COVID-19-related brain inflammation and support cell death linked to viral protein activating ion channels

Previous histopathologic examinations of brain tissues from SARS-CoV-2-infected patients with COVID-19 and human brain organoids demonstrated neurodegeneration characterized by neuroinflammation and damage to neurons, astrocytes, and other cells (15, 50–52). However, the virologic causes underlying these neuroinflammatory and neurodegenerative effects remain largely unknown. Given the neuroinflammatory role of Sur1-regulated ion channels in various neuropathologies, we investigated whether a link exists between SARS-CoV-2 ORF3a protein expression, Sur1 expression, and neuroinflammation in COVID-19-affected brain tissues. To explore this potential connection, immunohistochemistry (IHC) was performed on postmortem brain tissues of patients with COVID-19 (Fig. 1). In COVID-negative control (CTR) brain tissues, no immunolabeling was detected with the anti-ORF3a antibody (Fig. 1A–a, left). However, strong immunoreactivity to the anti-ORF3a antibody was observed in COVID-positive (C-19) brain tissues, suggesting specific detection of the ORF3a protein (Fig. 1A–a, right). To validate the viral infection status of ORF3a-positive and control brain tissues, we examined the presence of the viral nucleocapsid (NCp) protein by co-immunostaining with S100B, an astrocytic marker of gray matter (53). NCp protein was clearly detected in C-19 brain tissues but was absent in CTR tissues (Fig. S1). As a structural component of SARS-CoV-2 and a widely used marker of active viral infection (54), the detection of NCp confirms the presence of viral infection in C-19 brain tissues. To provide a broader visualization of ORF3a and NCp in astrocytes of the cortical gray matter from postmortem brain tissues compared to controls, low-resolution images are presented in Fig. S1. Notably, ORF3a expression was closely associated with Sur1 (Fig. 1A–a). Widespread reactive astrocytes indicated by glial fibrillary acidic protein (GFAP) immunopositivity were observed in COVID-positive postmortem brain tissues. Consistent with neuroinflammation, NF-kB (p65) and proinflammatory cytokines, such as TNF and IL-6, were co-localized with reactive astrocytes (Fig. 1A–b–d). Sur1 was elevated in reactive astrocytes in COVID-positive brain tissues but not in COVID-negative controls (Fig. 1B–a). A detailed examination showed strong Sur1 expression in astrocytes, particularly in perivascular endfeet, a specialized subcellular compartment that ensheathes cerebral vasculature and plays crucial roles in maintaining the blood-brain barrier (BBB) (55). Since Sur1-regulated ion channels are activated in response to brain injury through the transcriptional activation of Abcc8/Sur1, we used RNAscope, a highly sensitive in situ hybridization method (56), to detect Abcc8 RNA. Abcc8 RNA was clearly detected in astrocytes and co-localized with its pore-forming partner, Trpm4 (Fig. 1B–b–c). This observation was corroborated by co-immunolabeling of Sur1 and Trpm4 protein expression in astrocytes (Fig. 1B–d). These findings suggest a correlation between ORF3a expression, Sur1-Trpm4 channel expression, and neuroinflammation induced in COVID-19 brain tissues.

To assess whether ORF3a itself could drive the increase in Sur1, we developed an ORF3a-expressing adenovirus 5 (Ad5-ORF3a) system and investigated the impact of ORF3a in various astrocytic cell lines. In SNB19 cells, a human glioblastoma cell line derived from astrocytes, transducing Ad5-ORF3a at increasing multiplicities of infection (MOIs) resulted in a proportional increase in ORF3a mRNA levels by 5-day post-viral transduction (dpt) (Fig. 2A–a). The ORF3a protein became detectable at MOIs 50 and higher (Fig. 2A–b), while neither mRNA nor protein expression was detected in Ad5 empty vector controls, confirming an ORF3a-specific expression. Note that we used a hexon-based adenovirus titration assay to measure the titer of Ad5-ORF3a. This assay quantifies the hexon protein, a major structural component of the adenovirus capsid. However, it does not distinguish between infectious and non-infectious viral particles that contain hexon. As a result, the measured titer may overestimate the actual number of infectious particles capable of initiating a productive infection in SNB19 cells, which may explain the relatively high MOI required in these experiments.

