Herpes Simplex Virus 2 Counteracts Neurite Outgrowth Repulsion during Infection in a Nerve Growth Factor-Dependent Manner

Herpes simplex virus 2 (HSV-2) is a prevalent human pathogen that establishes lifelong latency in neurons of the peripheral nervous system. Colonization of neurons is required for HSV-2 persistence and pathogenesis. The viral and cellular factors required for efficient infection of neurons are not fully understood. We show here that nonneuronal cells repel neurite outgrowth of sensory neurons, while HSV-2 infection overcomes this inhibition and, rather, stimulates neurite outgrowth. HSV-2 glycoprotein G and nerve growth factor contribute to this phenotype, which may attract neurites to sites of infection and facilitate virus spread to neurons. Understanding the mechanisms that modulate neurite outgrowth and facilitate HSV-2 infection of neurons might foster the development of therapeutics to reduce HSV-2 colonization of the nervous system and provide insights on neurite outgrowth and regeneration.

H erpes simplex virus 1 (HSV-1) and HSV-2 establish lifelong latency in sensory factors secreted by two cell lines, HEK-293T and ARPE-19, into the culture medium. We incubated primary neurons obtained from mouse DRG with NGF plus conditioned medium from HEK-293T or ARPE-19 cells collected at 72 h postseeding. As a control, we used nonconditioned cell culture medium with or without NGF. After 24 h of incubation, we labeled neurites with antibodies against the neuronal marker tubulin ␤-III (Fig.  1A). We counted the neurites and neurons in regions of interest (ROI). Figure 1B shows the number of neurites per neuron in an equal number of ROI per condition after log 2 transformation. Neurons cultured for 24 h in nonconditioned medium without NGF had Representative images of primary DRG-derived mouse neurons exposed to different media with or without NGF. Neurons were labeled with an anti-tubulin ␤-III antibody (Tuj1; green), and the DNA was labeled with DAPI (blue). Bars, 50 m. (B) Graph showing the number of neurites per neuron (log 2 transformed) following incubation of cells with conditioned medium from HEK-293T or ARPE-19 cells or nonconditioned medium with or without NGF. The results are from one representative experiment out of two independent ones. The means for the population and the standard deviation (error bars) are depicted. P values were calculated using ANOVA with Browne-Forsythe and Welch corrections with Dunnett's T3 multiple-comparison posttest for indicated groups. about 5 neurites (mean, 5.03 neurites; limits of the 95% confidence interval [CI], 4.07 to 6.22 neurites) (Fig. 1B). NGF induced neurite outgrowth (mean, 10.18 neurites; 95% CI, 7.56 to 13.72 neurites). Conditioned supernatants from ARPE-19 cells (mean, 4.06 neurites; 95% CI, 3.42 to 4.82 neurites) or HEK-293T cells (mean, 2.99 neurites; 95% CI, 2.28 to 3.90 neurites) abolished the stimulating effect of NGF on neurite outgrowth. These results indicate that NGF induces neurite outgrowth of mouse DRG neurons and that factors secreted by both ARPE-19 and HEK-293T cells prevent this induction, results comparable to previously obtained results with sympathetic neurons (15).
Generation of a recombinant reporter HSV-2. To monitor HSV-2 infection, we generated HSV-2 reporter viruses using the MS strain containing the bacterial artificial chromosome (BAC) [HSV2(MS)BAC], provided by Meseda and colleagues (33). A BAC cassette in HSV2(MS)BAC had been inserted into the UL23 gene, to disrupt thymidine kinase (TK) expression ( Fig. 2) (33). We repaired the UL23 gene and introduced a self-excisable Cre recombinase-expressing cassette into pHSV2(MS)BAC by en passant mutagenesis (34) to generate HSV2(MS)Lox ( Fig. 2A to H). Enzymatic restriction analyses and Southern blot analysis confirmed the complete excision of the BAC cassette after HSV-2 reconstitution in eukaryotic cells ( Fig. 2I and J). We confirmed the repair of the UL23 gene and the expression of functional TK with an acyclovir (ACV)-based plaque reduction assay in U2OS cells (Fig. 2K) and determined a half-maximal inhibitory concentration of ACV of 2.9 g/ml (12.9 M), within the expected range (35). ACV at 24 g/ml completely arrested the replication of HSV2(MS)Lox in U2OS cells, whereas that of HSV2(MS)BAC, lacking a functional TK, was unaffected. Thus, in contrast to HSV2(MS)BAC, HSV2(MS)Lox expressed a functional TK.
We used HSV2(MS)Lox to generate the reporter HSV2(MS)Lox-pHCMV mCheGLuc (named MS-CheGL for short) expressing a bicistronic mRNA under the control of the major IE promoter (MIEP) of human cytomegalovirus (HCMV) and encoding mCherry and Gaussia luciferase (GL) separated by a picornavirus 2A (P2A) site (36) (Fig. 3A). We introduced the reporter cassette in the intergenic locus between the open reading frames (ORFs) of UL55 and UL56, a locus that does not affect HSV-1 replication (37). We sequenced the virus obtained upon reconstitution in mammalian cells (the sequences have been uploaded into GenBank and may be found under accession no. MH796783). The replication of the MS-CheGL virus was similar to that of its parental HSV2(MS)Lox virus in HEK-293T cells (Fig. 3B), indicating that the insertion of the reporter cassette did not affect HSV-2 replication, as reported before for HSV-1 (37). We also detected secreted Gaussia luciferase (GLuc) in the supernatant. The level of GLuc activity was similar to the overall kinetics of virus replication (Fig. 3B).
