Initiation of O-glycan synthesis in IgA1 hinge region is determined by a single enzyme, UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 2.

The hinge region of human immunoglobulin A1 (*IgA1) possesses multiple O-glycans, of which synthesis is initiated by the addition of GalNAc to serine or threonine through the activity of UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferases (pp-GalNAc-Ts). We found that six pp-GalNAc-Ts, pp-GalNAc-T1, -T2, -T3, -T4, -T6, and -T9, were expressed in B cells, IgA-bearing B cells, and NCI-H929 IgA myeloma cells. pp-GalNAc-T activities of these six enzymes for a synthetic IgA hinge peptide, which has nine possible O-glycosylation sites, were examined using a reversed phase-high performance liquid chromatography, a matrix-assisted laser desorption ionization time of flight mass spectrometry, and peptide sequencing analysis. pp-GalNAc-T2 showed the strongest activity transferring GalNAc to a maximum of eight positions. Other pp-GalNAc-Ts exhibited different substrate specificities from pp-GalNAc-T2; however, their activities were extremely weak. It was reported that the IgA1 hinge region possesses a maximum of five O-glycans, and their amino acid positions have been determined. We found that pp-GalNAc-T2 selectively transferred GalNAc residues to the same five positions. These results strongly suggested that pp-GalNAc-T2 is an essential enzyme for initiation of O-linked glycosylation of the IgA1 hinge region.

ferase (pp-GalNAc-T). 1 O-Glycans were found in many glycoproteins, particularly in secretory glycoproteins such as mucins. Eight core structures, core 1-8, that are basal structures initiated by the GalNAc addition to peptides, are recognized in the mucin-type O-glycan. Each core structure is differentially expressed in conjunction with the differentiation and malignant transformation of various cells and tissues (1)(2)(3)(4). pp-GalNAc-Ts are biologically important because they determine the number and position of mucin-type O-glycans in a protein. To date, at least 11 human pp-GalNAc-Ts , -T1, -T2,  -T3, -T4, -T6, -T7, -T8, -T9, -T10, -T11, and -T12, have been identified (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). The proteins are 40 -60% identical in their sequence and are therefore homologous. In particular, the predicted catalytic domains are highly conserved. Their substrate specificities have been examined using a variety of peptides, of which sequences are derived from native glycoproteins. They have shown different substrate specificities, different kinetic properties, and different tissue distributions, although some of them showed substantial overlaps in the assessed catalytic specificities and tissue distribution. The positions of O-glycans in proteins are determined by a variety of pp-GalNAc-Ts expressed in the cells and their substrate specificities. Some pp-GalNAc-Ts exhibit strong "primary activity" toward peptides that have no GalNAc, whereas others, such as pp-Gal-NAc-T4 and -T7, prefer peptides having GalNAc residue(s) as acceptor substrates rather than the corresponding peptides with no GalNAc (11,17,18). The "secondary activity," i.e. GalNAc addition to a peptide already having GalNAc, of the latter may be directed by a lectin domain in their catalytic region. It is supposed that multiple GalNAc transfers to a single peptide are completed in the cells by a combination of multiple pp-GalNAc-Ts having primary activity and ones having secondary activity.
