Generation of a glycogene knock out library in HaCaT keratinocytes

HaCaT is a human keratinocyte cell line capable of forming a stratified squamous epithelium, and thus allows evaluating the infection of the skin tropic HSV-1 in both cell and organotypic tissue culture. In order to address the role of specific glycan structures in the HSV-1 infectious cycle, we used precise gene editing to target glycosyltransferases involved in the early steps of core structure synthesis, and in major elongation and branching steps of the main glycosylation pathways, including N-linked glycosylation, mucin type O-linked glycosylation, as well as GSL and GAG synthesis (Fig.1a, Supplementary Table1).

For N-linked glycans we generated MGAT1, MGAT4A, MGAT4B, MGAT5, and MGAT5+4B knock outs (KO). MGAT1 adds the first N-acetylglucosamine to the C-2 of core 3-linked mannose, and lack of this enzyme results in elimination of all N-glycan maturation steps, yielding high-mannose type N-glycans as confirmed by MS-glycoprofiling (Fig.1a, Supplementary Fig.1). MGAT4A, MGAT4B and MGAT5 are responsible for N-glycan branching, where MGAT4A or MGAT4B initiate a 4-linked antenna on the 3-linked mannose, and MGAT5 performs 6-linked branching from the core 6-linked mannose. Lack of MGAT5 results in loss of tetra-antennary N-glycans, and loss of MGAT4 isoforms also strongly diminishes the content of tetra-antennary N-glycans (Supplementary Fig.1). In addition, KO of each of the three branching enzymes resulted in increased relative abundance of hybrid type N-glycans, whereas double KO of MGAT5 and MGAT4B increased the relative abundance of biantennary glycans (Supplementary Fig.1).

For mucin type O-linked glycans, we knocked out core 1 synthase (C1GALT1), its obligate chaperone COSMC (C1GALT1C1), core 2 synthase (GCNT1), as well as the major core 1-capping glycosyltransferase ST3GAL1. Loss of C1GALT1 or COSMC eliminates the 3-linked galactose (core 1 structure), results in truncation of O-linked glycans to the initiating -GalNAc, and prevents assembly on secreted -benzyl GalNAc precursor used in CORA O-glycoprofiling (Fig.1a, Supplementary Fig.2). GCNT1 is the predominant enzyme creating the branched core 2 structure by addition of 6-linked GlcNAc to the GalNAc. Loss of GCNT1 nearly abolished all the disialylated core 2 structures, though some structures matching the composition of monosialylated core 2 could still be detected. Such structures cannot be discriminated from isobaric core 1 structures, and a minor contribution from other GCNTs to core 2 synthesis cannot be excluded either. Finally, loss of ST3GAL1 significantly reduces the 3-linked sialic acid content and results in predominantly non-capped core 1 structures (Supplementary Fig.2). We also targeted the synthesis of GSLs and GAGs. Through KO of B4GALT5 or ST3GAL5, we generated cells with the truncated GSLs, glucosylceramide (GlcCer) and lactosylceramide (LacCer), respectively (Fig.1a). Furthermore, we knocked out B4GALT7, which adds a 4-linked galactose to the initiating xylose in GAG biosynthesis, effectively truncating all classes of GAGs on membrane proteoglycans (Fig.1a). The generated keratinocyte library represents a screening platform for roles of defined cell-surface presented glycan structures in HSV-1 biology in the context of natural infection.

