsin B (Fig. 1H) into the media of astrocytes. No cytotoxic effect was observed for S1 or mutant S1 at the concentrations used (1–400 ng/ml), as assessed by LDH release assay (Fig S1). Collectively, these findings suggest that the multibasic motif is critical for S1-induced endolysosome dysfunction in astrocytes.

Fig. 1.

The multibasic motif is critical for S1-induced endolysosome dysfunction in primary human astrocytes. A Schematic of the SARS-CoV-2 Spike protein highlighting the furin S1/S2 cleavage site and the multibasic motif (RRAR), which is deleted in mutant S1. B Immunoblotting showing internalization of S1 and mutant S1 (MS1) into human astrocytes after 24 h treatment. C Confocal microscopy images showing the uptake of iFluo-488–labeled S1 and Alexa-488–labeled mutant S1 into endolysosomes (LysoTracker) after 1 h of treatment. Scale bars: 10 μm. D S1 (100 ng/mL for 10 min), but not mutant S1, increased endolysosome pH (n = 3). E S1 (200 ng/mL for 48 h), but not mutant S1, increased mass of LAMP1 positive endolysosome (n = 6). Scale bars: 10 μm. F Galectin-3 puncta assay showing S1 (200 ng/mL for 30 min), but not mutant S1, induces endolysosome membrane damage as indicated by the co-localization of galectin-3 and LAMP1 (n = 4). Scale bars: 10 μm. G S1 (200 ng/mL for 24 h), but not mutant S1, increased the release of galectin-3 (n = 4). H S1 (200 ng/mL for 48 h), but not mutant S1, increased the release of Cathepsin B (n = 6). Data are expressed as Mean ± SEM and analyzed by one-way ANOVA with Tukey’s post hoc test (D-H), n represents independent cultures

The multibasic motif is critical for S1-induced senescence-like phenotype in human astrocytes

Cellular senescence is a state of stable cell-cycle arrest with secretory features in response to various cellular stresses. A key feature of cellular senescence is endolysosome dysfunction [74–77], which includes endolysosome enlargement, endolysosome de-acidification, endolysosome membrane leakage, and increased endolysosome content with senescence‐associated β‐galactosidase (SA‐β-gal), the most widely employed marker of the senescent state [85, 86]. Given our findings that S1 induces endolysosome dysfunction, we next determined the extent to which S1 induces cellular senescence in human astrocytes. Cellular senescence is characterized by the secretion of cytokines, chemokines, matrix remodeling proteins, and growth factors called senescence-associated secretory phenotype (SASP) into the tissue; therefore, we first determined whether S1 induces the release of SASP factors. Our initial cytokine and chemokine array identified IL-6 as a key SASP factor elevated by S1, and our time- and concentration-dependent studies (Fig S2) confirmed that IL-6 as a key SASP factor released in response to S1 treatment. Next, we determined the role of the multibasic motif in S1-induced senescence-like phenotype. We demonstrated that S1, but not mutant S1, increased the release of IL-6 (Fig. 2A). We then assessed whether the multibasic motif of S1 affects senescence-associated β-galactosidase (SA-β-gal) activity [87]. We demonstrated that S1, but not mutant S1, increased the percentage of SA-β-gal-positive cells as measured by a colorimetric assay (Fig. 2B). A similar increase in SA-β-gal activity was observed with S1 treatment, but not mutant S1, using a fluorometric assay (Fig. 2C). Cell cycle arrest is another hallmark of cellular senescence [88], primarily governed by two pathways: the p16^Ink4a/RB and p53/p21^CIP1 axis. Increased levels of the antiproliferative proteins p16 and/or p21 lead to cell cycle arrest [89]. Thus, we examined protein levels of p16 and p21 by immunoblotting and we found that S1, not mutant S1, significantly increased p21 (Fig. 1D) and p16 (Fig. 1E). Collectively, our findings suggest that the multibasic motif is critical for SARS-CoV-2 S1-induced a senescence-like phenotype in human astrocytes.

Fig. 2.

The multibasic motif is critical for S1-induced senescence-like phenotype in human astrocytes. A S1 (200 ng/mL for 24 h), but not mutant S1 (MS1), increased the release of IL-6 (n = 3). B-C S1 (200 ng/mL for 72 h), but not mutant S1, increased the percentage of SA-β-gal–positive cells (B, n = 3) and enhanced SA-β-gal enzymatic activity (C, n = 3). D-E S1 (200 ng/mL for 48 h), but not mutant S1, increased protein levels of p21 (D, n = 3) and p16 (D, n = 3). Data are expressed as Mean ± SEM and analyzed by one-way ANOVA with Tukey’s post hoc test (A-E), n represents independent cultures

