Structural Basis for the Action Mechanism of Legionella Glycosyltransferase

Pathogenic bacterium Legionella pneumophila, the causative agent of Legionnaires’ disease, secretes hundreds of effectors into host cells that subvert cell pathways during pathogenesis. The Lgt family effectors, containing Lgt1, Lgt2, and Lgt3, from L. pneumophila are glycosyltransferase that shut down protein synthesis in human cells by specific glycosylation of a serine residue in the eukaryotic elongation factor 1 A (eEF1A), but the action mechanism remains poorly understood. Herein, the structural basis of the action mechanism is unveiled. Lgt family effectors catalyze the transfer of glucose moiety of UDP‐glucose in a conserved retaining mechanism, but exhibit different substrate recognition mechanisms. Lgt2 bears the positive‐charged catalytic cleft to interact with the negatively charged patches in helices α2 and α3 of eEF1A; instead, Lgt1 employs a negatively charged surface of E445 and E446, which are proposed to interact with residue Lys51 in eEF1A. And Lgt family effectors inhibit host unfolded protein response by shutting down host protein synthesis. These results provide a structural basis of action mechanism of glycosyltransferase and highlight the role of glycosyltransferase in Legionella's lifecycle.

family effectors suppress host unfolded protein response. Our data provide insight into the substrate recognition mechanism by bacterial glycosyltransferase and highlight the important role of bacterial effectors against host immune response.

Structure of Legionella Glycosyltransferase Lgt2
Glycosyltransferases from L. pneumophila, namely Lgt family effectors, contain three members Lgt1, Lgt2, and Lgt3. [18] Although Lgt1 has been characterized structurally, its action mechanism remains poorly understood. To reveal their action mechanism, we first resolved the Lgt2 structure at a resolution of 2.27 Å. Similar to Lgt1, the Apo-form structure reveals four distinct domains N-terminal domain (NTD), glycosyltransferase domain (GT), and protrusion domain (PD), and C-terminal domain (CTD), encompassing 29 of α-helices and 9 of β-sheets. The NTD divides into two subdomains NTD-I (α1-α5) and NTD-II (α6-α10) (Figure 1a). A linker region connects NTD-II to the central part of the GT domain. The GT possesses a double Rossmann fold-like signature typical of the GT-A fold and a conserved DxD motif (D338-X-D340). And the PD consisting of α-helices (α16-α25), forms an extended α-helical protrusion from the GT domain. The CTD, encompassing four α-helices (α26-α29), returns to GT from the PD domain. The NTD-II and PD form the protrusion from the GT domain as two pliers, holding NTD-I between them, resulting in a compactable overall conformation. Helices α1-α3 from NTD-I principally participate in contact with other domains by extensive polar and hydrophobic interactions (Figure 1b-d).

UDP-Glucose Binds to Glycosyltransferase Domain
As is well-known, Lgt family effectors are the UDP-glucosedependent glycosyltransferases that employ UDP-glucose as glucose moiety donor. To wonder about UDP-glucose recognition by Lgt2, we subsequently resolved the Lgt2-UDP-glucose binary complex structure. In binary complex, UDP-glucose (UPG) binds to a deep binding cleft in glycosyltransferase domain, where is covered by CTD (α28-α29 region), thus locking UPG in place ( Figure 2a). Structural alignment shows that UPG-bound Lgt2  . UDP-glucose binds to GT domain of Lgt2. a) Crystal structure of Lgt2-UDP-glucose binary complex shown in cartoon representation. UDP-Glucose (UPG) is buried into a groove in GT domain, which is covered by the very C-terminal region of CTD domain. UPG is depicted in orange stick representation. The close-up view shows the omitted map of UPG (grey mesh) contoured at 2.0σ. UPG is depicted as a stick. b) Structural alignment of Lgt2 in the presence or absence of UPG by α C atoms. The structure of Lgt2 in UPG-free