Involvement of N-glycans in binding of Photorhabdus luminescens Tc toxin

Photorhabdus luminescens Tc toxins are large tripartite ABC-type toxin complexes, composed of TcA, TcB and TcC proteins. Tc toxins are widespread and have shown a tropism for a variety of targets including insect, mammalian and human cells. However, their receptors and the specific mechanisms of uptake into target cells remain unknown. Here, we show that the TcA protein TcdA1 interacts with N-glycans, particularly Lewis X/Y antigens. This is confirmed using N-acetylglucosamine transferase I (Mgat1 gene product)-deficient Chinese hamster ovary (CHO) Lec1 cells, which are highly resistant to intoxication by the Tc toxin complex most likely due to the absence of complex N-glycans. Restoring Mgat1 gene activity, and hence complex N-glycan biosynthesis, recapitulated the sensitivity of these cells to the toxin. Exogenous addition of Lewis X trisaccharide partially inhibits intoxication in wild-type cells. Additionally, sialic acid also largely reduced binding of the Tc toxin. Moreover, proteolytic activation of TcdA1 alters glycan-binding and uptake into target cells. The data suggest that TcdA1-binding is most likely multivalent, and carbohydrates probably work cooperatively to facilitate binding and intoxication.

The Tc toxins are large tripartite ABC-type toxin complexes (Ffrench-Constant & Waterfield, 2005;Sheets & Aktories, 2017;Sheets et al., 2011), composed of TcA, TcB and TcC components, which combine into a biologically active holotoxin complex (Ffrench-Constant & Waterfield, 2005;Lang et al., 2010;Sheets & Aktories, 2017). The TcA complex is a homopentamer and forms the binding and translocation component of the holotoxin (Gatsogiannis et al., 2013). The TcB-TcC complex forms a hollow cocoon-like structure, that encases an active enzyme formed by the hypervariable C-terminal end of TcC (TcChvr) (Busby, Panjikar, Landsberg, Hurst, & Lott, 2013;Meusch et al., 2014). The toxins are also produced in a variety of isoforms likely to target a wide array of host species (Ffrench-Constant & Waterfield, 2005;Forst, Dowds, Boemare, & Stackebrandt, 1997). For instance, up to seven TcA-and TcC-type genes and three TcB-type genes have been identified in P. luminescens strain TTO1 (Ffrench-Constant & Waterfield, 2005;. The holotoxin binds to target cells through a yet unknown receptor and is probably taken up by receptor-mediated endocytosis. Acidification of the toxin-containing endosome induces conformational changes in TcA, which then perforates the endosomal membrane in a syringe-like mechanism, forming a pore through which the toxic TcChvr translocates into the host cell cytosol. Here, it re-folds with the help of cytosolic host chaperones into its active conformation . Two TcChvrs from P. luminescens have been characterized, TccC3 and TccC5. Both are ADP-ribosyltransferases, targeting actin and small GTPases of the Rho family, respectively (Lang et al., 2010;Lang, Schmidt, Sheets, & Aktories, 2011;Pfaumann, Lang, Schwan, Schmidt, & Aktories, 2015). While TccC3 modifies actin at T148, resulting in increased actin polymerisation .
ADP-ribosylation of Rho proteins (Q63 of RhoA and Q61 of Rac and Cdc42) locks the GTPases in a persistently active state. Thus, both toxins act in concert to induce actin polymerisation and clustering (Lang et al., 2010. Recently, we showed that TcdA1 (one of the TcA proteins from P. luminescence) is processed by the Photorhabdus metalloprotease PrtA1 and by collagenase, leading to increased toxin activity (Ost, Ng'ang'a, Lang, & Aktories, 2019). This increased toxin activity was in part due to increased cell-surface binding of the cleaved TcdA1. Additionally, we and others showed that the Tc toxin could be used to deliver foreign peptides into target cells (Ng'ang, Ebner, Plessner, Aktories, & Schmidt, 2019;Roderer, Schubert, Sitsel, & Raunser, 2019). Here, we sought to identify cell-surface interactors of the cell binding component TcdA1 of P. luminescens Tc toxin. We report that TcdA1 interacts with N-glycans on target cells, particularly Lewis antigens as possible interactors. Moreover, we identified Chinese hamster ovary (CHO) cells devoid of complex N-glycans (Lec1 cells) as being completely resistant towards the P. luminescens holotoxin, whereas parental CHO cells are sensitive.

