Establishment of a CPER reverse genetics system for Powassan virus defines attenuating NS1 glycosylation sites and an infectious NS1-GFP11 reporter virus

ABSTRACT Powassan virus (POWV) is an emerging tick-borne Flavivirus that causes lethal encephalitis and long-term neurologic damage. Currently, there are no POWV therapeutics, licensed vaccines, or reverse genetics systems for producing infectious POWVs from recombinant DNA. Using a circular polymerase extension reaction (CPER), we generated recombinant LI9 (recLI9) POWVs with attenuating NS1 protein mutations and a recLI9-split-eGFP reporter virus. NS1 proteins are highly conserved glycoproteins that regulate replication, spread, and neurovirulence. POWV NS1 contains three putative N-linked glycosylation sites that we modified individually in infectious recLI9 mutants (N85Q, N208Q, and N224Q). NS1 glycosylation site mutations reduced replication kinetics and were attenuated, with 1–2 log decreases in titer. Severely attenuated recLI9-N224Q exhibited a 2- to 3-day delay in focal cell-to-cell spread and reduced NS1 secretion but was lethal when intracranially inoculated into suckling mice. However, footpad inoculation of recLI9-N224Q resulted in the survival of 80% of mice and demonstrated that NS1-N224Q mutations reduce POWV neuroinvasion in vivo. To monitor NS1 trafficking, we CPER fused a split GFP11-tag to the NS1 C-terminus and generated an infectious reporter virus, recLI9-NS1-GFP11. Cells infected with recLI9-NS1-GFP11 revealed NS1 trafficking in live cells and the novel formation of large NS1-lined intracellular vesicles. An infectious recLI9-NS1-GFP11 reporter virus permits real-time analysis of NS1 functions in POWV replication, assembly, and secretion and provides a platform for evaluating antiviral compounds. Collectively, our robust POWV reverse genetics system permits analysis of viral spread and neurovirulence determinants in vitro and in vivo and enables the rational genetic design of live attenuated POWV vaccines. IMPORTANCE Our findings newly establish a mechanism for genetically modifying Powassan viruses (POWVs), systematically defining pathogenic determinants and rationally designing live attenuated POWV vaccines. This initial study demonstrates that mutating POWV NS1 glycosylation sites attenuates POWV spread and neurovirulence in vitro and in vivo. Our findings validate a robust circular polymerase extension reaction approach as a mechanism for developing, and evaluating, attenuated genetically modified POWVs. We further designed an infectious GFP-tagged reporter POWV that permits us to monitor secretory trafficking of POWV in live cells, which can be applied to screen potential POWV replication inhibitors. This robust system for modifying POWVs provides the ability to define attenuating POWV mutations and create genetically attenuated recPOWV vaccines.

Reverse genetics systems are needed to genetically modify POWVs, define determi nants of POWV pathogenesis, and create live attenuated recombinant POWV vaccines. Typical reverse genetics approaches are based on cloning viral RNA genomes into bacterial plasmids or artificial chromosome vectors with the toxicity and stability issues of large viral inserts that magnify the complexity of genetic virus modification (51)(52)(53)(54)(55)(56)(57). The circular polymerase extension reaction (CPER) was originally devised to clone complex genes (58,59) and subsequently used to bypass the plasmid instability of large viral inserts and to rapidly generate mutant and reporter viruses (38,(60)(61)(62)(63)(64)(65). CPER has now been used to investigate virulence determinants in West Nile virus (WNV), Zika virus, yellow fever and dengue virus (DENV), and JEV (38, 60-64, 66, 67) and has the potential to be applied to other positive-stranded RNA viruses (62,68).
Here, we used CPER to develop a robust POWV reverse genetics system and generate an infectious LI9 POWV from cDNA (recLI9) that mirrors wild-type (WT) LI9 in sequence, replication, and focal cell-to-cell spread in VeroE6 cells. CPER was used to generate recLI9 mutant viruses lacking individual NS1 glycosylation sites (N85Q, N208Q, and N224Q). Mutating LI9 NS1 glycosylation sites reduced replication kinetics, viral titers and delayed focal cell-to-cell spread in vitro. The recLI9-NS1-N224Q mutant was severely attenuated with foci formation delayed by 2-3 days in vitro and in vivo resulting in the survival of 80% of peripherally inoculated mice. Our findings define a key role for the NS1-N224 glycosylation site in POWV replication, cell-to-cell spread, and pathogenesis and suggest N224Q as an attenuating mutation to be considered in the development of live attenuated POWV vaccines. In order to monitor NS1 functions, we used CPER to add a split-GFP tag to the NS1 C-terminus and generate an infectious, GFP expressing, recLI9 reporter virus (69)(70)(71)(72)(73). Infecting cells expressing ER-GFP 1-10 cells with the POWV GFP11 reporter virus (recLI9-NS1-GFP11) reconstituted GFP fluorescence that permits live cell analysis of NS1 spread during infection and a fluorescence-based assay for screening potential inhibitors of POWV replication and maturation (69,70,72,73). The novel accumulation of NS1-GFP in large intracellular vesicles suggests a potential mechanism for POWV NS1 to regulate viral release and cell-to-cell spread (69,74). Collectively, our robust POWV reverse genetics system permits us to define determinants of POWV pathogenesis, evaluate attenuating POWV mutations, and create genetically modified recPOWVs for use as live attenuated vaccines.

