CDP-glycerol inhibits the synthesis of the functional O-mannosyl glycan of α-dystroglycan

α-Dystroglycan (α-DG) is a highly glycosylated cell-surface laminin receptor. Defects in the O-mannosyl glycan of an α-DG with laminin-binding activity can cause α-dystroglycanopathy, a group of congenital muscular dystrophies. In the biosynthetic pathway of functional O-mannosyl glycan, fukutin (FKTN) and fukutin-related protein (FKRP), whose mutated genes underlie α-dystroglycanopathy, sequentially transfer ribitol phosphate (RboP) from CDP-Rbo to form a tandem RboP unit (RboP–RboP) required for the synthesis of the laminin-binding epitope on O-mannosyl glycan. Both RboP- and glycerol phosphate (GroP)-substituted glycoforms have recently been detected in recombinant α-DG. However, it is unclear how GroP is transferred to the O-mannosyl glycan or whether GroP substitution affects the synthesis of the O-mannosyl glycan. Here, we report that, in addition to having RboP transfer activity, FKTN and FKRP can transfer GroP to O-mannosyl glycans by using CDP-glycerol (CDP-Gro) as a donor substrate. Kinetic experiments indicated that CDP-Gro is a less efficient donor substrate for FKTN than is CDP-Rbo. We also show that the GroP-substituted glycoform synthesized by FKTN does not serve as an acceptor substrate for FKRP and that therefore further elongation of the outer glycan chain cannot occur with this glycoform. Finally, CDP-Gro inhibited the RboP transfer activities of both FKTN and FKRP. These results suggest that CDP-Gro inhibits the synthesis of the functional O-mannosyl glycan of α-DG by preventing further elongation of the glycan chain. This is the first report of GroP transferases in mammals.

Here, we report that, in addition to having RboP transfer activity, FKTN and FKRP can transfer GroP to O-mannosyl glycans by using CDPglycerol (CDP-Gro) as a donor substrate. Kinetic experiments indicated that CDP-Gro is a less efficient donor substrate for FKTN than is CDP-Rbo. We also show that the GroP-substituted glycoform synthesized by FKTN does not serve as an acceptor substrate for FKRP and that therefore further elongation of the outer glycan chain cannot occur with this glycoform. Finally, CDP-Gro inhibited the RboP transfer activities of both FKTN and FKRP. These results suggest that CDP-Gro inhibits the synthesis of the functional O-mannosyl glycan of ␣-DG by preventing further elongation of the glycan chain. This is the first report of GroP transferases in mammals.
Recently, Yagi et al. (12) reported that there was a glycerol phosphate (GroP)-substituted phospho-core M3 unit in addition to a RboP-substituted phospho-core M3 unit based on their analysis of the glycoforms of a truncated recombinant ␣-DG expressed in HEK293T cells and HCT116 colon cancer cells. The same group also reported that GroP modifications as well as RboP modifications were not observed in FKTN-deficient cells, implying that FKTN may be involved in transferring not only RboP but also GroP to the phospho-core M3 structure. However, whether FKTN itself has GroP transferase activity and whether GroP substitution affects the synthesis of the functional core M3-type glycan remain unclear.
In Gram-positive bacteria, several GroP transferases are known to transfer the GroP moiety from the activated precursor CDP-glycerol (CDP-Gro) in the biosynthesis of teichoic acid, a major cell wall component (13). However, these GroP transferases are not conserved in mammals, and to date, no GroP transferases have been identified in mammals. In the present study, we report that FKTN and FKRP have GroP transfer activity using CDP-Gro as a donor substrate. We also demonstrate that GroP substitution of a phospho-core M3 unit by FKTN inhibits further elongation of the outer glycan chain. Furthermore, we show the inhibitory effect of CDP-Gro on the RboP transferase activity of FKTN and FKRP. The present data suggest the inhibitory function of CDP-Gro in the synthesis of ␣-DG glycan.

