Autopolysialylation of polysialyltransferases is required for polysialylation and polysialic acid chain elongation on select glycoprotein substrates

Polysialic acid (polySia) is a large glycan polymer that is added to some glycoproteins by two polysialyltransferases (polySTs), ST8Sia-II and ST8Sia-IV. As polySia modulates cell adhesion and signaling, immune cell function, and tumor metastasis, it is of interest to determine how the polySTs recognize their select substrates. We have recently identified residues within the ST8Sia-IV polybasic region (PBR) that are required for neural cell adhesion molecule (NCAM) recognition and subsequent polysialylation. Here, we compared the PBR sequence requirements for NCAM, neuropilin-2 (NRP-2), and synaptic cell adhesion molecule 1 (SynCAM 1) for polysialylation by their respective polySTs. We found that the polySTs use unique but overlapping sets of PBR residues for substrate recognition, that the NCAM-recognizing PBR sites in ST8Sia-II and ST8Sia-IV include homologous residues, but that the ST8Sia-II site is larger, and that fewer PBR residues are involved in NRP-2 and SynCAM 1 recognition than in NCAM recognition. Noting that the two sites for ST8Sia-IV autopolysialylation flank the PBR, we evaluated the role of PBR residues in autopolysialylation and found that the requirements for polyST autopolysialylation and substrate polysialylation overlap. These data together with the evaluation of the polyST autopolysialylation mechanism enabled us to further identify PBR residues potentially playing dual roles in substrate recognition and in polySia chain polymerization. Finally, we found that ST8Sia-IV autopolysialylation is required for NRP-2 polysialylation and that ST8Sia-II autopolysialylation promotes the polymerization of longer polySia chains on SynCAM 1, suggesting a critical role for polyST autopolysialylation in substrate selection and polySia chain elongation.

Polysialic acid (polySia) 3 is a glycan polymer composed of large linear chains of 8 -400 ␣2,8-linked sialic acids, which in mammalian cells caps the N-linked or O-linked glycans of a small set of glycoproteins (1)(2)(3). polySia serves as both a "global regulator of cell adhesion" (4) and as a modulator of select signaling pathways (1,5). The latter function is likely due to its ability to modulate protein-protein interactions and could also be a result of its ability to serve as a reservoir for biologically active molecules like neurotrophins, neurotransmitters, and growth factors (1,2,5). polySia is independently synthesized by two polysialyltransferases (polySTs), ST8Sia-II and ST8Sia-IV, whose expression is transcriptionally regulated (6,7).
The most abundant and well-studied polysialylated protein is the neural cell adhesion molecule, NCAM. polyST and NCAM knock-out mice demonstrated that temporal control of NCAM adhesion and associated signaling is not only critical for proper brain formation but also for the maintenance and function of other parts of the nervous system (8 -11). Other studies demonstrate its role in developmental and regenerative processes in extraneural tissues like immune cells, lung, placenta, testis, and liver (1,12) and suggest its contribution to cancer cell invasion and metastasis (13).
The small number of polyST substrates, and the inefficient polysialylation of free glycans relative to those on proteins (26), suggested to us that polysialylation is a protein-specific modification that requires an initial protein-protein interaction between enzyme and substrate. Work in our labora-tory has abundantly supported this notion (1,3,27). We have shown that an acidic patch on the first fibronectin type III repeat of NCAM (FN1) is required for the recognition, binding, and polysialylation of two N-glycans in the adjacent Ig5 domain (28 -30). The same pattern holds for NRP-2, where a pair of acidic residues on the surface of the meprin-A5 antigen--tyrosine phosphatase (MAM) domain are essential for polysialylation of O-glycans in the adjacent linker region (31).
With these data in mind, we identified a conserved amino acid stretch in the polySTs that is enriched in basic residues that might act as a complementary binding region for the acidic patches in the NCAM FN1 domain and NRP-2 MAM domain. This polybasic region (PBR) is composed of residues 86 -120 in ST8Sia-II and residues 71-105 in ST8Sia-IV (32). We found that the ST8Sia-IV PBR was required for NCAM binding, and replacing PBR residues Arg 82 and Arg 93 substantially decreased NCAM polysialylation and blocked the ability of an inactive ST8Sia-IV enzyme (H331K) to serve as a competitive inhibitor of NCAM polysialylation by wild-type ST8Sia-IV (33). These results suggested that these basic residues are required for ST8Sia-IV recognition of NCAM (33). Recently, we confirmed this notion using biophysical techniques to demonstrate a direct interaction between the isolated PBR region and NCAM FN1 that depended upon PBR residues Arg 82 and Arg 93 and FN1 residues Asp 520 , Glu 521 , and Glu 523 (34).
