Polypeptide N-acetylgalactosaminyltransferase-Associated Phenotypes in Mammals

Mucin-type O-glycosylation involves the attachment of glycans to an initial O-linked N-acetylgalactosamine (GalNAc) on serine and threonine residues on proteins. This process in mammals is initiated and regulated by a large family of 20 UDP-GalNAc: polypeptide N-acetylgalactosaminyltransferases (GalNAc-Ts) (EC 2.4.1.41). The enzymes are encoded by a large gene family (GALNTs). Two of these genes, GALNT2 and GALNT3, are known as monogenic autosomal recessive inherited disease genes with well characterized phenotypes, whereas a broad spectrum of phenotypes is associated with the remaining 18 genes. Until recently, the overlapping functionality of the 20 members of the enzyme family has hindered characterizing the specific biological roles of individual enzymes. However, recent evidence suggests that these enzymes do not have full functional redundancy and may serve specific purposes that are found in the different phenotypes described. Here, we summarize the current knowledge of GALNT and associated phenotypes.


Introduction
Initiation of mucin-type O-glycosylation is controlled by a large family of UDP-GalNAc: polypeptide N-acetylgalactosaminyltransferases (EC 2.4.1.41; also referred to as GalNAc-Ts, ppGalNAc-Ts, and ppGaNTases) in animals. These enzymes catalyze formation of GalNAcα1-serine (Ser)/threonine (Thr) linkages in glycoproteins. In humans, 20 isoforms of GalNAc-Ts have been identified and several reviews have described these enzymes in detail [1][2][3]. The importance of each isoform is unclear because phenotypes lacking one or two isoforms have not been observed in experimental animals, except in Drosophila [4][5][6][7][8]. However, more recent studies have described GALNT mutation-related phenotypes in mammals. The purpose of this review is to summarize these phenotypes and to correct unintentional errors and nomenclature in the literature. tissues suggests that disruption of the gene would severely affect organisms. However, disruption of murine Galnt1 is not fatal and does not cause infertility, although Galnt1 deficiency does have moderate lethality and one-fourth of homozygous null mice, but not heterozygous mice, die in utero beyond E12.5 or before the age of one month for unknown reasons [19]. The moderate lethality observed in Galnt1 null mice might indicate that other GalNAc-Ts could partially compensate for GalNAc-T1.
In 2007, utilizing the Cre-loxP recombination system, Tenno et al. showed altered innate and adaptive immune cell trafficking in Galnt1-inactivated mice. This correlates with decreased expression of E-and P-selectin ligands on neutrophils and decreased expression of L-selectin ligands on lymph node high endothelial venules (HEVs). Increased apoptosis of B cells in the germinal center, resulting in impaired IgG production, also occurs in GalNAc-T1-deficient mice. A bleeding disorder was also observed in the mice, due to reduction of plasma levels of coagulation factors V, VII, VIII, IX, X and XII, even though platelet homeostasis was unaffected [20]. Immunological impairment of Galnt1-deficient mice has also been reported [21]. P-selectin-dependent and E-selectin-mediated leukocyte rolling were significantly impaired and L-selectin-dependent leukocyte rolling was completely abolished in Galnt1-deficient mice. P-selectin glycoprotein ligand-1 (PSGL-1) has been suggested to be a target protein for GalNAc-T1 O-glycosylation.
Loss of Galnt1 affects the early stages of murine organogenesis by disrupting secretion of components, especially laminin and collagen IV, of the basement membrane, which is a specialized extracellular matrix (ECM) that is important for mammalian development. Reduced cell proliferation and embryonic submandibular gland (SMG) growth, resulting in decreased integrin and FGF signaling, occur in Galnt1-deficient mice. Interestingly, induction of endoplasmic reticulum (ER) stress and the unfolded protein response (UPR), distinct from those observed in N-linked glycan deficiency, were observed in SMGs of these mice [22]. The same group also found impaired cardiac function in Galnt1-deficient adult mice. In this case, aberrant valve formation caused by increased cell proliferation of developing hearts resulted in aortic and pulmonary valve stenosis and regurgitation, left ventricular hypertrophy, cardiac dilation and valve thickening. In the developing valve tissues of Galnt1-deficient mice, transforming growth factor (TGF)-beta associated bone morphogenetic protein (BMP) signaling and mitogen-activated protein kinase (MAPK) signaling were increased. As an antagonist to BMP signaling, epidermal growth factor receptor (EGFR) signaling was decreased in these tissues. These alterations of signaling pathways resulted in increased cell proliferation and larger heart valves in the early stages of development. Expression of ADAMTS (a disintegrin and metalloproteinase with thrombospondin motifs) proteases, ADAMTS1 and ADAMTS5, which have important roles in valvulogenesis, was reduced at the protein level in developing valves of Galnt1-deficient mice, resulting in reduction of proteoglycan versican processing. For ECM proteins involved in cardiac development, reduced expression of cartilage link protein 1 (Crtl1) and increased accumulation of collagen I and fibronectin occurred in the mice. Unlike in embryonic SMGs of Galnt1-deficient mice, there was no ER stress in the developing heart tissue [19]. These observations are interesting, but the target proteins and underlying mechanisms in the GALNT1-related phenotypes are still unclear.