HSV-2 infection reduces the repelling effect of HEK-293T cells. Based on the results shown in Fig. 1, we decided to address the impact of HSV-2 infection on the repelling effect of HEK-293T cells. We chose HEK-293T cells over ARPE-19 cells because they are easier to transfect and express repellents of neurite outgrowth, such as semaphorin 3A and class A ephrin (38,39). We seeded mouse DRG neurons in microfluidic chambers (MFC) to separate the cell bodies and neurites (Fig. 4A). These devices have two compartments connected by microgrooves that allow the passage of neurites but not neuronal cell bodies and have been widely used to study alphaherpesvirus neurobiology (32,(40)(41)(42). We placed dissociated DRG neurons from mice aged between 33 and 38 days in the somal compartment (SC) and seeded mock-or HSV-2infected HEK-293T cells into the neurite compartment (NC). NGF was added to both the NC and the SC. We seeded a mix of 20% infected cells (mCherry positive) and 80% noninfected cells into the NC to mimic the situation at the initial phase of infection, when most of the cells are not infected. We seeded the same number of dissociated DRG cells and HEK-293T cells under all conditions in every experiment (5 ϫ 10 4 and 2 ϫ 10 5 , respectively). As a control, we cultured dissociated neurons without HEK-293T cells. The HEK-293T cells were infected for 24 h and added to the NC 1 day after neuronal seeding. At this time, the neurites had already entered the NC from the SC through the microgrooves. We added the HEK-293T cells with aphidicolin to inhibit any further mitosis, as this would lead to high cell confluence, precluding the observation of neurites. Aphidicolin also inhibits HSV replication (43)(44)(45). Aphidicolin reduced HSV-2 titers by about 1 log unit after 24 h in HEK-293T cultures compared to that for the controls (data not shown).
After 3 days of coculturing HEK-293T cells and neurons, we detected the neurons with an antibody to tubulin ␤-III, the nuclei with DAPI (4=,6-diamidino-2-phenylindole), and infected cells by HSV-2-mediated mCherry expression. There were many long neurites in the absence of HEK-293T cells (Fig. 4B, left), whereas the presence of mock-treated HEK-293T cells reduced the number and length of neurites (Fig. 4B, middle). This suggests that the HEK-293T cells expressed repellents of neurite outgrowth, as shown before ( Fig. 1) (38,39,46). Infection of HEK-293T cells with HSV-2 (MS-CheGL) partially overcame the repulsion, with more and longer neurites growing toward the HEK-293T cells (Fig. 4B; compare the middle and right panels). To measure neurite outgrowth, we traced their length with the Simple Neurite Tracer tool in ImageJ software (47). We measured the neurite length and number from the edge of the microgrooves into the NC (Fig. 5 A and B). Quantification of the neurite length showed shorter neurites in the presence of HEK-293T cells (median, 90.88 m; limits of the interquartile range [IQR], 62.5 to 144.4 m) than in neurons alone (median, 811 m; limits of IQR, 487.9 to 1.202 m) (Fig. 5A). This negative effect on neurite length was reduced when the NC contained infected HEK-293T cells (median, 314.7 m; limits of IQR, 154.3 to 647.5 m). The same trend was observed when comparing the total number of neurites (Fig. 5A, top line) and the number of neurites per image (Fig. 5B). There were 442 neurites in total and 77.98 neurites/image (mean; 95% CI, 42.72 to 142.32 neurites/image) in the NC without HEK-293T cells, while the presence of HEK-293T cells resulted in 177 neurites and 28.74 neurites/image (mean; 95% CI, 22.19 to 37.25 neurites/image). The presence of infected cells in the NC increased the total number of neurites to 327 and the number of neurites per image to 53.01 (mean; 95% CI, 40.76 to 68.93 neurites/image). Infection of HEK-293T cells with HSV-2 reduced their repulsion over that for mock-infected cells in four independent experiments with primary neurons, resulting in longer neurites (mean length, 2.23-fold increase; 95% CI, 0.97-to 5.14-fold increase) and more neurites (mean number, 1.3-fold increase; 95% CI, 1.1-to 1.44-fold increase) ( Fig. 5C and D). The increase in neurite number was more robust than the increase in neurite length. These results show that the presence of NGF We seeded dissociated mouse DRG neurons in the somal compartment (SC). After 24 h, we added HSV-2-infected (MOI, 0.1) or mock-treated HEK-293T cells to the neurite compartment (NC) in the presence of aphidicolin. Seventy-two hours later the cells were fixed and stained. We acquired images of the channel on the NC side (indicated by the boxed area) and quantified neurite outgrowth. (B) Representative images showing neurite outgrowth from the SCs through the microgrooves to the NCs. The left side of each image shows the neurites and HEK-293T cells in the NCs, whereas the right part contains the neurons seeded in the SC. Neurons were labeled with an anti-tubulin ␤-III (Tuj1 antibody; green), DNA was detected with DAPI (blue), and infected cells were identified by mCherry expression (magenta). The number and length of neurites entering the NCs from the microgrooves were quantified with the Simple Neurite Tracer plug-in of ImageJ software. without HSV-2 infection did not revert the inhibitory phenotype and that HSV-2 infection of HEK-293T cells partially inhibited their repulsion on neurite outgrowth. gG2 expression contributes to inhibit the repulsion of HEK-293T cells during infection. Repellents secreted from HEK-293T cells into the conditioned medium, exogenous NGF, and HSV-2 infection modulated neurite outgrowth in our model. Purified rSgG2 increases NGF-induced neurite outgrowth (15). Whether the gG2 expressed during infection has a similar function was not known. To address the role of gG2 in neurite outgrowth during infection, we generated an HSV-2 mutant lacking gG2 expression [HSV2(MS)Lox-Che-GLuc-ΔgG2, or MS-CheGL-ΔgG2 for short]. We mutated the US4 start codon and introduced a stop codon and a frameshift to prevent gG2 expression (Fig. 6A). This strategy reduces the possibility of gG2 expression, even if there were a restoration of the ATG codon. We sequenced MS-CheGL-ΔgG2 upon reconstitution in mammalian cells (GenBank accession no. MH796784). There were no undesired mutations in the recombinant virus. We could not detect gG2 in lysates or supernatants from MS-CheGL-ΔgG2infected cells, while the levels of the major capsid protein (MCP) were unaffected (Fig.  6B). We also addressed whether translation could start at methionine 286, producing a shorter form of gG2 of 418 amino acids. This would correspond to a protein with a molecular weight of approximately 45 kDa. We did not detect any part or precursor form of the gG2 protein using an antibody targeting residues 290 to 321 or the gG2 ectodomain, which would have been included in an N-terminally truncated gG2 (Fig. 6C). The lack of gG2 expression did not affect replication in MeWo and HEK-293T cells (Fig. 6D).
