Structural basis for the recognition of complex-type N-glycans by Endoglycosidase S

Endoglycosidase S (EndoS) is a bacterial endo-β-N-acetylglucosaminidase that specifically catalyzes the hydrolysis of the β-1,4 linkage between the first two N-acetylglucosamine residues of the biantennary complex-type N-linked glycans of IgG Fc regions. It is used for the chemoenzymatic synthesis of homogeneously glycosylated antibodies with improved therapeutic properties, but the molecular basis for its substrate specificity is unknown. Here, we report the crystal structure of the full-length EndoS in complex with its oligosaccharide G2 product. The glycoside hydrolase domain contains two well-defined asymmetric grooves that accommodate the complex-type N-linked glycan antennae near the active site. Several loops shape the glycan binding site, thereby governing the strict substrate specificity of EndoS. Comparing the arrangement of these loops within EndoS and related endoglycosidases, reveals distinct-binding site architectures that correlate with the respective glycan specificities, providing a basis for the bioengineering of endoglycosidases to tailor the chemoenzymatic synthesis of monoclonal antibodies.

The manuscript by Trastoy et al. describes primarily the crystal structure of a complex of the IgGspecific endoglycosylase S (EndoS) with the G2 product of its reaction with the glycosylation on Asn127 of the Fc domain of IgG. The interactions between enzyme and product are described in detail. SAXS studies are used to show that there are no large-scale conformational changes in solution upon product binding. Mutations in loops shown from the crystal structure to be important for enzyme-product interactions are made and their effect on enzymatic activity against the model IgG Rituximab are analysed, highlighting the importance of loops 1, 6 and 7.
In summary, while the results are undeniably interesting, it is uncertain to me to what extent they would influence thinking in the field. The results are similar to those obtained for EndoF3 more than 15 years ago. I feel that detailed comparison with EndoF3, which has very much overlapping substrate specificity, is lacking. How do the differences between EndoS and EndoF3 explain how EndoF3 accommodates the same product as EndoS while also accepting triantennary glycosylations? Finally, the manuscript is in places poorly written, with poor figure legends and inconsistent reference to figures. I would recommend either that the work is sent to a more specialised journal or resubmitted here with major revision. Some comments to help improve the manuscript: I think that the picture of the glycosylation in Fig. S2 should be in the main manuscript.
line 49: "between the first two N-acetylglucosamine residues" line 53: "glycoforms and" line 74: how was the volume of the G2 glycan binding site estimated? lines 84 and onwards: the numbering of the saccharide rings used in this description is not shown in any figure. line 96: "protrude away" line 99: " Figure S5" line 101: what is the S2G2 substrate? line 108: "comprised" is unnecessary lines 108-110: It's unnecessary to write "Ala" at the end of every mutation, as it was specified in the previous sentence that they are all Ala mutations. line 121: I would include references to Figs. 2d and 3a here. Why do references to the original unbound crystal structure (ref. 18) appear for the first time here? line 121: Supplemental Figure S6 is used to support a statement about the position of Trp153 at the bisection point of the two binding grooves. However the main aim of this figure seems to be to describe differences between the bound and unbound structures. No mention of this is made in the main text. Furthermore, Figure S6 is in contradiction to Figure S4, where the crystal structure of the complex is superposed with the envelope of both the bound and unbound forms to assert that there is no difference in conformation. In Fig. S6b one can clearly see a change in the conformation of two domains. The fitting of both forms to the SAXS data of both forms (i.e. 4 comparisons) should be investigated. What is the chi-squared value for the crystal structures(s) using CRYSOL? line 124: what program produced the Z-score? Was it DALI? If so, it should be mentioned and cited. Online methods: line 262: the reference 21 should be moved to just after "FastCloning method" on line 261 and an address should be given for GeneWiz. line 265: I guess 50 micrograms/ml is meant line 275: Coomassie Blue line 279: What happened after the protein was loaded onto the Superdex column? I guess it was eluted in some way? How was it concentrated and stored? line 283: sodium line 284: amino acid mixuture (L-Na-glutamate,... lysine HCl) line 290: How were the data scaled, with XSCALE or AIMLESS in CCP4? line 294: Was 4NUZ used unmodified as a search model, or was it split into domains? line 301: SAXS is not diffraction. line 303: batch mode line 305: The description of SAXS data collection is unsatisfactory. What was the sample size? Capillary volume? Temperature? Was the sample flowed or oscillated to avoid radiation damage? Was the I(0) normalised? line 311: Were any attempts made to do ab initio reconstruction using DAMMIN/DAMMIF? line 319: The lack of conformational change is a result and should not be mentioned only in the Online Methods section, but in the main text.
The supplementary/supplemental/supporting material (the authors use all three terms) is a bit sloppily prepared, with above all inadequate figure legends.