| TcdA1 interacts with N-glycans and Lewis antigens
To identify interactors of TcdA1, a glycan microarray screen was performed with the help of the Consortium for Functional Glycomics (CFG) (www.functionalglycomics.org). In line with the previously described varying binding capabilities of TcdA1 due to cleavage by metalloproteases like collagenase (Ost et al., 2019), we studied processed TcdA1 (TcdA1 FPLC/Collag ) and the respective unprocessed TcdA1 (TcdA1 FPLC ) as control. In addition, a TcdA1 preparation (TcdA1 aff ) was employed that was not pre-treated for processing (see Methods and Figure S1 for cleavage protocol and TcdA1 nomenclature). The screen data were further analysed using the online Glycan Array Dashboard (GLAD) software (https://glycotoolkit.com/GLAD/) (Mehta & Cummings, 2019). First, a force graph showing interaction of the protein samples with the top glycan hits was generated using default parameters ( Figure 1a). All hits with an average relative fluorescence unit (RFU) above 1,000 were included in the analysis. The hits were then grouped according to their interaction with each sample. As shown in Table 1, 10 hits out of 585 tested met these criteria ( Figure 1a). Two hits interacted with all samples (red oval), five hits interacted with three samples, that is, the cleaved toxins (5 and 50 μg) and 50 μg of TcdA1 FPLC (blue oval). Finally, three hits interacted with two samples, the collagenase-cleaved samples only (green oval) ( Figure 1a and Table 1). Binding profiles containing the top 20 hits of the 5 μg samples of TcdA1 FPLC and TcdA1 FPLC/Collag are shown in Figure 1b,c, respectively, while the 50 μg samples are shown in Figure S2a,b, respectively. Additionally, each glycan present in the top 10 hits (Figure 1a and Table 1) was labelled with the respective coloured circle in Figure 1b,c, as well as Figure S2a,b.
As shown in Table 1, Figures 1b and S2a, the overall top hits interacting with all toxin preparations (red circles) were N-glycans with terminal N-acetylglucosamine (GlcNAc) or galactose residues (Figure 1d,e). A single sialylated-N-glycan shared by three samples (the two cleaved-TcdA1 samples and the 50 μg uncleaved-TcdA1 sample) was also observed in the top hits ( Figure 1f). Highly represented in the top hits were fucosylated (Lewis X and Y) structures (Figure 1g-i), whose different variations interacted with at least two toxin samples (blue and green circles). See Table S1 for the list of all the glycans used in the screen.
The binding profiles also showed that cleavage of TcdA1 by collagenase (TcdA1 FPLC/Collag ) altered its pattern of interaction with the glycans ( Figure S2c). Different glycans were enriched on cleaved and uncleaved proteins (compare Figures 1b,c and S2a,b Figure S2c), suggesting saturation of glycan-binding sites. Taken together, these data suggest cleavage-related structure and glycan-binding changes on TcdA1.
To get a picture of the possible glycan-binding sites and domains on TcdA1, an in silico analysis was also performed using 3DLigandSite (Wass, Kelley, & Sternberg, 2010) and I-TASSER (Yang et al., 2015).
3DLigandSite is a web server for prediction of ligand-binding sites on full protein and/or domain structures. Similarly, I-TASSER is a software suite that models protein structure and function from sequences and structural similarity with characterized proteins in protein databases. For this analysis, TcdA1 was split into its various domains as described by Meusch et al., 2014 (Figure S3a), and then the different domains were submitted to 3DLigandSite and I-TASSER for prediction of potential ligands. Splitting of TcdA1 was necessitated by its large size, which could not be accommodated by the software.
Ligands bound to similar structures were fitted onto the TcdA1 domains. Subsequently, all the predicted ligands on each domain were ranked according to the total number of hits found for each site, that is, the total number of ligand-binding proteins found for each ligand that showed structural similarity to the respective domain on TcdA1 (Tables S2 and S3).
T A B L E 1 Top 10 hits (out of 585 tested) of a glycan microarray screen that are shared by all tested samples of TcdA1 (TcdA1 FPLC and TcdA1 FPLC/Collag )