In this work, to understand how the two polySTs differentially recognize their substrates, we have compared the polyST PBR residues required for polysialylation of NCAM, SynCAM 1, and NRP-2, as well as enzyme autopolysialylation. A competition assay is then used to determine whether loss of substrate polysialylation reflects the loss of polyST recognition. We find that distinct sets of overlapping PBR residues in each polyST are required for the recognition and polysialylation of NCAM, NRP-2, or SynCAM 1. We also provide evidence for the mechanism of polyST autopolysialylation and make a surprising observation that polyST autopolysialylation is required for NRP-2 polysialylation and promotes SynCAM 1 polySia chain elongation.

Results and discussion
Previous work established the role of the ST8Sia-IV PBR residues Arg 82 and Arg 93 in NCAM recognition and polysialylation (32)(33)(34). In addition, we found that Arg 82 plays a role in NRP-2 polysialylation but Arg 93 does not (33). In this work, our overall goal was to determine whether each polyST-substrate pair requires a unique combination of ST8Sia-IV or ST8Sia-II PBR basic residues and how those residues contribute to the polysialylation process.
To do this, each arginine or lysine residue within the ST8Sia-II and ST8Sia-IV PBR regions was individually mutated to alanine (Fig. 1, underlined). Additionally, all ST8Sia-II and ST8Sia-IV PBR alanine mutants, as well as these PBR mutants of ST8Sia-II H346K and ST8Sia-IV H331K, inactive enzymes that are used in competition assays (see below), were examined for proper cellular localization by individually expressing each Myc-tagged protein in COS-1 cells and staining the cells with anti-Myc antibody. All of the polyST mutants co-localize with the GM130 Golgi marker, indicating that these proteins are correctly folded, exit the endoplasmic reticulum, and are transported to the Golgi where they can fulfill their function (data not shown).
Previous attempts to use a co-immunoprecipitation approach to evaluate the role of specific PBR residues in substrate binding provided inconclusive results. This is most likely due to multiple interactions between the polySTs and their substrates and the inability to distinguish productive and non-productive interactions (35). To isolate those interactions that lead to protein-specific polysialylation, we have taken a competition approach in which PBR residues are replaced in a catalytically inactive polyST, ST8Sia-II H346K or ST8Sia-IV H331K, and those inactive mutants that exhibit a reduced ability to compete with the active enzyme to block substrate polysialylation identify PBR residues that are likely to play a role in substrate recognition (33). In other words, if competition is decreased or lost for an inactive PBR mutant, and polysialylation of the substrate is recovered, then that residue is likely to be part of the enzyme recognition site for that substrate or alternatively influence the structure of the enzyme recognition site. Here, we present the comparisons of the impact of replacing basic PBR residues in the wild-type and inactive polySTs for each substrate-polyST pair to define residues involved in substrate recognition and polysialylation.
Next, to evaluate whether the ST8Sia-IV PBR residues that contributed to NCAM and NRP-2 polysialylation also play roles A, left panels, V5-tagged NCAM was co-expressed with Myc-tagged ST8Sia-IV or its mutants in COS-1 cells. After 24 h, NCAM was recovered from cell lysates by immunoprecipitation and subjected to SDS-PAGE and immunoblotting with the 12F8 anti-polySia antibody to analyze the level of NCAM polysialylation (upper panel). Relative NCAM and ST8Sia-IV expression levels were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as % of wild-type enzyme polysialylation. B, left panels, V5-tagged NCAM was co-expressed in COS-1 cells with untagged wild-type ST8Sia-IV and Myc-tagged inactive ST8Sia-IV H331K or its PBR mutants in a ratio of 1:1:6. NCAM polysialylation was determined as described above (upper panel). The relative expression levels of NCAM and ST8Sia-IV H331K or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as fold recovery from competition with ST8Sia-IV H331K. A and B (right scatterplots), results from four different repeats were averaged, and standard deviation and significance were assessed using a one-way ANOVA test with a Dunnett's post hoc test, where **, 0.001 Ͻ p Ͻ 0.01; ***, 0.0001 Ͻ p Ͻ 0.001; ns, p Ͼ 0.05.

