ER chaperones use a protein folding and quality control glyco-code

SUMMARY

N-glycans act as quality control tags by recruiting lectin chaperones to assist protein maturation in the endoplasmic reticulum. The location and composition of N-glycans (glyco-code) are key to the chaperone selection process. Serpins, a class of serine protease inhibitors, fold non-sequentially to achieve metastable active states. Here, the role of the glyco-code to assure successful maturation and quality control of two human serpins, alpha-1 antitrypsin (AAT) and antithrombin III (ATIII), is described. We find that AAT, which has glycans near its N-terminus, is assisted by early lectin chaperone binding. In contrast, ATIII, which has more C-terminal glycans, is initially helped by BiP then later by lectin chaperones mediated by UGGT reglucosylation. UGGT action is increased for misfolding-prone disease variants, and these clients are preferentially glucosylated on their most C-terminal glycan. Our study illuminates how serpins utilize N-glycan presence, position, and composition to direct their proper folding, quality control and trafficking.

eTOC Blurb

N-glycans act as quality control tags by recruiting endoplasmic reticulum lectin chaperones to assist protein maturation. The location and composition of N-glycans (glyco-code) are key to the molecular chaperone selection process. Our study illuminates how serpins utilize N-glycan presence and location to direct their proper folding, quality control and trafficking.

Graphical Abstract

INTRODUCTION

One third of the human proteome is targeted for secretion by the presence of an endoplasmic reticulum (ER) signal sequence ¹. Secretory proteins are folded, modified and assembled in the ER prior to trafficking to their destinations ². To support the efficient maturation of these structural diverse clients, the ER is temporally and spatially organized. Clients enter co-translationally through the Sec61 translocon found in the rough ER where an assembly line of luminal and membrane factors await their arrival ^3–5. As the nascent chain matures and moves further into the ER, it is subjected to several stages of quality control that assesses the structural integrity of the product and its readiness to exit the ER. Proteins that are not exit-ready are retained to provide further time to achieve their native conformation. After ample chances for proper folding have been exhausted, non-native cargo is targeted for destruction to maintain cellular homeostasis.

The vast majority of secretory pathway clients are N-glycosylated in the ER ^6,7. N-glycans act as protein maturation and quality control tags; they serve as reporters of the fitness of the glycoprotein and aid in the trafficking of the cargo ^2,7. These glycans recruit protein folding and quality control factors, which in turn facilitate glycoprotein maturation and trafficking in the early secretory pathway. N-glycans are primarily attached in mammalian cells co-translationally by the oligosaccharyltransferase (OST) associated with the Sec61 (OST-A; Figure 1A, step 1) ^8,9. A subset of N-glycosylation sites skipped by the OST-A complex are modified by a second OST complex (OST-B) that is not associated with the translocon and catalyzes post-translational N-glycoyslation. The OSTs transfer a pre-assembled Glc₃Man₉GlcNAc₂ glycan to the consensus sequence N-X-T/S (X ≠ P). The glycan is then remodeled by a series of glycosidases and transferases along the secretory pathway such that the maturation status of the protein is indicated by its N-glycan composition ^2,7,10–12.

Figure 1. Efficient secretion of AAT and ATIII in HEK293 cells requires their N-glycans.

A) See text for explanation of the individual steps of the lectin chaperone substrate binding cycle. All cartoons are created with BioRender.com. B) Non-sequential folding of serpins. Efficient folding of serpins is accomplished by initial folding of the C-terminal region (green), resulting in a solvent exposed RCL (purple), before N-terminal region (orange) fully folds resulting in active folded protein ²⁸. C) Features and modifications of AAT and ATIII. AAT and ATIII contain N-terminal cleavable signal sequences (grey box). AAT contains three glycosylation sites (red) and one free Cys (yellow). ATIII contains four glycosylation sites and six Cys paired into three disulfides bonds. Both AAT and ATIII possess a gate region (green), five β-strands (blue) comprising the central β-sheet. D) Tertiary structures of AAT (PDB: 1ATU) and ATIII (PDB: 1E05) with glycans (red), disulfides (yellow), and features from C denoted. E) Cells were transfected with AAT or ATIII, radiolabeled for 30 min and chased for the indicated times. F) Quantification of secreted AAT and ATIII from D. Standard deviations are displayed for three independent biological replicates for all plots.

More specifically, the Glc₃Man₉GlcNAc₂ glycan is rapidly and sequentially trimmed by glucosidases I and II (GlsI and GlsII) to generate a monoglucosylated protein (Figure 1A, step 2). The monoglucosylated protein is a substrate for the carbohydrate-dependent (lectin) ER chaperones calnexin (CNX; type I membrane protein) and its soluble paralogue calreticulin (CRT) ^2,7. This trimming can commence during translation and ER translocation to support co-translational intervention by the lectin chaperones (Figure 1A, 1x Initial) ^3,13–15. Substrate binding to the lectin chaperones continues until the terminal glucose is trimmed by GlsII (Figure 1A, step 3). The released glycoprotein is then free to fold to its native state. If the protein continues to adopt an incompletely folded or non-native structure, it is recognized by the UDP-glucose: glycoprotein glucosyltransferase (UGGT) and reglucosylated to send it back for additional rounds of lectin chaperone binding and refolding (Figure 1A, step 4, Rebinding) ^2,7. In this respect, UGGT acts as a gatekeeper of the early secretory pathway determining which proteins can proceed towards exit from the ER and which must be retained because they possess non-native elements. Terminally misfolded proteins are eventually extracted from the lectin chaperone binding cycle by a series of mannosidases that further modify glycans and target misfolded proteins for degradation ^16–18. Given the central importance of the lectin chaperone binding cycle to the trafficking of secretory pathway clients, it is crucial to understand how it evaluates and assists maturing proteins in the ER.

Serpins, or serine protease inhibitors, comprise a family of proteins that modulate a wide array of activities inside and outside the cell. Inhibitory serpins are especially sensitive to misfolding because they must achieve a metastable state that stores the potential energy to drive a large conformational change that is essential for their protease inhibitory activity ¹⁹. In addition, the native serpin structure, shared among family members, is topologically complex with two non-sequential discontinuous domains and numerous sequence-distant contacts (Figure 1B). During the normal folding of serpins, an unstructured sequence near the C-terminus called the reactive center loop (RCL; Figure 1B–D, purple loop) is exposed. The RCL acts as bait for the target serine or cysteine proteases, which form a covalent bond with the RCL upon docking. The consequent cleavage of the RCL triggers a large conformational change whereby the RCL inserts as a sixth strand into the large central A beta-sheet and displaces the attached protease approximately 70 Å relative to the serpin body. This process mechanically distorts the protease active site and the resulting protease-serpin conjugate traps the protease acyl-enzyme intermediate ^20,21.

The serpin alpha-1 antitrypsin (AAT) as expressed in E. coli is a robust folder under optimal in vitro experimental conditions ²². Both equilibrium and kinetic refolding studies and folding simulations found that the C- terminal portion of the serpin must fold first for AAT to attain its functionally required metastable state with the RCL exposed (Figure 1B) ^23–27. Folding in the cell provides additional folding and maturation challenges. We have previously shown by monitoring the order of formation of its three intramolecular disulfide bonds, that the serpin antithrombin III (ATIII), which is structurally homologous to AAT, also requires the C-terminus to fold first ²⁸. These results point to a conserved in vitro and in cellulo folding trajectory for serpins to successfully obtain their metastable state.

The efficient folding of some purified serpins when expressed in bacteria and thus without N-glycans suggests that the amino acid sequence without glycan modifications can be sufficient to direct proper folding ^23,27,29. By contrast, here we demonstrate that for successful folding of the serpins AAT and ATIII in the cells, the vectorial arrival of the nascent polypeptide into the ER requires N-glycans. We observe that AAT, which is N-glycosylated in its N-terminal region, engaged the lectin chaperones early, presumably to prevent the premature folding of the N-terminal region and permit the initial folding of the C-terminal region. In contrast, serpins such as ATIII and neuroserpin (NS), which lack N-glycans near their N-termini, engaged BiP to block premature folding of their N-terminal regions. Moreover, we found that early folding of the serpin C-terminus is aided by the absence of N-glycans or lectin chaperone binding sites in this region.

Disease-associated mutations also have a significant impact on glycosylation and lectin chaperone associated quality control. In the case of NS, a C-terminal glycosylation site was found to be modified by the post-translationally acting OST-B complex only in the context of a C-terminal, disease-associate destabilizing mutation. In the later stages of serpin maturation in the ER, reglucosylation of N-glycans closest to the C-terminus by the protein sensing gatekeeper UGGT was abnormally elevated in the presence of disease-associated mutations, offering insight into the underlying mechanism of terminal misfolding and disease pathology. Taken together, our findings on serpin maturation provide new insights into the importance of the glyco-code for secretory proteins to successfully navigate a complex folding landscape. Strikingly, both the location and the timing of glycan modifications dynamically orchestrate chaperone binding and thus the molecular choreography of the maturing protein in such a way that they directly influence the folding mechanism.