state (Apo form) is colored in the cyan cartoon, and that of the Lgt2-UPG complex is shown in the green cartoon. UPG is indicated in an orange stick. c,d) Interactions of Lgt2 with UPG. Residues are implicated in contact with glucose and uridine moieties of UPG (c) and interact with a pyrophosphoric moiety of UPG (d). All contacting residues are shown in stick representation, which is from the GT domain (in green) and CTD (in light blue), respectively. UPG is depicted in an orange stick. Hydrogen bonds are indicated by blue dashed lines, and the water molecules as gray spheres. The distance is shown in black dashed lines. e) The binding abilities of Lgt2 and its mutants to UPG. The binding of Lgt2 to UPG was determined by ITC to obtain their dissociation constant (Kd). Kd values of wild-type Lgt2 and its indicated mutants are listed, calculated from Figure S1, Supporting Information. DD is the Lgt2 mutant D338A/D340A; /, no expression. f ) Structurebased sequence alignment of Lgt2 with other members of glycosyltransferase GT-A family (See also Figure S3, Supporting Information). This alignment is performed by PROMALS3D [35] and ESPript server. [36] PDB ID of glycosyltransferase structures used for alignment as follow: Lgt1:2WZF; TscL:2BV1; TcdB:2VKH. Structures of Lgt3 and Lpg1961 were generated by the AlphaFold server. Residues marked with a triangle are implicated in UPG-binding (black), catalysis (red), and substrate-binding (green). g) Glycosylating activities of Lgt2 and its mutants. The glycosylating activity was measured by incubating Lgt2 or its mutants with UPG and released UDP was measured by UDP-Glo Glycosyltransferase assay kit. Results shown were from three independent experiments. Data are presented as means AE SD. *** P < 0.001. h) Lgt2 requires divalent metal ions. The glycosylating activity was performed with the addition of the indicated various ions and the released UDP was measured by UDP-Glo Glycosyltransferase Assay Kit. Results shown were from three independent experiments. Data are presented as means AE SD. ***P < 0.001. www.advancedsciencenews.com www.small-structures.com was highly similar to the UPG-free form with a root-mean-square deviation (RMSD) of 0.53 Å for the C α atoms (Figure 2b). Moreover, UPG contacts the glycosyltransferase domain extensively via polar and hydrophobic interactions (Figure 2c, d). In detail, the nucleotide base of UPG stacks against the indole moiety of W224, and the protein assures base specificity through three hydrogen bonds from uridine-N3 to the backbone carbonyl of F225 and from the backbone amide nitrogen of the same residue to uridine-O2, and from the side chain of residue Q631 to uridine-O1. The ribose moiety forms two hydrogen bonds from its 2 0 -OH group and the side chain of S319 and from its 3 0 -OH group and the backbone amide nitrogen of A339, while the 3 0 -OH group also contacts one water molecule coordinating with the backbone of residues I223 and A339. The glucopyranose moiety of UPG is firmly locked in the binding pocket, with the 2 00 ÀOH, 3 00 ÀOH, 4 00 ÀOH, and 6 00 ÀOH groups of the sugar in direct hydrogen-bonding contact with the side chains of residues D320 and R323. The pyrophosphate moiety of UPG forms hydrogen bonds with S625 and the indole nitrogen of W626 from CTD.
To verify the structural observations, we constructed the Alasubstituted mutants and detected their UPG-binding ability by Isothermal Titration Calorimetry (ITC) method. As expected, mutants I223A, S319A, D320A, S625A, W626A, and Q631A greatly impaired the UPG-binding ability, especially D320A causing a nearly 20-fold loss of affinity compared with the wild-type (Figure 2e and S1, Supporting Information). And these residues implicated in UPG-binding are highly conserved in glycosyltransferase GT-A family (Figure 2f and S2, Supporting Information). Consistently, those mutants impaired or obliterated glycosyltransferase activity (Figure 2g).