S1 interacts with endolysosome-resident TLR7

To investigate the mechanism by which S1 induces endolysosome damage and cellular senescence, we assessed whether the multibasic motif influences the interaction between the spike S1 receptor-binding domain (RBD) and ACE2. An ACE2:S1 RBD inhibitor screening assay was performed using increasing concentrations of S1 and a mutant S1. Both proteins exhibit similar inhibition on S1 RBD binding to ACE2 (Fig. 3A), indicating that the multibasic motif may not induce endolysosome dysfunction and cellular senescence by interfering with S1 binding to ACE2 receptors at the plasma membrane. Given that S1 internalization and S1-induced endolysosome dysfunction represents an early step in S1-induced cellular senescence, we postulated that S1-induced endolysosome dysfunction plays an upstream role in cellular senescence. We further speculated that internalized S1 could induce endolysosome dysfunction directly at the site of the endolysosomes, which could lead to cellular senescence. To test this, we explored the possibility that S1 induces endolysosome dysfunction and cellular senescence via interacting with endolysosome-resident proteins, focusing on those proteins that mediate pathogen sensing such as TLR3, TLR7, TLR8, and TLR9. Using biotinylated S1 as bait, we specifically pulled down TLR7, but not TLR3, TLR8, or TLR9 (Fig. 3B). Reciprocal immunoprecipitation with anti-TLR7 antibodies confirmed an interaction between TLR7 and S1, but not with mutant S1 lacking the multibasic motif (Fig. 3C). In parallel, immunofluorescence analysis revealed co-localization of S1 with TLR7 within the same cellular compartments (Fig. 3D). Together, these data identify TLR7 as a candidate endolysosome sensor for S1-induced endolysosome dysfunction and cellular senescence.

Fig. 3.

The multibasic motif is critical for the interaction of S1 with TLR7. A ACE2:S1 receptor-binding domain (RBD) inhibitor assay showing S1 and mutant S1 exhibit similar binding to ACE2 (n = 6). B Biotinylated S1 pull-down assay demonstrating its specific interaction with TLR7, but not TLR3, TLR8, or TLR9. C TLR7 immunoprecipitation assay showing S1, but not mutant S1, interacts with TLR7. D Confocal images showing the co-localization of S1 and TLR7 within the same intracellular compartments. Data are expressed as Mean ± SEM and analyzed by two-way ANOVA with Tukey’s post hoc test (A), n represents independent replicates

S1 induces endolysosome dysfunction and cellular senescence viaTLR7

If SARS-CoV-2 S1 induces endolysosome dysfunction and a senescence-like phenotype via interaction with TLR7, then pharmacologic activation of TLR7 might recapitulate these effects. Indeed, stimulation of TLR7 with the synthetic agonist Imiquimod (R837) triggered endolysosome dysfunction in astrocytes, as evidenced by endolysosome de-acidification (Fig. 4A), increased endolysosome mass (Fig. 4B), the formation of galectin-3 puncta (Fig. 4C), and the release of galectin-3 (Fig. 4D). TLR7 activation with R837 also induced key features of cellular senescence, including elevated IL-6 release (Fig. 4E), increased percentage of SA-β-gal–positive cells (Fig. 4F), enhanced SA-β-gal enzymatic activity (Fig. 4G), and upregulation of the antiproliferative proteins p21 (Fig. 4H) and p16 (Fig. 4I). At the concentrations used, TLR7 activation with R837 did not induce cytotoxicity as assessed by LDH release assay (Fig S3).

Fig. 4.

Activation of TLR7 with Imiquimod (R837) induces endolysosome dysfunction and senescence-like phenotype in human astrocytes. A R837 (10 ug for 10 min) induced endolysosome de-acidification (n = 3). B R837 (10 ug for 48 h) increased mass of LAMP1 positive endolysosome mass (n = 5). C R837 (10 ug for 30 min) increases the formation of galectin-3 puncta in LAMP1 positive endolysosomes (n = 3). D R837 (5–10 ug/ml for 24 min) increased Galectin-3 secretion (n = 4). E R837 (5–10 ug for 48 h) increased IL-6 secretion (n = 3). F-G R837 (10 ug for 72 h) increased the percentage of SA-β-gal–positive cells (F, n = 3) and enhanced SA-β-gal enzymatic activity (G, n = 5). H-I R837 (10 ug for 24 h) increased protein levels of p21 (H, n = 3) and p16 (I, n = 3). Data are expressed as Mean ± SEM and analyzed by t-test (A-C, F-I) and one-way ANOVA with Tukey’s post hoc test (D-E), n represents independent cultures

Next, we determined whether TLR7 is required for S1-induced responses. To do this, we performed siRNA-mediated knockdown of TLR7 in astrocytes (Fig. 5A) and evaluated the resulting impact on endolysosome function and senescence markers. TLR7 silencing significantly reduced S1-induced galectin-3 puncta formation (Fig. 5B) and galectin-3 release (Fig. 5C). Likewise, knockdown of TLR7 attenuated S1-induced IL-6 secretion (Fig. 5D), SA-β-gal activity (Fig. 5E), and expression of p16 (Fig. 5F). These findings collectively support the critical role of TLR7 in mediating S1-induced endolysosome dysfunction and a senescence-like phenotype in human astrocytes.

