PREreview of Siglec-7 and -15 recognize repeated clustered sulfo-sialo O-glycan motifs on select O-glycoproteins

Goerdeler and colleagues investigate how sialoglycan–Siglec interactions achieve selectivity, focusing on Siglec-7 and Siglec-15. The study makes an important contribution by showing that binding in their assays is influenced by both sulfation and the presentation of O-glycans on selected protein carriers. These include MAdCAM1, identified here as a Siglec-binding carrier, and CD43 and PSGL-1, previously reported as Siglec-7 ligands. The clustered saccharide patch hypothesis holds that selectivity arises not only from single glycan structures but from multiple glycan moieties presented together, potentially combined with multivalent engagement by receptors in oligomeric states. The hypothesis has been difficult to test, since many binding studies rely on isolated oligosaccharides that lack the protein context in which glycans are normally displayed. The glycoengineering approach used here allows glycan repetition, sulfation and carrier context to be varied and compared in reporter assays, making protein-carrier context experimentally tractable.

A second contribution is methodological. Precomplexed Siglec-Fc chimeras detected binding across a wide range of reporters, including non-sulfated glycoforms. In reverse assays, the tested non-sulfated forms did not bind cell-displayed Siglec-7, whereas selected sulfated forms did. Sulfated MAdCAM1 also bound primary monocytes expressing Siglec-7, but not appreciably Siglec-7-positive NK cells. These findings highlight that Fc-chimera binding does not necessarily predict binding by cell-displayed Siglec-7.

Overall, the assays provide solid support for a contribution of sulfation and carrier context to binding, particularly for Siglec-7. The authors propose that repeated glycan motifs support multivalent recognition, an interpretation consistent with the mutagenesis and clustering results. The binding geometry and the contribution of receptor oligomerization remain unresolved, as do the precise glycan structures in the binding epitopes, leaving open whether sulfation acts directly or indirectly. In my assessment the experimental binding findings are well supported; most of the points below concern how particular claims are framed and what is reported alongside them rather than the substance of the findings.

Major points

M1. The CHO reverse-assay results indicate a ligand-affinity threshold for detecting binding.

Fig. 5b (p. 10, legend p. 30), Supplementary Fig. 8b (SI p. 13), SI p. 5, and Discussion pp. 14 and 17.

Values in Fig. 5b are obtained by subtracting mean fluorescent intensity (MFI) of CHO^Sig-7R124A from CHO^Sig-7 WT cells, so the comparison depends on the two lines expressing Siglec-7 comparably. Gating on anti-Siglec staining (SI p. 5) addresses part of this, since it restricts both lines to expressing cells, but per-cell density within the gate is not reported and I could not find an expression histogram or MFI for any CHO line in the supplement. Relatedly, the interpretation on p. 17 that these results indicate a ligand affinity threshold rests on receptor density, which is not reported, and on affinity, which is not measured; the same applies to the description on p. 14 of Siglecs in their natural oligomeric state, given the authors’ observation on p. 17 that this state remains to be determined. The authors may wish to consider showing the anti-Siglec staining already acquired for both lines, and stating how the lines were generated, since the supplementary information (SI) describes generation of the HEK293 knockout lines but not the CHO lines. Citing an earlier characterization, if one exists, would equally address this. Presenting the p. 17 threshold as consistent with the data rather than indicated by them would align it with the measurements shown. This would let the subtracted signal be attributed confidently to binding-site-dependent interaction and bring the threshold and oligomeric-state statements into line with what was measured.

M2. Siglec-15 recognizes the selected sulfated O-glycoprotein reporters.

Title and Abstract; Supplementary Fig. 7 (SI p. 13); Figs. 3d and 4b; Introduction (Fig. 1) and Discussion, on precomplexed Fc chimera.

CHO cells displaying Siglec-15 showed no appreciable binding to sulfated MAdCAM1 or fragment1, in contrast to the precomplexed Siglec-15-Fc results, so the positive evidence for Siglec-15 comes from the Fc format alone. The Introduction raises the concern that precomplexing favors multivalent engagement and may detect glycan arrangements that do not occur naturally, and the Discussion notes that the precomplexed chimera, with four sialic acid binding sites, may be more prone to bind high densities of the simple O-glycan epitope. Since Siglec-15 appears in the title and abstract, this difference in support seems worth addressing, particularly as the two explanations offered, low and unstable expression and a binding site closer to the membrane, are not tested, and the first rests on expression data not reported for any CHO line (see M1). Quantifying Siglec-15 surface expression alongside Siglec-7, with a ligand known to bind as a positive control, would distinguish them. If expression is adequate, noting in the abstract and discussion that the cell-level assay was negative would reflect the data; if it proves low or unstable, reporting that would equally resolve the ambiguity.

