Dengue Virus Degrades USP33–ATF3 Axis via Extracellular Vesicles to Activate Human Microglial Cells Free

An alarming increase in dengue fever has been recorded worldwide, with repeated sporadic or large outbreaks occurring globally (https://www.who.int/). Dengue fever is caused by dengue virus (DENV), a member of family Flaviviridae, and has been recently recognized as the second leading cause of acute febrile disease in travelers (1, 2). DENV circulates mainly as four different serotypes as DENV-1, -2, -3, and -4, causing a wide spectrum of disease (3). Dengue infection is known to cause various secondary illnesses in the central and peripheral nervous systems (4). Recently, Murthy (5) has proposed that based on its pathogenesis, neurologic manifestations of DENV infections can be majorly categorized into three types: 1) metabolic disturbances such as encephalopathy; 2) meningitis, myelitis, myositis, and encephalitis due to viral invasion; and 3) autoimmune overreactions such as acute disseminated encephalomyelitis, optic neuritis, neuromyelitis optica, and Guillain-Barré syndrome (6). Recent studies also suggest that dengue-mediated neurologic involvements could also be displayed in the form of peripheral nervous system syndromes, optical abnormalities (6, 7) and post-dengue immune-mediated syndromes upon recovery of patients from dengue fever (7, 8). However, their precise molecular mechanisms are still poorly defined. Initially, DENV neurotropism in the human host was considered to be an opportunistic scenario (9, 10), but direct neurotropism and immunological mechanisms are now held responsible for these neurologic manifestations (9).

Microglia represent only 10% of total cells in the normal adult brain (11). Because microglia are brain-resident macrophages and their reactive state have been linked with many neuropathological changes, they emerge as prime suspect in viral neurotoxicity (12, 13). Among all kind of brain cell populations such as neurons, astrocytes, and oligodendrocytes, microglia are predominantly responsible for the inflammatory immune reactions in the CNS (14). They are well documented to play a central role during neuroviral infections as they can be productively infected by the virus, and their immune activation can cause neuroinflammation (15). During neuronal infections, microglia have been shown to physically surround and phagocytose dying neurons and their processes in vivo and ex vivo (16). Modulation of innate immune signaling within human microglia and subsequent neuroinflammation are well documented during Japanese encephalitis virus (JEV) (17), HIV-1 Tat-mediated neurotoxicity (18, 19), CMV, and other viral infections (20). However, DENV-mediated disruption of microglial biology is not dissected in much detail yet.

DENV neuropathogenesis has many intricacies. The time lagging between peak viremia and the onset of neuropathological manifestation is one of them. Additionally, the percentage of DENV-infected cells in host blood is less and has often reported between 2 and 6% of total PBMCs, which is comprised of many different types of cells (21, 22), but still DENV infection is capable of generating the profound “cytokine storm” phenomenon (23). This is suggestive of a widespread disruption in host innate immune responses and altered intercellular communications within the host. Virus-infected cells are known to secrete microvesicles/microparticles (24) for intercellular communication (25, 26). These secreted extracellular vesicles (EVs) have been shown to help them positively for sustaining their life cycle and successful pathogenesis but simultaneously can also act as a messenger for the activation of host antiviral pathways (24–26). The role of these secreted EVs in causing neuroinflammation is increasingly reported now (27, 28). Therefore, the roles of exosomal pathway and release of EVs and their loaded cargo seem instrumental in understanding DENV neuropathogenesis. Exosomes/microvesicles are 30–200 nm small membrane-bound vesicles, generated via endosomal secretory pathway, and can transport multiple functional RNAs, microRNA (miRNA), proteins, and lipids (29–31). The role of packaged miRNA inside exosomes/microvesicles specifically during viral infections has been studied for many viruses like HIV, JEV, influenza, and hepatitis C virus, etc. (32–34). However, the knowledge of the exact role of exosomal miRNA in causing DENV pathologies is still in infancy.

Proteasomal pathway plays a pivotal role in maintaining the cellular levels of almost all proteins, including immune signaling molecules (35, 36). The proteasomal pathway comprises many E1, E2, and E3 ligases, which add the ubiquitin to target substrate molecule and multiple deubiquitinases (DUBs), which constantly remove the ubiquitin molecule from its substrate protein and thereby maintain the homeostasis of cellular protein levels (36). Ubiquitination/deubiquitination is a central housekeeping process that primarily regulates protein turnover but also controls protein localization, protein functions, and protein–protein interactions. This makes the ubiquitination/deubiquitination process a favorite pathway to be extensively exploited by many viruses (35, 37, 38). Host innate immunity and, in general, proinflammatory pathways are known to be regulated via modulating DUB levels in host cells (39). However, studies dissecting the role of the proteasomal pathway, especially DUBs, during DENV pathogenesis are less explored. We have previously reported how DENV infection and specifically NS5 protein could alone degrade a DUB protein USP42 in the human microglial cell (40).

In this study, we have examined the role of EVs secreted from DENV-infected monocytic cells, identified miR-148a being transferred via EVs, and its regulatory impact on ubiquitin-specific peptidase 33 (USP33), a DUB protein of recipient human microglia. We also investigated the novel function of USP33 protein in influencing cytokine regulatory pathways via stabilizing ATF3 in human microglia.