POWV spreads cell-to-cell in the presence of neutralizing antibody
The direct isolation of POWV LI9 in VeroE6 cells revealed a novel focal spread phenotype that occurred in the absence of restrictive overlays and without lytic plaque formation (20). POWVs bud into the lumen of the ER and are trafficked via secretory vesicles to the cell surface where they remain largely cell associated (6,7). Foci formation suggested the potential for POWV to spread cell-to-cell similar to hepatitis C virus (HCV), a distant FV family member, that spreads focally cell-to-cell in a neutralizing antibody-independent manner (75)(76)(77). To investigate whether LI9 foci are formed by cell-to-cell spread, we evaluated foci formation in the presence of POWV neutralizing antibody (ATCC-HMAF) or control antibody added daily to infected cell supernatants. Anti-POWV antibody added to cells during viral adsorption prevented infection, while the post-adsorption addition of control or anti-POWV antibody at a 1:250 dilution (neutralizing titer 1:6,400) failed to inhibit the formation of POWV-infected cell foci (Fig. 1A). This establishes that POWV LI9 initially spreads cell-to-cell and explains the focal spread of LI9 during isolation from ticks and during passage in epithelial cells in liquid culture. These are novel findings for tick-borne FVs that may permit POWV to bypass immune surveillance (78) and provide potential mechanisms for POWVs to spread at tick bite sites or into the CNS.