sFKTN has GroP transfer activity using CDP-Gro as a donor
To determine whether FKTN has GroP transferase activity as well as RboP transferase activity to the phospho-core M3 unit, we prepared a soluble form of FKTN with a His/Myc tag at the N terminus (sFKTN) and a synthetic phospho-core M3 peptide (AT*PAPVAAIGPK) modified with phospho-core M3 at the Thr* as an acceptor substrate. sFKTN was expressed in HEK293T cells and immunoprecipitated from the culture medium using an anti-c-Myc antibody. The immunoprecipitated sFKTN was found to be of the appropriate molecular weight by Western blotting (Fig. 1A), and the purity of sFKTN was checked by Coomassie Brilliant Blue staining (Fig. S1A). To confirm the enzymatic activity of sFKTN, sFKTN was incubated with CDP-Rbo (Fig. 1B) and the phospho-core M3 peptide, and the products were separated by HPLC. In the HPLC chromatogram, shown in Fig. 1C, a small peak of the acceptor substrate (phospho-core M3 peptide; indicated as S) and a large peak of the RboP-transferred product (RboP-phospho-core M3 peptide; indicated as P1) were detected as reported previously (4). Because CDP-Gro is a known donor substrate for GroP transferase in bacterial teichoic acid biosynthesis (13), we examined whether sFKTN has GroP transferase activity using CDP-Gro as a donor substrate. Commercially available CDP-Gro (Sigma) is a mixture of enantiomers CDP-1-Gro and CDP-3-Gro (Fig. 1B), and these isomers can serve as the donors for sn-glycerol 1-phosphate (Gro1P) and sn-glycerol 3-phosphate (Gro3P), respectively. Because whether the GroP on the core M3-type glycan of recombinant ␣-DG (12) is Gro1P or Gro3P remains unclear, we first used the mixture of CDP-Gro enantiomers (mix-CDP-Gro). sFKTN was incubated with mix-CDP-Gro and the phospho-core M3 peptide, and the products were analyzed by HPLC. As shown in Fig. 1D, a new small peak was detected in the presence of sFKTN (lower panel, indicated by P2) but not when sFKTN was absent (upper panel). The fraction containing peak P2 was isolated and analyzed using MALDI-TOF-MS. The m/z value of P2 was 1935.9, which corresponds to the [M ϩ H] ϩ ion of the GroP-transferred phospho-core M3 peptide (GroP ϩ GalNAc-GlcNAc-(P)Man peptide) (Fig. 1E)  , and m/z 1133.6 (peptide), indicating that the peak at m/z 1935.9 is the GroP-substituted phosphocore M3 peptide. The peak at m/z 1763.8 corresponds to the dehydrated form of m/z 1781.8. There were no obvious peaks in the MS/MS spectrum containing GroP, which suggested that GroP is attached to the terminal GalNAc of the phospho-core M3. The position to which the GroP is transferred is further analyzed below.

sFKTN transfers GroP to the terminal GalNAc of phospho-core M3 peptide
To determine the position on the phospho-core M3 peptide to which sFKTN transfers GroP, we used exo-␤-N-acetylhexosaminidase (HexNAcase) from jack bean. Because HexNAcase can remove nonreducing, terminal ␤-linked HexNAc (GlcNAc and GalNAc) residues but cannot cleave modified HexNAc (nonterminal HexNAc), HexNAcase can be used to determine whether the terminal GalNAc is modified with GroP. To confirm the enzymatic activity of HexNAcase toward ␤1-3-linked GalNAc, the phospho-core M3 peptide was treated with HexNAcase, and the product was analyzed by HPLC and MS. A new peak (indicated as SЈ) was observed in the HPLC chromatogram after HexNAcase treatment ( Fig. 2A), and the fraction containing the substrate (S) and product (SЈ) were subjected to MS analysis. In addition to the substrate (S) peak at m/z 1781.9 (GalNAc-GlcNAc-(P)Man-peptide), two major peaks at m/z 1578.7 (GlcNAc-(P)Man-peptide) and m/z 1375.7 ((P)Manpeptide) were observed after HexNAcase treatment (Fig. 2B), demonstrating that the terminal GalNAc was removed and that the resulting terminally exposed GlcNAc was also removed. Under these experimental conditions, the GroP-substituted phospho-core M3 peptide (P2) was treated with HexNAcase, and the product was analyzed by HPLC. The intensities of peak P2 with or without HexNAcase treatment were comparable (Fig. 2C), although a very small product peak (SЈ) was observed

CDP-Gro inhibits O-mannosyl glycan synthesis
after HexNAcase treatment in Fig. 2C because the P2 fraction contained a small amount of phospho-core M3 peptide (S) as a contaminant during the isolation of the GroP-phospho-core M3 peptide synthesized from the phospho-core M3 peptide by sFKTN. MS analysis of the eluates around peak P2 showed that the substrate (P2) peak is the only GroP-containing peak regardless of HexNAcase treatment (Fig. 2D), indicating that the GroP substituent on the phospho-core M3 peptide was not  Figure 2. sFKTN transfers GroP to the terminal GalNAc of the phospho-core M3 peptide. A, the phospho-core M3 peptide was treated with or without HexNAcase, and the products were analyzed by HPLC. S, unreacted substrate (phospho-core M3 peptide); SЈ, product of the reaction with HexNAcase. Eluates with the retention times indicated by dotted lines were subjected to MS analysis (B). B, the MALDI-TOF-MS spectra of S without HexNAcase (upper) and S and SЈ in A with HexNAcase (lower). C, the GroP-transferred phospho-core M3 peptide was treated with or without HexNAcase, and the products were analyzed by HPLC. P2, GroP-phospho-core M3 peptide; S, phospho-core M3 peptide (a small amount of this peptide was present in the P2 fraction as a contaminant); SЈ, product of the HexNAcase reaction with S. Eluates with the retention times indicated by dotted lines were subjected to MS analysis (D). D, the MALDI-TOF-MS spectra of the eluates in C without HexNAcase (upper) and with HexNAcase (lower). Asterisks, fragment ions of P2 and S formed during the MS experiment. The m/z 1375.7 and 1578.8 peaks (lower) also contain the HexNAcase reaction products with S.

