O -glycosylation disorders pave the road for understanding the complex human O -glycosylation machinery

Over 100 human Congenital Disorders of Glycosylation (CDG) have been described. Of these, about 30% reside in the O - glycosylation pathway. O -glycosylation disorders are characterized by a high phenotypic variability, reﬂecting the large diversity of O -glycan structures. In contrast to N - glycosylation disorders, a generic biochemical screening test is lacking, which limits the identiﬁcation of novel O -glycosylation disorders. The emergence of next generation sequencing (NGS) and O -glycoproteomics technologies have changed this situation, resulting in signiﬁcant progress to link disease phenotypes with underlying biochemical mechanisms. Here, we review the current knowledge on O -glycosylation disorders, and discuss the biochemical lessons that we can learn on 1) novel glycosyltransferases and metabolic pathways, 2) tissue-speciﬁc O -glycosylation mechanisms, 3) O -glycosylation targets and 4) structure-function relationships. Additionally, we provide an outlook on how genetic disorders, O - glycoproteomics and biochemical methods can be combined to answer fundamental questions regarding O -glycan synthesis, structure and function.


Introduction
Glycosylation, the addition of carbohydrate chains to proteins, is the most common post-translational and co-translational modification. It is initiated by the cytosolic synthesis of activated sugars (with the exception of CMP-N-acetylneuraminic acid) that are subsequently transported to the endoplasmic reticulum (ER) and Golgi apparatus, where glycans are assembled and modified on proteins. Glycosylation affects many aspects of protein function, including protein folding, enzyme activity and cell-to-cell and cell-to-extracellular matrix (ECM) interactions. Therefore, it is not surprising that glycosylation disorders present with a broad range of clinical phenotypes.
Currently, over 100 different Congenital Disorders of Glycosylation (CDG) have been described [1,2], the majority affecting the N-glycosylation pathway. Broad availability of an adequate screening assay for abnormal N-glycosylation, isofocusing of serum transferrin (TIEF), has resulted in the identification of defects in glycosyltransferases, nucleotide sugar transporters and enzymes involved in sugar metabolism, which are all directly linked to glycosylation. In recent years, more complex mechanisms have been identified underlying abnormal N-glycosylation related to Golgi trafficking, homeostasis and vesicular transport [3 ,4,5].
In contrast to N-glycosylation defects, the identification of O-glycosylation disorders is much more challenging. In humans, O-glycans are initiated by seven different monosaccharides that can be further extended to complex Oglycan structures. For mucin O-glycosylation (O-linked Nacetylgalactosamine, O-GalNAc), the most common form of O-glycosylation, over 20 polypeptide GalNAc transferases are known with tissue and substrate-specific activities [6,7]. Isofocusing of ApoC-III was developed to detect defects in mucin type O-glycosylation [8]. Although many of the Golgi homeostasis disorders showed abnormal mucin type O-glycosylation of ApoC-III, only mutations in polypeptide GalNAc transferase 2 (GALNT2) could be detected with this test. So far, the complexity of O-glycan structures renders it impossible to design a single screening test for diagnostics of O-glycosylation disorders.
O-glycans are important for protein structure, folding, stability, recognition, expression, and processing, and they are known to modulate enzyme activity [9,10,11 ,12-15]. Furthermore, highly negatively charged O-mucin glycans can bind water, forming protective layers and preventing bacterial adhesion [16]. The function of an O-glycan can be tissue, protein, and sitespecific, alongside mediating different functions throughout development [17]. That, O-glycans play not only important, but also complex roles, is illustrated by the vast amount of O-glycan enzymes that upon knockout, caused embryonic lethality or tissue-specific phenotypes in mice [18,19 ]. Mice knockout systems have provided invaluable lessons about O-glycan function, for example, the role of O-fucosylation of thrombospondin type 1 repeats (TSRs) by POFUT2 in epithelial organization and expression of signaling factors during gastrulation [20].
In humans, a more complete understanding of the human O-glycosylation machinery can be accomplished by studying genetic defects in O-glycosylation. Identification of an increasing number of genetic O-glycosylation disorders has been facilitated by the emergence of next generation sequencing (NGS) [2]. Furthermore, recent developments in glycopeptide analysis revealed previously unidentified O-glycosylation enzymes and their targets, which can be linked to disease. 3D structural models of human glycosyltransferases are rare, especially since these types of proteins are embedded in the membrane of the ER and Golgi apparatus making crystallization extremely daunting. However, in recent years, some structures have been resolved and modeled. Taken together, new opportunities arise to link findings from genetic disease with fundamental research to increase our understanding of the mechanisms of O-glycosylation. In this review, we illustrate the importance of inherited Oglycosylation disorders to elucidate the structural aspects of the O-glycosylation machinery (Figure 1). Glycosaminoglycan biosynthesis disorders are not discussed and have been described in great detail by others [21]. For elaborate descriptions of O-glycosylation disorder phenotypes, we recommend the reviews of Wopereis et al.