Number
Glycan ID IUPAC name Common name Complex-type N-glycan Neo Lactosamine, N-glycan basic, Lewis type 1 Lactosamine, N-glycan basic, Lewis type 2 Sialylated, fucosylated, N-glycan basic, Lewis type 2 Fucosylated, N-glycan basic, Lewis X, Lewis Y Lactosamine, N-glycan basic, Lewis type 2 Fucosylated, N-glycan basic, Lewis X, Lewis Y Fucosylated, Lewis X, Lewis Y Note: The hits are colour-coded according to the groups in Figure 1a.
F I G U R E 1 Glycan microarray screen of TcdA1. (a) Force graph of the top hits interacting with the four TcdA1 samples (5 and 50 μg TcdA1 FPLC ; 5 and 50 μg TcdA1 FPLC/Collag ). All hits with an average relative fluorescence unit (RFU) above 1,000 (grey circles) were linked to each TcdA1 sample to reveal shared interactions. The hits marked in red (red oval) were shared by all four samples, those in green (green oval) were shared by three samples and those in blue (blue oval) were shared by two samples. Interestingly, as shown in Table S2 and Figure S3, a variety of predicted glycan-binding sites were found on TcdA1, particularly on the putative receptor binding domains, which all had at least one predicted glycan-binding site. No sites were found on the α-helical shell domains, indicating that glycan-binding may be restricted to the putative receptor binding domains. Overall, the large variety of predicted glycan-binding sites mirrored the glycan screen, in that, many sites accommodated interactions with mostly Nacetylglucosamine (GlcNAc), galactose (Gal), fucose (Fuc) and, to a lesser extent, mannose (Man) (Table S2) toxin B, which to the best of our knowledge lacks Le X interactions, was used. As shown in Figure 2b,d, the Le X trisaccharides delayed cell intoxication by PTC3 in a concentration-dependent manner, while C.
difficile toxin B was not affected by addition of the sugar (Figure 2c,e).
Furthermore, exogenous addition of a mixture of the individual sugars constituting the Le X trisaccharide (Gal, GlcNAc and Fuc) at the same concentrations did not affect cell intoxication ( Figure 2f). These results further supported the hypothesis of the involvement of Le X trisaccharides in the interaction of PTC3 with the cell surface. Additionally, the data suggested that the glycans must be in a defined complex, rather than single monosaccharides in order to interact with the toxin.

| N-glycan-deficient cells are resistant to intoxication by PTC3
To gain more insight into N-glycan binding, Chinese hamster ovary (CHO) Lec1 cells, which do not produce complex and hybrid N-glycans, were used. These cells have a mutated MgatI gene, which encodes N-acetylglucosaminyltransferase I (GlcNAc-TI) and hence lack N-glycans (Puthalakath, Burke, & Gleeson, 1996). However, they have short oligomannose trees and produce normal O-linked glycans (North et al., 2010). The cells were intoxicated with affinity-purified PTC3 (PTC3 aff ) ( Figure S1) and, as a positive control, with Clostridium botulinum C2 toxin, which was shown to interact with N-glycans (Eckhardt, Barth, Blöcker, & Aktories, 2000). As a negative control, C.
difficile toxin B was used. As shown in Figure 3a,b, Lec1 cells were resistant to PTC3 and C. botulinum C2 toxin, but not to C. difficile toxin B, indicating a dependence on complex N-glycans for PTC3 binding.
Comparatively, cells deficient in various monosaccharides that constitute N-glycans were sensitive to PTC3. Specifically, CHO Lec2 and Lec8 cells lacking sialic acid and galactose, respectively (North et al., 2010), as well as fucose-deficient leukocyte-adhesion deficiency II (LADII) cells (Wild, Luhn, Marquardt, & Vestweber, 2002) were all susceptible to PTC3. All cells were intoxicated with PTC3 for 5 hr, and the percentage of rounded cells was quantified. As shown in Figure S4a-c, none of the cell lines were resistant to intoxication by PTC3. Furthermore, neither Lec2 nor Lec8 cells showed decreased toxin binding compared to the wild-type CHO-K1 ( Figure S4d). These data further supported the involvement of glycan complexes rather than single monosaccharides in toxin interaction with the target cells.
We sought to confirm that the gained resistance of Lec1 cells to PTC3 is due to deficiency in GlcNAc-TI activity. Accordingly, we