Fig. 5.

TLR7 knockdown attenuates S1-induced endolysosome dysfunction and senescence-like phenotype in human astrocytes. A Western blot confirming siRNA-mediated knockdown of TLR7 (n = 5). B Confocal images showing TLR7 knockdown blocked S1 (200 ng/ml for 30 min)-induced formation of galectin-3 puncta (n = 3). C TLR7 knockdown blocked S1(200 ng/ml for 24 h)-induced galectin-3 release (n = 4). D TLR7 knockdown blocked S1(200 ng/ml for 48 h)-induced IL-6 release (n = 4). E TLR7 knockdown blocked S1(200 ng/ml for 72 h)-induced increases in SA-β-gal activity (n = 2). F TLR7 knockdown blocked S1 (200 ng/ml for 48 h)-induced increases protein levels of p16 (n = 5). Data are expressed as Mean ± SEM and analyzed by t-test (A) and two-way ANOVA with Tukey’s post hoc test (B-F), n represents independent cultures

TLR7/p38 MAPK Signaling underlies S1-induced cellular senescence

To elucidate the signaling pathway through which S1-mediated TLR7 activation drives cellular senescence, we investigated the three canonical TLR7 signaling pathways: NF-κB, IRF7, and MAPK. We first assessed NF-κB activity and found no measurable NF-κB activation in S1-treated astrocytes (Fig. S4). Similarly, we examined interferon-α (IFN-α) production as a readout of IRF7 activation and detected no significant induction upon S1 exposure (Fig. S4). We next assessed the MAPK pathway by evaluating phosphorylation of p38, ERK, and JNK. Among these MAPK pathways, S1 treatment significantly increased phosphorylation of both p38 and ERK (Fig. S4). Notably, TLR7 knockdown abolished S1-induced phosphorylation of p38 MAPK (Fig. 6A) and ERK MAPK (Fig. 6B), suggesting that activation of p38 and ERK occurs downstream of the S1-TLR7 interaction.

Fig. 6.

p38 MAPK, but not ERK, mediates S1-induced senescence-like phenotype in human astrocytes. A-B TLR7 knockdown blocked S1 (200 ng/ml for 30 min)-induced phosphorylation of p38 MAPK (A, n = 3) and ERK1/2 MAPK (B, n = 4). C-E Inhibiting ERK with ERK1/2 inhibitor (5 µM 1 h pretreatment) attenuated (200 ng/ml for 48 h)-induced increases in p21 (C, n = 4) but did not attenuate S1-induced increases in p16 (D, n = 3) and the release of IL-6 (E, n = 6). F–H Inhibiting p38 MAPK with SB203580 (1 µM 1 h pretreatment) blocked S1 (200 ng/ml for 48 h)-induced increases in p21 (F, n = 4), p16 (G, n = 4), and the release of IL-6 (H, n = 3). Data are expressed as Mean ± SEM and analyzed by two-way ANOVA with Tukey’s post hoc test (A-H), n represents independent cultures

To determine the functional relevance of these two pathways in the senescence-like phenotype, we tested the extent to which pharmacological inhibitors of p38 and ERK affect S1-induced senescence-like phenotype. We demonstrated that ERK inhibition with ERK1/2 inhibitor 1 attenuated S1-induced increases in p21 (Fig. 6C) but did not attenuate S1-induced increases in p16 (Fig. 6D) or IL-6 release (Fig. 6E). In contrast, inhibition of p38 MAPK with SB203580 abolished S1-induced increases in p21 (Fig. 6F), p16 (Fig. 6G), IL-6 secretion (Fig. 6H). Together, these findings indicate that p38 activation plays a central role downstream of S1-TLR7 interaction in driving the senescence-like phenotype in human astrocytes.

Discussion

Prominent findings of our studies are that the multibasic motif is critical for S1-induced endolysosome dysfunction and cellular senescence in human astrocytes. Mechanistically, S1 interacts with endolysosome-resident TLR7 via the multibasic motif, and such an interaction plays a key role in S1-induced endolysosome dysfunction and cellular senescence. Furthermore, we identified that p38 MAPK activation downstream of S1-TLR7 interaction drives the development of cellular senescence. Our findings, therefore, provide mechanistic insights into the development of cellular senescence in neuro-PASC.

