Mammalian cells lacking either the cotranslational or posttranslocational oligosaccharyltransferase complex display substrate-dependent defects in asparagine linked glycosylation

Asparagine linked glycosylation of proteins is an essential protein modification reaction in most eukaryotic organisms. Metazoan organisms express two oligosaccharyltransferase complexes that are composed of a catalytic subunit (STT3A or STT3B) assembled with a shared set of accessory subunits and one to two complex specific subunits. siRNA mediated knockdowns of STT3A and STT3B in HeLa cells have shown that the two OST complexes have partially non-overlapping roles in N-linked glycosylation. However, incomplete siRNA mediated depletion of STT3A or STT3B reduces the impact of OST complex loss, thereby complicating the interpretation of experimental results. Here, we have used the CRISPR/Cas9 gene editing technology to create viable HEK293 derived cells lines that are deficient for a single catalytic subunit (STT3A or STT3B) or two STT3B-specific accessory subunits (MagT1 and TUSC3). Analysis of protein glycosylation in the STT3A, STT3B and MagT1/TUSC3 null cell lines revealed that these cell lines are superior tools for investigating the in vivo role and substrate preferences of the STT3A and STT3B complexes.

Scientific RepoRts | 6:20946 | DOI: 10.1038/srep20946 is associated with the protein translocation channel 6,9 , and mediates cotranslational glycosylation of NXT/S sites on nascent polypeptides 10,11 . The analysis of glycosylation in STT3B depleted cells indicates that the major cellular role for the STT3B complex is to maximize sequon occupancy in glycoproteins by modification of sites that are skipped by the STT3A complex 11 . Expression of both OST complexes in human cells is necessary to achieve full glycosylation of the N-glycoproteome and is essential for normal human health and development as exemplified by the recent discovery of patients with novel forms of congenital disorders of glycosylation-I (CDG-1) that are caused by mutations in the STT3A or STT3B genes that reduce, but do not eliminate, STT3A or STT3B expression 12 .
MagT1 and TUSC3 are orthologues of the yeast Ost3 and Ost6 proteins 13 . Human STT3B complexes contain either MagT1 or TUSC3 5 , just as the yeast OST complex contains either Ost3 or Ost6 [14][15][16] . The structures of the lumenal domains of Ost6 and TUSC3 have been solved revealing a thioredoxin fold with an active site CXXC motif that is required for function 17,18 . Mutations in the human MagT1 gene cause X-linked immunodeficiency syndrome (XMEN) 19 , but are no longer thought to be linked to X-linked intellectual disability (XLID) 20 . Mutations in TUSC3 cause non-syndromic autosomal recessive mental retardation (ARMR) a disease characterized by intellectual impairment [21][22][23] .
Pulse-chase labeling of human glycoproteins in siRNA treated HeLa cells revealed important information about STT3A-and STT3B-dependent glycosylation sites. Acceptor sites that are skipped at a high frequency by the STT3A complex include sequons adjacent to the signal sequence cleavage site 11 and sequons located within the last 50 residues of proteins 24 . Closely spaced NXS acceptor sites 25 , and certain internal acceptor sites with sub-optimal sequons 26 including a subset of acceptor sites that have internal cysteine residues (NCT/S sites) are also skipped by the STT3A complex 5 . Local sequence features that promote site skipping by the STT3A complex remain poorly defined.
Although siRNA mediated depletion of STT3A, STT3B or MagT1 has been a valuable tool for glycosylation analysis; there are several major limitations. The reduction in STT3A or STT3B protein levels seldom exceeds more than 4-fold in HeLa or CHO cells 5,11,27 presumably due to the long half-life of OST subunits (> 24 h 28 ). Incomplete depletion of the STT3A or STT3B complexes almost certainly reduces the extent of protein hypoglycosylation, thereby hindering the interpretation of experimental results. Glycosylation of many NXT/S sites in glycoproteins is not reduced by siRNA-mediated depletion of STT3A or STT3B 11,25 . It is not clear whether these insensitive acceptor sites can be modified efficiently by either OST complex, or whether these sites are simply not susceptible to incomplete depletion of STT3A or STT3B. Notably, simultaneous siRNA mediated knockdown of STT3A and STT3B does not eliminate N-glycosylation of any substrate we have tested to date due to incomplete depletion of STT3A and STT3B. Finally, cell lines that divide less rapidly than HeLa cells will likely display lower extents of siRNA mediated depletion of STT3A or STT3B, further limiting the utility of siRNA mediated knockdowns.