RESULTS

N-linked glycosylation is required for AAT and ATIII secretion

The secretion of AAT and ATIII was followed in HEK293 cells using a radioactive pulse-chase. Both serpins were entirely in the cell lysate fraction for the 10-min of chase (Figure 1E, lanes 1–2 and 8–9). After chasing for 30-min, AAT and ATIII became detectable in the media (Figure 1E, lane 10), and the level of secreted serpin increased with time. AAT and ATIII reached a maximum level of secretion of 65% (± 12%) and 51% (± 4%), respectively (Figure 1F). The half-times of secretion between the proteins were similar (63 min AAT; and 68 min ATIII).

We next asked whether N-glycosylation was important for the maturation and secretion of AAT and ATIII. The three (AAT) or four (ATIII) consensus glycosylation sites were mutated to create glycan null versions. All constructs were solely detected in the cell lysate immediately after the pulse with the glycan null constructs migrating faster due to the lack of N-glycan modifications (Figure 2A, lanes 1 and 5). Both glycosylated AAT and ATIII appeared in the media after a 90-min chase while, neither of the glycan-null constructs was detected in the media (Figure 2A, lanes 7–8). The glycan-null AAT accumulated in the detergent insoluble fraction over time, and the ATIII glycan null accumulated in the cell lysate (Figure 2A, lanes 10 and 12). Therefore, both AAT and ATIII require their N-glycans for secretion from cells.

Figure 2. AAT is helped by early lectin chaperone assistance whereas ATIII benefits from later, UGGT directed lectin chaperone binding.

A) Secretion of AAT, ATIII and glycan null variants in HEK cells. AAT or AAT^{glycan null} (ATT^gn, top), or ATIII or ATIII^gn (bottom) were transfected into HEK cells before radiolabeling. AAT, ATIII and glycan null variants were chased before harvesting the fractions. AAT (B) and ATIII (D) was exogenously expressed in ALG6^−/− and UGGT1/2^−/− HEK cells. After incubation, DNJ was added to the indicated plates for 1-hr before pulsing for 30 min and chasing for indicated times. * and ** represents P ≤ 0.05 and ≤ 0.01, respectively.

The lectin chaperone cycle differentially assists AAT and ATIII maturation

To determine whether binding to the carbohydrate-dependent lectin chaperones plays a role in serpin maturation and secretion, secretion was monitored in wild type (WT) HEK293 cells in the absence or presence of the GlsI and GlsII inhibitor, n-butyldeoxynojirimycin (DNJ) using a pulse-chase approach. DNJ prevents the trimming of the nascent triglucosylated N-glycans thereby preventing interactions with lectin chaperones (Figure S1A)^30,31. Due to the accumulation of the triglucosylated glycoform caused by glucosidase inhibition (Figure 2B and 2D, lanes 1–2), AAT and ATIII synthesized in the presence of DNJ treatment migrated slower than protein in untreated cells. After 2-hr of chase with DNJ treatment, 58% of AAT arrived in the media as compared to 72% without DNJ (Figure 2B, lane 7–8 and 2C). For ATIII secretion into the media, treatment with DNJ caused a reduction from 42% to 10% (Figure 2D, lanes 3–4 and 7–8, and 2E). Therefore, lectin chaperones assist with the efficient secretion of both AAT and ATIII.

Glycoproteins synthesized in the ALG6^−/− cells attach a Man₉GlcNAc₂, rather than the mature Glc₃Man₉GlcNAc₂ carbohydrate (Figure S1B) ^32–34, and thus lectin chaperone binding can only occur in ALG6^−/− cells if the protein is glucosylated by UGGT. Conversely, proteins synthesized in the UGGT1/2^−/− cells can undergo only a single early round of lectin chaperone binding initiated by the trimming of the two outer glucoses to generate a monoglucosylated protein (Figure S1C). With UGGT absent, once the remaining glucose is removed by GlsII, the protein cannot rebind CNX/CRT and re-enter the cycle. Using these cell lines, the impact of the different stages of lectin chaperone binding on AAT and ATIII secretion was determined.

AAT secretion diminished from 80% in WT cells to 58% in the ALG6^−/− cells after 2-hr of chase, which is equivalent to the level observed in WT cells upon treatment with DNJ (Figure 2C). In contrast, the secretion efficiency of AAT was not affected by the absence of the UGGTs. In untreated UGGT1/2^−/− cells, 76% of total AAT was secreted, compared to 72% in WT cells. Regardless of whether UGGT is present, AAT was efficiently secreted into the media. The decrease in AAT secretion seen in WT cells treated with DNJ therefore results from an inability of AAT to bind lectin chaperones at the initial stage of chaperone binding rather than rebinding stages directed by UGGT reglucosylation. In addition, AAT in the media was largely active demonstrating the integrity of protease inhibitor folding (Figure S2A–B). Taken together, these results support the idea that AAT does not require the persistent lectin chaperone binding throughout late stages of maturation that is facilitated by the UGGTs. However, early chaperone intervention mediated by the initial glucosidase trimming of N-glycans significantly contributes to the efficient folding and secretion of AAT.

By contrast, in ATIII transfected ALG6^−/− cells, where initial lectin binding was inhibited, no significant change in secretion was observed when compared to WT cells (Figure 2D–E, 37%). However, in UGGT1/2^−/− cells, ATIII secretion fell to 21% of total ATIII, a ~50% reduction in secretion when compared to WT cells. Treatment with DNJ further lowered the amount secreted from UGGT1/2^−/− cells to 6%. DNJ treatment of ALG6^−/− cells traps proteins in a monoglucosylated state if they are recognized and modified by UGGT (Figure S1B). This is expected to lead to persistent lectin chaperone binding and subsequent trapping of the protein in the cell, which was observed for both AAT and ATIII (Figure 2B–E, lanes 15–16). Altogether, these results indicate that initial lectin chaperone assistance is the key for efficient AAT secretion while lectin chaperone rebinding directed by UGGT was necessary for efficient ATIII maturation.

BiP binds ATIII and NS but not AAT

While AAT and ATIII are similar with regard to their structure and function, one main difference is the positioning of glycans within the protein. Unlike ATIII, which contains no glycan modification sites in its N-terminal 50 amino acids, AAT is glycosylated on N46 (Figure 3A). As noted above, the lectin chaperones assist the efficient folding and secretion of AAT early, but initial lectin chaperone interaction does not play a role for ATIII. It was previously proposed that proteins that lack an N-glycan in approximately the first 50 amino acids after the signal peptide cleavage site have the opportunity to engage BiP, the ER Hsp70 family member, prior to being passed over to the lectin chaperones ¹⁴. Based on the position of the glycans, this model predicts that BiP could bind ATIII but not AAT. To assess BiP binding to AAT and ATIII, we used BiP^Flag knock-in HEK293 cells ³⁵.

Figure 3. BiP binds ATIII, NS and NS-G376E but not AAT.

A) N-terminal alignment of AAT, ATIII and NS with glycan locations indicated (red) and G376E mutation in NS (brown). Gold box (early chaperone selection zone) indicates the first 50 residues of each mature serpin. Proteins with and without N-glycans within this box are predicted to bind lectins and BiP, respectively. B) AAT and ATIII were exogenously expressed in HEK cells possessing BiP^FLAG. Cells were pulsed for 30 min. Lysates were incubated in the presence of ADP/apyrase, or ATP/MgCl₂. Lysates split between αAAT, αMyc, or αFLAG antibodies. C) Quantification of percent AAT or ATIII bound to BiP. Total protein was calculated by quantifying the amount of AAT precipitated from αAAT beads (lanes 4 and 6) and αMyc for ATIII (lanes 8 and 10). **** indicated P ≤ 0.0001. D) NS and NS-G376E were expressed, processed, and analyzed as indicated in B. Both WT NS and the G376E variant were immunoprecipitated using an αMyc antibody. E) Quantification of percent NS or NS-G376E bound to BiP from D. *, **, and *** indicated P ≤ 0.05, 0.01, and 0.001, respectively.

The BiP^Flag cells were transfected with either AAT or ATIII, and labeled with [³⁵S]-Met/Cys. A significant fraction of ATIII co-immunoprecipitated with BiP when ADP/apyrase was included in the lysis buffer to trap interactions (80%, Figure 3B, lane 9 and 3C). This interaction was substantially reduced when ATP was present indicating that the interaction was adenine nucleotide-dependent (Figure 3B, lane 11). By contrast, very little AAT co-immunoprecipitated with BiP (Figure 3B, lanes 5 and 7). These results demonstrate compellingly that ATIII interacts with BiP while AAT does not and supporting the hypothesis that the folding of the N-terminus of ATIII may be delayed by early BiP intervention in order to permit early C-terminal folding while AAT appears to use the lectin chaperones for this purpose.