CPER reverse genetics: a mechanism for defining determinants of POWV pathogenesis
Currently, there are no established mechanisms for attenuating POWVs or defining determinants of POWV neurovirulence. To generate recombinant POWVs, we used a CPER approach (38,(60)(61)(62)(63)(64)(65) to develop a robust POWV reverse genetics system (Fig. 1B through D). POWV LI9 RNA was initially used to generate first-strand cDNA that was amplified into five fragments comprising the full-length POWV genome and a sixth circularizing UTR-linker fragment (Fig. 1B). Amplified fragments contain 26 nucleotide overlapping sequences, with the UTR-linker fragment containing 26 nt from the LI9 3′UTR, a HDVr, SV40 polyadenylation signal, CMV promoter, and the first 33 nt of the 5′UTR sequence (Fig. 1C). To generate circularized CPER DNAs, LI9 F1-F5 cDNA fragments and the UTR-linker DNA were amplified using high-fidelity Phusion DNA polymerase (Fig.  1D). CPER-amplified DNA was purified and directly transfected into HEK293T or VeroE6 cells. Three days post-transfection, supernatants from transfected cells were inoculated into VeroE6 cells, and infectious viral rescue was evaluated by immunoperoxidase staining of LI9 infected cells ( Fig. 1E and F). Cells infected by recLI9 retained a focal cellto-cell spread phenotype, replication kinetics, and titers, with complete recLI9 sequences identical to WT LI9 (Fig. 1G). This demonstrates the authenticity of the CPER-generated recLI9 POWV and the development of a robust reverse genetic system for generating POWVs from recombinant DNA.
Here, we used POWV reverse genetics to modify individual POWV NxS/T sites (N85Q, N208Q, or N224Q) to determine whether NS1 glycosylation affects viral replication and focal cell-to-cell spread. Using CPER, we split F2, containing NS1, into F2A and F2B using overlapping primers that contain glycosylation site-specific N->Q mutations (Fig. 2C). Amplified F2A and F2B fragments were substituted for F2 in CPERs and used to generate recLI9-NS1 NxT mutants, as in Fig. 1. Sequencing of rescued recLI9-NS1 mutants verified site-specific N->Q residue changes within NS1 proteins of each individual POWV (N85Q, N208Q, and N224Q) (Fig. S1). To investigate roles for NS1 NxT sites in NS1 expression and secretion, we infected VeroE6 cells at an MOI of 10 with the LI9-NS1-N85Q, LI9-NS1-N208Q, or LI9-NS1-N224Q mutant viruses. Intracellular and secreted NS1 proteins were evaluated 6 dpi by immunoblotting using an anti-NS1 mouse monoclonal antibody ( Fig.  2D and E). Compared to WT LI9, mutating N208Q and N224Q glycosylation sites increased migration of NS1 proteins by Western Blot, while migration of NS1 from the N85Q mutant was dramatically increased from both intracellular and secreted NS1 sources (Fig. 2D). Evaluation of the relative amounts of intracellular and secreted NS1 proteins revealed a 200%-300% increase in the secretion of N85Q and N208Q NS1 proteins and a ~50% decrease in the secretion of the N224Q NS1 protein compared to NS1 from WT LI9 (Fig. 2E). Relative to WT LI9 infection, titers of recLI9 NS1 glycosylation mutants revealed slower viral replication kinetics with maximal titers reduced by 1-3 logs from 2 to 6 dpi ( Fig. 2F). While each NS1 glycosylation mutant spread focally, the formation of foci in monolayers was delayed by 1 day in N85Q and N208Q NS1 mutants (Fig. 2G). In contrast, foci formation was reduced by 2-3 days in the LI9-NS1-N224Q mutant, which failed to spread uniformly throughout monolayers by 6 dpi (Fig. 2G). These findings indicate that mutating any single LI9 NS1 glycosylation site reduced viral replication and spread, while mutating the NS1-N224Q glycosylation site resulted in a highly attenuated recLI9 mutant that was severely restricted in replication and spread in vitro.

Research Article mBio
Attempts to CPER generate recPOWV LI9 mutants with two glycosylation site changes (N85Q/N208Q, N85Q/N224Q, or N208Q/N224Q) failed to result in recovery of any double mutants. It remains unclear whether additional glycosylation site mutations critically impact NS1 functions required for POWV replication and virion assembly, and whether double mutants could be rescued by transcomplementation in NS1 protein-expressing cells.

POWV NS1 mutant dimerization, glycosylation, and secretion
Changing NS1 protein glycosylation may impact dimer formation, glycan processing, and NS1 protein functions. To determine whether dimer formation is altered in POWV NS1 glycosylation mutants, we compared untreated or heat-treated (2 min at 100°C) NS1 dimer stability prior to SDS-PAGE analysis. We found that all mutants formed dimers in unheated samples and that both dimers and monomers of glycosylation mutants were formed but with increased migration on SDS gels vs WT LI9 NS1 (Fig. 3A). These findings are consistent with unaltered dimer formation by DENV NS1 glycosylation site mutants (25).
We evaluated changes in high mannose or complex glycan addition to N-linked glycans of LI9 NS1 glycosylation mutants using endoglycosidases. Cell-associated and secreted NS1 proteins from WT and NxT mutants were treated with Endo H to remove high-mannose N-linked glycans or PNGase F to remove a combination of high mannose, hybrid, and complex oligosaccharides. Digestion profiles demonstrate that intracellular NS1 from WT LI9 or recLI9 glycosylation mutants possess mainly high-mannose N-linked glycans as digestion with either Endo H or PNGase F resulted in similar NS1 protein migration (Fig. 3B). However, secreted NS1 proteins of WT, N208Q, and N224Q are only partially reduced in size following Endo H digestion compared to PNGase (Fig. 3B), suggesting that N208Q and N224Q mutant NS1 glycans are further modified in the secretory pathway containing complex or hybrid sugars (Fig. 3B). In contrast, the secreted NS1 N85Q protein migrates similarly after Endo H or PNGase F digestion indicating that glycans on NS1 N85Q mutant proteins are not further processed in the Golgi (Fig. 3B). The large size difference of N85Q mutants, compared to NS1 proteins from LI9 WT or N208Q, and N224Q mutants suggests that the N85 glycosylation site deter mines whether N208 and N224 glycans are further processed during NS1 secretion.