CDP-Gro inhibits O-mannosyl glycan synthesis
cleaved by HexNAcase digestion. These results show that sFKTN transferred GroP to the terminal GalNAc of the phospho-core M3 peptide, and this result is consistent with the previous report that GroP is linked to the terminal GalNAc of the phospho-core M3 unit on recombinant ␣-DG (12).

Both CDP-1-Gro and CDP-3-Gro serve as donors for sFKTN
Next, to determine which enantiomer, CDP-1-Gro or CDP-3-Gro (or both), serves as the donor for sFKTN, we synthesized both isomers using glycerol-phosphate cytidylyltransferase (GCT), which catalyzes the production of CDP-Gro from CTP and GroP in microorganisms (14). For the GCT reaction, we used a GCT from the bacterium Aquifex aeolicus, AQ1368, which has been shown to synthesize CDP-1-Gro and CDP-3-Gro using Gro1P and Gro3P, respectively ( Fig. S2A) (15). We expressed His-tagged recombinant AQ1368 in Escherichia coli. The purified His-tagged AQ1368 (Fig. S2B) was incubated with CTP and either Gro1P or Gro3P. The reaction products were separated by HPLC, and the peak corresponding to CDP-Gro was collected (Fig. S2C). The purities of CDP-1-Gro and CDP-3-Gro were verified by HPLC (Fig. S2D). HPLC-MS analysis was used to confirm the production of CDP-Gro (Fig. S3). Then, sFKTN was incubated with the phospho-core M3 peptide and CDP-1-Gro or CDP-3-Gro, and the products were analyzed by HPLC. As shown in Fig. 3, A and B, a new small peak (indicated by an open triangle) was detected in the chro-matograms after the reaction with CDP-1-Gro or CDP-3-Gro as was the case with mix-CDP-Gro (Fig. 1D). The fractions containing the reaction product with CDP-1-Gro or CDP-3-Gro were subjected to MS analyses, and the m/z value of the products was 1936.1 or 1936.0, respectively, corresponding to the [M ϩ H] ϩ ion of the GroP-transferred phospho-core M3 peptide (GroP ϩ GalNAc-GlcNAc-(P)Man peptide) (Fig. 3, C and D). These results indicate that sFKTN transfers both Gro1P and Gro3P from CDP-1-Gro and CDP-3-Gro, respectively.

Kinetic analysis of sFKTN with CDP-Rbo, CDP-1-Gro, and CDP-3-Gro
To compare the efficiency of CDP-Rbo, CDP-1-Gro, and CDP-3-Gro as donor substrates for sFKTN, kinetic experiments were performed using increasing concentrations of each donor and 100 M phospho-core M3 peptide. As shown in Fig.  4A, the transfer reactions of RboP, Gro1P, or Gro3P in the presence of CDP-Rbo, CDP-1-Gro, or CDP-3-Gro, respectively, followed typical Michaelis-Menten kinetics. Analysis of the results using a Lineweaver-Burk plot (
Although CDP-1-Gro appears to be a better substrate for sFKTN than CDP-3-Gro, both CDP-1-Gro and CDP-3-Gro serve as significant donor substrates. In addition, when a mixture of equal amounts of CDP-1-Gro and CDP-3-Gro was used as the donor substrate, medial GroP (Gro1P/Gro3P) transfer activity was observed (Fig. S4). This result suggests that CDP-1-Gro and CDP-3-Gro do not affect the Gro3P and Gro1P transfer activities of sFKTN, respectively. Therefore, we used mix-CDP-Gro (Sigma) in subsequent experiments, and hereafter, we refer to mix-CDP-Gro simply as CDP-Gro except where otherwise indicated.