We monitored Sur1 expression in SNB19 cells transduced by Ad5-ORF3a through examining the transcriptional activation of Abcc8/Sur1 over 6 days. Expression of ORF3a triggered activation of the Abcc8/Sur1 promoter activity, which was measured by an Abcc8 promoter luciferase assay, and showed a peak of the promoter activity at 4.5 dpt (Fig. 2B–a) (57). Following the Abcc8/Sur1 promoter activation, the maximal transcriptional activity of Sur1, as measured by quantitative reverse transcription PCR (qRT-PCR), was achieved around 4.5 dpt and plateaued by 6 dpt (Fig. 2B–b). Alongside the ORF3a-induced Sur1 upregulation, GFAP, a marker of reactive astrocytes and brain injury (58, 59), and NOS2, a downstream Sur1 effector (60), also showed progressive increases within this timeframe (Fig. 2B–c–d). This upregulation coincided with a marked reduction in cell proliferation and viability, as indicated by MTT and trypan blue assays, accompanied by elevated cell death (Fig. 2C–a–b). The ORF3a-induced effects in SNB19 cells were consistent with our findings in other astrocyte cell lines, including human neuroblastoma SH-SY5Y (Fig. S2) and mouse neuroblast neuro-2a (N2a) cells (Fig. S3), demonstrating a conserved ORF3a impact on astrocyte-derived cell lines across species.

We next investigated the molecular mechanisms underlying ORF3a-induced Sur1 elevation. To minimize potential artificial effects from ORF3a protein overproduction, we generated SNB19 cells that stably express ORF3a or an enhanced green fluorescent protein (EGFP) control using a doxycycline (Dox)-inducible pLVX-TetOne-Puro plasmid system. The addition of Dox (100 ng/mL) enabled inducible expression of ORF3a or EGFP, and cells were collected for analysis at 48 h post-induction (hpi). The inducible ORF3a expression in SNB19 cells resulted in a significant increase of Sur1 expression (Fig. 3A–a). Given that Sur1-regulated ion channels facilitate Ca²⁺ influx (35, 60), we evaluated ORF3a-induced Ca²⁺ influx through Ca²⁺ imaging with a Fluo-8 assay (61). ORF3a expression resulted in elevated intracellular Ca²⁺ levels (Fig. 3A–b), leading to cell death by apoptosis and necrosis (Fig. 3A–c–e).

While ORF3a expression alone had minimal activation effect on NF-κB and only modestly elevated cytokines, such as TNFα, IL-1β, and IFNβ1 (Fig. 3B), the addition of TNFα or IFNγ to ORF3a-producing cells mimicking secreted cytokines from targeted or immune cells further augmented Sur1 and cytokine levels (Fig. 3C). IFNγ induced a more pronounced upregulation of Sur1 than TNFα (Fig. 3C–a). Supporting the role of TNFα in NF-κB activation (62), TNFα significantly enhanced NF-κB levels in ORF3a-producing cells, while IFNγ did not (Fig. 3C–b). TNFα induced greater TNFα and IL-1β production than IFNγ (Fig. 3C–c–d), whereas IFNγ prompted a stronger IFNβ1 response indicative of a potential NF-κB-independent pathway (Fig. 3C–e). These data suggest that the ORF3a-mediated Sur1 upregulation is in synergy with secreted cytokines, such as TNFα and IFNγ, likely released by targeted and infiltrating immune cells, such as infected glial cells in SARS-CoV-2-infected brain tissues (63), exacerbating neuroinflammation through both NF-κB-dependent and -independent signaling pathways.

