A Cytosolic Reductase Pathway is Required for Complete N-Glycosylation of an STT3B-Dependent Acceptor Site

N-linked glycosylation of proteins entering the secretory pathway is an essential post-translational modification required for protein stability and function. Previously, it has been shown that there is a temporal relationship between protein folding and glycosylation, which influences the occupancy of specific glycosylation sites. Here we use an in vitro translation system that reproduces the initial stages of secretory protein translocation, folding and glycosylation under defined redox conditions. We found that the efficiency of glycosylation of hemopexin was dependent upon a robust NADPH-dependent cytosolic reductive pathway, which could also be mimicked by the addition of a membrane impermeable reducing agent. The identified hypoglycosylated acceptor site is adjacent to a cysteine involved in a short range disulfide bond, which has been shown to be dependent on the STT3B-containing oligosaccharyl transferase. We also show that efficient glycosylation at this site is dependent on the STT3A-containing oligosaccharide transferase. Our results provide further insight into the important role of the ER redox conditions in glycosylation site occupancy and demonstrate a link between redox conditions in the cytosol and glycosylation efficiency.


Introduction 33
Proteins entering the secretory pathway are subject to a variety of modifications, the most prevalent of 34 which include N-linked glycosylation and disulfide formation (Bulleid, 2012;Cherepanova et al., 35 2016). N-glycosylation is catalysed by one of two oligosaccharyl transferases (OST) that transfer a 36 pre-formed oligosaccharide from a dolichol-phosphate intermediate to asparagine residues on the 37 polypeptide chain within the consensus sequence -N-X-S/T where X is any amino acid other than 38 proline (Kelleher et al., 2003). The two OST isoforms are multi-subunit complexes characterised by 39 the catalytic subunits STT3A or STT3B with common subunits as well as complex-specific subunits 40 including DC2 and KCP2 for the STT3A, and the thioredoxin-domain containing proteins MagT1 or 41 TUSC3 for the STT3B complex (Blomen et al., 2015;Roboti and High, 2012;Shibatani et al., 2005). 42 It has been demonstrated previously that the STT3A complex associates with the ER translocon 43 (Braunger et al., 2018;Shibatani et al., 2005) and catalyses the co-translational glycosylation of 44 proteins, whereas the STT3B complex glycosylates sites skipped by STT3A acting predominantly 45 post-translationally (Cherepanova et al., 2014;Ruiz-Canada et al., 2009). Because of their distinct 46 specificities, some substrates require the STT3A or STT3B complexes for efficient glycosylation 47 (Cherepanova and Gilmore, 2016;Cherepanova et al., 2014). Indeed, recent proteomic analysis of 48 glycoproteins synthesised in either STT3A or STT3B depleted cells identify classes of STT3A and 49 STT3B dependent N-glycosylation sites (Cherepanova et al., 2019). Deficiency of the STT3B 50 complex cannot be compensated by the STT3A complex, resulting in hypoglycosylation of substrates 51 affecting their function and leading to disease pathologies linked to immunodeficiency (Blommaert et 52 al., 2019;Matsuda-Lennikov et al., 2019). 53 Utilisation of potential glycosylation sites or sequons is not guaranteed and is dependent on the 54 position within the chain (Nilsson and von Heijne, 2000;Ruiz-Canada et al., 2009;Shrimal et al., 55 2013), or the amino acid context of the site (Shrimal and Gilmore, 2013) with the kinetics of the 56 folding or collapse of the polypeptide chain affecting glycosylation. Sequons buried within a protein 57 structure, present at the amino or carboxy terminus or close to cysteines involved in disulfide 58 formation may be underutilised giving rise to heterogeneity in glycoprotein forms. 