Analysis of attenuated recLI9-NS1-N224Q in mice
POWV LI9 is lethal when intracranially or subcutaneously inoculated into C57BL/6 mice (20). Here, we assessed the lethality of in vitro attenuated recLI9-NS1-N224Q in murine models and found that like WT LI9, recLI9-NS1-N224Q is 100% lethal when i.c. inoculated into suckling mice (N = 5) 5-7 dpi. Although both WT LI9 and recLI9-N224Q are lethal when i.c. inoculated, there was a 1-day difference in the timing of lethality between WT and mutant ( Fig. 4A and B). As i.c. inoculation of suckling mice reflects neurovirulence, but not neuroinvasion, we s.c. inoculated C57BL/6 mice with LI9 or recLI9-NS1-N224Q (N = 8-10) and evaluated weight loss and lethality from 1 to 20 dpi. WT LI9 caused weight loss and was lethal in 40% of C57BL/6 mice 9-12 dpi. In contrast, in mice, s.c. inoculated with recLI9-NS1-N224Q lethal disease onset was delayed, with 80% of mice surviving infection ( Fig. 4B and C). Similar to attenuating replication and spread kinetics in vitro, our findings suggest that mutating the N224Q glycosylation site in infectious LI9 is attenuating in vivo and partially reduces lethal POWV outcome.

Research Article mBio
As a means of analyzing POWV NS1 protein functions, we used CPER to generate NS1 Cterminally tagged with a split GFP11 fusion protein in recLI9 (recLI9-GFP11) (Fig. 5B). The recLI9-GFP11 reporter virus forms infected cell foci by immunostaining and replicates with similar kinetics and titers 1-3 dpi as WT LI9 in VeroE6 cells ( Fig. 5C and D). HEK293T and VeroE6 cells were transduced with retroviruses expressing cytoplasmic GFP 1-10 and mCherry (mCh-GFP 1-10 ) or an ER-translocated GFP 1-10 -KDEL and cytoplasmic mCherry (mCh-ER-GFP 1-10 ). We infected mCh-GFP 1-10 or mCh-ER-GFP 1-10 -expressing HEK293T cells with the recLI9-NS1-GFP11 virus and found GFP-positive infected cell foci in ER-GFP 1-10 but not in GFP 1-10 -expressing cells 2-3 dpi ( Fig. 5E; Fig. S3). Similarly, in VeroE6 cells expressing mCh-ER-GFP 1-10 , infection with recLI9-NS1-GFP11 reconstituted NS1-GFP fluorescence with a perinuclear localization consistent with NS1-GFP11 translocation to the ER (Fig. 5F). In addition, we found abundant expression of NS1-GFP11 in large fluorescent vesicles (6 dpi) that may reflect NS1 protein accumulation in intracellular secretory or specialized vesicles. Intracellular NS1 accumulation provides a rationale for reduced NS1 secretion during POWV infection and suggests that novel vesicular trafficking may foster cell-to-cell spread in POWV-infected cells. Collectively, using CPER reverse genetics, we have generated an infectious recLI9-GFP11 reporter virus that provides a mechanism for analyzing the role of NS1 in restricting POWV secretion or directing cell-to-cell spread and provides an infectious fluorescent reporter virus that can be used to screen POWV inhibitors.