DENV serotype 2 (DENV2/DV2), a prevalidated strain of New Guinea C strain, have been propagated in C6/36 cells, and cultured in Leibovitz L‐15 medium (catalog no. 11415‐049; Life Technologies) supplemented with 10% FBS and 1× penicillin–streptomycin–glutamine mix (catalog no. A002A; HiMedia Laboratories) at 28°C in the absence of an external CO₂ supply. Cell supernatants were harvested from infected C6/36 on every fourth day for up to 4 wk. Supernatants were centrifuged at 14,000 Rpm to settle down cellular debris, filtered through 0.22-μM filters, and stored at −80°C until further analysis. DENV concentration was measured via standardized viral plaque assay. DENV infection has been given at 5 multiplicity of infection (MOI) to THP1 cell lines in serum-free media. After 4 h of virus incubation, cells were washed with serum-free media and replaced with complete media with the usual 10% FBS and 1× penicillin–streptomycin–glutamine mix.

Human microglial cell line CHME3 has been used for studying the impact of secreted EVs. HEK293T cells have been used for luciferase assay, in vivo ubiquitination assay, and chase assay. C6/36 cells have been used for DENV propagation. THP1 (human monocytic cell line) cells have been used for DENV infection experiments. CHME3 and HEK293T and Huh7 hepatocytes cells have been cultured in DMEM (catalog no. 12-604F; Lonza) supplemented with 10% FBS (catalog no. RM1112; HiMedia Laboratories) and 1× penicillin–streptomycin–glutamine (catalog no. A002A; HiMedia Laboratories) in a humidified chamber at 37°C temperature with a constant flow of 5% CO₂. THP1 cells have been cultured in RPMI 1640 (catalog no. 12-702D; Lonza) with 10% FBS and 1× penicillin–streptomycin–glutamine. Transfection of DNA plasmids have been done with Lipofectamine 2000 (catalog no. 11668019; Invitrogen) as per the manufacturer’s instructions. pCDNA 3.1‐His‐NS1 plasmids were a kind gift from Dr. R. Mohana‐Borges (Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil) (41). Flag-USP7 was a kind gift provided by A. Wani from Ohio State University (Columbus, OH). pRK-ATF3 (catalog no. 26115; Addgene) (42), Flag-HA-USP33 (catalog no. 22601; Addgene) (43), and IFN-β-pGL3 (catalog no. 102597; Addgene) (44) constructs were purchased from Addgene. Renilla luciferase plasmid used for transfection control was a kind gift given by Dr. V. Natrajan (Institute of Genomics and Integrative Biology, Delhi, India). The 6× His-ubiquitin plasmids were a kind gift provided by Prof. D. Xirodimas (University of Dundee, Dundee, UK). NF-κB-Luc and TNF-α-luc plasmids was obtained from Stratagene.

For Western blotting, all treated cells have been lysed with RIPA lysis buffer (20 mM Tris [pH 7.5], 150 mM NaCl, 1 mM EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM glycerophosphate, 1 mM Na₃VO₄, and 1 g/ml leupeptin). The 1× protease inhibitor mixture (catalog no. S8820; Sigma‐Aldrich) was added to protect cell lysate from degradation. Total protein estimation has been done with the Bradford assay (catalog no. 500‐0006; Bio‐Rad Laboratories). Thirty to forty micrograms of total protein samples were run in 8% polyacrylamide gel followed by transfer onto a nitrocellulose membrane (catalog no. SCNJ8101XXXX101; Advanced Micro Devices) at 100 V for 2 h. Blocking was done with 5% skimmed milk powder (catalog no. GRM1254; HiMedia Laboratories) dissolved in 1× Tris‐buffered saline with Triton X‐100 (TBST) for 1 h at room temperature with slow rocker shaker movement. Primary Abs were given at 1:1000 dilutions overnight at 4°C temperature on slow rocker shaker. Membranes were washed with three washes of TBST for 15 min each. Respective secondary Abs were incubated onto the membrane for 1 h at room temperature, followed by three washes with TBST (15 min each). Anti-rabbit IgG conjugated to HRP and anti-mouse IgG conjugated to HRP were purchased from Jackson ImmunoResearch Laboratories. Blots were developed with ECL Western Blotting Substrate (catalog no. 32106 Pierce; Thermo Fisher Scientific) while capturing the images on x-ray films supplied by Carestream Health (catalog no. 6568307). The density of Western blot images has been quantified by ImageJ software version 1.52. The image density of the corresponding GAPDH lanes of the same blot have been used as a normalizer in all the Western blotting experiments.