GALNT2-Related Phenotypes
The second GalNAc transferase, GalNAc-T2, was identified and purified from human placenta [15]. Based on the amino acid sequence, polymerase chain reaction (PCR) primers were generated and GALNT2 was obtained from a cDNA library of the gastric tumor cell line MKN45. Galnt2 and its protein are ubiquitously expressed in tissues and cell types at high levels in mice [18]. Recent genome-wide association studies (GWASs) have revealed associations of single nucleotide polymorphisms (SNPs) of GALNT2 with levels of plasma high-density lipoprotein cholesterol (HDL-C) and triglycerides (TG) in humans [23][24][25][26]. Teslovich et al. further showed that liver-specific overexpression of Galnt2 using an adenoassociated virus (AAV) vector expression system reduced HDL-C levels, while shRNA-mediated knockdown of hepatic Galnt2 raised HDL-C in mice [26]. These results are supported by a study showing that a GalNAc-T2 D314A (p.[Asp314Ala]) heterozygous mutation raised plasma HDL-C levels in humans. This may be due to a loss-of-function (LOF) in GalNAc-T2, caused by the mutation reducing O-glycosylation and activity of Apo-CIII, a protein that inhibits lipoprotein lipase (LPL), with a consequent increase in HDL-C [27].
In contrast, other GWAS and in vitro studies have shown that a modest increase in GALNT2 expression increases HDL-C levels [28]. Furthermore, two cases of human homozygous LOF mutations in hepatic GALNT2 with low levels of HDL-C and moderate reduction of TG in plasma were recently reported while searching for causes of heritable neurological traits and intellectual disability [29]. These mutations are a T>C in exon 3 of GALNT2 (p.[Phe104Ser]) and a nonsense mutation in exon 9 (p.[Gln289*]). Both are within a catalytic domain or between the catalytic domains of GalNAc-T2, leading to expression of a non-functional protein [29].
An explanation of the contrasting results among these studies was suggested by Khetarpal et al; overdose of AAV and AAV shRNA used in the earlier study caused GalNAc-T2-independent reduction and shRNA sequence-specific elevation in HDL-C levels, respectively [29]. For the mutated GalNAc-T2 D314A enzyme, normal activity of the protein was confirmed [30] and a GalNAc-T2 D314A homozygote had low HDL-C levels in humans, indicating that the higher HDL-C in heterozygotes in the earlier study was GALNT2-independent [29].
The major target proteins involved in these human phenotypes are angiopoietin-like protein 3 (ANGPTL3), apolipoprotein C-III (ApoC-III) and phospholipid transfer protein (PLTP) [29,31,32]. GalNAc-T2 selectively glycosylates Thr226 of ANGPTL3 adjacent to the furin-like proprotein convertase processing site, and O-glycosylation of this site blocks protein cleavage in vitro. Cleavage of ANGPTL3 at this site leads to activation of the protein [33] as an inhibitor of LPL [34] and endothelial lipase (EL) [35]; therefore, LOF of GalNAc-T2 will increase ANGPTL3 activity, lower LPL activity and reduce HDL-C levels. This has been shown clearly in cultured cells and in mice [36]. GalNAc-T2 glycosylates Thr74 of ApoC-III, and if O-glycosylation stabilizes the protein, elevated Apo-CIII will lower LPL and increase plasma TG, although species-specific regulation of plasma TG levels has also been reported [29,[36][37][38]. The effect of O-glycosylation on Thr74 of ApoC-III clearly requires further study, because the absence of O-glycosylation at Thr74 did not affect Apo-CIII secretion or binding affinity for lipoprotein in vitro [39], and resulted in normal TG, HDL and ApoC-III levels in humans [40].
Plasma PLTP activity is negatively associated with HDL-C levels and positively associated with TG levels [41]. PLTP activities are partially lowered in rodents and in humans, even though protein expression in plasma was not affected by GALNT2 deficiency [29]. This partial inhibition of activity can be explained by the ability of another major GalNAc-T expressed in liver-GalNAc-T1-to glycosylate the protein, although not as efficiently as GalNAc-T2 [29,31].
There are many players among lipases and in lipoprotein metabolism, and molecular regulation of lipoproteins is still not fully understood [42]. However, it is clear that GALNT2 deficiency lowers HDL-C levels and affects lipoprotein metabolism in rodents, nonhuman primates and humans. Novel congenital disorders with global development delay, intellectual disability and decreased HDL-C caused by GALNT2 deficiency in humans have recently been reported [43], and these multiple phenotypes also occur in Galnt2-deficient rodent models. This indicates that LOF of GalNAc-T2 affects the functions of protein substrates in different tissues. A summary of published GALNT2 mutations and associated phenotypes is shown in Table 1.  (1) The phenotype for GALNT2-congenital disorder of glycosylation (CDGIIt) refers to OMIM #61885 and that for affecting HDL-C levels refers to sources [27,29].