Next, we characterized the impact of HEK-293T cells infected with parental or gG2-deficient HSV-2 on neurite outgrowth (Fig. 7). As before (Fig. 5), mock-infected HEK-293T cells inhibited neurite outgrowth, while infection with the parental virus partially reverted the phenotype, inducing the formation of longer and more neurites ( Fig. 7A  . This suggested that the observed effects were not simply caused by HSV-2-infection but were at least partially due to gG2 expression. To test whether gG2 contributed to counteract the repressive effect on neurite outgrowth, we complemented MS-CheGL-ΔgG2 in trans with gG2-expressing constructs. We transfected HEK-293T cells with an empty control or plasmids expressing SgG2 (pSgG2) or fulllength gG2 (FLgG2) (pFLgG2) prior to infection with the MS-CheGL-ΔgG2 virus (Fig. 7C Transfection with pFLgG2 complemented MS-CheGL-ΔgG2 more efficiently than transfection with pSgG2. The expression of gG2 and MCP was assessed by immuno-   7D). Transfection of pFLgG2 resulted in more SgG2 expression than transfection of pSgG2, despite equal molarities of the transfected plasmids ( Fig. 7D) (48). This difference in protein levels may have been responsible for the higher neurite outgrowth observed when pFLgG2 was transfected. However, we cannot exclude the possibility that other domains of gG2, such as the ectodomain, the transmembrane region, or the cytoplasmic tail, also contribute to neurite outgrowth. We compared the results from four independent experiments after normalizing the average length or the average number of neurites per image to those for the mockinfected sample for each data set ( Fig. 7E and F). We obtained results similar to those presented in Fig. 7A, with MS-CheGL-ΔgG2 infection leading to shorter neurites (mean length, 1.14-fold increase; 95% CI, 0.75-to 1.70-fold increase) than infection with MS-CheGL (mean length, 1.90-fold increase; 95% CI, 0.91-to 3.99-fold increase) in all four experiments ( Fig. 7E and F). Transfection with pSgG2 partially complemented the MS-CheGL-ΔgG2 phenotype (mean length, 1.98-fold increase; 95% CI, 1.42-to 2.74-fold increase). Complementation was more efficient with pFLgG2 (mean length, 2.28-fold increase; 95% CI, 1.65-to 3.16-fold increase) than with pSgG2. When comparing the number of neurites, we observed the same trend. In all four independent experiments, there were fewer neurites when cells were infected with MS-CheGL-ΔgG2 (mean number, 2.45-fold increase; 95% CI, 0.94-to 6.39-fold increase) than when they were infected with MS-CheGL (mean number, 4.86-fold increase; 95% CI, 2.74-to 8.62-fold increase). Transfection with pSgG2 or pFLgG2 did not significantly increase the neurite number (for pSgG2, the mean was a 2.45-fold increase [95% CI, 1.44-to 9.51-fold increase]; for pFLgG2, the mean was a 5.20-fold increase [95% CI, 1.84-to 14.64-fold increase]). Overall, our results showed that the MS-CheGL-ΔgG2 virus was less efficient than its parental MS-CheGL virus in overcoming the repulsion of HEK-293T cells. Complementation with gG2 in trans partially reverted this phenotype, indicating that HSV-2 gG contributed to reduce the repulsion of neurite outgrowth, but it was not the only factor involved.
