Studying the cell surface glyco-code

Most human secreted and cell surface proteins are post-translationally modified by Ser/Thr(O)-linked glycosylation with N-acetylgalactosamine (O-GalNAc). On the cell surface, these glycoproteins play a central role in modulating signal transduction, cell-cell interactions, and biophysical properties of the plasma membrane.

Transfer of GalNAc to Ser/Thr sidechains is accomplished in the Golgi by a 20-member family of isozymes, the GalNAc transferases. The protein substrate specificity of each isozyme is poorly studied, due to a dearth of reagents able to selectively detect the products of a particular GalNAc transferase, or to distinguish between O-GalNAc modifications and other glycan subtypes. Overcoming this knowledge deficit is important, as the O-GalNAc glycosylation signature is differentially regulated indifferent cell types or activation stages, and diseases, including cancer, are associated with errors in the process.

Recently, so-called “bump-and-hole” tactics have been used to deconvolute glycosylation pathways. A “hole” is engineered into the active site of an individual enzyme to accommodate a modified carbohydrate donor carrying a chemical “bump”. By virtue of a clickable tag on the bumped donor, any proteins it glycosylates can be detected by bioconjugation of the tag to a reporter moiety, allowing detection by such techniques as advanced microscopy or mass spectrometry (glyco)proteomics. As a further bonus, such bioorthogonal enzyme-substrate pairs can act as reporter systems for individual enzymes in vivo, with minimal perturbation of the glycosylation network in which the wild-type enzyme functions.

As a postdoc in Stanford with Carolyn Bertozzi, one of the world’s leading glycobiologists, Ben Schumann spearheaded the first application of a glycosyltransferase bump-and-hole engineering approach in living cells, making it possible to contemplate an in vivo dissection of the cell surface glyco-code. He has extended this work in the first publication from his new laboratory at the Crick, by developing a bioorthogonal reagent specific for O-GalNAc glycans. This work makes the Crick one of only a few labs in the world with both the experimental capacity and expertise to site-selectively analyse the products of individual glycosyltransferases, crucial for a full understanding of glycobiology.

The substrate for GalNAc transferases is UDP-GalNAc, which is transferred onto a recipient protein, and then elaborated in a series of downstream glycosylations. Whilst at Stanford, Schumann and his colleagues mutated the active site gatekeeper residues in GalNAc transferase 2 (GalNAc-T2) to open up a hole that could be filled by chemically modified versions of the UDP-GalNAc substrate carrying a “bump” — an alkyne side chain (Choi et al, 2019). A co-crystal structure of the mutant, BH-T2, with its bumped substrate showed that the 3-D structure was virtually identical to wild type; the enzyme was therefore functional, but was blinded to its normal substrate, instead gaining reactivity for chemically modified GalNAc analogues. Mutation of GalNAc transferase 1 (GalNAc-T1) yielded similar in vitro results, and in a peptide glycosylation assay, the two mutant enzymes, BH-T1 and BH-T2, retained the specificity of their wild type counterparts (Schumann et al, 2020).

Having generated bioorthogonal enzyme-substrate pairs that worked in vitro, the group sought to move the system into a living cell, a substantial technical challenge; the cell must synthesise the bumped substrate from an exogenously delivered precursor, and this must be delivered into the Golgi compartment, where the correctly localised and folded mutant enzyme has to be expressed.

To generate substrate, bumped GalNAc analogues were fed to cells as caged GalNAc-1-phosphate precursors and turned into UDP-GalNAc analogues by a mutated version of the pyrophosphorylase AGX1. Chromatographic analysis showed that bumped UDP-GalNAc was being made, but a small proportion of it was epimerised into UDP-GlcNAc. This could be prevented by CRISPR-mediated knockout of UDP-galactose 4-epimerase (GALE).