DISCUSSION
We isolated the LI9 strain of POWV directly from ticks in VeroE6 cells in order to avoid potential neuroadaptive changes of prototype POWV strains (LB and SP) that were isolated and adapted to growth in murine brains (5,6,14,20). The LI9 POWV isolate nonlytically infects VeroE6 cells, unexpectedly spreading focally in cells lacking restrictive Research Article mBio overlays (20). LI9 spread focally in media containing high levels of anti-POWV-neutraliz ing antibodies, revealing that LI9 spreads cell-to-cell without requiring apical release into cell supernatants (Fig. 1). This both confirms the large cell-associated nature of POWVs (6), while uniquely mirroring data establishing that HCV, a distantly related FV, spreads cell-to-cell in hepatocytes (75)(76)(77)(78). The neutralizing antibody-independent, cell-to-cell spread of HCV is proposed to occur at partially sealed cell junctions that permit HCV to evade immune surveillance (75)(76)(77)(78). Measles virus (MV) was also shown to spread cell-tocell in epithelial cells, without syncytia formation, using a GFP reporter virus (87). The MV-GFP reporter was visualized flowing to the cytoplasm of adjacent cells through lateral pores and was suggested as a mechanism for MV transmission from epithelial cells to neurons (78,87). Implying that POWV cell-to-cell spread is epithelial cell specific, we found that POWV fails to spread cell-to-cell in hBMECs or pericytes. At later times postinfection, POWV titers increase and spreads conventionally by the diffusion of cell-free viral particles (78). Cell-to-cell spread is a known viral mechanism for immune evasion Research Article mBio (78) that may contribute to dissemination in the skin or foster CNS transmission via epithelial cell barriers that comprise the choroid plexus (88)(89)(90). The mechanism of POWV cell-to-cell spread remains to be defined, and our recLI9-GFP reporter virus may permit analysis of POWV trafficking and spread to recipient cells. In order to define mechanisms for attenuating POWVs, we developed a CPER reverse genetics system that authentically generates infectious recombinant POWVs. The CPER approach bypasses stability issues of bacterial FV clones and creates recombinant POWVs from amplified POWV genomic cDNAs and clones (38,(60)(61)(62)(63)67). The recLI9 POWV was CPER generated by circularizing five to seven fragments of LI9 cDNAs with defined CMV promoter and HDVr-directed 5′ and 3′ termini. CPER-assembled LI9 DNA transfected into HEK or VeroE6 cells results in infectious recLI9 progeny that differs only by cell transfection efficiency. The authenticity of recLI9 viruses with WT LI9 was verified by sequencing, replication kinetics, and phenotypically by focal cell-to-cell spread ( Fig. 1  and 2).
Our newly developed POWV CPER reverse genetics system permits us to define determinants of POWV spread, neuroinvasion, and neurovirulence, providing a mechanism for genetically attenuating POWVs and rationally designing live attenuated POWV vaccines. Reverse genetics has been accomplished for several FVs, initially using RNA transcription or plasmid approaches and more recently by employing CPER systems (38,(60)(61)(62)(63)67). Reverse genetics has also been used to generate chimeric FVs with prM-Env-specific neutralization determinants and NS protein backgrounds derived from attenuated FVs or the Yellow Fever Virus (YFV) 17D vaccine (23,46,52,68). Insertion of the POWV prM-Env in a YFV-17D background retained lethal neurovirulence with only 43% of inoculated mice surviving this virulent chimera (51).
POWV contains two novel NS1 glycosylation sites, not found in mosquito-borne FVs, and adds complex glycans to unique sites on the periphery of putative NS1 hexamers (Fig. S2). To define POWV NS1 glycosylation functions, we modified individual NS1 glycosylation sites (N85Q, N208Q, and N224Q) and isolated infectious recLI9 NS1 mutant POWVs (Table S1). Each glycosylation site mutant spread focally cell-to-cell in VeroE6 cells like parental LI9 but with a 1-2 log decrease in NS1 mutant titers and reduced replication kinetics. A pronounced 2-to 3-day reduction in the timing of foci formation by the recPOWV-NS1-N224Q mutant coincides with reduced replication and reveals a severely attenuated in vitro phenotype. Although i.c. inoculation of mice with the NS1-N224Q glycosylation mutant was lethal, peripheral inoculation of the recLI9-NS1-N224 mutant resulted in 80% survival, elicited antibody responses to POWV Env proteins, and a partially attenuated phenotype relative to WT LI9 infection. Further analysis is needed to determine if the NS1-N224Q glycosylation mutant is attenuated at peripheral spread or neuroinvasive levels and how the unique structural position of the POWV NS1 N224 glycosylation site restricts replication and delays POWV cell-to-cell spread. Irrespective of the mechanism, our findings rationalize including N224Q mutations in strategies for developing a live attenuated POWV vaccine.
