Herpes simplex virus 1 protein pUL21 alters ceramide metabolism by activating the interorganelle transport protein CERT

Herpes simplex virus (HSV)-1 dramatically alters the architecture and protein composition of cellular membranes during infection, but its effects upon membrane lipid composition remain unclear. HSV-1 pUL21 is a virus-encoded protein phosphatase adaptor that promotes dephosphorylation of multiple cellular and virus proteins, including the cellular ceramide (Cer) transport protein CERT. CERT mediates nonvesicular Cer transport from the endoplasmic reticulum to the trans-Golgi network, whereupon Cer is converted to sphingomyelin (SM) and other sphingolipids that play important roles in cellular proliferation, signaling, and membrane trafficking. Here, we use click chemistry to profile the kinetics of sphingolipid metabolism, showing that pUL21-mediated dephosphorylation activates CERT and accelerates Cer-to-SM conversion. Purified pUL21 and full-length CERT interact with submicromolar affinity, and we solve the solution structure of the pUL21 C-terminal domain in complex with the CERT Pleckstrin homology and steroidogenic acute regulatory–related lipid transfer domains using small-angle X-ray scattering. We identify a single amino acid mutation on the surface of pUL21 that disrupts CERT binding in vitro and in cultured cells. This residue is highly conserved across the genus Simplexvirus. In addition, we identify a pUL21 residue essential for binding to HSV-1 pUL16. Sphingolipid profiling demonstrates that Cer-to-SM conversion is severely diminished in the context of HSV-1 infection, a defect that is compounded when infecting with a virus encoding the mutated form of pUL21 that lacks the ability to activate CERT. However, virus replication and spread in cultured keratinocytes or epithelial cells is not significantly altered when pUL21-mediated CERT dephosphorylation is abolished. Collectively, we demonstrate that HSV-1 modifies sphingolipid metabolism via specific protein–protein interactions.

Herpes simplex virus (HSV)-1 dramatically alters the architecture and protein composition of cellular membranes during infection, but its effects upon membrane lipid composition remain unclear. HSV-1 pUL21 is a virus-encoded protein phosphatase adaptor that promotes dephosphorylation of multiple cellular and virus proteins, including the cellular ceramide (Cer) transport protein CERT. CERT mediates nonvesicular Cer transport from the endoplasmic reticulum to the trans-Golgi network, whereupon Cer is converted to sphingomyelin (SM) and other sphingolipids that play important roles in cellular proliferation, signaling, and membrane trafficking. Here, we use click chemistry to profile the kinetics of sphingolipid metabolism, showing that pUL21-mediated dephosphorylation activates CERT and accelerates Cer-to-SM conversion. Purified pUL21 and full-length CERT interact with submicromolar affinity, and we solve the solution structure of the pUL21 C-terminal domain in complex with the CERT Pleckstrin homology and steroidogenic acute regulatory-related lipid transfer domains using small-angle X-ray scattering. We identify a single amino acid mutation on the surface of pUL21 that disrupts CERT binding in vitro and in cultured cells. This residue is highly conserved across the genus Simplexvirus. In addition, we identify a pUL21 residue essential for binding to HSV-1 pUL16. Sphingolipid profiling demonstrates that Cer-to-SM conversion is severely diminished in the context of HSV-1 infection, a defect that is compounded when infecting with a virus encoding the mutated form of pUL21 that lacks the ability to activate CERT. However, virus replication and spread in cultured keratinocytes or epithelial cells is not significantly altered when pUL21mediated CERT dephosphorylation is abolished. Collectively, we demonstrate that HSV-1 modifies sphingolipid metabolism via specific protein-protein interactions.
Herpes simplex virus (HSV)-1 is a human pathogen that is estimated to infect the majority of the human population, causing a lifelong latent infection (1). Latent HSV-1 resides in sensory neurons or sympathetic neurons, migrating to the periphery in periodic reactivation events throughout the lifetime of the host. In order to sustain acute (lytic) infection, HSV-1 drastically modifies infected cells (2)(3)(4)(5). In particular, the virus extensively remodels the composition and architecture of cellular membranes to facilitate virus assembly and spread. Nascent capsids leave the nucleus via sequential envelopment and de-envelopment at the inner and outer nuclear membranes (6). These cytosolic capsids acquire a proteinaceous layer termed "tegument" and bud into the lumen of post-Golgi membranes that are studded with viral glycoproteins (so called "secondary envelopment") (7). The resultant virus-containing vesicles are transported to cell contact sites where they release the mature virus particles to disseminate the infection (8).
HSV-1 pUL21 is a tegument protein that is conserved in all alphaherpesviruses (9). This multifunctional protein and its homologs are known to interact with multiple cellular and viral partners, including pUL16 (10,11), pUL11 (12), gE (12), tubulin (13), and Roadblock-1 (14), and it has been implicated in a number of important processes including capsid nuclear egress (15), viral cell-to-cell spread (12,16), and retrograde transport along axons to the neuronal cell bodies where latency is established (14,17). Mutant HSV-1 lacking pUL21 expression exhibits a 10-fold to 100-fold replication defect and severely impaired cell-to-cell spread, both of which can be at least partially ascribed to the phosphomodulatory role of pUL21 as a protein phosphatase 1 (PP1) adaptor (18). PP1 is a highly active and abundant cellular phosphatase (19), and pUL21 recruits PP1 to promote dephosphorylation of multiple substrates, including the viral protein pUL31 that is implicated in viral nuclear egress (18,20) and the cellular protein CERT that regulates sphingomyelin (SM) metabolism (18).
The cytoplasmic ceramide (Cer) transport protein CERT (a.k.a. Goodpasture's antigen-binding protein, GPBP, encoded by the gene COL4A3BP) mediates the nonvesicular trafficking of Cer from the endoplasmic reticulum (ER) to the trans-Golgi network (TGN) and, in doing so, defines the rate of SM synthesis (21)(22)(23). CERT contains two well-folded globular domains: an N-terminal Pleckstrin homology (PH) domain that mediates its interaction with TGN membranes by binding phosphatidylinositol 4-phosphate and a C-terminal steroidogenic acute regulatory-related lipid transfer (START) domain that directly binds Cer to mediate its transfer (23). Crystal structures of both domains have been solved (24,25). The "middle region" (MR) that connects the PH and START domains of CERT does not adopt a globular fold, instead containing a coiled-coil region that is likely to mediate CERT self-association (26), a "two phenylalanines in an acidic tract" (FFAT) motif that recruits CERT to the ER membranes via binding proteins VAPA and VAPB (27), and a serine-rich motif, hyperphosphorylation of which represses CERT activity (28). In cultured HeLa cells, the vast majority of CERT is in an inactive hyperphosphorylated state (CERT P ), available to be mobilized via dephosphorylation to increase the rate of ER-to-TGN Cer transport in response to stimuli such as SM depletion (28).
Cer and sphingolipids like SM are essential for mammalian cell growth, and they play important roles in cell signaling, apoptosis, and membrane trafficking (29). Furthermore, sphingolipids and cholesterol participate in the formation of membrane microdomains (including "lipid rafts") that compartmentalize the lipid and protein composition of cellular membranes (30). This compartmentalization is especially important in highly polarized cells such as neurons (31). Although sphingolipids are crucial to host cell biology, relatively little is known about the interaction between pathogens and cellular sphingolipid metabolism. The bacteria Chlamydia trachomatis is known to directly recruit CERT to bacteriacontaining intracellular inclusions, thereby increasing the abundance of SM in the inclusion membrane (32), and CERT activity is necessary for the biosynthesis of double-membrane vesicles that serve as the sites of hepatitis C virus and polio virus replication (33). However, to the best of our knowledge, the binding of CERT to HSV-1 pUL21 (18) is the only known example of a direct interaction between CERT and a virus protein.
Previous studies have shown that HSV-1 infection increases Cer synthesis (34) and that depletion of sphingomyelinase causes a >30-fold decrease in HSV-1 replication (35). Sphingomyelinase treatment of cultured epithelial cells has only a very modest effect upon HSV-1 entry (36), suggesting that SM is not absolutely required for virus infection. In contrast, recent studies in macrophages demonstrated that acid ceramidase activity is potently antiviral, helping sequester incoming virus particles by promoting their neutralizing association with sphingosine (Sph)-rich intraluminal vesicles within endocytic multivesicular bodies (37). Furthermore, siRNA depletion of CERT has been shown to promote secretion of usually cell-associated HSV-1 virions (38). It is therefore increasingly clear that cellular sphingolipid metabolism is important for HSV-1 biology and that further studies are required to obtain a full picture of how specific sphingolipids support and/or prevent HSV-1 infection.
We sought to define the functional consequences of the pUL21-CERT interaction for HSV-1 infection. Using click chemistry, we demonstrated that pUL21-mediated CERT dephosphorylation increases the rate of Cer to SM conversion in cultured cells. We characterized the solution structure of the pUL21 C-terminal domain in complex with the PH and START domains of CERT, identifying a specific pUL21 amino acid required for the interaction. Functional characterization of mutant HSV-1 encoding this mutated form of pUL21 confirmed that pUL21-mediated CERT dephosphorylation alters the rate of SM synthesis in infected cells. Furthermore, we observed a dramatic increase in the rate of Cer accumulation in infected cells, which was only partly counteracted by pUL21-stimulated CERT activity, but this increase in Cer abundance does not significantly alter virus replication or spread in cultured keratinocytes or epithelial cells.