| TcdA1 may interact with glycosaminoglycans (GAGs)
GAGs are linear, anionic polysaccharides that provide an ideal landing pad for proteins, macromolecules, bacteria, viruses and parasites (Kamhi, Joo, Dordick, & Linhardt, 2013;Tao et al., 2019). Therefore, to test whether they may be involved in TcdA1 binding, TcdA1 aff was premixed with heparin (HP), heparan sulfate (HS), the semi-synthetic GAG analogue dextran sulfate (DexS) and chondroitin sulfate (CS) and incubated for 2-5 min at RT. Then, together with BC3, it was used to intoxicate HeLa cells. HP, HS and DexS partially reduced cell intoxication ( Figure S5a-c). The effect of CS was minimal ( Figure S5d). Lower concentrations of HP and HS than of DexS were required to inhibit intoxication. Nevertheless, these results suggested some influence of GAGs on binding and intoxication by PTC3.
To study the influence of SA on PTC3, we first assessed its effect on HeLa cell intoxication. In this regard, TcdA1 aff was premixed with increasing concentrations of SA as previously described and then added to cells together with BC3. As shown in Figure 7a To resolve the discrepancy between toxin binding and intoxication, we used a more sensitive assay to assess toxin binding. To this end, we analysed a BC3 complex where C3hvr (the enzyme contained in BC3) was replaced with MLuc7, a small luciferase from the marine copepod Metridia longa (Markova, Golz, Frank, Kalthof, & Vysotski, 2004). We had previously shown that Mluc7 was efficiently translocated through TcdA1, and that it subsequently provided a intoxication in high concentrations, which most likely is due to its negative charge and not due to specific interference with the toxinreceptor interaction. The saccharides, however, reduced toxin activity as part of the Le X -trisaccharide, suggesting that they probably need to be incorporated into glycan structures to have an effect.
Very recently, while our manuscript was in preparation, the group of Stefan Raunser reported on the interaction of glycans with TcdA1 (Roderer et al., 2020). They also identified Lewis antigens as interaction sites of TcdA1. Moreover, using molecular docking, they could demonstrate the interaction of BSA-Lewis X with the so-called receptor binding domain D. This interaction was of low affinity and supports the view of a co-receptor function. In contrast, our findings with Toxins premixed with the indicated concentrations of sialic acid were allowed to bind on cells for 30 min at 4 C. Then, the pH of the surrounding medium was reduced to pH 5 to induce injection. Subsequently, the cells were washed and lysed before analysis of luminescence. (f) Bioluminescence analysis of PTC3-Mluc7 (1 nM) premixed with the indicated concentrations of sialic acid, as well as GlcNAc (3 mM) and pH 5.6 buffer. Unpaired, two-tailed t test (*p < .05; **p < .01; ***p < .001, ns, not significant, n = 3, ±SEM). The scale bar is 80 μm CHO-Lec1 cells, which are only marginally sensitive towards the Tc toxin, suggest that N-glycans play a leading and essential role in the interaction of the toxin with host cells. This was also supported by increased glycan binding by the collagenase-cleaved toxin (TcdA1 FPLC/Collag ), which has also shown increased cell binding and intoxication kinetics (Ost et al., 2019). CHO-Lec1 cells were also resistant to TcdA1 FPLC/Collag . Importantly, glycan structures on Lec1 cells are much reduced, leaving only Man5GlcNAC2 core structures, but does not affect O-glycans (North et al., 2010) ( Figure S7a). Therefore, binding to N-glycans may require, at least, the next level of GlcNAc addition, which is catalysed by the missing Mgat1 gene. After the action of Mgat1 and mannosidase II, additional GlcNAc-transfererases create more terminal GlcNAc residues, which are not further extended by galactoses in Lec8 cells (Kawar et al., 2005) ( Figure S7b). As these Lec8 cells are sensitive to PTC3, the critical step that makes cells sensitive to the toxin seems the addition of GlcNAc residues to N-glycans. Exactly an N-glycan terminating with three GlcNAc residues was the top hit in the glycan array, followed by several other N-glycans further extended. This is also the last step that is common between mammals and insect.
We and Roderer and coworkers (Roderer et al., 2020) showed the binding of heparins and fucosylated structures of Lewis-type antigens onto various types of TcA toxin components. In each case, we observed a relatively moderate inhibition of the toxin effect by the addition of these compounds, with heparin showing the strongest effects. However, CHO mutant (pgsA-745) cells, which lack or have reduced CS and HS content (Esko, Stewart, & Taylor, 1985), exhibited no reduced intoxication, suggesting cell-type-dependent variation in the effect of GAGs.
Roderer et al. (Roderer et al., 2020) suggested that because GAGs form long negatively charged polymers on the cell surface, extending 20-150 nm from the plasma membrane, they could act as primary low-affinity receptors that provide a landing pad for the toxin. This initial binding on GAGs could then be followed by binding to N-glycans, such as the Lewis antigens, which themselves are also not essential, as they are absent in sensitive Lec8 and LADII cells. Notably, of all the TcA proteins they studied, only TcdA1 from P. luminescence, showed significant binding to N-glycan fragments. For none of the other TcA proteins, a significant binding to the glycan microarray was evident, even at higher protein concentrations.
Interestingly, for a related Yersinia entomophaga ABC toxin, GlcNAc-containing glycans were the main binding partners (Piper et al., 2019).
Like the Le X antigens, the marginal effects observed indicate that GAGs and sialic acids are also likely not high-affinity targets, further supporting their role as an initial landing pad for the toxin. Notably, Lec1 cells expressed more than the double amount of GAG compared to wild-type cells (Fujitani et al., 2013), yet they were resistant to TcdA1.
The appearing picture of our experiments and of related studies by others as described above is that Tc toxins might bind with low affinity to a variety of different glycans, possibly by different binding sites, but none of these structures alone are essential for binding. We, as well as Roderer et al.,