Human serum IgA consists of two structurally and functionally distinct subclasses, IgA1 and IgA2, of which the ratio is about 85 and 15% of total IgA, respectively (19 -21). Immuno-* This work was supported by the New Energy and Industrial Technology Development Organization and performed as part of the R&D project of the Industrial Science and Technology Frontier Program (R&D for Establishment and Utilization of a Technical Infrastructure for Japanese Industry). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
IgAN is characterized by the selective deposition of IgA1 in the renal glomerular mesangium (23), although the mechanism for this is unclear. There is some debate regarding the cause of the deposition. Several investigators have proposed that the elevation of the serum IgA1 concentration is the cause of IgAN (24,26). To explain why IgA1, but neither IgA2 nor IgG is deposited, it was proposed that O-glycan structures in the hinge region of IgA1 are profoundly involved in the deposition. IgA1 is characterized by a long extended polyproline structure and distinctive O-glycan side chains in its hinge region (27,28). The other immunoglobulins, except for IgA1 and IgD, do not have O-glycans in their hinge regions. Five O-glycans have been identified to attach to the hinge region of IgA1 derived from healthy individuals (28,29). The amino acid positions of the O-glycans attached were determined, and the O-glycan structure was defined to be SA␣2-3Gal␤1-3GalNAc␣1-Ser/Thr (29). There are many reports describing incomplete structures of O-glycans in the IgA1 hinge region of IgAN patients compared with those of normal individuals (20, 30 -35). According to such papers, the number of O-glycans is decreased or some O-glycans are shortened in the IgA1 hinge region of IgAN patients. Iwase et al. (36,37) showed by an in vitro experiment that removal of sialic acid enzymatically resulted in the selfaggregation of IgA1, which may be the cause of the deposition of IgA1.
Based on the hypothesis that the disruption of O-glycan synthesis in the IgA1 hinge region is the cause of IgAN, we started an investigation of the molecular basis of O-glycan synthesis in the IgA1 hinge region. In this study, we examined, as a first step, which pp-GalNAc-T(s) determine the initiation of O-glycan synthesis among the many members cloned to date.  Table I. cDNAs were prepared from human lymphocytes and a cell line as follows. Peripheral blood mononuclear cells (PBMCs) of healthy volunteers were isolated from heparinized venous blood by density gradient sedimentation over Lymphoprep (Nycomed Pharma-AS, Oslo, Norway). B cells were isolated from PBMCs using CD19 Dynabeads  Table I. The amplified fragments were inserted into pFBIF as described previously (38). Recombinant proteins were purified with anti-FLAG M1 antibody resin (Sigma). Expression and purification of recombinant protein were performed as described in detail previously (14,38).
Estimation of the Amount of FLAG-tagged Recombinant pp-Gal-NAc-T Proteins-The amount of recombinant protein recovered was estimated by immunoblotting using the method reported in our previous study (39). The protein was separated by SDS-PAGE (10% gel), transferred onto polyvinylidene difluoride membranes (Immobilon, Millipore), and probed with an anti-FLAG peptide antibody BioM2 (Sigma). The FLAG-tagged bacterial alkaline phosphatase (met-FLAG-BAP, molecular mass of 49.2 kDa) was used as a standard to estimate the relative amount. The pp-GalNAc-T and BAP protein bands were quantified by densitometric scanning of the digitized image using NIH image (version 1.61) software on a Macintosh computer (Apple Computer Inc., Cupertino, CA). The standard curve for each substrate was plotted by increasing the amount of FLAG-tagged BAP protein on the same blotting membrane as the pp-GalNAc-T samples. As the band intensity and the concentration of the recombinant pp-GalNAc-T protein in the medium exhibited a linear correlation, the amounts of recombinant pp-GalNAc-T protein were estimated from the standard curve, using known amounts of FLAG-tagged BAP protein (49.2 kDa). The amount of recombinant pp-GalNAc-T protein is expressed in arbitrary units, with each unit of intensity equivalent to 10 ng of FLAG-tagged BAP protein.
Enzyme Reaction of pp-GalNAc-T with a Peptide of the IgA1 Hinge Region (HRP) as a Substrate-The amino acid sequence of the human IgA1 hinge region is reported to be VPSTPPTPSPSTPPTPSPSC (29) which contains two mucin boxes, XTPXP (40). The cysteine residue at the C terminus was replaced by a lysine in order to be labeled with 5-carboxyfluorescein (5-FAM). This peptide labeled with 5-FAM was purchased from Sawady Co., Ltd. (Tokyo, Japan), and used as an acceptor substrate.