Results
Human cells typically maintain lipidome homeostasis via robust feedback and control mechanisms, which respond to perturbations via local activation of specific signaling pathways (39,40). We hypothesized that pUL21, expressed at late points of infection (3), would likely alter the rate of sphingolipid synthesis and thereby affect lipid-mediated signaling events, rather than altering the overall steady-state lipid composition of infected cells. Synthetic lipids with alkyne-containing acyl chains are efficiently processed by mammalian lipid-modifying enzymes and represent powerful tools to probe lipid metabolism (41). The impact of pUL21-directed CERT dephosphorylation on sphingolipid metabolism was therefore probed using a clickable analog of Sph, alkyne-Sph, to monitor the rate of SM biogenesis in immortalized human keratinocyte (HaCaT) cells. Exogenous Sph is efficiently incorporated into cellular metabolic pathways, being rapidly converted into Cer and then SM or hexosylceramides like glucosylceramide or galactosylceramide (42). It is also converted into phosphatidylcholine (PC) via a so-called "salvage" pathway ( Fig. 1A) that directs Sph to palmitoyl-CoA, which serves as substrate for reacetylation of lysophosphatidylcholine (43).
To monitor Sph metabolism, HaCaT cells, either parental or stably expressing pUL21 (HaCaT21), were incubated with alkyne-Sph for 5 min (pulse), and the rate of alkyne-Sph incorporation into the competing metabolic pathways was monitored for 2 h (chase) by high-performance TLC (HPTLC) separation and detection of lipids conjugated to coumarinazide via a "click" reaction ( Fig. 1, B and C). Alkyne-Sph was very efficiently converted to alkyne-Cer in both cell types, the levels of alkyne-Cer remaining relatively stable throughout the chase, and for both cell types, synthesis of alkyne-PC and alkyne-SM was observed but not synthesis of alkynehexosylceramide. HaCaT21 cells exhibited significantly reduced rates of alkyne-Sph conversion, and the level of alkyne-Cer is significantly lower (Fig. S1). The rate of alkyne-PC synthesis was also reduced, although this reduction was not significant owing to the high interexperiment variability of the alkyne-PC signal for the parental HaCaT cells. Taken together, these results indicated that lipid metabolism is altered in HaCaT cells constitutively expressing pUL21. However, it remained unclear whether these differences reflected adaptations of cellular lipid metabolism in response to constitutive pUL21 expression rather than a direct effect upon CERT activity. Since CERT-mediated nonvesicular transport of Cer from the ER to the TGN defines the rate of Cer-to-SM conversion (21)(22)(23), pUL21-directed CERT dephosphorylation (and thus activation) should increase the rate of SM synthesis. Consistent with this hypothesis, the abundance of alkyne-SM as a fraction of total alkyne-Cer plus alkyne-SM was significantly higher in HaCaT21 versus HaCaT cells at early time points (0-30 min) during the chase (Fig. 1D). By the 60 min chase time point, the rate of alkyne-SM accumulation slows and the difference in relative abundance between cell lines was diminished, consistent with the alkyne-SM levels in both cell types approaching equilibrium. This increase in alkyne-SM accumulation as a fraction of alkyne-Cer plus alkyne-SM abundance was observed despite similar overall signal for alkyne-SM and alkyne-Cer in HaCaT21 cells when compared with parental cells at early time points (0-30 min), the abundance of these lipids being significantly lower in HaCaT21 cells at later time points (Fig. 1E). Reduced overall levels of SM plus Cer in HaCaT21 cells likely result from adaptation to constitutive pUL21 expression (and thus constitutive CERT hyperactivation) via stimulation of feedback pathways that reduce Sph to Cer conversion and/or increased back conversion of SM and/or Cer to Sph. Treatment of HaCaT cells with the highly specific CERT inhibitor HPA-12 (44) confirmed that the rate of Cer-to-SM conversion Herpes simplex virus 1 (HSV-1) pUL21 alters the rate of ceramide (Cer) to sphingomyelin (SM) conversion in cultured cells. A, simplified schematic diagram of sphingolipid biosynthetic pathways that lead from sphingosine (Sph) to SM and glycosphingolipids (hexosylceramide; HexCer), or to phosphatidylcholine (PC) via a "salvage" pathway. In this salvage pathway, Sph-1-phosphate is converted to trans-2-hexadecenal, then trans-2hexadecenoic acid and trans-2-hexadecenoyl-coA before being converted to palmitoyl-coA (112), which is in turn conjugated to glycerol-3-phosphate to form lysophosphatidic acid before being converted to phosphatidic acid, diacylglycerol, and then PC (113). Cer to SM conversion is accelerated by the transport protein CERT, which is selectively inhibited by HPA-12. B, rate of Sph conversion to Cer, SM, or PC was measured in HaCaT cells, untreated or treated with 1 μM HPA-12, and in HaCaT cells stably expressing pUL21 (HaCaT21). Cells were incubated with 5 μM "clickable" alkyne-Sph (pulse) for 5 min and harvested for lipid extraction either immediately (0 min) or at the indicated times (chase). Extracted lipids were bioconjugated to 3-azido-7hydroxycoumarin, separated by HPTLC and detected using UV light. Separated lipids were identified using clickable standards (Stds) and previous literature (42). Data are from one representative experiment of two independent repeats. C, quantitation of the lipid intensities from (B) as determined by densitometry and represented as percentage fraction of total signal. D, rate of SM synthesis expressed as its fraction in the cumulative signal for SM and Cer. E, the ratio of alkyne-Sph incorporated into either Cer or SM as a proportion of total alkyne-lipid signal, representing the influx of Sph into the SM biosynthetic pathway. For D and E, the data represent two independent experiments (mean ± SD). Data points are labeled if significantly different to parental HaCaT cells: *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001 (two-way ANOVA with Dunnett's multiple comparisons test). F, the resolubilized proteins precipitated during lipid extraction were analyzed by SDS-PAGE and immunoblotting using the antibodies listed. Where indicated, the gel was supplemented with PhosTag reagent to retard the migration of phosphorylated proteins, thus enhancing the separation of CERT that is hypophosphorylated (CERT O ) or hyperphosphorylated (CERT P ). GAPDH serves as a loading control.
HSV-1 pUL21 modulates cellular ceramide transport was defined by CERT activity: SM synthesis was significantly decreased in HPA-12-treated cells when compared with untreated cells, and there was concomitant accumulation of alkyne-Cer in treated cells (Figs. 1, B-D and S1). Immunoblot analysis of protein samples that were resolubilized following their precipitation during lipid extraction (Fig. 1F) confirmed that pUL21 was expressed and that dephosphorylated CERT (CERT O ) predominates in HaCaT21 cells, in contrast to the parental and HPA-12-treated HaCaT cells.
pUL21 has multiple functions during infection, promoting the PP1-mediated dephosphorylation of not just CERT but of multiple different cellular and viral proteins (18). It was therefore essential to identify a point mutant of pUL21 with impaired binding to CERT, but not to other cellular or viral partners, to dissect the functional significance of increased CERT activity during infection. Defining the interaction surfaces of transient protein complexes presents a considerable challenge (45), made more difficult in the case of pUL21 and CERT by their containing multiple domains and regions of intrinsic disorder (18). The minimal region of CERT required for pUL21 binding was thus probed using pull-down experiments where pUL21-glutathione-S-transferase (GST) purified following bacterial expression (bait) was used to capture CERT or truncations thereof (preys) expressed by in vitro coupled transcription/translation in wheat germ extract ( Fig. 2A). The longer cytoplasmic isoform of CERT (CERT L ), where expression of exon 11 extends the amino-terminal region of the START domain by 26 amino acids (START L ), was used since both CERT and CERT L have been shown to possess Cer transfer activity (23), and a contribution of these additional 26 amino acids to pUL21 binding could not be excluded. Fulllength CERT L was efficiently captured by pUL21-GST. The PH and START L domains were also captured by pUL21-GST when expressed individually, albeit less efficiently, suggesting that the CERT L possesses multiple pUL21-binding motifs. The CERT MR was not captured by pUL21-GST and is thus likely to be dispensable for binding. A truncated CERT L construct that retained the pUL21-binding PH and START L domains, but lacked the majority of the highly flexible MR, was thus designed for use in subsequent biochemical and structural studies (miniCERT L ) (Fig. 2B). The extent of the MR retained to link the two domains was informed by the crystal structure of the CERT PH-START domain complex (46), defining a minimal distance of at least 55 Å between the PH domain C terminus and the START domain N terminus, and the desire to exclude the predicted coiled-coil region that may cause oligomerization (26). MR residues 132 to 350 were thus excluded from miniCERT L . The first 19 amino acids of the PH domain were also omitted from miniCERT L as this region is predicted to be highly disordered and was excluded in previous structural studies (24,47).
Previous studies identified that the MR mediates CERT trimerization (26,48). Size-exclusion chromatography with inline multiangle light scattering (SEC-MALS) (Fig. 2C) confirmed that hyperphosphorylated StrepII-tagged CERT L P , purified following expression in mammalian (Freestyle 293F) cells, has an observed mass (184.3 kDa) approaching that expected for a trimer (215.7 kDa). Similarly, a constitutively hypophosphorylated form of CERT L where S132 is mutated to alanine, and thus cannot become phosphorylated to initiate the serine-rich motif phosphorylation cascade (28), had an observed mass (199.2 kDa) approaching that expected for a trimer. In contrast, H 6 -miniCERT L purified following bacterial expression was predominantly monomeric and monodisperse (observed mass of 43.9 kDa, expected monomeric mass of 44.9 kDa), consistent with trimerization of CERT L being driven by the MR and not being dependent upon CERT L phosphorylation.
Isothermal titration calorimetry (ITC) demonstrated that pUL21-H 6 (18) binds StrepII-CERT L with approximately micromolar affinity (Fig. 2, D and E, Tables 1 and S1), the observed affinity not differing significantly between the hyperphosphorylated (CERT L P ) and hypophosphorylated (CERT L S132A ) forms of the protein. While StrepII-CERT L S132A forms a 1:1 complex with pUL21, the observed binding stoichiometry (N) was consistently lower for StrepII-CERT L P (0.74), consistent with a proportion of the pUL21 binding sites on CERT L being sterically occluded when the protein is hyperphosphorylated. H 6 -miniCERT L and pUL21-H 6 form an equimolar complex, binding with micromolar affinity similar to the full-length protein ( Fig. 2F and Table 1). However, the thermodynamics of binding differ significantly; whereas pUL21 binding to CERT L is endothermic and entropically driven, binding to miniCERT L is exothermic and enthalpically driven with minimal change in overall entropy (Table 1). This is consistent with CERT L undergoing significant conformational rearrangement upon binding to pUL21, whereas the conformational changes to miniCERT L (if any) upon pUL21 binding are likely to be much more modest. Small-angle X-ray scattering (SAXS) in batch mode was used to probe the solution structure of H 6 -miniCERT L ( Fig. 2G and Table S2). The frequency distribution of real-space distances (p(r) profile) is largely symmetric (Fig. 2H), and the dimensionless Kratky plot (Fig. 2I) has a bell-shaped peak at sR g = √3, consistent with miniCERT L having a compact globular conformation in solution. Ab initio modeling of the H 6 -miniCERT L scattering profile using GASBOR (49) yields a dummy-residue model (Fig. 2, J and K) that closely resembles the crystal structure of CERT START and PH domain complex (46). A pseudoatomic model of the H 6 -miniCERT L solution structure, generated using the crystal structures of the CERT PH and START domains determined in isolation (24,25) as rigid bodies, confirms the similarity of the H 6 -miniCERT L solution model to the crystal structure of the CERT PH-START domain complex (Fig. 2, L and M). Collectively, the MALS, ITC, and SAXS analyses demonstrate that miniCERT L forms a compact monomeric protein with high affinity for pUL21, representing an excellent tool for structural studies.
Structural characterization via X-ray crystallography and SAXS has revealed pUL21 to comprise two domains joined by a highly flexible linker (Fig. 3A) (18,50,51), and immunoprecipitation experiments mapped CERT binding to the pUL21 C-terminal domain (18). The C-terminal domain of pUL21 (pUL21C), spanning amino acids 275 to 535, was HSV-1 pUL21 modulates cellular ceramide transport Figure 2. pUL21 binds to the PH and START L domains of CERT L and to a monomeric form of CERT L comprising just the PH and START L domains (miniCERT L ). A, minimal binding elements of CERT L were determined via pull down (PD) experiment using immobilized purified pUL21-GST or GST alone to capture myc-tagged full-length CERT L or truncations thereof expressed via in vitro transcription/translation. Captured proteins were subjected to SDS-PAGE and immunoblotting using an anti-myc antibody. Ponceau S (PonS) staining of the immunoblot membrane before blocking shows equal and efficient capture of the bait proteins across the tested conditions. B, schematic representation of miniCERT L . Dotted lines indicate the regions of full-length CERT L that were omitted. C, SEC-MALS elution profiles (normalized differential refractive index, dRI) of StrepII-CERT L (black dotted), StrepII-CERT L S132A (gray solid), and H 6 -miniCERT L (green solid). Weight-averaged molecular masses (colored solid lines) are shown across the elution peaks. The expected molecular masses for trimeric StrepII-CERT L and StrepII-CERT L S132A , and for monomeric H 6 -miniCERT L , are shown as dotted horizontal lines. Inset shows Coomassie-stained SDS-PAGE of the samples used for SEC-MALS (left), plus SDS-PAGE supplemented with PhosTag reagent (right) of the StrepII-CERT L proteins. D-F, representative ITC titration curve of pUL21-H 6 binding to (D) StrepII-CERT L , (E) StrepII-CERT L S132A , and (F) H 6 -miniCERT L . Insets show normalized binding curves with integrated changes in enthalpy (ΔH) as a function of molar ratio. The affinity (K D ), ΔH, and stoichiometry (N) for the presented titrations are displayed below. G, SAXS profile measured for H 6 -miniCERT L . The reciprocal-space fit of the p(r) profile to the SAXS data is shown as a yellow line. χ2, fit quality; p, Correlation Map (CorMap) probability of systematic deviations between the model fit and the scattering data (92). Inset shows the Guinier plot (sR g < 1.3), which is linear as expected for an aggregate-and repulsion-free system. H, the real-space distance distribution function, p(r), calculated from the SAXS profile. I, dimensionless Kratky plot of the SAXS data. Gray dotted lines indicate the expected maximum of the plot for a compact protein (sR g = √3, (sR g ) 2 I(s)/I(0) = 3e −1 ) (114). J, fit of an ab initio dummy-residue model calculated using GASBOR to the SAXS profile and (K) representative GASBOR dummy-residue model. White spheres indicate modeled water beads of the hydration shell. L, fit to the SAXS profile of the pseudoatomic model of H 6 -miniCERT L obtained using CORAL. M, CORAL pseudoatomic model of H 6 -miniCERT L (green ribbons) superimposed onto the crystal structure of the complex between the CERT PH and START domains (cyan ribbons, PDB ID: 5JJD) (46) by aligning the START/START L domains. Superposition is shown in orthogonal orientations where the linker HSV-1 pUL21 modulates cellular ceramide transport expressed with an N-terminal hexahistidine tag and purified following bacterial expression. In contrast to a previous report (51), we did not observe copurification of nucleic acid with H 6 -pUL21C, the purified protein having an absorbance ratio of 0.6 at 260/280 nm. SEC-MALS and SAXS analysis (Fig. S2) confirmed that H 6 -pUL21C is monomeric and monodisperse, adopting a compact structure in solution that matches the previously determined crystal structure (χ 2 = 0.99, CorMap p = 0.133) (51). SEC analysis of a preformed complex between purified H 6 -miniCERT L and a 1.34-fold molar excess of H 6 -pUL21C confirmed that the two proteins form a stable complex (Fig. 3B), suitable for structural studies. ITC demonstrated that H 6 -pUL21C forms an equimolar complex with H 6 -miniCERT L with a dissociation constant (K D ) of 4.3 ± 1.2 μM (Fig. 3C, Tables 1 and S1). The approximately fourfold reduction in miniCERT L binding by pUL21C versus full-length pUL21 is consistent with previous immunoprecipitation results, where transfected pUL21C-GFP captured endogenous CERT slightly less efficiently than did pUL21-GFP in human embryonic kidney 293T (HEK293T) cells (18). These ITC experiments confirm that the C-terminal domain of pUL21 is the major determinant of CERT binding. Attempts to predict the structure of pUL21C in complex with miniCERT L using AlphaFold-Multimer (52) did not yield high confidence models, presumably because of a paucity of coevolutionary signal between these viral and human proteins, and structural characterization of this complex thus required an experimental approach.