GroP-phospho-core M3 peptide cannot serve as an acceptor substrate for a soluble form of FKRP (sFKRP)
Because sFKTN can transfer both RboP and GroP, we were curious to determine the effect of GroP modification of phospho-core M3 on the synthesis of core M3-type glycan. To this aim, we examined whether the GroP-phospho-core M3 peptide serves as an acceptor substrate for FKRP, the enzyme immediately downstream of FKTN. sFKRP was expressed in HEK293T cells and immunoprecipitated from the culture medium. The immunoprecipitated sFKRP was confirmed to have the appropriate molecular weight by Western blotting (Fig. 5A), and the purity of sFKRP was checked by Coomassie Brilliant Blue staining (Fig. S1B). To confirm the enzymatic activity of sFKRP, sFKRP was incubated with CDP-Rbo and the RboP-phosphocore M3 peptide, which is the usual acceptor for FKRP, and the products were analyzed by HPLC. As shown in Fig. 5B, an sFKRP product (RboP-RboP-phospho-core M3 peptide) was detected in the chromatogram (peak P3), and the content of the substrate (P1; RboP-phospho-core M3 peptide) was markedly reduced by the presence of sFKRP (lower panel). MS analysis of the fraction containing peaks P1 and P3 confirmed the production of RboP-transferred RboP-phospho-core M3 peptide (P3) (Fig. 5C). Then, sFKRP was incubated with CDP-Rbo and the GroP-phospho-core M3 peptide, and the products were analyzed by HPLC. As shown in Fig. 5D, two peaks were detected in both the absence and presence of sFKRP; the large peak (P2) corresponds to the GroP-phospho-core M3 peptide, and the small peak (S) corresponds to the phospho-core M3 peptide, which is present as a contaminant during the isolation of the GroP-phospho-core M3 peptide. The peak pattern in the chromatogram after the reaction with sFKRP was almost entirely the same as that of the reaction in the absence of sFKRP, and no new peaks were detected (Fig. 5D, lower panel), demonstrating that RboP was not transferred to the GroP-phospho-core M3 peptide by sFKRP. MS analysis of the eluate around peak P2 (Fig. 5D, lower panel) confirmed that RboP-transferred GroPphospho-core M3 peptide (calculated m/z value, 2149.9) was not produced (Fig. 5E). These results indicate that the GroPphospho-core M3 peptide cannot serve as an acceptor for sFKRP.

sFKRP also has GroP transfer activity using CDP-Gro
Both FKTN and FKRP have RboP transfer activity using CDP-Rbo as a donor substrate, and the catalytic domain of FKRP shows significant sequence similarity to that of FKTN (16), indicating that FKRP might also have GroP transfer activity using CDP-Gro. Therefore, sFKRP was incubated with CDP-Gro and its usual acceptor, the RboP-phospho-core M3 peptide, and the products were analyzed by HPLC. As shown in Fig.  6A, a new small peak (indicated as P4) was detected in addition to the substrate peak (P1; RboP-phospho-core M3 peptide) in the chromatogram after the reaction in the presence of sFKRP

CDP-Gro inhibits O-mannosyl glycan synthesis
(lower panel) but not in that after the reaction without sFKRP (upper panel). The fraction containing peaks P1 and P4 was subjected to MS analysis, and the results showed that the m/z value of P4 was 2149.8, which corresponds to the [M ϩ H] ϩ ion of the GroP-transferred RboP-phospho-core M3 peptide (GroP ϩ RboP-GalNAc-GlcNAc-(P)Man peptide) (Fig. 6B). The components of P4 were further confirmed by MS/MS analysis of the peak at m/z 2149.8. As shown in Fig. 6C, five fragment ions were detected at m/z 1995.8 (RboP-GalNAc-GlcNAc-(P)Man-peptide), m/z 1781.8 (GalNAc-GlcNAc-(P)Man-peptide), m/z 1578.8 (GlcNAc-(P)Man-peptide), m/z 1375.7 ((P)Man-peptide), and m/z 1133.7 (peptide), indicating that the peak at m/z 2149.8 is the GroP-substituted RboP-phospho-core M3 peptide. No obvious peaks containing GroP were detected in the MS/MS spectrum, suggesting that GroP is attached to the terminal RboP. These results indicate that sFKRP also has GroP transfer activity using CDP-Gro as a donor. In addition, the HPLC data show that the GroP-transferred product peak (Fig.   Figure 5. The GroP-phospho-core M3 peptide cannot serve as an acceptor substrate for sFKRP. A, immunoprecipitation of the Myc-tagged sFKRP. sFKRP in the culture supernatant was immunoprecipitated with anti-Myc-agarose (rabbit polyclonal) and subjected to Western blotting with an anti-Myc antibody (goat polyclonal). B, enzymatic activity of sFKRP with CDP-Rbo and the RboP-phospho-core M3 peptide. Upper, without sFKRP; lower, with sFKRP. The products formed after 2-h reaction at 37°C were analyzed by HPLC. P1, unreacted acceptor substrate (RboP-phospho-core M3 peptide); P3, product of the reaction of sFKTN with CDP-Rbo. Eluate with the retention times indicated by dotted lines were subjected to MS analysis (C). C, MALDI-TOF-MS spectrum of P1 and P3. Asterisks, fragment ions of P1 and P3 formed during the MS experiment. D, no enzymatic activity was observed for sFKRP with CDP-Rbo and the GroP-phosphocore M3 peptide. The products formed after 2-h reaction at 37°C were analyzed by HPLC. Upper, without sFKRP; lower, with sFKRP. P2, unreacted acceptor substrate (GroP-phospho-core M3 peptide); S, the phospho-core M3 peptide (a small amount of this peptide was in the P2 fraction during the HPLC fractionation process). Eluates with the retention times indicated by dotted lines were subjected to MS analysis (E). E, MALDI-TOF-MS spectrum of P2 and S. Asterisks, fragment ions of P2 and S formed during the MS experiment.

