Structural Basis of Chemokine CXCL8 Monomer and Dimer Binding to Chondroitin Sulfate: Insights into Specificity and Plasticity
, Balaji Nagarajan 3,†, Nehru Viji Sankaranarayanan 3,4, Prem Raj B. Joseph 1,2, Umesh R. Desai 3 and Krishna Rajarathnam 1,2,5,6,*Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
In this study, the authors investigated the structural basis of chondroitin sulfate (CS) binding by monomeric and dimeric CXCL8 using an integrated approach combining NMR spectroscopy, experimentally driven molecular docking, and molecular dynamics simulations. The findings provide important insights into the CS-mediated fine-tuning of CXCL8 receptor signaling and contribute to a more comprehensive understanding of neutrophil trafficking and effector functions. However, the manuscript in its current form requires substantial and critical revisions before it can be considered for publication.
Major points #1
The authors employed NMR data–based molecular docking and molecular dynamics simulations to identify basic amino acid residues involved in the interaction between CS oligosaccharides and both the monomeric and dimeric forms of CXCL8. However, these predictions are not supported by any experimental validation. As a result, the functional significance of the identified residues remains largely speculative.
To address this critical limitation, the authors should provide experimental evidence to substantiate the proposed roles of these amino acid residues. For instance, does substitution of Lys54 differentially affect the interaction between CS oligosaccharides and the monomeric versus dimeric forms of CXCL8? Furthermore, does binding of CS oligos to CXCL8 actually interfere with the interaction of CXCL8 with the corresponding receptor? In addition, do mutations of these predicted basic residues restore or otherwise modulate the interaction between CXCL8 and its receptor?
Experimental validation addressing these points is essential to support the conclusions drawn from the computational analyses.
Major points #2
The CS oligosaccharides purchased from Iduron appear to be a mixture of oligosaccharides with different degrees of polymerization, ranging from dp4 to dp20. In many figures presented in this manuscript, CS oligosaccharides with dp14 are used. It is therefore unclear whether the dp14 species was further separated and purified from the purchased CS oligosaccharide mixture. If so, this procedure should be clearly described in the Methods section. If not, the authors should explain how the chain length–dependent binding of CS oligosaccharides to CXCL8 was assessed.
Furthermore, it should be noted that the C4–C5 double bond described in lines 129–130 is formed via an elimination reaction rather than by the endolytic action of chondroitinase ABC.
Major points #3
The legends for the Supplementary Figures are missing. In addition, the descriptions of the CS oligosaccharides in Figures 6A and 8A are incomplete. Moreover, CS residues 1, 3, 5, and 7 appear to be GalNAc4S rather than GalNAc6S.
Author Response
Major points #1. The authors employed NMR data–based molecular docking and molecular dynamics simulations to identify basic amino acid residues involved in the interaction between CS oligosaccharides and both the monomeric and dimeric forms of CXCL8. However, these predictions are not supported by any experimental validation. As a result, the functional significance of the identified residues remains largely speculative. To address this critical limitation, the authors should provide experimental evidence to substantiate the proposed roles of these amino acid residues. For instance, does substitution of Lys54 differentially affect the interaction between CS oligosaccharides and the monomeric versus dimeric forms of CXCL8? Furthermore, does binding of CS oligos to CXCL8 actually interfere with the interaction of CXCL8 with the corresponding receptor? In addition, do mutations of these predicted basic residues restore or otherwise modulate the interaction between CXCL8 and its receptor? Experimental validation addressing these points is essential to support the conclusions drawn from the computational analyses.
Response:
We appreciate the reviewer’s suggestion that mutational analyses of the individual basic amino acids involved in GAG recognition and receptor signaling would be valuable for further validating our predictions. We agree that such experiments could provide important complementary insights. However, undertaking these studies would require substantial additional resources, personnel, and time, and therefore falls beyond the scope of the present manuscript. The primary goal of this work is to describe the structural basis for CXCL8 monomer and dimer recognition of CS, to place these findings within the context of current understanding in the field, and to examine how the structural features of the monomer and dimer, as well as differences in GAG structure and sulfation patterns, influence binding. We believe that we have addressed these objectives thoroughly and that our results offer meaningful new insights into GAG recognition and chemokine function.