Genome editing in human cells provides a powerful tool for creating knock-out cell lines. Clustered regulatory interspaced short palindromic repeat (CRISPR)/Cas9-based technology has been widely applied due to its affordability and ease of use [29][30][31] . In the present study, we describe the generation and initial characterization of HEK293-derived cell lines that do not express one of the OST catalytic subunits (STT3A or STT3B) or do not express one or both of the accessory subunits (MagT1 and TUSC3) of the STT3B complex. We have compared protein hypoglycosylation in these cell lines to siRNA treated HEK293 and HeLa cells. Human cells with null mutations in STT3A or STT3B, or complex specific accessory subunits will provide a superior model system for further investigation of the in vivo role of the STT3A and STT3B complexes in protein glycosylation.

Results and Discussion
siRNA knockdowns of OST subunits in HEK293 cells is relatively ineffective. A 72 h treatment of HEK293 cells with our previously validated siRNAs for human STT3A, STT3B or MagT1 caused roughly a 2-fold reduction in expression levels of the targeted proteins as detected on protein immunoblots (Fig. 1a). OST subunit depletion is likely limited by the long half-life of OST subunits 28 . Expression of the catalytic subunits (STT3A or STT3B), shared OST subunits (e.g., ribophorin I), or the protein translocation channel (Sec61) was not reduced by treating cells with the MagT1 siRNA (Fig. 1a,b). Depletion of STT3B was accompanied by a reduction in MagT1 expression as observed previously in HeLa cells 5 . MagT1 mRNA is widely expressed in human tissues while TUSC3 mRNA expression is more restricted 21,32 . The cultured cell lines we have tested to date (HeLa, CHO, HepG2, U97, NSC-34 and human fibroblasts) express MagT1 protein, but do not express sufficient TUSC3 to readily detect by protein immunoblotting using anti-TUSC3 sera 5 , (Cherepanova, Shrimal and Gilmore, unpublished observations). Lack of detectable TUSC3 in the HEK293 cells is particularly remarkable given that the anti-TUSC3 sera cross-reacts strongly with endogenous MagT1 (Fig. 1b) due to the high amino acid sequence identity between MagT1 and TUSC3 4,5 . The siRNA mediated reduction in MagT1 expression in HEK293 cells, but not in HeLa cells 5 , was accompanied by the appearance of an anti-TUSC3 immunoreactive product that migrated between MagT1 and transiently expressed TUSC3-DDKHis (Fig. 1b). The protein of intermediate mobility was identified as TUSC3 based upon sensitivity to treatment with TUSC3 siRNA (Fig. 1b).
After 48 hours of siRNA treatment, the HEK293 cells were transfected with expression plasmids for prosaposin (pSAP, Fig. 1c) or sex hormone binding globulin (SHBG, Fig. 1d), the two most extensively validated substrates of the STT3A and STT3B complexes. A pulse-chase labeling experiment revealed a 20% reduction in pSAP glycosylation in cells that were treated with the STT3A siRNA (Fig. 1c). SHBG has two extreme C-terminal glycosylation sites that are glycosylated by the STT3B complex in a MagT1 dependent manner 5,24 . Pulse labeling of SHBG in the HEK293 cells revealed no reduction in glycosylation of SHBG in the cell treated with the MagT1 siRNA presumably due to enhanced stable expression of TUSC3. We detected a minor reduction in SHBG glycosylation in HEK293 cells treated with the STT3B siRNA (Fig. 1d). These results indicate that siRNA mediated depletion of OST subunits in HEK293 cells, unlike HeLa cells, is not efficient enough to markedly reduce Scientific RepoRts | 6:20946 | DOI: 10.1038/srep20946 glycosylation of these well-characterized substrates of the STT3A and STT3B complexes. Additional glycoprotein substrates were not tested in the siRNA treated HEK293 cells given the minor impact of the partial STT3A or STT3B depletions on pSAP and SHBG glycosylation.