Neuro-PASC affects millions, but the mechanisms underlying neuro-PASC remain unclear. Among many factors implicated in neuro-PASC, the persistence of SARS-CoV-2 viral factors such as the persistence of S proteins, especially the S1 subunit, could induce altered neuroimmune responses that lead to neuro-PASC [18–23]. As a released viral protein present in the brain [22, 38, 40, 41], S1 contains the binding domain for receptors, such as ACE2 [44], neuropilin-1 [45] and other entry factors [54–58, 90], that could mediate its internalization into endolysosomes via receptor-mediated endocytosis [42, 55, 72]. Consistent with our findings in neurons [69], we demonstrated that S1 traffics to endolysosomes and induces endolysosome dysfunction in astrocytes, where S1 induces endolysosome de-acidification, endolysosome enlargement, perhaps due to impaired clearance of undegraded substrates [69, 91], endolysosome membrane permeabilization, and the release of endolysosome content such as cathepsin B and galectin-3, likely via lysosomal exocytosis [92–94].

Endolysosomes, comprising endosomes, lysosomes, and autolysosomes, are critical for the degradation of macromolecules or damaged organelles delivered to lysosomes via endocytosis or autophagy. Endolysosomes are important for the degradation of internalized pathogens including SARS-CoV-2 [42], and central to metabolism and cellular homeostasis [95]. Endolysosomes in astrocytes play a central role in maintaining a healthy nervous system [96], and endolysosome dysfunction in astrocytes alone leads to neurodegeneration [97]. Endolysosome dysfunction could lead to abnormal accumulation of undegraded materials (macromolecules and mitochondria), leading to the enlargement of endolysosomes [69, 91], mitochondrial dysfunction [98–100], impaired clearance of pathogens [91, 101–103], and augmented release of endolysosome luminal contents via exocytosis [93, 94] which contribute to inflammation [104–108]. As such, dysfunction of endolysosomes contributes to the development of neurodegenerative disorders [109] including Alzheimer’s disease [110, 111], Parkinson’s disease [112], Amyotrophic lateral sclerosis [113], Frontotemporal dementia [114], Lysosomal storage disorders [115], and HIV-associated neurocognitive disorders [116].

Recently, endolysosome dysfunction has been strongly linked to cellular senescence [74–77] that is often accompanied by deacidification, leakage, lipofuscin accumulation, and dysregulated lysosomal enzyme activity [85, 86]. Importantly, we have shown in the present study that S1 induces cellular senescence as indicated by enhanced release of SASP factors (IL-6, galectin-3 and cathepsin B) [76, 88], elevated p16 and p21, and increased SA-β-gal activity. Importantly, cell senescence is present in the COVID-19 brain [32], and the accumulation of senescent cells, which secrete pro-inflammatory factors that elicit deleterious paracrine-like effects [33–37], could lead to neurovascular injury [24], neuroinflammation [25], and neurodegeneration [26], all of which are deemed to underlie the pathogenesis of neuro-PASC. Given that S1 internalization and S1-induced endolysosome dysfunction represent an early step in S1-induced cellular senescence, S1-induced endolysosome dysfunction may drive the development of cellular senescence within the brain of COVID-19 patients [32].

A novel finding of this study is that S1 interacts with TLR7 and induces endolysosome dysfunction and cellular senescence via a TLR7-dependent mechanism. As an endolysosome-resident pathogen-associated molecular pattern recognizing protein, TLR7 recognizes single-stranded RNA (ssRNA) from viruses [117, 118]. Besides its interaction with ssRNA, TLR7 has been shown to interact with proteins [119]. Thus, it is not surprising that we demonstrate the interaction between S1 and TLR7. Moreover, S1 interaction with TLR7 requires the multibasic motif, which is critical for SARS-CoV-2 pathogenicity [46, 47]. Deletion of this motif abolishes S1’s ability to bind TLR7, induce endolysosome damage, and trigger senescence. This interaction raises the possibility of such binding with multibasic motifs in other viral glycoproteins, such as those of HIV-1 [120, 121], influenza [122], and human cytomegalovirus [123, 124], all of which have been shown to induce senescence-like phenotype. Indeed, we have shown that HIV-1 Tat, which contains a similar arginine-rich motif, is capable of binding to and activating TLR7 and inducing senescence-like phenotype [125].