Regarding whether binding of CS oligos to CXCL8 actually interfere with the interaction of CXCL8 with the corresponding receptor — As describe in the manuscript, our current studies show extensive overlap between receptor and CS binding interfaces. In addition, previous studies examining receptor and heparin binding have demonstrated unambiguously that these interactions are mutually exclusive, such that heparin-bound CXCL8 monomers and dimers cannot bind the CXCR1 and CXCR2 receptors, and vice versa (ref. 18). Given that our results indicate that the binding interfaces in HS/heparin–CXCL8 and CS–CXCL8 complexes are largely similar, it is therefore reasonable to expect that CS-bound CXCL8 would likewise be impaired in its ability to interact with its receptors.
Major points #2. The CS oligosaccharides purchased from Iduron appear to be a mixture of oligosaccharides with different degrees of polymerization, ranging from dp4 to dp20. In many figures presented in this manuscript, CS oligosaccharides with dp14 are used. It is therefore unclear whether the dp14 species was further separated and purified from the purchased CS oligosaccharide mixture. If so, this procedure should be clearly described in the Methods section. If not, the authors should explain how the chain length–dependent binding of CS oligosaccharides to CXCL8 was assessed. Furthermore, it should be noted that the C4–C5 double bond described in lines 129–130 is formed via an elimination reaction rather than by the endolytic action of chondroitinase ABC.
Response:
We purchased dp14 oligosaccharides (Catalogue no. CSO14) from Iduron, rather than unfractionated CS. The product sheet for dp14 oligosaccharide states that ‘this oligosaccharide has been prepared by high resolution gel filtration of partial chondroitin ABC lyase digestion of mixed isomer chondroitin sulfate’. Further, the product sheet states that ‘Uronic acid (ΔUA) at the non-reducing ends of the oligosaccharides has a C4-C5 double bond as a result of endolytic scission.’ Chondroitinase ABC is a polysaccharide lyase, not a hydrolase. It cleaves CS chains by an endolytic β- elimination reaction, which generates an unsaturated uronic acid residue (ΔUA) at the non-reducing end of the product. This reaction specifically introduces the C4–C5 double bond in the uronic acid.
Major points #3. The legends for the Supplementary Figures are missing. In addition, the descriptions of the CS oligosaccharides in Figures 6A and 8A are incomplete. Moreover, CS residues 1, 3, 5, and 7 appear to be GalNAc4S rather than GalNAc6S.
Response:
We thank the reviewer for bringing this to our attention. In the original manuscript, the legends for the supplementary figures were placed beneath the legends for the main figures. We have now placed the legends under each of the supplementary figures. In addition, we have now corrected the figure legends for figures 6 and 8.
Reviewer 2 Report
Comments and Suggestions for Authors
Summary:
The study by Mahler et al. examines the molecular interactions between monomeric and dimeric forms of the chemokine CXCL8 (IL-8) and the glycosaminoglycans (GAGs) chondroitin sulfate (CS) and heparan sulfate (HS). Using solution NMR spectroscopy in combination with molecular docking and molecular dynamics simulations, the authors show that the CXCL8 dimer is the high-affinity CS-binding species, whereas monomer binding is comparatively weak. Both monomer and dimer engage CS through essentially the same set of basic residues clustered in the N-loop, 40s loop, and C-helix. The CS-binding interface corresponds to a contiguous positively charged surface on the CXCL8 monomer and accommodates both CS4S and CS6S variants with similar binding modes despite differences in sulfation patterns. Notably, the same residues mediate binding to CS and to heparin/HS, indicating that CXCL8 employs a conserved binding platform to interact with structurally distinct GAGs. Several CS-binding residues overlap with known CXCR1/2 receptor-interacting sites, supporting the conclusion that GAG-bound CXCL8 is likely impaired in receptor engagement. Overall, the study provides informative experimental and computational insights into how CXCL8 achieves specificity and adaptability in GAG recognition. However, interactions between CXCL8 and chondroitin sulfate and other GAGs have been investigated previously, and the novelty of the present work would benefit from a clearer articulation of the specific knowledge gaps left by earlier studies that this work aims to address. In addition, the authors are encouraged to consider the following points.