Requirements for polysialyltransferase substrate recognition
in substrate recognition, V5-tagged NCAM or NRP-2 were expressed with the untagged active polyST and Myc-tagged inactive ST8Sia-IV H331K and its PBR mutants in COS-1 cells at a ratio of 1:1:6 (substrate/wild-type ST8Sia-IV/ ST8Sia-IV competitor) (Figs. 2B and 3B). V5-tagged substrates were immunoprecipitated from cell lysates, and their polysialylation was assessed by immunoblotting with an anti-polySia antibody (Figs. 2B and 3B, upper panels). Relative expression levels of NCAM, NRP-2, as well as wild-type and competitor polyST expression were determined as described under "Experimental procedures" (Figs. 2B and 3B, middle and bottom panels). Rel-ative substrate polysialylation observed in the presence of ST8Sia-IV H331K and its PBR mutants was quantified using ImageJ software and reported as polysialylation recovery (fold recovery) observed with each PBR mutant as compared with the substrate polysialylation observed in the presence of the competitive inhibitor, ST8Sia-IV H331K.
In the case of NCAM, R82A and R93A mutations that significantly reduce its polysialylation also reduce the ability of the inactive H331K mutant to compete with the active enzyme, leading to an increase in NCAM polysialylation (fold recovery,   A, left panels, V5-tagged NRP-2 was co-expressed with Myc-tagged ST8Sia-IV or its mutants in COS-1 cells. After 24 h, NRP-2 was recovered from cell lysates by immunoprecipitation using an anti-V5 antibody and subjected to SDS-PAGE and immunoblotting with the 12F8 anti-polySia antibody to analyze the level of NRP-2 polysialylation (upper panel). Relative NRP-2 and ST8Sia-IV expression levels were determined described under "Experimental procedures" (middle and bottom panels). Data are reported as % of wild-type enzyme polysialylation. B, left panels, V5-tagged NRP-2 was co-expressed in COS-1 cells with untagged wild-type ST8Sia-IV and Myc-tagged inactive ST8Sia-IV H331K or its PBR mutants in a ratio of 1:1:6. NRP-2 polysialylation was determined as described above (upper panel). The relative expression levels of NRP-2 and ST8Sia-IV H331K or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as fold recovery from competition with ST8Sia-IV H331K. A and B, right scatterplots, results from four and three repeats, respectively, were averaged, and standard deviation and significance were assessed using a one-way ANOVA test with a Dunnett's post oc test, where *, 0.01 Ͻ p Ͻ 0.05; **, 0.001 Ͻ p Ͻ 0.01; ***, 0.0001 Ͻ p Ͻ 0.001; ns, p Ͼ 0.05.

Requirements for polysialyltransferase substrate recognition
and Table 2). This observation is in accordance with our recent data demonstrating that these residues are essential for the direct interaction between an isolated PBR peptide and a recombinant NCAM FN1 domain (34). Of the other PBR mutants that show a lesser impact on NCAM polysialylation (replacement of Lys 83 , Arg 87 , or Lys 103 ), only the K83A mutation in the ST8Sia-IV H331K protein led to a reduction in competition and a significant recovery of NCAM polysialylation (fold recovery, K83A ϭ 2.58 Ϯ 0.57 (S.D.)) ( Fig. 2B and Table 2).
These results demonstrate the common role of Arg 82 in the recognition and polysialylation of NCAM and NRP-2. However, recognition and polysialylation requirements then deviate. NCAM requires Arg 93 for recognition and polysialylation, whereas Lys 99 plays a key role in NRP-2 polysialylation but may only weakly contribute to recognition. The mismatched contribution of Lys 83 (NCAM) or Lys 83 and Arg 87 (NRP-2) to recognition and polysialylation of these substrates is difficult to explain, but it could reflect a change in local structure that impacts the availability of key interacting residues that is more readily detected in the competition assay. For example, any change in interaction surface of the inactive competitor, even a local structural change that slightly compromises access to key binding residues, would make it easier for the wild-type enzyme to replace the competitor and polysialylate the substrate. In contrast, the same mutant in the wild-type enzyme may have less impact on substrate polysialylation if the enzyme could "hold on" to the substrate using a partial recognition site.