M3. Punctate MAdCAM1 binding indicates Siglec-7 clustering, and the reporter is taken up by Siglec-7-expressing CHO cells.

Supplementary Fig. 10 shows AF488 reporter and DAPI, so Siglec-7 is not visualized and clustering is inferred from the distribution of the ligand alone. For uptake, the protocol does not include permeabilization, a surface strip, or z-stack acquisition, so the 0–60 min progression establishes that signal accumulates over time and depends on a functional binding site, but not whether it accumulates inside the cell or at the surface; replication is stated as representative images without an n. Co-staining the receptor with a non-blocking anti-Siglec-7 antibody would show whether receptor and ligand colocalize in the puncta, and one orthogonal view or z-stack with reported n would resolve the localization. If these are not feasible, describing punctate binding and accumulating intracellular signal, without asserting oligomerization, would match these claims to the images shown.

M4. Only the strongest ligands identified with precomplexed Siglec-7-Fc bind cell-displayed Siglec-7.

Discussion p. 14; Results p. 12; Fig. 5 and Supplementary Fig. 8; Fig. 6; against the reporter panel of Fig. 3b.

The Discussion concludes that only the best ligands identified with precomplexed Siglec-7-Fc ultimately gave detectable binding to Siglec-7 naturally expressed on cells. This would mean that some ligands reported from Fc-based assays may not bind Siglec-7 as it is presented on cells, so it seems worth noting two aspects of Fig. 6 that affect how far the conclusion can be taken at present. First, on primary cells the binding is attributed to Siglec-7 from anti-Siglec-7 co-staining, while the CHO experiments have the R124A line to compare against; co-staining shows that binding and receptor expression occur on the same cells, not that the binding requires the receptor. The Results reflect this, describing the cells as expressing Siglec-7 and concluding that sulfo-dST MAdCAM1 appears to serve as a ligand for Siglec-7 on monocytes, whereas the Discussion describes these reporters as the first natural ligands demonstrated to interact with Siglecs on cells, to the authors’ knowledge. Second, as presented the comparison rests on four reporters on CHO cells (MAdCAM1 variants, CD43, PSGL-1 and MUC5AC, the last the only negative carrier in this assay) and on MAdCAM1 alone for primary monocytes. Adopting the Results’ wording in the Discussion would keep the claim within what co-staining shows; alternatively, competition with Siglec-7 blocking antibody or excess unlabeled sulfo-dST reporter in the existing PBMC assay would provide evidence that the binding requires Siglec-7. Saying in the Discussion how many reporters the comparison is based on would allow readers to weigh the conclusion appropriately. Where feasible, extending the reverse assay to further reporters scoring positive in Fig. 3b, and testing CD43 or PSGL-1 on primary monocytes, would show whether the pattern holds beyond MAdCAM1. This would show that the monocyte binding depends on Siglec-7 and clarify what has been demonstrated for exogenously versus endogenously expressed Siglec-7.

Minor points

m1. The authors compare Siglec binding across mucin reporters and MAdCAM1 variants (Figs. 3b, 3d and 4b). Reporter-expressing cells are identified using GFP, with anti-FLAG staining also used to verify surface expression of the MAdCAM1 and Glycocarrier constructs, but it is unclear whether reporter abundance was comparable across constructs or accounted for in the comparisons. It would help interpretation if the authors showed or summarized the relevant expression data and clarified whether binding was assessed in expression-matched populations. Where expression differed, noting this would help readers interpret the reduced binding observed for some MAdCAM1 mutants.

m2. The concluding sentence of the abstract states that Siglec-7 on immune cells employs multivalent interactions with repeated motifs of clustered sulfo-sialyl O-glycans to drive selectivity in ligand interactions and Siglec-induced immunosuppression. Immunosuppression is not assayed, and the Discussion notes this twice. The reference to immune cells is also somewhat broader than the demonstration, since binding was shown on monocytes while very few Siglec-7⁺ NK cells bound at comparable receptor expression. The authors could consider describing the work as identifying candidate ligands rather than as indicating immunosuppression, noting that functional consequences remain to be tested, and specifying monocytes, which would make the limitations they already note visible to readers who see only the abstract.

m3. Fig. 4 is well represented graphically, but the reader would benefit from more guidance on the results, and splitting the panels would make it easier to follow what each represents in the text. The radar format also limits readability, since exact values cannot be read. Replotting as grouped bars with individual replicate points, ordered to match the sequence schematic, would make the dataset directly interpretable, which is useful here because Fig. 4 carries the comparison between cluster number and spacing.

I reviewed this preprint as part of the Transparent and Accountable Peer Review training program. This report was written with the support of my research mentor.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.