Moreover, like most of the GT-A family members, Lgt2 also shares the conserved DxD motif (D338-X-D340) (Figure 2d,f ), enabling the coordinate of the divalent cations (usually Mn 2þ or Mg 2þ ) to facilitate the departure of the nucleoside diphosphate leaving group by electrostatically stabilizing the developing negative charge. [22] Additionally, the double mutant D338A/D340A (DD) had no detectable glycosyltransferase activity but a negligible effect on the UPG-binding ability (Figure 2e,g). Interestingly, Lgt2 showed a highly selectivity to Mg 2þ but not Mn 2þ (Figure 2h), which differs from other known GT-A enzymes.

Lgt2 Shows a Low Catalytic Activity Toward the eEF1A-Derived Decapeptide
The structural alignment reveals the overall shape of Lgt2 differs drastically from Lgt1 and TcdB (Figure 3a and S3a, Supporting Information). Lgt1 NTD folding into α-helical bundle laterally packs together with the glycosyltransferase domain, yielding an elongated overall shape. However, Lgt2 NTD divides into two subdomains Lgt2 NTDÀI and Lgt2 NTDÀII . Lgt2 NTDÀII forms an extended α-helical protrusion and vertically hangs on the glycosyltransferase domain, resulting in the formation of a huge groove between NTD-II and PD domain to position and stabilize NTD-I. Furthermore, the very C-terminal α-helix α29 from CTD is locked in the glycosyltransferase domain by several hydrogen bonds from residue S635 to N228, from residues L634 and Q631 to N227, and from residue W626 to N314 ( Figure S3b, Supporting Information). Whereas the equivalent region of Lgt1 undergoes a disordered loop, endowing it with more flexibility that is necessary for the correct binding of UPG. [20,23] Otherwise, we also have tried to obtain the full-length Lgt3 structure. Unexpectedly, we only obtained the part structure of Lgt3 lacking NTD-I and CTD, and this portion is highly similar to the Lgt2 structure (Figure 3b), suggesting that the overall structure of Lgt3, particularly its NTD and CTD, is more flexible than Lgt2.
To our knowledge, Lgts suppress host unfolded protein response (UPR) by blocking host protein synthesis. To wonder whether the aforementioned differences affect Lgt family members' functions in cells, we then tested the effect of Lgts on host UPR. However, ectopic expression of Lgts in HEK293T cells treated with thapsigargin showed no difference in the inhibition of the UPRassociated factors CHOP and ATF4 expressions (Figure 3c). Given the intracellular complexity, we subsequently detected their glycosyltransferase activities in vitro. Consistent with the structural observations, Lgt2 indeed showed a much lower activity towards a minimal decapeptide derived from eEF1A than Lgt1 and Lgt3 ( Figure 3d). Consistently, the catalytic efficiency of Lgt2 showed an approximately 10-fold lower than Lgt1 and Lgt3 ( Figure 3e).

Lgt2 Bears a Positive-Charged Substrate-Entrance Surface
We next analyzed the structural conservatism of the Lgt family and found that their UDP-glucose-binding pocket is highly conserved, but the putative substrate-entrance pocket appears to be variable ( Figure S4aÀc, Supporting Information). The substrateentrance sites E445 and E446 in Lgt1, located at the funnel-like entrance to the active site, which is proposed to play a role in the recognition and binding of two lysine residues K51 and K55 in eEF1A. [20] Interestingly, the residue in Lgt2 equivalent to residue E445 in Lgt1 is replaced by a methionine (Figure 2f ). However, mutant Lgt1 E445M and Lgt2 M547E did not show a significant effect on glycosyltransferase activities of Lgt1 and Lgt2 ( Figure S4d, Supporting Information). We further examined the surface potential distribution of Lgts' structures and found that the putative substrate-entrance of Lgt1 is around together with negative charges, instead, Lgt2 and Lgt3 with extensive positive charges ( Figure 4aÀc). Hence, the positively charged residues K51 and K55 make the minimal decapeptide ( 50 GK 51 GS 53 FK 55 YAWV 59 ) positively charged, leading to a charge repulsion effect. It appears to be one of the reasons that result in a lower activity of Lgt2.
