HEK-293T and human microglial cell line (CHME3) were cultured in DMEM medium (Himedia #AL219A) with 10%FBS (Himedia #RM1112) and 1% antibiotic/antimycotic solution (Himedia #A002A). Human microglial cells CHME3, have been used for studying inflammatory pathways gene expression and HEK-293T cells have been utilized for luciferase reporter assay, in-vitro-ubiquitination assay, cycloheximide chase assay and for spike gene transfection purposes. All the cells have been grown in 37°C culture-incubator with constant supply of 5% CO₂. All plasmid DNA transfections have been performed with Lipofectamine 2000 (#11668019, Invitrogen™) as per manual provided with kit.

SARS-CoV-2 Spike gene (S) plasmid (pTwist-EF1α-nCoV-2019-S-2×Strep) was a kind gift from Nevan J. Krogan, QBI COVID-19 Research Group (QCRG), San Francisco, CA 94158, USA. Flag-USP7 plasmid was a kind gift by Altaf Wani (Ohio State University, OH, USA). 6X His-Ubiquitin plasmids were a gift from Prof. Dimitris Xirodimas, University of Dundee, UK. Renilla luciferase construct utilized for normalization process in luciferase reporter assay were given generously by Dr. Vivek Natrajan, IGIB, Delhi, India. NF-kB-Luc and TNFα-luc constructs were obtained from Stratagene. Flag-HA-USP33 (Addgene plasmid # 22601) (48), IFN-beta-pGL3 (Addgene plasmid #102597) (49–51). pLV-IRF9 was a gift from George Stark (Addgene plasmid # 71452) (52–54). psiCHECK IRF9 3’UTR was a gift from Thomas Tuschl (Addgene plasmid # 19863) (48). All the Addgene constructs were commercially purchased from Addgene, USA.

2µg of SARS-CoV-2 Spike gene (S) plasmids (pTwist-EF1α-nCoV-2019-S-2×Strep) were transfected in 1x10⁵ of HEK-293T cells in 6-well plate format with Lipofectamine 2000. Regular passage of transfected cells were done to expand their number and yield of harvested conditioned media from the cells that contained all the secreted extracellular vesicles. Upon every third day, the supernatant were collected from Spike transfected cells. Mock transfected cells were considered as control and supernatants were collected similarly. Supernatants were pooled up and exosomes isolation were performed as per previously established methods described by Miranda et al. (55). In this method, the supernatants were harvested and “salting-out” procedure were performed for the isolation of exosomes. After clearing the cell debris by centrifugation at 2000 rpm, 10% sodium acetate buffer (1 M and pH 4.75) were added to the harvested supernatant and incubated on ice for almost ~60 minutes. A short heating at 37°C for 5 minutes were given which leads to turbid solutions indicating the aggregation of microvesicles. These turbid solutions were then centrifuged at 10000g for 60 minutes to obtain the pellets of extracellular vesicles. These pellets were then washed with 0.1M Na-acetate buffer followed by a final high-speed centrifugation and final exosomes pellets were dissolved in PBS and stored at 4°C till further analysis. Total yield of exosomes were measured as protein content with the help of Bradford reagent. TSG101 protein expressions were considered as tetraspanin marker and have been confirmed in all exosomes preparations. Additionally, CD63 has also been checked as exosomal marker in exosome preparation. Absence of cellular endosomal marker protein ‘Calnexin’ have been confirmed in all the isolated exosomes. These exosomes pellets were then subjected to various miRNA analysis by qPCR and protein analysis via western blotting. Human microglial cells CHME3 were treated with various amounts of exosomes (2 µg and 4µg as shown in their respective experiments) for 24 or 48 hours as indicated in individual experiments.