To explore whether ORF3a expression correlates with neuroinflammation and astrocytic apoptosis in an animal model system, we employed K18-hACE2 transgenic (Tg) mice, which are susceptible to SARS-CoV-2 infection (23). Following infection with SARS-CoV-2, we collected brain tissues of both SARS-CoV-2-infected and mock-infected K18-hACE2 Tg mice at 8-days post-infection (dpi). Compared to mock controls, infected mice displayed widespread ORF3a expression throughout the brain (Fig. 4A). Consistent with the high levels of reactive astrocytes seen in the brain tissues of COVID-19 patients (Fig. 1), infected mice also showed elevated levels of reactive astrocytes compared to the mock controls (Fig. 4B). This increase was indicated by the higher expression of GFAP, a well-established marker for astrocyte reactivity and brain injury (58, 59). Additionally, proinflammatory cytokines TNF and IL-6 were observed within GFAP-positive astrocytes in infected tissues but were absent in mock controls, highlighting a strong inflammatory response (Fig. 4C through D). In parallel with the widespread ORF3a expression in infected mouse brains, there was extensive labeling of cleaved Caspase-3 (cCasp3; Fig. 4E), which marks apoptotic cell death in astrocytes. Sur1 was observed to co-reside with these cCasp3-positive astrocytes, establishing a link between ORF3a expression, Sur1 elevation, and astrocytic apoptosis (Fig. 4F). Collectively, these findings suggest that ORF3a contributes to neuroinflammation and astrocyte death in the infected mouse brain, reflecting similar neuropathological patterns seen in human COVID-19 cases (Fig. 1).

As an error-prone RNA virus (64), SARS-CoV-2 continues to mutate, and certain fixed ORF3a mutations are present in natural SARS-CoV-2 variants, especially those defined as variants of interest (VOI) or variants of concern (VOC) by the World Health Organization (WHO) that are often associated with higher mortality (25, 65, 66) and severe disease outcomes (65, 67–69). To examine whether the observed ORF3a effects are also present in those emerging ORF3a mutants, we extended our study to include the unique ORF3a T223I mutant associated with Omicron variants (21, 25, 70). The ORF3a T223I mutant was selected because some of the ongoing circulating viral variants are Omicron-derived viruses, which are in many ways different from the original WT virus both in viral pathogenicity and transmission (71). Additionally, our previous mutagenesis studies identified two distinct types of lysosome-associated (L-ORF3a, such as WT) and ER-associated (E-ORF3a like T223I) ORF3a proteins based on subcellular localization and cytotoxic effects (22, 72).

To examine whether subcellular location of WT and T223I mutant ORF3a proteins remains similar in SNB19 cells to that in other human cells as we reported previously (21), we measured their presence in the ER compartment in SNB19 cells using the immunofluorescence assay. As expected, there was a significantly (P < 0.007) higher percentage of T223I mutant ORF3a protein residing in the ER with a median of Pearson's correlation coefficient of 0.70 compared to WT ORF3a with a median of Pearson's correlation coefficient of 0.48 (Fig. 5A). These observations are similar to what we observed in other human cells (21, 22, 72).

To examine whether the T223I mutant also induces Sur1 upregulation, we transfected SNB19 cells with plasmids expressing WT ORF3a, T223I mutant ORF3a, or an empty control vector. qRT-PCR analysis confirmed that both WT and T223I ORF3a proteins significantly elevated the mRNA level of Sur1 (Fig. 5B–a). Promoter activity assays for Accb8/Sur1 demonstrated that the T223I mutant elicited a higher level of Accb8/Sur1 expression than WT ORF3a (Fig. 5B–b). Western blot analysis further confirmed the production of WT and T223I ORF3a proteins with elevated Sur1 expression (Fig. 5B–c). In alignment with the observed Sur1 elevation, both WT and T223I ORF3a induced proportional increases in intracellular Ca²⁺ measured by Fluo-8 imaging (Fig. 5C–a–b). Additional qRT-PCR analysis showed corresponding transcriptional profiles of GFAP, IL-1β, and IFNβ1 (Fig. 5D), which corresponded to subsequent apoptosis and necrosis of SNB19 cells (Fig. 5E).

To explore potential strategies for blocking Sur1-mediated neuroinflammation and neurotoxicity induced by ORF3a, we tested the inhibitory effects of glibenclamide (GBC, aka glyburide), an FDA-approved antidiabetic drug that specifically targets Sur1 channels (42). A seven-dose range of GBC from 5.0 to 320.0 µM was employed to determine its half-maximal effective concentration (EC₅₀) against ORF3a-induced cell death in SNB19 cells using the MTT assay. GBC had an EC₅₀ of 41.93 ± 5.61 µM, a 50% cytotoxic concentration (CC₅₀) of 256.0 ± 8.49 µM, and a therapeutic index (TI) of 6.11 ± 0.81 (Fig. 6A–a–b). Additionally, our previous studies identified glycyrrhizin (GL), a natural compound, as an inhibitor of ORF3a effects in renal cells (21). In this study, we further evaluated the therapeutic potential of GL in SNB19 cells. The MTT analysis revealed that GL exhibited an EC₅₀ of 20.13 ± 2.39 µM, a CC₅₀ of 276.3 ± 9.66 µM, and a TI of 13.73 ± 1.44 (Fig. 6A–c–d).