To test the hypothesis that the absence or presence of N-glycans in the first 50 amino acids of the mature sequence dictates initial chaperone selection, glycan positioning on other secretory serpins was analyzed (Figure S3). The first N-glycan on NS is found at N141, therefore it is predicted to bind to BiP. A G376E missense mutation at the C-terminus of NS is associated with neurodegenerative dementia ³⁶. This mutation in the early folding C-terminal region causes misfolding and ER retention.

BiP bound efficiently to WT NS when apyrase/ADP was present and the binding level increased significantly with the ER-retained NS-G376E variant (Figure 3D, lanes 4 and 8, and 3E). The efficient binding of BiP to both ATIII and NS, and not AAT, is consistent with the hypothesis that N-glycans in the first ~50 amino acids, termed the N-terminal chaperone selection zone, determines the initial chaperone engaged by the nascent protein. This model predicts if there is no glycan in the first 50 amino acids of the N-terminal portion of the mature client sequence, then BiP would initially bind the client; whereas if an N-glycan is present in this region, the lectin chaperones would be the first to intervene.

The C-terminal early folding region of NS is glycosylated post-translationally when misfolded

The folding of serpins to their active state requires the C- terminus to fold early helping to hold the RCL in its solvent exposed active position ^{24,27,28,37,38}. In addition to the delay in folding of the N-terminal region facilitated by either the lectin chaperones or BiP, serpin folding is predicted to be further helped by the absence of chaperone binding sites in the C-terminal region so that it efficiently folds early. Analysis of the C-terminal regions of the human secretory serpins found that only three of fifteen have a glycan in their final 65 amino acids (Figure S4). This is the zone, which we term the C-terminal lectin chaperone exclusion zone where glycans are expected to be added post-translationally by the OST-B complex and therefore are unlikely to impact early folding events ⁸. Although NS possesses an N-linked consensus site at N385, nine amino acids away from its C-terminus (Figure 4A), this site is not glycosylated. However, following NS WT and NS-G376E expression in HEK293 cells ³⁶, the G376E mutant migrated more slowly than WT NS due to glycosylation of the N385 site (Figure 4B, lanes 1 and 3). Using a small molecule inhibitor (C19) of the catalytic subunit of OST-B ³⁹ to determine whether the additional glycosylation site recognized with NS-G376E was modified by OST-B, we found that mobility was reduced for NS-G376E and there were no effects on WT NS (Figure 4B, lanes 2 and 4). Furthermore, when NS-G376E was expressed in STT3B^−/− cells, it migrated similarly to WT-NS, which has two N-glycans modified further supporting additional glycosylation of NS-G376E by OST-B (Figure 4C, lanes 1 and 5).

Figure 4. Destabilization of the C-terminus of NS results in nearby glycosylation site being recognized by the OST-B.

A) Features and glycosylation sites for NS as described in Figure 1B. B). NS or G376E disease mutants were exogenously expressed in HEK cells. C19 (STT3B inhibitor) was added to indicated plates (lanes 2 and 4). NS^3G and NS^2G indicates three and two glycans present, respectively. C) WT NS (lanes 1–4) and G376E (lanes 5–8) with individual glycan mutants were expressed in WT HEK (top) or STT3B^−/− cells (bottom). NS^3G, NS^2G, NS^1G indicated the presence of three, two, or one glycan, respectively. A fraction of the lysates was treated with PNGaseF (lanes 9–16, deglycosylated bands indicated by NS^DG). D) WT NS (top) and G376E (bottom) were expressed in WT (lanes 1–8), STT3B^−/− (lanes 9–16) or WT cells in the presence of C19 inhibitor (lanes 17–24) with N385 glycosylation site present or mutated to alanine (N385A). Cells were radiolabeled and chased for indicated times.

To verify which additional N-glycans was being modified with NS-G376E, the glycosylation sites were individually mutated. For WT NS, the mutation of the N141 and N305A increased the mobility of NS regardless of whether they were expressed in WT or STT3B^−/− cells. No shift was observed with the N385A mutation. This indicated that the N141 and N305 sites are modified efficiently by OST-A, the co-translational transferase. In contrast, the slower mobility of NS-G376E was uniformly increased by the mutation of each of the individual sites indicating that all three sites were modified (Figure 4C, lanes 5–8). However, no shift was observed for NS-N385A in the STT3B^−/− cells demonstrating that the OST-B complex was responsible for the modification of N385 in the context of NS-G376E.

Since the timing of glycosylation could impact the accessibility of lectin chaperone binding sites and thus the opportunity of the C-terminal region to fold, the timing of the glycosylation of NS was next analyzed using a radioactive pulse/chase approach. WT NS with the two N-terminal glycans modified efficiently appeared in the media after 90 min of chase regardless of whether STT3B was genetically knocked out or inhibited by C19, or if the C-terminal glycosylation site was mutated (Figure 4D). In contrast, the N385 site was inefficiently recognized with NS-G376E after a 10-min chase. The level of glycosylation increased with further time of chase of up to 90-min unless the third glycosylation site was mutated, or STT3B was inhibited genetically or chemically. The absence of NS-G376E in the media regardless of whether the third glycan was modified or not, demonstrated that ER retention of the mutant was caused by the missense mutation rather than the additional glycan. This supports the model in which delayed or inhibited folding of the C-terminus caused by the G376E mutation permits prolonged exposure of N385 and its post-translational glycosylation by OST-B.

Prolonged glucosylation of ATIII supports the later action of lectin chaperones

A key component of the carbohydrate-dependent chaperone folding and quality control cycle is the ability of a client protein to rebind the lectin chaperones as directed by the protein folding sensor UGGT if the protein is deemed non-native. The results from Figure 2 demonstrated that lectin chaperones assisted AAT early after emergence into the ER lumen and ATIII later in ER maturation. Later assistance by the lectin chaperone cycle is mediated by UGGT. To test this conclusion, an assay was developed to delve into the difference in the timing of glucosylation between AAT and ATIII.

AAT and ATIII were expressed in ALG6^−/− cells and pulsed labeled. Monoglucosylated proteins were measured in ALG6^−/− cells (to ensure that monoglucosylation was a result of UGGT and not glucosidase trimming) by their binding to a recombinantly produced CRT fused to glutathione-S-transferase (rCRT). Newly synthesized and monoglucosylated proteins were trapped using different time windows of DNJ treatment to determine the amount of glucosylation occurring at specific time periods (Figure 5A). Using sequential pulldowns, first with rCRT followed by the anti-serpin antibodies (αSerp, either αAAT or αMyc for ATIII), monoglucosylated AAT and ATIII were isolated from the pool of unmodified serpin. Both AAT and ATIII were detected in each 15 min window (Figure 5B, lanes 1, 3, 5). Monoglucosylated AAT and ATIII were similarly detected when DNJ was added during the pulse in window 1 (Figure 5B, lanes 2, and 5C). However, UGGT modified ATIII was also observed in window 2 after the pulse and 30-min of chase, whereas glucosylation of AAT was not seen at this time (Figure 5B, lanes 4, and 5C). Therefore, ATIII spent more time being modified by UGGT when compared to AAT consistent with the later role of lectin chaperones in assisting ATIII maturation.

Figure 5. AAT and ATIII are modified by both paralogues of UGGT.

A) Overview of trapping monoglucosylated AAT and ATIII with DNJ. AAT and ATIII were expressed in HEK-ALG6^−/− cell and a pulse-chase approach was used. DNJ was added for 15 min to trap monoglucosylated AAT or ATIII with different time windows prior to or during the chase with nonradioactive media. Lysates were split and incubated with either αAAT or αMyc antibodies or incubated with rCRT initially before a subsequent pulldown with either the αAAT or αMyc antibody to isolate monoglucosylated AAT or ATIII, respectively. B) Autoradiograph of lysates from AAT (top) and ATIII (bottom) from DNJ treatment window 1 (lanes 1 and 2), window 2 (lanes 3 and 4) and window 3 (lanes 5 and 6). αSerp indicates immunoprecipitation of serpins with αAAT or αMyc (ATIII) antibodies. C) Quantification of percent glucosylation from B. * indicates P ≤ 0.05. D) Workflow to isolate monoglucosylated proteins from cells. ALG6^−/− cells are incubated with DNJ for 5-hr. Lysate is split between a WCL, a rCRT and a lectin deficient variant of rCRT (rCRT*). E) AAT and ATIII were expressed in ALG6^−/−(lanes 1–3), ALG6/UGGT1^−/− (lanes 4–6), ALG6/UGGT2^−/− (lanes 7–9) or ALG6/UGGT1/2^−/− (lanes 10–12). F) Quantification of percent glucosylation of AAT and ATIII from E in each cell line. * and *** indicates P ≤ 0.05 and 0.001, respectively.