Infection with HSV-2 strain 333 increases neurite outgrowth in an NGFdependent manner. To determine whether the HSV-2-enhanced neurite outgrowth was strain specific, we constructed another HSV-2 mutant lacking gG2 expression using CRISPR/Cas9 (Fig. 8). First, we generated a reporter virus, HSV2(333) pHCMV TurboFP635-FRFL (333-TFP-FL, in short), by inserting TurboFP635 (TFP) and red firefly luciferase (FL) under the control of the MIEP promoter of HCMV between the UL26 and UL27 genes, a locus previously shown to permit the insertion of foreign DNA without affecting virus replication (49). We replaced the US4 locus, coding for gG2, with a cassette expressing enhanced green fluorescent protein (EGFP), generating HSV2(333) pHCMV TurboFP635-FRFL-FLuc-ΔgG2, or 333-TFP-FL-ΔgG2 in short (Fig. 8A). These viruses were plaque purified and sequenced (the sequences have been uploaded to the European Nucleotide Archive and may be found under accession no. ERS3367584 [https://www.ebi .ac.uk/ena/browser/view/ERS3367584] and ERS3367585 [https://www.ebi.ac.uk/ena/ browser/view/ERS3367585]). There were no undesired mutations in these viruses. Infection with 333-TFP-FL-ΔgG2 led to a lack of gG2 protein expression. However, there   . This result differed from that obtained with the MS strain, in which the MS-CheGL-ΔgG2 virus induced shorter neurites than the MS-CheGL virus. To determine the functional relevance of NGF, we added an anti-NGF neutralizing antibody. We used the antibody and did not simply remove NGF from the culture because DRG cells also express and release NGF. Treatment of the culture with this NGF antibody prior to and during incubation with infected HEK-293T cells prevented the slight increase in neurite length induced by 333-TFP-FL (mean, 759.55 m; 95% CI, 701.84 to 822.57 m).
Infection of HEK-293T cells with 333-TFP-FL increased the number of neurites in the presence of NGF compared to the number for mock-infected cells ( Fig. 9A and B). This effect was similar to that of the HSV-2 MS strain ( We compared the results from four independent experiments after normalizing the average length and number of neurites per image to those for the mock-infected sample for each data set ( Fig. 9C and D). We did not observe a significant effect on neurite length (Fig. 9C). The effects on the number of neurites were reproducible and followed the same trend that was seen upon infection with HSV-2 MS. In all four experiments, infection with the 333-TFP-FL virus increased the number of neurites, whereas in three out of four experiments, the 333-TFP-FLΔgG2 virus failed to do so (Fig. 9D). We also observed that the neutralization of NGF prevented the increase in neurite number by 333-TFP-FL infection in all experiments. These results indicate that gG2 of both HSV-2 MS and 333 increased the neurite number during infection in an NGF-dependent manner. They also suggest that there might be strain-specific differences in the effect sizes between HSV-2 MS and 333.

DISCUSSION
The first, main objective of this project was to determine whether HSV-2 modulates the neurite outgrowth of sensory neurons during infection of nonneuronal cells. Second, we wanted to characterize the role of NGF and gG2 in this process. Our results show that NGF increased the neurite outgrowth of DRG neurons and that repellents secreted by HEK-293T and ARPE-19 cells inhibited this growth. Infection with the HSV-2 MS or 333 strain counteracted the repelling effect of HEK-293T cells in an NGFdependent manner, having a stronger impact on neurite number than on neurite length. Infection with HSV-2 MS lacking gG2 expression resulted in fewer and shorter neurites than infection with the parental virus. Complementation experiments showed that gG2 contributed to the inhibition of the repulsive effect on neurite outgrowth but that it was not the sole factor involved. The neurites were longer and more numerous in the presence of gG2 and NGF, suggesting that these proteins cooperated in counteracting the HEK-293T cell-mediated neurite repulsion. Interestingly, infection with HSV-2 strain MS increased both neurite length and neurite number, whereas strain 333 increased the number of neurites but not their length. Moreover, gG2 did not contribute to neurite outgrowth during infection with HSV-2 strain 333, pointing to potential strain-specific differences. Addition of NGF-neutralizing antibodies reduced HSV-2-induced neurite outgrowth, demonstrating the contribution of NGF to this process.
Complementation with FLgG2 in trans confirmed that gG2 counteracted the HEK-293T cell-mediated repulsion of neurite outgrowth and that other potential mutations in the HSV-2 genome were not responsible for this phenotype. Supporting this, we did not find significant differences between the genomes of MS-CheGL and MS-CheGL-ΔgG2 or 333-TFP-FL and 333-TFP-FL-ΔgG2 by next-generation sequencing. FLgG2 is proteolytically cleaved, leading to the secretion of SgG2 and retention of the remaining domains of the gG2 type I transmembrane protein in the cells (21)(22)(23)48). The fact that FLgG2 complements CheGL-ΔgG2 more efficiently than SgG2 is interesting and requires further investigation. This may be due to higher protein levels upon transfection of pFLgG2. However, the ectodomain, transmembrane region, or cytosolic tail of gG2 may also actively contribute to overcoming the repulsion of neurite outgrowth by modifying the expression or activity of AGMs and neurotrophic factors. The ectodomain of gG2 is 298 amino acids long and may interact with other cellular proteins involved in neurite outgrowth. There are many AGMs and neurotrophic factors that could be modulated by FLgG2. It will be interesting to determine whether FLgG2 modulates their expression or activity.
Our results also suggest that gG2 is not the only factor overcoming the repelling effect of nonneuronal cells during infection. HEK-293T cells express repellents, such as semaphorin 3A and class A ephrins (38,39,46). The possibility that HSV-2 infection may also modulate the expression or activity of attractive or repulsive cues affecting neurite outgrowth, in addition to the gG-mediated modulation reported here, requires further investigation. One could also postulate the presence of alternative indirect or direct effects of other viral factors on the expression of AGMs. For example, HSV modifies gene expression and protein translation by several mechanisms, including disruption of transcription termination (50,51). Moreover, the virus host shutoff (VHS) protein, encoded by UL41, induces both transcription and translation shutoff (52,53). A potential direct effect of a viral gene could be mediated by HSV-2-infected cell protein 0, since this protein induces IL-17c expression during infection of human keratinocytes (14).