The standard enzyme reaction mixture contained 25 mM Tris-HCl (pH 7.4), 5 mM MnCl 2 , 0.1% Triton X-100, 250 M UDP-GalNAc (Sigma), 15 M acceptor substrate, and the purified enzyme in a final volume of 20 l. The reaction mixture was incubated at 37°C for various periods, and the reaction was terminated by boiling.
Separation of Enzyme Reaction Products by Reversed Phase-High Performance Liquid Chromatography (HPLC)-The reaction products were filtered with an Ultrafree-MC column (Millipore, Bedford, MA) and subjected to reversed phase-HPLC using a Cosmosil (5C 18 -AR, 4.6 ϫ 250 mm; Nakalai Tesque, Kyoto, Japan) column. The reaction products were eluted with solvent A (0.05% trifluoroacetic acid) and solvent B (0.05% trifluoroacetic acid in 2-propanol/acetonitrile, 7:3). As a standard separation, the column was eluted with a linear gradient ranging from 10 to 15% of solvent B in solvent A at a flow rate of 1 ml/min for 40 min. With a fluorescence detector, RF-10A XL (Shimadzu, Kyoto, Japan), FAM-labeled glycopeptides were detected by the fluorescence intensity at 520 nm (excitation, 492 nm).

Identification of Reaction Products by Mass Spectrometry (MS) and
Peptide Sequencing-A method for the identification of pp-GalNAc-T products has been reported (41)(42)(43)(44). In order to determine the number of GalNAc residues attached, the glycopeptides isolated by reversed phase-HPLC were subjected to matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) MS (Reflex IV; Bruker Daltonics, Billerica, MA) in the positive ion mode. The acceleration voltage was 20 kV, and positive ion spectra were recorded in a reflectron mode. The concentrated glycopeptides in a 0.1% trifluoroacetic acid and matrix solution were mixed on a stainless steel plate at 1:1. The matrix solution contained 0.3 mg/ml of ␣-cyano-4-hydroxycinnamic acid in ethanol/ acetone, 2:1. The calibration of the molecular size (m/z) was carried out using bombesin (m/z, 1619.8), ACTH-(1-17) (m/z, 2093.1), ACTH-(18 -39) (m/z, 2465.2), and somatatostatin-28 (m/z, 3147.5). For the identification of positions of attached GalNAc residues, Edman degradation amino acid sequencing of glycopeptides was performed on a PPSQ-23A (Shimadzu). With this sequencing system, a phenylthiohydantoin (PTH) derivative of GalNAc-attached Thr (Calbiochem) was detected as two peaks eluted at 3.07 and 3.25 min, which were close to the DTT peak, and that of GalNAc-attached Ser was identified as a peak at 2.96 min which overlapped the PTH-Glu peak.

RT-PCR
Initiation of O-Glycan Synthesis in IgA1 Hinge Region 5615 scripts of 10 pp-GalNAc-Ts were amplified by RT-PCR. As shown in Fig. 1, human B cells expressed six pp-GalNAc-Ts, pp-GalNAc-T1, -T2, -T3, -T4, -T6, and -T9, but not pp-GalNAc-T7, -T8, -T10, or -T12. IgA-positive cells showed a similar expression pattern, although pp-GalNAc-T9 was not detected. Activity Analysis of pp-GalNAc-Ts Expressed in Human B Cells by HPLC-By the RT-PCR analysis, six pp-GalNAc-Ts were identified as candidates responsible for O-glycosylation of HRP in IgA-positive cells. Therefore, the six enzymes, pp-Gal-NAc-T1, -T2, -T3, -T4, -T6, and -T9, were produced as a soluble form tagged with a FLAG peptide in insect cells and purified with anti-FLAG M1 antibody agarose gel. The amounts of pp-GalNAc-T proteins were determined by Western blotting with anti-FLAG M2 antibody. An equal amount of each enzyme was employed for the pp-GalNAc-T assay using a FAM-labeled IgA1 hinge region peptide (VPSTPPTPSPSTPPTPSPSK-FAM; HRP) as an acceptor substrate. The enzymatic reaction was performed for 15-360 min using each enzyme, and then the reaction products of pp-GalNAc-Ts were separated by reversed phase-HPLC (Fig. 2). The separation profiles of reaction products differed dependent on each enzyme. In particular, the profile of pp-GalNAc-T2 differed from those of the other five pp-GalNAc-Ts. pp-GalNAc-T2 exhibited extremely strong ac-tivity for GalNAc transfer compared with the other enzymes. After the incubation with pp-GalNAc-T2 for only 15 min, the acceptor substrate was completely consumed, and multiple peaks around a retention time of 20 -30 min were observed. The elution pattern of pp-GalNAc-T2 at 120 min incubation was almost the same as that at 360 min. This indicates that the peaks at 120 min are the