Primary Abs used in this study are as follows: anti-calnexin rabbit mAb (catalog no. 2679; Cell Signaling Technology), GAPDH (catalog no. Sc‐32233; Santa Cruz Biotechnology), USP7 (catalog no. D17C6; Cell Signaling Technology), TSG101 (catalog no. Sc-13611; Santa Cruz Biotechnology), ATF3 (catalog no. Sc-81189; Santa Cruz Biotechnology), USP33 (catalog no. Sc-100632; Santa Cruz Biotechnology), and DENV type 2 envelope mAb GT643 (catalog no. MA5-17291; Invitrogen). MG132 (catalog no. C2211; Sigma‐Aldrich) is a membrane-permeable proteasome inhibitor. It has been used at 10 μM final concentration for an in vivo ubiquitination assay. GW4869 is a neutral sphingomyelinase and exosome biogenesis inhibitor (catalog no. D1692-5MG; Sigma-Aldrich), used at 10 μM final concentration as indicated in respective experiments. PR-619 is a DUB inhibitor (catalog no. SML0430-1MG; Sigma-Aldrich), used at 5 and 10 μM final concentration as indicated in respective experiments. USP33 small interfering RNA (siRNA) were purchased as FlexiTube siRNA (catalog no. SI00109123; QIAGEN).

pCDNA 3.1‐His‐NS1 were transfected in HEK293T cells in a 90-mm dish with Lipofectamine 2000. Cells were expanded after transfection, and culture supernatants were collected after every third day for up to 1 wk. Empty vector transfected cells were treated similarly. All the supernatants were pooled, and EV isolation was done as described by Miranda et al. (45). Briefly, cell supernatants were centrifuged at 2000 Rpm to remove cell debris. These supernatants were collected and subjected to the “salting out” procedure for the isolation of EVs. Then, 0.1 vol of sodium acetate buffer (1 M and pH 4.75) were added to cleared supernatant and incubated on ice for 30–60 min, followed by thermal incubation of supernatant at 37°C for 5 min. The resulting turbid suspension was centrifuged for 10 min at 5000 × g, and the pellets were washed with 0.1 M Na acetate buffer. This suspension was again centrifuged at 5000 × g for 60 min, and the final EV pellet was “solubilized” in PBS. Concentrations of EVs were measured via Bradford reagent. TSG101 expressions were confirmed in all harvested EV preparations. The absence of calnexin protein was also confirmed in all EVs harvested by this method. Harvested EVs were further subjected to various miRNA and protein analysis. For treatment, an equal amount (2 or 4 μg as indicated in respective experiments) of control and NS1-derived EVs were treated on CHME3 cells for 24 h.

RNA isolation was done with miRNeasy Kit (catalog no. 217004; QIAGEN). Protocol was followed as per the instructions given in the kit manual. TaqMan Reverse Transcription Kit (catalog no. 4366596; Applied Biosystems) was used for cDNA preparation by using miRNA-specific primers. Thermal incubations for reverse transcription were as follows: 16°C for 30 min, 42°C for 30 min, and 85°C for 5 min. Universal PCR Master Mix (catalog no. 4324018; Applied Biosystems) was used for miRNA quantitative PCR (qPCR) analysis. Cellular levels of miR-148a were estimated with commercial TaqMan assay (TaqMan Assay identifier: 000470 for human miR-148a and RNU24 no. 001001; Applied Biosystems). RNU24 was used as a qPCR internal loading control. Thermal cycles were as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Applied Biosystems 7500 Fast Real-Time PCR cycler has been used for all qPCR experiments.

MiRNA bioinformatics prediction tools such as Pictar, Target Scan 7.1, and microRNA.org have been used to search the targets of human miR-148a. Seed sequences of miR-148a were found to have a strong complementary binding site in 3′ untranslated region (3′UTR) of USP33 gene.

To analyze the posttranscriptional regulation of USP33 gene by miR-148a, miR-148a mimics (catalog no. MC10263; Thermo Fisher Scientific) and antago-miR-148a (catalog no. AM10263; Thermo Fisher Scientific) have been used. CHME3 cells were seeded in a six-well plate, with 1 × 10⁵ cells in each well, 1 d before transfection. Mimics and anti–miR-148a transfection were given at 100 pmol per well with the help of specific Lipofectamine RNAiMAX Transfection Reagent (catalog no. 13778150; Invitrogen) as per the manufacturer’s instructions. After 48 h of transfection, CHME3 cells were harvested and subjected to RNA isolation and immunoblot analysis.

For luciferase promoter activity assay of TNF-α, IFN-β, and NF-κB activity, HEK293T cells were seeded in a six-well plate prior to cotransfection. One microgram plasmids of TNF-α, IFN-β, and NF-κB were cotransfected with 1 and 2 μg of USP33 plasmids and different doses of 1, 2, and 3 μg of ATF3 plasmids as well as 500 ng of Renilla luciferase plasmids in various experiments as shown in their respective results. After 24 h of transfection, cells were lysed and processed for dual luciferase assay as per the kit’s instructions (catalog no. E1910; Promega). Luminescence readings were captured on Synergy H1 Multi-Mode Reader (BioTek Instruments). Renilla luminescence have been taken as a normalizer to get the final luciferase activity in all the experiments.

Total RNA were extracted from the HEK293T cells infected with DENV2 by using QIAzol Lysis Reagent according to the manufacturer’s instructions (QIAGEN). RNA was quantified by using NanoDrop, and 500 ng of total RNA were subjected to estimate viral RNA copy number. Quantitative RT-PCR assay was based on the amplification of a DENV2 capsid region–specific primers and probe (46). The reactions were performed using the TaqMan Fast Virus 1-Step Master Mix kit (Applied Biosystems) on a StepOne Plus Real-Time PCR system (Applied Biosystems). The concentrations of DENV2 viral RNA (copies/milliliter) were estimated by using the standard curve generated from DENV2 transcripts.