GALNT3-Related Phenotypes
Human GALNT3 was identified using the common sequences of GALNT1 and GALNT2 [44], and murine Galnt3 was identified from mouse testis [45]. In contrast to GALNT1 and GALNT2, expression of GALNT3 is highly regulated. Human GALNT3 is mainly found in the pancreas and testis, with lower expression in the kidney, prostate and intestine. Murine Galnt3 is highly expressed in the testis and moderately in other organs, including the sublingual gland, uterus, cervix, stomach, colon, kidney, submandibular gland and parotid gland [44,45].
Topaz et al. first reported GALNT3 mutation-associated hyperphosphatemic familial tumoral calcinosis (HFTC) cases in humans [46], and phenotype-related GALNT3 mutations have subsequently been reported in 62 cases from 39 families. Some of these cases were also summarized by Sun et al. [47] and Rafaelsen et al. [48]. Hyperostosis-hyperphosphatemia syndrome (HHS) is another clinical feature of GALNT3 deficiency; however, as concluded by Ichikawa et al. [49], we consider HFTC and HHS to be a continuous spectrum of the same disease and we describe both conditions as HFTC in this review. The mechanism underlying HFTC caused by GALNT3 mutation is a lack of O-glycosylation on Thr178 in a furin-like proprotein convertase processing site of phosphaturic factor FGF23; cleavage leads to inactivation of the factor [50].
The 62 reported HFTC cases in 39 families are summarized in Table 2. The disease is caused by both compound heterozygous and homozygous mutations of GALNT3. As shown in Table 2, there is no apparent difference in ethnicity, with 11 different ethnicities and a group of unknown ethnicities. A total of 36 different GALNT3 mutations are represented in the families, of which 16 are missense mutations, 6 are stop mutations, 9 are insertions or deletions causing a frameshift and 5 are splice site mutations ( Table 3). The mutations are in all exons, except for exon 1 and exon 11. Exon 1 is a non-coding region of GALNT3 (NM_004482.4) and exon 11 contains the γ region of the lectin domain ( Figure 1). In some previous studies, exon 1 has been skipped in numbering of exons and introns of GALNT3, but in this review, exon 1 is included in this numbering. The HFTC-related GALNT3 mutations are found in all domains, including the initiating ATG start codon, stem region, catalytic domain, linker region and lectin domain of the enzyme. Two of six cysteine residues involved in disulfide bridges in lectin domains of GalNAc-T3 are affected by missense mutations (p.     Galnt3-deficient or N-ethyl-N-nitrosourea (ENU)-induced GalNAc-T3 W589R (p. [Trp589Arg])mutated mice show moderate phenotypes of hyperphosphatemic tumoral carcinosis with decreased alkaline phosphatase activity and intact FGF23 levels [77,78]. Interestingly, Galnt3-deficient mice do not develop ectopic calcification, which is a phenotype in human HFTC, but GalNAc-T3 W589R mice have this phenotype. Ichikawa et al. further showed that Galnt3-deficient male mice have growth retardation, infertility and increased bone mineral density, compared with wild-type mice [77]. These findings also occur in GalNAc-T3 W589R mice [78], indicating that GalNAc-T3 activity is important for male infertility, consistent with the high GALNT3 expression in testis.
Mandel et al. showed strong expression of GalNAc-T3, but not GalNAc-T1, in human spermatozoa [79]. The same group also showed GalNAc-T3 expression in spermatocytes and spermatids, where T antigen (Gal-GalNAc-Ser/Thr) is expressed. Importantly, GalNAc-T3 was the only GalNAc-T isoform expressed in ejaculated spermatozoa (although GalNAc-T17 may also be present; see below) and had ring-shaped equatorial expression, indicating that this isoform has a specific role in germ cells [80]. The fraction of spermatozoa with equatorial expression of GalNAc-T3 was significantly lower in men with oligoteratoasthenozoospermia. Therefore, GalNAc-T3 expression seems to be related to the quality of spermatozoa, and GalNAc-T3 deficiency may lead to impaired O-glycosylation of proteins and abnormal maturation and function of spermatozoa [81]. One candidate protein is zonadhesin, a cell membrane adhesion molecule that plays a role in sperm-egg binding and has a mucin-like tandem repeat region that is heavily O-glycosylated [82,83]. Miyazaki et al. [84] showed that spermatozoa in Galnt3-deficient mice are rare and immotile, and have deformed round heads. A glycoprotein localized at the equatorial segment of the acrosome, equatorin, was identified as another candidate target protein specifically glycosylated by GalNAc-T3, possibly at Thr138. This site of O-glycosylation on equatorin is important for sperm-egg interaction in vitro and in vivo [85,86]. GALNT3 deficiencyrelated HFTC cases provided the first evidence for non-redundant functions of GalNAc-Ts in mammals. It is still unclear if HFTC males have reduced fertility, but O-glycosylation of sperm proteins by GalNAc-T3 seems to be important for sperm-egg interaction.