Table S1: The table seems to have been copied from phenix.table_one without too much afterthought. The space group is more simply written P2(1), with the 1 as a subscript. The use of two decimal places for unit cell dimensions and for the B-factors is not necessary. The difference between Rmodel and Rfree seems rather small for a 2.9 Å structure: were the Rfree flags used to crossvalidate the unbound structure transferred to the dataset for the bound structure presented here? It isn't stated anywhere now many reflections were used for the Rfree calculation. Table S2: The symbol ≤ should presumably be ±. There seems to be an extra digit, "1" after the end of the second Porod volume estimate. The units of I(0) cannot be reciprocal Ångström, as this is the unit of q, on the x-axis of the scatterig curve. I(0) is on the y axis and has arbitrary units unless normalised using a known standard (was this done?) Likewise, why is the unit reciprocal cm used for the I(0) estimate from Guinier? "Comparation" should be "comparison". Why is the CRYSOL chisquared value so high? Did the authors consider EOM? The program "SCATER" should be spelled "SCÅTTER" In Fig. S1, the figure gives the misleading impression that the glycosylations are highly exposed on the exterior of the Fc domain, when they are in fact on the inside. While I appreciate that this can be hard to draw, it should at least be indicated in the figure legend. It should be explained what the right-hand reactive group of the molecule used to attach uniform glycosylations to IgG is. The cleavage point leading to the G2 product should be indicated. The symbols for the different types of saccharide unit should be defined. There should be a legend for this figure! Fig. S2 legend: "996-1191 residues" should read "residues 996-1191". The sequence should be numbered.
In Fig. S4, the same colour scheme should be used for the ab initio reconstructions as for the scattering curves and P(r) functions. As it is, the orange colour is used for the complexed protein in panel a and for the uncomplexed protein in panel b.
In Fig. S6, the legend states that the complexed structure is orange in both panels a and b, when it is only orange in panel A. What colour scheme is used for panel b? Is the uncomplexed structure still light grey?
In Fig. S7, the structure-based sequence alignment cannot be completely correct, because e.g. the second of the catalytic residues marked by red dots is not conserved in EndoF1 and EndoBT. This manuscript describes the structure of a complex between the IgG-specific endoglycosidase, EndoS, and a biantennary complex type N-glycan. The resulting structure describes the manner by which the endoglycosidase recognizes the branched glycan product in the enzyme active site.
While the manuscript is an effective structure presentation of the enzyme-glycan product complex, there are some deficiencies that need to be resolved in the description of the complex, including: 1) The authors completely under-represent the fact that they already published the overall structure of apo-EndoS (reference 18, Trastoy et al (2014) PNAS). The reference of this prior paper was cryptically cited during the structure description. However, it is so buried that a cursory read of the present paper would have left a reader with an impression that this was the first structural description of this protein. As such, the novelty of the present paper is the structure of the glycan substrate complex, not the initial description of the protein. The details of the complex are well described in the manuscript and the mutagenesis data, coupled with kinetic analysis, are an effective description of how the recognition of the biantennary complex glycan structure is achieved. 2) In Fig 3C the authors attempt to make a structural comparison with other endoglycosidase structures, but this visual representation is not effective. The nature of the isolated loop regions in cartoon representation in the illustration provide no meaningful insight to the differences in structural basis of substrate specificity. The authors state that "Residues involved in the interaction with the conserved Manβ1-4GlcNAcβ1-4GlcNAcβ1 13; core are essentially preserved …" This conservation in interactions with these glycan residues are not evident from this figure. What are the PDB files used in this comparison along with the respective citations of the structures? How many of the structures have substrate complexes? The authors make cursory statements about the comparison with the substrate complex for EndoF3. The previous PNAS paper presented a structural comparison of EndoS with the EndoF3-glycan complex, but the present manuscript needs to show a detailed structural comparison of the EndoF3 complex with the EndoS-glycan complex. What is novel about the present glycan structural complex by comparison to the prior work on EndoF3? 3) In the prior PNAS paper the authors present a model of the docked EndoS complex with IgG and how the EndoS accessory domains enable the insertion of the glycan into the active site. Is this model still relevant based on the present glycan complex? 4) The authors model in the GlcNAc(-1) and NeuAc(+6) and (+10) monosaccharides in Fig. S5, but never describe how that modeling was performed nor is it described in either the legend to Fig. S5 or the body of the paper that these monosaccharides were modeled and not a part of the empirically derived structure. 5) It would be very helpful to show the positions of the catalytic residues that were mutated and provide a model for the catalytic mechanism that resulted in the formation of the enzymatic product complex that is present in the active site. This would best be presented in conjunction with Fig. 2D. 6) In regard to the N-3HB domain, the authors state that: "these data suggest that the N-3HB domain is a key element involved not only in N-linked glycan substrate recognition but also in EndoS-Fc fragment interaction." The data do not really suggest these conclusions. An equally likely interpretation is that the N-3HB domain stabilizes the GH domain structure. The IgG domain would be expected to engage the active site from the opposite side of the GH domain from the N-3HB domain and would not likely directly interact with the IgG domain.
Other minor issues: 1) Please move the Man+7 label out into white space in Fig 2D. 2) Show residue D233A in Fig 2D. Overall, the manuscript is a nice structural presentation of the enzyme-glycan product complex for EndoS and provides tremendous insights into how this enzyme is able to recognize biantennary complex glycan structures. Additional detail is required to provide insights into the novelty of this structure in compared to the prior PNAS paper and to the structure of the EndoF3-glycan complex. No additional experimentation is required, but a more effective presentation of comparative data with other structures would greatly improve the manuscript.