Targeting and selection of HEK293 derived cell lines. The CRISPR/Cas9 system was used to generate HEK293 derived cell lines that do not express STT3A, STT3B, MagT1 or TUSC3. Two to three CRISPR targets were selected in the MagT1, TUSC3, STT3A and STT3B genes (Supplementary Table S1). HEK293 cells were transfected with the pSpCas9(BB)-2 A-GFP vector containing sgRNA expression constructs designed to target the MagT1, STT3A and STT3B genes. After 24 h, cells were analyzed by fluorescence activated cell sorting (FACS) to obtain cells that express green fluorescent protein (GFP). Growth of single cells in 96 well plates allowed the isolation of HEK293-derived cell clones that were candidates for the desired knockout cell lines. Protein immunoblot analysis of cell extracts using antisera specific for MagT1, STT3A or STT3B was used as the primary screen to select potential knockout cell lines for further analysis. Cell lines that lacked a detectable protein immunoblot signal for the targeted OST subunit were subjected to a secondary immunoblot screen using antisera that recognize OST subunits that are specific for the STT3A complex (KCP2 6  screened for expression of TUSC3 using protein immunoblots. Target sequences selected for further analysis are indicated with an asterisk in Supplementary Table 1. DNA sequences of targeted OST subunit genes. Potential positive clones for each knockout were expanded and subjected to several assays to characterize the knockout cell lines. Heteroduplexes derived from PCR amplicons flanking the target site were subjected to T7E1 (T7 endonuclease 1) digestion (Fig. 2a). All knockout cell lines displayed heterozygosity at the target site, since cleaved DNA fragments were produced by T7E1 digestion. PCR products from WT HEK293 cells were mixed with PCR products from the mutant, re-annealed and treated with T7E1. The samples showed similar digestion patterns further indicating that the clonal cell lines are heterozygous at the target sites.
DNA sequencing of the PCR products (15-35 sequences/cell line) was used to identify the mutations in the knockout cell lines (Fig. 2b). We did not recover the wild type OST subunit sequence from any of the knockout cell lines. It has been shown that HEK293 cells are hypotriploid 33 . The genome of HEK3293 cells contains seven copies of the MagT1 gene, three copies of the STT3A and STT3B genes and two copies of the TUSC3 gene (http:// hek293genome.org/v2/). Sequencing of the PCR amplicons flanking the MagT1 target site revealed three amplification products (Fig. 2b) lacking 6 bp, 11 bp or 17 bp present at a ratio of 7:23:5 among the 35 sequences that were obtained (Fig. 2b). Sequencing of the STT3A(− /− ), STT3B(− /− ) and MagT1(− /− )TUSC3(− /− ) cell lines identified two independent alleles for the STT3A, STT3B and TUSC3 genes (Fig. 2b). The observed number of alleles for MagT1, STT3A and STT3B could be an underestimate of the full spectrum of indels since the PCR-based method would not identify large insertions and deletions, inversions or structural aberrations 34 . For example, one of the sequenced STT3A alleles contains a large insertion (112 bp) that was intially detected by gel electrophoresis of the PCR amplification products (Fig. 2a). With one noteworthy exception, all of the identified genomic sequences contained indels that will create a frameshift mutation in the mRNA. The 6 bp in-frame deletion in the MagT1 gene will eliminate two residues from the N-terminal signal sequence. Immunoblot analysis and further functional analysis revealed that the MagT1(− /− ) cells do not express MagT1 despite having this small in-frame deletion. The most likely explanation for the MagT1 null phenotype is that the MagT1 gene is located on the X-chromosome 19,33 , hence the allele with the in frame deletion is likely silent due to X chromosome inactivation.
Genomic sites with fewer than 5 mismatches to the sgRNA that are adjacent to a PAM site (protospacer adjacent motif, NGG) can be recognized by the sgRNA and act as off-target sites 35 . The CRISPR DESIGN webserver (http://crispr.mit.edu) was used to search for potential off-target sites for each of the sgRNAs. The three most probable off-targets sites for the sgRNAs are listed in Supplementary Table S2. PCR fragments containing potential off-target sites were amplified from WT HEK293 cell DNA as well as from the STT3A(− /− ), STT3B(− /− ), MagT1(− /− ) and MagT1(− /− )TUSC3(− /− ) cell lines. T7E1 analysis did not disclose indel mutations in any of the potential off-target sites ( Supplementary Fig. S1).

Analysis of OST protein levels in knock-out cell lines.