final products of pp-GalNAc-T2. In the case of pp-GalNAc-T1, -T3, -T4, -T6, and -T9, the peak of acceptor substrate, peak "S," still remained even after 120 min of incubation, and these five pp-GalNAc-T products were retained longer on a C18 column than the products of pp-GalNAc-T2. Three enzymes, pp-GalNAc-T1, -T4, and -T9, showed a similar profile but different from the profiles of pp-GalNAc-T2, -T3, and -T6. pp-GalNAc-T1, -T4, and -T9 mainly produced a single peak of reaction product under the conditions used, and the strength of their O-glycosylation activity was very low. pp-GalNAc-T3 and -T6 produced several peaks that were apparently different from those of pp-GalNAc-T2. As the activity of pp-GalNAc-T2 was extremely strong, the peak S and the peaks close to peak S may have already shifted to other peaks after 15 min of incubation. A more detailed experiment on pp-GalNAc-T2 products was again performed with a shorter incubation time (Fig. 3A). One-tenth as much as the amount of pp-GalNAc-T2 used in the previous experiment was employed, and the two peaks close to peak S were primarily detected after 1 min of incubation. Eleven major peaks, numbered P1 to P11, were observed in total during 120 min of incubation. Up to 360 min of incubation, the amount of P11 still remained very small (data not shown). To confirm the 11 peaks in total for pp-GalNAc-T2, we searched for reaction conditions to obtain all these peaks in a single reaction mixture. As seen in Fig. 3B, the reaction mixture of pp-GalNAc-T2 gave the 11 peaks when 22.5 M of the acceptor substrate was used, and the incubation was terminated at 5 min (Fig. 3B).
To examine the specificity of each pp-GalNAc-T in more detail, we improved the separation conditions with HPLC, and we focused on the initial products of each enzyme (Fig. 4). Fig.  4 shows initial reaction products of each enzyme early in the incubation. pp-GalNAc-T2 produced two peaks, P1 and P2, in a minute almost at the same time. pp-GalNAc-T1 and -T4 produced a single peak, "PX," indicated by an arrow, of which the retention time (88.0 min) was apparently different from those of P1 and P2. This peak was commonly produced by pp-Gal-NAc-T1, T3, -T4, -T6, and -T9 but not by -T2, as indicated by arrows in Fig. 4. pp-GalNAc-T3, -T6, and -T9 produced an additional peak, "PY," as indicated by arrows in Fig. 4, at the 82.5-min retention time that is different from that of P1 (83.0 min). These results demonstrated that pp-GalNAc-T2 has a unique specificity toward HRP different from that of other enzymes. Table II, the acceptor specificity of each pp-GalNAc-T was determined with a variety of peptides and mono-glycosylated peptides. The amount of each recombinant pp-GalNAc-T was determined as described under "Experimental Procedures," and the relative activity of each enzyme for producing the first product was presented as pmol/ min/unit. The reaction was terminated with a short incubation when the first product, of which the amount was 5-10% of the substrate, appeared. Among six enzymes examined, pp-Gal-NAc-T4 is known to have a preference toward a peptide partially glycosylated with GalNAc residues, i.e. the secondary activity. Therefore, two mono-glycosylated peptides, HRP-P1 and -P2, were employed to measure the secondary activity. Muc1aЈ and Muc7, which are unrelated to HRP, were used as control substrates. pp-GalNAc-T2 showed very weak activity toward Muc1aЈ, about one-hundredth less than the activity toward three HRP peptides. The relative activity of pp-Gal-NAc-T2 for Muc1aЈ (0.172) was only 2.9 and 1.6 times stronger than those of pp-GalNAc-T1 (0.069) and -T3 (0.105). In contrast, pp-GalNAc-T2 exhibited extremely strong activity not only for non-glycosylated HRP (primary activity) but also for mono-glycosylated HRP (secondary activity). The relative activities of pp-GalNAc-T2 for the three HRP peptides were almost 100 -500 times stronger than those of the other enzymes. Interestingly, pp-GalNAc-T4 did not show any preference toward mono-glycosylated HRP. A significant preference of the secondary activity of pp-GalNAc-T3 was observed for HRP-P2 but not for HRP-P1. Muc7 was also a very good substrate for pp-GalNAc-T2, and the other enzymes, i.e. pp-GalNAc-T1, -T3, and -T4, also preferred Muc7 as a better substrate to Muc1aЈ.