To probe the structural basis of the CERT recruitment by pUL21, a preformed complex of H 6 -miniCERT L and H 6 -pUL21C was subjected to SEC with inline SAXS measurement (SEC-SAXS, Fig. 3D). SAXS data were processed by averaging frames with a consistent calculated radius of gyration (R g ) and then subtracting averaged buffer frames to yield the H 6 -miniCERT L -H 6 -pUL21C complex scattering profile (Fig. 3E). The probable frequency of real-space distances (p(r) profile) of the complex is moderately asymmetric (Fig. 3F), in contrast to the highly symmetric p(r) profiles of H 6 -pUL21C ( Fig. S2D) or H 6 -miniCERT ( Fig. 2H) alone, suggesting a less spherical particle, and the dimensions of the complex (R g = 3.5 nm, D max = 13.6 nm) are substantially larger than for H 6 -pUL21C (R g = 2.2 nm, D max = 8.5 nm) or H 6 -miniCERT (R g = 2.7 nm, D max = 9.1 nm). The peak of the dimensionless Kratky plot is slightly higher, with its peak away from sR g = √3 (Fig. 3G), suggesting some flexibility in the system granted either by modest dissociation of the complex or some mobility of the H 6 -miniCERT L domains with respect to each other and H 6 -pUL21C. Ab initio modeling using GASBOR indicated an elongated molecule (Fig. 3H). While initial pseudoatomic models of the complex generated using a fixed conformation of H 6 -miniCERT L did not fit the SAXS profile acceptably, allowing the PH and START L domains freedom to move with respect to each other and to H 6 -pUL21C yielded three pseudoatomic models with high-quality fits to the SAXS profile (Figs. 3, I and J and S3). In all three top models, miniCERT L binds "end-on" to the pUL21 molecule, forming an ellipsoidal "rugby ball"-like particle. In two of the top three models, miniCERT L binds the "left wing" of the dragonfly-shaped pUL21C domain (51), whereas in the other, it binds the "right wing" (Figs. 3, I and J and S3). All these models have a similar overall shape, and thus all explain the SAXS scattering data well, but the relative orientations of H 6 -miniCERT L and H 6 -pUL21C differ. All three top models were thus used to design specific pUL21 mutations that might disrupt (mini) CERT binding.
Mutations in pUL21C were designed to identify whether miniCERT L binds the left wing or right wing of this domain. Amino acids in helix α4 or the subsequent loop of pUL21C were mutated to test binding to the left wing ( Fig. 3I), whereas amino acids in helices α5 and α9 were used to test binding to the right wing (Fig. 3J). Immunoprecipitation experiments in transfected HEK293T cells that had been infected with HSV-1 lacking pUL21 expression (HSV-1 ΔpUL21) demonstrated that four of five substitution in the left wing of pUL21C (P365D, V368E, R373E, and V382E) disrupted the ability of CERT to coprecipitate with pUL21-GFP (Fig. 4A), whereas none in the pUL21C right wing disrupted CERT coprecipitation (Fig. 4B). These results are consistent with CERT binding the left wing of pUL21C. They are also consistent with observations that CERT binding is lost when pUL21C has a bulky N-terminal GFP tag but retained when the GFP tag is C terminal (18), as the left-wing models place the pUL21C amino terminus but not the carboxy terminus in close proximity to miniCERT L (Figs. 3I and S3).
Of the pUL21 substitutions that disrupted CERT binding (Fig. 4A), V382E appeared to cause the largest decrease in regions or termini that were modeled by CORAL are depicted as yellow spheres. DP, differential power; GST, glutathione-S-transferase; ITC, isothermal titration calorimetry; PDB, Protein Data Bank; PH, Pleckstrin homology; SAXS, small-angle X-ray scattering; SEC-MALS, size-exclusion chromatography with inline multiangle light scattering; START, steroidogenic acute regulatory-related lipid transfer. As quantitated by ITC. Experiments were performed n (replicates) times and mean ± SD values are shown. Data for individual titrations are presented as supporting information (Table S1).
HSV-1 pUL21 modulates cellular ceramide transport  Figure 2F. The affinity (K D ), ΔH, and stoichiometry (N) for the presented titration are displayed below. D, SEC elution profile (partial integrated scattered X-ray intensity versus data frame number) obtained during SEC-SAXS analysis of H 6 -pUL21C-H 6 -miniCERT L complex. Dashed line indicates the calculated radius of gyration (R g ) across the frames averaged for structural analyses. E, averaged SAXS profile of the H 6 -pUL21C-H 6 -miniCERT L complex. The reciprocal-space fit of the p(r) profile to the SAXS data is shown as a yellow line. χ 2 , fit quality; p, Correlation Map (CorMap) probability of systematic deviations between the model fit and the scattering data (92). Inset displays the Guinier plot (sR g < 1. 3), which is linear as expected for an aggregate-and repulsion-free system. F, real-space distance distribution function, p(r), calculated from the SAXS profile. G, dimensionless Kratky plot of the SAXS data. Gray dotted lines indicate the expected maximum of the plot for a compact protein (sR g = √3, (sR g ) 2 I(s)/I(0) = 3e −1 ). H, ab initio modeling of H 6 -miniCERT L -H 6 -pUL21C using GASBOR. Fit of the calculated scattering (yellow) to the SAXS profile is shown, as is a representative dummy-residue model (blue spheres) with modeled water beads of the hydration shell (white spheres). I and J, pseudoatomic models of the H 6 -miniCERT L -H 6 -pUL21C complex generated using CORAL. The fits of the computed scattering (yellow) to the H 6 -pUL21C-H 6 -miniCERT L SAXS profile (blue) are shown. High-quality fits are obtained with models where miniCERT L binds either the left or right "wings" of the dragonfly-like pUL21C domain (51). pUL21C is shown as a violet ribbon with "wings" and termini labeled, and miniCERT L as a green molecular surface (PH, light green; START L , dark green). For clarity, regions absent from the crystal structures that were modeled by CORAL are not displayed. Residues from the left and right "wings" of pUL21C that were selected for further investigation are shown as sticks with red (I) or pink (J) carbon atoms, respectively. An additional pseudoatomic model plus its fit to the H 6 -pUL21C-H 6 -miniCERT L SAXS profile is shown in Figure S3, where the fits presented in (I) and (J) are also reproduced for ease of comparison. DP, differential power; ITC, isothermal titration calorimetry; PDB, Protein Data Bank; PH, Pleckstrin homology; SAXS, small-angle X-ray scattering; SEC, size-exclusion chromatography; START, steroidogenic acute regulatory-related lipid transfer.  CERT binding while maintaining the ability of pUL21 to coprecipitate its other known binding partners, PP1 (18) and pUL16 (10). The presence of a valine residue at this position is absolutely conserved in HSV-1 and HSV-2 and is highly conserved across members of the simplexvirus genus (Fig. S4). H 6 -pUL21 V382E was purified following bacterial expression (Fig. 4C). Differential scanning fluorimetry (a.k.a. Thermofluor) showed pUL21 V382E to be well folded as its thermal stability is similar to WT H 6 -pUL21 (Fig. 4D). ITC analysis demonstrated that H 6 -pUL21 V382E has approximately eightfold reduced binding affinity for H 6 -miniCERT L when compared with WT pUL21 (Fig. 4E, Tables 1 and S1). The effect of the V382E substitution upon the ability of pUL21 to promote CERT dephosphorylation, converting CERT P to CERT O , was probed using an in vitro dephosphorylation assay with allpurified reagents (Fig. 4, F  While mutation of the pUL21C right wing did not prevent CERT binding, one substitution (L490E) abolished binding to pUL16 (Fig. 4B). pUL21 L490E retains the ability to bind CERT and PP1, suggesting that the mutant is stable and maintains a correct overall fold. A model of pUL16 in complex with pUL21C was generated using AlphaFold-Multimer (52), revealing that pUL16 is predicted to bind the right wing of pUL21C, with helix α9 spanning residues 486 to 497 being central to this interaction (Fig. 4H). The structures of pUL16 and pUL21C, and their relative orientations in the complex, were predicted with high overall confidence (Fig. 4I). pUL16 is predicted to comprise two domains, an N-terminal domain (residues 35-172) that is joined by a short linker to a C-terminal domain (residues 179-371) (Fig. S5A). A structural homology search using DALI (53) revealed that the pUL16 C-terminal domain shares greatest structural similarity to the N-terminal domain of pUL21 ( Fig. 5B and Table S3), whereas the pUL16 C-terminal domain is similar in fold to a variety of bacterial and eukaryotic proteins that have diverse annotated functions ranging from phosphatase activity to lipid binding (Fig. 5, C and D and Table S3). Residues 1 to 34 of pUL16 were predicted with low confidence and are likely to be intrinsically disordered. Residue L490 of pUL21C is predicted to bind a hydrophobic pocket formed by L91, R362, and residues 73 to 78 of pUL16 (Fig. 4H). Loss of pUL16 binding by pUL21C L490E is consistent with the introduction of a charged glutamic acid side chain into this pocket being energetically unfavorable. pUL21 R497 is in close proximity to pUL16 residues T359 and R362. Inverting the charge of pUL21 residue 497 by substituting arginine for glutamate did not prevent pUL16 binding, presumably because the negative glutamic acid side chain could still form polar interactions with these residues. pUL21 E409 is not predicted to lie at the pUL16 binding interface, consistent with mutation of this residue not affecting the pUL21C-pUL16 interaction.
To probe the ability of the pUL21 mutant with reduced CERT binding (pUL21 V382E ) to stimulate CERT dephosphorylation during infection, a mutant strain of HSV-1 encoding pUL21 V382E was generated using two-step Red recombination (54). Dephosphorylated CERT (CERT O ) is significantly less abundant in HaCaT cells infected with HSV-1 expressing pUL21 V382E or lacking pUL21 expression (ΔpUL21) when compared with WT HSV-1-infected cells (Fig. 5, A and B). In addition to CERT, pUL21 expression reduces the phosphorylation of multiple substrates of the viral kinase pUS3 in HSV-1-infected cells (18), the phosphorylated forms of these substrates being detectable using an antibody that recognizes phosphorylated substrates of the cellular kinase Akt (55). While infection with ΔpUL21 HSV-1 causes a dramatic increase in the abundance of multiple phosphorylated pUS3 substrates, the abundance of these phosphoforms is indistinguishable between WT and pUL21 V382E HSV-1 (Fig. 5A). This confirms that the V382E substitution specifically disrupts CERT dephosphorylation, rather than generally inhibiting the ability of pUL21 V382E to recruit PP1 to substrates. Similar changes in CERT dephosphorylation, but not in the dephosphorylation of other pUS3 substrates, are observed when Vero cells are infected with pUL21 V382E HSV-1 (Fig. S6, A and B). Immunocytochemistry confirms that both WT and V382E pUL21 have the similar subcellular localization, being observed predominantly at the nuclear rim of infected Vero cells (Fig. 5C).
Metabolic labeling was used to monitor the impact of pUL21-mediated CERT dephosphorylation on sphingolipid biogenesis during infection. A pulse-chase experiment was performed where HaCaT cells infected with WT or pUL21 V382E HSV-1, or mock infected, were incubated for 5 min with alkyne-Sph at 14 h postinfection (hpi), and its metabolites were monitored for 2 h. Cer accumulates, and the rate of SM synthesis is significantly decreased in cells infected with WT and pUL21 V382E HSV-1 when compared with uninfected cells (Fig. 5D). Although a substantial decrease in the rate of SM synthesis is seen for both WT and mutant HSV-1 infection, the defect is significantly larger in cells infected with HSV-1 pUL21 V382E (Fig. 5E). The overall abundance of SM plus Cer is similar between infected and uninfected cells, suggesting that Cer to SM conversion is specifically impaired rather than influx of alkyne-Sph into the SM biogenesis pathway being defective (Fig. 5F). Taken together, this suggests that pUL21-mediated activation of CERT accelerates Cer to SM conversion during infection, albeit from a much lower base than in uninfected cells.
Having confirmed that pUL21 V382E HSV-1 specifically lacks the ability to stimulate CERT dephosphorylation, and that HSV-1 encoding pUL21 V382E has a reduced rate of Cer to SM conversion, the impact of this deficit upon virus replication and spread in cultured cells was assessed. WT and pUL21 V382E HSV-1 form similar-sized plaques on HaCaT and Vero cells (Fig. 6A), suggesting that CERT dephosphorylation is dispensable for efficient viral cell-to-cell spread. A single-step HSV-1 pUL21 modulates cellular ceramide transport growth curve, where Vero and HaCaT cells are infected at a high multiplicity of infection (MOI) and the production of infectious progeny is monitored over time, was used to compare the replication of WT and pUL21 V382E HSV-1 (Fig. 6B). Both viruses produce similar abundance of infectious progeny by 24 h postinfection. In two biologically independent experiments performed for each cell type, the kinetics of virus replication appeared to be accelerated, with higher titers of pUL21 V382E HSV-1 being observed between 6 and 12 h postinfection, but the difference in growth rate was not statistically significant for either cell line (two-way ANOVA with Sidak's multiple comparisons test).
While HSV-1 preferentially remains cell attached, spreading via direct cell-cell contacts, the cell-free secretion of virions from infected cells is altered when CERT is depleted or overexpressed (38). Cell-free release of pUL21 V382E HSV-1 was determined by quantitating the amount of infectivity secreted into the medium as a percentage of overall infectivity (secreted plus cell associated), normalized for each independent replicate to the secretion of a WT virus in control experiments performed at the same time. HaCaT cells infected for 12 h with various MOI of pUL21 V382E HSV-1 did not release significantly more or less infectivity when compared with WT HSV-1 infection (Fig. 6C). Similarly, the cell-free release of infectivity at 12 hpi did not differ between Vero cells infected with pUL21 V382E or WT HSV-1 (Fig. S6C).