Total cell extracts were prepared from wild type and mutant cells and resolved by SDS-PAGE for protein immunoblot analysis using antibodies raised against the OST subunits (Fig. 3a). Glyceraldehyde 3'phosphate dehydrogenase (GAPDH) was used as a protein loading control to normalize the calculated expression values for the OST subunits. Calculated expression values below 10% in the knockout cell lines represent non-specific background. Consistent with the DNA sequence analysis, the diffusely migrating STT3A and STT3B proteins are absent in the STT3A(− /− ) are STT3B(− /− ) cell lines, respectively. We observed a 30% increase in STT3A expression in the STT3B(− /− ) cell line and a 30% increase in expression of STT3B and MagT1 in the STT3A(− /− ) cells. These increases in expression level are consistent with our previous observations for siRNA treated HeLa cells 5,11 , and are likely explained by competition during OST assembly in wild type cells between STT3A and STT3B for a limited quantity of one or more of the shared OST subunits (Rb1, Rb2, OST48, DAD1 or OST4). The STT3A specific accessory subunit KCP2 was undetectable in extracts prepared from STT3A(− /− ) cells. Likewise, the STT3B complex specific accessory subunits MagT1 and TUSC3 were not detected in STT3B(− /− ) cells. We conclude that the accessory subunits (KCP2, MagT1 and TUSC3) are degraded when their cognate catalytic subunit is absent.
The complete absence of MagT1 immunoreactivity in the MagT1(− /− ) cells was accompanied by increased stable expression of the TUSC3 protein (Fig. 3a) as observed for MagT1 siRNA treated HEK293 cells (Fig. 1b). MagT1 and TUSC3 compete for a shared binding site in the STT3B complex 5 , hence when MagT1 is absent TUSC3 is able to occupy the vacant binding site. It is not clear why we detect elevated amounts of KCP2 in the MagT1(− /− ) TUSC3(− /− ) cell line, as we do not detect incorporation of KCP2 into the STT3B complex as assayed by blue native gel electrophoresis (data not shown).
The morphology and growth rate of the knockout cell lines was indistinguishable from the parental HEK293 cells during roughly 25 passages after the clones were isolated. Hence, loss of the STT3A or STT3B complex is tolerated in cultured cells. Importantly, fibroblasts obtained from the STT3A-CDG and STT3B-CDG patients showed reductions in STT3A and STT3B protein expression respectively, and displayed glycosylation defects that were similar to siRNA treated HeLa cells 12 .
Hypoglycosylation of proteins in the ER causes an induction of the unfolded protein response (UPR) that can be detected by increased expression of lumenal chaperones including BiP 36  Reverse transcription PCR was performed to evaluate mRNA levels for TUSC3 and STT3A in the knockout cell lines (Fig. 3b) (Fig. 3a) is not explained by enhanced transcription of the TUSC3 mRNA, but is instead explained by enhanced stability of the TUSC3 protein upon incorporation into the STT3B complex. Even though TUSC3 and MagT1 are homologues, incorporation of MagT1 into the STT3B complex is strongly favored relative to TUSC3.

Glycosylation of proteins in knockout cell lines.
The knockout cell lines were transiently transfected with expression vectors for a panel of glycoproteins that contain previously characterized STT3A and STT3B dependent glycosylation sites. Pulse-chase labeling of SHBG in HEK293 cells yields a mixture of monoglycosylated SHBG and diglycosylated SHBG (Fig. 4a) as observed previously for other cell lines and in human sera 24,37 . The absence of both oxidoreductase subunits (MagT1(− /− ) TUSC3(− /− )), or elimination of STT3B, causes a dramatic reduction in SHBG glycosylation. Single glycosylation sites mutants of SHBG were analyzed to determine which site is partially glycosylated in the STT3B(− /− ) cell line (Fig. 4b). Glycosylation of the N 380 RS site (N396Q mutant) was incomplete in wild type cells and strongly inhibited in STT3B(− /− ) cells. Glycosylation of the N 396 GT site (N380Qmutant) was efficient in wild type cells, but not completely blocked in STT3B(− /− ) cells. Taken together with previous results 24 , we can conclude that the frequency of acceptor site skipping of SHBG sequons by the STT3A complex is influenced by sequon type (NXS sites are skipped more frequently than NXT sites). The efficiency of posttranslational glycosylation by the STT3B complex is also higher for NXT sites than NXS sites, consistent with our bioinformatic analysis of extreme C-terminal sites in the murine glycoproteome 24 . OST assays using synthetic peptide acceptors indicate that the OST has a higher affinity for NXT sequons than NXS sequons 38 .