Two major pathways mediate cell cycle arrest during senescence [126]: the p53–p21 axis and the p16–RB pathway [127]. The p53–p21 pathway is typically activated by DNA damage caused by telomere attrition [128, 129], genotoxic stress [130], oncogene activation [131], or tumor suppressor loss [132], while the p16–RB pathway can be triggered by epigenetic alterations [133] or oxidative stress, independently of DNA damage [127]. We observed that S1 protein increases both p16 and p21 protein levels in human astrocytes. Our findings suggest that S1 interaction with TLR7 at the site of endolysosomes could lead to cell cycle arrest and the development of cellular senescence. To delineate the upstream signaling mechanisms responsible for initiating these pathways downstream of TLR7 activation, we examined the canonical TLR7 signaling, including NF-κB, IRF7, and the MAPK cascade [134, 135]. We found that S1 selectively activated the p38 and ERK branches of the MAPK pathway downstream of TLR7 activation. Interestingly, inhibition of p38, but not ERK, prevented the upregulation of p16, p21, and IL-6 secretion, suggesting that p38 is the major signaling pathway downstream of S1-TLR7 interaction responsible for the development of cellular senescence. Our findings are consistent with others’ findings that p38 MAPK is known to regulate transcriptional programs related to inflammation, cell cycle arrest, metabolism, and the development of cellular senescence [136–138]. While the precise mechanism of p38 activation downstream of S1-TLR7 interaction remains to be elucidated, it is possible that p38 activation may reflect a response to ROS production [139] downstream of TLR7 activation [140]. TLR7 mediated ROS production may also lead to endolysosome dysfunction, as high levels of ROS could lead to endolysosome membrane damage [141, 142] likely via peroxidation of endolysosome membrane lipids [143–145]. Additionally, S1-induced endolysosome dysfunction may lead to impaired mitophagy, mitochondrial dysfunction, and elevated ROS production in mitochondria [146], which is known to engage both p16- and p21-mediated cellular senescence. Additionally, since lysosomal function is required for p16 turnover [147], S1-induced endolysosome disruption may further contribute to its accumulation. Our findings support a model in which S1 interacts with TLR7 at the site of the endolysosomes, and subsequent p38 MAPK activation promotes the development of cellular senescence in astrocytes.

Astrocytes, the most abundant glial population in the brain, regulate neuronal function, blood–brain barrier integrity, and neuroimmune communication [148]. Disruption of astrocyte homeostasis contributes to chronic inflammation, neurodegeneration, and cognitive decline [149, 150]. Endolysosomes are critical mediators of astrocytic immune responses [96, 151], and their dysfunction in astrocytes alone can drive neurodegenerative pathology [97]. Notably, signs of astrocyte injury are consistently observed in individuals with COVID-19 [41, 152–155]. The development of cellular senescence in astrocytes may not only result in the loss of their physiological support to neurons, but also could result in the release of SASP that elicits deleterious paracrine-like effects on neighboring cells such as neurons, contributing to neurovascular injury, neuroinflammation, and neurodegeneration [33–37]. Our findings that S1 induces endolysosome dysfunction and secretion of pro-inflammatory SASP factors position astrocytes as critical effectors of neuroinflammation in neuro-PASC [3, 24, 156, 157]. Functionally, S1–TLR7 interaction drives the release of inflammatory mediators from astrocytes, including galectin-3, cathepsin B, and IL-6, all of which are elevated in the CNS and serum of individuals with COVID-19 [158–162]. S1-induced IL-6 release from astrocytes can activate microglia [163], enhance their pro-inflammatory state [164], induce neuronal injury [165], and disrupt synaptic function and neuronal excitability [166]. The increase in extracellular galectin-3, typically confined to the cytosol but released following endolysosome membrane damage, may further amplify neuroimmune responses through microglial activation [167–169]. Similarly, elevated cathepsin B in the extracellular space, also released in response to S1, has been implicated in neurodegenerative processes [170]. Importantly, these inflammatory emissions occurred without apparent astrocyte death, suggesting that S1 promotes a sustained, astrocyte-driven inflammatory environment that may contribute to the chronic neuroinflammation observed in neuro-PASC.

Conclusions

In summary, we show that S1 induces endolysosome dysfunction and cellular senescence via a multibasic motif–dependent mechanism. Our findings underscore the critical role of TLR7 in S1-induced endolysosome dysfunction and cellular senescence in human astrocytes. These findings provide mechanistic insight into how persistent viral antigens contribute to neuro-PASC and highlight TLR7 as a potential therapeutic target. Looking forward, the following studies warrant further investigation: (1) Exploring the precise signaling cascade linking S1–TLR7 activation to the establishment of the senescence-like phenotype in astrocytes. (2) Investigating the interactions between astrocytes and other cell types, such as neurons and microglia. (3) Testing whether our mechanistic findings in primary human astrocytes can be replicated in vivo. (4) Assessing clinically the role of TLR7 mutation or polymorphism in neuro-PASC. These studies would help translate the current mechanistic findings into a broader understanding of neuro-PASC pathogenesis.

Materials and methods

Cell culture

Human astrocytes were purchased from (1800, ScienCell) and cultured in in the following specialty medium from ScienCell: Astrocyte medium (1801, ScienCell) with FBS (0010, ScienCell), astrocyte growth supplement (1852, ScienCell), and penicillin/streptomycin solution (0503). Cultured astrocytes were maintained at 37 °C in 5% CO2 incubator. The purity (> 95%) of the astrocytes was determined using double immunofluorescence staining with GFAP and DAPI. U87MG human glioblastoma cells were purchased from ATCC and cultured in 1X DMEM supplemented with 10% fetal bovine serum (FBS) and 1X penicillin and streptomycin antibiotic at 37 °C in 5% CO2 incubator. U87MG cells were only used for Co-IP and pull-down assays.