Polysialylation of NCAM and SynCAM 1 by ST8Sia-II requires the contribution of a larger set of PBR basic residues
We took the same approaches described above to evaluate the recognition and polysialylation of NCAM and SynCAM 1 by ST8Sia-II. At first, we used membrane-anchored substrates but realized that ST8Sia-II was co-precipitating with NCAM and SynCAM 1, something we had not observed with ST8Sia-IV. Consequently, we were concerned that the inconsistent results we initially obtained related to how much autopolysia-lylated ST8Sia-II was co-precipitating with the polysialylated substrate. To address this challenge, we used Fc-tagged soluble forms of substrates, denoted henceforth as NCAM-Fc and SynCAM-Fc. We reasoned that because these soluble forms are polysialylated in the Golgi by the Golgi-localized ST8Sia-II and then secreted into the cell culture medium, we could recover them without lysing the cells and thereby avoid co-precipitation with membrane-anchored enzyme.
The Fc-tagged substrates were precipitated from the cell culture media using protein A-Sepharose beads, and their polysialylation was assessed by immunoblotting with an anti-polySia antibody (Figs. 4A and 5A, left upper panels). The relative expression levels of the soluble substrates and the membraneassociated polySTs were determined as described under "Experimental procedures" (Figs. 4A and 5A, left middle and bottom panels). We confirmed the absence of any co-precipitated ST8Sia-II following protein A-Sepharose precipitation of Fc-tagged substrates from the cell culture medium by performing an immunoblot using the anti-V5 antibody (data not shown).

Requirements for polysialyltransferase substrate recognition
To evaluate the role of ST8Sia-II PBR residues in the recognition of NCAM-Fc and SynCAM-Fc, we created the PBR mutants in the ST8Sia-II H346K catalytically inactive enzyme and co-expressed these and the wild-type enzyme with NCAM-Fc or SynCAM-Fc in COS-1 cells. The impact of these enzyme competitor mutants on competition and recovery of NCAM-Fc and SynCAM-Fc polysialylation was determined as described above. In the case of NCAM-Fc, replacing several PBR residues led to a loss of competition by the ST8Sia-II H346K mutant. Replacing ST8Sia-II Arg 97 and Lys 108 , which are homologous to the ST8Sia-IV Arg 82 and Arg 93 that are key for NCAM recognition and polysialylation by this polyST, led to the most loss of competition in the H346K background and maximum polysialylation recovery (fold recovery, 11.48 Ϯ 2.54 (S.D.) and 9.54 Ϯ 1.95 (S.D.)) ( Fig. 4B and Table 4). Replacing other ST8Sia-II PBR basic residues that also reduced NCAM polysialylation, such as Lys 102 , Lys 114 , and Lys 118 , also allowed polysialylation recovery in competition assays (fold recovery, K102A ϭ 5.83  Table 4). A, left panels, NCAM-Fc was co-expressed with V5-tagged ST8Sia-II or its PBR mutants in COS-1 cells. After 24 h, NCAM-Fc was recovered from cell medium using protein A-Sepharose beads and its polysialylation was assessed by SDS-PAGE and immunoblotting with the 12F8 anti-polySia antibody (upper panel). The relative expression levels of NCAM-Fc, ST8Sia-II, or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as % of wild-type enzyme polysialylation. B, left panels, NCAM-Fc was co-expressed in COS-1 cells with V5-tagged wild-type ST8Sia-II and Myc-tagged inactive ST8Sia-II H346K or its PBR mutants in a ratio of 1:1:6. NCAM-Fc polysialylation was determined as described above using the anti-polySia 735 antibody (upper panel). The relative expression levels of NCAM-Fc and ST8Sia-II H346K or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as fold recovery from competition with ST8Sia-II H346K. A and B, right scatterplots, results from six and five repeats, respectively, were averaged, and standard deviation and significance were assessed using a one-way ANOVA test with a Dunnett's post hoc test, where *, 0.01 Ͻ p Ͻ 0.05; ***, 0.0001 Ͻ p Ͻ 0.001; ns, p Ͼ 0.05.