To define the capacity of HSV-1 to complete the infectious cycle in glycoengineered keratinocytes, we infected confluent monolayers of the KO cell lines with HSV-1, and quantified HSV-1 DNA and infectious particles in the growth media at 17h post infection (hpi) by qPCR and plaque titration, respectively. As a measure for viral replication fitness, we calculated the ratio of genome copies/infectious particles for each KO. When infecting cells with truncated O-glycans (C1GALT1C1 KO) a decrease in viral titers was detected (Fig.1b, e). In contrast, the same cells generated close to normal levels of viral DNA (Fig.1c, f), suggesting decreased fitness of virions lacking elongated O-glycans (Fig.1d, g). This feature was unique to complete truncation, and not seen when eliminating branching or sialylation of O-glycans. In cells lacking N-glycan maturation (MGAT1 KO) we also found a lower number of infectious particles (Fig.1b, e) with an apparent decreased fitness as indicated by an increase in the ratio of DNA/infectious particles (Fig.1d, g). This apparent decrease in fitness was not detected in cells with loss of N-glycan branching, and in MGAT4A KO cells we even observed an overall increased viral output (Fig.1b, c). When analysing cells with GSL synthesis defects, we found that lack of LacCer sialylation (ST3GAL5 KO) accelerated virus production (Fig.1b, c, e, f), but without any change in viral fitness (Fig.1d, 1g). Finally, loss of cellular GAGs increased the production of viral particles (Fig.1c). In conclusion, most of the tested glycogene disruptions permitted HSV-1 replication, and only disruption of N- or O-glycan maturation impaired viral fitness. We next evaluated the impact of defined glycan classes to distinct stages of the HSV-1 infectious cycle, including binding and entry, viral assembly and properties of progeny virus, and cell-to-cell spread.

HSV-1 binds and enters human keratinocytes very rapidly, with around 30% of virions bound after 20min on ice, and 80% after 2h28. Most of the bound virions enter keratinocytes within 5min after warming28. Perturbations in each of the investigated glycosylation pathways modulated early virus-host interactions (Fig.2ae). Diminished core 2 O-glycan branching resulted in increased binding also reflected in subsequent entry experiments (Fig.2b, c). Lack of complex N-glycans and reduced 4-antenna branching (MGAT1 KO and MGAT4B KO) showed reduced binding, again also reflected in the entry experiments (Fig.2ac). Interestingly, deletion of MGAT4A, another isoform catalyzing the 4-antenna synthesis on N-glycans, likely on another subset of proteins or sites in proteins29,30, selectively affected viral entry (Fig.2b, c, e). Cells displaying truncated glycolipids showed a reduction in binding to around 50% of that of WT (Fig.2a, b). A similar effect was observed in both B4GALT5 and ST3GAL5 KOs, controlling the consecutive steps in the biosynthesis of the GSL GM3. In addition, an incremental reduction in entry was observed for B4GALT5 KO cells, suggesting involvement of glycolipids in both viral binding and entry to host cells (Fig.2c, e).

HSV-1 binding (20min (a) or 120min (b) on ice) and entry (5min at 37C after 120min on ice (c)) to KO cell lines. Data is shown as WT-normalized mean+SEM of 3 independent experiments for each KO cell line. Two-way ANOVA followed by Dunnetts multiple comparison test was used on raw data to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Proportion of virus bound at 20min compared to 120min (d) or proportion of virus entered at 5min compared to virus bound at 120min (e) is shown as mean+SEM of 3 independent experiments for each KO cell line from a total of 15 experiments. One-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). f HaCaT WT and B4GALT7 KO cells were probed for HSV-1 gC binding. Enzymatic treatments were included to evaluate contributions of HS and CS GAG chains. Representative of 2 independent experiments. g A simplified overview of GAG synthe sis. h HSV-1 gC binding to a panel of CHO KO cell lines. Data is shown as WT-normalized geometrical means of 3 independent experiments for each KO cell line, and the bar heights indicate mean+SEM. One sample t test was used to evaluate differences from 1. FDR at 5% was controlled by two-stage step-up method of Benjamini, Krieger and Yekutieli (*q<0.05, **q<0.01, ***q<0.001, ****q<0.0001). i Nectin 1 and HVEM surface expression. Data points show background subtracted median fluorescence intensity (MFI) from two independent experiments, and the bar heights indicate the mean. j Percentages of total quantified CS disaccharides in HaCaT WT (Supplementary Table2). k Percentages of total quantified HS disaccharides in HaCaT WT (Supplementary Table3). Source data are provided as a Source Data file for all graphs.