Comments:
(1) Comparison with previous studies
The authors should expand the comparison of their findings with previous studies that have examined CXCL8–GAG interactions. In particular, Pichert et al. (2012) (https://pmc.ncbi.nlm.nih.gov/articles/PMC3230280/) investigated CXCL8 binding to several GAGs, including CS4 and CS6, using NMR spectroscopy and molecular modeling. A more explicit comparison with these results would be valuable. How do the binding modes identified in the present study relate to those reported by Pichert et al.? Do the current data support the previously proposed binding poses for CS4 and CS6, or do they suggest alternative interaction geometries? In addition, Pichert et al. reported stronger binding of CS6 relative to CS4, based on both NMR chemical shift perturbations and MM/GBSA binding free energy calculations. Do the present results similarly indicate differences in binding affinity or energetics between CS4 and CS6?
(2) Quality of data presentation
To facilitate evaluation of the agreement between the NMR data and the MD-derived binding models for CS4 and CS6, a more direct visual comparison would be helpful. For example, Figure 5 could be expanded to include a mapping of NMR chemical shift perturbations onto the CXCL8 molecular surface or ribbon representation, allowing the reader to directly compare the locations of CSPs with basic residues and predicted GAG-binding sites. Furthermore, in Figure 8, which presents the interaction models of CS4/CS6 with the CXCL8 dimer, it would improve clarity to display both chains of the dimer. This would help determine whether interactions are strictly confined to one monomer or whether residues from the second chain also contribute to CS binding.
In addition, captions are missing for all Supplementary Information figures and should be added. The x-axis labels in Figures S1A and S1B should also be checked, as they appear to indicate [CXCL8] rather than the expected [CS dp14].
(3) Missing method descriptions
The manuscript does not sufficiently describe the free energy calculations (MM/PBSA or MM/GBSA). A brief explanation of the methodology used to compute binding free energies, as well as the per-residue energy decomposition, should be included in the Methods section. Related to this point, the term “MM/PB(GB)SA” used in line 296 is potentially confusing, as typically only one solvation model - either a PBSA or GBSA - is applied. Please clarify which approach was used.
(4) Missing data
In lines 212–216, the authors state that titrations with CS oligosaccharides of different lengths (dp8 to dp14) yield similar CSP profiles with varying magnitudes. To substantiate this claim, it would be useful to include CSP data for the shorter CS oligosaccharides, either in the main text or in the Supplementary Information, in addition to the dp14 data already shown.
(5) Clarification of conclusions and interpretation
In lines 345–347, the authors conclude that, for the CXCL8 dimer, binding across the dimer interface is not supported by the NMR data and that only intra-monomer binding is consistent with the experimental observations. The basis for this conclusion is not entirely clear from the current presentation. Additional explanation of how the NMR data shown in Figure 2 lead to this inference would be helpful. Related to this point, the rationale for using CS dp8 in the modeling, while the NMR experiments were performed with CS dp14, should be explained more explicitly.
The authors also report differences in residue-specific interactions between CS4 and CS6. For example, K15 appears to form more hydrogen bonds with CS4 than with CS6, whereas R68 interacts more prominently with CS6, and S44 appears to engage selectively with CS6. Similarly, in the dimer, R47 is suggested to contribute primarily to CS4 binding, while H18 and R68 appear more relevant for CS6 binding. It would be valuable for the authors to discuss whether these observations reflect an underlying structural or physicochemical basis for differential specificity between CS4 and CS6, or whether they may arise from limited sampling in the MD simulations. Clarification of whether additional MD replicates may alter or equilibrate these interaction patterns would strengthen the interpretation.