Depletion of STT3B or MagT1 reduces glycosylation of a single acceptor site in cathepsin C 5,11 . The STT3B-dependent N 29 CT site in cathepsin C (pCatC) is four residues from the signal sequence cleavage site and contains a cysteine that is disulfide-bonded in the mature protein. Previously, we showed that the signal sequence proximal location of the N 29 CT site and the internal cysteine residue are both required for STT3B and MagT1 dependence in HeLa cells 5 . Glycosylation of a pCatC derivative (pCatCΔ 234) that has a single acceptor site (N 29 CT) was strongly inhibited in cells lacking either STT3B or both oxidoreductase subunits (Fig. 4c).
Prosaposin and progranulin are cysteine-rich glycoproteins that are composed of four (pSAP) or seven (pGran) small repeat domains separated by spacer segments. Glycosylation of prosaposin and progranulin was greatly reduced in STT3A (− /− ) cells but not reduced when cells lack the STT3B complex (Fig. 4d), consistent with previous results obtained using siRNA treated HeLa cells 11 .
Glycosylation of certain acceptor sites was not reduced in HeLa cells that were treated with STT3A or STT3B siRNA. For example, glycosylation of haptoglobin (Hp) is insensitive to knockdown of STT3A 25 . A single acceptor site in haptoglobin (N 241 YS) was reduced by treatment with the STT3B siRNA 25 . The insensitivity of the haptoglobin sites to siRNA mediated depletion of STT3A or STT3B might indicate that the Hp sequons can be efficiently glycosylated by either STT3A or STT3B, or it might be explained by the incomplete depletion of STT3A and STT3B that is achieved by siRNA treatment. Pulse labeling of haptoglobin in the STT3B(− /− ) and MagT1(− /− ) TUSC3(− /− ) cell lines caused less than a 10% reduction in glycan occupancy (Fig. 4e).
Two of the five glycosylation sites in hemopexin (Hpx) show intermediate STT3B-dependence due to the presence of an internal cysteine residue (N 187 CS) or a C-terminal location (N 453 VT) 25 . Glycosylation of Hpx was not reduced in the STT3A(− /− ) cell line, but was reduced by 20% in the STT3B(− /− ) and MagT1(− /− ) TUSC3(− /− ) cell lines (Fig. 4f). The accumulation of glycoforms that lack one or two glycans is consistent with the presence of two acceptor sites that show partial STT3B dependence. The other three sites in hemopexin can be glycosylated by either STT3A or STT3B.
The results obtained by pulse-labeling of the knockout cell lines (Fig. 4) were compared to results obtained with siRNA treated HEK293 cells (Fig. 1) or siRNA treated HeLa cells 5,11,24,25 to evaluate the utility of the knockout cell lines (Table 1). Several conclusions can be drawn from the data in Table 1. Glycosylation of a protein with previously identified STT3A dependent sites (pSAP) was more severely reduced in the STT3A(− /− ) cell line than in siRNA treated HEK293 or HeLa cells. On average, the STT3B complex can modify two of the five sites in pSAP when the STT3A complex is absent. The heterogeneity of pSAP glycoforms in STT3A(− /− ) cells suggests that STT3B complex does not have a strong preference for a subset of the five acceptor sequons. Glycosylation of the previously identified STT3B dependent sites in SHBG and pCatCΔ 234 was more severely reduced in STT3B(− /− ) cells than in siRNA treated HeLa or HEK293 cells. Based upon the residual glycosylation of these three sites in STT3B(− /− ) cells, acceptor site skipping by STT3A can approach 100%. Analysis of haptoglobin and hemopexin glycosylation in STT3B(− /− ) or STT3A(− /− ) cells yielded results that were quite similar to what we obtained with siRNA treated HeLa cells. Analysis of Hp and Hpx glycosylation provides strong evidence for a class of glycosylation sites that can be glycosylated by either OST complex with high efficiency. A more global analysis of acceptor site occupancy in STT3A (− /− ) and STT3B(− /− ) cells will be required to identify additional features that mandate glycosylation by one of the two OST complexes. Consistent with our previous analysis 5 , STT3B complexes that lack an oxidoreductase subunit have little or no residual glycosylation activity, at least with the substrates we have analyzed to date. The oxidoreductase