The spike-ACE2 binding assay

The binding of S1 or mutant S1 lacking the multibasic domain to ACE2 was assessed using the ACE2: Spike S1 RBD (SARS-CoV-2) Inhibitor Screening Assay Kit (BPS Bioscience, 79,936). ACE2-His was diluted and added to a 96-well nickel-coated plate, followed by a one-hour incubation at room temperature. After washing and blocking, various concentrations of either S1 (Abcam, ab273068) or mutant S1 (Sigma, AGX818), along with positive controls and blanks, were added to the wells and incubated for one hour. Subsequently, the SARS-CoV-2 Spike protein receptor binding domain (RBD)-mFc (1 μg/ml) was introduced and incubated for an additional hour at room temperature. Following washing and blocking, secondary HRP-labeled antibodies were added and incubated for one hour. After final washing and blocking steps, ELISA ECL Substrate A and B were added, and chemiluminescence was detected using a microplate reader (Synergy H1).

siRNA knockdown

For siRNA knockdown of TLR7, target siRNA (J-004714–05, Dharmacon) (20 nM) and control siRNA (D-001810–10–20, Dharmacon) (20 nM) were dissolved in ddH₂O and Lipofectamine 3000 transfection reagent (L3000001, Invitrogen) and Opti-MEM (31,985,062, Gibco) were used as transfection reagent and medium for human astrocyte cells. Knockdown efficiency for TLR7 was determined with immunoblotting. The TLR7 siRNA SMART-pool sequences were as follows: UGGUUUACAUGUCGACUAA, UGGUUUACAUGUUGUGUGA, UGGUUUACAUGUUUUCUGA, UGGUUUACAUGUUUUCCUA.

Lysosome pH measurement

Total endolysosome pH was measured using a combination of dextran labelling as done previously [171]. Cells were plated on 35 mm glass bottom Poly-D-lysine dishes, and after 24 h loaded with10 mg/mL each of pH sensitive pHrodo Green Dextran (P35368, Thermo Fisher) and pH insensitive dextran, Texas Red (D1863, Thermo Fisher) for another 24 h. Following washing with PBS, and cells transferred to Hibernate E Low Fluorescence Medium (HELF, Brainbits) at 37 °C for imaging. S1 or mutant S1 was added at mentioned concentrations and fluorescence emission at 533 nm and 615 nm measured for Green and Texas Red dextran respectively. The ratio of 615/533 was converted to pH using an intracellular pH calibration kit (P35379, Thermo Fisher) with the addition of 10 μM nigericin and 20 μM Monensin in Hibernate E Flow Fluorescence (HELF) Media adjusted to different pH with HCl or NaOH. Additionally, endolysosome pH was measured with a lysosomal acidic pH detection kit (Dojindo, item code L268-10). Following treatment, human astrocytes cultured in 35 mm dishes were washed twice with serum-free medium and incubated with LysoPrime Deep Red working solution (1000X) for 30 min at 37 °C. Following this, cells were washed again and incubated with pHLys Green working solution (1000X) for an additional 30 min at 37 °C. After the final washes, a cell growth medium was added with nuclear stain, and the cells were observed under a confocal microscope. The fluorescence intensities of pHLys Green (excitation at 488 nm, emission at 500–600 nm) and LysoPrime Deep Red (excitation at 633 nm, emission at 640–700 nm) were measured on an Andor DragonFly 200 platform using a cf40 Zyla camera attached to a Leica DMi8 confocal microscope using the Fusion software. Images were exported to tiff format and fluorescence intensity ratios were calculated using the ROI function in Image J. For all pH imaging and measurements, a total of 5 fields under 40X on a Zeiss LSM800 confocal microscope comprising of at least 5–10 cells/field were imaged, and three independent experiments were carried out.

LAMP-1 Transfection and imaging

Cells were plated on to Poly-L-lysine coated glass coverslips in a 12-well plate at a density of 3 × 10⁴ cells/ml. and transduced with 30 uL of CellLight LAMP1-GFP (ThermoFisher, C10596). After transduction, the cells were allowed to rest overnight. The following day, the cells were treated with either S1 (ab273068, Abcam), mutant S1 (AGX818, Sigma), or the TLR7 agonist R837 (tlrl-imqs-1, InvivoGen) for 48 h. Cells were rinsed once with PBS and fixed in 4% PFA for 15 min, followed by three additional washes with PBS. The coverslips were then transferred to slides and mounted in ProLong Diamond Antifade with DAPI (ThermoFisher, P36971). Images were acquired using a Zeiss LSM 800 confocal microscope with 63X and 40X objectives, utilizing 1 μm z-stack intervals, and were analyzed with Imaris software.