GALNT4-Related Phenotypes
Murine Galnt4 [16] is highly expressed in the sublingual gland, stomach and colon, and moderately expressed in the lung, small intestine, cervix and uterus. Human GALNT4 [87] is highly expressed in the liver, small intestine, stomach, pancreas, thyroid, spleen, lymph node and bone marrow. GalNAc-T4 V506I (p.[Val506Ile]) carriers were shown to be associated with reduced risk of acute coronary syndrome (ACS), although the association was not significant in the haplotypic test [88].

GALNT5-Related Phenotypes
Rat Galnt5 is highly expressed in the sublingual gland and colon, and moderately in the stomach and small intestine [89]. The partial human GalNAc-T5 amino acid sequence was published in 1999 [90]. GALNT5-related phenotypes have not been reported in mammals so far.

GALNT6-Related Phenotypes
Human GALNT6 was identified as a close homolog of GALNT3 [83]. Human GALNT6 is expressed in the placenta and trachea, with weak signals in the brain and pancreas. Fibronectin is a good substrate for O-glycosylation by human GalNAc-T3 and GalNAc-T6, but not by GalNAc-T1 and GalNAc-T2, in vitro. However, GalNAc-T3 is not expressed in fibroblasts, in contrast to GalNAc-T6. These findings indicate that fibronectin is specifically glycosylated by GalNAc-T6 in vivo. Unlike GalNAc-T3, GalNAc-T6 is not expressed in human spermatozoa, whereas expression of GALNT6, but not GALNT3, occurs in the brain [83]. Related to this observation, Akasaka-Manya et al. showed that expression of several GALNTs, including GALNT6, is altered in sporadic Alzheimer's disease progression in human brain. More prominent amyloid precursor protein (APP) O-glycosylation occurred through GalNAc-T6 compared to GalNAc-T1 or GalNAc-T4, and β-amyloid (Aβ1-40 and Aβ1-42) generation was reduced by overexpression of GalNAc-T6 without affecting the activities of secretases in vivo. This suggests that O-glycosylation of APP by GalNAc-T6 inhibits APP cleavage, and thereby decreases Aβ production [91]. Furthermore, downregulation of both GalNAc-T3 and GalNAc-T6 in the ectopic endometrium contributes to the development of endometriosis [92].