To check the impact of USP33 on the levels of ubiquitinated ATF3, HEK293T have been used for all the transfection experiments. Cells were grown in a 100-mm dish for up to 70% confluency at the time of transfection. An equal amount of 7.5 μg ATF3 and 5 μg of 6× His-ubiquitin plasmids were transfected along with two doses of USP33 (5 and 10 μg). After 30 h of incubation, MG132 were treated at 20 μM final concentration for 8 h in all the dishes. An in vivo ubiquitination assay was done as described in detail in our previous publication (47). After final elution, all the samples were boiled with 2× Laemmli buffer and immunoblotted with anti-ATF3 Ab.

HEK293T cells were transfected with the respective genes such as ATF3 only and ATF3 and USP33 cotransfection with the help of Lipofectamine 2000. After 24 h of transfection, cycloheximide (catalog no. 01810; Sigma-Aldrich) was given at a final concentration of 100 μg/ml to stop further translation processes in cells. Cells were harvested at different time points. Cell lysates were prepared in RIPA buffer, and immunoblottings were done for the indicated proteins.

All the comparative results in graph bars are shown as mean plus SEM from three independent biological repeat experiments. Real-time PCR results are displayed as relative levels of miR‐148a, as calculated by ΔΔCt algorithms. The levels of significance (p values) between untreated and treated samples were estimated by ANOVA, and p < 0.05 has been taken as significant.

Monocytes and macrophages are naturally permissive cells for productive DENV infection (21, 48). To assess the changes in the release of EVs and to study their cargo, we infected THP1 (1 million cells) with 1 MOI of DENV virus (serotype 2), also referred to as DV2. After every third day, culture supernatant from infected cells were harvested, pooled, and further processed for EV isolation (Fig. 1A, schematic). In every batch, EVs yield ranged from 1.8 to 2.0 μg per million cells over a 3-d harvest in control uninfected cells. We observed up to ∼1.5-fold increase in the total amount of released EVs from DENV-infected THP1 cells (Fig. 1B). The purity of released EVs was confirmed via immunoblotting of endosomal protein calnexin (Fig. 1C), which should be absent in any EV preparation. The presence of TSG101 was confirmed in isolated EVs to show the characteristic tetraspanins in secreted vesicles (Fig. 1C) via Western blotting analysis. Secreted EVs are reported to carry many messenger molecules in the form of miRNA, mRNA, DNA, protein, etc. We sought to identify potential miRNAs in EVs secreted from DENV-infected cells. Viruses have been potentially known to modulate cellular miRNA profiles upon infection (49–51). We have also previously reported that even single viral toxin proteins such HIV-1 Tat and DENV-NS5 are capable of perturbing the cellular miRNA profile in neuronal cells as well as macrophages (18, 40). Harvested EVs from DENV-infected THP1 were used for total RNA isolation, and miR-148a levels were checked and compared with uninfected THP1-secreted EVs. We found almost 3-fold higher levels of miR-148a loaded in EVs secreted from DENV-infected monocytes (Fig. 1D). We observed a sharp decline in cellular USP33 expression levels in human microglial cells after treatment with EVs released from DENV-infected cells (Fig. 1E, 1G). GW4869-treated DENV-infected cells, which would release a reduced amount of exosomal vesicles, showed almost no change in USP33 protein levels compared with mock recipient microglial cells (Fig. 1E, 1G). ATF3 protein expression levels followed a similar pattern as cellular USP33 levels (Fig. 1F, 1G). DENV-infected cell–released EVs also showed decreased ATF3 levels in microglial cells, whereas EVs received from GW4869-treated DENV-infected cells showed almost no change in ATF3 levels (Fig. 1F).

We aimed to investigate the bystander effect of DENV infection on uninfected host cells and organs located distant from the site of infection. Specifically, the mechanism of DENV-mediated neuroinflammation is observed when DENV viral titers declines in the host body. Because EVs that are released from virus-infected cells might carry viral particles within themselves, we wanted to rule out any direct role of viral RNA or viral particle in modifying recipient microglial cells in this study. For this purpose, we transfected DENV-NS1 gene in HEK293T cells and similarly collected the secreted EVs (schematic shown as Fig. 2A). We obtained ∼2.0–2.5 μg of EVs per million cells over a 3-d culture supernatant harvest. Multiple flasks were harvested and pooled to get the bigger yield of EVs for characterization and treatments. Before exposing these EVs on human microglial cells, we analyzed the secreted EVs for their size distribution on NanoSight (nanoparticle tracking analysis [NTA] 3.2 Dev Build 3.2.16; Malvern Panalytical. NS1-transfected cell–released EVs were falling in size range above 100 nm (mode = 107.1 nm), which confirmed that they belonged to the microvesicles category (Fig. 2B). Pelleted EVs were subjected to RNA isolation, and loaded miR-148a levels were checked via qPCR. Because total EVs pellet have been used for RNA isolation and miR-148a qPCR analysis, it is unspecified how much percentage of EVs would specifically contain miR-148a inside them. As compared with control/mock-transfected cells, DENV-NS1–transfected cells secreted significantly higher levels (∼5-fold) of miR-148a (Fig. 2C). Upon fusion with target recipient cells, they release all of their content to target cells, which further modifies the phenotype of recipient cells. We exposed the human microglial cells CHME3, with these EVs and miR-148a levels, which were checked within microglial cells. We found almost ∼7-fold increase in levels of miR-148a in microglial cells as compared with control EVs-treated cells (Fig. 2D). To find out the potential target of miR-148a, we used various bioinformatics prediction tools such as microRNA.org, Targetscan, and Pictar. USP33, a DUB, emerged as a potential target candidate with a mirSVR (miRNA support vector regression algorithm) score of −1.1025 (Fig. 3A). Therefore, we checked the cellular levels of USP33 protein in microglia treated with these EVs. In a dose-dependent manner, USP33 levels went down to 70% in EV-treated microglia compared with mock treated microglia (Fig. 2E, 2F). This result suggested a plausible regulatory role of miR-148a over USP33 gene expression. We also blocked the EV secretion from NS1-transfected cells by applying GW4869. GW4869 is a commonly used pharmaceutical inhibitor to block the ceramide-mediated inward budding of multivesicular bodies. Microglial cells were treated with the EVs harvested from these GW4869-treated NS1-transfected cells. When vesicle secretion was blocked from donor cells (DENV-NS1–transfected cells), the receiving microglial cell did not show any significant changes in USP33 expression levels (Fig. 2G). This observation strengthened the notion that the regulation of cellular USP33 levels in microglia was due to secreted EVs from DENV-NS1–transfected cells.