Cell lysis have been done with RIPA buffer (20mM Tris [pH 7.5], 150mM NaCl, 1g/mL leupeptin, 1mM-glycerophosphate, 1mM Na3VO4, 2.5mM sodium pyrophosphate,1mM EGTA, 1% NP-40 and 1% sodium deoxycholate). 1X Protease inhibitor cocktail from Sigma (#S8820, Sigma‐Aldrich) was used for inhibiting protein degradation. Bradford assay (#500‐0006; Bio‐Rad Laboratories) has been used for total protein estimation. 30µg of total protein lysates were boiled with 4X loading dye and run on 8-10% polyacrylamide gel at 120 volts. Proteins were transferred on nitrocellulose membrane (#SCNJ8101XXXX101 MdI, advanced microdevices Ltd) by running at 100 Volts for 2 hours in a wet-transfer apparatus. Skimmed milk powder (# GRM1254, Himedia) have been used for blocking the transferred membrane for one hour at room temperature on slow rocker shaker. 5% BSA solution have been used for primary and secondary antibody incubations. Primary antibodies were incubated with membrane for overnight at 4°C. Primary antibody have been given at various dilutions, 1:500 upto1:5000 depending on antibodies and genes in respective experiments described individually in their legends. Membrane received three washings to remove excess primary antibodies followed by one hour incubation with respective secondary antibodies, given in 5% BSA. After 1 hour incubation, excess secondary antibodies were removed with three TBST washings given 15 minutes each. HRP conjugated secondary antibodies for anti-rabbit IgG and anti-Mouse IgG were from Jackson Immunoresearch, USA. ECL western blotting substrate (#32106 Pierce, Thermo Scientific) were used to develop signals on X-ray films (#6568307, Carestream Health Inc., USA). Equal loading of protein samples have been assessed through images density measured by densitometry, performed on ImageJ software version-1.52q. Image densities of GAPDH protein in the respective lanes have been utilized as normalizer in all the western blotting experiments.

Antibodies used in this study are; anti-Calnexin, (#2679, Cell Signaling technology), anti-CD63, (#sc-5275), USP42 (sc-390604, Santacruz Biotech), USP33 (#sc-100632, Santacruz Biotech), GAPDH (#sc‐32233, Santacruz Biotech), USP7 (#D17C6, Cell Signaling Technology), TSG101 (#sc-13611, Santacruz Biotech), IRF9 (#76684, Cell Signaling Technology), SARS-CoV-2 Spike Protein S2 Monoclonal Antibody (1A9), # MA5-35946, Invitrogen). Proteasome inhibitor MG132 (#C2211, Sigma‐Aldrich) has been used for in vivo ubiquitination assay in HEK-293T cells. Final concentration of 10µM MG132 dissolved in DMSO have been applied on cells for 8 hours. For exosomes biogenesis inhibition, GW4869; a neutral sphingomyelinase (#D1692-5MG, Sigma Aldrich), have been used at 10µM final concentration. For general deubiquitinase inhibition, PR-619 (#SML0430-1MG, Sigma Aldrich) have been used at 10µM final concentration. siRNA for negative control (# 1027280, Qiagen), USP33 (#SI00109123, Qiagen) and IRF9 (# 1027417, Qiagen) were purchased as Flexitube siRNA from Qiagen.

RNA extraction was performed with miRNeasy Mini kit (#217004 Qiagen), protocol followed as per manufacture’s instruction. microRNA specific primers have been used for cDNA synthesis provided by TaqMan MicroRNA assay system with TaqMan reverse transcription kit (#4366596, applied Biosystem). Reverse transcription was done at thermal incubations as follows; 16°C for 30 min, 42°C for 30 min, and 85°C for 5 min. MicroRNA qPCR analysis have been done by using universal PCR master mix (#4324018, applied Biosystem). Changes in cellular levels of microRNAs were checked with commercial TaqMan MicroRNA assay (TaqMan Assay ID: 000470 for human miR-148a, # 002677 for miR-590-3p and RNU24 #001001; Applied Biosystem). RNU24 has been used as qPCR internal normalization control. Thermal cycles for qPCR were as follows; 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. All qPCR reactions have been performed on ABI 7500 fast Real Time thermal cycler from Applied Biosystem.

To search the potential targets of miR-148a and miR-590-3p, microRNA bioinformatics prediction tools such as microRNA.org, miRDB and TargetScan 7.1 have been employed. Complementary binding sites in 3’UTR of IRF9 and USP33 genes for miR-590-3p and miR-148a respectively have been identified with the help of miR-Base and microRNA.org.