Discussion
HSV-1 extensively remodels the proteome of infected cells, altering both the abundance (3) and post-translational Lysates were harvested at 16 hpi in the presence of phosphatase inhibitors and subjected to SDS-PAGE plus immunoblotting using the antibodies listed. Where indicated, the gel was supplemented with PhosTag reagent to enhance separation of CERT phosphoforms. The antibody recognizing phosphorylated Akt substrates (Akt sub) illustrates activity of the HSV-1 kinase pUS3, several substrates of which are dephosphorylated in a pUL21-dependent manner (arrowheads) (18). VP5, infection control. B, quantitation of the CERT dephosphorylation level (ratio of CERT O to total CERT) in cells infected with WT or mutant HSV-1, as determined by densitometry. Results are presented as mean ± SD from three independent experiments. Oneway ANOVA with Tukey's multiple comparisons test was used for the statistical analysis (ns, nonsignificant; **p < 0.01; ***p < 0.001). C, Vero cells were infected at MOI = 1, fixed at 14 hpi, and stained with DAPI (blue) plus an antibody recognizing pUL21 (green). D, pulse-chase experiment to measure the rate of Sph conversion to Cer, SM, and PC. HaCaT cells were infected with WT or pUL21 V382E HSV-1 at MOI = 5 or mock infected. Cells were incubated with 5 μM alkyne-Sph (pulse) at 14 hpi for 5 min and harvested for lipid extraction either immediately (0 min) or at the indicated times (chase). Extracted lipids were bioconjugated to 3-azido-7-hydroxycoumarin using click chemistry, separated by HPTLC, visualized using UV light, and relative lipid abundances were quantitated by densitometry. Data are from one representative experiment of two independent repeats. E, rate of SM synthesis expressed as its fraction in the cumulative signal for SM and Cer. F, the proportion of alkyne-Sph incorporated into either Cer or SM, representing the temporal influx of Sph into the SM biosynthesis pathway. E and F, the data represent two independent experiments (mean ± SD). Data points are labeled if significantly different to WT HSV-1: **p < 0.01; ****p < 0.0001 (two-way ANOVA with Dunnett's multiple comparisons test). G, the resolubilized proteins precipitated during lipid extraction were analyzed by SDS-PAGE and immunoblotting using the antibodies listed. Cer, ceramide; DAPI, 4 0 ,6-diamidino-2-phenylindole; hpi, hours postinfection; HPTLC, high-performance TLC; HSV-1, herpes simplex virus 1; MOI, multiplicity of infection; PC, phosphatidylcholine; SM, sphingomyelin; Sph, sphingosine. modification status (2, 18, 56) of multiple cellular proteins. We know much less about HSV-1-mediated changes to the cellular lipidome, although previous studies have identified that HSV-1 infection changes phosphoinositide levels (57) and increases the rate of de novo phospholipid synthesis (58). By combining biochemical and structural studies with cell-based models of infection, we show here that pUL21 accelerates the conversion of Cer to SM in infected cells by promoting dephosphorylation and activation of the Cer transport protein CERT (Fig. 5). This is the first observation of a viral protein directly binding CERT and altering its activity. While pUL21-mediated CERT activation causes an apparent modest acceleration in the rate of virus replication, this effect is not statistically significant and CERT activation does not contribute to replication or cell-tocell spread of HSV-1 in cultured keratinocytes or epithelial cells (Fig. 6).
Our previous work showed that pUL21 is a phosphatase adaptor with multiple different targets (18). Mutation of the pUL21 TROPPO motif, required for PP1 binding and thus stimulation of dephosphorylation, dramatically reduced the replication and spread of HSV-1. While the TROPPO motif was identified via its conservation across alphaherpesvirus sequences, the molecular basis of CERT recruitment remained unknown. Our previous immunoprecipitation experiments demonstrated that the N-terminal domain of pUL21 was dispensable for CERT binding (18). Biochemical mapping and SAXS structural characterization presented here have now identified a specific mutation, V382E, that decreases the affinity of pUL21 for miniCERT L by approximately eightfold (Fig. 4E). While this reduction in affinity is moderate when assayed in dilute biochemical solution, it is sufficient to disrupt coprecipitation of CERT from transfected cells (Fig. 4A). The V382E substitution abolishes pUL21-mediated dephosphorylation of CERT during infection (Fig. 5, A and B) and significantly decreases the rate of Cer to SM conversion (Fig. 5E). These results are consistent with other binding partners competing to bind CERT and/or pUL21 within the context of infected cells, these competing interactions amplifying the effect of reduced pUL21-CERT binding affinities. The V382E mutation does not prevent pUL21 binding to pUL16 (Fig. 4A) nor the ability of pUL21 to stimulate dephosphorylation of other targets (Fig. 5A), confirming that pUL21 binds CERT via a molecular surface that is distinct from the binding site(s) of other targets. The WT levels of virus replication and cell-tocell spread observed for pUL21 V382E HSV-1 are consistent with our previous in vitro evolution studies of HSV-1 mutants where the pUL21 TROPPO motif was mutated (18). Adaptation of the virus to the loss of pUL21 PP1 binding, via suppressor mutations that reduce the activity of the kinase pUS3, restored virus replication and spread without restoring enhanced CERT dephosphorylation (18). We therefore conclude CERT is not a critical substrate of pUL21 for virus replication and spread in cultured fibroblasts and keratinocytes. The C-terminal domain of pUL21 is highly conserved across HSV-1 and HSV-2 (>93% and 80-84% identity, respectively). V382 is absolutely conserved across these species, and a small hydrophobic residue (valine, leucine, or alanine) at this position is very highly prevalent across the genus Simplexvirus (Fig. S4A). The C-terminal domain of pUL21 is much less well conserved when considering the entire subfamily Alphaherpesvirinae (>20% sequence identity across the subfamily), with the amino acid residue at position 382 being correspondingly variable (Fig. S4B). This is consistent with our previous observation that the pUL21 homolog ORF38 from varicella-zoster virus (genus: Varicellavirus) does not bind CERT (18). It is therefore likely that CERT binding and activation is a conserved feature of HSV-1, HSV-2, and other simplexviruses but does not extend to other genera of alphaherpesvirus. Given the lack of phenotype observed for viruses encoding the CERT-binding pUL21 V382E mutant in cultured cells, it is possible that this interaction plays a role in immune regulation or transmission fitness (59). Alternatively, Figure 6. pUL21-mediated dephosphorylation of CERT does not enhance virus replication, secretion, or spread in cultured cells. A, monolayers of HaCaT or Vero cells were infected with 100 plaque-forming units of WT or pUL21 V382E HSV-1. Following infection, cells were overlaid with medium containing 0.6% carboxymethyl cellulose and incubated for 48 h, and then fixed and immunostained with chromogenic detection. Relative plaque areas (pixels) were measured using Fiji (102,103). Bars represent mean plaque sizes, which were compared using unpaired t test (n = 100; ns). B, single-step (high MOI) growth curves of WT and pUL21 V382E HSV-1. Monolayers of HaCaT (continuous line) or Vero (dotted line) cells were infected (MOI = 5) with the viruses shown. Samples were harvested at the indicated times, and titers were determined by plaque assay using Vero cells. Data are presented as mean values ± SD of technical duplicates from one representative experiment. Difference in replication kinetics across two biological replicates is not statistically significant (twoway ANOVA with Sidak's multiple comparisons test). C, virus release into the culture supernatant from HaCaT cells infected with WT or pUL21 V382E HSV-1 at various MOIs. Samples were harvested at 12 hpi and virus infectivity in the cells versus the culture medium was measured by titration on Vero cells. The fold change in secretion of infectivity into the culture medium for pUL21 V382E versus WT HSV-1 is shown as mean values ± SD of two (MOI = 1 or 3-5) or three (MOI = 10) independent experiments. Technical replicates for each independent experiment are grouped by shape. For MOI 3 to 5, the data represent one independent experiment performed at MOI = 3 (squares) and one at MOI = 5 (circles). For each MOI, the extent of virus secretion was compared using an unpaired two-tailed t test with Welch's correction (ns). HSV-1, herpes simplex virus 1; MOI, multiplicity of infection; ns, nonsignificant.
HSV-1 pUL21 modulates cellular ceramide transport the interaction may enhance virus replication in other cell types. Specifically, we note that HSV-1 is a neurotropic virus and that sphingolipids like SM are highly enriched in neurons, where they are critical for correct neuronal development and function (31). Studying whether pUL21-mediated CERT dephosphorylation affects virus replication in neurons is an important question for future investigation.
Studies of cellular lipid metabolism are complicated by robust cellular feedback mechanisms (60) and challenges in accurately detecting changes in lipid abundance (61). Here, we successfully employed lipid labeling, click chemistry, and HPTLC to monitor the kinetics of Sph metabolism, revealing that HSV-1 pUL21 significantly increases the rate of CERTmediated Cer to SM conversion both outside (Fig. 1) and within (Fig. 5) the context of infection. Furthermore, we observe a highly significant increase in the abundance of labeled Cer when cells are infected with HSV-1, this increase being greater when pUL21-mediated CERT activation has been abolished. The dramatic change in the rate of Cer to SM conversion is most likely explained by the known propensity of HSV-1 to promote dispersal of the Golgi and TGN (4,62). Such dispersal would alter ER-TGN contact sites, where CERT-mediated transport enables Cer to SM conversion by TGN-resident SM synthase (63). Side-by-side kinetic comparison of lipid metabolic enzymes including Cer synthase confirms that the presence of alkyne groups within acyl tails does not significantly alter lipid catalysis when compared with radiolabeled or natural substrates (41). While the CERT START domain has been demonstrated to bind Cer containing an alkyne group plus a photoactivatable diazirine group (64,65), we are unaware of any comparisons between CERT lipid transfer activity with natural or radiolabeled Cer versus alkyne-Cer. It is possible that the presence of an alkyne group may alter the affinity of CERT for Cer, and hence that pUL21mediated CERT activation may cause a greater or lesser change to the rate of unlabeled Cer to SM conversion. Furthermore, our experimental strategy to probe CERT activity exploited the uptake of exogenous alkyne-Sph and its conversion to alkyne-Cer via the salvage pathway mediated by Cer synthase. This contrasts to the de novo synthesis pathway, where Cer is generated from serine and palmitoyl-coA via intermediates including dihydroceramide (66). While both the de novo and salvage Cer synthesis pathways are known to occur within cells, and the efficient conversion to alkyne-Cer and alkyne-SM clearly indicated efficient salvage of alkyne-Sph in our experiments (Figs. 2, B and C and 5D), it is possible that the effect of pUL21-mediated CERT activation could differ in magnitude for Cer synthesized via the de novo pathway. Future experiments could use radiolabeled serine to study how pUL21-mediated CERT activity affects the metabolism of Cer produced via the de novo pathway (67).
HSV-1 utilizes protein kinase D-mediated trafficking from the TGN to the plasma membrane (68). Disruption of protein kinase D-mediated trafficking has been shown to alter virus secretion, with siRNA-mediated depletion of CERT increasing the secretion of extracellular (cell-free) HSV-1 particles and CERT overexpression reducing virus secretion (38). We do not observe changes in the secretion of pUL21 V382E HSV-1, which lacks the ability to dephosphorylate CERT and thus activate CERT-mediated Cer transport. While these results may appear at first glance to be contradictory, it should be noted that in our experiments the cells contained WT levels of CERT, and cells infected with either ΔpUL21 or pUL21 V382E HSV-1 have similar levels of active (CERT P ) and inactive (CERT O ) protein compared with uninfected controls (Fig. 5A). In contrast, siRNA depletion would lead to a complete absence of CERT and of CERT-mediated lipid transfer. Furthermore, the previous study demonstrated that treatment of infected cells with HPA-12 did not stimulate virus particle secretion (38). This insensitivity of HSV-1 secretion to CERT pharmacological inhibition, combined with our observation that pUL21mediated CERT activation does not alter HSV-1 secretion, strongly suggests that CERT catalytic activity is not directly linked to the regulation of virus secretion.
Our observation of labeled Cer accumulation during HSV-1 infection (Fig. 5) is consistent with an earlier report that HSV-1 infection causes an approximately twofold increase in Cer abundance within BHK-21 cells (34). Cer has many distinctive physical properties that set it apart from other membrane lipids: it has negative intrinsic curvature, it increases the order of phospholipids in membranes, and it makes biological membranes more permeable to even large solutes such as proteins (69). Nascent HSV-1 particles must traverse multiple biological membranes during virus assembly, most notably during capsid egress from the nucleus. Being too large to exit via nuclear pores, capsids leave by budding into and then out of the perinuclear space, the former step being catalyzed by the herpesvirus nuclear egress complex (NEC) (70). A recent study identified that the NEC induces lipid ordering to generate the negative curvature required for capsid budding (71). It is tempting to speculate that pUL21 regulates nuclear capsid egress by modulating Cer abundance, an effect that would be distinct from the previously demonstrated role of pUL21 in promoting nuclear egress via regulating the phosphorylation of NEC components (18,20). In addition to potentially promoting virus nuclear egress, accumulation of Cer can lead to caspase 3 activation and apoptosis via increased mitochondrial outer membrane permeabilization (72). HSV-1 is known to encode proteins that defend against Cer-induced apoptosis (73), and it is possible that pUL21-stimulated acceleration of Cer to SM conversion serves to limit the proapoptotic activity of Cer, although confirmation of this requires further study.
In the process of probing the molecular basis of pUL21 binding to CERT, we serendipitously discovered a single amino acid substitution (L490E) that prevents pUL21 binding to HSV-1 pUL16 (Fig. 4B). Both pUL16 and pUL21 are known to facilitate efficient release of virus capsids from the nuclei of infected cells (74), and a recent study suggests that the pUL16-pUL21 interaction may help regulate capsid maturation and stability (75). Our mutagenesis is consistent with previous biochemical experiments that had mapped the binding of pUL16 to the C-terminal region of pUL21 (residues 268-535) and shown that the N-terminal 40 amino acids of pUL16 were dispensable for this interaction (10). Recent advances in deep learning structure prediction enable the accurate prediction of protein structures and complexes, even in challenging cases such as multidomain viral proteins with no known sequence homologs (76). We therefore generated a model of the pUL16-pUL21C interaction using AlphaFold-Multimer (52). pUL21 residue L490 lies at the center of the predicted interaction interface, consistent with its substitution to glutamate inhibiting pUL16 binding (Fig. 4, B and H). The structure of pUL16 has not been solved experimentally, and inspection of the pUL16 structural model provides two interesting observations. First, the closest structural homolog of the predicted pUL16 C-terminal domain is the N-terminal domain of pUL21 ( Fig. 5B and Table S3). This structural correspondence (DALI Z-score 5.0; 3.6 Å rmsd across 121 C α atoms) is striking given the low sequence identity (7%) shared by the domains, although it is known that viral proteins can share similar folds despite having unidentifiable sequence similarly (77). While identifiable pUL21 homologs are restricted to alphaherpesviruses (9), homologs of pUL16 can be found across Herpesviridae (78). The similarity of the predicted pUL16 C-terminal domain structure to that experimentally observed for the pUL21 N-terminal domain suggests that the two proteins are distant homologs, having arisen via a gene duplication event in an ancestral (alpha)herpesvirus. The second observation is that several proteins with structural similarity to the predicted pUL16 N-terminal domain have been associated with the transport of lipids or other bulky hydrophobic molecules ( Fig. 5D and Table S3). For example, the VtrC protein from Vibrio parahaemolyticus was crystallized in complex with the large hydrophobic molecule taurodeoxycholate (79). Given the association between pUL21 and the lipid transport protein CERT, it is tempting to speculate that pUL16 may also have a role in modifying host or viral lipid membranes, but further biochemical and experimental structural analyses are required assess to whether the pUL16 N-terminal domain has any capacity to bind and/or transport lipids.
In summary, we demonstrated that pUL21 dephosphorylates and activates the cellular lipid transport protein CERT, stimulating conversion of Cer to SM. Characterizing the solution structure in complex with the PH and START domains of CERT allowed us to identify a single amino acid mutation of pUL21 that disrupts CERT dephosphorylation in infected cells. HSV-1 encoding this pUL21 mutant had similar replication kinetics, virus yields, and plaque sizes, confirming that dephosphorylation of other cellular and/or viral targets underpins the important role of pUL21 in HSV-1 replication and spread. The functional rationale for pUL21-mediated modulation of CERT activity by HSV-1 remains elusive, but we have defined the molecular tools that will allow its dissection in other cell types and/or animal models of infection.