In sum, ST8Sia-II recognition and polysialylation of NCAM-Fc require Arg 97 and Lys 108 that are homologous to Arg 82 and Arg 93 , the key residues for ST8Sia-IV substrate rec-ognition and polysialylation, but other residues, Lys 102 , Lys 114 , and Lys 118 , may also play a role in recognition. SynCAM-Fc polysialylation appears to be quite different with modest contributions by Arg 97 and Lys 108 , and stronger contributions by Lys 102 , Lys 114 , and Lys 118 . Strikingly, for SynCAM-Fc, only the H346K K102A and K114A mutants exhibit any significant loss of competition and recovery of polysialylation, suggesting these residues are involved in SynCAM 1 recognition. Based on their impact on NCAM-Fc and SynCAM-Fc polysialylation, we might have expected a larger recovery of polysialylation in the competition assays for the K114A and K118A mutants, and this A, left panels, SynCAM-Fc was co-expressed with V5-tagged ST8Sia-II or its PBR mutants in COS-1 cells. After 24 h, SynCAM-Fc was recovered from cell medium using protein A-Sepharose beads, and its polysialylation was assessed by SDS-PAGE and immunoblotting with the 12F8 anti-polySia antibody (upper panel). Relative expression levels of SynCAM-Fc, ST8Sia-II, or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as % of wild-type enzyme polysialylation. B, left panels, SynCAM-Fc was co-expressed in COS-1 cells with V5-tagged wild-type ST8Sia-II and Myc-tagged inactive ST8Sia-II H346K or its PBR mutants in a ratio of 1:1:6. SynCAM-Fc polysialylation was determined as described above (upper panel). The relative expression levels of SynCAM-Fc and ST8Sia-II H346K or its mutants were determined as described under "Experimental procedures" (middle and bottom panels). Data are reported as fold recovery from competition with ST8Sia-II H346K. A and B, right scatterplots, results from three repeats were averaged, and standard deviation and significance were assessed using a one-way ANOVA test with a Dunnett's post hoc test, where *, 0.01 Ͻ p Ͻ 0.05; **, 0.001 Ͻ p Ͻ 0.01; ***, 0.0001 Ͻ p Ͻ 0.001; ns, p Ͼ 0.05.

Requirements for polysialyltransferase substrate recognition
mismatch might reflect additional roles for these residues in the polysialylation process.

Potential roles of polyST PBR residues in substrate polysialylation
The loss of polysialylation versus loss of competition/gain of polysialylation numbers for some PBR mutants do not precisely match. In an effort compare the impact of these mutations for different polyST-substrate pairs, we set the highest fold recovery for each competition experiment to 10 and adjusted the other fold recovery numbers accordingly (see numbers in parentheses in Tables 2 and 4). Three categories were created (please see Fig. 6 legend for precise parameters for each category). Category 1 (Fig. 6, green) includes those residues that when replaced show a parallel loss of substrate polysialylation and competition suggesting that they play a role in substrate recognition. These include ST8Sia-IV Arg 82 and Arg 93 (NCAM), Arg 82 (NRP-2), and ST8Sia-II Arg 97 , Lys 108 , Lys 114 , and Lys 118 (NCAM), and Lys 102 and Lys 114 (SynCAM 1). Category 2 (Fig. 6, red) includes those residues that when replaced reduce competition but have a lesser impact on substrate polysialylation, suggesting that their replacement may impact local structure and the presentation of key recognition residues, an effect that may be more apparent in a competition assay. These include ST8Sia-IV Lys 83 (NCAM) and Lys 83 and Arg 87 (NRP-2). Finally, category 3 (Fig. 6, blue) includes those residues that when replaced led to a loss of substrate polysialylation that was not associated with an accompanying loss of competition, suggesting that they are involved in something besides substrate recognition. These include Lys 99 (NRP-2) and Lys 118 (SynCAM 1). Other residues such as ST8Sia-II Lys 102 (NCAM) and Arg 97 (SynCAM 1), which might be included in the green and blue groups, respectively, have intermediate impact on polysialylation and competition, making it difficult to categorize them.
What function could the residues in category 3 play? In addition to making contacts with substrates, basic residues in the PBR could serve as part of an extended basic surface that engages the growing polySia chain and promotes its polymerization and even the continued engagement of the substrate and enzyme once the protein-protein interaction no longer is possible (36 -38), or they could play structural roles and/or be essential for catalytic activity. We created a model structure of ST8Sia-IV using the SWISS-MODEL server based on the crystal structure of another ␣2,8-sialyltransferase (ST8Sia-III) solved by Volkers et al. (38 -42). This structure reveals the disposition of the PBR residues, their proximity to the two sites of ST8Sia-IV autopolysialylation, and to other basic residues of the polysialyltransferase domain (PSTD), identified by Nakata et al. (43), which together with the PBR could form a basic surface to promote polySia chain elongation (Fig. 7) (38). This model depicts the ST8Sia-IV PBR as a broken helix that has three parts. The first section is composed of Ser 75 -Ser 76 -Leu 77 -Val 78 -Leu 79 -Glu 80 -Ile 81 -Arg 82 -Lys 83 -Asn 84 -Ile 85 -Leu 86 -Arg 87 -Phe 88 , and is directly preceded by Asn 74 a site of ST8Sia-IV glycosylation and autopolysialylation (22). The second section of the PBR is composed of Ala 91 -Glu 92 -Arg 93 -Asp 94 -Val 95 . The third helical section of the PBR is composed of Lys 99 -Ser 100 ; however, the PBR region extends through Gly 105 . Interestingly, another site of ST8Sia-IV glycosylation and autopolysialylation (Asn 119 ) is found close to this turn (22). The proximity of the PBR to the two sites of ST8Sia-IV autopolysialylation suggested that replacing the PBR residues might impact enzyme autopolysialylation.