Sequence alignment shows that the N-terminal region (34-76 amino acids) of eEF1A is highly conserved in different species ( Figure S5a, Supporting Information). This portion containing the decapeptide sequence and two helices α2 and α3, forms the helix-loop-helix structure, here termed HLH ( Figure S5b, Supporting Information). HLH harbors two negatively charged patches Glu40-Glu43 and Glu45-Glu48 (Figure 4d and S5b,c, Supporting Information). We thus speculated the negatively charged HLH is much easier to access the positive-charged substrate-entrance of the catalytic cleft in Lgt2 and Lgt3. Consistent with the structural observations, Lgt2 and Lgt3 displayed a stronger activity than Lgt1 when HLH was a substrate (Figure 4e). Meanwhile, Lgt2 also exhibited a higher glycosylating activity when HLH was a substrate (Figure 4f ). On the contrary, Lgt1 has been reported to show much lower activity towards HLH than the minimal decapeptide. [19]

NTD-I Domain is Required for Intracellular Localization
To uncover the role of Lgt2 NTDÀI , we constructed a truncated mutant Lgt2 ΔN lacking an NTD-I domain and first detected its activity. We found that deletion of NTD-I brought a negligible effect on the glycosyltransferase activity towards eEF1A-derived decapeptide ( Figure 5a). Subsequently, we detected the effect of NTD-I on the regulation of UPR. After treatment with thapsigargin, the cells bearing empty vector or expressing Lgt2 DD showed the prototypical robust up-regulation of the UPR response regulators BiP, ATF4, and CHOP translationally. Instead, we did not observe the appreciable up-regulation of these regulators in cells expressing wild-type Lgt2 or Lgt2 ΔN (Figure 5b). Consistently, the mRNA levels of BiP, CHOP, and ATF4 in cells expressing Lgt2 ΔN also had no difference from that of cells expressing the wild-type (Figure 5c-e). Thus, these data indicated that Lgt2 NTDÀI does not play an essential role in the glycosyltransferase activity and inhibition of the host UPR.
Given the inhibitory effect of Lgt2 on the host unfolded protein response, we suspected that Lgt2 might be relevant to the host Lgt2 is colored in a cyan cartoon and Lgt1 is shown in an orange cartoon. b) Structural alignment of Lgt2 and Lgt3 by α C atoms of GT-I domain. Lgt2 is colored in a cyan cartoon and Lgt3 is shown in an orange cartoon. c) The components of the UPR, ATF4, and CHOP are blocked translationally in the presence of Lgts. After 24 h transfection with Lgts, HEK293T cells were untreated or treated with thapsigargin (Tg; 1 μg•μL À1 ) for 6 h. Protein expressions of ATF4 and CHOP were monitored by their specific antibodies via immunoblotting. GAPDH was used as the loading control. d) Glycosylating activity of Lgt family effectors. The glycosylating activity was measured by incubating 0.1 μM glycosylating enzymes with UPG, and the released UDP was measured with a UDP-Glo Glycosyltransferase assay kit. The eEF1A-derived decapeptide was used as substrate. Results shown were from three independent experiments. Data are presented as means AE SD. **** P < 0.0001; ns, not significant. e) Enzymatic kinetics of Lgt family members. A concentration gradient of UPG was incubated with 1 μM Lgts on a 96-well plate. The released UDP measured by UDP-Glo Glycosyltransferase assay kit was read every 30 s for 30 min at 37°C. The steady-state kinetic parameters were derived by fitting to the Michalis-Menten equation. endoplasmic reticulum and further tested whether Lgt2 locates at the endoplasmic reticulum. However, we did not find co-localization of Lgt2 with the endoplasmic reticulum, instead, located at both cytosol and nucleus. Interestingly, we observed a significant increase in the proportion of cytosolic Lgt2 mutant lacking NTD-I (Lgt2 ΔN ) (Figure 5f,g). Instead, neither mutant Lgt1 ΔN not Lgt3 ΔN appeared to affect their intracellular distribution ( Figure S6, Supporting Information). Taken together, these data suggested that NTD-I is responsible for the intracellular localization of Lgt2.