The different impact of HSV-2 MS or 333 infection on neurite outgrowth may have been due to strain-specific effects on the expression or activity of IL-17c or AGMs. This seems more likely than differences in gG2 from these two strains, as their amino acid identity is 98% (only 9 out of 698 residues are different). Moreover, we showed previously that gG2 from HSV-2 333 enhances NGF-induced neurite outgrowth (15). However, we cannot formally exclude the possibility that these differences at the amino acid level are relevant to enhance NGF function during infection. The fact that NGF activity was required to inhibit the repelling effect of HEK-293T cells on the number of neurites, even during HSV-2 infection, suggests that the observed effect is not only due to a nonspecific effect on gene expression in the infected nonneuronal cells.
Our results have several implications for HSV-2 persistence and pathogenesis. Following cell entry at the neurites, HSV-2 establishes lifelong latency in peripheral neurons. Upon reactivation, newly generated virions exit the neurons via the neurites and cause genital herpes, characterized by itch, burn, and pain (54,55). Proinflammatory cytokines expressed during infection by epithelial cells and infiltrating leukocytes may damage neurites and induce neuronal activity and death. However, some of these cytokines-e.g., tumor necrosis factor alpha, IL-1␤, and IL-6 -also induce the expression of NGF by epithelial cells, fibroblasts, and neurons (56-59). Neurotrophic factors send survival signals to the neuronal cell body, reducing neurotoxicity. In this context, HSV-2 reactivation induces the expression of IL-17c in keratinocytes, increasing neurite density and, potentially, neuronal survival in the human genital tract (14). NGF promotes neurite outgrowth, neuronal survival, and neuronal differentiation upon binding to tropomyosin receptor kinase A (TrkA) (60,61). NGF and other neurotrophins also bind to the p75 neurotrophin receptor (p75NTR), a member of the tumor necrosis factor receptor superfamily (62,63). Our previous results indicated that rSgG2 acts mainly on the NGF-TrkA axis (15). Activation of the NGF-TrkA axis increases the survival of responsive neurons, but it may also have pathological consequences, such as the induction of pain (60).
Sensory ganglia contain a diverse population of neurons with different expression and functional profiles (64). TrkA is often coexpressed with the calcitonin gene-related peptide (CGRP) in peptidergic neurons that express the transient receptor potential vanilloid 1 (TRPV1), involved in the sensation of painful stimuli (65,66). NGF induces pain through enhancing the expression and activation of TRPV1 and sodium channels, and injection of rSgG2 in the mouse footpad increases the sensation of thermal pain through TrkA and TRPV1 activation (30,(67)(68)(69). Interestingly, HSV-2 replicates more efficiently in neurons that express CGRP and TrkA (70,71), and its replication is restricted in neurons lacking TrkA and expressing the GDNF receptor, probably leading to the establishment of latency (72,73). Moreover, NGF and GDNF signaling are required to maintain HSV-1 and HSV-2 latency, respectively (74)(75)(76). Therefore, enhanced NGF activity and the subsequent increase in neurite outgrowth may facilitate HSV-2 spread from infected epithelial cells to sensory neurons and promote the survival of neurons expressing TrkA. This could favor initial HSV-2 spread and may contribute to the induction of acute, local pain.
Overall, we show that HSV-2 infection reduced the repulsion of neurites and increased their growth in an NGF-dependent manner. The contribution of gG2 was strain specific, indicating that this was not the sole factor involved. Whether other alphaherpesviruses modulate the activities of neurotrophic factors or AGM to increase neurite outgrowth is currently unknown. We propose that HSV-1 and HSV-2 have developed several strategies to modulate neurite outgrowth and neuronal survival. The discovery and characterization of such mechanisms of neuronal modulation may provide novel therapeutic targets for intervention to either limit the infection of neurons by HSV or reduce virus-induced tissue damage. were cultured in Eagle minimal essential medium supplemented with 8% FCS. All mammalian cells were grown at 37°C in a CO 2 buffered cell incubator.

MATERIALS AND METHODS
The HSV2-BAC MS strain [termed HSV2(MS)BAC in this report] was kindly provided by Jerry Weir (Food and Drug Administration, Bethesda, MD, USA) and has been previously described (33). The generation and characterization of HSV2(MS)Lox, HSV2(MS)Lox-pHCMV mCheLuc (termed MS-CheGL in this report), and HSV2(MS)Lox-pHCMV mCheLuc-ΔgG2 (termed MS-CheGL-ΔgG2 in this report) are described below. HSV-2 strain 333 was kindly provided by Helena Browne (University of Cambridge, Cambridge, UK). The generation of recombinant viruses using this strain is explained below.

Generation of recombinant viruses. (i) Construction of the self-excisable, TK-positive BAC pHSV2(MS)Lox.