Then, we analysed cells impaired in GAG biosynthesis and found an almost complete loss of binding to cells presenting only the initiating xylose on proteoglycans (B4GALT7 KO). Although we still lack a clear understanding of finer structural requirements of GAGs presented on their core proteins in the context of total cellular glycome, this fits well with the known importance of heparan sulfate (HS) in the initial attachment of HSV-1 (Fig.2a, b)26. To further dissect the importance of GAG binding determinants we investigated the binding of recombinant HSV-1 gC to our HaCaT KO cells. The use of recombinant HSV-1 gC limited the interactions to a single viral protein known to bind to synthetic GAGs in vitro, similarly to intact HSV-127,31,32. As expected, no binding was detected on B4GALT7 KO cells (Fig.2f), and to further confirm the selectivity for HS we treated HaCaT WT cells with heparinases 1, 2, and 3. Loss of HS completely abolished gC binding suggesting minimal interaction with chondroitin sulfate (CS) or dermatan sulfate (DS) presented on the cell surface (Fig.2f). Minimal changes in cell staining for bound HSV-1 gC after chondroitinase ABC treatment further supported this interpretation. Next, we analyzed a library of glycoengineered CHO cells delineating the GAG biosynthesis pathways (Fig.2g) and quantified gC binding by flow cytometry (Fig.2h)33. This library included selective elimination of HS or CS (Extl2+Extl3 KO and Csgalnact1 KO+Csgalnact2 KO+Chsy1 KO, respectively), reduction in chain polymerization of HS or CS (Ext1 KO+Ext2 KO and Chpf KO, respectively), elimination of HS N-sulfation, also effectively diminishing follow-up O-sulfation (Ndst1 KO+Ndst2 KO), as well as elimination of 4-O sulfation of CS and DS units of CS chains (Chst11 KO+Chst12 KO+Chst13 KO+Chst14 KO). In addition, we used B4galt7 KO and B3gat3 KO cells to truncate all GAGs to the initiating xylose and a short linker trisaccharide, respectively (Fig.2g). In agreement with the HaCaT cell staining data, manipulation of HS synthesis and chain length substantially decreased gC binding showing that the interaction was entirely dependent on HS sulfation and not compensated by the presence of CS (Fig.2h). Accordingly, manipulation of CS synthesis only slightly decreased gC binding independent of the predominant 4-O sulfation (Fig.2h). As expected, truncation to the linker also eliminated gC binding (Fig.2h). To our surprise, some binding was retained upon complete GAG truncation, possibly representing unspecific binding due to gross changes in the glycocalyx. In conclusion, by using cell surface presented GAGs, we were able to identify sulfated HS as the major contributor to HSV-1 gC binding and show that CS sulfation is not necessary for interaction with CS, at least in the presence of HS. More generally, the binding and entry assays show that perturbations in the cellular glycome landscape have immediate effects to early virus-cell interactions, which can be further dissected as demonstrated for the interaction between gC and HS.

To follow up on our binding and entry data, we aimed to investigate the cellular landscape of HSV-1 entry receptors and other surface molecules that may have an impact on the early virus-cell interactions in the different knock out cells. We first quantified the surface expression levels of Nectin 1 and HVEM in WT HaCaT cells and found very low levels of the latter (Fig.2i, Supplementary Fig.3a, b). MGAT1 KO and B4GALT7 KO cells expressed significantly lower levels of Nectin 1 on the cell surface, whereas MGAT4 KO and GCNT1 KO expressed higher levels (Fig.2i). These results correlate well with the virus binding data, and may help explain the altered proportion of virus bound to cells with alterations in N-glycosylation and O-glycosylation pathways. Importantly, the selective effect on entry to MGAT4 KO was not correlated to availability of the receptor.