Lgt2 Employs a Retaining Catalytic Mechanism
Our results show that Lgt2 is a metal-dependent glycosyltransferase (Figure 2h). Like other GT-A enzymes, Lgt2 structure demonstrates a conserved DxD motif, nucleotide-sugar binding sites, and a key residue Asn384, positioned %5 Å away from the anomeric carbon, in a similar manner to Asn293 in Lgt1 (Figure 3c) [20,21] and Asn384 in TcdB, [24] which are proposed to be involved in a back-side nucleophilic push. TcdB glucosyltransferase domain and Lgt1 belong to the GT-A fold family of glycosyltransferases with a retaining catalytic character. [22] For the retaining enzymes, a glutamine or glutamic acid residue is proposed as the catalytic nucleophile involved in a D N Â A Nss ion-pair mechanism. [22] Analogously, Ala substitution in residues close to the anomeric carbon, e.g., Asn384, Thr385, and Asn602 in Lgt2, indeed resulted in severely reduced enzyme activity but not affected the UDP-glucose-binding (Figure 2e,g). Consistently, Lgt3 also contains two equivalent residues Asn392 and Asn603 in its glycosyltransferase domain ( Figure S3d, Supporting Information). Taken together, our structural and  ) shows the substrate entrance sites. In Lgt1, two negatively charged residues, E445 and E446 (as sticks), line the entrance to the binding site for UDP-Glc (UPG). In Lgt2, two residues M547 and E548 (as sticks) position at a place equivalent to that of residues E445 and E446 from Lgt1. In Lgt3, two residues M548 and E549 (as sticks) at a place equivalent to that of residues E547 and E548 from Lgt2. d) Surface charge distribution of eEF1A-derived HTH contoured from À7kBT (red) to þ7kBT (blue). Residues E43/E45 and E45/E48 in sticks are composed of two negatively charged patches. K51 and S53 are the residues that are recognized and glycosylated by the Lgts family, respectively. e) Glycosylating activity of Lgt family effectors. The GST-fused eEF1A-derived HLH (GST-HLH) was used as a substrate and GST protein as a blank control. Results shown were from three independent experiments. Data are presented as means AE SD. ****P < 0.0001; ns, not significant. f ) Glycosylating activity of Lgt2. The decapeptide and GST-fused eEF1A-derived HLH were used as a substrate, respectively. WT: wild-type protein Lgt2; DD: Lgt2 mutant D338A/D340A. Results shown were from three independent experiments. Data are presented as means AE SD. ****P < 0.0001; ns, not significant. biochemical studies suggest that Lgt2 employs a retaining mechanism following an S N i-like mechanism, in which Asn384 likely acts as the nucleophile involved in pushing the anomeric carbon during the Sn i (substitution nucleophilic internal)-like mechanism through a D N Â A Nss ion-pair and Asn602 may play an important role in stabilizing the transition state and the leaving group.

Lgt Family Effectors Recognize the Substrate in Different Manners
Our structures have shown that Lgt family effectors undergo different conformations, except from GT and PD domains (Figure 3a,b). Meanwhile, the biochemical results also showed c-e) Components of the UPR, BIP, ATF4, and CHOP were blocked transcriptionally. After 24 h transfection with Lgt2, HEK293T Cells were then either untreated or treated with thapsigargin (Tg; 1 μg•αL À1 ) for 6 h, and mRNA was harvested and qRT-PCR was performed to assay the abundance of components of the UPR mRNA transcripts. GAPDH was used for the endogenous normalization of gene expression. In both cases, two biological replicates were used each containing three technical replicates. Data are presented as means AE SD. ***P < 0.001, ns, not significant. f ) NTD-I domain is relevant to Lgt2 intracellular localization. HKE293T cells transfected to express Lgt2-mCherry (WT) or Lgt2-mCherry lacking NTD-I domain (ΔN) were fixed and stained with ER-specific protein Calreticulin antibodies and nuclei were labeled by DAPI. Images acquired by a confocal microscope were pseudocolored with a Leica software package. Scale bar: 20 μm. g) The quantification of intracellular localization (f ) was calculated for three times using more than 100 randomly selected cells as one group. *** P < 0.001.