Plasmids
The StrepII-CERT transient mammalian expression construct was described previously (18). For generation of stable cells, a synthetic gene encoding human CERT L (UniProt ID: Q9Y5P4-1) was cloned into a modified version of plasmid PB-T (80) that encodes an N-terminal StrepII tag and a woodchuck hepatitis virus post-transcriptional regulatory element (81) in the 3 0 untranslated region. The S132A substitution was introduced into this vector using QuikChange mutagenesis (Agilent) according to the manufacturer's instructions. CERT L destined for expression using in vitro transcription/translation system from wheat germ extract was cloned into the plasmid pF3A WG BYDV (Promega) with an N-terminal myc epitope tag. The truncations of CERT L were generated by inverse PCR (MR, residues 123-364; START L , residues 365-624) or by introduction of a stop codon (PH, residues 1-122) using QuikChange mutagenesis. For purification following bacterial expression, miniCERT L (residues 20-131 plus 351-624) was cloned into pOPTH (77), encoding an N-terminal MAH 6 tag. The generation of pUL21-H 6 and pUL21-GFP was described previously (18), and single amino acid substitutions were introduced by QuikChange mutagenesis. pUL21-H 6 was subcloned into pOPTH, and inverse PCR was used to generate H 6 -pUL21C, encoding pUL21 amino acids 275 to 535. A plasmid (UK622) encoding mouse PP1γ (UniProt ID: P63087) with an N-terminal GST tag (82) was a kind gift from David Ron (Cambridge Institute for Medical Research). To generate pUL21 V382E HSV-1, pEPkan-S containing an I-SceI/KanR selection cassette was used (54).
Doxycycline-inducible stably transfected Freestyle 293F cells expressing StrepII-CERT L and StrepII-CERT L S132A were generated using a piggyBac transposon-based system (80). A 30 ml suspension culture of Freestyle 293F cells at 1 × 10 6 cells/ml was transfected with a 5:1:1 mass ratio of PB-T-CERT L (S132A) :PB-RN:PBase (35 μg total DNA) using Freestyle MAX transfection reagent (Invitrogen) as per the manufacturer's instructions. After 2 days, the cells were transferred to fresh media supplemented with 500 μg/ml geneticin (Gibco), and the drug selection was continued for 2 weeks with media replenishment every 3 days.