IgG glycans regulate antibody effector functions. EndoS from Streptococcus pyrogenes cleaves specifically complex-type N-linked glycans of IgG and contributes to the immune evasion of the bacterium. Interestingly, EndoS glycosynthase variants can be used to transfer pre-defined complex type N-glycans to intact IgG, which could be very useful to generate therapeutic antibodies with the desired glycoforms. With the aim to better understand the substrate specificity of this enzyme, Trastoy and collaborators have crystallized an inactive variant of endoS in complex with its oligosaccharide product (G2 glycoform). Following the identification of EndoS domains interacting with the G2 glycosform, authors have performed key mutations of amino acid residues to demonstrate their involvement in the recognition of the glycan product. Finally, by comparing the 3D structure of EndoS domains with those of other glycosylhydrolases members of the GH18 family, authors proposed that the substrate specificity of EndoS relies on the unique groove 2 of the enzyme. Although the manuscript is clearly written, of great interest and provides several novel findings, additional experiments are required to fully support the conclusions made by authors: Major comments: 1) My main concern is related to the last part of the manuscript. Based on the functional data depicted in figure 3 and comparison of endoglycosidase 3D structures, authors proposed that the EndoS specificity for complex-type N-linked oligosaccharide relies on its exclusive groove 2 structure (loops 1, 2 and 7), which is significantly different than those of other GH18 endoglycosidases. But authors did not directly demonstrate it. Authors mentioned (Lines 141-143 and Fig. 3C) that loops 1 and 2 of high-mannose-type specific GH18 endoglycosidases are "ordered" and adopt "a betahairpin conformation" but not that of endoS. However, a counterexample of this is the lack of order of loops 1 and 2 of EndoT, which is also a high-mannose-type specific endoglycosidase. Likewise (Line 144), authors mentioned that the Loop7 is "markedly shorter" in high-mannose-type specific GH18 endoglycosidases as compared to that of EndoS. However, EndoF3 that recognizes complextype N-glycans, also have a very short loop 7. A direct and more convincing proof of the involvement of loop 1, 2 and 7 roles in the substrate specificity of EndoS would be the substitution or introduction of residues (or even entire loops) from corresponding loops in high-mannose-typespecific GH18 endoglycosidases (Ex. EndoH), following by the measurement of binding of different glycan products (G2, High-mannose-type glycans,….) or, if possible, measurement of activity of EndoS variants on different substrates.
Minor comments 1) Figures 3a and b-mutations are presented in terms of amino acids (ex. R119A/E130A/K133A) whereas the corresponding text (Lines 104-128) refers to Loops, making the demonstration hard to follow. Please clearly present the Loops also in both figures. 2) Figure 3c: For Endo S, Loop-1 is indicated to interact with two of the core mannoses whereas Figure 3a suggest that Loop-1 interacts with terminal glucosamine and galactose. Please, correct or explain this discrepancy. 3) Figure 3c: Loops 3, 4, 5 and 6 are not at all commented into the text, compromising the interest of presenting them into the table. I would suggest moving them in supplementary data 4) Figure 3b-A presentation of mutant ΔN is lacking in the legend. 5) Line 53-typo « glycoformsand and » 6) Line 69-define or explain « hybrid Ig domain » 7) Line 95-Correct: "residues of each arm adopt two "alternative" (instead of different) conformations" and precise "into the crystal". Likewise, in Figure S5-Correct "the two alternative conformations of G2 product". 8) Line 125-the mode of interaction of the SpA C domain (similar to endoF N-3HB) with IgG or IgM is not clearly explained. The reader cannot even discriminate between protein-protein or proteinglycan interfaces … 9) Figure S1-For a better understanding, please add some explanations on the fig S1 legend. I also don't understand why the item "Immune evasion" is below the second arrow and not below the first arrow. 10) Figure S6-Line 110-Correct "a" and "b" instead of "A" and "B". 11) Figure S6A-Loop 2 should be highlighted in the figure. 12) Lines 138-139-steric hindrance is hypothesized for EndoS, and compared to tridimensional permissiveness in EndoF3 or EndoH, but no 3D illustration of EndoF3 or H groove 2 corroborate the comments (as Figures 3c focuses on EndoS and Figure S7 presents truncated loops in a series of Endo enzymes). This should be illustrated.
1. The manuscript by Trastoy et al. describes primarily the crystal structure of a complex of the IgGspecific endoglycosylase S (EndoS) with the G2 product of its reaction with the glycosylation on Asn127 of the Fc domain of IgG. The interactions between enzyme and product are described in detail. SAXS studies are used to show that there are no large-scale conformational changes in solution upon product binding. Mutations in loops shown from the crystal structure to be important for enzyme-product interactions are made and their effect on enzymatic activity against the model IgG Rituximab are analysed, highlighting the importance of loops 1, 6 and 7. In summary, while the results are undeniably interesting, it is uncertain to me to what extent they would influence thinking in the field. The results are similar to those obtained for EndoF3 more than 15 years ago. I feel that detailed comparison with EndoF3, which has very much overlapping substrate specificity, is lacking. How do the differences between EndoS and EndoF3 explain how EndoF3 accommodates the same product as EndoS while also accepting triantennary glycosylations? ANSWER: We very much appreciate that you consider our study of undeniable interest, and all the constructive comments/suggestions you made to improve the article.