An increased level of miR-148a and concomitant decreased cellular levels of USP33 gave a strong indication of regulatory interaction between miR-148a and USP33 gene expression in microglial cells. Bioinformatics prediction tools also suggested a strong binding affinity between seed sequences of miR-148a and to complementary 3′UTR region of USP33 (Fig. 3A). For experimental validation of this interaction, we performed all standard experiments such as miR-148a overexpression via mimics and blocking of cellular miR-148a by antago-miR-148a. Upon 100 pmol of miR-148a mimic transfection, USP33 protein levels in human microglia significantly decreased to 60% (Fig. 3B, 3C). Contrary to that, antago-miR-148a transfection in microglial cells sequestered out the cellular miR-148a levels. This pulling away of cellular miR-148a via antagomirs rescues the 3′UTR and allows the translation of the target gene to happen at higher levels. We observed this effect upon transfection of antago-miR-148a, which elevated the expression levels of USP33 protein up to 2-fold (Fig. 3D, 3E).

To assess the functional implications of decreased USP33 levels in modulating immune responses of microglial cells, we checked cellular ATF3 levels in microglia upon DENV-NS1 EV exposure. Because ATF3 is a key regulator of macrophage IFN responses (52) and microglia are basically brain-resident macrophages, we wanted to investigate if ATF3 is being targeted to alter the inflammatory state of the CNS. Microglia were exposed to DENV-NS1–released EVs for 24 h, harvested, and subjected to immunoblot analysis to check ATF3 expression levels. EVs treatment on microglia (which already decreases USP33 levels in microglia) showed a significantly decreased (∼60%) ATF3 level (Fig. 4A). Because hepatocytes are also well known to be affected upon DENV infection, we checked human hepatocytes (Huh7 cell line) to check the effects of EV exposure on ATF3 levels; both type of cells (Huh7 and CHME3) displayed the decrease in cellular ATF3 levels to the same extent (Fig. 4A, 4B). GW4869 inhibitor–treated EVs were also tested on microglia to confirm that ATF3 modulation is being done via secreted EVs. When EV donor cells (DENV-NS1–transfected cell) are treated with exosomal inhibitor, the release of EVs significantly went down, and that is why EVs collected from such cells have no effect on ATF3 expression levels in recipient microglia (Fig. 4C, 4D).

Our observations so far indicated that exposure with DENV-NS1 EVs have similar effects on cellular USP33 and ATF3 levels. To explore this relation further and to explore whether cellular USP33 might be a regulatory factor for ATF3 protein levels in human microglial cells, we performed many experiments. We exogenously overexpressed miR-148a via miR-148a mimic and checked the ATF3 levels in microglia. Immunoblot analysis clearly showed that a decrease in the USP33 protein level is accompanied by a decrease in cellular ATF3 protein levels (Fig. 4E, 4F). In the next proof-of-principle experiment, we inhibited the cellular miR-148a via transfecting anti–miR-148a in microglia. We showed in previous results that anti–miR-148a sequestered the endogenous miR-148a and thereby rescues the USP33 expression (Fig. 3D). Cellular ATF3 proteins followed a similar pattern of rescued expression levels as shown by immunoblotting (Fig. 4G, 4H). Additionally, we also used USP33-specific siRNA to block the endogenous USP33 protein expression and analyzed the cellular ATF3 protein expression levels in microglia. USP33 siRNA effectively blocked the USP33 expression levels up to ∼95% (Fig. 4I). In the same line, USP33 siRNA was able to diminish cellular ATF3 protein levels in microglia (Fig. 4I). All of these experimental evidences strongly suggested that USP33 is required to maintain cellular ATF3 levels in human microglia.