GFP affinity capture
Monolayers of HEK293T cells grown in 9 cm dishes were transfected with TransIT-LT1 (Mirus) using 7.7 μg of pEGFP-N1 (for GFP alone), pUL21-GFP, or point mutants thereof, following the manufacturer's instructions. At 24 h post-transfection, the HSV-1 pUL21 modulates cellular ceramide transport cells were harvested by scraping into the medium, pelleted (220g, 5 min, 4 C), washed three times with cold PBS, and lysed at 4 C in 1 ml lysis buffer (10 mM Tris [pH 7.5], 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40, and 1:100 diluted EDTA-free protease inhibitor cocktail [Merck]) for 45 min before clarification (21,000g, 10 min, 4 C). After immunoprecipitation with GFP-Trap beads (ChromoTek) performed in accordance with the manufacturer's protocol, the samples were eluted by incubation at 95 C for 5 min in 45 μl 2× SDS-PAGE loading buffer. Input and bound samples were separated by SDS-PAGE and analyzed by immunoblot.

GST pulldown
GST pull-down experiments were carried out in 96-well flat-bottomed plates (Greiner) using magnetic glutathione beads (Thermo Fisher Scientific) to capture the GST-tagged pUL21 (bait protein). First, 0.5 nmol of purified bait protein was incubated with the beads at 4 C for 30 min. The beads were washed three times with wash buffer (20 mM Tris [pH 8.5], 200 mM NaCl, 0.1% NP-40, 1 mM DTT, and 1 mM EDTA) and then incubated at 4 C for 60 min with various truncations of myc-CERT L (prey proteins), expressed using the TNT SP6 High Yield Wheat Germ in vitro transcription/ translation system (Promega) in accordance with the manufacturer's protocol, followed by four washes in wash buffer. Protein was eluted using wash buffer supplemented with 50 mM reduced glutathione before being analyzed by SDS-PAGE and immunoblotting.