Fluorescent labeling of recombinant mutant-S1

Fluorescent labeling of recombinant mutant-S1 (AGX818, Sigma) was conducted using the Alexa Fluor 488 microscale protein labeling kit (Thermo Fisher, cat# A3006) according to the manufacturer’s instructions. Briefly, mutant-S1 (1 mg/ml) and a molar ratio of 2:1 (dye: mutant-S1) were used to achieve selective labeling of the amine termini at room temperature for 15 min. Following labeling, conjugate purification was done using the provided spin filter to remove the unconjugated dye.

Live cell imaging

Human astrocyte cells were treated with iFluor-488-S1 labeled SARS-CoV-2 S1 (4 μg/mL, BPS Bioscience, #100,936) or Alexa Fluor 488 labeled-mutant S1 (4 μg/mL) for 1 h and LysoTracker Red (10 nM) for 15 min at 37 °C. Live cell images were acquired under the 63X objective of a Zeiss LSM 800 confocal microscope using 0.5 mm z-stack intervals. Co-localization was analyzed with Fiji, ImageJ.

Fixed cell imaging

Cells were rinsed once with PBS and fixed in 4% PFA for 15 min, followed by permeabilization using either 0.5% Triton X in PBS for 10 min or 0.1% saponin for 30 min. Blocking was performed with 3% goat serum for 1 h. Subsequently, cells were incubated with primary antibodies at 4 °C overnight. Following three washes with PBS, cells were incubated with secondary antibodies for 2 h at 37 °C. Cell images were acquired under the 63X objective of a Zeiss LSM 800 confocal microscope using 0.5–1 μm z-stack intervals and analyzed with Fiji, ImageJ.

Galectin-3 puncta assay

Following treatment with S1 (ab273068, Abcam), mutant S1 (Sigma, AGX818), or the TLR7 agonist R837 (tlrl-imqs-1, InvivoGen) for 30 min, cells were fixed in 4% PFA for 15 min. Next, cells were permeabilized using either 0.5% Triton X in PBS for 10 min or 0.1% saponin for 30 min and washed 3 times in PBS plus 0.01% TWEEN. Blocking was performed with 3% goat serum for 1 h. Subsequently, cells were incubated at 4 °C overnight with primary antibodies: LAMP1 (H4A3, Sata Cruz), Galectin-3 (sc-23938, Santa Cruz). Following three washes with PBS, cells were incubated with secondary antibodies for 2 h at 37 °C. Cell images were acquired under the 63X objective of a Zeiss LSM 800 confocal microscope using 0.5–1 μm z-stack intervals and analyzed with Fiji, ImageJ.

Immunoprecipitation

For the pulldown assay, the Pierce Pull-Down Biotinylated Protein: Protein Interaction Kit (21,115, Thermo Scientific) was used according to the manufacturer’s instructions. Briefly, the Streptavidin gel slurry was equilibrated, and 50 μg of SARS-CoV-2 S1 proteins (R&D, BT10569) were diluted in 200 μl of PBS and added to the Streptavidin gel, followed by overnight incubation. Next, the spin columns were washed, and the flow-through was collected. Total U87MG cell lysates, prepared in a cell lysis buffer (50 mM Tris [pH 8.0], 10 mM EDTA, 125 mM NaCl, 1% Triton X-100, protease inhibitor mixture), were then added to the columns and incubated overnight at 4 °C. After washing with PBS-T, target proteins were eluted and analyzed by immunoblotting. For the immunoprecipitation assay, Streptavidin was pre-incubated with 1 μg of biotin-labeled anti-TLR7 (AA 27–348, antibodies online), with a biotin-labeled secondary antibody (ThermoFisher, A18907) as a control, followed by incubation with total U87MG cell lysates. After washing, the Streptavidin was incubated with either SARS-CoV-2 S1 proteins (GenScript, Z03501) or mutant SARS-CoV-2 S1 proteins lacking the multibasic (RRAR) domain (Sigma, AGX818) overnight at 4 °C. Following further washes with PBS-T, target proteins were eluted and examined by immunoblotting.

Immunoblotting

SDS-PAGE (20% or 4–12% gel) was used to separate total cell lysate proteins and blots were transferred to nitrocellulose membranes using the iBlot 2 dry transfer system (Invitrogen). Membranes were incubated overnight at 4 °C with antibodies against TLR7 (AA900-950, antibodies online), p16 (10,883–1-AP, proteintech), p21 (28,248–1-AP, proteintech), SARS-CoV-2 S1 (40,591-MM43, SinoBiological), actin (ab8226, Abcam), GAPDH (ab8245, Abcam). Blots were developed with IRDye 680 or 800 secondary antibodies (926–68,070, 926–68,071, 926–32,210, 926–32,213, LiCor), and the density of antibody-positive protein bands was determined using an Odyssey Fc Imaging System (LiCor).