We earlier established a correlation between expression levels of USP33 and ATF3 in human microglia (Fig. 4). Because USP33 is a DUB protein, we explored the possibility whether USP33 could be responsible for stabilizing ATF3 in microglia through its DUB activity. We begin the experiment by analyzing the impact of MG132 (proteasomal inhibitor) on ATF3 stability. MG132 treatment (10 μM for 24 h) significantly stabilized the ATF3 levels (Fig. 5A). In contrast, microglia treated with general DUB inhibitor PR-619 showed an accelerated degradation of ATF3 protein (Fig. 5B, 5C). These two observations confirmed the turnover of ATF3 to be predominantly regulated via proteasomal degradation pathway. To check the specificity of USP33 upon ATF3 stability, we overexpressed the USP33 and another irrelevant USP such as USP7 in a simultaneous experiment. Only USP33 was able to stabilize the ATF3 protein levels and not the USP7 (Fig. 5D). We also performed a chase assay with cycloheximide to confirm the role of USP33 as a stabilizer of ATF3 protein levels. This assay was performed in HEK293T cells via a standard cotransfection experiment with ATF3 and USP33 plasmid constructs. ATF3 solo transfection showed its natural turnover pattern in cells with a half-life between 10 and 12 h (Fig. 5F). When cotransfected with USP33, ATF3 levels were significantly stabilized for up to 24 h and beyond (Fig. 5G). This experiment confirmed the role of USP33 in stabilizing ATF3 protein levels.

To establish the role of USP33 in deubiquitinating ATF3 protein, we performed an in vivo ubiquitination assay. In this experiment, we transfected pRK-ATF3– and His-ubiquitin–expressing plasmids with two different doses of Flag-HA-USP33 plasmids (5 and 10 μg) in HEK293T cells within T-25 cell culture flasks. After 36 h of transfection, MG132 was added to all the flasks, and after an 8-h treatment, cells were harvested for ubiquitination assay. Ubiquitination assay was performed as described in 2Materials and Methods, and immunoblotting was carried out to assess the ATF3 ubiquitination levels. We observed a dose-dependent impact of USP33 protein on the ubiquitinated levels of ATF3 protein (Fig. 6A–C). USP33 overexpression significantly decreased the ubiquitinated ATF3 proteins in the cells, which explains why USP33 overexpression and anti–miR-148a treatment were able to protect the ATF3 from proteasomal degradation. This also explains the concomitant change in the ATF3 proteins along with the changes of USP33 levels in human microglia.

The impact of USP33 in stabilizing ATF3 was also suggestive that it would modulate the expression levels of proinflammatory genes. To test the impact of USP33 and ATF3 gene expression levels on DENV replication, we performed a qPCR assay by using DENV2 capsid region–specific primers and probes. HEK293T cells were transfected with 2 μg of pRK-ATF3 and 2 μg of Flag-HA-USP33 plasmids in separate wells. After 12 h of transfections, these cells were infected with 5 MOI of DENV2 virus for a further 24 h. ATF3 overexpression downregulated the DENV2 replication up to ∼60%, and USP33 overexpression could reduce the DENV2 replication up to ∼50% (Fig. 6C). For confirming these results, we also evaluated the impact of USP33 overexpression on DENV2 replication through Western blotting by using the envelope gene Ab. Western blotting analysis showed that USP33 overexpression (which increases the cellular ATF3 levels too) has indeed suppressed the DENV2 replication (Fig. 6D–F).

ATF3 is a member protein of the ATF/CREB family of transcription factors. It is a well-known negative regulator of inflammation (53, 54). ATF3 gets induced in multiple cellular stress responses such as serum variations and modulation of protein synthesis and is extensively involved in cell growth and survival pathways (55). Because DENV-secreted EVs were able to modulate microglial ATF3 levels via the modulation of USP33 expression levels, we were interested in investigating the impact of ATF3 perturbation on major representative cytokine pathway genes, such as TNF-α, IFN-β, and NF-κB. We performed luciferase promoter activity assay with promoter-luciferase constructs and their cotransfection with various experimental variables, such as USP33 overexpression, ATF3 overexpression, miR-148a overexpression, etc., in HEK293T cells. Mimicking the EV cargo that delivers miR-148a to microglial cells, we overexpressed miR-148a in HEK293T cells along with different promoter-luciferase constructs. miR-148a transfection significantly enhanced the cytokine expression levels for TNF-α and NF-κB (Fig. 7D, 7E); however, IFN-β promoter activity was not much affected (Fig. 7F). This can be explained as type I IFN pathways are regulated by multiple routes; besides, miR-148a is also capable of modulating multiple target genes altogether. We also checked the impact of USP33 overexpression on all three (TNF-α, IFN-β, and NF-κB) cytokine expression patterns via a promoter-luciferase assay. USP33 was overexpressed at two different doses, 1 and 2 μg, and it effectively downregulated the TNF-α as well as NF-κB activity (Fig. 7D, 7E). USP33 also downregulated the IFN-β promoter activity but to an insignificant extent (Fig. 7F). Because we showed earlier that USP33 is directly regulating the stability of cellular ATF3 levels, we checked the impact of ATF3 separately on the cytokine promoter activity (TNF-α, IFN-β, and NF-κB). ATF3 was overexpressed in a dose-dependent manner as 1, 2, and 3 μg, and the respective promoter–luciferase constructs were cotransfected. Luciferase data clearly showed that ATF3 is indeed a negative regulator of inflammatory gene expression. All three (TNF-α, IFN-β, and NF-κB) cytokine promoter activities were attenuated by ATF3 overexpression (Fig. 7A–C).