To compare the level of glucosylation of the two serpins, an established assay that measures UGGT activity in cells was used ^33,34,40. AAT and ATIII were overexpressed in ALG6^−/− cells to isolate monoglucosylated proteins modified by UGGT using rCRT (Figure 5D). AAT (23% glucosylation) and ATIII (31% glucosylation) are both substrates of UGGT with ATIII being modified more efficiently (Figure 5E, lane 2 and F).

Two human paralogues of UGGT exist ⁴¹, UGGT1 and UGGT2. To determine if one paralogue of UGGT was responsible for glucosylation, cells that lacked UGGT1, UGGT2, or both in the ALG6^−/− background cells were used ³². When expressed in ALG6/UGGT1^−/− cells, the amount of monoglucosylated AAT decreased by more than a half, while ATIII glucosylation decreased by ~25% (Figure 5E, lane 5 and F). In contrast, when UGGT2 was absent, both AAT and ATIII glucosylation slightly increased (Figure 5E, lane 8 and F). When both UGGT1 and 2 were absent, no glucosylation was observed (Figure 5E, lane 11 and F). While UGGT1 appears to predominantly glucosylate both serpins, UGGT2 does act on AAT and ATIII when UGGT1 is absent although to a lesser extent than UGGT1. Taken together, these results demonstrate that both AAT and ATIII are recognized and modified by UGGT.

Disease-associated variants of AAT and ATIII are more efficiently glucosylated on their most C-terminal glycan

UGGT has been proposed to preferential modify on-pathway clients to focus the actions of lectin chaperones toward rescuable clients ⁴². This conclusion was largely based on studies using purified components. To test this hypothesis in cells and further elucidate the role of UGGT in writing the glyco-code, we asked if disease-associated mutants of AAT and ATIII are modified by UGGT in cells. Four disease-associated mutations were selected for each serpin. These included missense mutants AAT-Z (E342K) and AAT-S (E228V), as these are the most common mutations associated with AAT deficiency, as well as AAT-Siiyama (S53F) and null Hong Kong (NHK) variants (Figure 6A) ^43–45. Four disease variants of ATIII, C430F, Y63C, C128Y, and F229L, were also tested for glucosylation (Figure 6D) ²⁸.

Figure 6. Levels and sites of UGGT modification in AAT, ATIII, and their disease variants.

Location of selected disease mutations for AAT (A): AAT-NHK (Δ), AAT-Z (green), AAT-Siiyama (magenta), and AAT-S (tan) or ATIII (D): ATIII-C7430F (cyan), ATIII-Y63C (orange), ATIII-C128Y (light green) and ATIII-F229L (pink). B) Determining levels of glucosylation for AAT and disease-associated variants. AAT (lanes 1–3), NHK (lanes 4–6), Z (lanes 7–9), Siiyama (lanes 10–12) and S (lanes 13–15) were expressed in in ALG6^−/− cells and processed and analyzed as described in Fig 5E. C) Quantification and calculating percent glucosylation for AAT and disease variants. E) Determining levels of glucosylation for ATIII and disease mutants. ATIII (lanes 1–3), ATIII-C430F (lanes 4–6), ATIII-Y63C (lanes 7–9), ATIII-C128Y (lanes 10–12) and ATIII-F229L (lanes 13–15) were expressed in in HEK-ALG6^−/− cells and processed and analyzed. F) Quantification and calculating percent glucosylation for AAT and disease-associated variants. For (C) and (F), *, **, ***, and **** indicate P ≤ 0.05, 0.01, 0.001, and 0.0001, respectively. G) Workflow to identify site of UGGT reglucosylation for AAT, ATIII and disease mutants. (1) Constructs were expressed in ALG6^−/− before incubating with DNJ. Cells were lysed and incubated with either rCRT (2) or rCRT* (9) to subtract lectin independent binding. Proteins precipitated from rCRT or CRT* were eluted and trypsinized (3). Peptides were incubated with either rCRT or rCRT*. Unbound peptides were washed away using a spin filter (4). Peptides released from rCRT and rCRT* with heat and PNGaseF allowing deamidated peptide to flow through the filter (5). The resulting eluate was labeled with isobaric TMT labels (6) and combined (7) before being analyzed by tandem mass spectrometry (8). Quantifications from rCRT* pathway were subtracted from rCRT. Identification and quantification of UGGT modification sites in AAT (H) and ATIII (I). Data is indicative of two independent biological replicates *, ** and *** indicate P ≤ less than 0.05, 0.01, and 0.001 respectively. † indicates N135 hypoglycosylation site.

ATIII WT was more efficiently glucosylated than WT AAT as observed previously (Figures 5E–F, 6C and 6F). Glucosylation of all AAT and ATIII variants significantly increased when compared to WT (Figure 6C and F). These results showed that commonly observed disease-associated variants of AAT and ATIII were more efficiently modified by UGGT, providing a potential explanation for why these mutants are ER retained and associated with disease phenotypes.

As WT AAT and ATIII along with their disease-associated mutants are clients of UGGT, we next sought to identify which glycans are modified by UGGT. Recent work from our lab has identified and quantitated substrates that are modified by UGGT1 and UGGT2 in modified HEK293 cells ³². This workflow was expanded and combined with a filter aided sample preparation (FASP) protocol to isolate peptides containing the modified glycans ⁴⁶. Here, the monoglucosylated specific lectin rCRT was substituted for the non-specific lectins used in the study by Zelinska, et al to identify consensus sites modified by the OSTs. These enriched peptides were then labeled with isobaric tandem-mass tags (TMT), which allowed for the relative quantitation of the peptide samples (Figure 6G, steps 6–8).

WT serpins and their disease-associated variants were transfected into ALG6^−/− cells and treated with DNJ to enrich for monoglucosylated substrates. The lysates were processed according to the novel glycoproteomic workflow (Figure 6G). Two peptides containing the glycosylation sites N83 and N247 were identified and quantified for AAT (table S1, related to Figure 6H). The peptide containing N46 was not detected (Figure 6H). To ensure that the enrichments were not biased for these sites, the flowthrough from the FASP method was incubated with concanavalin A (conA), a lectin that binds high mannose glycans. When analyzed by mass spectrometry, N46 was detected in the sample. This result suggests that our glycoproteomics method was specific for isolating glycopeptides containing all glycosylation sites and that N46 was not targeted by UGGT. For all five AAT constructs tested N247 was significantly more modified by UGGT compared to N83. As expected, the reporter ion intensity for all four AAT disease-associated variants was larger suggesting they were targeted to a higher degree by UGGT (Figure 6H).

ATIII and its disease-associated mutants were tested in the same manner (table S1, related to Figure 6I). Three of the four glycosylation sites were detected by mass spectrometry, with the hypoglycosylation of N135 under normal physiological conditions ^28,47,48. When quantified N192, the most C-terminal but medial-located glycan, had the highest levels of glucosylation (Figure 6I). This was followed by N96, and finally N155 for ATIII. Similar to AAT, the four disease-associated variants had higher levels of glucosylation for each glycan compared to the WT protein, supporting the previous results (Figure 5). Taken together these data suggest the most C-terminal, but medial glycan, is highly selected and modified by UGGT for both AAT and ATIII. This may indicate that these regions are more exposed thereby supporting efficient UGGT modification and subsequent prolonged lectin chaperone engagement especially for the highly modified AAT mutants.

DISCUSSION

While the cast of characters in the ER quality control pathway is largely known, much less is known about ER chaperone-assisted folding compared to the relatively well-studied chaperone-client interactions in the cytoplasm. The feature of ER folding that is specific to this compartment and the clients that fold there is the presence of N-linked glycans and the connection between their attachment and interactions with ER chaperones. The position and composition of N-linked glycans on secretory pathway clients help direct the folding and quality control of nascent chains in the ER by supporting region-specific binding of the carbohydrate-dependent chaperones, CNX and CRT. The central principle of the ER folding and quality control glyco-code is that the lectin chaperones bind clients possessing monoglucosylated glycans. The lectin chaperones temporally and spatially control the folding of glycoproteins by slowing or sterically hindering the folding of chaperone-bound regions ^3,15,49.