59 Hypoglycosylation of sequons due to disulfide formation can be dependent upon STT3A or STT3B 60 and is reversed when proteins are prevented from forming disulfides under highly reducing conditions 61 (Allen et al., 1995;Cherepanova et al., 2014). In addition, STT3B-dependent glycosylation of 62 cysteine-proximal sites requires the oxidoreductase activity of the thioredoxin-domain containing 63 subunits MagT1 or TUSC3 (Cherepanova and Gilmore, 2016;Cherepanova et al., 2014). Structural 64 analysis of TUSC3 indicates its direct binding to cysteine-containing peptides, suggesting direct 65 binding to the polypeptide to slow down protein folding and disulfide formation to allow 66 glycosylation to occur (Mohorko et al., 2014). The fact that MagT1 is mainly oxidised in cells 67 (Cherepanova et al., 2014) would suggest that it acts as a reductase, thereby preventing disulfide formation prior to glycosylation. Taken together these observations indicate a crucial role for the 69 STT3B complex in coupling disulfide formation and glycosylation to maximise utilisation of 70 cysteine-proximal acceptor sites. 71 The temporal relationship between disulfide formation and glycosylation suggests that the redox 72 status of the ER may contribute towards sequon utilisation (Cherepanova et al., 2016). ER redox 73 reactions are balanced to allow both disulfide formation and reduction resulting in the formation of 74 the correct disulfides within folding proteins (Bulleid and van Lith, 2014). Members of the protein 75 disulfide isomerase (PDI) family are thioredoxin-domain containing proteins that catalyse disulfide 76 exchange reactions (Bulleid, 2012). Their oxidation is catalysed by Ero1, which couples the reduction 77 of oxygen to the formation of a disulfide in PDI (Cabibbo et al., 2000). Specific members of the PDI 78 family, such as ERp57 (Jessop et al., 2007) and ERdj5 (Oka et al., 2013;Ushioda et al., 2008) 79 catalyse the reduction of non-native disulfides either allowing the correct disulfides to form or 80 targeting misfolded proteins for degradation. Exactly how these PDI enzymes are reduced is 81 unknown but recent evidence suggests a role for the cytosolic reductive pathway in correct disulfide 82 formation, driven by the reduction of thioredoxin reductase (Cao et al., 2020;Poet et al., 2017). 83 It is likely that the ER oxidative and reductive pathways influence the STT3B subunits, MagT1 and 84 TUSC3 during oxidoreductase activity towards cysteines proximal to sequons. Hence, the correct 85 utilisation of sequons may well be regulated by the prevailing redox conditions within the ER. To 86 address the role of ER redox conditions on utilisation of STT3B-dependent acceptor sites we 87 capitalised on a recently described in vitro translation system that reproduces the early stages of 88 secretory protein ER translocation and modification under defined redox conditions (Poet et al., 2017;89 Robinson and Bulleid, 2020). In this system, the redox conditions can be manipulated simply by the 90 addition of glucose 6-phosphate (G6-P) which recycles NADPH thereby driving the cytosolic 91 reductive pathway. When a source of ER is included during translation, the newly synthesised 92 proteins are translocated across the ER membrane and can undergo both disulfide formation and N-93 linked glycosylation (Wilson et al., 1995). We chose to translate the STT3B-dependent substrate 94 hemopexin (Cherepanova et al., 2014) in such a system, and show that it is hypoglycosylated in the 95 absence of added G6-P, an effect that is reversed upon G6-P inclusion. Our results highlight the role 96 of ER redox in the efficiency of sequon glycosylation and reveals an unexpected role for the NADPH-97 dependent cytosolic reductive pathway in the function of both the STTA and STT3B-containing OST 98 complex.