To confirm the IRF9 and USP33 gene regulation via microRNAs, miR-148a mimics (# MC10263, Thermo Fisher Scientific), antago-miR-148a (#AM10263, Thermo Fisher Scientific), miR-590-3p mimic (# MC12644, Thermo Scientific) and anti-miR-590-3p (# AM17000) have been used. A day before transfection, human microglial cells, CHME3 were seeded at 1x10⁵ cells in each well (6-well plate format). 100 picomoles of mimics and anti-miR-148a were transfected per well with Lipofectamine RNAi Max transfection reagent (# 13778150, Invitrogen) as per manufacturer’s instructions. CHME3 cells were harvested post 48 hours of transfection and proceeded for RNA and western blot analysis.

HEK-293T cells have been utilized for all the luciferase assays. Promoter activity assays for TNFα, IFN-β and NF-κB were done by co-transfection methods in 6 well plate format. One day before transfection, 1x 10⁵ cells were seeded to reach almost 60% confluency at the time of transfection. 1µg TNFα, IFN-β and NF-κB plasmids were co-transfected with 1µg of Spike plasmids, USP33 plasmids and IRF9 plasmids. Similarly in IRF9 siRNA experiments, 100 picomoles of siRNA against IRF9 were transfected 12 hours before TNFα, IFN-β and NF-κB plasmids transfections. For transfection normalization, 500 ng of Renilla luciferase plasmids were transfected in all wells. Post 24 hours of transfection, HEK-293T cells were harvested and proceeded for dual luciferase assay as per manufacturer’s instructions (# E1910, Promega). Synergy H1 multi-mode reader, Biotek have been employed for capturing luminescence readings followed by capturing Renilla luminescence in all the wells. Renilla luminescence reading have been used as denominator to get the final luciferase promoter activities.

In vivo ubiquitination assay was performed as described in our previous publication (29). Briefly, HEK-293T cells were seeded in 90 mm culture dishes to reach for almost 70% confluency before transfection. 5µg of pLV-IRF9 plasmids and 5µg of 6X-His-ubiquitin plasmids were transfected along with 5µg of Flag-HA-USP33 in co-transfection experiments to check the impact of USP33 on ubiquitination levels of IRF9. After 24 hrs of transfection, all cells were treated with 20µM final concentration of MG132 and incubated for at least 8 hours. Cells were then harvested and sonicated and lysed with Buffer A (6M guanidine-HCL, 0.1M Na2HPO4, 10mM imidazole at pH 8.0). The lysates were then incubated with Ni-NTA agarose beads overnight on rotor shaker at room temperature. The beads were washed next morning with buffer A, followed by Buffer A+Ti (1 volume of buffer A and 3 volume of buffer Ti (25mM Tris-HCL, 20mM imidazole at pH 6.8) and finally with buffer Ti. Final elutions were done with 50µl of His-Ubiquitin elution buffer (0.2M imidazole, 5% w/v SDS, 0.15M Tris-Cl at pH 6.8) also known as 2X Laemmli buffer and boiled for 5 minute at 100°C. Samples were run on SDS gel and probed with anti-IRF9 antibody.

Cycloheximide is a well-known translation inhibitor to analyze the half-life of cellular proteins. HEK-293T cells were used for cycloheximide chase assay in 6 well plate format with 1x10⁵ cells per well. At ~70% confluency, cells were transfected with IRF9 only and IRF9 plus USP33 co-transfection in different plates with the help of Lipofectamine 2000. Post-24 hrs of transfection, Cycloheximide (#01810, Sigma-Aldrich) were given to all wells at 100µg/mL concentration. Cells were then harvested at various time points (2, 4, 8, 12 and 24 hours) followed by lysis with RIPA buffer and western blot analysis for indicated proteins.

All the experiments have been repeated independently as three biological replicates to obtain means, standard deviation and standard error of means (S.E.M). The levels of significance (p values) have been obtained via student’s t-test and one-way ANOVA whichever applicable. p<0.05 have been taken as statistically significant and displayed as * or ** in their respective experiments. Real time (qPCR) results are displayed as relative changes in miR quantity (RQ), calculated by double delta Ct (ΔΔCt) algorithms.