The BAC replication origin was inserted into the pHSV2(MS) genome by homologous recombination; thereby, the thymidine kinase (TK)-encoding gene (UL23) was disrupted (33) ( Fig. 2A and  B). We flanked the nonviral sequences with loxP sites and inserted a eukaryotic Cre recombinase expression cassette to remove the BAC replication origin from the viral backbone upon transfection of eukaryotic cells (78,79). To this end, we performed en passant BAC mutagenesis (34) using a DNA fragment amplified from the plasmid pEP-Cre-in using primers TAT TGC CGT CAT CGC CGG GAG GCC TTC CGT TCG GGC TTC CGT GTT TGA AAT AAC TTC GTA TAA TGT ATG C and CCA AGC TAT TTA GGT GAC ACT ATA GAA TAC TCA AGC TTG ACC CCC CAG GCT ATA GGG CGA ATT GGA GCT C (Fig. 2C). The template pEP-Cre-in was generated by inserting an I-SceI site, a kanamycin resistance cassette (aphAI), and a 50-bp duplication into pUC18LC (78). This DNA fragment was amplified from pEP-Kan-S (80) with TTA AGG ATC CAC AAG GAT GAC GAT AAG and TTA AGG ATC CGA AAA GAA AAC GTT GAT GCC GGT GAA CGT GCA AAA CAG GCT CTA GCG TTC CAG GGT AAT GCC AGT GTT AC and inserted via BamHI. After en passant mutagenesis, the remaining 3= end of the disrupted UL23 had been removed (Fig. 2E). In the resulting BAC, pHSV2(MS)LoxCre, the wrongly oriented loxP site at the 3= end of the BAC origin insertion was replaced by bacteriophage Red recombination with a rpsLneo cassette amplified from pRpsL-neo (Gene Bridges, Dresden, Germany) with primers TCT CTG TTT TTG TCC GTG GAA TGA ACA ATG GAA GTC CGA GCT CAT CGC TAG CCT GGT GAT GAT GGC GGG ATC G and CTA TCG CCT CCC TGC TGT GCT ACC CGG CCG CGC GGT ACC TCA TGG GAA GCT CAG AAG AAC TCG TCA AGA AGG (Fig. 2F). This cassette was replaced under streptomycin counterselection with a DNA fragment containing a correctly oriented loxP site and the complete UL23 ORF, which was amplified from HSV-2 DNA with the primers TCT CTG TTT TTG TCC GTG GAA TGA ACA ATG GAA GTC CGA GCT CAT CGC TAA TAA CTT CGT ATA ATG TAT GCT ATA CGA AGT TAT CTA AAC TCC CCC CAC CTC GCG and ATG GCT TCT CAC GCC GGC CAA C (Fig. 2G). In the resulting BAC, pHSV2(MS)Lox, the UL23 ORF encoding TK was restored, and the nonviral sequences were flanked with loxP sites for Cre-mediated excision after the transfection of eukaryotic cells (Fig. 2H). All oligonucleotides were synthesized by Sigma. (
To mutate the US4 ORF in the HSV-2 MS genome, a kanamycin resistance cassette flanked by overhangs containing the stop codon and frameshift mutations was produced by PCR. Primers with overhangs containing the 40-bp sequences targeting US4, the stop codon and frameshift mutations, and a short repeat for the seamless removal of the kanamycin resistance gene cassette were used. The kanamycin resistance gene cassette was amplified from plasmid pEP-KanS2 (80)  After confirming the successful mutagenesis of US4, this BAC was also tagged with mCherry-GLuc using the same approach described above, producing HSV2(MS)Lox-mCherry-GLuc-ΔgG2, named MS-CheGL-ΔgG2 in this study. The successful incorporation of the BAC cassette and the successful mutagenesis of the US4 ORF were checked by PCR, enzymatic restriction analysis, and next-generation sequencing of the viral genome (with an Illumina MiSeq system) (see below).
(iv) Production of virus stocks. The virus stocks for this study were produced from the supernatant and lysed infected Vero cells. In short twenty 15-cm plates with Vero cells were infected at a multiplicity of infection (MOI) of 0.01 and incubated for at least 72 h until a full cytopathic effect was visible. Then, the cells were scraped off and pelleted into 50-ml Falcon tubes. The supernatant was kept on ice, while the cell pellets were disrupted by 3 freeze-thaw cycles. The cell debris was then pelleted in a Heraeus tabletop centrifuge (860 ϫ g, 5 min, 4°C), and the supernatant was then combined with the supernatant from the first round of centrifugation. Virus was then pelleted using a Beckman L8-70 ultracentrifuge (type 19 rotor, 12,000 rpm, 4°C, 90 min), and the pellet was resuspended in 1 ml of cell culture medium.
Analytical preparation of BAC DNA. BAC DNA for analytical purposes was isolated from transformed Escherichia coli GS 1783 cells (80) overnight culture using a NucleoBond XBAC kit from Macherey-Nagel (Düren, Germany).
Southern blot analysis. DNA was digested and separated by electrophoresis. The gels were blotted onto nylon membranes (GE Healthcare, Little Chalfont, UK) and hybridized with an [alpha-32 P]dCTPlabeled probe directed against the chloramphenicol resistance gene encoded in the BAC replication origin (78). The radioactive probe was generated using a random primed DNA labeling kit (Roche, Mannheim, Germany). The hybridization signals were visualized by autoradiography.
Acyclovir plaque reduction assay. A total of 5 ϫ 10 5 U2OS cells/well were seeded into 6-well plates and infected with HSV-2 at an MOI of 0.001 PFU/cell for 1 h at room temperature. Afterwards, the virus inoculum was removed and 2 ml culture medium containing 25 g/ml immunoglobulin G (IgG) and ACV was added. At 2 dpi, the viral plaques were counted to determine the viral titer. Experiments were conducted in triplicate. ACV dilutions with concentrations ranging from 0 to 27 g/ml were prepared in culture medium. The 50% inhibitory concentration value was determined with the calculated virus titer at the given ACV concentration. HSV-neutralizing IgG (25 g/ml; Sigma-Aldrich Chemie GmbH) was added to the culture medium to prevent reinfection of the cells with progeny virions secreted into the medium.