CDP-Gro inhibits O-mannosyl glycan synthesis
6A, lower panel, P4) is smaller than the RboP-transferred product peak (Fig. 5B, lower panel, P3) using the same enzyme assay conditions, suggesting that CDP-Gro is a less efficient donor for sFKRP than CDP-Rbo, which is the case with sFKTN.
We further examined whether sFKRP can transfer GroP to the GroP-transferred phospho-core M3 peptide. sFKRP was incubated with CDP-Gro and the GroP-phospho-core M3 peptide, and the products were analyzed by HPLC. As shown in Fig.  6D, the substrate peak (GroP-phospho-core M3 peptide; indicated by P2) and the associated peak (phospho-core M3 peptide; indicated by S) were detected, as shown in Fig. 5D, but no new peaks were detected, indicating that sFKRP cannot transfer GroP to the GroP-transferred phospho-core M3 peptide. MS analysis of the eluate around peak P2 confirmed that GroPtransferred GroP-phospho-core M3 peptide (calculated m/z value, 2089.9) was not produced (Fig. 6E). The above results demonstrated that the GroP-transferred phospho-core M3 unit cannot be further elongated with RboP or GroP by sFKRP, and a tandem GroP-GroP unit cannot be formed on the core M3 glycan.

RboP transfer activities of sFKTN and sFKRP are inhibited by CDP-Gro
Because both RboP-and GroP-modified O-mannosyl glycans of recombinant ␣-DG were observed in cultured cells (12), it is conceivable that CDP-Rbo and CDP-Gro (CDP-1-Gro, CDP-3-Gro, or both) coexist in the cells. Kinetic data of sFKTN (Fig. 4) suggest that sFKTN exhibits comparable affinity for CDP-Rbo and CDP-Gro, but the GroP transfer rate is noticeably lower than the RboP transfer rate, suggesting that CDP-Gro may competitively inhibit RboP transfer from CDP-Rbo. Then, we examined the effect of the coexistence of CDP-Rbo and CDP-Gro on the RboP transfer activity of sFKTN. Because the RboP-phospho-core M3 peptide and GroP-phospho-core M3 peptide, both of which are produced by sFKTN, are eluted at almost the same HPLC retention time, it is impossible to monitor only the RboP transfer activity in the presence of both CDP-Rbo and CDP-Gro by HPLC analysis. Therefore, we used 3 H-labeled CDP-Rbo (CDP-[ 3 H]Rbo) to determine only the RboP transfer activity. sFKTN was incubated with 100 M phospho-core M3 peptide; 500 M CDP-[ 3 H]Rbo; and 0, 500, or 2,500 M CDP-Gro (molar ratio of 1:0, 1:1, or 1:5 CDP-Rbo to CDP-Gro). The reaction mixtures were analyzed with HPLC, and the radioactivities incorporated into the acceptor peptide were measured. As shown in Fig. 7A, the product peak (indicated by an open triangle) became small as the molar ratio of CDP-Gro to CDP-Rbo increased. In accord with this observation, the RboP transfer activity decreased with an increasing molar ratio of CDP-Gro to CDP-Rbo (Fig. 7B). The RboP transfer activity decreased to 63 (CDP-Rbo:CDP-Gro, 1:1) or 30% (CDP-Rbo:CDP-Gro, 1:5) of that without CDP-Gro (CDP-Rbo: CDP-Gro, 1:0). The reduction in the RboP transfer activity may not be due to the concomitant increase in the GroP transfer activity because of the lower amount of the total reaction product generated in the presence of CDP-Gro (Fig. 7A) and the low GroP transfer activity of sFKTN using CDP-1-Gro or CDP-3-Gro (Fig. 4). These results indicate that the RboP transfer activity of sFKTN is inhibited by CDP-Gro. In addition, we observed that the RboP transfer activity of sFKTN was not inhibited by ribitol, RboP, or Gro3P (Fig. S5), suggesting that the CDP moiety is important for the inhibition of sFKTN.
In addition, we examined whether the presence of CDP-Gro also affects the RboP transfer activity of sFKRP. sFKRP was incubated with 50 M RboP-phospho-core M3 peptide; 250 M CDP-[ 3 H]Rbo; and 0, 250, or 1,250 M CDP-Gro (molar ratio of 1:0, 1:1, or 1:5 CDP-Rbo to CDP-Gro). The product mixtures were subjected to HPLC analysis, and radioactivities incorporated into the acceptor peptide were determined. As was the case with sFKTN, in the HPLC chromatograms, the product peak (indicated by an open triangle) became small as the molar ratio of CDP-Gro to CDP-Rbo increased (Fig. 7C). In accord with this observation, the RboP transfer activity also decreased with an increasing molar ratio of CDP-Gro to CDP-Rbo (Fig.  7D). The RboP transfer activity decreased to 65 (CDP-Rbo: CDP-Gro, 1:1) or 19% (CDP-Rbo:CDP-Gro, 1:5) of that without CDP-Gro (CDP-Rbo:CDP-Gro, 1:0). These results indicate that the RboP transfer activity of sFKRP is inhibited by CDP-Gro much like it is for sFKTN.