How does this network of N-glycan-directed lectin chaperone interactions facilitate the productive folding of clients? We selected proteins that have intrinsic challenges in their folding as case studies to gain insight into this question. The physiologically important and widespread family of serine-protease inhibitors—the serpins–must overcome several challenges in their folding. Serpin are non-sequential domain proteins and it is important for their C-terminal region to fold before the N-terminal region, to reach their active state ^25,27,28. This folding mechanism enables these proteins to fold to a metastable state and act on protease substrates in a mousetrap-like way. We hypothesized that chaperone binding to the early synthesized N-terminal region would be required to delay its folding and allow the C-terminal region to fold first. Indeed, we found this to be a key feature of serpin folding in the ER. Strikingly, in examining two different serpins, AAT and ATIII, we discovered that they utilize distinct mechanisms to delay folding of the N-terminal regions, both of which rely on important components of the ER quality control pathway. The folding of AAT, which has N-terminal glycans (N46 and N83), is aided early by interactions with the lectin chaperones (Figure 7A, top). In contrast, ATIII (first glycan site at N96) was found to efficiently bind BiP (Figure 7A, bottom). An additional serpin, NS, for which the first glycan consensus sequence is N141, also bound BiP. This result is consistent with the finding that upon viral infection of CHO cells, viral membrane glycoproteins engaged BiP if they lacked a glycan in approximately the first 50 amino acids ¹⁴. Whereas if an N-terminal glycan was present, the client initially associated with the lectin chaperones. This hypothesis has gone untested in human cells with non-pathogenic proteins due to the lack of suitable reagents to follow BiP binding.

We emphasize the utility of a CRISPR BiP^Flag cell line as a means to efficiently isolate BiP clients ³⁵. The level of ATIII binding to BiP exceeded 80% when ATP was depleted, arresting BiP in a high affinity form and trapping bound substrates. Binding was not observed in the presence of ATP indicating its adenine nucleotide dependence. In addition, no binding was observed when BiP^Flag cells were mixed post-lysis with WT HEK cells expressing radiolabeled ATIII, demonstrating that the interaction occurred in the lumen of the ER pre-lysis (Figure S5). Evidently either BiP or the lectin chaperones can serve to delay the folding of the earlier translated N-terminus to permit the C-terminus to fold first. Chaperone binding is not needed for AAT folding in vitro as the C-terminal region is already present in the full length protein and can rapidly fold when folding is initiated ^25,27,29.

To enable the C-terminal region of serpins to fold before the N-terminal region, it would be advantageous to exclude N-glycan modifiable (lectin-chaperone binding) sites in the serpin C-terminal region. This led us to the model in which the C-terminal region is a lectin chaperone-exclusion zone. Supporting the concept of a C-terminal lectin chaperone exclusion zone was the observation that the majority of secretory serpins lack a C-terminal glycosylation consensus sequence (Figure S4). Moreover, for two of the three serpins that have an N-glycan consensus sequence in the C-terminal region, the C-terminal site is known to be unmodified: N329 on plasminogen activator inhibitor I is not utilized ⁵⁰, and N385 in NS is only found to be modified when the protein is destabilized by a nearby disease missense mutation that disrupts folding ³⁶ (7B). N-glycans in the C-terminal 65-amino acids of proteins rapidly enter the lumen after the nascent chain is released from the ribosome, are not available for OST-A glycosylation, and are therefore post-translationally modified by OST-B ⁵¹. Modification of NS-N385 by STT3B (OSTB) was demonstrated by the lack of modification in STT3B^−/− cells or upon treatment with the STT3B specific inhibitor C19 (Figure 4). These results support our model that WT NS has an effective C-terminal lectin chaperone exclusion zone devoid of N-glycans or lectin chaperone binding, which permits rapid folding of this region.

Late interactions or rebinding to lectin chaperones is directed by UGGT reglucosylation. For the WT serpins, the disulfide-containing ATIII was found to be more reliant on later lectin chaperone binding and UGGT modification (Figures 2D–F). Moreover, ATIII reglucosylation was prolonged compared to AAT (Figure 5B–C). This is consistent with the modestly slower secretion observed for ATIII when compared to AAT (Figure 1E and F). The formation of the three intramolecular disulfides in ATIII likely provides an additional quality control hurdle and thus ATIII requires additional time to fold properly, prolonging the exposure of folding intermediates to UGGT. Alternatively, exposure of free thiols for on-pathway folding intermediates might make the client more susceptible to modification resulting in added rounds of lectin chaperone binding.

UGGT has been proposed to favor the recognition of on-pathway molten globule-like folding intermediates, thereby focusing the attention of the lectin chaperone cycle on potentially rescuable clients ⁴². This hypothesis was supported by studies using purified components. However, we found that off-pathway disease-associated variants of both AAT and ATIII were more efficiently modified by UGGT than their WT proteins (Figure 6B, C, E and F). Therefore, in the cellular context, UGGT efficiently glucosylated slower folding or terminally misfolded disease mutants of both AAT and ATIII.

In an advance towards understanding UGGT selectivity, we developed a quantitative glycoproteomics platform to identify reglucosylated sites on UGGT clients. A filtered-based procedure coupled with rCRT as a monoglucosylated specific lectin was used to isolate monoglucosylated glycopeptides, which were then identified by LC-MS/MS (Figure 6G). Interestingly, for AAT the most C-terminal glycan N247 was the only glycan significantly glucosylated in WT AAT and in its disease-associated variants (Figure 6H and 7C). The level of modification increased significantly for the disease-associated variants. UGGT modification was observed for both the N-terminal (N96) and the most C-terminal (N192) glycans for WT ATIII, and the level of glucosylation increased for all mutants (Figure 6I and 7C). Our working hypothesis is that if the early folding C-terminal region of serpins remains exposed due to disruptions in their normal folding trajectory, it is recognized by UGGT as non-native, resulting in the glucosylation of the nearby C-terminal glycan (Figure 7C). This supports additional lectin chaperone intervention. These results do not align with the observation that the AAT-Z glycan at Asn46 is required for persistent binding to calnexin for lysosomal targeting⁵². However, this study relied on the deletion of individual glycosylation sites and monitored protein trafficking and turnover rather than glucosylation.

Figure 7. Model for serpin folding and quality control.

Efficient maturation of serpins is achieved by initial folding of the more C-terminal B and C beta sheets followed by structural consolidation of the N-terminal region. A) Serpins with N-glycans within the N-terminal chaperone selection zone (yellow box) (AAT) utilize lectin chaperones to delay folding (top) while those that are devoid of glycans in this region (ATIII and NS) engage BiP (bottom). Binding delays N-terminal folding, allowing the C-terminus to fold first, tethering the C-terminal of the RCL in a solvent exposed position. Chaperones then dissociate from the N-terminus, so it can fold, and the serpin can achieve its metastable active conformation. B) The C-terminus must be devoid of N-glycans and other chaperone binding sites to remain unencumbered so efficient serpin folding can be achieved. Glycosylation sites within the last 65 amino acids are not recognized by co-translational glycosylation machinery creating the C-terminal lectin chaperone exclusion zone (box). The C-terminus folds within minutes, burying the site and not allowing it to be recognized by the post-translational machinery (top). Destabilization of the C-terminal region by a disease-associated mutation (G376E, brown star) results in inefficient folding, allowing the C-terminal site to be recognized by the post-translational glycosylation machinery (OST-B) (bottom). C) Summary of sites and levels of UGGT modification. The presence of glycan indicates the site was detected by mass spectrometry and glycan size designates the level of glucosylation. Faded glycans were not detected in rCRT precipitation but were detected using a general lectin. D) The non-redundant human proteome taken from Uniprot was segmented to isolate the total soluble proteins and those soluble proteins that contain a signal peptide. The signal peptide containing proteins are further split into those that do not contain a consensus site for an N-linked glycan (blue) or those that do (orange). The glycosylated proteins are further segmented into those that contain an N-terminal glycan within the first 50 amino acid of the mature protein sequence (glycan, pink) and those that do not (purple). Mean and median lengths shown in respective boxes. Histogram of protein length are shown for non-glycosylated (blue) and glycosylated (red) proteins.

How general are our observations about the glyco-code roles in facilitating successful folding of serpins in the ER? We performed a computational analysis of the soluble proteome and found that ~10% of the proteome is comprised of soluble secretory pathway clients that possess an N-terminal signal sequence (Figure 7D). Two-thirds of these proteins have at least one N-linked glycosylation consensus sequence. Moreover, the median length for glycosylated proteins was 402 amino acids while for non-glycosylated proteins it was only 129 amino acids. Therefore, we speculate that the presence and position of N-glycans might be most important in the folding and maturation of larger more complex proteins. One-third of the soluble secretory pathway clients had a glycan in the first 50 amino acids, consistent with early lectin chaperone intervention. In contrast, the other two-thirds which lacked an N-terminal glycan and are therefore presumably early BiP substrates are on average a larger group with a median length of 431 amino acids compared to 308 amino acids for the proteins with a glycan modification site in their first 40 amino acids. This suggests that it is an advantage for the longer more complex clients to engage both the classical Hsp70 and lectin chaperone families to help with their folding.