Recombinant protein purification following bacterial expression
All recombinant proteins were expressed in Escherichia coli T7 Express lysY/I q cells (New England Biolabs). Except for GST-PP1γ, cells were grown in 2× TY medium at 37 C to an absorbance at 600 nm of 0.8 to 1.2 before cooling to 22 C and inducing protein expression by addition of 0.4 mM IPTG. At 16 to 20 h postinduction, cells were harvested by centrifugation, and pellets were stored at −70 C until required. For GST-PP1γ, the 2× TY medium was supplemented with 1 mM MnCl 2 , and the cultures were cooled to 18 C upon reaching an absorbance of 0.8 at 600 nm, followed by induction using 1 mM IPTG. For all recombinant proteins, cells were resuspended in lysis buffer (see later) at 4 C before lysis using a TS series cell disruptor (Constant Systems) at 24 kpsi. Lysates were cleared by centrifugation (40,000g, 30 min, 4 C) and incubated with the relevant affinity resins for 1 h at 4 C before extensive washing (≥20 column volumes) and elution using the relevant elution buffer (see later). Samples were concentrated and subjected to SEC (see later). Fractions containing the desired protein as assessed by SDS-PAGE were pooled, concentrated, snap-frozen in liquid nitrogen, and stored at −70 C.

Mutagenesis of viral genomes and generation of recombinant HSV-1
All HSV-1 strain KOS viruses used in this study were reconstituted from a bacterial artificial chromosome (BAC) (87), and the mutated strain was generated using the two-step Red recombination method (54) with the following primers: Forward: 5 0 -CGGCTCGTAGGCCGGTACACACAGCGCC ACGGCCTGTACGAACCTCGGCCCGACGACCCAGTAG GATGACGACGATAAGTAGGG.
The generation of pUL21 deletion mutant (ΔpUL21) was described previously (18). To generate the P0 stocks, Vero cells were transfected with the recombinant BAC DNA together with pGS403 encoding Cre recombinase (to excise the BAC cassette) using TransIt-LT1 (Mirus) following the manufacturer's instructions. After 3 days, the cells were scraped into the media, sonicated at 50% power for 30 s in a cup-horn sonicator (Branson), and titrated on Vero cell monolayers. The subsequent stocks were generated by infecting either Vero (HSV-1 WT) or HaCaT pUL21 cells (HSV-1 mutants) at an MOI of 0.01 for 3 days. The cells were then scraped and isolated by centrifugation at 1000g for 5 min. Pellets were resuspended in 1 ml of complete DMEM supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, and freeze/ thawed thrice at −70 C before being aliquoted, titered on Vero cell monolayers, and stored at −70 C until required. The presence of the desired mutation in the reconstituted virus genomes was confirmed by sequencing the pUL21 gene.

Metabolic labeling and lipid extraction for TLC
Metabolic labeling was performed using subconfluent (60-80% confluence) HaCaT or HaCaT21 cells grown in a 6-well plate. For analysis of stable cell lines, the cells were pretreated for 30 min with complete DMEM containing 1 μM N-(3-hydroxy-1-hydroxymethyl-3-phenylpropyl)dodecanamide (HPA-12; Tokyo Chemical Industry) dissolved in 0.1% dimethyl sulfoxide (DMSO; Merck), or 0.1% DMSO alone, and HPA-12 or DMSO was retained at the same concentrations throughout the subsequent incubation steps. For infection, cells were infected (see later) 14 h before metabolic labeling.
For metabolic labeling, cells were washed twice with warm PBS before incubation in 500 μl prewarmed DMEM with 1% (v/v) Nutridoma (Merck) supplemented with 5 μM alkyne-Sph (Cayman Chemical) for 5 min (pulse). Next, cells were washed twice with warm PBS, and 1 ml of prewarmed DMEM with 1% (v/v) Nutridoma was added to each well, followed by incubation at 37 C for the indicated times of chase. Alkyne-Sph was stored as a 3.3 mM ethanolic stock solution at −20 C.
At the indicated times of chase, the plate with cells was transferred onto the ice, washed twice with 1 ml ice-cold PBS, scraped into 300 μl of ice-cold PBS, and transferred into appropriate 1.5 ml microcentrifuge tubes containing 600 μl of methanol. To each tube, 150 μl of chloroform was added, followed by vigorous vortexing. The precipitated protein was pelleted (20,000g, 1 min) at room temperature (RT). The organic supernatant was transferred to separate 2 ml tubes containing 300 μl chloroform. About 600 μl of 0.1% acetic acid in water was subsequently added to each tube to induce formation of two phases. Following extensive vortexing, the phases were separated by centrifugation (20,000g, 1 min, RT), and the lower phase was transferred to a new 1.5 ml microcentrifuge tube. This lipid-containing solvent phase was dried in a UniVapo centrifugal vacuum concentrator (UniEquip) at 30 C for 20 min.

Virus infections
Monolayers of indicated cells were infected by overlaying with the appropriate viruses and diluted in complete DMEM to the specified MOI. The time of addition was designated 0 hpi. Cells were incubated with the inoculum at 37 C in a humidified 5% CO 2 atmosphere for 1 h, rocking the tissue culture plate housing the monolayers every 15 min, before complete DMEM supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin was added to dilute the original infection medium fivefold. Infected cells were incubated at 37 C in a humidified 5% CO 2 atmosphere and harvested at the specified times.