subunit in the STT3B complex is thought to form mixed disulfides with glycoprotein substrates explaining why MagT1 is necessary for glycosylation of sequons with an internal cysteine residue (e.g., N 29 CT sequon in pCatC). However, the oxidoreductase subunits are also required for glycosylation of the extreme C-terminal sequons in a SHBG derivative that lacks the C-terminal disulfide 5 , suggesting that MagT1 or TUSC3 are needed for substrate recognition. The MagT1(− /− ) cells had a less severe defect in glycosylation of STT3B dependent substrates than MagT1 siRNA treated HeLa cells due to enhanced stable expression of TUSC3 in the MagT1 deficient HEK293 cells. In contrast, the MagT1(− /− )TUSC3(− /− ) HEK293 cells had a more severe glycosylation defect that MagT1 siRNA treated HeLa or HEK293 cells. We have not identified a substrate that has a specific requirement for one of the two oxidoreductases, suggesting that these gene products are functionally redundant. An overlapping function of MagT1 and TUSC3 would help explain why TUSC3 [21][22][23] or MagT1 19 deficient human patients have less severe symptoms than the STT3B-CDG patient 12 . Restoration of glycosylation in knockout cell lines. The HEK293 derived knockout cell lines were cotransfected with a previously characterized expression plasmid for an OST subunit 5,27 and a glycoprotein reporter to determine whether normal glycosylation could be restored by transient expression of the missing OST subunit (Fig. 5). As expected, expression of MagT1-V5His had no impact on glycosylation of pCatCΔ 234 in the MagT1(− /− ) cells, since the enhanced stable expression of TUSC3 mitigates any glycosylation defect caused by loss of MagT1 (Fig. 5a). More importantly, coexpression of MagT1-V5His and TUSC3-DDKHis in the (MagT1(− /− )TUSC3(− /− ) mutant allows near-normal glycosylation of pCatCΔ 234 (Fig. 5a). Expression of STT3B-mycDDK in the STT3B(− /− ) cell line was somewhat less effective in restoring normal glycosylation of pCatCΔ 234 (Fig. 5a). Although we observed lower incorporation of radiolabel into both pCatCΔ 234 glycoforms in the cells expressing STT3B-mycDDK in this experiment, it is the ratio of the two glycoforms rather than total pCatCΔ 234 synthesis that is used to evaluate the efficiency of glycosylation. Expression of STT3A-DDKHis in the STT3A(− /− ) cell line improved, but did not fully restore glycosylation of prosaposin (Fig. 5b). Restoration of glycosylation by expression of STT3A or STT3B in the null cell lines is a more complicated process, because it entails de novo assembly of a seven to eight subunit integral membrane protein. Moreover, the exogenously expressed catalytic subunit (e.g. STT3B-mycDDK) will compete with newly synthesized endogenous STT3A for assembly with the shared subunits (Rb-1, Rb-2, OST48, DAD1 and OST4) of the STT3A and STT3B complexes.

Substrates (sequons)
The HEK293 derived null cell lines described here will be an important tool for further analysis of OST complex function in mammalian cells. While the results obtained with the knockout cell lines validates our previous conclusions about STT3A and STT3B dependent glycosylation sites, the interpretation of results was greatly simplified in the null cell lines relative to the siRNA treated cells. The null cell lines will be useful for experimental approaches where transient siRNA mediated knockdowns are not optimal including generation of cells with deficiencies in two or more proteins. These HEK293 derived null cell lines should also prove useful for mass spectrometric analysis of sequon occupancy of glycoproteins with multiple glycosylation sites like prosaposin or progranulin. The null cell lines are good models for complementation experiments where wild type and mutant alleles of an OST subunit will be tested for the ability to restore glycosylation to the mutant cells. The latter experiments are more challenging in siRNA treated cells due to the incomplete depletion of the genome-encoded wild type subunit and the need to generate an expression construct that is insensitive to the siRNA.

Materials and Methods
Cell culture, siRNA transfection and protein expression. HEK293 cells (ATCC CRL-1573) were cul-