Acceptor Substrate Specificity of Each pp-GalNAc-T toward Various Substrates-As shown in
Determination of the Number of GalNAc Residues Attached to HRP by MALDI-TOF-MS-As described above, pp-Gal-NAc-T2 showed the strongest activity toward HRP and was predicted to be the most effective pp-GalNAc-T for O-glycosylation in the IgA1 hinge region. To identify the number of GalNAc residues attached to HRP, each HPLC peak of the products of the pp-GalNAc-T2 was subjected to MALDI-TOF-MS analysis.
As seen in Fig. 3B, at least 11 peaks, named P1 to P11, were obtained. Several minor peaks were also observed, but their amounts were too small to fractionate under any of the conditions tested. Therefore, we analyzed 11 products of pp-GalNAc-T2. To recover enough of the small peaks, such as P11, we adopted a different incubation time or amount of enzyme for the reaction. MS spectra of all 11 peaks are summarized in Fig.  5. [M ϩ Na] ϩ ions were observed as the major ion from all derivatives. Molecular sizes of the acceptor substrate (HRP) and GalNAc are 2338.5 and 203.2, respectively. By the MS analysis, P1 and P2, having a molecular mass of 2540.8, were defined as a mono-glycosylated HRP. Thus, P3 was defined to have two GalNAc residues, and then the results are interpreted as P4 and P5 (three GalNAcs), P6 (four GalNAcs), P7 and P8 (five GalNAcs), P9 (six GalNAcs), P10 (seven GalNAcs), and finally, P11 (eight GalNAcs). Although the number of possible O-glycosylation sites in HRP is nine, the MS analysis revealed that the maximum number of GalNAc transferred to HRP was eight as determined in P11.
Identification of the Positions of GalNAc Attached to HRP by Peptide Sequencing Analysis-Each peak of the 11 peaks in Fig. 3B was isolated and subjected to peptide sequencing to determine the positions where GalNAc was attached in HRP (data not shown). Based on 1) Fig. 3A which gives the order of appearance of each peak, 2) Fig. 5 which shows the number of GalNAcs, and 3) the results of peptide sequencing which determined the positions of the GalNAc attached, the pathways of O-glycan initiation by pp-GalNAc-T2 are schematically depicted in Fig. 6. In the HPLC analysis, P1 and P2 appeared at the same time at the beginning of the reaction, and the MS analysis revealed that both peaks carried a single GalNAc residue. Peptide sequencing revealed that the position where GalNAc was attached in P1 was Ser-11 and that in P2 was Thr-7, as shown in Fig. 6. After the transfer of the initial GalNAc residue at Thr-7, the second GalNAc residue was transferred to Thr-15 (P3). Three GalNAcs in P4 and P5 were identified to be positioned at Thr-7, Thr-12, and Thr-15 (P4) and at Thr-7, Ser-11, and Thr-15 (P5), respectively. P6, having four GalNAcs at Thr-7, Ser-11, Thr-15, and Ser-19, was generated by the addition of GalNAc to Ser-19 of P5. P7 was a mixture of two products, both of which possessed five GalNAcs at Ser-3 or Thr-4, Thr-7, Ser-9, Ser-11, and Thr-15. P8 was a product from P6 by the addition of GalNAc to Ser-9. Thereafter, P9 and P10 were produced from P7 or P8 by the addition of GalNAc to Ser-17. The final product of pp-GalNAc-T2, P11, was fully glycosylated at eight positions, Ser-3, Thr-4, Thr-7, Ser-9, Ser-11, Thr-15, Ser-17, and Ser-19.