In summary this study provides vital insights into the orchestration of chaperone action in the ER and for the importance of the glyco-code for specific aspects of protein folding. Using the highly abundant and physiologically significant family of serpins, which have both non-sequential folding pathways and energetically complex folding landscapes due to the requirement that they fold to a metastable state, this study revealed alternative ways in which ER chaperones can facilitate productive folding. Heavy reliance on lectin chaperone binding or alternatively use of the ER Hsp70 system could lead to an “ordered” folding process, key to successful folding outcomes for serpins. We posit that the observations on serpins, which are ~400 residues in length, will foreshadow the mechanisms utilized by most ER folding clients, in particular those that have complex topologies and lengths over 400 amino acids. These are also likely to be pathologically important ER folding clients as their folding is easily disrupted by mutation or deficiencies in the ER quality control system due to stress or other misfolding events that trigger the unfolded protein response. This is an avenue for future work. An additional exciting future direction is the spatial organization of the chaperone actions we have observed. There is no doubt that initial and rebinding chaperone actions will be spatially distinct, with the former closer to the Sec61 translocation apparatus and the latter closer to ER exit sites. We seek to observe directly the spatial arrangement of the different chaperone-binding events we have characterized.

Limitations of the Study

This study focused on using serpins as model proteins given the extensive understanding of their folding pathway. The lectin chaperone clients studied should be expanded to test the generality of the principles established. HEK cells were used because of the wide range of reagents available rather than natural host of hepatocytes where AAT and ATIII secretion is more efficient⁵³. It would be of interest to characterize the utilization of the glyco-code in hepatocytes. Our studies utilized CRISPR edited HEK cell lines. The knockout of genes can have compensatory effects on other protein levels, as we have previously observed with the knockout of UGGT2 increasing the level of UGGT1³².

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, D. N. Hebert (dhebert@biochem.umass.edu).

Materials availability

All reagents produced from this manuscript will be available on request to academic laboratories and non-profitable organizations without any restrictions. All materials are available upon request to the lead contact D. N. Hebert (dhebert@biochem.umass.edu).

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines and culture

HEK293–6E cells were used as the parental line to create all CRISPR/Cas9 edited lines (Tom et al., 2008). The generation of ALG6^−/−, UGGT1^−/−, UGGT2^−/−, ALG6/UGGT1^−/−, ALG6/UGGT2^−/−, and ALG6/UGGT1^−/−UGGT2^−/− cells has been described previously ³². The UGGT1^−/−, UGGT2^−/−, and UGGT1/2^−/− cells were validated by immunoblotting (Figure S6). BiP^Flag knock in HEK293 cells were previously described ³⁵. Cells were cultured in DMEM (Gibco, 11965–092) supplemented with certified 10% fetal bovine serum (Gibco, 16000044). STT3B^−/− cells were cultured in DMEM (Gibco, 11965–092) and supplemented with qualified 10% fetal bovine serum (Gibco, A3160901) ⁸. All cells were cultured at 37 °C, 5% CO₂. Cells were tested for the presence of mycoplasma using a universal mycoplasma detection kit (ATCC, Cat # 30–012K).

METHOD DETAILS

Reagents and plasmid sequences

Antibodies used include: rabbit polyclonal alpha-1 antitrypsin antibody (Dako, A0012), mouse monoclonal anti-myc antibody 9B11 (Cell Signaling, 2276S), mouse monoclonal anti-FLAG antibody (Sigma, F3165), mouse monoclonal anti-GAPDH (EMD Millipore, MAB374), IRDye goat anti-rabbit secondary 680RD (LiCor, 926–68071) and IRDye goat anti-mouse secondary 800CW (LiCor, 926–32210). Chemicals include C19 (N-(5-(morpholinosulfonyl)-2-(pyrrolidin-1-yl)phenyl)benzamide) STT3B inhibitor (Enamine, Z26531254) ³⁹and 1-deoxynojirimycin (DNJ) (Cayman, 10011718). DNA and protein sequences used can be found in Supplementary information (Table S2).

Transfection of AAT, ATIII and disease mutants

Prior to transfection, cells from the indicated cell line were plated into a cell culture dish and incubated for 24 hr at 37 °C, 5% CO₂. The next day, cells were transfected with cDNA using amounts described below based on the size of the plate used.

The DNA was diluted into the appropriate amount of Opti-MEM (Gibco, 31985–062). Polyethyleneimine (PEI) “Max” (MW 40,000) (Polyscience, Inc., 24765) was used as the transfection reagent for all cell lines and was diluted in Opti-MEM. The PEI:Optimem mixture was incubated for 5 min before the indicated amount was added to the tube containing the diluted DNA. This mixture was incubated for an additional 20 min at room temperature before the DNA:PEI mixture was added dropwise to the cells. The cells were grown for an additional 16–18 hr at 37 °C, 5% CO₂ prior to downstream analysis.

Metabolic labeling and immunoprecipitation

Cells were cultured and transfected in a 3-cm dish prior to metabolic labeling. Where indicated 500 μM DNJ was added to the plates and incubated for 1 hr at 37 °C, 5% CO₂. The culture media was removed and the cells were washed twice with 1 mL PBS before [³⁵S]-Met/Cys labeling for 30 min with completed media (1 mL DMEM -L-Met/L-Cys (Gibco, 21013), 10% (v/v) sodium pyruvate (Gibco, 11360), 10% (v/v) GlutaMAX^™ supplement (Gibco, 35050) and 6.25 μL EasyTag Express Protein Labeling Mix, [³⁵S]-Met/Cys (Promega, NEG772007MC) per plate).

After a radioactive pulse the media was removed, and the cells were washed with 1 mL of PBS, before adding normal growth media for indicated chase times. DNJ was added throughout the chase to indicated plates. After chasing the media was collected or removed and the cells were washed once with cold PBS and placed on ice. Cells were lysed using 500 μL of lysis buffer (20 mM MES, 100 mM NaCl, 30 mM Tris-HCl, 0.5% Triton-X 100 (v/v), pH 7.5) completed with Halt^™ Protease Inhibitor Cocktail (Thermo, 1861278) and 20 mM N-ethylmaleimide (NEM). Cell lysates were shaken for 10 min at 4 °C before being centrifuged at 20,000xg for 10 min at 4 °C. Supernatants were saved and, where indicated, 20 μL of 1% SDS in 100 mM Tris-HCl, pH 8 was added to dissolve the insoluble pellet. The insoluble fraction was treated for 10 min at 95 °C before 980 μL of lysis buffer was added to quench the SDS. To the media, lysate, and insoluble fraction 10 μL of Protein-A-Sepharose^® 4B (Invitrogen, 101042) and 1 μL of anti-myc or anti-AAT antibody was added to immunoprecipitate ATIII or AAT, respectively. Samples were incubated at 4 °C overnight.

The next day the tubes were spun at 250xg for 5 min at 4 °C to pellet the resin. The supernatant was removed, and the beads were washed with 500 μL of lysis buffer containing no protease or NEM. The beads were washed twice more. After the final wash the supernatant was removed and 30 μL of gel loading buffer (30 mM Tris-HCl, 9% SDS (w/v), 15% (v/v) glycerol, 0.05% bromophenol blue (w/v), pH 6.8) containing 100 mM dithiothreitol (DTT) was added. The samples were treated for 10 min at 95 °C and shaken vigorously before being resolved on a 9% SDS-PAGE gel. Gels were dried and incubated with a phosphor screen exposure cassette 8” x 10” (Molecular Dynamics, 184093943970) for 2 days before being imaged using a GE Typhoon FLA 9500 phosphorimager (GE Healthcare) and quantified using ImageQuant TL. The percent secreted was calculated by dividing the amount of protein in the media fraction by the total amount of protein (media plus lysate) and multiplying by 100.

BiP co-immunoprecipitation

In a 6-cm plate, 2.25×10⁶ BiP^Flag HEK293 cells were added and grown for 24 hr at 37 °C, 5% CO₂. The cells were then transfected with the indicated cDNA and incubated for 24 hr. The next day the media was removed and replaced with 1.5 mL of completed pulse media containing 12.5 μL of [³⁵S]-Met/Cys and incubated for 30 min. After labeling, the media was removed and 1 mL of cold PBS was added and the cells were scraped off the cell culture dish and added to a 1.5 mL tube. The tubes were then spun to pellet the cells at 10,000xg for 2 min at 4 °C and the supernatant was removed. The cells were then lysed in either 200 μL ADP/apyrase lysis buffer (1% digitonin (w/v), 1x protease inhibitor cocktail, 20 mM NEM, 2.5 mM adenosine di-phosphate (Sigma, A2754), 10 U/mL apyrase (New England Biolabs, M0398) in BiP binding buffer (50 mM Tris, 150 mM NaCl, 1 mM CaCl₂, pH 7.5 buffer) or 200 μL ATP lysis buffer (1% digitonin (w/v), 1x protease inhibitor cocktail, 20 mM NEM, 5 mM ATP (Sigma, A2383), 5 mM MgCl₂ in BiP binding buffer) to trap or inhibit substrate binding to BiP, respectively.