Research studies suggest the critical role of a dysregulated host immune responses during DENV pathogenesis (29). The host immune responses are usually antiviral in the beginning; however, in later stages, it can turn proviral because of various modulations enforced by viruses.

EVs, also referred to as microparticles, which are an extremely heterogeneous population of cell-derived vesicles, have been extensively studied as a moderator of viral pathogenesis (56, 57). Their biogenesis is an energy-dependent, ubiquitous cellular process termed as “vesiculation” or “ectocytosis,” which happens spontaneously as well as in response to various stresses and pathogenic stimuli (57). These changes have been used as an indicator of various pathological conditions, such as cancer, hyperinflammation, tissue injury, cardiovascular, hematological, and infectious diseases (58–60).

In this study, the rationale behind investigating the loaded cargo of EVs secreted after DENV-NS1 transfection was to find out the specific molecular messengers in terms of miRNA or transcription factors, etc., which are acting as toxic trails of DENV infection. It has been quite evident by recent epidemiological studies that DENV peak viremia does not corroborate the neurologic perturbations and hemorrhagic outcomes in patients (41, 61). Circulating DENV-NS1 protein alone is widely reported to contribute to dengue pathogenicity and disease severity (62, 63).

A big part of DENV pathogenesis involves the heavy secretion and circulation of inflammatory cytokines and chemokines (TNF-α, IFN-β, and ILs, etc.). These secreted signals of hyperactivation can circulate in blood/plasma and travel throughout the host bodies and ultimately enter the CNS. This idea has shaped our experimental design to study the impact of DENV-secreted EVs on human microglial cells. Because microglial cells are the most critical executor of immune responses and a defender in the CNS, they became our cellular model for studying DENV-mediated neuropathogenesis. We sought to mimic a situation in which DENV peak viremia has passed, and modified circulating host cellular factors are now in charge of bringing and sustaining the pathogenesis. We transfected HEK293T cells with DENV-NS1 and harvested loaded cargo in the secreted EVs. We also blocked the exosomal secretion from donor cells via GW4869 application, which showed that most of selected phenotypes like USP33 expression and ATF3 expression levels in recipient microglial cells remained unaffected (Figs. 2G, 4C, 4D). These results confirmed the idea that the route of regulatory signals is predominantly via the exosomal pathway. Because EVs are a heterogenous population of cell-secreted vesicles (size ranging from 30 to 200 nm) (30, 31), it is presumed to contain multiple of miRNAs/TFs in a different subset of the exosomal population (64). In our preliminary studies, we checked multiple miRNAs (miR-590, miR-181, miR-374, etc.), which could be potentially loaded in these secreted EVs. miR-181, miR-374, etc., were present in EVs but did not show any regulatory interactions with the USP33 gene when examined via bioinformatics prediction tools. Only miR-148a showed a strong binding interaction with USP33. We found that miR-148a was the common miRNA that was present in both DENV-infected monocyte secreted EVs as well as NS1-transfected cell–secreted EVs. That is why we chose miR-148a because single DENV-NS1 transfection was sufficient to potentially load it into secreted EVs. Another reason for focusing on the function of miR-148 was that in recipient microglial cells, the levels of miR-148 were highly upregulated after absorption of EVs as compared with other EV-loaded miRs like miR-590. The reason for focusing on the impact of DENV-NS1 on exosomal cargo and downstream regulatory cascade was that NS1 is already reported to have a major role in DENV pathogenesis, and it circulates in the plasma/blood of DENV patients for a longer duration in stable hexameric forms (65). Some clinical studies also demonstrated higher NS1 proteins to be associated with more severe disease in patients (66, 67). Previous studies have also demonstrated that aberrant miRNAs expression profiles are in fact a remarkable characteristic of patient’s plasma postinfection with DENV (68). Our results indicated that DENV-NS1–transfected cells release higher level of miR-148a in EVs compared with DENV-infected cells. This is plausible because NS1-transfected cells released more amount of total EVs compared with DENV-infected cells. This can also be partially explained by the fact that during total virus infection, all of DENV viral proteins, such as NS5, NS4a, NS2a, envelope protein, and capsid, etc., would be expressed and play their individual role in overall DENV pathogenesis, which sometimes could be antagonistic to each other. We have previously reported in another study that two different HIV-1 proteins might antagonize each other’s functions. HIV-1 Rev was shown to downregulate Tat expression and HIV-1 replication via NAD(P)H:quinine oxidoreductase 1 function (69). This explains the observation that in a specific protein transfection scenario such as NS1 overexpression, the magnitude of the host response could vary if compared with a complete virus infection.

A recent study of EVs secreted from DENV3-infected dendritic cells demonstrated that the EV pathway can be effectively exploited by DENV for different immune modulations (68). Another study in platelets also demonstrated that platelet-derived microparticles or EVs play a crucial role in DENV-induced lethality via CLEC5A/TLR2 (70). In our study, we have chosen the EVs secreted from DENV-infected monocytes as well as DENV-NS1–transfected cells. Because the International Society for EVs has updated their new nomenclature of small EVs of 40–120 nm size to be collectively called EVs (31, 71), we have taken these sized ranged particles as EVs and used them in our analysis.