In the new version of the manuscript, we have now included an extensive and detailed comparison and discussion on the structural differences of the EndoS and EndoF 3 glycosidase domain, and the importance of these differences in defining the glycan specificity of each enzyme. Please see 'Structural comparison of EndoS with GH18 family of endoglycosidases' in the new 'Results and Discussion' section, and new Figures 8 and 9.
2. Finally, the manuscript is in places poorly written, with poor figure legends and inconsistent reference to figures. I would recommend either that the work is sent to a more specialised journal or resubmitted here with major revision. ANSWER: The original manuscript was transferred from another NPG journal with a considerably more condensed format. In the process of reshaping the manuscript, the comments made by the reviewers have been invaluable.
3. Some comments to help improve the manuscript: I think that the picture of the glycosylation in Fig. S2 should be in the main manuscript. 12. lines 108-110: It's unnecessary to write "Ala" at the end of every mutation, as it was specified in the previous sentence that they are all Ala mutations. 16. Furthermore, Figure S6 is in contradiction to Figure S4, where the crystal structure of the complex is superposed with the envelope of both the bound and unbound forms to assert that there is no difference in conformation. In Fig. S6b one can clearly see a change in the conformation of two domains. 7 42. Table S2: The symbol ≤ should presumably be ±. There seems to be an extra digit, "1" after the end of the second Porod volume estimate. Table 2 accordingly. 43. The units of I(0) cannot be reciprocal Ångström, as this is the unit of q, on the x-axis of the scatterig curve. I(0) is on the y axis and has arbitrary units unless normalised using a known standard (was this done?) Likewise, why is the unit reciprocal cm used for the I (0)  44. "Comparation" should be "comparison".