Author Response
Reviewer #2.
Comments:
(1) Comparison with previous studies
The authors should expand the comparison of their findings with previous studies that have examined CXCL8–GAG interactions. In particular, Pichert et al. (2012) (https://pmc.ncbi.nlm.nih.gov/articles/PMC3230280/) investigated CXCL8 binding to several GAGs, including CS4 and CS6, using NMR spectroscopy and molecular modeling. A more explicit comparison with these results would be valuable. How do the binding modes identified in the present study relate to those reported by Pichert et al.? Do the current data support the previously proposed binding poses for CS4 and CS6, or do they suggest alternative interaction geometries? In addition, Pichert et al. reported stronger binding of CS6 relative to CS4, based on both NMR chemical shift perturbations and MM/GBSA binding free energy calculations. Do the present results similarly indicate differences in binding affinity or energetics between CS4 and CS6?
Response:
We observed significant chemical shift perturbations for residues K15, H18, K20, K23, R47, K54, R60, K64, K67, and R68 (Fig. 2). In the manuscript by Pichert et al (PMC3230280), the bar graph shows perturbations for H18, K20, K23, K54, and K64 for both CS variants and perturbations for R60, K67 and R68 only for CS6. Moreover, the perturbations for K67 and R68 were minimal. They do not observe perturbations for K15 and R47 and also do not discuss the perturbations observed for residues K23 and K54. They also observe the highest CSPs for A69 and E70 that we do not.
However, there are critical differences between the two NMR studies. Whereas they carried out their NMR titrations using a 1 mM sample (final protein: CS ratio of 1:0.7), we used a 50 μM sample (final protein: GAG ratio of 1:4). We could reach higher protein to CS ratios (1:4) due to our low starting protein concentration, could obtain a binding isotherm, and were not limited by precipitation issues. In addition, we used a longer oligosaccharide – dp14 vs. dp6. As a result, the extent of chemical shift perturbations and the overall CSP profiles differ between our studies, as evident from the bar graphs (Fig. 4 in their paper and Fig. 2 in our paper). Consequently, it is difficult to determine whether the
differences in our CSP profiles are real or due to differences in experimental variables. For example, we observe that the perturbation for K67 is significant only for dp14 and not for shorter oligosaccharides. Finally, note that they used the 1-77 version of the protein, and so residue numbering will vary by 5 (our R60 corresponds to R65 in their paper). We have made the necessary corrections of their numbering for meaningful comparison.
Regarding docking studies — Whereas we used an octasaccharide (dp8) for our docking studies, Pichert et al. used the tetrasaccharide (dp4). Furthermore, they conclude that all of the different GAGs used in their study (16 different derivatives, Table 5) essentially interact with the same set of core residues – H18, K20, R60, K64, K67, and R68. More importantly, they do not explicitly discuss the differences at a single residue level in CXCL8 binding to CS4 and CS6.
Unfortunately, the extent and depth of our investigation - oligosaccharide length and residue-specific results – precludes direct comparison differences between the two studies.
Regarding binding affinities — Considering that ionic interactions play an important role in GAG interactions and that the measured binding constant varies between techniques and between buffer conditions, these values must be interpreted with caution. The focus of our study is not as much on the actual binding constant values but on the binding modes and the implications for their in vivo function. Pichert et al. report binding constants measured from changes in fluorescence intensity of tryptophan and not from chemical shift changes. For NMR studies, they used a 1 mM protein sample that precluded observing a binding isotherm for calculating a binding constant. For this reason, they mention that they used fluorescence methods. For their Kd measurement studies, they used a 1 μM protein sample, a concentration that is 1000-fold lower than that used in their NMR studies and also a concentration at which CXCL8 is very likely to be a monomer.