Specific PBR mutants have an effect on ST8Sia-IV and ST8Sia-II autopolysialylation
To evaluate the possibility that changes to the polyST PBRs may alter autopolysialylation, we individually expressed Myctagged ST8Sia-II and ST8Sia-IV PBR mutants in COS-1 cells, immunoprecipitated them using an anti-Myc antibody, and assessed their polysialylation by immunoblotting with an anti-polySia antibody (Fig. 8, A and B). For ST8Sia-IV, replacing Figure 6. Summary of the roles of polyST PBR residues in substrate recognition and polysialylation and polyST autopolysialylation. To make comparisons between competition assays that varied in their fold recovery of polysialylation, we set the highest fold recovery to 10 for each enzyme/substrate pair and adjusted the other numbers to that scale. We then grouped residues impacting polysialylation and/or competition into three groups. Residues colored green are those that when replaced led to substrate polysialylation of between 0 and 50% that seen with the wild-type enzyme and rated between 5 and 10 on the loss of competition/recovery of polysialylation scale. For these residues, loss of polysialylation matches the loss of substrate recognition. Residues colored red are those that when replaced led to substrate polysialylation of between 51 and 100% of that seen with the wild-type enzyme and rated between 5 and 10 on the loss of competition/recovery of polysialylation scale. For these residues, loss of recognition/competition was greater than observed loss in polysialylation. Residues colored blue are those that when replaced led to substrate polysialylation of between 0 and 50% that seen with the wild-type enzyme and rated between 0 and 4.99 on the loss of competition/recovery of polysialylation scale. For these residues loss of polysialylation was greater than the observed loss of recognition/competition. Those residues that impact autopolysialylation when replaced with alanines are indicated by a *.
How could these mutations be impacting polyST autopolysialylation? For Arg 82 and Lys 99 in ST8Sia-IV, their proximity to the two sites of autopolysialylation suggests that replacing these residues might cause a local structural change that hinders the enzyme N-glycan access to the active site, and thus its autopolysialylation. This is unlikely to be the case for Lys 114 and Lys 118 because ST8Sia-II autopolysialylation occurs on three N-glycans that are not close to these two residues (23).
Because the polyST PBR residues that are essential for enzyme autopolysialylation are also involved in substrate recognition and polysialylation, we considered what processes could be common to both types of polysialylation. Both autopolysialylation and substrate polysialylation would certainly require a catalytically active enzyme. This possibility could be definitively ruled out for the ST8Sia-IV K99A mutant that shows a dramatic decrease in NRP-2 polysialylation (12 Ϯ 12% (S.D.)), a smaller decrease in autopolysialylation (24 Ϯ 11% (S.D.)), and very efficient NCAM polysialylation (97 Ϯ 13% (S.D.)) ( Table 1). The ST8Sia-IV R82A mutant appears to equally impact autopolysialylation and NCAM polysialylation reducing these to 47 Ϯ 13% (S.D.) and 49 Ϯ 16% (S.D.) of that seen with the wild-type enzyme. However, this mutation has a much greater impact on NRP-2 polysialylation reducing it to 9 Ϯ 6% (S.D.) of that seen with the wild-type enzyme (Table 1).