Discussion
In the present study, we showed that FKTN can transfer GroP from CDP-Gro, in addition to RboP from CDP-Rbo, to the core M3 glycan. These results are consistent with the previous report that the GroP-substituted phospho-core M3 unit is not found in FKTN-deficient cells (12), and the present study provides direct evidence that FKTN itself is the GroP transferase. In addition, we showed that FKRP also has GroP transfer activity using CDP-Gro as well as RboP transfer activity. To date, no GroP transferases have been reported in mammals, and therefore, this is the first report of mammalian GroP transferases using CDP-Gro as a donor substrate.
Based on the kinetic data of sFKTN (Fig. 4), the K m values for CDP-Rbo, CDP-1-Gro, and CDP-3-Gro are similar, suggesting that glycerol being shorter than ribitol does not substantially affect its affinity for sFKTN. Meanwhile, the V max value for CDP-Rbo is much higher than those for CDP-1-Gro and CDP-3-Gro, suggesting that the larger size of ribitol is required for efficient transfer. These kinetic properties indicate that CDP-Rbo is a preferred substrate for FKTN. The V max value for CDP-1-Gro is higher than that for CDP-3-Gro. Because the orientation of the hydroxyl group at the ␤-carbon of the phosphate group is the same in CDP-Rbo and CDP-1-Gro but not in CDP-3-Gro, the orientation of this hydroxyl group may affect the transfer activity of FKTN to some degree. Conversely, when CDP-Gro (CDP-1-Gro, CDP-3-Gro, or both) and CDP-Rbo coexist in the cells, CDP-Gro may act as a competitive inhibitor of the RboP transfer activity of FKTN due to its substantial affinity for FKTN. Actually, CDP-Gro inhibits the RboP transfer activity of sFKTN from CDP-Rbo in vitro (Fig. 7, A and B). A similar inhibitory effect of CDP-Gro was also observed in the case of sFKRP (Fig. 7, C and D), suggesting that FKRP and FKTN have similar kinetic properties in their reactions with CDP-Rbo and CDP-Gro.
How CDP-Gro (CDP-1-Gro, CDP-3-Gro, or both) is synthesized in mammalian cells remains unclear. In microorganisms, GCT is known to synthesize CDP-Gro from GroP and CTP (14) as mentioned above. However, GCT is not conserved in mammals. In addition, CDP-Rbo is synthesized from RboP and CTP by ISPD (4, 5, 17), but GroP does not serve as a sub-strate for ISPD (18), and ISPD-deficient cells also contain the GroP-substituted core M3 unit (12). These reports indicate that ISPD does not synthesize CDP-Gro. Therefore, mammals may

CDP-Gro inhibits O-mannosyl glycan synthesis
have a different enzyme that catalyzes the synthesis of CDP-Gro.
The data presented herein showed that CDP-Gro inhibits the synthesis of core M3-type glycan by two mechanisms. First, when FKTN transfers GroP from CDP-Gro to the phosphocore M3 peptide, the GroP-transferred product does not then serve as a substrate for FKRP, and further elongation of the outer glycan chain cannot occur. In fact, further glycosylation of the GroP substituent was not observed in the glycoform of a truncated recombinant ␣-DG (12), which is consistent with our results. Second, CDP-Gro inhibits the RboP transfer activities of FKTN and FKRP. This result also suggests that potential CDP-Rbo-using enzymes other than FKTN or FKRP are commonly inhibited by CDP-Gro. However, it remains unclear whether CDP-Gro actively acts as a stop signal of core M3-type glycan synthesis or whether FKTN and FKRP erroneously recognize CDP-Gro, and both of these mechanisms result in the inhibition of core M3-type glycan synthesis. The following issues will be key to revealing whether CDP-Gro functions as a regulator of core M3-type glycan synthesis: 1) the amount of CDP-Gro and the ratio of CDP-Rbo to CDP-Gro in the cells, 2) the biosynthetic machinery of CDP-Gro, and 3) the regulatory mechanisms of the production and/or clearance of CDP-Rbo and CDP-Gro in the cells.
Loss of ␣-DG glycosylation is observed not only in ␣-dystroglycanopathy but also in a variety of cancers (19 -21). Hypoglycosylation of ␣-DG hinders interactions between the cells and extracellular matrix, and it can affect intracellular signaling, resulting in invasive and proliferative phenotypes in cancer cells (19,21). In several cancers, reduced expression levels of enzymes involved in the glycosylation of ␣-DG, such as LARGE in breast cancer (22), have been reported. Because CDP-Gro may also induce hypoglycosylation of ␣-DG, CDP-Gro may be involved in cancer progression.

Cell culture and transfection
HEK293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Nakalai Tesque, Kyoto, Japan) supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Transfection of the plasmid DNA into cells was performed using Lipofectamine 3000 (Life Technologies) according to the manufacturer's protocol.