O-glycosylation disorders: current status and screening methods
Most of the currently known O-glycosylation disorders have been identified through genetic techniques. The clinical phenotypes are highly variable, which is linked to the large number of different O-glycan types. O-glycosylation defects have now been identified for each type of Oglycan, and an overview of the known O-glycosylation disorders is provided in Figure 1 and Table 1. Thus far, assays for functional validation of mutations are largely lacking, except for the dystroglycanopathies. This is a group of disorders affecting the O-mannosyl glycan on the a-dystroglycan (aDG) protein that is essential for binding to extracellular matrix components ( Thus, together with NGS, functional tests are highly warranted for a more rapid identification of inherited O-glycosylation disorders, and to increase our understanding of O-glycosylation mechanisms. O-glycomics, the profiling of the complete set of glycans produced by specific cell types, offers potential as a generic functional test. Methods have been developed for the comparative analysis of O-glycans from complex samples [27][28][29][30]. Unfortunately, O-glycomics has thus far not contributed to the functional confirmation of O-glycosylation disorders. This can be explained by the fact that O-glycosylation is highly dependent on the specific attachment site, and O-glycans do not have a general consensus sequence with the exception of O-fucose glycans (C 2 X 3-5 S/TC 3 and WX 5 CX 2/3 S/TCX 2 G; C = conserved cysteines of epidermal growth factor (EGF)-like or TSRs, S/T = serine or threonine, X = any residue) and O-glucose glycans (C 1 XSXPC 2 ). Therefore, it is essential to study O-glycan structures in their protein context. Identification of aber-

Current Opinion in Structural Biology
Characterization of O-glycosylation disorders is indispensable to accomplish a better understanding of the human O-glycosylation mechanisms. Phenotypic heterogeneity of the O-glycosylation disorders reflects the high diversity of O-glycan structures with a high tissue-specificity. Phenotypic characterization and modern omics techniques such as genomics, glycomics, and glycoproteomics complement each other in the each type of discovery in the O-glycosylation field, covering the majority of the O-glycosylation disorder core types.

Discovery of new glycosyltransferases and metabolic pathways
Firstly, genetic defects in O-glycosylation with a characteristic phenotype have aided the discovery of new Oglycosylation gene candidates. For example, NGS has resulted in the identification of new genes causing dystroglycanopathy that is characterized by muscular dystrophy and, in severely affected individuals, eye and brain abnormalities.   Table 1.
cholesterol (HDL-C) levels in human, mice, rats and cynomolgus monkeys. GALNT2 exhibited species-specific glycosylation targets, including PLTP, a regulator of HDL metabolism in plasma [58 ]. PLTP activity was altered by absence of GALNT2 O-GalNAc modifications, explaining the findings in GALNT2 patients. The involvement of additional GALNT2 targets in the disease phenotype remains to be investigated.  [63][64][65]. All five patient mutations that have been described so far reside in the N-terminal tetratricopeptide (TPR) repeats of OGT, which are involved in the substrate recognition and specificity of OGT [66]. OGT patient-derived cells and model cell lines with patient mutations showed normal O-GlcNAcylation [59 ,60 ,62 ]. This homeostasis was suggested to be maintained by a reduced expression of OGA [59 ,60 ] or by temporal dynamics in O-GlcNAcylation kinetics [62 ]. In addition, OGT is involved in proteolytic maturation of HCF1 [14,67], and it has been suggested that the Finally, B3GLCT deficiency leads to Peter's Plus syndrome, a severe disorder characterized by anterior eye chamber defects (Table 1: O-fucose). B3GLCT attaches glucose via a b-1,3 linkage to O-fucose (synthesized by POFUT2) on TSRs of proteins. In search for B3GLCT targets linked to the eye defects, Dubail et al. [77 ] found that ADAMTS9 haploinsufficient mice showed a similar eye phenotype [77 ]. Glycosylation with glucose-b-1,3fucose by POFUT2 and B3GLCT ensures proper secretion of ADAMTS9 during development. Taken together, the identification of new genetic O-glycosylation disorders can provide important insights about the targets and functions of specific O-glycans.