Coronavirus spike protein (S Protein) is the outermost ‘crown like’ structural protein that mediates coronavirus entry into the host cell (56). Coronavirus spike gene have been the prime target for developing an effective vaccine and therapy. Since S protein is supposed to play a crucial role in inducing neutralizing antibody, T-cell response and protective immunity (57–59); we were interested to evaluate the impact of spike protein in perturbing the immune response specifically in context of CNS. We transfected HEK-293T cells with S gene plasmids (pTwist-EF1α-nCoV-2019-S-2×Strep) and harvested the released exosomes on every 24 hours, pooled them and followed the protocol of exosome isolation as displayed in schematic ( Figure 1A ). Exosome characterization were performed as per established standards. The particle size distribution analysis showed that harvested exosomes from control and S gene transfected cells were in size range of a typical exosomal population (mode: 85.9 nm for control exosomes, mode: 107.2 nm for S exosomes ( Figures 1B, C ). Purity of exosomes and exclusion of cellular debris were checked by absence of endosomal protein Calnexin in exosome population ( Figure 1D ) and tetraspanin marker protein TSG101 were also confirmed in harvested exosomes ( Figure 1D ). We have also confirmed the presence of another exosomal marker CD63 in harvested exosomes ( Figure 1I ). Confirmation of SARS-CoV-2 spike protein expression in transfected HEK-293T cells were also done ( Figure 1H ). Disruption of exosomal cargo and manipulation of target recipient cell signaling by SARS-CoV-2 is a completely unexplored area. This prompted us to examine the potentially loaded cargo in the form of microRNAs, mRNA or proteins. We chose to focus on actively loaded microRNAs in these released exosomes because viral infection and inflammation nexus are intricately related with modulation of microRNAs (34). Viral refashioning of host microRNAs that target the expression levels of inflammatory gene expressions are a well-known phenomenon (33, 34, 37, 38, 60). Levels of total secreted exosomes were significantly higher from S-transfected cells compared to mock transfected cells ( Figure 1E ). We observed a significant higher loading of miR-590 (~ 6.6 folds) and ~4.2 folds higher loading of miR-148a in S-transfected cell released exosomes ( Figures 1F, G ).

We wanted to understand the bystander impact of SARS-CoV-2 infection on disruption of CNS homoeostasis. Multiple organ dysfunctions including neurological sequelae during SARS-CoV-2 infection is a widespread observation even when the organ is not directly infected with the virus. To dissect this scenario further, we have used exosome released from spike transfected cells instead of using total SARS-CoV-2 infection. We transfected Spike gene in HEK-293T cells and harvested the released exosomes, which we found full of miR-148a and miR-590 among other cargo ( Figures 1E–G ). We treated human microglia with these loaded exosomes for 24 hours and assessed the protein expression levels of target genes USP33 and IRF9 ( Figure 2A ). USP33 and IRF9 were chosen because they emerged as the potential target of miR-148a and miR-590 respectively, in multiple bioinformatics prediction tools. We checked the miR-148a and miR-590 levels in both, the donor cells (S-transfected HEK-293T cells) and recipient human microglia. In donor cells, level of miR-148a were found slightly higher by only 1.3 folds ( Figure 2B ) indicating that miR-148a is actively loaded in released exosomes. At two different doses of 2µg and 4 µg of S-Exo treatment, level of miR-148a were significantly increased up to 3.5 folds and 16 folds respectively ( Figure 2C ) in recipient human microglia. However, levels of miR-590 in recipient microglia increased only slightly at 4 µg of S-Exo treatment ( Figure 2H ). The cellular levels of USP33 and IRF9 in S-transfected cells (exosomes donor) were found slightly increased compared to control cells ( Figure 2D ).