Preparation of the phospho-core M3 peptide and RboP-phospho-core M3 peptide
The synthesis of the core M3 peptide (H-Ala-Thr(Man-GlcNAc-GalNAc)-Pro-Ala-Pro-Val-Ala-Ala-Ile-Gly-Pro-Lys-NH 2 ) was performed via the Fmoc solid-phase peptide synthesis using a glycosylated threonine derivative that was synthesized as shown in the supporting experimental procedures. In Fmoc solid-phase peptide synthesis, each Fmoc-amino acid was coupled by 1,3-diisopropylcarbodiimide and N-hydroxybenzotriazole hydrate. The Fmoc group was deprotected by treatment with 20% (v/v) piperidine/N-methylpyrrolidone for 15 min. Acetic anhydride (2 ml) and pyridine (4 ml) were used for the acetyl capping. After elongation of the peptide, the resin was washed with piperidine/N-methylpyrrolidone, dichloromethane, methanol, and diethyl ether and then dried in vacuo.
To cleave the peptide from the resin and deprotect it, the resin was treated with a TFA:triisoproylsilane:H 2 O mixture (95:2.5: 2.5) at room temperature. After being stirred for 2 h, the resulting mixture was filtered. The filtrate was concentrated under reduced pressure, and then diethyl ether was added to give a white precipitate. The resulting solid was separated by decantation, washed with diethyl ether, and then dried under reduced pressure to give the crude, O-acetyl-protected glycopeptide. To a solution of the crude glycopeptide in acetonitrile:H 2 O (1:3; 4 ml), the aqueous solution of tetrabutylammonium hydroxide (40%, v/v; 258 l) was added at 0°C. The resulting mixture was stirred at room temperature. After 5 h, TFA (50 l) was added, and purification by preparative HPLC (column, Inertsil ODS-3; gradient, CH 3 CN:H 2 O 15:85 to 30:70) gave the core M3 peptide (21.1 mg; 39%) as a white powder. Low-resolution MALDI-TOF-MS: m/z calculated for [M ϩ H] ϩ , 1701.9; observed, 1700.9. The phospho-core M3 peptide was synthesized from the core M3 peptide using soluble POMK as described previously (4). The RboP-phospho-core M3 peptide was synthesized as described previously (4).

Enzyme assay for sFKTN and sFKRP
The sFKTN or sFKRP expression plasmids (4) were transfected into HEK293T cells. The recombinant proteins were immunoprecipitated from the culture supernatant with anti-c-Myc antibody-agarose (rabbit polyclonal, Sigma). The proteins bound to the agarose were used as the enzyme sources. sFKTN and sFKRP protein expression levels were determined by Coomassie Brilliant Blue staining and Western blotting as described previously (10). sFKTN enzymatic reactions were performed on 20 l of solution containing 100 mM MES (pH 6.5), 500 M donor substrate, 100 M acceptor peptide, 10 mM MnCl 2 , 10 mM MgCl 2 , 0.5% Triton X-100, and 5 l of the enzyme-bound agarose at 37°C for 15 h except where otherwise indicated. For the kinetic experiments on sFKTN, 12.5, 25, 100, 250, and 500 M CDP-Rbo, CDP-1-Gro, or CDP-3-Gro and 100 M acceptor peptide were used, and the mixture was incubated for 4 h. For the sFKTN enzymatic reactions in the presence of both CDP-Rbo and CDP-Gro, 0, 500, or 2,500 M mix-CDP-Gro (Sigma); 500 M CDP-[ 3 H]Rbo (50,000 dpm/nmol); and 100 M phospho-core M3 peptide were used, and the mixture was incubated for 4 h. For sFKTN enzymatic reactions in the presence of both CDP-Rbo and ribitol, RboP, or Gro3P, 500 M CDP-Rbo; 5,000 M ribitol, RboP, or Gro3P; and 100 M phospho-core M3 peptide were used, and the mixture was incubated for 2 h. sFKRP enzymatic reactions were performed on 20 l of solution containing 100 mM MES (pH 6.5), 250 M CDP-Rbo or mix-CDP-Gro (Sigma), 50 M acceptor peptide, 10 mM MnCl 2 , 10 mM MgCl 2 , 0.5% Triton X-100, and 5 l of the enzymebound agarose at 37°C for 2 h except where otherwise indi-