Modeling mutations to study structure-function relations of O-glycosylation proteins
In the last few years, crystal structures have been solved of enzymes related to O-glycosylation disorders, for example of OGT [78], POMK [79], POMGNT1 [80] and ISPD [38]. Known disease-causing mutations can be modeled in 3D structures, helping to understand the function of specific enzymatic domains and with it, underlying disease mechanisms. For example, the crystal structure of ISPD revealed a N-terminal cytidyltransferase domain and a C-terminal domain connected via a linker helix [38]. Surprisingly, the C-terminal domain did not share homology with any known enzyme domains. No missense mutations have been reported in the C-terminal domain, but the c.1114_1116del (p.Val372del) mutation is reported for five patients. The absence of the Val residue leads to relatively mild phenotype (LGMD) compared to larger deletions like a deletion of exon 6-8 or 9-10 (WWS). Taken together, this demonstrates that the C-terminal domain is important for ISPD function, either contributing to the stability of the enzyme, or having a enzymatic function on its own [38], a question that so far remains unanswered. For POMGNT1, one study has reported a correlation between mutations closer to the 5 0 end of the gene with more severe hydrocephalus than mutations near the 3 0 end. However, correlations with enzymatic activity or structure have not been established yet [81]. Taken together, much work remains to elucidate the 3D structure of many O-glycosylation enzymes. However, if such models are accomplished, structure-function relationships can be studied utilizing described O-glycosylation patient mutations. Additionally, this will lead to a better understanding of disease mechanisms, and will hopefully be accompanied by the emergence of new treatment opportunities.

Conclusions
We illustrated that studying the complex phenotypes of Oglycosylation disorders has enabled the elucidation of Oglycosylation proteins, targets, and O-glycan structure and function. Nevertheless, many questions remain to be answered about the O-glycosylation machinery. Although we know in many diseases which O-glycan core structure is affected, for most, their exact attachment site and tissuespecific protein targets remain to be elucidated. In the future, the development of more advanced O-glycopeptide profiling methods is essential to facilitate these discoveries. Ideally, untargeted O-glycoproteomics LC-MS/MS technology will evolve to enable robust high-throughput analysis for the in-depth characterization of intact O-glycopeptides in biological samples. The screening of patient groups with similar clinical presentations or with different genetic O-glycosylation defects (e.g. in different GALNTs) with genomics and O-glycoproteomics will lead to the discovery of glycosylation genes and tissue-specific targets, respectively. As illustrated in this review, comparing the phenotypes of other known disorders to the phenotype of Oglycosylation disorders can hint to the respective targets.
So far, most O-glycosylation defects that have been identified affect the core sugar of O-glycans. In the last few years, NGS has been applied more frequently, and probably will lead to the identification of additional disorders that affect more distal monosaccharides on O-glycan structures. Functional validation of these disorders will require developments in the glycoproteomics field, since large scale in-depth characterization of the exact glycan structure of intact glycopeptides is still challenging. Furthermore, it is important to develop in-silico approaches to identify differential Oglycopeptides and interpret complex glycobiology by novel bioinformatic approaches. Combined analysis of O-glycopeptide data and patient meta data by machine learning is of particular interest to associate protein specific O-glycosylation changes to the physiopathology of O-glycosylation disorders. Taken together, understanding the disease mechanisms of the O-glycosylation disorders will contribute to our understanding of O-glycosylation mechanisms, while vice versa, new mechanistic insights are highly warranted to develop new therapeutic strategies.

Conflict of interest statement
Nothing declared.

19.
Stanley P: What have we learned from glycosyltransferase knockouts in mice? J Mol Biol 2016, 428:3166-3182. This review summarizes the glycosylation genes that are not only necessary for mouse embryonic development, but also for pathway-specific glycosylation genes, of which deletion causes a much milder phenotype. Stanley pinpoints the lessons that we have learned from these mouse models and also discusses general strategies for generating and interpreting the phenotype of mice in relation to human CDG. This was the first paper that demonstrated the presence of Rbo5P moieties on the functionalO-mannosyl glycan of the a-dystroglycan protein using MALDI-TOF mass spectrometry. Major finding of this paper is that FKTN and FKRP, that were long known to cause dystroglycanopathy when mutated, are the Golgi transferases adding Rbo5P to this glycan. 29. Wang C, Zhang P, Jin W, Li L, Qiang S, Zhang Y, Huang L, Wang Z: Quantitative O-glycomics based on improvement of the onepot method for nonreductive O-glycan release and simultaneous stable isotope labeling with 1-(d0/d5)phenyl-3methyl-5-pyrazolone followed by mass spectrometric analysis. J Proteomics 2017, 150:18-30.