The positions of GalNAc residues of PX and PY in Fig. 4 were also determined. PX or PY possessed a single GalNAc at Thr-12 or Thr-15, respectively. Thus, the five enzymes showed different specificities from pp-GalNAc-T2. pp-GalNAc-T2 did not produce a product having a single GalNAc at Thr-12 or Thr-15 (PX or PY).

DISCUSSION
IgAN is defined as predominant IgA1 deposits in the renal mesangium (45). Although there is some argument regarding the pathogenesis of IgAN, incompleteness of O-glycosylation in the IgA1 hinge region is considered a possible cause of the deposition (20,31,33,46). Unlike IgA2 or IgG, IgA1 has a structurally exceptional hinge portion that comprises a prolinerich sequence and possesses multiple O-glycans. In the past few years, with the progress of MS analysis, capillary electrophoresis and so on, it has been well elucidated that the number of carbohydrate residues in the IgA1 hinge region is reduced in IgAN patients compared with normal controls (47)(48)(49). Furthermore, it has been reported that incompletely glycosylated IgA1 tends to aggregate in vitro (36,50).
ined, and transferred GalNAcs to selective Ser/Thr residues that had been identified to be glycosylated in vivo (29). Thus, pp-GalNAc-T2 was determined to be merely responsible for the initiation of O-glycan synthesis in the IgA1 hinge region in this study.
Six pp-GalNAc-Ts, pp-GalNAc-T1, -T2, -T3, -T4, -T6, and -T9, were found to be expressed in IgA-positive cells. This expression profile was almost similar to that of the B cells and an IgA myeloma cell line. We examined the specificity and the relative activity of the six enzymes toward a hinge peptide (HRP) that mimicked the human IgA1 hinge region.
The retention time of HRP glycosylated by pp-GalNAc-Ts on reversed phase-HPLC became shorter dependent on the number of GalNAcs because of the hydrophilicity. As the number of incorporated GalNAc residues increased, the retention time decreased. In addition, we could separate the reaction products having the same number of GalNAcs dependent on the positions where GalNAc was added, such as P1 versus P2, P4 versus P5, and P7 versus P8.
As summarized in Fig. 6, two initial products, P1 and P2, were identified to have a GalNAc at Ser-11 or Thr-7, respectively, followed by the addition at Thr-15 as a secondary reaction. Thr-7 and Thr-15 are typical threonine residues encoded in the mucin box, XTPXP. Thus, the specificity of pp-Gal-NAc-T2 is somehow in agreement with the mucin box rule (40), although Ser-11 is not consistent with the rule. The existence of these two initial products suggested that there are at least two synthetic pathways. The next product from P2 was P3; however, the product derived from P1 was not found, and similarly some products lacked intermediates. These transient products were thought to exist as small peaks that we could not frac-tionate, and such small transient peaks may be immediately glycosylated to the following product. P5 was considered to be generated from both of the initial products. P4, which has a GalNAc at Thr-12, was produced from P3; however, the products from P4 were not detected until the end of the reaction. P4 may be a dead-end product of the reaction.