| Protein expression and purification of Photorhabdus toxins
Proteins were expressed and purified as previously described (Gatsogiannis et al., 2013;Lang et al., 2017). In short, E. coli

| Cleavage of TcdA1
In some instances, affinity purified TcdA1 (TcdA1 aff ) was further processed by cleavage using collagenase as described by (Ost et al., 2019). For this, 3 μg of TcdA1 was first incubated for 1 hr with ). The Äkta purification involved a run of gel filtration/sizeexclusion chromatography in a Superose 6 Increase column (GE healthcare, Freiburg, Germany). As a control, affinity purified TcdA1 was incubated in the same conditions in buffer only without the enzyme and then purified by FPLC (TcdA1 FPLC ). This procedure is summarised schematically in Figure S1. Dylight488 or unlabelled. In the latter case, a TcdA1-specific mouse monoclonal antibody, followed by a fluorescently labelled secondary antibody were used to probe the toxin bound onto the microarray. Subsequently, the screen data were first tabulated and analysed in Excel, followed by visualisation and analysis using the online Glycan Array Dashboard (GLAD) software (https://glycotoolkit.com/GLAD/) (Mehta & Cummings, 2019) provided by the CFG.

| Cell cytotoxicity assays
For cytotoxicity experiments, cells were seeded in culture dishes and incubated in starvation medium (0.5% FCS) together with the respective toxins. Where pharmacological inhibitors were used, the cells were pre-incubated with the inhibitors for the stated duration before intoxication. For intoxication in the presence of glycans, TcdA1 was premixed with each sugar at the indicated concentrations, pre-incubated for 2-5 min at room temperature, then, added to the cells together with the BC3 complex. After the indicated incubation periods, the cells were visualised and pictures taken. Toxin effects were quantified by counting the number of rounded cells using the ImageJ/Fiji software, followed by subsequent analysis using statistics programs.

| In silico prediction of glycan binding sites on TcdA1
To predict potential glycan binding sites on TcdA1, the protein was split into its various domains as described by (Meusch et al., 2014).

| Post-intoxication ADP-ribosylation of Actin
CHO-K1 and Lec1 cells intoxicated for 3 hr by PTC3 were collected and lysed (lysis buffer; 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 1 mM PMSF and 2.5 mM DTT). Then, after lysis, in vitro ADP-ribosylation of actin was performed as previously described . Briefly, BC3 (0.7 μM) and biotin-labelled NAD + (70 μM) were added to the lysates and incubated for 20 min at 37 C. As a control, recombinant βγ-actin (1.9 μM) was used. The reactions were stopped by addition of Laemmli buffer. Afterwards, the samples were subjected to SDS-PAGE, blotted and the biotin-labelled, that is, ADP-ribosylated actin, was detected with horseradish peroxidase-avidin and the ECL system. The amount of total actin is shown as loading control.

| Transfection of CHO Lec1 cells
Cells were seeded on 12-well plates and allowed to grow overnight. The

| pH-dependent bioluminescence assays
For pH-dependent bioluminescence assays, HeLa cells were first incubated with 100 nM of Bafilomycin A1 (Enzo life sciences) for 30 min at 37 C. Then, they were transferred to ice, precooled to 4 C. Afterwards, BC3-Mluc7 only or PTC3-Mluc7, treated with, or without sialic acid, was added and incubated for 1 hr in 1× Hank's balanced salt solution (HBSS) buffer supplemented with 20 mM Hepes, pH 7.5, to allow for toxin binding. This experimental setup ensured toxin binding at the cell surface without uptake into the cells. Then, to induce injection of Mluc7 into the cells by the surface-bound toxin complex, the pH of the surrounding medium was changed to acidic pH 5 medium (DMEM medium with 0.5% FCS and 20 mM MES, pH 5). This pH change mimicked the endosomal pH required to induce pore formation and translocation of the payload into the cytosol. These cells in pH medium were incubated for 1 min at 37 C before the medium was shifted back to pH 7.5. Afterwards, the cells were washed and treated with trypsin/EDTA for 2 min at 37 C to degrade extracellular protein and to simultaneously detach them from the plate. Finally, the cells were harvested and lysed in passive lysis buffer (Promega) and transferred into an infinite M200 microplate reader (Tecan) for detecting luciferase activity in the presence of coelenterazine H (Biotium).