www.advancedsciencenews.com www.small-structures.com that Lgt family effectors display different affinity to ligand UDPglucose and different activity toward eEF1A-derived fragments (Figure 3d,e). Lgt1 has been reported that two negatively charged residues E445 and E446 [20] close to the anomeric carbon form a local negatively charged patch (Figure 4a), which are proposed as a prominent position to interact with the positively charged lysine K51 in the recognition peptide of eEF1A. [25] Instead, the replacement of residue Glu with Met in Lgt2 and Lgt3 results in a neutrally charged surface (Figure 4b,c). Meanwhile, the replacement of residue Glu with Met in Lgts showed a negligible effect on the glycosyltransferase activity, consistent with their substrate substrate-entrance pockets being variable ( Figure S4d, Supporting Information). Moreover, Lgt2 and Lgt3 bear a positive-charged substrate entrance, instead, a negatively-charged entrance in Lgt1 (Figure 4a-c). We deduce that these features impair the recognition ability of Lgt2 to the decapeptide with positive charges. Consistently, Lgt2 displays a lower activity toward Lgt1. On the contrary, Lgt2 showed a higher activity toward eEF1A-derived HLH with the negative-charged patches, but Lgt1 did not (Figure 4e). These aforementioned indicated that Lgt2 undergoes a different substrate recognition mechanism to Lgt1. In addition, although Lgt3, similar to Lgt2, also harbors the positively charged entrance, it showed an equivalent activity towards eEF1A-derived HLH and decapeptide, which might result from more flexibility of its NTD and CTD, suggesting a different recognition mechanism against the substrate. Taken together, our structural studies show that Lgt family effectors recognize their substrate eEF1A in a different manner.

Inhibition of Unfolded Protein Response is Conducive for Bacterial Survival in Host
There are growing evidences that Legionella has evolved to regulate host innate immune response such as interferon production [26] and activation of NF-κB pathway [27] by the secreted effectors, aiding survival in host cells. In this study, our data suggest that Lgt family effectors have the remarkable ability to inhibit the UPR pathway by blocking the production of UPR-response factors ATF4 and CHOP (Figure 3c and 5b-e). And the PERK/ATF4/CHOP signaling pathway functions in ER stress-induced cell apoptosis to defend against microbial infection. [28] This observation reflects Legionella's strategy to utilize the ER without inducing UPR-driven apoptosis. The cGAS-STING pathway is well known to promote the eventual production of interferon against exogenous pathogens, [29] which selectively control the translation program preferential to inflammatory and survival signaling via direct activation of PERK-eIF2α pathway recently. [30] Given the close interplays among UPR, the inflammatory response pathway, and innate immune signaling, [11,31] it would be exciting to further investigate how L. pneumophila uses the UPR to modulate the host innate immune system. Taken together, we favor the hypothesis that protein synthesis inhibition mediated by Lgt family effectors plays a role in the modulation of host innate immunity, possibly providing a cellular environment more conducive to bacterial replication.

Conclusions
Lgt family effectors Lgt1, Lgt2, and Lgt3, a redundant set of virulence factors, glycosylate host eEF1A, thus blocking host protein synthesis. Here, we reveal that Lgt2 which is localized by its N-terminal domain harbors a conserved retaining action mechanism but exhibits a different substrate recognition mechanism derived from the positive-charged substrate-entrance surface.

Experimental Section
Bacterial and Cell Culture: Escherichia coli strains were grown using LB broth or agar, supplemented with 100 μg mL À1 ampicillin, or 50 μg mL À1 kanamycin if the selection of plasmids was required.
All cell lines were authenticated by microscopic morphology evaluation. Human embryonic kidney 293T cell (HEK293T) was grown in DMEM (4.5 g L À1 glucose and sodium pyruvate) supplemented with 10% v/v heat-inactivated fetal bovine serum (FBS, Gibco) and 4 mM L-glutamine at 37°C and 10% CO 2 .