ANSWER: We have modified the text accordingly.
45. Why is the CRYSOL chi-squared value so high? Did the authors consider EOM? ANSWER: We interpret that the chi-squared is high because the EndoS shows some flexibility in solution as depicted in the normalized Kratky plot (please see new Figure S4b). Following the reviewer's suggestion, we have used EOM and SREFLEX to evaluate the flexibility of the system. We have obtained the best chi-squared values using SREFLEX and we have included this new data in the new version of the manuscript.
46. The program "SCATER" should be spelled "SCÅTTER" ANSWER: We have modified the text accordingly. Fig. S1, the figure gives the misleading impression that the glycosylations are highly exposed on the exterior of the Fc domain, when they are in fact on the inside. While I appreciate that this can be hard to draw, it should at least be indicated in the figure legend. It should be explained what the right-hand reactive group of the molecule used to attach uniform glycosylations to IgG is. The cleavage point leading to the G2 product should be indicated. The symbols for the different types of saccharide unit should be defined. There should be a legend for this figure! ANSWER: We have modified the Supplementary Figure 1 (now Figure 1) and its legend accordingly. 8 48. Fig. S2 legend: "996-1191 residues" should read "residues 996-1191". The sequence should be numbered.

ANSWER: We have modified the text accordingly.
49. In Fig. S4, the same colour scheme should be used for the ab initio reconstructions as for the scattering curves and P(r) functions. As it is, the orange colour is used for the complexed protein in panel a and for the uncomplexed protein in panel b. Figure 4 accordingly. Fig. S6, the legend states that the complexed structure is orange in both panels a and b, when it is only orange in panel A. What colour scheme is used for panel b? Is the uncomplexed structure still light grey?  Fig. S7, the structure-based sequence alignment cannot be completely correct, because e.g. the second of the catalytic residues marked by red dots is not conserved in EndoF1 and EndoBT. 1. This manuscript describes the structure of a complex between the IgG-specific endoglycosidase, EndoS, and a biantennary complex type N-glycan. The resulting structure describes the manner by which the endoglycosidase recognizes the branched glycan product in the enzyme active site. While the manuscript is an effective structure presentation of the enzyme-glycan product complex, there are some deficiencies that need to be resolved in the description of the complex, including:

In
The authors completely under-represent the fact that they already published the overall structure of apo-EndoS (reference 18, Trastoy et al (2014) PNAS). The reference of this prior paper was cryptically cited during the structure description. However, it is so buried that a cursory read of the present paper would have left a reader with an impression that this was the first structural description of this protein. As such, the novelty of the present paper is the structure of the glycan substrate complex, not the initial description of the protein. The details of the complex are well 9 described in the manuscript and the mutagenesis data, coupled with kinetic analysis, are an effective description of how the recognition of the biantennary complex glycan structure is achieved.

ANSWER: We very much appreciate your considerations about the novelty of our study, and all the constructive comments/suggestions you made to improve the article.
The original manuscript was transferred from another NPG journal, with a considerably more condensed format. In the process to reshape the manuscript, the comments made by the reviewers have been invaluable. In that sense, (i) we have included the original unliganded crystal structure from reference 18 in the new 'Introduction' and 'Results and Discussion' sections and (ii) extended the comparison between both crystal structures in the new 'Results and Discussion' section and new Supplementary Figures 3 and 4. Fig 3C the authors attempt to make a structural comparison with other endoglycosidase structures, but this visual representation is not effective. The nature of the isolated loop regions in cartoon representation in the illustration provide no meaningful insight to the differences in structural basis of substrate specificity. The authors state that "Residues involved in the interaction with the conserved Manβ1-4GlcNAcβ1-4GlcNAcβ1-core are essentially preserved …" This conservation in interactions with these glycan residues are not evident from this figure. The authors make cursory statements about the comparison with the substrate complex for EndoF3. The previous PNAS paper presented a structural comparison of EndoS with the EndoF3-glycan complex, but the present manuscript needs to show a detailed structural comparison of the EndoF3 complex with the EndoS-glycan complex. What is novel about the present glycan structural complex by comparison to the prior work on EndoF3?

ANSWER: In this new version of the manuscript, we have described in detail the structural differences between members of the GH18 family of endoglycosidase to support the substrate specificity. Please see 'Structural comparison of EndoS with GH18 family of endoglycosidases' in the 'Results and Discussion
ANSWER: In the new version of the manuscript, we have now included an extensive and detailed comparison and discussion on the structural differences of the EndoS and EndoF 3 glycosidase domains, and the importance of these differences to define the glycan specificity of each enzyme. Please see 'Structural comparison of EndoS with GH18 family of endoglycosidases' in the new 'Results and Discussion' section, and new Figures 8 and 9. 3. In the prior PNAS paper the authors present a model of the docked EndoS complex with IgG and how the EndoS accessory domains enable the insertion of the glycan into the active site. Is this model still relevant based on the present glycan complex? ANSWER: Although we expect the IgG to accommodate into the concave interior of EndoS (please see Figure S6 in PNAS paper), the new EndoS-G2 crystal structure strongly support the occurrence of an important conformational change on the Fc fragment (i) to expose the glycan/s and (ii) to accommodate them into the deep grooves of the glycoside hydrolase domain.
4. The authors model in the GlcNAc(-1) and NeuAc(+6) and (+10) monosaccharides in Fig. S5, but never describe how that modeling was performed nor is it described in either the legend to Fig. S5 or the body of the paper that these monosaccharides were modeled and not a part of the empirically derived structure.