The cellular expression levels of both targets USP33 and IRF9 were decreased up to 50% and 60% in S-Exo treated human microglia ( Figures 2E, F ). S-transfected HEK293T cells were also treated with GW4869, a pharmaceutical inhibitor of neutral sphingomyelinase, which blocks the inward budding of microvesicles and thereby generation of exosomes. When such blocked conditioned media were used as treatment on recipient microglia, they couldn’t decrease the levels of USP33 and IRF9 ( Figures 2E, F ). This confirmed our idea that released S-Exo is largely responsible for modulating USP33 and IRF9 expression levels. When human microglia were directly transfected with S gene plasmids, the expression levels of USP33 and IRF9 remained largely unaffected ( Figure 2G ).

The direct targeting of USP33 protein expression by exosomal miR-148a were established in our earlier study (29). So we focused for investigating miR-590 mediated regulation of IRF9. Bioinformatics prediction tools such as TargetScan, MicroRNA.org, miRDB; all suggested a potential binding site of miR-590 seed sequences onto 3’UTR of IRF9 ( Figure 3A ). The mirSVR score were found as -1.2841 suggesting a strong free energy change upon this binding which means a strong binding affinity between miR-590 and complementary IRF9 3’UTR sequences. We performed a luciferase reporter assay for this target validation. Co-transfection of IRF9 3’UTR and miR-590 mimics showed a significantly reduced luminescence (~60%) suggesting a binding and blockage of luciferase expression ( Figure 3B ). On the other hand, ATF3, a transcription factor and negative regulator of miR-590 (61), were also blocked with siRNA against ATF3, to elevate the cellular levels of miR-590. In this co-transfection, elevated miR-590 again suppressed the luminescence activity, which suggested that IRF9 expression level is indeed regulated by miR-590 ( Figure 3B ). miR-590 overexpression with the help of commercial miR-590 mimics significantly reduced the target IRF9 protein expression levels (~50%) ( Figures 3C, D ). Contrary to that, transfection of anti-miR-590 significantly rescued the expression levels of IRF9 ( Figures 3E, F ). These standard validation experiments confirmed the regulatory relationship between miR-590 and IRF9 gene.

We initially hypothesized that miR-590 loaded in S-exo would be internalized and significantly elevate the levels of miR-590 in recipient human microglia and therefore would reduce the target IRF9 expression levels. However, qPCR analysis of miR-590 in S-exo recipient microglia showed that upon 2 µg of S-Exo treatment, there was not enough elevation in miR-590 levels in recipient microglia ( Figure 2H ). A small increase in miR-590 in recipient microglia was only observed at 4 µg of S-Exo treatment ( Figure 2H ). This forced us to consider additional regulatory pathways which might be operating prominently in downregulating IRF9 levels in S-exo treated human microglia. Simultaneously, we observed a consistent trend that cellular levels of IRF9 was always following the cellular levels of USP33 in human microglia. Upon S-exo treatment, USP33 levels went down, followed by similar downregulation in IRF9 protein levels ( Figure 4A ). We blocked the cellular levels of USP33 by transfecting siRNA against USP33 and we were able to observe a reduced levels of cellular IRF9 levels ( Figure 4B ). We also transfected human microglia with exogenous miR-148a mimic, which we earlier observed to be loaded and transported within S-exo. Downregulation of USP33 via miR-148a mimic were also followed by reduced levels of IRF9 in human microglia ( Figures 4C, D ). This gave us a primary clue that IRF9 levels and its stability might be under the regulation of USP33, which is a deubiquitinase protein. To confirm this regulatory role of USP33 over IRF9, we performed further experiments. We overexpressed USP33 with Flag-HA-USP33 plasmid. Along with, we also transfected some other USPs, such as USP42, USP7. We observed a specific impact of only USP33 on IRF9 stabilization ( Figure 4E ). USP42 and USP7 were not able to stabilize or increase the cellular levels of IRF9 protein ( Figure 4E ). We also gave different increasing doses of USP33 (1µg, 2µg and 4µg) and observed a dose dependent stabilization of IRF9 protein levels in co-transfection experiments ( Figure 4F ). This confirmed the positive regulatory function of USP33 upon IRF9 protein levels.