Results 100
Cytosolic reductive pathway determines the extent of sequon usage in a STT3B-dependent 101

glycoprotein. 102
Our initial experiments aimed to determine whether the redox conditions within the ER had any effect 103 on the fidelity of sequon usage within a model protein, hemopexin, that had previous been shown to 104 undergo STT3B-dependent hypoglycosylation. Hemopexin has 5 potential sequons and forms six 105 disulfides (Fig. 1A). For these experiments we adjusted the redox status of our in vitro translation 106 reactions by adding specific components to the rabbit reticulocyte lysate. We have previously shown 107 that a commercial reticulocyte lysate that has no added DTT allows disulfides to form in proteins 108 synthesised even in the absence of semi-permeabilised (SP) cells as a source of ER (Poet et al., 2017). 109 Supplementing this lysate with G6-P to drive G6-P dehydrogenase (G6PDH) and thioredoxin 110 reductase (TrxR1) activity, renders this lysate sufficiently reducing to prevent disulfide formation in 111 proteins synthesised without SP cells but allows disulfide formation in translocated proteins when SP-112 cells are present. When hemopexin was translated in the absence of added G6-P and presence of SP-113 cells we noted the appearance of two potential glycoforms giving rise to a doublet after SDS-PAGE 114 only the slower migrating glycoform was synthesised when translations were carried out in the 118 presence of the membrane permeable and impermeable reducing agent DTT or TCEP respectively 119 (lanes 3 and 4). Hence it would appear that hemopexin is hypoglycosylated in the absence of G6-P, 120 an effect that is reversed when translations were carried out under more reducing conditions. As G6-P 121 most likely alters the redox conditions by recycling NADP to NADPH in the cytosol and TCEP is 122 membrane impermeable these results suggest that the redox conditions on the cytosolic side of the ER 123 membrane are affecting the glycosylation efficiency of ER translocated hemopexin. 124 To verify that the two translation products seen after synthesis of hemopexin are indeed glycoproteins 125 and to identify the status of the faster migrating band, we carried out a limited digestion of the protein 126 with endoglycosidase (endo) H (Fig. 1C). Digestion of the translation products with the highest 127 enzyme concentration resulted in a single band corresponding to the fully deglycosylated protein 128 indicating that the two products are indeed glycoforms (lanes 1 and 2). Addition of limiting amounts 129 of endo H to the reaction allowed partial digestion revealing all 5 potential glycoforms that arise from 130 variable digestion of the five oligosaccharide side chains on hemopexin (lane 5). From this analysis 131 we can conclude that the two apparent glycoforms seen when hemopexin is translated in the absence 132 of added G6-P are indeed the 5 and 4 glycan forms. These results are consistent with our previously 133 observed hypoglycosylation of hemopexin when expressed in mammalian cells (Shrimal and Gilmore, 134

2013). 135
To verify that the reversal of hemopexin hypoglycosylation by G6-P is mediated by the recycling of 136 NADP, we supplemented the translation reactions with NADPH (Fig. 1D, E). As with G6-P, we 137 could reverse the hypoglycosylation of hemopexin just by adding NADPH confirming that the effect 138 is not due to G6-P directly influencing the glycosylation machinery or synthesis of the 139 oligosaccharide side chain.

156
Defining the sequon giving rise to G6-P-dependent hypoglycosylation 157 To determine which sequon within hemopexin is hypoglycosylated, we mutated hemopexin N187 and 158 N453 to glutamine, as it has been noted before that these sequons can frequently be skipped by the oligosaccharyl transferases (Shrimal and Gilmore, 2013). We found that hypoglycosylation in the 160 absence of G6-P occurred with wild type hemopexin and hemopexin N453Q, which can be resolved 161 by the addition of G6-P ( Fig. 2A, lanes 1, 2 and 5, 6). No hypoglycosylation was observed with 162 hemopexin when N187 is mutated in either the single or double mutant (lanes 3 and 7). 163 Quantification of the level of hypoglycosylation from three separate experiments supports the 164 qualitative gel analysis (Fig. 2B). These results demonstrate that N187 is the acceptor site that is