SA-β-Galactosidase staining

Changes in β-Galactosidase were assessed through staining and verified by measuring SA-β-gal activity using a fluorometric substrate. Initially, the Senescence β-Galactosidase Staining Kit (9860, Cell Signaling) was used following the manufacturer’s instructions. Astrocyte cells were cultured in 6-well polystyrene plates and exposed to either S1 (ab273068, Abcam), mutant S1 (Sigma, AGX818), or the TLR7 agonist R837 (tlrl-imqs-1, InvivoGen) for 72 h. Subsequently, the cells were fixed, stained with 1 ml of β-Galactosidase Staining Solution, and incubated overnight. Cells were imaged under a microscope (200X total magnification) to observe the development of a blue color. For the SA-β-gal staining assay, five randomly selected fields, each containing approximately 30–50 cells, were imaged in three independent experiments.

SA-β-Gal activity assay

SA-β-gal activity was measured using the Cellular Senescence Activity Assay (ENZ-KIT129-0120, ENZO) as per the manufacturer’s protocol. Astrocyte cells were grown in 6-well polystyrene plates and treated with either S1, mutant S1, or TLR7 agonist R837 for 72 h. The cells were then washed, lysed with 1X Lysis Buffer (provided in the kit), and total protein concentration was determined using the Bradford assay. Fifty microliters of cell lysate were transferred to a 96-well plate, combined with 50 µL of 2X Assay Buffer (provided in the kit), and incubated at 37 °C for 3 h. The mixture (50 μl) was then transferred to a 96-well plate suitable for fluorescence measurement. Two hundred microliters of Stop Solution (provided in the kit) were added, and fluorescence was measured using a fluorescence plate reader at 360 nm excitation and 465 nm emission (Synergy H1). Measurements were normalized to the total protein concentration in the cell lysate.

Enzyme-Linked Immunosorbent Assay (ELISA)

The release of inflammatory factors from human astrocytes into the media was quantified using several ELISA kits: Human IL-6 ELISA kit (Abcam, cat#ab100572), Proteome profiler human cytokine array kit (R&Dsystem, cat#ARY005B), Human Cathepsin B ELISA kit (Abcam, cat#ab119584), Human Galectin-3 ELISA kit (Abcam, cat#ab269555), Human IFN alpha ELISA kit (Thermo Fisher, cat#BMS216), Human IFN beta ELISA kit (R&D system, cat# QK410). Briefly, following treatment, cell culture supernatants were collected, and centrifuged at 1500 g for 2 min to remove cellular debris. Following the manufacturer's protocol, cell culture supernatants in triplicates or standards in duplicates were added to the precoated wells and incubated overnight at 4 °C. After washing, biotinylated detection antibodies were added to each well, followed by incubation with HRP-conjugated streptavidin. TMB substrate was then added, and color development was allowed to develop. The reaction was stopped by the addition of a stop solution, and absorbance was measured at 450 nm using a microplate reader (BioTek). Concentrations of inflammatory factors were determined from standard curves prepared with known concentrations of specific inflammatory factors using a four-parameter logistic curve fitting in Gen5 software (BioTek Instruments, Inc).

LDH cytotoxicity assay

An LDH cytotoxicity assay kit (Invitrogen, cat#C20300) was used to assess the cytotoxicity of various reagents on human astrocytes. Briefly, human astrocytes were treated with various concentrations of recombinant S1 (ab273068, Abcam), mutant S1 (Sigma, AGX818), or the TLR7 agonist R837 (tlrl-imqs-1, InvivoGen) for 48 h at 37 °C, with a 10X lysis buffer as positive control. Following treatment, the cell culture medium was collected and LDH activity was measured following the provided protocol. The absorbance was measured at 490 nm and 680 nm using a microplate reader (BioTek(. Calculated absorbance by subtracting 680 nm value from 490 nm value was used as relative cytotoxicity.

Statistical analysis

All data are presented as means ± standard deviations. For comparisons between two groups, statistical significance was determined using two-tailed Student's t-test. Comparisons among multiple groups with one factor were conducted using one-way ANOVA followed by Tukey's post hoc test to adjust for multiple comparisons. Comparisons among multiple groups with two factors were conducted using two-way ANOVA followed by Tukey's post hoc test to adjust for multiple comparisons. A p-value of less than 0.05 was considered statistically significant.

Authors’ contributions

WAH and XC designed the research. WAH, EM, GD, SJ, and NR performed experiments and analyzed data. WAH drafted the manuscript, and XC revised the manuscript.

Funding

This work was supported by the National Institute of Mental Health [MH119000], National Institute of Mental Health [MH134592], National Institute on Drug Abuse [DA059280].

Data availability

Datasets reported in this paper are not composed of standardized datatypes. No original code was reported in the paper. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Declarations

Ethics approval and consent to participate

This article does not contain any studies with human participants performed by any of the authors.

Consent for publication

This article does not contain any studies with human participants performed by any of the authors.

Competing interests

The authors declare no competing interests.