Given the comparable EC₅₀ values of GBC and GL, their inhibitory effects were tested at equal concentrations (80 µM) against ORF3a-induced Sur1 and GFAP expressions, cytokine production, and apoptotic cell death in SNB19 cells. As the control, ORF3a expression in SNB19 cells increased Sur1 levels by approximately ninefold. Treatment with GBC reduced Sur1 levels by 33.1% (Fig. 6B–a, left), while doubling the concentration of GBC further decreased Sur1 expression to a 50.8% reduction, indicating a dose-dependent response (Fig. S4A). Treatment with GL reduced Sur1 levels by 67.0% (Fig. 6B–a, right). GL also showed stronger inhibitory effects on GFAP and IFNβ1 expressions compared to GBC (Fig. 6B–b–d). Interestingly, while GBC treatment did not affect IL-1β levels, GL significantly downregulated IL-1β, though the ORF3a-induced increase in IL-1β was marginal (Fig. 6B–d). However, the differential effects of GBC and GL on IL-1β were reproducible (Fig. S4D). Regarding ORF3a-induced cell death, GL also demonstrated a stronger reduction effect compared to GBC (Fig. 6C).

To assess the combined inhibitory effects of GBC and GL, we carried out a reciprocal dosing experiment, where one drug's concentration was fixed at 80 µM, while the other was varied incrementally from 20 to 80 µM. Both GBC and GL independently reduced ORF3a-induced cell death in a dose-dependent manner. However, the combination of GBC and GL resulted in a small but significant additional reduction in ORF3a-induced cell death compared to GBC or GL alone (Fig. 6D). These findings suggest an additive effect of GBC and GL on the suppression of ORF3a-induced cell death.

SNB19 (RRID:CVCL_0535 from National Cancer Institute) is a human glioblastoma cell line that develops from astrocytes. It was maintained in RPMI 1640 medium supplemented with L-glutamine and 10% FBS (Quality Biological/Fisher Scientific Cat#: 112025101) and 100 units/mL penicillin plus 100 µg/mL streptomycin (41, 115). The SNB19 cell line was also used to establish in this study cell clones that stably produce ORF3a or EGFP by using a DoX-inducible pLVX-TetOne-Puro plasmid system (Takara Bio). Dox-inducible production of ORF3a or EGFP (as a control) proteins was achieved by adding Dox at a concentration of 100 ng/mL, and cells were collected at 24–48 h post-Dox induction. Human neuroblastoma SH-SY5Y cells were cultured in a medium composed of a 1:1 mixture of Eagle's minimal essential medium (EMEM) and F12K. Mouse N2a cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% heat-inactivated FBS, 0.3% glutamine, and antibiotics (100 U/mL penicillin, 100 µg/mL streptomycin) at 37°C and 5% CO₂ (116, 117).

We employed three different approaches to express ORF3a in cells. First, an ORF3a-expressing Ad5-ORF3a construct was generated using the Adeno-X Adenoviral System 3 (Takara Bio) following the manufacturer's instructions, with correct insertion of the ORF3a nucleotide sequence confirmed via restriction digestion and Sanger sequencing and viral titers quantified using the hexon-based adenovirus titration kit (Takara Bio). Second, a Dox-inducible ORF3a expression system in SNB19 cells was established and used to achieve stable, low-level expression of ORF3a, minimizing potential artifacts from protein overproduction. Third, WT and T223I mutant ORF3a constructs were expressed using pCAG plasmids as we previously described (21, 72). For transfections, SNB19 and N2a cells were transfected with pCAG-ORF3a plasmids using Invitrogen lipofectamine stem transfection reagent (Thermo Fisher, Cat# STEM00008), while SH-SY5Y cells were transfected with Lipofectamine 3000 (Invitrogen), with all transfections performed following the respective manufacturer's protocols.