174
The N187 sequon is NCS with the C188 forming a short-range disulfide with C200 in the native 175 structure of hemopexin (Fig. 1A) (Paoli et al., 1999). As it has been observed previously that 176 cysteine-proximal glycosylation sites are often skipped (Cherepanova et al., 2014), we evaluated the 177 role of hemopexin C188 in hypoglycosylation. In addition, to determine if the formation of the C188-178 C200 disulfide prevents efficient glycosylation, we mutated both cysteines individually and together 179 to serine. We found that mutation of the more distal C200 did not prevent hypoglycosylation, which 180 was resolved by inclusion of G6-P (Fig. 3A, lanes 5 and 6). In contrast, mutation of C188, in either C188S or C188S/C200S mutant, resulted in almost complete loss of hypoglycosylation (lanes 3 and 182 7). 183 184

190
Preventing the native disulfide from forming by mutating C200 did not stop hypoglycosylation of the 191 acceptor site whereas mutating C188 did, suggesting that the presence of the cysteine restricts 192 glycosylation rather than the disulfide per se. Alternatively, C188 could be oxidised or form a non-193 native disulfide to an alternate cysteine to C200. To test these two possibilities, we created a 194 construct where we mutated all the cysteines in the sequence apart from the cysteine at C188. Upon 195 translation the protein was fully glycosylated either in the presence or absence of G6-P (Fig. 3B, lanes  196   3 and 4). This result suggests that the formation of either a native or non-native disulfide via C188 197 restricts the ability of the OST from glycosylating N187 resulting in hypoglycosylation. In addition,it 198 shows that it is a change in the redox conditions during synthesis that is reversed by the inclusion of 199 G6-P maintaining the C188 in a reduced state to allow efficient glycosylation. 200

Hemopexin forms distinct disulfide-bonded species during translation in the absence or 201 presence of added G6-P 202
The ability of C188 to form a native or non-native disulfide affected N187 occupancy, which suggests 203 a role for G6-P in modulating disulfide formation in our translation system. Indeed, we have 204 previously shown that addition of G6-P prevents non-native disulfide formation in a variety of 205 proteins by recycling NADPH and maintaining cytosolic thioredoxin in a reduced state (Poet et al., 206 2017). To determine the redox status of hemopexin following translation, we prevented disulfide 207 rearrangement following synthesis using an alkylating agent and separated the translation products 208 under non-reducing conditions. Typically, long-range disulfides formed in proteins affect their 209 electrophoretic mobility by altering the hydrodynamic volume of the denatured protein. When the 210 hemopexin translation products were analysed this way, we observed several oxidised species with a 211 greater mobility than the reduced protein (Fig. 4A, compare lane 1 and 5). To rule out any 212 contribution of hypoglycosylation to the pattern under non-reducing conditions, the samples were also 213 treated with endo H to remove all glycans (even numbered lanes). Multiple oxidised species were still 214 observed indicating that hemopexin forms distinct and incompletely disulfide-bonded species in our

228
When the translations were carried out in the presence of G6-P, most of the oxidised species migrated 229 as those in untreated lysates (lane 1 and 2 vs 3 and 4). One species appeared after G6-P addition, seen 230 in the endoH treated samples (lane 2 versus 4, indicated by an arrow). This translation product was 231 digested by proteinase K, indicating that it corresponds to untranslocated hemopexin (Fig. 4B, lane 4, arrow). The removal of untranslocated material is most clearly seen when the samples are separated 233 under reducing conditions (lane 5 and 7, down arrows). The remainder of the bands are protected 234 from proteinase K digestion, indicated that all these species are translocated into the ER lumen. The 235 dramatic change to the redox status of untranslocated protein is consistent with our previous studies 236 indicating that the addition of G6-P restores a robust reducing pathway in the cytosol, but does not 237 prevent correct disulfide formation in proteins translocated across the ER membrane (Poet et al., 238 2017;Robinson and Bulleid, 2020). Interestingly, the oxidised species migrating with intermediate 239 mobility become less diffuse after G6-P addition (lanes 2 and 4, vertical line and asterisk) suggesting 240 some rearrangement of disulfides.