Mutating WNV NS1 glycosylation sites (N130, N207) produced stable attenuated strains with NS1 accumulation in the ER and restricted NS1 secretion (48,95). In DENV NS1, N207 mutants showed reduced dimer stability and secretion, with complex glycans attached to N130 and high mannose glycans added to N207 (23,25). The NS1 N85 glycosylation site was reported to be broadly required for high mannose or complex glycan processing, suggesting that N85Q mutations broadly restrict Golgi processing of N208/N224 glycans and alter NS1 secretion (30). Although all single POWV LI9 NS1 glycosylation mutants formed dimers and were secreted from cells, NS1 secretion was reduced in N224Q mutants and increased in NS1 N85Q and N208Q mutants relative to WT LI9. It remains to be determined how the position of individual glycosylation sites ( Fig. 2B; Fig. S2) impacts higher order NS1 glycan processing or hexamer formation or constrains NS1 secretion and POWV release from cells.
Why some NS1s are highly secreted, while others remain highly cell associated, and how NS1 proteins regulate the assembly and secretion of FVs remains an enigma linked to viral spread and pathogenic mechanisms (2,23,32,41,96). POWV virions (6) and NS1 proteins remain highly cell associated with low levels of secreted NS1, and this novel NS1 attribute may have a role in directing the unique cell-to-cell spread phenotype of POWVs. Many functions ascribed to NS1 do not reflect that NS1 proteins are cotranslationally translocated into ER compartments and unavailable for cytoplasmic interactions (2,23,32). Further, NS1 and Env proteins that are both present in secretory vesicles, yet discretely released from cells, suggest intertwined roles for NS1 and Env glycosylation in virion assembly or secretion that have yet to be discovered (27,38,41,96).
As an initial approach to study POWV NS1 translocation and secretion, we used CPER to generate a recLI9-NS1-GFP11 reporter virus for analysis of NS1 localization in cells. Fluorescence of the recPOWV NS1 split GFP11 protein was reconstituted in the ER of cells expressing ER-GFP 1-10 protein. This demonstrates the ability of recLI9-NS1-GP11 to serve as a fluorescent reporter virus and permits real-time analysis of POWV infection and NS1 functions in the lumen of the ER (69,71,97). Initial studies demonstrate the ER localization of NS1-GFP fluorescence and the novel accumulation of NS1-GFP in large intracellular vesicles in infected cells 6 dpi. These findings reveal NS1 lined vesicles and suggest a mechanism for NS1 retention that may restrict assembled POWV particle release (6) or contribute to a vesicular mechanism of POWV cell-to-cell spread. The nature of NS1 fluorescing vesicles (secretory, lipid droplet, autophagosome, convoluted membranes) remain to be defined (24,25,32,41,47,48,95,96) but may explain how NS1 proteins are excluded from virions within secretory vesicles, yet determine FV spread in mammalian cells (38,98,99). Combining NS1 glycosylation mutants in a LI9-NS1-GFP reporter background may reveal roles for NS1 glycosylation sites in POWV secretion that impact attenuation. Further, the production of an infectious POWV NS1-GFP reporter virus provides a platform for high-throughput screening of POWV replication inhibitors (64,66,86,97).
Our findings provide a robust mechanism for generating recombinant POWV mutants and defining determinants of cell-to-cell spread, neuroinvasion, and neurovirulence in vitro and in vivo. We generated a viable infectious POWV reporter virus that permits analysis of NS1 functions required for viral replication, maturation, and neurovirulence and revealed an NS1-N224Q glycosylation mutation that attenuates POWV replication in vitro and partially attenuates POWV in mice. Studies presented provide a mechanism for assessing attenuating POWV mutations and genetically designing live attenuated POWV vaccines.