USP33 is a deubiquitinating enzyme and plays an important role in a variety of cellular processes. The role of USP33 has been extensively studied in various cancer and tumor models such as gastric adenocarcinoma, colorectal cancer, and epithelial–mesenchymal transitions, etc. (72, 73). However, the role of USP33 in viral pathogenesis, particularly flavivirus infections and associated neuroinflammation, is almost nonexisting. To our knowledge, this is the first study to demonstrate the role of USP33 in neuroinflammation. Similarly, ATF3 is a well-known suppressor of proinflammatory responses and has been reported in various diseases and pathogenic stimuli, such as multiple cancers and JEV infection (74–76). ATF3 is rightfully considered a hub of the cellular adaptive–response network in various diseases in which inflammation regulation becomes an important contributing factor (75). A previous study has elegantly showed that ATF3 is a negative regulator of antiviral signaling, especially type I IFN pathway genes in the case of JEV infection (76). ATF3 is also reported to block the reactivation of HSV triggered by neuronal stress (77). It plays an important role in modulating the IFN responses of macrophages via attenuating basal and inducible levels of IFN-β (52). The type I IFN is known to suppress virus infections by creating a nonpermissive state in infected and uninfected cells (78, 79). During Neisseria gonorrhoeae, upregulated ATF3 is known to inhibit proinflammatory IL-6 expression levels (78). DENV infection is previously reported to inhibit type I IFN production in infected myeloid cells by cleaving human STING protein (79). In dendritic cell infection scenario, DENV is again reported to suppress type I IFN production (80). In this study, the author could show that DENV negatively influences the dendritic cell capacity to prime naive T cells toward Th1 immunity, therefore attenuating or skewing the adaptive immune responses (80).

Our study is, however, presenting a new bystander regulatory pathway beginning with the DENV-secreted EVs. These EVs are carrying mature miR-148a, which gets internalized by target recipient microglial cells. Imported miR-148a in microglia begins a regulatory cascade for suppressing USP33 via binding with complementary 3′UTR sequences of USP33. USP33, by virtue of being DUB, protects the ATF3 protein from degradation. Our study is the first, to our knowledge, to report this stabilizing effect of USP33 on ATF3 protein levels. We could show by various experiments that USP33 effectively deubiquitinates ATF3 (Fig. 6A, 6B). We have demonstrated that any regulatory milieu that can modify the cellular USP33 levels, such as overexpressed USP33, USP33 siRNA, upregulated miR-148a, blocked out miR-148a, EV exposure, and blockage of EVs secretion, can modify the ATF3 protein levels concomitantly (Figs. 4–6).

We also demonstrated that miR-148a/USP33/ATF3 axis triggered upon the absorption of DENV-secreted EVs could ultimately influence the expression levels of various proinflammatory cytokines (Fig. 7). Our luciferase reporter assays for TNF-α, NF-κB, and IFN-β could show that via modulating miR-148a/USP33/ATF3 cascade, DENV-secreted EVs can effectively alter the proinflammatory gene expression levels. We also demonstrated that USP33 and ATF3 could diminish the DENV replication level (Fig. 6D–F). Overall, these results establish that DENV infection uses the EV secretion route to modify the neighboring cells in a bystander fashion. It sends out a signal in the form of miR-148a to suppress USP33/ATF3 to prepare the uninfected cells for more favorable DENV infection. However, at the same time, USP33 and ATF3, being a negative regulator of inflammation, suppression by DENV can also trigger a massive production of proinflammatory genes such as TNF-α, NF-κB, and IFN-β genes. Considering the special location and protective role of microglia inside the CNS, this pathway becomes a challenging double-edged sword that can create uncontrolled neuroinflammation and related neuropathological consequences.

In this study, we are reporting for the first time, to our knowledge, that stability of ATF3 protein, a central inflammation suppressor molecule, is regulated via DUB USP33. Diminished levels of USP33/ATF3 can trigger the uncontrolled expression of proinflammatory genes, making the microglia hyperactivated and leading to a neuroinflammatory state inside the CNS. Our study offers many crucial intermediate immunoregulatory check points, which can be explored further, for finding new treatment modalities to combat DENV pathogenesis.

We express sincere gratitude for receiving various research materials from different laboratories. We sincerely thank Dr. Ronaldo Mohana-Borges (Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil) for providing pCDNA 3.1-His-NS1 plasmids. We acknowledge all the plasmid constructs purchased via Addgene, and their respective references have been mentioned in the manuscript. We also thank the laboratory of Duane Gubler, Adjunct Professor of International Health (Johns Hopkins School of Hygiene and Public Health) for providing the DENV virus. We are highly obliged to receive the kind gift of human microglial cell line (CHME3) from Prof. Anirban Basu (National Brain Research Center, Manesar, India). We are thankful to Translational Health Science and Technology Institute (Faridabad, India) for providing the NanoSight-NTA 3.2 Dev Build 3.2.16 version instrument facility for EV analysis.

The authors have no financial conflicts of interest.