They report Kds of 1.4 and 4.8 μM for CS6S and CS4S, and also that Kd for HA (which does not carry any sulfates) as 5.5 μM, a value comparable to both CS and DS (1.5 to 5.5 μM). We used a 50 μM sample for our NMR studies that is 20 times lower than that used in the other study, and this allowed us to obtain a binding isotherm, as shown in the supplementary figures. Whereas they used a CS hexasaccharide, we used a 14mer. Our NMR studies show that the affinity decreases (higher Kd) with decreasing oligosaccharide length. Therefore, our observations of weaker binding in terms of actual Kd values and between the monomer and dimer disagree with those reported by Pichert et al.
(2) Quality of data presentation. To facilitate evaluation of the agreement between the NMR data and the MD-derived binding models for CS4 and CS6, a more direct visual comparison would be helpful. For example, Figure 5 could be expanded to include a mapping of NMR chemical shift perturbations onto the CXCL8 molecular surface or ribbon representation, allowing the reader to directly compare the locations of CSPs with basic residues and predicted GAG binding sites. Furthermore, in Figure 8, which presents the interaction models of CS4/CS6 with the CXCL8 dimer, it would improve clarity to display both chains of the dimer. This would help determine whether interactions are strictly confined to one monomer or whether residues from the second chain also contribute to CS binding. In addition, captions are missing for all Supplementary Information figures and should be added. The x-axis labels in Figures S1A and S1B should also be checked, as they appear to indicate [CXCL8] rather than the expected [CS dp14].
Response.
We gave considerable effort to determining how best to present the GAG-binding residues, their topology, and potential differences in the binding interfaces. Figure 5 shows all of the basic residues implicated in binding from NMR studies.
We chose this particular pose because it best highlights the relative orientation of the basic residues on the threedimensional scaffold. In Figures 6 and 8, we selected poses that clearly show the basic residue side chains and their interactions with the GAG sulfates and carboxylates. We now also show the interactions of CS4/CS6 are located within the monomer of the dimer as a supplementary figure.
Regarding the x-axis label in Figures S1A and S1B – We thank the reviewer for bringing this to our attention. In the original manuscript, the legends for the supplementary figures were placed beneath the legends for the main figures. We have now placed the legends under each of the supplementary figures. In addition, we have now corrected the x-axis labels in the supplementary figures.
(3) Missing method descriptions. The manuscript does not sufficiently describe the free energy calculations (MM/PBSA or MM/GBSA). A brief explanation of the methodology used to compute binding free energies, as well as the per-residue energy decomposition, should be included in the Methods section. Related to this point, the term “MM/PB(GB)SA” used in line 296 is potentially confusing, as typically only one solvation model - either a PBSA or GBSA - is applied. Please clarify which approach was used.
Response.
We have now added details of the MM-GBSA binding free energy calculations to the Methods section and have also corrected line 296.
(4) Missing data. In lines 212–216, the authors state that titrations with CS oligosaccharides of different lengths (dp8 to dp14) yield similar CSP profiles with varying magnitudes. To substantiate this claim, it would be useful to include CSP data for the shorter CS oligosaccharides, either in the main text or in the Supplementary Information, in addition to the dp14 data already shown.
Response.
We now show CSP bar graphs for the different oligosaccharides as a supplementary figure.
(5) Clarification of conclusions and interpretation. In lines 345–347, the authors conclude that, for the CXCL8 dimer, binding across the dimer interface is not supported by the NMR data and that only intra-monomer binding is consistent with the experimental observations. The basis for this conclusion is not entirely clear from the current presentation. Additional explanation of how the NMR data shown in Figure 2 lead to this inference would be helpful. Related to this point, the rationale for using CS dp8 in the modeling, while the NMR experiments were performed with CS dp14, should be explained more explicitly.