GALNT7-Related Phenotypes
Rat Galnt7 (first reported as Galnt6) was identified by Ten Hagen et al. and human GALNT7 was published by Bennet et al. in 1999 [90,93]. Rat and murine Galnt7 are expressed in the sublingual gland, stomach, small intestine and colon, with trace amounts in the ovary, cervix and uterus. Human GALNT7 is expressed ubiquitously, including in the stomach, thyroid, spinal cord, lymph node, trachea and adrenal gland, but not in bone marrow. This enzyme has a preference for GalNAc-glycosylated substrates and functions as a follow-up enzyme [90]. Genome-wide linkage analysis revealed that human GALNT7 is positively associated with schizophrenia, although the association was not statistically significant in quantitative PCR analyses [94]. A GWAS showed that neuritic plaques, a core neuropathologic feature of Alzheimer's disease formed by β-amyloid deposits, are positively associated with GALNT7 [95]. Together with GalNAc-T6, GalNAc-T7 may control β-amyloid generation in Alzheimer's disease.

GALNT8-Related Phenotypes
Human GALNT8 was cloned as an autosomal dominant hypophosphatemic rickets (ADHR) candidate gene from human fetal brain [96]. The tissue distribution of GALNT8 is high in the heart, skeletal muscle, kidney and liver; moderate in the placenta, small intestine, leukocyte and lung; and weak in the brain, colon, thymus and spleen. GALNT8 is highly polymorphic, but no substitutions are ADHR-related. The enzymatic activity of GalNAc-T8 was unclear until several studies showed that it might be a non-functional enzyme in vitro, despite weak expression of GALNT8 in the fetal brain, testis, colon and small intestine found by quantitative real-time PCR, in contrast to a previous study [97,98].

GALNT9-Related Phenotypes
Human GALNT9 shows brain-specific expression, especially in the cerebellum, frontal lobe, temporal lobe and putamen; weak expression in the cerebral cortex; and no expression in the medulla, occipital pole and spinal cord [99]. This distribution is consistent with quantitative real-time PCR results reported by Raman et al. [97]. Its enzymatic activity was shown by Zhang et al. [100] to be similar to GalNAc-T1 in terms of substrate specificities. GWAS in chickens showed that GALNT9 in the liver differed significantly between fat and lean broilers, and correlated with abdominal fat traits, but it is unclear if this is also true in mammals [101].

GALNT10-Related Phenotypes
Rat Galnt10 was identified (but referred to as Galnt9 at the time) by Ten Hagen et al., together with Galnt5 from rat sublingual gland [102]. The highest levels of rat Galnt10 occur in the sublingual gland, testis, small intestine, colon and ovary; with lower levels in the heart, brain, spleen, lung, stomach, cervix and uterus. GalNAc-T10 resembles GalNAc-T7, in that it also serves as a follow-up enzyme and does not utilize non-glycosylated peptides as substrates. Murine Galnt10 is predominantly expressed in several hypothalamic, thalamic and amygdalar nuclei in the brain. The broad expression of Galnt10 in mouse brain might indicate that the enzyme has a role in the central nervous system (CNS) [103]. Human GALNT10 was cloned as a homolog of human GALNT7. Quantitative real-time PCR revealed that human GALNT10 is ubiquitously expressed, with the highest levels in the small intestine, and intermediate levels in the stomach, pancreas, ovary, thyroid gland and spleen. Human GalNAc-T10 displays strong activity for glycosylated peptides, but negligible catalytic activity toward non-glycosylated peptides [104,105]. Interestingly, a further study showed that the catalytic domain of GalNAc-T10 is involved in site selection on glycopeptides, but the lectin domain is not required for catalysis [106]. Expression of Galnt10 in the brain in mice might be related to human schizophrenia, and a recent GWAS showed an unambiguous association between schizophrenia and GALNT10 with other glycosyltransferases [107]. Together with GalNAc-T7, follow-up or glycopeptide-preferring GalNAc-Ts may be associated with serious mental diseases, but further work is needed to draw a clear conclusion.