G6-P does not act via an ER NADPH pool or PDIs involved in non-native disulfide reduction 253
The results presented so far would suggest a requirement to maintain a robust cytosolic reductive 254 pathway to ensure the efficient glycosylation of hemopexin. Alternatively, G6-P can be transported 255 into the lumen of the ER by the glucose-6-phosphate transporter (G6PT), where it can be used by the 256 ER-localised hexose-6-phosphate dehydrogenase (H6PDH) to locally generate NADPH (Clarke and 257 Mason, 2003). The NADPH could be used to provide reducing equivalents for the promotion of 258 hemopexin glycosylation by a yet unidentified pathway. To determine if the ER NADPH pool is 259 involved, we evaluated the effect of G6-P on hemopexin glycosylation using SP-cells derived from a 260 H6PDH knock out (KO) cell-line (Fig. 5A). G6-P addition still reversed the hypoglycosylation of 261 hemopexin in this cell line ruling out a role for H6PDH (Fig. 5B). 262 The fact that hypoglycosylation of hemopexin can be influenced by a disulfide formed via C188 led 263 us to determine whether previously characterised PDIs that have reductase activity might be involved 264 directly or via disulfide exchange with the STT3B-subunits MagT1 or TUSC3. We focused on the 265 PDIs ERp57 and ERdj5 as they are known to catalyse the reduction of non-native disulfides (Jessop et 266 al., 2007;Oka et al., 2013;Ushioda et al., 2008). We created KO cell-lines for the individual proteins 267 as well as a combined ERdj5-ERp57 double KO (Fig. 5A). For each of these cell-lines there was no 268 effect on the reversal of hypoglycosylation facilitated by G6-P (Fig. 5B). Hence the reductive 269 pathway maintained by the addition of G6-P functions independently from the known ER PDI

278
Role of the OST complexes containing either STT3A or STT3B catalytic subunits. 279 To determine whether STT3A or STT3B are required for the G6-P effect on hemopexin glycosylation, 280 we used two previously characterised KO cell lines for STT3A and STT3B (Cherepanova and 281 Gilmore, 2016). To look specifically at glycosylation of the cysteine proximal N187 we used the 282 hemopexin N453Q mutant for in vitro translation in SP cells derived from these KO cell lines. 283 Previously, the N453 site had been shown to be STT3B-dependent (Shrimal and Gilmore, 2013). 284 Hemopexin N453Q translated into STT3A KO SP cells was glycosylated similarly to that in wild type 285 cells, with both G6-P and DTT able to partially resolve the hypoglycosylation (Fig. 6A, lanes 4-6 286 versus 1-3 and Fig. 6C). These results indicate that the effect of G6-P seen in wild type cells is not 287 due to the STT3A OST. 288 In contrast, hemopexin hypoglycosylation is more pronounced in STT3B KO cells in the absence of 289 G6-P with levels of 35-40% hypoglycosylation as compared with 20-25% in wild type cells (Fig. 6B  290 and D compared to Fig. 6A). There was an effect of adding G6-P or DTT which reduced the 291 hypoglycosylation to wild-type levels seen in the absence of reducing agent (Fig. 6D). In STT3B knockout cells, not only is the STT3B catalytic activity abolished, but these cells have also lost 293 expression of the STT3B-specific subunits MagT1/TUSC3 (Cherepanova and Gilmore, 2016). To see 294 if there is a direct contribution of MagT1/TUSC3, we used SP cells derived from MagT1/TUSC3 KO 295 cells for in vitro translation of hemopexin. Like with the STT3B KO cells, hemopexin 296 hypoglycosylation was more pronounced in the absence of G6-P and, when either G6-P or DTT was 297 included during translation, hypoglycosylation was reduced to the levels seen in wild type cells in the 298 absence of G6-P (Fig. 6B, lanes 4-6 and Fig. 6D). Thus, efficient glycosylation of hemopexin N187 299 is dependent on the STT3B complex and specifically the TUSC3 or MagT1 subunits.