Figure 8. Arg 82 and Lys 99 in the ST8Sia-IV PBR region and Lys 114 and Lys 118 in the ST8Sia-II PBR region are crucial for polyST autopolysialylation. A,
Myc-tagged ST8Sia-IV PBR mutants were expressed in COS-1 cells. After 24 h, ST8Sia-IV PBR mutants were recovered from cell lysates by immunoprecipitation using an anti-Myc antibody and subjected to SDS-PAGE and immunoblotting with the 12F8 anti-polySia antibody to analyze the level of ST8Sia-IV autopolysialylation (upper panel). An aliquot of cell lysate was boiled with Laemmli sample buffer to remove polySia, subjected to SDS-PAGE, followed by immunoblotting using anti-Myc antibody (lower panel). B, autopolysialylation of Myc-tagged ST8Sia-II and its PBR mutants was analyzed as described above. A and B, lower scatterplots, results from four repeats were averaged, and S.D. and significance were assessed using a one-way ANOVA test with a Dunnett's post hoc test, *, 0.01 Ͻ p Ͻ 0.05; ***, 0.0001Ͻ p Ͻ 0.001; ns, p Ͼ 0.05.

Requirements for polysialyltransferase substrate recognition
If there is a reduction in activity for the R82A mutant, it would only be ϳ50%. In contrast, the ST8Sia-II K114A and K118A mutants dramatically reduce SynCAM-Fc polysialylation to 3 Ϯ 5% (S.D.) and 2 Ϯ 3% (S.D.) of that seen with the wild-type enzyme, but have a smaller impact on NCAM-Fc polysialylation (42 Ϯ 18% (S.D.) and 31 Ϯ 18% (S.D.), respectively) ( Table  3). Their impact on autopolysialylation is mixed with the K114A mutant reducing ST8Sia-II autopolysialylation to 32 Ϯ 9% (S.D.) of that of the wild-type enzyme and the K118A mutant reducing ST8Sia-II autopolysialylation to 15 Ϯ 7% (S.D.) of that of the wild-type enzyme (Table 3). Again, these mutants are not completely inactive and at least retain 30 -40% of wild-type enzyme activity.
Two other processes might be required for both autopolysialylation and substrate polysialylation. First, autopolysialylation might require a protein-protein interaction like the one observed to initiate substrate polysialylation. For the autopolysialylation process, this would be an interaction between two enzyme monomers that would then polysialylate their partner's glycans. Second, both autopolysialylation and substrate polysialylation may require the same or overlapping basic surfaces to optimally elongate growing polySia chains to lengths recognized by most anti-polySia antibodies (usually Ն8 units) (1,2). It is important to note here that it is difficult to separate reduced catalytic activity, which would be involved in both chain initiation and elongation, from disruption of the elongation process due to lack of an adequate basic surface to stabilize the growing chain. To get insight into the roles of these PBR residues in substrate polysialylation and enzyme autopolysialylation, below we evaluate the mechanism of autopolysialylation and ask whether polyST autopolysialylation is required for NRP-2 and SynCAM 1 polysialylation.

Autopolysialylation of ST8Sia-IV appears to be self-and not cross-polysialylation
To evaluate whether ST8Sia-IV autopolysialylation requires one enzyme monomer to polysialylate a partner's N-glycans (cross-polysialylation), we used two non-autopolysialylated but active mutant enzymes, ST8Sia-IV mut2.3, in which the second and third glycosylation sites, Asn 74 and Asn 119 that carry the polysialylated glycans, are mutated to serine and glutamine, respectively (22), and ST8Sia-IV L151A that localizes to the Golgi but is also not autopolysialylated. 4 Both of these mutant enzymes are non-autopolysialylated but able to polysialylate NCAM (Fig. 9, A and C). We co-expressed V5-tagged mut2.3 and L151A with Myc-tagged inactive ST8Sia-IV H331K, immunoprecipitated the Myc-tagged protein from the cell lysates, and immunoblotted using an anti-polySia antibody to assess the ability of the active but non-autopolysialylated mutants to polysialylate the inactive polyST. We found that the non-autopolysialylated ST8Sia-IV mutants were not capable of polysialylating co-expressed ST8Sia-IV H331K (Fig. 9B). These results strongly suggest that autopolysialylation is a self-and not a cross-modification event.
So, despite the fact that both ST8Sia-IV Arg 82 and ST8Sia-II Lys 114 are involved in substrate recognition, these results sug-gest that the decreased autopolysialylation of the ST8Sia-IV R82A and ST8Sia-II K114A mutants is not due to reduction in the interaction of two enzyme monomers and may be more likely due to nearly equivalent reductions in catalytic activity/ chain elongation. The decreased autopolysialylation of the K118A mutant that decreases NCAM recognition but not Syn-CAM 1 recognition likely reflects an ϳ30% decrease in catalytic activity plus an additional ϳ15% decrease possibly due to a disruption of the basic surface required for polySia chain elongation. Finally, the reduction seen in K99A autopolysialylation may be explained by a local structural alteration that hampers polyST N-glycan polysialylation. Another possibility is that this residue is an essential part of the basic surface required for chain elongation. We think that this possibility is unlikely because the small proportion of the K99A mutant that is polysialylated migrates as very high molecular mass forms suggesting that chain elongation per se is not compromised (Fig. 8A).