ANSWER: We agree with the reviewer. We have included this information in the Methods section.
5. It would be very helpful to show the positions of the catalytic residues that were mutated and provide a model for the catalytic mechanism that resulted in the formation of the enzymatic product complex that is present in the active site. This would best be presented in conjunction with Fig. 2D.
ANSWER: We agree with the reviewer. As suggested, we have included a new Figure 5 displaying (i) the proposed catalytic mechanism of the reaction and (ii) the position of the mutated residues in the active site.
6. In regard to the N-3HB domain, the authors state that: "these data suggest that the N-3HB domain is a key element involved not only in N-linked glycan substrate recognition but also in EndoS-Fc fragment interaction." The data do not really suggest these conclusions. An equally likely interpretation is that the N-3HB domain stabilizes the GH domain structure. The IgG domain would be expected to engage the active site from the opposite side of the GH domain from the N-3HB domain and would not likely directly interact with the IgG domain. 9. Overall, the manuscript is a nice structural presentation of the enzyme-glycan product complex for EndoS and provides tremendous insights into how this enzyme is able to recognize biantennary complex glycan structures. Additional detail is required to provide insights into the novelty of this structure in compared to the prior PNAS paper and to the structure of the EndoF3-glycan complex. No additional experimentation is required, but a more effective presentation of comparative data with other structures would greatly improve the manuscript.
ANSWER: We very much appreciate that you consider our study provides a tremendous insight into how this enzyme is able to recognize biantennary complex glycan structures, and all the constructive comments/suggestions to improve the article. Following the reviewer suggestions, we have made a strong effort in order to effectively present the comparison of the structural data and improve the readability of the entire manuscript.
Reviewer #3: 1. IgG glycans regulate antibody effector functions. EndoS from Streptococcus pyrogenes cleaves specifically complex-type N-linked glycans of IgG and contributes to the immune evasion of the bacterium. Interestingly, EndoS glycosynthase variants can be used to transfer pre-defined complex type N-glycans to intact IgG, which could be very useful to generate therapeutic antibodies with the desired glycoforms. With the aim to better understand the substrate specificity of this enzyme, Trastoy and collaborators have crystallized an inactive variant of endoS in complex with its oligosaccharide product (G2 glycoform). Following the identification of EndoS domains interacting with the G2 glycosform, authors have performed key mutations of amino acid residues to demonstrate their involvement in the recognition of the glycan product. Finally, by comparing the 3D structure of EndoS domains with those of other glycosylhydrolases members of the GH18 family, authors proposed that the substrate specificity of EndoS relies on the unique groove 2 of the enzyme. Although the manuscript is clearly written, of great interest and provides several novel findings, additional experiments are required to fully support the conclusions made by authors: ANSWER: We very much appreciate that you consider our study to be of great interest, and all the constructive comments/suggestions you made to improve the article.

Major comments:
12 My main concern is related to the last part of the manuscript. Based on the functional data depicted in figure 3 and comparison of endoglycosidase 3D structures, authors proposed that the EndoS specificity for complex-type N-linked oligosaccharide relies on its exclusive groove 2 structure (loops 1, 2 and 7), which is significantly different than those of other GH18 endoglycosidases. But authors did not directly demonstrate it. Authors mentioned (Lines 141-143 and Fig. 3C) that loops 1 and 2 of high-mannose-type specific GH18 endoglycosidases are "ordered" and adopt "a betahairpin conformation" but not that of endoS. However, a counter example of this is the lack of order of loops 1 and 2 of EndoT, which is also a high-mannose-type specific endoglycosidase.
We have modified Figure 3c (new Figure 7) to show the correct loop 2 of EndoT and confirm that it is also a β-hairpin like in the other high-mannose-type specific endoglycosidases.
3. Likewise, (Line 144), authors mentioned that the Loop7 is "markedly shorter" in high-mannosetype specific GH18 endoglycosidases as compared to that of EndoS. However, EndoF3 that recognizes complex-type N-glycans, also have a very short loop 7.
EndoF 3 has a shorter loop 7 than EndoS because EndoF 3 can accept both triantennary and bianntenary complex type oligosaccharides as substrates (please see Figure 8). EndoS is not able to hydrolyze trianntenary complex type oligosaccharides. As depicted in Figure 8, the long loop 7 observed in EndoS would block the entrance of the third antenna.