With the GalNAc-T bump-and-hole system and a method for substrate delivery in hand, the group set out to probe isozyme-specific glycosylation in the living cell. Human cells were transfected with mutant AGX1 and either BH-T1 or BH-T2 (the mutant versions of GalNAc-T1 or GalNAc-T2) and fed bumped GalNAc-1-phosphate precursor. Cell surface glycosylation was then monitored by clicking a fluorophore onto the bumped O-GalNAc, following which the cells were lysed and analysed. Pleasingly, using the orthogonal pairs, it was possible for the first time to delineate distinct glycosylation signatures for GalNAc-T1 or GalNAc-T2, demonstrating the utility and importance of the approach (Schumann et al, 2020).

Upon moving to a joint Imperial College-Crick appointment in late 2018, Schumann and his new group focussed on the other major experimental handicap in studying O-GalNAc glcosylation: the lack of a completely O-GalNAc-specific reporter molecule. As discussed above, unless GALE is removed from cells, a proportion of UDP-GalNAc substrate is epimerised to UDP-GlcNAc, leading to background N-glycan labelling. For glycoproteomics experiments, it is possible to account for this by treating lysates with the amidase PNGase F, which cleaves GlcNAc from glycoproteins, but this approach is not suitable for imaging and other such experiments. GALE knockout, while effective, causes excessive disruption of the cellular glycosylation machinery, so is not a good solution. Accordingly, Schumann’s lab, still collaborating with his former colleagues, sought to develop an O-GalNAc reporter tool that would be resistant to epimerisation.

Comparison of the crystal structures of the active site of GalNAc transferases with that of the epimerase GALE yielded a potential way forward. In GalNAc transferases, the acetamide group of the GalNAc donor molecule is bound in a globular pocket with some three-dimensional freedom in it, in contrast to the narrow, tunnel-like active site of GALE. Schumann reasoned that if the side chain of the UDP-GalNAc molecule were bulkier, this might be enough to prevent it being inserted into the GALE active site, blocking epimerisation.

A panel of UDP-GalNAc analogues with varying side chains had already been made in the Bertozzi lab (Choi et al, 2019), so the group fed them into an in vitro glycosylation experiment, to which they added purified human GALE. For all the analogues with linear side chains, UDPGlcNAc, and hence epimerisation, was detectable, but for the bulky, branched side chain analogues, there was no epimerisation. One of the structurally simplest of the branched analogues, UDPGalNAzMe 5, was shown to be a substrate for all GalNAc transferases tested, and was therefore chosen for further work in living cells.

Using their previous strategy to deliver substrate, the group transfected cells with mutant AGX1, and then fed them caged, membrane-permissive GalNAzMe-1- phosphate 11 precursor to produce UDPGalNAzMe 5. There was no detectable epimerisation, and GalNAzMe specifically entered the O-GalNAc glycosylation pathway, confirming that the new analogue could function as a highly specific O-GalNAc labelling reagent.

As proof of GalNAzMe’s utility as a reporter for the biology of O-GalNAc glycosylation, K-562 cells fed the analogue were assessed by chemical mass spectrometry glycoproteomics, in collaboration with the Bertozzi lab and the Proteomics STP at the Crick. GalNAzMe could substitute for peptide-proximal O-GalNAc residues, and could also be extended by the downstream glycosylation machinery. When bioconjugated to a fluorescent reporter, GalNAzMe could also be used in superresolution microscopy to visualise the mucin-covered tubules on the surface of K-562 cells.

Even more ambitiously, GalNAzMe-1-phosphate 11 was used as a reporter in a fluorescence-based genome-wide CRISPR knockout screen for genes involved in glycan biosynthesis. GalNAzMe labelling was sensitive to knockout of genes mediating cell surface O-glycan presentation such as mucins, but crucially, not those involved in N-glycan synthesis, unlike control epimerisable GalNAc probes, which were affected by both. The results validate GalNAzMe as a potent reporter tool for future genetic profiling experiments.

As a final flourish, Schumann collaborated with his Crick colleague Vivian Li to image intestinal organoids, important tools for understanding fundamental principles of both bowel cancer and normal gut development and homeostasis. Production of O-GalNAc glycans in organoids is often probed with backbone-directed antibodies or carbohydrate-binding lectins, but GalNAzMe offers a detection tool independent of both protein backbone and the complexities of glycan elaboration, reporting instead on the first link in the chain — the invariant peptide-proximal GalNAc moiety.