The human protein GLT25D1 was reported to have galactosyltransferase activity towards mannose binding lectin (MBL), transferring galactose to hydroxylysine residues in the Gly-X-Lys repeats . In this report we examined the GLT25D1 localisation by immunofluorescence and made deletion mutants of GLT25D1 to ascertain whether the predicted subcellular targeting signal sequences were functional (Figure 2 and Figure 3A). It should be noted that exogenous over-expression might influence localization due to saturation of the transport machinery. We performed our studies with a human hepatoma cell line, Huh7, because MBL is produced mainly by liver cells . Additionally, we confirmed by mass spectrometry analysis that GLT25D1 is normally expressed in these cells (Observed peptides are shown in Additional file 3). Moreover identical results were obtained in an additional cell line VERO, derived from monkey kidney epithelial cells. We found that GLT25D1, after being targeted to and cleaved in the ER (Figure 1 and Figure 3C), appears primarily in the ER most likely due to a functional ER retention signal, RDEL (Figure 3C). Although GLT25D1 gets secreted when a C-terminal tag is present, no secretion was found when the extreme four amino acids are RDEL, showing that these represent a functional ER retention signal (Figure 3C). This modified but related to KDEL carboxyl-terminal tetrapeptide has been shown to direct intracellular retention for several other proteins . Moreover GLT25D1 is highly sensitive to Endoglycosidase H, which is able to cleave non-complex N-linked oligosaccharides present in the ER . This not only shows targeting to the ER lumen, but also ER-retention of the soluble GLT25D1 (Figure 4A).
We furthermore demonstrate by partial digestion with Endoglycosidase H or F that at least three asparagine residues become N-glycosylated (Figure 4B). In accordance with the number of glycosylated forms we distinguish, the NetNglyc server exactly predicted three asparagines to be N-glycosylated (residues 96, 184, 404) (Figure 2), suggesting that all these three residues are modified.
GLT25D1 is able to galactosylate MBL and other collagens . Interestingly, we observe colocalization between GLT25D1 and MBL. Intracellularly MBL forms oligomers before it moves from the ER to the Golgi apparatus  and is secreted to serve as an activator of the lectin complement pathway. Binding of MBL to carbohydrates on pathogens not only can mediate an innate immune response towards microbes, but more and more data indicate a potential defense against viruses [24, 25]. The envelope glycoproteins of human immunodeficiency virus (HIV), Ebola and Influenza A virus were demonstrated to attach to serum MBL [26–28]. Additionally a subcellular interaction between HIV glycoprotein gp120 and MBL was shown . The presence of GLT25D1 mainly in the ER, where it colocalizes with MBL and LH3, suggests that galactosylation by GLT25D1 occurs early in bio-synthesis before being transported to the Golgi. This would be in line with the results of Heise et al., which demonstrate that glycosylation of MBL continues while transport to the Golgi complex is blocked by Brefeldin A treatment .
Before galactosylation of the lysine in the Gly-X-Lys repeat, the lysine is hydoxylated. LH3 has hydroxylase, galactosyltransferase and glucosyltransferase activity (Reviewed in ). When performing domain prediction analysis of GLT25D1 (see Methods), we noticed that GLT25D1 is a modular glycosyltransferase (GT) composed of two GT family members, with the C-terminal one related to glycosyltransferase of family 25 (hence its name) and an additional N-terminal domain which displays distant relation to family GT2 transferases (Figure 2 and Additional file 4). In both domains the conserved DxD motif, which binds one of the ribose hydroxyl groups, could be observed (See Additional file 4; Bernard Henrissat personal communications). Other examples of modular GTs are Heparin synthase, Chondroitin synthase and hyaluronan synthase, each involved in addition of alternating sugars, and each containing an N-terminal GT2 domain . Part of this GT2 domain of GLT25D1 surprisingly shows sequence homology to LH3, which suggests that they might function similarly either in substrate recognition or in glucosyltransferase activity.
It is interesting to note that in a search for tumor-specific markers both GLT25D1 and LH3 genes were upregulated in a large majority of human malignancies , but the biological pathway involved is unknown. The location of a particular glycosyltransferase in the cell clearly defines its function and biological importance. Most glycosyltransferases reside in the Golgi, where glycan synthesis takes place in a sequential order [12, 13]. The presence of GLT25D1 early in the secretory pathway indicates its enzyme activity is displayed there, confirming that collagens and collectins are likely to be glycosylated in the ER before they move to the Golgi apparatus and are secreted. Yet the exact role of glycosylated hydroxylysine residues in collagenous proteins is still poorly understood. Nonetheless, our results strengthen the hypothesis that glycosylation of collagens and collectins occurs early in their biosynthesis.