For B4GALT7 KO, Nectin 1 presentation decreased by approximately 60%, but this does not explain the complete loss of HSV-1 binding, which is likely a combination of a decrease in GAG and protein receptors. While gC mediates early virus-GAG interactions, facilitating subsequent interactions between gD and its cognate protein entry receptors, 3-O-sulfated HS has also been identified as an independent entry receptor for gD34,35. In order to evaluate the potential contribution of 3-O-sulfated HS to HSV-1 entry in skin cells, we performed disaccharide analysis of HaCaT WT and B4GALT7 KO cells, using our recently developed method, which allows detection of 3-O-sulfated HS36 (Fig.2 j, k, Supplementary Fig.4, Supplementary Table2 and 3). Except for hyaluronan, which is synthesized by a distinct family of enzymes, we did not detect any CS or HS disaccharides in B4GALT7 KO cells (Supplementary Fig.4). HaCaT WT cells expressed high levels of 4-O-sulfated or 6-O-sulfated CS, hyaluronan, as well as N-sulfated, N-/2-O-sulfated, N-/2-O/6-O-sulfated, and non-sulfated HS. We detected very low levels of 3-O-sulfated HS disaccharides, demonstrating that usage of these receptors for HSV-1 entry in human keratinocytes is limited. We therefore suggest that Nectin 1 is the most widely available HSV-1 entry receptor for gD in HaCaT keratinocytes.

The disaccharide expression profiles in skin cells provided additional insight into the gC binding data on the CHO cell library. Namely, N-sulfated GAG motifs required for gC binding to CHO cells were abundantly found on human keratinocytes, and likely play a significant role in vivo. On the contrary, 4-O-sulfated CS, abundantly found on skin cells, is unlikely to be a critical receptor for gC, as seen from CHO data.

We next looked into GSLs expressed in skin cells (Fig.3). We saw comparable levels of Nectin 1 on the surface of WT, B4GALT5 KO and ST3GAL5 KO cells (Fig.2i), and yet HSV-1 binding and entry to these cells was markedly decreased. We thus hypothesized that elongated GSLs may help deliver the viral entry receptors to membrane compartments accessible to incoming virus. We used antibodies and toxins recognizing various (glyco)lipid structures to illuminate their distribution in keratinocytes (Fig.3a). Ceramide and glucosylceramide, representing initial steps of GSL synthesis, were predominantly located intracellularly in WT cells, while some ceramide accumulation could be seen in B4GALT5 KO, devoid of elaborate GSLs (Fig.3b). Interestingly, expression of more complex GSLs was heterogeneous, and different cells appeared committed to a specific GSL subtype. Specifically, we detected Gb3 structures, synthesized from lactosylceramide precursor, in both WT, and ST3GAL5 KO cells with clear surface presentation, but not B4GALT
5 KO (Fig.3b, e). In contrast, GM3, the product of ST3GAL5, was only detected in WT cells (Fig.3b). GM3 partially co-localized with intracellular glucosylceramide-positive structures but were primarily expressed on the cell membrane (Fig.3c). Importantly, GM3 was abundantly found on apical cell surfaces accessible to the extracellular environment (Fig.3d). Gb3 and GM3 were expressed in mostly distinct subsets of cells, and a substantial proportion of skin cells remained unlabeled, presumably expressing more elaborate structures (Fig.3e). In conclusion, we show heterogeneous yet regulated expression of different GSLs in distinct cells and within different cellular compartments, which may be relevant for interaction with extracellular virus.

a The cartoon depicts a simplified human glycosphingolipid biosynthetic pathway. Glycolipid structures highlighted in magenta were probed by antibodies or fluorescently labeled toxins. be Cells grown on cover slips were fixed with 4% PFA and stained for different GSL structures. b Confocal micrographs show distribution of different GSLs in HaCaT WT, B4GALT5 KO and ST3GAL5 KO monolayers. Representative of two independent experiments, scale bars are indicated for each set of micrographs. c z-stack maximal intensity projection of HaCaT WT cells labeled with anti-GlcCer and anti-GM3 antibodies. Representative of 2 independent experiments, scale bar is indicated. d HaCaT WT cells labeled with anti-GM3 antibody. An individual z-slice within a stack is shown, with orthogonal cross sections of the z-volume included, and indicate apical expression of GM3. Nuclei are labeled with DAPI (blue). Representative of 2 independent experiments, scale bar is indicated. e The confocal micrograph shows spatially distinct distribution of Gb3 and GM3 GSLs in HaCaT WT, probed by FITC-labeled Shiga toxin B (StxB-FITC), and anti-GM3 antibody, respectively. Representative of 2 independent experiments, scale bar is indicated.