Plasmid Construction: All the primers used for this study were listed in Table S1, Supporting Information. All enzymes for DNA manipulation were obtained from New England Biolabs. Chromosomal DNA of L. pneumophila (strain Philadelphia 1/ATCC 33 152), was extracted using the Qiagen Kit and used to amplify lgt2 gene (lpg2862, protein accession number WP_010948548.1) and lgt3 gene (lpg1488, protein accession number WP_010947217.1) by PCR. To obtain the recombinant Lgts proteins, the open reading frame of gene lgt2 and lgt3 were cloned directly into the bacterial expression vector pET-His expressing an N-terminal hexahistidine tag by BamHI and NheI restriction sites. To generate Lgts in mammalian cells, the open reading frame of genes was cloned into the vector pCMV expressing an N-terminal 3x Flag by EcoRI and XhoI restriction sites. All mutations were generated using an improved QuickChange method with a KOD high-fidelity DNA polymerase (Takara). All plasmids were verified by sequencing.
Protein Expression and Purification: E. coli strain BL21 (DE3) transformed with an expression construct was cultivated in LB medium at 37°C till its optical density OD 600 of 0.8, and induced by 400 μM isopropyl β-D-1thiogalactopyranoside (IPTG) at 18°C overnight. The cell pellets were harvested by centrifugation at 4,000 rpm for 15 min and stored at À20°C.
His 6 -tagged proteins were purified using the following protocol. The cell pellets were suspended in a lysis buffer containing 20 mM Tris-HCl pH 8, 500 mM NaCl, 10% v/v glycerol, 5 mM β-mercaptoethanol, 20 mM imidazole, and protease inhibitors, and lysed using sonication on ice. The cell lysate was centrifuged for 30 min at 17 000 rpm and its supernatant was collected and incubated with nickel sepharose beads for 1-2 h at 4°C. The beads were washed with the lysis buffer and eluted by the lysis buffer added with a gradient of imidazole: 50, 100, 200, and 500 mM. The best fractions that contained our protein of interest were pooled, concentrated, and ran through a Superdex 200 size-exclusion column (GE Healthcare) in 20 mM Tris-HCl pH 8.0, 150 mM NaCl, and 2 mM DTT. The best fractions of protein peak were pooled, concentrated to 10 mg mL À1 with an Amicon Centrifugal filter (Millipore), flash-frozen in liquid nitrogen, and stored at À80°C for crystallization and activity assay. Protein concentration was measured at A 280 and calculated using their theoretical extinction coefficients.
Glycosyltransferase Activity Assay: The decapeptide ( 50 GKGSFKYAWV 59 ) of human eEF1A synthesized by a local company (SciLight Biotechnology Ltd., Beijing) was used as a substrate for glycosylation assay. Glycosylation reactions were performed in reaction buffer 150 mM NaCl, 20 mM Tris-HCl pH 8.0, 2 mM MgCl 2 . 100 μL of reaction mixture including 0.1 μM Lgt2 or its mutants, 15 μM decapeptide or GST-HLH, and 50 μM UDP-Glucose, was incubated at 37°C for 30 min. The released UDP produced by the reaction was determined using UDP-Glo Glycosyltransferase Assay Kit according to its manual and the results are indicated as luminescence.
To obtain the enzymatic kinetic parameters K m and V max , the reactions were set with 1 μM Lgts purified above, 15 μM decapeptide, and a UDP-Glucose gradient in a buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM MgCl 2 ) at 37°C. The released UDP produced by the reaction was monitored above continuously for 30 min. Initial velocity (V 0 ) was calculated using only the initial linear portion of the progress curve for each www.advancedsciencenews.com www.small-structures.com concentration of UDP-Glucose. And the enzymatic parameters such as K m and V max were derived by fitting the measurements to the Michaels-Menten equation.
To evaluate the impact of divalent metal ions on the glycosylating activity of Lgt2, the glycosylation reaction was conducted with various divalent metal ions at the same condition, respectively. EDTA was used as a negative control.