USP33, Ubiquitin specific peptidase 33 is a deubiquitinase enzyme (DUBs). By virtue of its deubiquitinase activity, it can remove the ubiquitin tagging and thereby stabilize its specific target proteins. We applied a cell-permeable pyridinamine class broad-spectrum DUB inhibitor, PR-619 on microglial cells and tested the cellular levels of IRF9. IRF9 levels were significantly decreased upon PR-619 treatment at 10µM final concentration ( Figure 5A ). We also treated microglial cells with MG132 (a proteasome inhibitor) at 10µM concentration for 8 hours and found the increased or stabilized cellular levels of IRF9 in microglial cells ( Figure 5A ) This strengthens our hypothesis that cellular IRF9 levels are under control of proteasomal degradation machinery and deubiquitinase enzymes play a vital role in maintaining the IRF9 stability. To establish the function of USP33 as IRF9 stabilizer, we performed multiple experiments. A chase assay with cycloheximide were performed to assess the half-life of IRF9. A co-transfection of USP33 and IRF9 were done along with solo IRF9 overexpression in HEK-293T cells in 6-well plate format. IRF9 solo transfection experiment showed that IRF9 half-life ranges between 3-4 hours ( Figure 5D ). USP33 co-transfection significantly improved the turnover time of IRF9 protein inside cells ( Figure 5E ). IRF9 levels were stabilized for up to 24 hours and beyond ( Figure 5E ).

Since our data suggested a positive regulation of cellular IRF9 levels by USP33, we tested the impact of USP33 on ubiquitination levels of IRF9 protein via in vitro ubiquitination assay. We found a significant decrease in ubiquitinated IRF9 levels when co-transfected with USP33 ( Figures 5B, C ). These results confirmed the role of deubiquitinase USP33 in stabilizing the IRF9 protein levels. These results explain our previous observations of why IRF9 protein levels were concomitantly changing with changing cellular levels of USP33. Any cellular change of USP33, be it by siRNA-USP33, miR-148a overexpression, plasmid mediated overexpression of USP33 and Spike-exosomes mediated downregulation of USP33, they all showed a downstream alterations in IRF9 protein level.

IRF9 expression levels have been considered as an important determinant of viral disease severity (49–51). Previous studies have demonstrated the crucial role of IRF9 in inflammation (62), autoimmune diseases like SLE (63), cardiovascular diseases (64), cell proliferation and immune cell regulation (65). In our experiments, we observed a sharp decline in IRF9 expression levels upon S-exo treatment in human microglia. Considering its multifunctional roles especially for inflammation and autoimmune regulation, we were interested to resolve its specific role in controlling inflammatory gene expression pathways such as NF-κB, TNFα and IFNβ. Firstly, we performed Spike gene transfection along with three major promoter luciferase plasmids of TNFα, IFNβ and NF-κB in a co-transfection experiment and checked the reporter luciferase expression levels as an indicator of these promoter activities. SARS-CoV-2 Spike gene transfection were able to induce all the three cytokine regulatory promoter activities of NF-κB, TNFα and IFNβ ( Figures 6A–C ). We mimicked our experimental set up of treatment with S-exo on cells and checked the promoter activities of these cytokine pathways. We could observe a significant elevation in promoter activity of all three pathways namely NF-κB, TNFα and IFNβ ( Figure 6D ). Since S-exo treatment on human microglia were causing the reduction in cellular IRF9 levels; we checked its nuclear translocation and we could observe a significant reduction of IRF9 nuclear concentration too ( Figure 6E ). In S-exo treated CHME3 cells, IRF9 levels were reduced in nuclear fraction ( Figures 6E, F ). Upon siRNA-USP33 and siRNA-IRF9 transfection too, its concentration in nucleus were found affected ( Figures 6E, F ) which would affect the transcriptional activation of its downstream cytokine expression levels. USP33 and IRF9 axis is supposed to control the inflammation as well as anti-viral state of cells. For investigating the impact of USP33 and IRF9 upon inflammatory gene regulatory network, we overexpressed USP33 and IRF9 in separate experimental set-up. Overexpression of USP33 and IRF9 both were significantly suppressing the promoter activities of NF-κB, TNFα and IFNβ ( Figures 6G, H ).