I. INTRODUCTION
Descriptions of phenotypic abnormalities in ontologies have
been useful for computational comparisons of
genotypephenotype associations across species (for example, via the
Human Phenotype Ontology, HPO [
1,2
] and Mammalian
Phenotype Ontology, MP[
3
]). Such ontologies have
supported the creation of new tools that help gene variant
prioritization based on phenotypes [
2
] (e.g. Exomiser [
4
]). We
have developed a prototype of a molecular glycophenotype
ontology (MGPO) that captures abnormality of glycosylation at
a molecular level (glycophenotypes) that may occur in human
genetic diseases [
5
]. Molecular phenotyping can enhance
precision medicine, for instance, fucosidosis ranks on Monarch
Phenotypic Similarity tools [6] were increased with
glycophenotypes terms [
7
]. We are currently using MGPO to
annotate glycophenotypes to hundreds of genetic
glycanopathies (e.g. diseases related to N-glycosylation
pathways). Many glycanopathies and glycosylation pathways
have been discovered using biological models (e.g. yeast [
8
]).
Hence, model organisms can inform human glycanopathies. To
improve cross-species comparison of genetic diseases,
enhancements to the current MGPO model are needed. In
Melissa Haendel*1
addition, many acquired diseases can show glycophenotypes.
For instance, hyposialylated transferrin is a glycophenotype for
chronic alcohol exposure [
9
]. Annotating glycophenotypes
from acquired diseases in MGPO in addition to anatomical
phenotypes could help enhance the disease description.
II. GLYCOBIOLOGY OF MODEL ORGANISMS CAN BE
INFORMATIVE FOR HUMAN DISEASES
Only 128 genes are known to be involved in human
glycosylation disorders (Congenital Disorder of Glycosylation,
CDG) [
10
]. While 21% of them are related to non-glycan
processes such as transport of small molecules and ions in the
Golgi/ER, the remaining 79% is related to glycan synthesis or
degradation. These account for only about 30% of known
human glycosyltransferases reported in the database for
Carbohydrate-Active Enzymes Database (CAZY,
http://www.cazy.org/) [
11
]. Mutations in ortholog of
glycosyltransferases in model organisms can be useful for
disease comparison, because in patients with undiagnosed
diseases, only the known clinical variants are reported from
genome/exome analyses, and thus, other relevant mutations can
be missed that may have been identified in model organisms.
Exomiser [
4
] can help variant prioritization using cross-species
genotypes-phenotypes association. In order to expand disease
comparison between species, curation and representation of
glycophenotype related to genes can help. For instance, Baker’s
yeast (Saccharomyces cerevisiae) has been used as a model to
understand N-glycan synthesis in eukaryotes and led to
understanding of many CDGs [
8
]. In yeast and fly models, a
defect in O-mannosylation was discovered and led to
understanding of many dystroglycanopathies (e.g. POMT1) [
8
].
Mouse knockout models can be useful as many knock-out
genes are associated with glycophenotypes and are available in
public databases like the Consortium for Functional Glycomics
(CFG, http://www.functionalglycomics.org/) [
12
]. However,
these data from CFG are not annotated with MGPO terms nor
represented in HPO yet. Finally, knock-out of glycan related
genes in zebrafish [
13
] have showed many anatomical
abnormalities that could be informative for comparison of many
human glycanopathies.
III. MANY ACQUIRED DISEASES HAVE GLYCOPHENOTYPES BUT
ARE NOT REPRESENTED YET IN HPO
Glycophenotypes can provide more precise disease
descriptions for acquired diseases and genetic diseases. For
instance, the phenotype ‘hepatic steatosis’ (HP:0001397) is
common in 188 diseases in the cross-species
phenotypegenotype data integration platform, Monarch Initiative [6]
(www.monarchinitiative.org) including non-liver fatty disease
(NLFD, MONDO:0021105) and DDOST-CDG
(MONDO:0013789). NLFD and DDOST-CDG share the
phenotype ‘Abnormal Protein N-Linked Glycosylation’
(HP:0012347). The main difference is the type of abnormal
glycoprotein, hypoglycosylation for DDOST-CDG and
hyperglycosylation for NLFD but these
hyper/hypoglycosylation are not yet integrated in HPO (Fig.1).
Thus capturing glycophenotypes for acquired disease can help
differentiate these diseases. Many cancers also show specific
glycosylation changes that can be measured in glycan binding
assays [
14
]. In the HPO, 5 types of cancers are referenced:
Cervix cancer (HP:00300790), Colon cancer (HP:0003003),
Merkel cell skin cancer (HP:0030447), Prostate cancer
(HP:0012125), Stomach cancer (HP:0012126). All of them
have glycophenotypes that are not represented yet in MGPO :
GalNAcα-1-3Gal (cervix cancer) [
15
], reduced expression of
core 3 and 4 glycans and increased Tn/sTn antigen (Colon
cancer)[
16
], increased sialylation and fucosylation (prostate
cancer) [
17
], decreased levels of high-mannose-type glycans
(Stomach cancer)[
18
]. On the other side of the spectrum of
glycan and cancer resistance, some rare animal models like
naked mole rats can help find molecular strategy to fight cancer
(as they present with high hyaluronan mass and cancer
resistance) [
19
].
IV. CONCLUSION
We believe that an integration of glycomics data in HPO and
MP will allow for a better understanding the interconnectivity
of molecules, and thus, better diseases comparison and
diagnoses. Indeed, to this day, 25% of patients have been
diagnosed with traditional methods like exome analysis
(Undiagnosed Diseases Network, personal communication);
adding molecular phenotyping could help those who remain
undiagnosed. Finally, molecular phenotyping with MGPO
could provide better insights toward possible treatments, e.g.
dietary supplementation of glycans [
20
].