4.
A direct and more convincing proof of the involvement of loop 1, 2 and 7 roles in the substrate specificity of EndoS would be the substitution or introduction of residues (or even entire loops) from corresponding loops in high-mannose-type-specific GH18 endoglycosidases (Ex. EndoH), following by the measurement of binding of different glycan products (G2, High-mannose-type glycans,….) or, if possible, measurement of activity of EndoS variants on different substrates. ANSWER: Switching loops between endoglycosidases might be a challenging task in order to keep a correct folding of the conserved (α/β) 8 barrel of the glycoside hydrolase domains. The proposed study is out of the scope of the current manuscript. However, following the reviewer suggestion, we provide additional support on the correlation between the architecture of the endoglycosidase loops surrounding the glycan binding pocket and the specific activity of this family of enzymes. Taking into account the analysis performed in Figure 7, the architecture of loops 1, 2 and 7 in EndoBT is similar to that observed in EndoT, EndoH and EndoF 1 , strongly suggesting that the enzyme is an endoglycosidase specific for high-mannose type oligosaccharides. It is worth mentioning that the crystal structure of EndoBT, a putative endoglycosidase of unknown function/enzymatic activity/glycan specificity, was solved in its unliganded form. We therefore determined the capacity of EndoBT to hydrolyse bianntenary complex-type N-linked glycans and/or high mannose type N-linked glycans from IgG1 antibodies. As depicted in Figure 9e, the hydrolytic assays showed that EndoBT is able to hydrolase high-mannose type IgG1 but not bianntenary complex-type IgG1.

Minor comments 13
Figures 3a and b-mutations are presented in terms of amino acids (ex. R119A/E130A/K133A) whereas the corresponding text (Lines 104-128) refers to Loops, making the demonstration hard to follow. Please clearly present the Loops also in both figures. Figures 3a and 3b (new Figures 6a and 6b) accordingly. Figure 3c: For Endo S, Loop-1 is indicated to interact with two of the core mannoses whereas Figure 3a suggest that Loop-1 interacts with terminal glucosamine and galactose. Please, correct or explain this discrepancy.

4.
ANSWER: The reason is the orientation of the G2 product in the Figure. Arg119 interacts with Man (-7) whereas Trp121 interacts with Man (-7) and Man (-2). The terminal GlcNAc (-10) and Gal (-9) residues of the G2 product protrude away from groove 2 of EndoS. Figure   9. Line 95-Correct: "residues of each arm adopt two "alternative" (instead of different) conformations" and precise "into the crystal". Likewise, in Figure S5-Correct "the two alternative conformations of G2 product". ANSWER: We have modified the text accordingly. 14 10. Line 125-the mode of interaction of the SpA C domain (similar to endoF N-3HB) with IgG or IgM is not clearly explained. The reader cannot even discriminate between protein-protein or protein-glycan interfaces … ANSWER: We agree with the reviewer. In the new version of the manuscript, we have made an effort to clearly explained the proposed mode of interaction of the SpA C domain with IgG or IgM and added references accordingly.

5.
11. Figure S1-For a better understanding, please add some explanations on the fig S1 legend. I also don't understand why the item "Immune evasion" is below the second arrow and not below the first arrow. Figure S1 (new Figure 1) and the associated legend.