GALNT11-Related Phenotypes
Human and murine GALNT11 were identified in 2002 [5]. Human GALNT11 is highly expressed in the kidney and moderately expressed in the brain, heart and skeletal muscle; in contrast, murine Galnt11 is highly expressed only in the kidney. A recent GWAS identified an association between human GALNT11 and deterioration of kidney function [108], and an assessment of GalNAc-T11 function in the kidney by Tian et al. [109] showed that Galnt11-deficient mice suffered from low-molecular-weight proteinuria. The same study also identified the endocytic receptor megalin/LRP2, a member of the low-density lipoprotein receptor (LDLR) family, as a specific substrate for GalNAc-T11 in the kidney. Megalin showed reduced binding to endogenous ligands in the absence of Galnt11. Another important aspect of GALNT11 is that its expression is related to heterotaxy, a congenital heart disease resulting from abnormalities in left-right body patterning [110]. A case report of a 22-month-old male indicated that GALNT11, together with GALNT20 (GALNTL5), which seems to be a non-functional protein (see below), was associated with developmental delay, distinctive facial features and multiple congenital anomalies [111]. These observations may be related to Notch signaling activation due to Notch1 O-glycosylation by GalNAc-T11 [112]. Galnt11 and NOTCH1 knockdown and gain-of-function produced similar left-right patterning in Xenopus, suggesting that GALNT11 affects the Notch signaling pathway [112].

GALNT12-Related Phenotypes
Human GALNT12, which is most homologous to GALNT4, was cloned and shown to be expressed in digestive organs, including the stomach, small intestine, pancreas and colon. Moderate expression of GALNT12 occurs in the testis, thyroid gland and spleen [113]. Recent GWASs have shown that GALNT12 is negatively associated with serum galactosedeficient IgA1 levels resulting in IgA nephropathy [114], and a single nucleotide polymorphism (SNP)-rs2295926, belonging to GALNT12-is strongly associated with rapid radiographic joint destruction in patients with rheumatoid arthritis [115], but the underlying molecular mechanisms are unclear.

GALNT13-Related Phenotypes
Murine Galnt13 was cloned and published as a homolog of Galnt1 in 1995 [116]. T-cell specific depletion of the gene in mice resulted in no phenotype. Human GALNT13 was cloned and shown to be specifically expressed in the brain [100]. In this study, Galnt13 null mice were generated; they had a decrease in the Tn (GalNAc-Thr/Ser) antigen in Purkinje cell bodies and internal granular layer cells of the cerebellum cortex, compared with wild-type mice. However, the mice were fertile and developed normally. A case of a 13-year-old girl showed a relationship of GALNT13 with minor facial and digital anomalies, mild developmental delay during infancy and behavioral disorders, which might reflect high expression of GALNT13 in the brain [117].

GALNT14-Related Phenotypes
Human GALNT14 was cloned based on GALNT2 and is expressed specifically in the kidney [118]. Using autozygosity mapping and whole-exome sequencing, GALNT14 with homozygous mutation in exon13 (c.[1273C>T]) resulting in truncation of >75% of the protein sequence (p.[Arg425*]) has been linked to embryonic lethality [119]. A heterozygous intragenic deletion of 123 kb in GALNT14 was identified in an intestinal malrotation case by analysis of array comparative genomic hybridization (aCGH) data [120]. A wholeexome sequencing study revealed that a homozygous frameshift in GALNT14 (c. [60del] or p.[Leu21Cysfs*6]) was identified as the genetic cause of Keratoconus in two cases [121]. However, the molecular mechanisms behind these observations are yet to be known.

GALNT15 (GALNTL2)-Related Phenotypes
Human GALNT15 was cloned using cDNA of human cerebellum as a template. The transcripts were ubiquitously expressed in human tissues, with high expression in the placenta and small intestine, and moderate expression in the spleen, ovary and cerebral cortex [122]. Even though its enzymatic activity using peptide substrates was confirmed in vitro, GALNT15-related phenotypes in mammals have not been reported.

GALNT16 (GALNTL1)-Related Phenotypes
Human GALNT16 was directly submitted to GenBank (AB078143.1), and the expression and catalytic activity of human GalNAc-T16 were studied by Raman et al. [97]. The enzyme transfers GalNAc to a wide range of peptide substrates. Human GALNT16 has a broad distribution, with high expression in the heart and moderate expression in the brain and spinal cord. A recent GWAS reported an association with risk for cardiovascular disease [123]. In mice, Galnt16 expression was very low in the tissues tested, including brain, colon, heart, kidney, liver, lung, skeletal muscle, ovary, prostate, spleen, sublingual gland, testis, thymus and thyroid [18 (as GalNAc-Ta)], but upregulation through an unknown mechanism was observed in diabetic mice [124].