Protein Crystallization and Structure Determination: Crystallization of 6 Â His-tagged Lgt2 or Lgt3 was performed at 16°C using the sitting-drop diffusion method by mixing 0.4 μL of protein (10 mg mL À1 ) with an equal volume of reservoir solution. Crystals of Lgt2 were observed in 0.2 M Magnesium acetate, 20% w/v polyethylene glycol (PEG) 3,350 after one week. After optimization, the best crystals were obtained in 0.2 M Magnesium acetate hydrate, 20% w/v PEG3,350. The crystals of Lgt3 were observed and optimized from buffer 0.2 M Potassium thiocyanate, 20% PEG w/v 3,350. The crystals were cryoprotected by brief soaking in the crystallization buffer with an additional 25% v/v glycerol. To obtain the structure of the Lgt2-UDP-glucose binary complex, the optimized crystals of wildtype Lgt2 were soaked for 5 min in the same cryoprotectant containing an additional 1 mM UDP-glucose. All crystals were flash-frozen in liquid nitrogen. Diffractions of crystals were collected at beamline BL02U1 of Shanghai Synchrotron Radiation Facility (SSRF) (Shanghai, China). Data sets were indexed, integrated, and scaled with xia2 software. [32] Structures of Lgt2 were determined by molecular replacement method with the program Phaser in PHENIX, [33] and the structure of Lgt1 (PDB ID: 3JSZ) was employed as a template. The structures of the Lgt2-UPG binary complex and Lgt3 were determined by the molecular replacement method using the Lgt2 structure as a template. The subsequent model was manually built using COOT [34] and refined with the program Refine in PHENIX. [33] The data collection and structure refinement statistics are summarized in Table 1. Structural figures were generated using PyMol (http://pymol.org/2/). Isothermal Titration Calorimetry: His 6 -Lgt2 was expressed and purified as described before and finally stored in 50 mM Tris-HCl pH 8.0 and 150 mM NaCl. UDP-glucose was dissolved in the same buffer. Concentrations of Lgt2 or its mutants and UDP-glucose were 0.1 and 1 mM, respectively. Isothermal titration calorimetry (ITC) experiments were performed using a microcalorimeter Affinity ITC (Waters, USA) at 25°C. Titrations were set for 20 injections and each of 2 μL with 200 s intervals. Data baseline subtraction and analysis were performed using NanoAnalyze. Peaks were integrated and fit with a 1:1 binding model. The first injection of each experiment was excluded from the analysis.
Confocal Imaging: For intracellular Lgt2 localization, HEK293T cells seeded on coverslips in a 12-well plate at a density of 1.0 Â 10 4 cells per well were transfected for 24 h with plasmids pCMV-mCherry harboring gene lgt1, lgt2, and lgt3, or their mutants, respectively. pCMV-mCherry was used as a control. Cells were then fixed with 4% w/v paraformaldehyde and permeabilized with 0.1% Triton X-100 for 10 min. Organelles were marked by their corresponding specific protein antibody (1:200). Endoplasmic reticulum was indicated separately by anti-Calreticulin antibodies, and nuclei of cells were stained by dye DAPI. More than 100 cells were collected and examined for nuclear localization analysis for each transfection and counted in ImageJ.
Quantitative Real-Time PCR Analysis: Total RNA was isolated from HEK293T cells using RNAiso Plus (Takara, 9109) as per the manufacturer's instructions. Then reverse transcribed into cDNA by PrimeScript RT reagent Kit with gDNA Eraser (Takara, RR047A). Real-time PCR was performed with TB Green Premix Ex Taq (Tli RNaseH Plus) (Takara, RR420A) in QuantStudio 6 Flex Real-Time PCR System (ABI) according to the standard procedure. The fold change expression (ÀΔΔCt) was calculated after normalization with GAPDH expression.
Data Analysis: Statistical analysis was calculated by one-way ANOVA by GraphPad Prism 8 software (SanDiego, CA, USA), and p < 0.05 was considered a significant difference.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.