We next investigated late stages of viral replication in KO cells with changes in protein glycosylation capacity and altered viral propagation dynamics. We probed the expression of gD and gB that promote virion envelopment. In WT most of gD signal was confined to the cell surface, partially overlapping with E-cadherin (Fig.4a), while gB primarily localized to the perinuclear compartment and secondary envelopment sites with some surface presentation, consistent with the literature (Fig.4b)37. In contrast, C1GALT1C1 KO, C1GALT1 KO, and MGAT1 KO cells exhibited a weaker and more dispersed gD immunostaining pattern with partial cytoplasmic accumulation suggesting issues with envelope glycoprotein trafficking (Fig.4a). In addition, gB exhibited poorer surface and perinuclear localization and presented in large clusters within the cells (Fig.4b). ST3GAL1 KO cells, which did not exhibit defects in viral propagation dynamics, displayed similar gB staining as WT (Fig.4b), while exhibiting stronger gD signal (Fig.4a). Overall, the results suggest that lack of core 1 O-glycans or mature N-glycans causes defects in viral particle formation due to suboptimal incorporation of viral proteins, which would fit with the observed diminished titers or loss of fitness. In addition, using an HSV-1 strain with GFP-labeled capsid protein VP26 allowed us to observe differences in the localization of viral capsids. The capsids were found in nuclear assembly compartments, outer nuclear rim, and transitioning through the cytosol in WT and ST3GAL1 KO cells. We observed lower numbers of capsid assembly sites in the nucleus and rare association with the outer nuclear rim in C1GALT1C1 KO, C1GALT1 KO and MGAT1 KO cells, with the most pronounced effect in C1GALT1C1 KO (Fig.4a, b).

a HaCaT cells grown on cover slips were infected with MOI10 of HSV-1 K26-GFP and fixed and permeabilized at 14hpi followed by co-staining for HSV-1 gD (magenta) and E-cadherin (cyan). Histograms on the left indicate intensities of gD and E-cadherin signals across the confocal images (marked with black arrowheads). Pixel overlap from the two channels is shown in white. GFP labeled capsid proteins (VP26) are seen in green. Nuclei were stained with DAPI (blue). Stainings of mock-infected cells are included. Scale bar: 10m. Images are representative of two independent experiments. b HaCaT cells grown on cover slips were infected with MOI10 of HSV-1 K26-GFP and fixed and permeabilized at 14hpi followed by staining for HSV-1 gB (magenta). GFP labeled capsid proteins (VP26) are seen in green. Nuclei were stained with DAPI (blue). Scale bar: 10m. Magnified regions of merged images are indicated with dashed white boxes. Images are representative of 2 independent experiments. c HaCaT WT and C1GALT1C1 KO cells were infected with MOI3 or MOI10 of HSV-1 K26-GFP and viral capsids imaged by live microscopy at 14 and 20hpi. Fluorescent images overlaid with bright field images are also shown. Scale bar: 10m. Images are representative of two independent experiments.

We further explored the viral replication dynamics in WT and C1GALT1C1 KO cells by live imaging of GFP-labeled HSV-1. Features seen in thin optical sections (Fig.4a, b) were also reflected in widefield images (Fig.4c). In WT cells at 14 hpi, multiple capsid assembly sites could be seen in the nucleus and capsids were also associating with the nuclear envelope in most cells irrespective of the viral load (Fig.4c). In C1GALT1C1 KO cells less and smaller assembly sites could be seen, and capsids were less frequently associating with nuclear envelope. This association slightly improved later in infection (20hpi), but the capsid production did not intensify, suggesting HSV-1 infection is generally less robust in C1GALT1C1 KO (Fig.4c).

To evaluate the contribution of viral glycans to fitness of progeny virus for early interactions with wild type host cells, we added equal numbers of infectious particles, produced in propagation experiments, to WT keratinocyte monolayers following the previously outlined strategy. No defects in binding or entry were found with virions lacking different glycan structures (Fig.5ae). In fact, virions lacking O-glycan elongation were capable of accelerated binding, despite low viral titers of HSV-1 produced in C1GALT1C1 KO or C1GALT1 KO (Fig. 5). This suggests the observed propagation defects are related to host and viral factors influencing the formation of infectious virions and not their efficiency in establishing a new infection.