Immunoblotting
For analysis of protein expression and phosphorylation, cells were washed two times with ice-cold 50 mM Tris (pH 8.5), 150 mM NaCl, and scraped into 100 μl of ice-cold lysis buffer (50 mM Tris [pH 8.5], 150 mM NaCl, 1% [v/v] Triton X-100, 1% [v/v] EDTA-free protease inhibitor cocktail [Merck], 10 mM tetrasodium pyrophosphate, 100 mM NaF, and 17.5 mM β-glycerophosphate). After 30 min of incubation on ice, the lysates were sonicated (two 15 s pulses at 50% power) in a cup-horn sonicator followed by centrifugation at 20,000g for 10 min. For all samples, protein concentrations were determined using a bicinchoninic assay (Pierce), and normalized protein lysates were analyzed by SDS-PAGE. For enhanced separation of phosphorylated proteins, 7% (w/v) acrylamide gels contained 50 μM MnCl 2 and 25 μM PhosTag reagent (Wako) where indicated. Separated proteins were then transferred to nitrocellulose membranes using the Mini-PROTEAN system (Bio-Rad) and analyzed by immunoblotting and signals being detected using Odyssey CLx Imaging System (LI-COR). For quantitation of CERT phosphoform relative abundance, the signal for CERT O band alone and for all CERT bands (CERT O + CERT P ) were quantitated using Image Studio Lite software with local background subtraction. Statistical tests were performed using Prism 7.
For SAXS in batch mode, scattering data (I(s) versus s, where s = 4πsinθ/λ nm −1 , 2θ is the scattering angle, and λ is the X-ray wavelength, 0.124 nm) were collected from the indicated batch samples and a corresponding solvent blank (20 mM Tris [pH 8.5], 500 mM NaCl, 3% [v/v] glycerol, and 1 mM DTT). The scattering profiles of H 6 -miniCERT L (1.73, 2.59, 3.46, and 6.91 mg/ml) and H 6 -pUL21C (0.76 mg/ml) were measured in continuous-flow mode using an automated sample changer (30 μl sample at 20 C; 1 mm path length). The sample and buffer measurements were measured as 100 ms data frames on a Pilatus 6M area detector (Dectris) for total exposure times of 2.1 and 2.6 s, respectively. For H 6 -miniCERT L , the resulting SAXS curves were extrapolated to infinite dilution.
For SEC-SAXS experiments, the scattering data were recorded using a Pilatus 6M detector with 1 s sample exposure times for a total of 3600 data frames spanning the entire course of the SEC separation. Proteins were mixed at 1:1 molar ratio, incubated on ice for 15 min, and then concentrated using a centrifugal concentrator to the desired concentration as estimated from absorbance using extinction coefficients (90) calculated assuming a 1:1 molar ratio. About 40 μl of purified H 6 -pUL21C-H 6 -miniCERT L (4.3 mg/ml) was injected at 0.35 ml/min onto a Superdex 200 Increase 5/150 GL column (Cytiva) equilibrated in 20 mM Tris (pH 8.5), 500 mM NaCl, 3% (v/v) glycerol, and 1 mM DTT. SAXS data were recorded from a single peak (frames 1245-1296 s), and solvent blank was collected from pre-elution frames (frames 247-411). Primary data reduction was carried out using CHROMIXS (91), and 2D-to-1D radial averaging was performed using the SASFLOW pipeline (92). Processing and analysis of the SAXS output was performed using the ATSAS 3.0.2 software package (93). The extrapolated forward scattering intensity at zero angle, I(0), and the radius of gyration, R g , were calculated from the Guinier approximation (lnI(s) versus s 2 , for sR g < 1.3). The maximum particle dimension, D max , was estimated based on the probable distribution of real-space distances, p(r), which was calculated using GNOM (94). A concentrationindependent estimate of molecular weight was determined using a Bayesian consensus method (95). All structural parameters are reported in Table S2.
Ab initio modeling was performed using GASBOR (49), and figures show the models that best fit their corresponding SAXS profiles (lowest χ 2 ). Pseudoatomic modeling was performed using CORAL (96). For H 6 -miniCERT L , crystal structures of the CERT PH (Protein Data Bank [PDB] ID: 4HHV) (24) and START (PDB ID: 2E3M) (25) domains were used, the latter including predicted secondary structure as modeled by I-TASSER (97) for intron 11 (residues 371-396), present only in CERT L . The miniCERT L 23 amino acid linker and the purification tag (MAH 6 ) were modeled using CORAL. The model with the best fit was superimposed onto the PH:START crystal structure (PDB ID: 5JJD) (46) by aligning the START domains using PyMOL, version 2.5.2 (Schrödinger). For H 6 -pUL21C, the N-terminal (MAH 6 QD) and C-terminal (HGQSV) elements missing from the crystal structure (PDB ID: 5ED7) (51) were modeled using CORAL. For H 6 -pUL21C-H 6 -miniCERT L , the heterodimer was modeled using the SAXS curve of the complex using a fixed conformation of H 6 -pUL21C and mobile CERT domains, with unstructured elements modeled as aforementioned. All the models were assessed for fit to the corresponding SAXS profiles (χ 2 and CorMap) using CRYSOL (98). Molecular images were generated using PyMOL.
Sequence analysis, deep learning structure prediction, and structural analysis Sequences of the C-terminal domain from pUL21 homologs were identified by querying the National Ceneter for Biotechnology Information nonredundant protein sequence database (99) with HSV-1 pUL21 (UniProt ID: F8RG07) residues 275 to 535 using BLASTP (100) for simplexviruses (taxid: 10294) and PSI-BLAST with two iterations for alphaherpesviruses (taxid: 10294). Truncated sequences (accession length <80% of HSV-1 pUL21) or those containing ambiguous amino acids ("X") were manually removed. Sequences were aligned using COBALT (101) and are provided as supporting information (Data S1).

Differential scanning fluorimetry
Differential scanning fluorimetry experiments were carried out using Viia7 real-time PCR system (Applied Biosystems) and 1× Protein Thermal Shift dye (Applied Biosystems). The assay buffer (20 mM Tris [pH 8.5], 500 mM NaCl, and 1 mM DTT) was mixed with dye stock solution and protein solution (H 6 -pUL21 or H 6 -pUL21 V382E ) in an 8:1:1 ratio, giving 1 ng protein in a final volume of 20 μl, and measurements were performed in triplicate. Samples were heated from 25 to 95 C at 1 per 20 s, and fluorescence was monitored at each increment. The melting temperature (T m ) is the inflection point of the sigmoidal melting curve, determined by nonlinear curve fitting to the Boltzmann equation using Prism 7.

In vitro dephosphorylation assays
Dephosphorylation assays were performed upon 0.5 μM of purified CERT P using a fixed concentration of GST-PP1γ (25 nM) in the presence of H 6-pUL21 or H 6 -pUL21 V382E at the indicated concentrations. Reactions (50 μl) proceeded in assay buffer (150 mM NaCl, 20 mM Tris [pH 8.5], 0.1% Tween-20, and 1 mM MnCl 2 ) for 30 min at 30 C before being stopped by the addition of 50 μl of 2× SDS-PAGE loading buffer and boiling at 95 C for 5 min. Samples were analyzed by SDS-PAGE using 7% (w/v) acrylamide gels supplemented with 25 μM PhosTag reagent and 50 μM MnCl 2 , and the protein was visualized using InstantBlue Coomassie stain (Expedeon). To measure the ratio of CERT O to total CERT, Coomassiestained gels were scanned using an Odyssey CLx Imaging System (LI-COR). The signals detected in the 700 nm channel for the CERT O band alone and for all CERT bands (CERT O + CERT P ) were quantitated using Image Studio Lite software with local background subtraction.

Plaque size analysis
Confluent monolayers of Vero or HaCaT cells in 6-well tissue culture plates were infected with 100 plaque-forming units of the indicated virus diluted in complete DMEM to a final volume of 500 μl. Following the adsorption for 1 h at 37 C in a humidified 5% CO 2 atmosphere, rocking the plate every 15 min, the cells were overlaid with plaque assay media (DMEM supplemented with 0.3% high viscosity carboxymethyl cellulose, 0.3% low viscosity carboxymethyl cellulose, 2% [v/v] fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin) and incubated for further 48 h. Next, cells were fixed with 3.7% (v/v) formal saline for 20 min, washed three times with PBS, and incubated for 1 h with mouse anti-gD (LP2) (86), diluted 1:50 in blocking buffer (PBS supplemented with 1% [w/v] bovine serum albumin and 0.1% Tween-20). Cells were washed thrice in PBS and incubated for 30 min with horseradish peroxidase-conjugated rabbit anti-mouse antibody (DaKo; catalog no.: P0161) diluted 1:5000 in blocking buffer, followed by two subsequent PBS washes and one wash with ultrapure water. Plaques were visualized using the TrueBlue peroxidase substrate in accordance with the manufacturer's instructions (Seracare). Plaques were scanned at 1200 dpi, and plaque areas were measured using Fiji (102,103).

Single-step (high MOI) virus growth assays
Virus growth assays were performed in technical duplicate for each independent experiment. Monolayers of HaCaT or Vero cells were infected with the indicated viruses diluted in complete DMEM to an MOI of 10. The time of virus addition was designated 0 hpi. After adsorption for 1 h at 37 C in a humidified 5% CO 2 atmosphere, rocking the plates every 15 min, extracellular virus particles were neutralized with an acid wash (40 mM citric acid, 135 mM NaCl, 10 mM KCl, pH 3.0) for 1 min, and cells were then washed three times with PBS before being subsequently overlaid with complete DMEM. At designated times postinfection, cells were harvested by freezing the plate at −70 C. When all plates were frozen, samples were freeze/thawed one more time before scraping and transferring to 1.5 ml tubes and storing at −70 C until titration. Titrations were performed on monolayers of Vero cells. Serial dilutions of the samples were used to inoculate the cells for 1 h, followed by overlaying with DMEM containing 0.3% high-viscosity carboxymethyl cellulose, 0.3% lowviscosity carboxymethyl cellulose, 2% (v/v) fetal bovine seum, 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. Three days later, cells were fixed in 3.7% (v/v) formal saline for 20 min, washed with water, and stained with 0.1% toluidine blue. Statistical tests were performed using Prism 7.

Virus release assays
Monolayers of HaCaT or Vero cells were infected as described previously for single-step virus growth assays, infections being performed in technical duplicate or triplicate for each independent experiment. At 12 hpi, the media (500 μl) were harvested to 1.5 ml Eppendorf tubes, centrifuged for 5 min at 1000g to remove any detached cells, and 300 μl of the supernatant was carefully transferred to fresh tubes and stored at −70 C until titration. The cells were overlaid with 500 μl of fresh DMEM, and the plates were immediately frozen at −70 C. Titrations were performed on monolayers of Vero cells as described previously. Statistical tests were performed using Prism 7. Figure 6C was generated using Super-PlotsOfData (104).

Immunocytochemistry
Cells grown on #1.5 coverslips were infected at an MOI of 1 as described previously. At 14 hpi, cells were washed with PBS and incubated with freezing cold (−20 C) methanol for 5 min at −20 C. Coverslips were washed with PBS, followed by incubation with blocking buffer (1% [w/v] bovine serum albumin in PBS) for 30 min at RT. Primary antibodies (aforementioned) were diluted in blocking buffer and incubated with coverslips for 1 h. Coverslips were washed 10 times with PBS before incubation for 45 min with the secondary antibodies (aforementioned) diluted in blocking buffer. Coverslips were washed 10 times in PBS and 10 times in ultrapure water before mounting on slides using Mowiol 4-88 (Merck) containing 200 nM 4 0 ,6-diamidino-2-phenylindole. Images were acquired using an inverted Olympus IX81 widefield microscope with a 60× Plan Apochromat N oil objective (numerical aperture of 1.42) (Olympus) and Retiga EXi Fast1394 interline CCD camera (QImaging).

Data availability
The SAXS data measured for each concentration, with an accompanying report, are available in the Small Angle Scattering Biological Data Bank (SASBDB) (105), entries SASDNB7, SASDNC7, and SASDND7. The AlphaFold-Multimer model of the HSV-1 pUL16-pUL21C complex, with associated raw output and quality statistics, has been deposited in the University of Cambridge Data Repository (https://doi.org/10.17863/CAM.88729).