A total of 12 K18-hACE2 Tg mice (strain B6.Cg-Tg(K18-ACE2)2Prlmn/J, Strain #: 034860, The Jackson Laboratory) were intranasally infected with SARS-CoV-2 New York strain at the dose of 2.5 × 10⁴ TCID₅₀ (21, 23). Two surviving mice were euthanized on 8 dpi. Brain tissues were fixed with formalin and processed for paraffin embedding and sectioning to evaluate the ORF3a protein level and brain tissue damage using IHC compared to controls. The K18-hACE2 mice were housed under pathogen-free conditions, and viral infection was conducted in an animal biosafety level 3 (ABSL-3) facility at the University of Maryland College Park.

The procedure to examine the subcellular location of ORF3a proteins has been described previously (21, 72). Briefly, the HA-tagged WT or T223I mutant ORF3a-carrying plasmids were transfected into SNB19 cells for 24 h. Immunofluorescent assay was performed by sequential incubation with primary antibodies and Texas red (TR)-labeled secondary antibodies (Vector Laboratories, Burlingame, CA). Mouse anti-HA tag antibody (Santa Cruz, sc7392) detecting HA-tagged ORF3a and rabbit anti-Calnexin antibody (Abcam: ab22595) detecting an ER membrane protein Calnexin were used to visualize co-localization of ORF3a with the ER membrane (22, 72). Cells were examined with a Leica TCS SPII confocal laser scanning system. Two or three channels were recorded simultaneously and/or sequentially and controlled for possible breakthrough between the fluorescein isothiocyanate and Texas red signals and between the blue and red channels. A Fiji image analysis software (https://imagej.net/↗) that combines ImageJ and JCoP plugin was used to quantify the subcellular location of ORF3a (22, 72). Random images were used for the analyses and the calculation of the mean and standard deviation of Pearson's correlation coefficients (P values).

Human brain sections were first deparaffinized and then subjected to antigen retrieval in citrate buffer (10 mM citric acid, 0.05% Tween 20, pH 6.0). After blocking sections in blocking buffer (10% donkey serum in 0.2% Triton-X in phosphate-buffered saline (PBS) for 1 h), sections were incubated at 4°C overnight with primary antibodies directed against anti-ORF3a (LS Bio: LS-C829863), anti-SARS-CoV-1/2NP (Sigma-Aldrich: ZMS1075), anti-s100B (Abcam: ab868), anti-GFAP (Sigma-Aldrich: AB5541), anti-TNF (Santa Cruz Biotechnology: sc-1350), and a custom-made anti-Sur1 antibody (118). After several rinses in PBS, sections were incubated with species-appropriate fluorescent secondary antibodies (Alexa Fluor 488 and 555, Molecular Probes, Thermo Fisher Scientific, Waltham, MA, USA) for 1 h at room temperature. Controls for IHC included the omission of primary antibodies.

Immunostaining of mouse brain tissues with the indicated antibodies used in this study has been previously described (41). For immunolabeling of mouse brain tissues, briefly, brains from C57BL/6 or K18-hACE2 transgenic mice and WT controls were fixed in formalin and embedded in paraffin. Sections, 10 µM thick, were cut from the paraffin-embedded blocks containing the entire brain using a microtome. These sections were deparaffinized and rehydrated through graded ethanol (100, 95, and 70%) and xylene washes. Following antigen retrieval, sections were blocked in 5% donkey serum in PBS with 0.2% Triton-X. Unbiased assessments of specific labeling of each image were obtained using NIS-Elements AR software (Nikon Instruments, Melville, NY, USA) from sections immunolabeled as a single batch. All images for a given signal were captured using uniform parameters of magnification, area, exposure, and gain.

To assess transcriptional activity of Abcc8/Sur1, an Abcc8 luciferase promoter assay was conducted based on our previous report (57). Briefly, 1.0 × 10⁴ SNB19 cells were seeded into individual wells of a white 96-well plate and incubated overnight. For the assay, pCAG-ORF3a carrying plasmid was co-transfected with pGL3-Abcc8 plasmid or PGL3-basic as the Vec control. Alternatively, SNB19 cells were transduced with Ad5-ORF3a and Ad5 control. To quantify luciferase activity, a dual-luciferase reporter assay system (Promega: Cat# E1910) was used with measurements taken using the Synergy H1 microplate reader (Agilent) at specified time intervals. Signal normalization was achieved using Renilla luciferase as an internal control. Fold changes were calculated relative to the control containing an empty vector.