Binding (20min (a) or 120min (b) on ice) and entry (5min at 37C after 120min on ice (c)) of HSV-1 produced in different KO cell lines to HaCaT WT. Data is shown as WT-normalized mean+SEM of three independent experiments for eachglycoengineered virus species. Two-way ANOVA followed by Dunnetts multiple comparison test was used on raw data to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Proportion of virus bound at 20min compared to 120min (d) or proportion of virus entered at 5min compared to virus bound at 120min (e) is shown as mean+SEM of three independent experiments for each glycoengineered virus species from a total of 14 experiments. One-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Source data are provided as a Source Data file for all graphs.

The effect of O-glycosylation on HSV-1 glycoprotein localization, and prior knowledge of O-glycosite modifications compelled us to investigate specific O-glycosites. Eliminating site-specific O-glycosylation may have a more profound effect on protein function than truncation of the O-glycan structure5, 38,39. Therefore, although O-glycan truncation had no deleterious effects on properties of infectious virions, it should not be excluded that individual O-glycosylation sites could play a functional role.

We have previously identified more than 70 O-glycosites on eight out of the 12 HSV-1 surface proteins, including the indispensable fusion machinery comprised of gB, gD, gH, and gL15. Based on available structural data and defined molecular mechanisms, we mutated five out of the identified 21gB O-glycosites and three out of five gD O-glycosites most likely to affect fusion and receptor binding, respectively (Figs.6a, 7a)15. We generated Ser/Thr to Ala substitutions alone or in combination to test cell-cell fusion efficiency using a split luciferase reporter assay as a proxy for viral entry (Supplementary Table4, Supplementary Fig.5). The assay quantifies fusion between two cell types, one (effector) lacking HSV-1 entry receptors and transfected with plasmids encoding the conserved fusion machinery, and the other (target) presenting HSV-1 entry receptors (Fig.6b)40. Each cell type is also transfected with plasmids encoding half of a split luciferase reporter, which upon cell fusion can form a functional enzyme generating luminescence. In addition, we quantified gB and gD surface expression by CELISA40. We used CHO cells, refractory to HSV-1 entry, as effector, and HEK293, an HSV-1 permissive epithelial cell line, as target. We quantified low levels of Nectin 1 and HVEM on HEK293 cells (Supplementary Fig.3c), suggesting other types of receptors and co-receptors may also be involved.

a HSV-1 gB structure (PDB: 2GUM) with select mutated O-glycan acceptor sites indicated within the dashed box. Respective previously identified O-glycans were drawn manually as yellow squares. Domains are numbered in roman numericals according to Heldwein et al., Science 2006. b The cartoon illustrates the principle of split luciferase assay. c Cell surface expression of gB O-glycosite Thr to Ala mutants evaluated by CELISA using mouse anti-gB antibodies. Data is shown as mean absorbance at 450nm+SD of three technical replicates and is representative of three independent experiments. One-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). d, e Cell-cell fusion activity over 240min using gB O-glycosite Thr to Ala mutants. Data from two independent experiments is shown, where mean normalized luminescence of three technical replicates at each time point is indicated by a dot. Mean values of the two independent experiments are shown as thin lines. Data is normalized to maximum luminescence reading at final time point using WT gB for each experiment. d Data related to gB domain I mutations. e Data related to gB domain V mutations. f Cell-cell fusion activity of gB mutants at t=120min. Data is shown as mean normalized luminescence from two independent experiments. Two-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). g Average percentages of cell surface expression and fusion efficiency at t=120min from two independent experiments are shown in side-by-side columns. Source data are provided as a Source Data file for all graphs.