The authors also report differences in residue-specific interactions between CS4 and CS6. For example, K15
appears to form more hydrogen bonds with CS4 than with CS6, whereas R68 interacts more prominently with CS6, and S44 appears to engage selectively with CS6. Similarly, in the dimer, R47 is suggested to contribute primarily to CS4 binding, while H18 and R68 appear more relevant for CS6 binding. It would be valuable for the authors to discuss whether these observations reflect an underlying structural or physicochemical basis for differential specificity between CS4 and CS6, or whether they may arise from limited sampling in the MD simulations. Clarification of whether additional MD replicates may alter or equilibrate these interaction patterns would strengthen the interpretation.
Response.
We now describe in more detail the rationale for excluding binding modes in which the GAG spans the dimer interface. For both CS variants, poses in which the GAGs spanned the dimer interface involved interactions only with the Cterminal helical residues R60, K64, K67, and R68, with no interactions with residues K15, H18, K20, or R47. Such poses are inconsistent with the NMR data and were therefore ruled out.
We carried out NMR titrations for both CS dp8 and dp14 and observed that the binding interactions are similar, with dp14 binding with higher affinity. We now present the corresponding NMR CSP profiles as a new supplementary figure. We used dp8 for the modeling studies because its dimensions sufficiently cover the binding interface. Moreover, computational docking and MD simulations with dp14 become limiting because of challenges associated with simulating long GAG chains.
Regarding residue-specific interactions between CS4 and CS6 — these are relevant questions. As reported in our previous work, the torsion density (probability) is higher for the CS-6S than for the CS-4S [https://www.mdpi.com/2218-273X/12/1/77], which gives rise to difference in inter- and intramolecular H-bonds. This difference in conformational sampling likely contributes to the observed differences in their binding affinities and interaction patterns. We believe that sub-microsecond simulations are sufficient to capture the relevant interaction behavior on the shortest time scale of interest; therefore, extending the MD simulation length is not expected to cause significant changes in the observed trends. To date, we have successfully reported multiple studies using CVLS in combination with MD simulations. The CVLS approach is a unique algorithm designed to identify both specific and non-specific interactions. In this study, we employed both NMR-restrained and non-restrained docking within the CVLS framework and subsequently filtered the best binding poses by applying both affinity and specificity criteria.
Round 2
Reviewer 1 Report
Comments and Suggestions for Authors
None
Author Response
the reviewer had no comments and I presume that the revised manuscript is acceptable to this reviewr
Reviewer 2 Report
Comments and Suggestions for Authors
The authors have satisfactorily addressed my comments, clarified several aspects of the data presentation, and added additional supporting information that facilitates interpretation of the results. They have also compared their findings with those of Pichert et al., who previously investigated the interaction between IL-8 and chondroitin sulfate using NMR spectroscopy and molecular modeling. Although differences in experimental design and computational protocols between the two studies limit the feasibility of a direct, quantitative comparison, I nonetheless find this comparison to be informative. At a qualitative level, it is possible to contrast, e.g., the sets of residues exhibiting pronounced chemical shift perturbations and those contributing most strongly to MM/GBSA binding energies in the two studies. The authors have already provided such a comparative analysis in their response #1 to my initial comments. To ensure that readers can fully benefit from this discussion, I recommend incorporating this comparative analysis into the Discussion section of the manuscript prior to final acceptance.
Author Response
Comments
The authors have satisfactorily addressed my comments, clarified several aspects of the data presentation, and
added additional supporting information that facilitates interpretation of the results. They have also compared
their findings with those of Pichert et al., who previously investigated the interaction between IL-8 and
chondroitin sulfate using NMR spectroscopy and molecular modeling. Although differences in experimental
design and computational protocols between the two studies limit the feasibility of a direct, quantitative
comparison, I nonetheless find this comparison to be informative. At a qualitative level, it is possible to
contrast, e.g., the sets of residues exhibiting pronounced chemical shift perturbations and those contributing
most strongly to MM/GBSA binding energies in the two studies. The authors have already provided such a
comparative analysis in their response #1 to my initial comments. To ensure that readers can fully benefit from
this discussion, I recommend incorporating this comparative analysis into the Discussion section of the
manuscript prior to final acceptance.
Response
We have now incorporated the comparative analysis in the Discussion section.