After resuspension, the cells were incubated on ice for 30 min before 800 μL of BiP binding buffer was added. Cells were spun at 20,000xg for 15 min at 4 °C before removing and saving the supernatant. 250 μL of the supernatant was added to 10 μL of Protein-A resin mixed with 1 μL of anti-myc for ATIII or NS immunoprecipitants or 1 μL an anti-AAT antibody to immunoprecipitate AAT. 250 μL of the supernatant was also added to a tube containing 10 μL of Protein-A resin mixed with 1 μL of anti-FLAG antibody to immunoprecipitated BiP and bound substrates. Samples were incubated overnight at 4 °C before washing the resin 3 times with 500 μL BiP binding buffer. Once washed, 30 μL of gel loading buffer was added containing 100 mM DTT. Samples were treated for 10 min at 95 °C and shaken vigorously before being resolved on a 9% SDS-PAGE gel and visualized by autoradiography. The percent bound to BiP was calculated by dividing the amount of serpin in the FLAG immunoprecipitation by the amount precipitated in the anti-myc or anti-AAT pulldown and multiplying by 100.

Endoglycosidase assay

In a 3-cm plate, HEK293–6E or STT3B^−/− cells were plated and transfected with the indicated cDNA. The next day the media was removed and the plate was washed with 1 mL of cold PBS before adding 500 μL of cell lysis buffer completed with protease inhibitor and NEM. The lysates were shaken for 10 min at 4 °C before being centrifuged at 20,000xg for 10 min at 4 °C. 100 μL of lysate was added to a tube containing 11 μL of 10x denaturation buffer (New England Biolabs, 0704) and treated for 10 min at 95 °C. Once cool, 11 μL of 10% NP-40, 11 μL of 10x Glycobuffer 2 and 1 μL of PNGaseF (New England Biolabs, 0704) was added. An untreated control was similarly prepared but 1 μL of Milli-Q water was used instead of PNGaseF.

Samples were incubated at 37 °C for 3 hr before being mixed with 500 μL of cold acetone to precipitate the protein. Samples were kept at −20 °C overnight. The next day, the tubes were spun at 20,000xg for 10 min at 4 °C to pellet the precipitated protein and the acetone was removed. Pellets were dried at 37 °C for 1 hr before 30 μL of gel loading buffer containing 100 mM DTT was added. The samples were treated for 10 min at 95 °C and shaken vigorously before being resolved on a 9% SDS-PAGE gel and transferred to a PVDF membrane (Millipore, IPFL00010) for Western blotting. Proteins were probed using an anti-myc antibody at a 1:1000 dilution and detected by IRDye goat anti-mouse secondary 800CW (1:20,000) when imaged using a Li-Cor. Quantification of immunoblots was done using ImageJ.

Cellular UGGT glucosylation assay

The UGGT glucosylation assay has been previously described ³³. Briefly, the indicated construct was transfected into the specified cell line using a 6-cm plate. The next day the media was removed, and 1.5 mL of fresh media was added containing 0.5 mM DNJ and incubated for 5 hr. After the incubation the media was removed and the plate was washed with 1.5 mL of PBS, and the cells were lysed in 750 μL of cell lysis buffer completed with protease inhibitor and NEM. The lysates were shaken for 10 min at 4 °C and centrifuged at 20,000xg for 10 min at 4 °C before the soluble fraction was removed and saved for later. Meanwhile recombinantly produced calreticulin (rCRT) or a lectin dead variant (rCRT*) ⁵⁶ glutathione coated beads (Cytiva, 17075601) were prepared. 25 μL of resin and 50 μg of either glutathione S-transferase (GST) tagged rCRT or rCRT* was added per pulldown. Beads were mixed with the GST-tagged protein for 3 hr before being added to the lysate to isolate monoglucosylated protein.

The supernatant containing the transfected protein was then split between a whole cell lysate (WCL) fraction (20% lysate) and a rCRT (35% lysate) or rCRT* (35% lysate) pulldown. The WCL was mixed with 750 μL of cold acetone and incubated at −20 °C overnight to precipitate the protein. The next day, the WCL fraction was spun at 20,000xg for 10 min at 4 °C to pellet the precipitated protein and the acetone was removed. Pellets were dried at 37 °C for 1 hr. The rCRT and rCRT* pulldowns were incubated overnight at 4 °C to isolate the monoglucosylated protein. The next day the beads were washed 3 times by first centrifuging the sample at 250xg for 5 min at 4 °C, removing the supernatant, and adding 500 μL of lysis buffer. After the last wash, 40 μL of gel loading buffer containing 100 mM DTT was added to both the WCL, rCRT and rCRT* samples. The tubes were treated for 10 min at 95 °C and shaken vigorously before being resolved on a 9% SDS-PAGE gel and transferred to a PVDF membrane for Western blotting.

AAT and mutants were probed using an anti-AAT antibody (1:1000) and visualized with IRDye goat anti-rabbit secondary 680RD (1:20,000). ATIII and mutants were probed with an anti-myc antibody (1:1000) and detected by IRDye goat anti-mouse secondary 800CW (1:20,000). Both membranes were visualized using a Li-Cor imaging system. Quantification of immunoblots was done using ImageJ. The percent glucosylation was calculated by first normalizing quantified bands by the percent of the input. The signal detected in the rCRT* samples was subtracted from that quantified in the rCRT sample before dividing the resulting number by the amount quantified in the WCL. The result number was multiplied by 100 to obtain a percentage.

Sample preparation for Identification of UGGT glucosylation sites

In three 10-cm plates, ALG6^−/− cells were plated and grown for 24 hr at 37 °C, 5% CO₂. The cells were then transfected with the indicated cDNA according to Table 1 and cultured overnight at 37 °C, 5% CO₂. The next day the media was removed and 5 mL of fresh media containing 0.5 mM DNJ was added to each plate and incubated for 10 hr. The media was removed and the plates were washed with 3 mL of cold PBS before adding 1 mL of cell lysis buffer completed with protease inhibitor. The lysates were pooled and shaken for 10 min at 4 °C and centrifuged at 20,000xg for 10 min at 4 °C before the soluble fraction was removed and saved. Meanwhile rCRT or rCRT* glutathione coated beads were prepared. Per pulldown, 100 μL of glutathione resin and 250 μg of either GST tagged rCRT or rCRT* was mixed for 3 hr at 4 °C before adding to the lysate. The lysate was split evenly between beads coated with rCRT and rCRT* and incubated overnight at 4°C.

Place size (cm)	Number of cells to plate	DNA/plate (μg)	Opti-MEM to dilute DNA (μL)	PEI/plate (μg)	Opti-MEM to dilute PEI (μL)	DNA:PEI to add to cells (μL)
3.5	8.00×105	2.4	40	4.8	40	80
6	1.75×106	4.8	80	12	80	160
10	3.50×106	8	200	20	200	400

The next day the beads were washed 3 times by centrifuging the sample at 250xg for 5 min at 4 °C, removing the supernatant, and adding 500 μL of cell lysis buffer. After washing, 100 μL of elution buffer (50 mM triethylammonium bicarbonate (TEAB) (Themo, 90114), 0.1 % (w/v) deoxycholate (Sigma, 30970), 10 mM DTT, pH 8.3) was added. The samples were treated for 10 min at 95 °C, shaken for 5 min and centrifuged at 250xg for 5 min. The supernatant was removed and saved. The elution process was repeated 2 more times before alkylating the pooled elutions with 20 mM iodoacetamide (Sigma, I6125). The protein elution was digested with 2.5 μg of Trypsin Gold (Promega, V5280) overnight with gentle agitation at 37 °C.

After digestion, the amount of peptide was quantified using the Pierce Quantitative Peptide Assay (Thermo, 2375). The remaining trypsin was inactivated by treating the sample for 10 min at 95 °C. The peptide solution was diluted 10x with 0.5 % (w/v) 3-((3-cholamidopropyl) dimethylammonio)-1-propanesulfonate (CHAPS) (Sigma, C3023) in HEPES buffered saline (50 mM HEPES, 200 mM NaCl, pH 7.5). 250 μg of rCRT or rCRT* was added to the corresponding tube to isolate monoglucosylated glycopeptides and incubated at 4 °C overnight.