To measure the effect of WT or T223I ORF3a on Sur1-regulated ion channel activities, Ca²⁺ influx was measured in two different ways using the Fluo-8 Calcium Assay Kit (AAT Bioquest) following the manufacturer's instruction. First, Ca²⁺ influx was measured in Dox-inducible SNB19 cells 48 h post-Dox gene induction. Alternatively, SNB19 cells were transduced with Ad5-ORF3a, and the Ca²⁺ influx was measured on day 5 post-transduction. Ca²⁺ influx was measured by following the manufacturer's instruction. Specifically, ORF3a-producing SNB19 cells were cultured in a 96-well plate and harvested at the indicated time. After removing the growth medium, add the dye in Fluo-8 AM working solution to the cells and incubate them at room temperature for 30–60 min. The fluorescence intensity changes at indicated time were recorded using the Synergy H1 microplate reader with excitation at 490 nm and emission at 525 nm. The change in fluorescence directly correlates to the influx of calcium ions into the cells. To ensure the observed changes are specific, the fluorescence changes were presented as the fold changes over the vector control.

The procedure of qRT-PCR has been previously described (21, 22). Briefly, cells were cultured in a six-well plate until they reached approximately 70% confluency. Cells were transfected as described above and collected at the indicated time. Total RNA was extracted using Trizol (Invitrogen, 448706). The extracted RNA was treated with RQ1 RNase-Free DNase (Promega: M6101) to eliminate genomic DNA contamination. Subsequently, the RNA was reverse-transcribed into cDNA using reverse transcriptase (Thermo Fisher: 4311235). Real-time PCR was conducted on a QuantStudio 3 Real-Time PCR System using gene-specific primers and SYBR Green Master Mix (Thermo Fisher: A46109) to detect mRNA expression levels. The amplification conditions included 40 cycles of 95°C for 10 s and 60°C for 30 s, followed by a melting curve analysis. Fold change in mRNA expression was calculated using the 2^−ΔΔCT method, with GAPDH mRNA serving as the internal control.

These procedures have been used in our laboratory previously (19, 21, 22). Briefly, to measure cell growth, viability, and apoptotic cell death, SNB19, SH-SY5Y, or N2a cells were introduced with pCAG-ORF3a plasmid or Ad5-ORF3a as described. At the indicated time, cell growth was counted using a TC20 automated cell counter (Bio-Rad) to determine the total cell count and the number of non-viable (stained) cells, and cell death was quantified by cell counting and trypan blue staining. Cell proliferation and viability were assessed using the MTT assay at the absorbance of 570 nm using an H1M microplate reader (Agilent). Cell apoptosis and necrosis were measured using a RealTime-Glo annexin V apoptosis and necrosis assay kit (Promega).

It was carried out as we described previously (21, 116). In brief, total proteins were extracted from transfected SNB19 cells at indicated time. Equal amounts of total protein were separated on SDS-PAGE gel by electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane (Bio-Rad). Target proteins were detected using anti-ORF3a (LS Bio: LS-C829863), a custom-made anti-Sur1 (118), and mouse anti-GAPDH antibody (cell signaling: 2118).

The FDA-approved antidiabetic drug and a Sur1 inhibitor, GBC (#G2539; meets USP testing; Sigma-Aldrich) and the ORF3a inhibitor GL (NSC234419) provided by the Developmental Therapeutics Program of the National Cancer Institute, the National Institutes of Health were used in the drug treatment study. Both drugs were stored in dimethyl sulfoxide (DMSO) as the stock concentrations of 50 mM at 4°C and diluted with DMSO for the indicated drug tests.

Statistical analyses were conducted using Prism 9 software (GraphPad, San Diego, CA, USA). Pairwise t-tests or one-way analysis of variance (ANOVA) were employed as appropriate. Statistical significance was defined as P < 0.05 at the 95% confidence level. Symbols *, **, ***, and **** were used to denote significance levels: P < 0.05, P < 0.01, P < 0.001, and P < 0.0001, respectively.