Viral RNA extraction and cDNA synthesis
RNA from POWV-infected VeroE6 cells was RLT extracted and RNeasy column (Qia gen) purified. cDNA synthesis was performed using the Transcriptor first-strand cDNA synthesis kit (Roche) following the manufacturer's protocol and a primer complimentary to the 3′UTR (Table S2, oligo F5R) (20,79,80).

PCR amplification of DNA fragments and CPERs
Five POWV-LI9 fragments were amplified from viral cDNA (Fig. 1A and B) using high-fidelity Phusion polymerase (NEB) and corresponding paired primers ( Table S2) that have a complementary 26-nucleotide overlap. An additional PCR fragment (UTR linker) was generated from plasmid pMiniT containing the CMV promoter, the first and the last 26 nts of the POWV-LI9 sequence, an HDVr, and SV40 polyadenylation site. POWV cDNA was amplified into six individual POWV DNA fragments: initial denaturation 98°C for 30 s; 32 cycles of 98°C for 20 s, 60°C for 30 s, and 72°C with 30 s per kb; and a final extension at 72°C for 5 min. Resulting fragments were gel purified, Monarch kit (NEB) extracted, and 0.1 pmol of each DNA fragment was CPER amplified in a 50-µL reaction containing 200 µM of dNTPs, 1× Phusion polymerase GC reaction buffer, and 1 µL Phusion polymerase. The following cycling conditions were used: initial denaturation 98°C for 30 s; 12 cycles of 98°C for 20 s, 60°C for 30 s, and 72°C for 10 min; and a final extension at 72°C for 10 min. CPERs lacking the UTR-Linker fragment were generated and used as negative controls.

LI9 POWV NS1 molecular modeling
The 3D model of dimeric LI9 NS1 protein was obtained using multimer prediction in AlphaFold2 using ColabFold within ChimeraX 1.5 software (100). Five models were created with the top-scoring model selected after relaxation. The POWV LI9 NS1 hexamer model was generated using cryoEM structures from DENV2 soluble NS1 hexamers (PDB 7WUV) scaffolding with the model relaxed using Rosetta Relax (101, 102).

In vivo LI9 POWV infection experiments
Anesthetized murine C57BL/6 pups (<2-week old, N = 3-5) were intracranially infected with 2 × 10 2 FFU POWV or buffer-only control in a volume of 2 L. Infected mice were monitored twice daily for signs of disease and daily for weight loss. Peripheral inocu lation: C57BL/6 mice (male, 41-to 51-week-old, N = 4-10, Jackson Laboratory) were anesthetized via intraperitoneal injection with 100 mg mL −1 (1) of ketamine and 20 mg mL −1 (1) of xylazine per kilogram of body weight. Animals were infected via subcutane ous footpad injection with 2 × 10 3 FFU POWV or buffer-only control in a volume of 20 L. Weights and neurovirulent sequelae were monitored daily as indicated.

Biosafety and biosecurity
Animal research was performed in accordance with institutional guidelines following experimental protocol review, approval, and supervision by the Institutional Biosafety Committee and the Institutional Animal Care and Use Committee at Stony Brook University. Animals were managed by the Division of Laboratory Animal Resources (Stony Brook University), which is accredited by the American Association for Accredi tation of Laboratory Animal Care and the Department of Health and Human Services. Animals were maintained in accordance with the applicable portions of the Animal Welfare Act and the DHHS "Guide for the Care and Use of Laboratory Animals. " Veterinary care was under the direction of full-time resident veterinarians boarded by the Ameri can College of Laboratory Animal Medicine. Experiments with infectious POWV were performed in an ABSL3 facility at SBU.

DIRECT CONTRIBUTION
This article is a direct contribution from Jorge Benach, a Fellow of the American Academy of Microbiology, who arranged for and secured reviews by Christopher Basler, Icahn School of Medicine at Mount Sinai, and Harry Greenberg, Stanford University.