The bound peptides were added to a Microcon-30 centrifuge filter (Millipore, MRCF0R030) and centrifuged at 14,000xg for 10 min at 4 °C until concentrated to 100 μL. The sample was washed 5 times with 400 μL of 50 mM TEAB buffer and centrifugation at 14,000xg for 10 min at 4 °C. The flow through was discarded. After the final wash the sample retained above the filter was collected and treated for 10 min at 95 °C before cooling to room temperature. The glycans were then removed using PNGase (New England Biolabs, P0704) and a 37 °C overnight incubation according to the manufacturer’s protocol. After removal of the glycans, the sample was added to a fresh Microcon-30 spin filter and centrifuged at 14,000xg for 10 min. The filter was washed twice with 25 μL of TEAB, and the flowthrough was collected, and the peptide amount was quantified. Equal amounts of peptides from each condition were labeled with TMT according to the manufacturer’s protocol (Thermo, 90309). The reaction was quenched by adding 2 μL of 5% hydroxylamine (Thermo, 90115). The labeled peptides were pooled and enriched using C18 tips (Thermo, 87784) prior to drying and analysis by mass spectrometry.

Analysis of samples by mass spectrometry

LC-MS analysis was performed, modified from an existing method using an Easy-nLC 1000 nanoLC chromatography system interfaced to an Orbitrap Fusion mass spectrometer (Thermo Scientific)⁵⁷. Samples were pre-concentrated and desalted on to a C18 trap column prior to separation over a 180-min gradient from 0% to 50% mobile phase B (A:0.1% formic acid in water, B:0.1% formic acid in acetonitrile) at a 300 nL/min flow rate with a 75 μm x 15 cm PepMap RLSC column (Thermo). Mass spectrometry parameters were as follows: ion spray voltage 2000V, survey scan MS1 120k resolution with a 2 s cycle time, interleaved with data-dependent ion trap MS/MS of highest intensity ions with charge state-dependent ion selection window (z=2:1.2 m/z, z=3:0.7 m/z, z=4–6:0.5 m/z) and collision induced dissociation (CID) at 35% normalized collision energy (NCE). Additionally the top 5 (z = 2) or 10 (z > 2) product ions were synchronously selected for HCD MS³ at 65% NCE with Orbitrap detection to generate TMT tag intensities ⁵⁷.

RAW data files were analyzed in Proteome Discoverer 2.4 (Thermo Scientific) using the SEQUEST search algorithm against the Homo sapiens (SwissProt TaxID=9606) database downloaded from uniprot.org. The search parameters used were as follows: 10 ppm precursor ion tolerance and 0.4 Da fragment ion tolerance; up to two missed cleavages were allowed; and dynamic modifications of methionine oxidation, N-terminal acetylation, and asparagine deamination. Peptide matches were filtered to a protein false discovery rate of 5% using the Percolator algorithm. Peptides were assembled into proteins using maximum parsimony and only unique and razor peptides were retained for subsequent analysis. Protein quantitation based on TMT ion abundance was performed using a co-isolation threshold of 75% and synchronous precursor selection match threshold 65%. The mass spectrometry data was generated by UMass Amherst Mass Spectrometry Core Facility, RRID:SCR_019063.

AAT activity assay

3.5-mm plates were seeded with 7.5×10⁵ HEK293, ALG6^−/−, or UGGT1/2^−/− cells for 24-hr at 37 °C, 5% CO₂. Cells in each plate were then transfected according to table above with AAT cDNA. The media was collected 20-hr post-transfection and cleared by centrifuging at 14,000xg for 5 min at 4°C and split into two fractions. To one fraction, 1 μL of human neutrophil elastase (HNE, Athens Research Technology,16–14-051200) (1.27 mg/mL) was added and incubated at 37 °C for 30 min. To the other fraction, no HNE was added to serve as a control to observe where AAT migrated on the blot. Samples were precipitated with 10% (vol/vol) TCA and spun at 14,000xg for 5 min at 4°C. The pellets were washed with 500 μL of cold acetone before dissolving with running buffer containing DTT and resolving by 9% reducing SDS/PAGE. Immunoblots were processed using a primary αAAT and αrabbit secondary. The AAT and HNE-AAT conjugate was detected at ~55 and 80 kDa, respectively. Band intensities were quantified using Image Studio Ver 4.0 from Li-Cor.

Analysis of human secretory pathway clients and N-glycan positioning

The reviewed human proteome from Uniprot (accessed 11–12-22) was used as the basis for the human proteome. Proteins containing a transmembrane region as per Uniprot accessions were removed from the data set, followed by those that do not contain a designated signal sequence. N-linked glycans were identified by the presence of an N-glycan acceptor site corresponding to N-X-T/S where X≠P. The first 50 amino acids were assessed based on the sequence of the mature protein. All analysis was done using R Studio.

QUANTIFICATION AND STATICAL ANALYSIS

For quantification of autoradiographs and Western blots standard deviation (S.D) is displayed and is representative of three independent biological replicates (n). TMT mass spectrometry data is representative of two independent biological replicates with standard deviation displayed. Prism v9 was used for all quantifications and statistical analyses. Statistical significance was determined by using an unpaired t-test with a minimal confidence interval of 95%. Levels of significance denoted in the figure legends and data presented is the average of all biological replicates tested.

Supplementary Material

3

Supplemental table S1. TMT mass spectrometry results for AAT, ATIII and disease mutants, related to Figure 6H and I.

For each protein the percent coverage, number of peptides and number of PSMs are provided. The identified peptides are noted in column G with the associated peptide modifications displayed in column H. For each glycosylation site, the most abundant peptide is noted in column G. TMT quantification values for the detected glycan from each substrate are noted in columns I through R and were used for analyses in Figure 6. Signal from alternative peptides containing glucosylation sites has been added.

4

Supplemental table S2. List of DNA and protein sequences, related to STAR Methods.

Disease-associated mutations are denoted in red with C-terminal linkers and affinity tags in green and blue, respectively.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Myc-Tag (9B11) mouse mAb	Cell Signaling Technologies	Cat# 2276 RRID:AB_331783
Monoclonal ANTI-FLAG® M2 mouse mAb	Millipore Sigma	Cat# F1804 RRID:AB_262044
Rabbit anti-alpha-1 antitrypsin pAb	Agilent	Cat# A0012 RRID:AB_2335672
Rabbit anti-UGGT1 pAb	GeneTex	Cat# GTX66459 RRID: AB_3068550
Rabbit anti-UGGT2 pAb	Novus Biologicals	Cat# NBP2–20803 RRID:AB_3068551
GAPDH (6C5) mouse mAb	EMD Millipore	Cat# MAB374 RRID:AB_2107445
Bacterial and Virus Strains
NEB® 10-beta Competent E. coli (High Efficiency)	New England Biolabs	C3019H
Chemicals, Peptides, and Recombinant Proteins
GST-CRT	Baksh and Michalak, 199154	N/A
GST-CRT*	Adams, et al, 202032	N/A
Human neutrophil elastase	Athens Research Technology	Cat# 16–14-051200
PNGaseF	New England Biolabs	Cat# P0709
DNJ: 1-deoxynojirimycin (hydrochloride)	Cayman Chemicals	Cat # 10011718
C19: N-(5-(morpholinosulfonyl)-2-(pyrrolidin-1-yl)phenyl)benzamide	Enamine	Cat# Z26531254
EasyTag™ 35S Protein Labeling Mix, [35S]	PerkinElmer	Cat #NEG772007MC
Critical Commercial Assays
TMT10plex™ Isobaric Label	Thermo Scientific	Cat# 90309
Experimental Models: Cell Lines
HEK293-EBNA1–6E	National Research Council, Canada	RRID:CVCL_HF20
HEK293-EBNA1–6E ALG6−/−	Narimatsu et al. 201855	N/A
HEK293-EBNA1–6E UGGT1/2−/−	Adams, et al, 202032	N/A
HEK293-EBNA1–6E ALG6/UGGT1−/−	Adams, et al, 202032	N/A
HEK293-EBNA1–6E ALG6/UGGT2−/−	Adams, et al, 202032	N/A
HEK293-EBNA1–6E ALG6/UGGT1/2−/−	Adams, et al, 202032	N/A
HEK293-Flp-In T-Rex BiPFLAG+/+	Sun et al. 202332	N/A
Software and Algorithms
ImageJ v1.5	National Institutes of Health	https://imagej.nih.gov/ij/download.html
Prism v9	GraphPad	N/A
Proteome Discoverer v2.4	Thermo Scientific	N/A

Highlights.

• N-glycans are required for alpha-1 antitrypsin and antithrombin secretion

• N-glycan positioning and composition (glyco-code) determines chaperone selection

• Temporal and region-specific chaperone binding directs glycoprotein folding

• UGGTs glucosylate serpin disease variants at specific sites more efficiently