Next-generation sequencing. (i) NGS of BAC-derived viruses (MS strain). Sequencing of BACderived strain MS viruses was performed at the Institute of Virology, Hannover Medical School. The integrity of all recombinant HSV-2 isolates was verified by whole-genome deep sequencing. Briefly, purified BAC DNA and isolated viral genomic DNA from virus reconstituted in mammalian cells were sheared by sonication. To avoid bias by overamplification, library preparation was performed using a Kapa real-time library preparation kit (Kapa Biosystems, Wilmington, MA, USA). Libraries were sequenced on a MiSeq system (Illumina), using the reagent kit (version 3) to generate 2 ϫ 300-base paired-end reads. The reads were mapped to HSV-2 HG52 as a reference strain (GenBank accession no. NC_001798). Genetic variants (variants with single-nucleotide polymorphisms and insertions/deletions) were identified by using the low-frequency variant detector function in CLC Genomics Workbench software. Variants were considered valid under the following conservative criteria: an overall read depth of Ն50, an average base call quality of Ն20, a forward/reverse read balance of 0.3 to 0.5, and a variant frequency of Ն2% (i.e., the relative frequency of a variant at a particular position). Variants detected within homopolymer regions were excluded. Sequence reads were de novo assembled, and the resulting contigs were oriented against an HSV-2 reference strain lacking the terminal repeat long (TRL) and terminal repeat short (TRS) regions. De novo-assembled sequences contain small gaps in repeat regions and are marked as a stretch of nucleotides (85). The sequences are listed as a partial genome in GenBank (MH796783 and MH796784). For replication kinetics with the HSV-2 333 wild-type strain and derived viruses, confluent Vero and HEK-293T cells grown in 12-well plates were infected with 400 l virus inoculum at an MOI of 0.01 for the multistep growth curves. Then, the cells were overlaid with 1 ml fresh DMEM containing 2% fetal bovine serum. At the time points specified above, supernatant or cells (subjected to 3 freeze-thaw cycles) from 2 wells per virus were harvested and titrated (10ϫ dilution series) on Vero cells in duplicate. The cells were stained with crystal violet, and the plaques were manually counted at 2 dpi.
Gaussia luciferase assay. To measure the GLuc activity in infected cell cultures, we used a microplate luminometer (Orion II; Berthold) with an injector system. Fifty microliters of culture supernatant was plated into a 96-well opaque white plate (Nunc) and, by using the injector, was mixed with 50 l 1-g/ml native coelenterazine (a 1-mg/ml stock in methanol [Sigma] diluted in PBS). Enzymatic activity was measured in the well mode with the following sequence: injection (50 l), with no shaking, delay for 0 s, measurement at 10 s. The values were analyzed with GraphPad Prism (version 5.0) software after using the remove baseline function.
Western blotting. Cells were lysed in radioimmunoprecipitation assay buffer (Pierce, Thermo Fisher), the nuclei were removed by centrifugation, and the supernatant was mixed with 5ϫ concentrated loading buffer (60 mM Tris, pH 6.8, 25% glycerol, 2% SDS, 5% ␤-mercaptoethanol, 0.1% bromophenol blue). The cell culture supernatants were directly mixed with 5ϫ loading buffer. Samples were heat denatured (100°C, 5 min), centrifuged for 1 min at 13,000 rpm with a tabletop centrifuge (Eppendorf), and loaded into SDS-PAGE gels. The gels were run at 25 mA/gel for 1 h, and the separated proteins were subsequently transferred onto nitrocellulose membranes with a wet transfer system at 100 V for 1 h. The membranes were then blocked in 5% milk in PBS-0.1% Tween 20 (PBS-T) and incubated overnight with primary antibody. After 3 washes in PBS-T, the membranes were incubated with secondary antibody (IRDye labeled; LI-COR) in 5% milk for 1 h, washed again 3 times in PBS-T and 2 times in PBS, and detected using an Odyssey imaging system (LI-COR).
Culture of primary neurons from dissociated DRG. Primary DRG neurons were dissociated from 35-day-old (Ϯ3 days) BALB/cJ-Ztm mice using a procedure adapted from previous work (31,32). The mice were sacrificed, and the vertebral column was removed and longitudinally split with two sagittal cuts. The spinal cord was then removed to expose the DRG, which were then pulled out, and the attached large axonal processes were trimmed. The DRG were collected in neuron medium (DMEM and Ham's F-12 medium supplemented with 10% FCS, Pen-Strep, and 50 ng/ml mouse NGF 2.5S [catalog no. G514A; Promega]) and then dissociated by sequential digestions with papain (20 min) and a dispasecollagenase mix (20 min). This was followed by disruption with 3 fire-narrowed Pasteur pipettes with decreasing diameters. The cells were washed and, finally, resuspended in 5 l neuron medium per MFC (ϳ1 animal for 4 MFC). The cells were seeded into the somal compartment (SC) of non-plasma-bound MFC with 450-m microgrooves (catalog no. RD450; Xona Microfluidics LLC, USA) according to the instructions in the manufacturer's manual, with the exception that 0.1 mg/ml poly-L-lysine (catalog no. P4832; Sigma) was used to coat the coverslips (24 by 32 mm; Roth), followed by coating with laminin overnight (30 l/MFC with 100 ng/l). The leftover cell suspension was used to count the number of seeded cells (which was usually about 3 ϫ 10 4 cells/compartment but ranged from 2 ϫ 10 4 to 5 ϫ 10 4 cells/compartment). Since we always added the same volume per experiment, we had an equal number of cells per condition in each experiment. The cells were incubated for 24 h before changing the medium to neuron medium supplemented with aphidicolin (4 g/ml; catalog no. A0781-1MG; Sigma) to inhibit the mitosis of nonneuronal cells. Note that aphidicolin also reduces the replication of DNA viruses. The neurons were then cultured for 3 days, if not stated differently.