a HSV-1 gD structure (PDB: 2C36) with select mutated O-glycan acceptor sites indicated. Positions after removal of signal peptide, often encountered in the literature, are indicated in brackets. Respective previously identified O-glycans were drawn manually as yellow squares. N-terminal region omitted in the crystal structure is drawn as a dashed line. b Cell surface expression of gD O-glycosite mutants evaluated by CELISA. Data is shown as mean absorbance at 450nm +SD of three technical replicates and is representative of three independent experiments. One-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT. c Cell-cell fusion activity over 240min using gD O-glycosite mutants. Data from two independent experiments is shown, where mean normalized luminescence of three technical replicates at each time point is indicated by a dot. Mean values of the two independent experiments are shown as thin lines. Data is normalized to maximum luminescence reading at final time point using WT gD for each experiment. d Cell-cell fusion activity of gD mutants at t=120min. Data is shown as mean normalized luminescence from two independent experiments. Two-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT. CHO cells stably expressing Nectin 1 (e, f) or HVEM (g, h) were used as target cells to evaluate cell-cell fusion activity using gD O-glycosite mutants. Cell-cell fusion activity over 180min using CHO-Nectin 1 (e) or CHO-HVEM (g) as target. Parental CHO cell line without entry receptors was use for background subtraction. Data is presented as in (c). Cell-cell fusion activity of gD mutants at t=120min using CHO-Nectin 1 (f) or CHO-HVEM (h) as target. Data is shown as mean normalized luminescence from two independent experiments. Two-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Source data are provided as a Source Data file for all graphs.

For gB single site O-glycan mutants, we focused on the domain directly involved in fusion (I), where we identified three sites on antiparallel beta strands (T169, T267, T268), as well as the arm domain (V) comprised of two alpha helices that undergo structural rearrangements upon fusion, where we found one O-glycosite on each (T690 and T703) (Fig.6a)15, 41,42. All mutations except for T169A and T268A affected gB cell surface expression; T267A and T703A showed moderate reduction, whereas T690A showed increased expression (Fig.6c). T268A exhibited reduced fusion activity, as did T267A. Double or triple mutations in domain I severely decreased surface presentation and fusion activity (Fig.6c, d, f, g). The activity of domain V single mutants did not correlate with changes in surface expression, where T690A exhibited very low fusion activity despite increased surface presentation (Fig.6c, e, f, g). Interestingly, concomitant mutation of T703 (T690A T703A) could partially compensate for the strongly decreased activity of the T690A mutant (Fig.6g).

Though gD does not directly execute fusion, it initiates entry by binding to several different host receptors and compromised interaction with gD would translate to reduced fusion efficiency. For gD, one O-glycan site on the N-terminal tail of the protein (S33 (8)), involved in interaction with both Nectin 1 and HVEM, and two O-glycan sites on an alpha helix undergoing structural changes upon interaction with HVEM (T255 (230) and S260 (235)), were mutated (Fig.7a)15,43,44,45. All mutants maintained close to normal levels of cell surface expression of gD and fusion activity (Fig.7bd). To inspect possible contributions of gD mutations to interactions with distinct HSV-1 entry receptors, we utilized CHO cells overexpressing Nectin 1 or HVEM as target (Fig.7eh, Supplementary Fig.3a, b). Here we saw a modest reduction in Nectin 1-initiated cell-cell fusion, when T255 and S260 were collectively mutated (Fig.7e, f). A more pronounced reduction in cell-cell fusion efficiency was seen in HVEM-mediated entry upon introduction of these mutations (Fig.7g, h).

In conclusion, we identified functionally relevant O-glycan acceptor amino acids on gB, directly executing fusion, but no effects were observed for the initial engager gD in the presence of multiple host entry receptors in HEK293 cells. However, compound mutations in gD affected isolated receptor-mediated entry.

Lastly, we investigated the roles of the specific classes of glycans in direct cell-to-cell spread mediated in part by gE/gI via cell contacts of 2D grown keratinocytes, and unrestricted spread in 3D skin culture, facilitated by tissue destruction and release of free virions (Fig.8a).