It has also been reported that the initial O-glycosylation of a peptide substrate influences the subsequent glycosylation, probably due to conformational change of the peptide and the accessibility of pp-GalNAc-Ts for particular acceptor sites (57). There are two consecutive Ser and Thr residues, such as Ser-3/Thr-4 and Ser-11/Thr-12, in HRP. Although one of the consecutive residues was glycosylated, no products in which both consecutive residues were glycosylated were found except for P11. This result indicated that pp-GalNAc-T2 cannot glycosylate the sites next to GalNAc-attached sites within HRP. This phenomenon is consistent with the previous result using a peptide substrate that mimicked the tandem repeat portion of MUC2 (41,58). This is contrast to the activity of pp-GalNAc-T13 which preferentially transfers GalNAc to all three consecutive Ser/Thr residues in a synthetic peptide derived from syndecan-3 (59).
Mattu et al. (29) analyzed O-glycans in the IgA1 hinge region that was prepared from serum IgA1 of healthy individuals. They indicated that O-glycans were attached at five positions, which correspond to Thr-4, Thr-7, Ser-9, Ser-11, and Thr-15 in our peptide, HRP. Our results indicated that pp-GalNAc-T2 is able to transfer GalNAc residues to almost all possible glycosylation sites in HRP. However, a tremendous amount of the enzyme and a very long incubation were required to glycosylate all such positions. Because a FAM-labeled short peptide was used in this study, the three-dimensional structure might be different from the physiological structure of the IgA1 molecule, and this may influence the enzyme specificities in an artificial manner. The enzyme reaction was performed in vitro; therefore, the acceptor substrate specificity of pp-GalNAc-T2 under in vitro conditions may not necessarily reflect those of pp-GalNAc-T2 in the cells. However, the glycosylated sites of P1 to P7 except for P4 and P6 were not inconsistent with the results obtained by analyzing natural IgA1 derived from human serum  (29). Supposing that the pathways up to P7 production reflect the in vivo O-glycosylation, this is consistent with the finding of Mattu et al. (29). P6 is not consistent with the results Mattu et al. (29) because it contained GalNAc at Ser-19; however, Ser-19 may be easily glycosylated because it exists at the end of an artificial peptide and may be difficult to be glycosylated in the hinge region of natural IgA1 molecule.
On the other hand, all five pp-GalNAc-Ts other than pp-GalNAc-T2 showed very weak and almost negligible activity toward HRP, and their specificities were totally different from those of pp-GalNAc-T2. In addition, they produced PX as an initial product that was mono-glycosylated at Thr-12. The Thr-12 glycosylation is not consistent with the previous report (29). These pp-GalNAc-Ts probably do not function in physiological O-glycosylation of the IgA1 in the cells.
From the above, we concluded that pp-GalNAc-T2 is merely responsible for the initiation of O-glycosylation in the IgA1 hinge region among the 10 pp-GalNAc-Ts that have been reported to date. A single enzyme, pp-GalNAc-T2, can transfer GalNAc to all five positions in the hinge region. This is quite an interesting exception from the general rule of O-glycosylation, because it has been believed that O-glycosylation on a protein is determined by a combination of multiple pp-GalNAc-Ts expressed in the cell. In the case of O-glycosylation for mucins that possess a large number of O-glycans, it is probably true that multiple pp-GalNAc-Ts are involved in the initiation of O-glycan synthesis on a single mucin.
The present study is the first report on the elucidation of O-glycosylation in the IgA1 hinge region. Up-regulation or down-regulation of the pp-GalNAc-T2 activity in the cells may lead to the increase or decrease in the number of O-glycans in IgA1, respectively. Based on the pathways in Fig. 6, we can predict that the decrease of pp-GalNAc-T2 activity will result in the loss of O-glycan in the order of Thr-4, Ser-9, Thr-15, and Thr-7 or Ser-11. To elucidate the molecular mechanisms of incomplete O-glycosylation of IgA1 of IgAN patients, more investigation will be required not only for pp-GalNA-Ts but also for core 1 Gal-T(s) and ST(s).