GALNT17 (GALNTL6)-Related Phenotypes
Human GALNT17 (described as GALNT20 at the time) was cloned based on the GALNT10 sequence, and showed high expression in the testis, brain and ovary [125]. In situ hybridization studies in mice indicated that Galnt17 might have functions in spermatogenesis. GalNAc-T17 is also a follow-up GalNAc-T and prefers glycosylated peptides as substrates [125]. These results were reconfirmed by Raman et al. [97]. Recent GWASs have revealed that GALNT17 (GALNTL6) polymorphism is associated with athletic performance through an unknown mechanism [126]. The data from the study suggested that GALNT17 (GALNTL6) rs558129 T allele carriers had significantly higher power values in a Wingate anaerobic test than those with the CC genotype. Furthermore, the T allele was overrepresented in power athletes compared with both endurance athletes and controls. Therefore, the GALNT17 (GALNTL6) rs558129 T allele could be favorable for anaerobic performance and strength of athletes [126]. A study in mice indicated that Galnt17 dysregulation is related to a phenotype in human Autism Susceptibility Candidate 2 (AUTS2) syndrome, although the underlying mechanism is unclear [127].

GALNT18 (GALNTL4)-Related Phenotypes
Human GalNAc-T18 is an interesting enzyme that is primarily distributed in the endoplasmic reticulum (ER), rather than in the Golgi apparatus, where other GalNAc-Ts are located [98,128,129]. Human GALNT18 is expressed ubiquitously, including in the lung, brain, uterus, placenta, testis and kidney [97,98], whereas mouse Galnt18 is expressed only in the lung [18 (as GalNAc-T8)]. Human GalNAc-T18 has highly specific substrate specificity and only two peptides (GTTAKPTTLKPTE and GAGAEAPTPAPAGAGK) have been identified that can be glycosylated by the enzyme [97,98]. A recent longitudinal and transancestral study showed a positive association between demethylation of a CpG site located within GALNT18 and development of active lupus nephritis [130]. In the study, a significant reduction in DNA methylation levels in a single CpG site (cg16204559) was observed during active nephritis in lupus patients. However, understanding the biological role of this demethylation in lupus nephritis will require further investigation [130].

GALNT19 (GALNTL3)-Related Phenotypes
Human and rat GALNT19 were cloned and published by Nakamura et al. as homologs of human GALNT9 [131]. As for GALNT9, human and rat GALNT19 are highly expressed in the brain, and human GALNT19 is also moderately expressed in the heart. Its enzymatic activities toward peptides are unclear [97, 98 (as GalNAc-T17), 131]. Interestingly, human GALNT19 was found to be identical to the gene WBSCR17, located in the critical region of patients with Williams-Beuren Syndrome, a neurodevelopmental disorder. Therefore, GalNAc-T19 may glycosylate brain-specific substrates in vivo [131].

GALNT20 (GALNTL5)-Related Phenotypes
Human GALNT20 is unique among GALNTs because it does not encode a lectin domain [97]. Highly specific expression of this gene in the testis has been detected in humans and mice [18,97]. Importantly, a heterozygous mutation of Galnt20 affected male fertility with impairment of sperm motility; however, enzyme activity of GalNAc-T20 against peptides could not be detected in vitro [97,132]. This might indicate that the panel of peptide substrates used for detecting GalNAc-T activity is not sufficient, and that there is a need to find "true substrates" and include these substrates in future assays.

Conclusions
Almost 10 years have passed since the 20th human GalNAc-T was cloned and published. Some of these enzymes seem to have roles in vivo, but their enzymatic activities have yet to be shown in vitro. Recent GWASs have revealed human phenotype-related GALNTs, and these studies may also be able to identify true substrates for each GalNAc-T in vivo. GalNAc-Ts may also have functions other than as GalNAc-transferases, which would be particularly interesting. Recently, the function of O-glycosylation of the proteins initiated by GalNAc-Ts has been reviewed in a broader perspective [133]. Most GALNTs are associated with carcinogenesis, as widely shown, but we have not included these findings in this review because the studies were reviewed by Hussain et al. recently [134]. Recent studies targeting each GALNT by microRNA (miRNA) were also not included, because a single miRNA targets several messenger RNAs and the publications reporting the studies are still few, although this would be an appropriate subject for a further review.