ANSWER: We have modified the
12. Figure S6-Line 110-Correct "a" and "b" instead of "A" and "B". 14. Lines 138-139-steric hindrance is hypothesized for EndoS, and compared to tridimensional permissiveness in EndoF3 or EndoH, but no 3D illustration of EndoF3 or H groove 2 corroborate the comments (as Figures 3c focuses on EndoS and Figure S7 presents truncated loops in a series of Endo enzymes). This should be illustrated. ANSWER: We agree with the reviewer. We have added two new figures illustrating the EndoS, EndoF 3 , EndoH, EndoF 1 , EndoT and EndoBT surfaces (Figures 8 and 9). We have also included new molecular docking experiments of high mannose-type oligosaccharide GlcNAc 1 Man 9 into the GH18 family endoglycosidases able to hydrolyze high mannose-type Nlinked glycans and tri-antennary complex-type oligosaccharide into EndoF 3 . In these figures we can observed that the binding pocket of EndoS is unique in accommodating bianntenary complex-type N-linked glycans while excluding other glycan types due to steric hindrance.
We wish to take this opportunity to thank the reviewers for her/his thoughtful suggestions that have made the manuscript so much better.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The revised manuscript by Trastoy et al. is a significant improvement on the original version. The authors have answered all of my original questions and I think that the article is now almost suitable for publication. The detailed comparison with EndoF3 is much appreciated. However there are still some errors in the manuscript that need to be corrected.
The first time that Z-scores are mentioned it should read "DALI Z-score" and a reference to DALI should also be inserted there.
On page 5, line 18 "the calculated buried surface area ... was" should be "is", as it presumably hasn't changed since it was calculated.
Same page, line 19: "form hydrogen bonds" On p.8 line 11 there is a mysterious bold text "BEA", which seems to be a request for the first author to complete the sentence. However this hasn't been done. p. 8: It's still unclear to me what evidence the current structure provides for the proposed reaction mechanism for GH18 enzymes apart from the fact that important residues are present. Does the product conformation say anything about the proposed distortion of the substrate?
The loop containing residues 303 and 305 is referred to as loop 6 in the text but loop 5 in Fig. 6. Also in Fig. 6, loop 3 is described as being light red when in fact it is some kind of gold colour. Please check all nomenclature and colours so that everything is consistent. p. 10: How can the cavity be bigger in EndoS (almost 3 times the size that it is in EndoF3) when EndoF3 has the ability to bind a triantennary glycosylation? Are the authors comparing like with like? p. 11: What do the authors mean by an "exclusive" groove 2? Groove 2 is present in both EndoF3 and EndoS, since both can bind the G2 product. EndoS excludes the triantennary substrate by sterically blocking a groove that is present in EndoF3.
One technical point that I missed the first time around: In the SAXS figures and tables, it is not clear whether the curves and values obtained were obtained from a single concentration or through the merging of data from the several concentrations used. This is an important detail, and it should be included in the Methods and figure and table legends.
One last point that is more a question of taste: pdb should be written PDB, as it is an abbreviation for Protein Data Bank.
Reviewer #2 (Remarks to the Author): This revised manuscript describes the structure of a complex between EndoS, and a biantennary complex type N-glycan substrate analog to map the structural determinants of substrate specificity. The revised manuscript was greatly improved in response to prior reviewer critiques, but still has several concerns that should be addressed prior to publication. 1) There are numerous places in the revised manuscript where the text description essentially catalogs structural information that would be better handled as a presentation in table form. Examples include manuscript lines 124-127, lines 131-142, lines 198-205, but similar issues are present in other parts of the manuscript. This catalog-like description breaks up the readability of the manuscript and the level of detail is not essential for the body of the manuscript text. 2) While the manuscript now mentions the structure determination of EndoS in their prior PNAS paper, the present paper has an extended description of the overall structure of the multi-domain protein (lines 83-91) that has already been described in the prior paper. A more appropriate presentation would to a focus on the differences in the present structure relative to the prior published structure and point readers to the prior paper for a description of the overall domain structure.
3) The present studies on full length EndoS employ a double mutant for (D233A/E235L) of the recombinant enzyme, but the authors never state why they chose these residues to mutate. These are not the mutants employed in the prior work and it is unclear why they needed to employ a mutant form of the enzyme at all, given the structure of the ligand resembling the enzymatic product in the structural studies. The authors should explain the basis for these mutations. There is an oblique comment on line 172 regarding the fact that the double mutant is inactive, but no description why these residues were employed. 4) Throughout the manuscript the authors state that they are employing a G2 structure as the substrate analog, but in reality the ligand is a truncated structure that represents the equivalent of the enzymatic product of EndoS action (cleaved between the two core GlcNAc residues). This should be clarified in Fig 5A. Again, it is not clear why the D233A/E235L mutant was employed and how these mutations impacted in the interaction with the reducing terminal GlcNAc residue. 5) On line 161 the authors should define BEA. 6) Line 181 states that mutations in loop 6 abolished hydrolytic activity and the data is shown in Fig.  6 A. Loop 6 mutants are not shown in Fig. 6A or any other figure. 7) The structural comparisons illustrated in Fig. 8, Fig. 9 and supplementary S5 are effective in examining substrate interactions in the respective enzyme catalytic domains. However, it would also be helpful to compare the overall structures of the respective proteins. Do the other enzymes have the equivalent of the additional accessory domains found in EndoS? If so, then how to they compare? If not, that is also relevant, because it implies that the additional domains in EndoS are employed for selective interactions with Igs that are not present nor relevant for the other enzymes.