a The cartoon illustrates different modes of HSV-1 cell-to-cell spread in 2D glycoengineered HaCaT cell monolayers in the context of a plaque assay, and spread in 3D organotypic skin models. b Plaque diameter on cell monolayers infected with HSV-1 Syn17+ at 48hpi. Data is presented as violin diagrams that include measurements from 3 independent experiments for each KO cell line, with 50 plaques measured for each experiment. Paired WT data includes measurements from 15 independent experiments. The dashed lines within the plots indicate median diameter, whereas the dotted lines indicate the interquartile range. One-way ANOVA followed by Games-Howells multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Source data are provided as a Source Data file. c Cell monolayers grown on cover slips were infected with 200 PFU (MOI<0.0005) of HSV-1 K26-GFP and overlaid with semi-solid media for 48h followed by fixation and staining for HSV-1 gE (magenta). GFP labeled capsid proteins (VP26) are seen in green. Confocal images at two different magnifications were taken to illustrate overviews of plaques (4 combined tiles at 10x, upper panels) as well as gE expression at higher resolution (63x, lower panels). Scale bars for the different magnifications are indicated.

We first performed plaque assays with 2D grown cells, where dissociation of progeny virions is impeded by the dense overlay media, making direct cell-to-cell spread as the predominant mode of spread. Perturbations in core 1 O-glycan biosynthesis resulted in increased plaque size, most notably in C1GALT1 KO and ST3GAL1 KO cells (Fig.8b). Upon plaque immunostaining, WT cells and KO cells exhibiting increased plaque size showed strong gE expression on the cell surface (Fig.8c). In cells lacking N-linked glycan maturation (MGAT1 KO) and those lacking MGAT4B (MGAT4B KO; MGAT5+MGAT4B KO), resulting in reduced 4-antenna branching, we found a markedly reduced cell-to-cell spread (Fig.8b). MGAT1 KO cells showed less pronounced and more punctate gE expression, which may be linked to N-glycosylation sites on gE and help explain the limited spread capacity. Surprisingly, accelerated spread was observed in MGAT4A KO cells, which also contributes to 4-antenna branching, and a similar tendency was observed for MGAT5 KO, devoid in 6-linked antenna branching.

To assess viral spread in tissue, we infected fully developed 3D epidermises built with the glycoengineered cells (Fig.9a). Different spread characteristics were observed, when viral spread was not limited to cell-to-cell contacts mediated by gE/gI complex. In wild type HaCaT skin equivalents trans epidermal lesions were observed at 36 hpi (Fig.9b, c). To avoid selection bias, we generated ten subsequent tissue sections separated by 30microns and scanned whole sections, which allowed to visualize and compare the extent of the viral lesions (Fig.9a, b, Supplementary Fig.6). We identified lesions spanning several sections and measured the cross-section areas corresponding to the central regions of those lesions (Fig.9d). Large lesions were seen in MGAT1 KO tissues, contrasting the small plaques observed in 2D (Fig.9c, d). Most N-glycan branching KO tissues, especially MGAT4A KO and MGAT5+4B KO, permitted only limited spread in the top layers of the epidermis. MGAT4B KO allowed formation of bigger lesions, but the tissue penetrance was limited, which was also the case for tissues with reduced core 1 sialylation (ST3GAL1 KO) (Fig.9c, d). No significant spread defects were noted for tissues with disruptions in GSL and GAG synthesis.

ad Fully differentiated 3D skin models built with glycoengineered cells were infected with HSV-1 Syn17+ for 36h followed by fixation in formalin and embedding in paraffin. a The cartoon illustrates the procedure for evaluating HSV-1 spread in organotypic skin tissues. FFPE tissues were sectioned every 30m for 10 consecutive slices containing two sections each, spanning from the center of the tissue outwards in two directions. b Consecutive sections were stained with a polyclonal FITC-labeled anti-HSV-1 antibody to visualize virus lesions and whole sections imaged with a microscope slide scanner. An example section series is shown with HSV-1 lesions outlined in purple. c Representative lesions were selected from the scanned section series for each KO tissue. Nuclei were labeled with DAPI. d Three lesions spanning several sections were identified for each KO tissue and lesion area measured at the centermost section. Data is shown as mean+SEM with individual measurement values indicated as black dots. One-way ANOVA followed by Dunnetts multiple comparison test was used to evaluate differences from WT (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Source data are provided as a Source Data file.

See the original post:
Glycoengineered keratinocyte library reveals essential functions of ... - Nature.com