Coculture of neurons and HEK-293T cells and neurite recruitment assay. Dissociated neurons were prepared and seeded as described above. HEK-293T cells were added to the neurite compartment at 24 h postseeding of the dissociated DRG cells. To do so, HEK-293T cells were scraped off 10-cm cell culture dishes (at 24 h postinfection [hpi]), pelleted, and resuspended in medium (ϳ150 l). Ten microliters of this concentrated cell suspension was then added to the top well of the neurite compartment (NC) part of the MFC, containing the same amount of neuron medium (with NGF and aphidicolin) as the somal compartment (SC), resulting in an 80% to 90% confluent cell monolayer. The MFC were then incubated for 3 days, before the cells in the MFC were fixed and stained.
For the infection experiments, the cells were infected at an MOI of 0.1 in 1 ml (6-well format) and harvested using a cell scraper at 24 hpi, as described above. The number of infected cells was estimated by counting the mCherry-positive cells under a microscope. The infected cell suspension was then mixed with noninfected cells to produce a cell suspension with 20% mCherry-positive cells. Twenty microliters of this cell suspension was subsequently seeded into the MFC, as described above. This corresponded to approximately 2 ϫ 10 5 cells. For transcomplementation experiments, cells were transfected with 0.5 g of the corresponding plasmids using the Lipofectamine 2000 reagent, as recommended by the manufacturer, on the same day that the neurons were prepared. The cells were washed with PBS at 4 h posttransfection and then infected at an MOI of 0.1. The treated cells were then harvested by scraping at 24 hpi and seeded in MFC, as described above.
For the neutralization of NGF, the medium of the neuronal cultures was changed at 24 h postseeding to medium containing 50 ng/ml NGF plus aphidicolin and 1:500-diluted anti-NGF antibody AB-09 (catalog no. MAB2561; R&D Systems), according to the manufacturer's instructions. Infected HEK-293T cells were collected and concentrated as described above, anti-NGF antibody AB-09 was added to the cell suspension (1:500), and the cells were incubated for 5 min before they were added to the MFC. The cultures were then incubated for a further 2 days before fixing and IF staining.
Immunofluorescence. Immunofluorescence staining of cells in MFC was performed as recommended by the manufacturer (Xona manual, non-plasma-bound devices). In short, cells were fixed using 3% paraformaldehyde at room temperature for 30 min and then washed with PBS, permeabilized using 0.2% Triton X-100 for 30 min, and then blocked with PBS plus 3% bovine serum albumin (IF grade, IgG free) and 5% goat or donkey serum, depending on the secondary antibodies used, for at least 1 h. The primary antibodies were diluted in blocking solution as stated above, and then a pressure gradient was created by adding 50 l to the NC and 100 l to the SC. The MFC were then incubated with the primary antibodies at 4°C overnight. On the next day, the MFC were washed with PBS 3 times and incubated with secondary antibodies (diluted 1:1,000) in blocking solution for at least 2 h. After the incubation, the MFC were washed again and 2 drops of ProLong antifade reagent were added to the top SC well and 1 drop was added to the top NC well. The MFC were then incubated at 4°C overnight to allow the antifade reagent to reach into the microgrooves. The MFC were imaged on the next day using a Zeiss Observer Z1 inverted microscope.
Quantification of neurite outgrowth. To quantify neurite outgrowth, we used the Simple Neurite Tracer (47) plug-in in FIJI software (version 1.5 and higher). In short, multidimensional microscopic images from the Zeiss Observer microscope (Zeiss ECPlan Neofluar 10ϫ objective; numerical aperture, 0.3 [∞/Ϫ]; catalog number 440330-9902) were loaded into the FIJI software and split into single-channel images, and the channel corresponding to tubulin ␤-III staining (Tuj1 antibody) was used for measuring neurites. Neurites were measured from the growth cone to their exit point from the microgroove, when possible, by following them and marking them with a path (count mask). To estimate the overall neurite density, all neurites visible in an image were covered with paths, including partially visible neurites. When forks in the path of a neurite were encountered, the shorter connection to the entry point was usually preferred; also, turning angles of more than 40°were avoided, unless there was a clear path. All endpoints were connected to the entry points, and any remaining neurites were measured as visible and connected to the closest entry point. To estimate neurite density, all marked paths were then measured by the plug-in, recording the number of paths and the length of every path.
Statistical analysis. For statistical analysis, the software GraphPad Prism (version 8.4.3 for Windows; GraphPad Software, San Diego, CA, USA) was used. We followed the recommendations of the journals Journal of Virology (89) and Infection and Immunity (90) regarding statistical analysis. Data on the neurite length and the number of neurites showed a lognormal distribution. Therefore, all data sets containing neurite length or the number of neurites were log 2 transformed to make the data more normally distributed. Since in some experiments the data groups also showed a significant difference in variances (Bartlett's test for equal variances) and the number of measurements (n) was different in many groups, several prerequisites for application of the standard one-way analysis of variance (ANOVA) were violated. Therefore, for all statistical analyses involving data sets with significantly different standard deviations (Bartlett's test and the Browne-Forsythe test), a Welch/Browne-Forsythe-corrected ANOVA was used with an appropriate posttest, depending on the number (Games-Howell's posttest for n Ͼ 50; Dunnett's T3 posttest for n Ͻ 50). For data sets with equal variances, an ANOVA with Sidak's posttest was applied. If the normality of the transformed data set could not be guaranteed (Fig. 5A)