CDP-Gro inhibits O-mannosyl glycan synthesis
cated. For sFKRP enzymatic reactions in the presence of both CDP-Rbo and CDP-Gro, 0, 250, or 1,250 M mix-CDP-Gro (Sigma); 250 M CDP-[ 3 H]Rbo; 100 M RboP-phospho-core M3 peptide; and 0.8 l of the enzyme-bound agarose were used, and the mixture was incubated for 15 min. Because the expression level of sFKTN in the HEK293T culture supernatant was lower than that of sFKRP and the amount of immunoprecipitated sFKTN from the culture supernatant was smaller than that of sFKRP, FKTN assays were performed with longer incubation time, and FKRP assays were done with shorter incubation time. For kinetic experiments of sFKTN (Fig. 4) and enzymatic reactions of sFKTN and sFKRP in the presence of both CDP-Rbo and CDP-Gro (Fig. 7), we confirmed that the amounts of the reaction products showed linearity with the incubation time. Each product was separated by reversed-phase HPLC with a Mightysil RP-18GP Aqua column (4.6 ϫ 250 mm) (Kanto Chemical, Tokyo, Japan). Solvent A was 0.085% TFA in distilled water, and solvent B was 0.085% TFA in acetonitrile. Peptides were eluted at a flow rate of 1 ml/min using a linear gradient of 0 -40% solvent B. Peptide elution was monitored by determining the absorbance at 215 nm. Each enzyme activity was calculated from the product peak area except where otherwise indicated. Additionally, the separated product peak was collected and lyophilized for MALDI-TOF-MS(/MS) analysis.
To determine the enzyme activity in the presence of both CDP-Rbo and CDP-Gro, the radioactivity of each fraction (1 ml) was measured using a liquid scintillation counter.

Preparation of CDP-1-Gro and CDP-3-Gro
CDP-1-Gro and CDP-3-Gro were produced as described previously with slight modifications (15). Briefly, a synthetic codon-optimized gene, aq_1368, a GCT-encoding gene from A. aeolicus, was cloned into the NdeI/BamHI sites of pET19b to generate pET19b_aq1368. E. coli BL21(DE3) harboring pET19b_aq1368 were grown at 37°C in LB medium with ampicillin (50 g/ml) and induced with 0.5 mM isopropyl ␤-D-1thiogalactopyranoside for 4 h. The cells were harvested and resuspended in 10 mM Tris-HCl (pH 7.4) containing 5 mM MgCl 2 and a protease inhibitor mixture (Nakalai Tesque), and the cells were disrupted by sonication. The homogenate was centrifuged at 20,000 ϫ g for 15 min, and the resulting supernatant was collected. His-tagged AQ1368 was purified using His SpinTrap (GE Healthcare) and dialyzed with 50 mM Tris-HCl (pH 8.6) containing 150 mM NaCl and 5 mM MgCl 2 . The purity of His-tagged AQ1368 was verified by SDS-PAGE followed by Coomassie Brilliant Blue staining. Then, CDP-1-Gro and CDP-3-Gro were produced by heating 20 l of solution containing 50 mM Tris-HCl (pH 8.6), 5 mM MgCl 2 , 50 g of His-AQ1368, 10 mM CTP, and 10 mM Gro1P or Gro3P at 37°C for 5 min. Each product was separated using reversed-phase HPLC with a COSMOSIL 5C18-AR-II column (4.6 ϫ 250 mm) (Nakalai Tesque) by isocratic elution with 20 mM acetic acidtriethylamine (pH 7.0). Product elution was monitored by determining the absorbance at 260 nm. The separated CDP-1-Gro or CDP-3-Gro was collected, repurified on the same HPLC system, and collected again. The purities of CDP-1-Gro and CDP-3-Gro were verified by HPLC analysis, and the production of CDP-Gro was confirmed by HPLC-MS analysis.

MALDI-TOF-MS(/MS) analysis
The peptides were desalted using GL-Tip SDB (GL Sciences, Tokyo, Japan) and reconstituted in 10 l of Milli-Q water prior to MS(/MS) analysis. 2,5-Dihydrobenzoic acid was dissolved in 50% acetonitrile containing 0.1% TFA; this solution was used as the matrix solution. The peptide solution was mixed with an equal volume of the matrix solution, and 2 l of this mixture was dropped onto a Focus MALDI plate (700 m; Hudson Surface Technology, Fort Lee, NJ) and left at room temperature to dry. MS(/MS) analysis was performed on an AB SCIEX TOF/ TOF 5800 system operated with TOF/TOF Series Explorer software version 4.1 (AB SCIEX). For each spot, MS spectra were acquired in positive ion mode between m/z 800 and 4,000 and accumulated from 1,000 laser shots in a random raster. MS/MS spectra were acquired using the following parameters and methods: acceleration voltage, 1 kV; collision-induced dissociation control, off; laser shots, 4,000; precursor mass window, 200 resolution (full width at half maximum); and metastable suppression, on.

HexNAcase treatment
HexNAcase digestion was performed on 25 l of solution containing 100 mM citrate-phosphate (pH 3.5), 20 M peptide, and 20 units/ml HexNAcase from jack bean (GKX-5003, Prozyme, Hayward, CA) at 37°C for 22 h. The product was analyzed by HPLC and MALDI-TOF-MS as described above.

Statistics
All experiments were performed at least three times with comparable results. Statistical analysis was performed using one-way analysis of variance followed by a Tukey test (SPSS Statistics version 22.0, IBM, Armonk, NY).