What was particularly striking in our comparisons of the impact of ST8Sia-IV Arg 82 and Lys 99 and ST8Sia-II Lys 114 and Lys 118 on enzyme autopolysialylation and substrate polysialylation was how replacing these residues had a much more dramatic impact on NRP-2 and SynCAM 1 polysialylation than it did on NCAM polysialylation (Tables 1 and 3). We have previously demonstrated that polyST autopolysialylation was not required for NCAM polysialylation, but these current data suggest that changes in the autopolysialylation of the PBR mutants may be able to explain, at least in part, the decreases observed in NRP-2 and SynCAM 1 polysialylation by the same mutants.

ST8Sia-IV autopolysialylation is required for NRP-2 polysialylation and ST8Sia-II autopolysialylation promotes SynCAM 1 polySia chain elongation
To determine whether enzyme autopolysialylation is required for NRP-2 polysialylation, we employed two ST8Sia-IV mutants used above that are catalytically active but not autopolysialylated. We found that although ST8Sia-IV mut2.3 and L151A can polysialylate NCAM, they cannot polysialylate NRP-2 (Fig. 9C). Consequently, the R82A and K99A mutations that compromise ST8Sia-IV autopolysialylation would in turn be expected to reduce NRP-2 polysialylation. Does the presence of polySia chains on ST8Sia-IV in some way promote polySia chain elongation on NRP-2 to lengths recognized by the anti-polySia antibodies (1,2)? In support of this potential role for polyST polySia, we found that a non-autopolysialylated mutant of ST8Sia-II, mut2.4.5, could not generate a high molecular mass population of polysialylated SynCAM-Fc that was synthesized by the wild-type enzyme (Fig. 9D). As the fourth and fifth glycosylation/polysialylation sites (Asn 219 and Asn 234 ) are not near ST8Sia-II Lys 114 and Lys 118 in the linear sequence or in the predicted structure, it seems very unlikely that alterations in these PBR residues would structurally impact glycosylation of these sites or the availability of glycans for autopolysialylation (and in fact we see no evidence of the molecular mass change reflective of a loss of glycosylation at one or more sites). Instead, we favor the idea that Lys 114 and Lys 118 serve as part of a basic surface that is used for both substrate recognition as well as polySia chain elongation.
How would polyST autopolysialylation promote substrate polysialylation? One possibility is that the polySia chains on the polySTs stabilize or direct substrate interactions by blocking interactions with inappropriate basic surfaces on the enzyme and in this way direct acidic surfaces of substrate's recognition domain to bind to the polyST's PBR sequences. This would necessitate the enzyme's polySia chains to bind back to the surface of the enzyme without binding to the PBR and blocking substrate access.
Another possibility is that the polySia chains on the polyST could bind to a basic surface on the substrate to prevent this surface from binding the growing polySia chain and hindering its elongation. For example, the O-glycans on NRP-2 that are polysialylated reside in a linker region between the second Factor V/Factor VIII homology domain and the MAM domain that serves as the NRP-2 recognition domain. The Factor V/Factor VIII domain is highly basic with a predicted pI of 8.31 (45), and it could interact with the growing, negatively charged polySia chain on the NRP-2 O-glycans. Could the ST8Sia-IV polySia chains block the interaction of the growing polySia chains with this basic domain, thereby promoting chain polymerization? This clearly would not be the case for SynCAM 1 that is polysialylated on one N-glycan in its very N-terminal Ig domain (25). Notably, for NCAM polysialylation that is not Figure 9. polyST autopolysialylation is self-and not cross-polysialylated and is required for NRP-2 polysialylation and optimal SynCAM 1 polysialylation. A, V5-tagged ST8Sia-IV or non-autopolysialylated ST8Sia-IV mutants (mut2.3 and L151A) were expressed in COS-1 cells. B, Myc-tagged, inactive ST8Sia-IV H331K was co-expressed with V5-tagged mut2.3 or L151A catalytically active but non-autopolysialylated enzyme mutants in COS-1 cells. Myc-tagged ST8Sia-IV was expressed alone as an autopolysialylation control. C, V5-tagged ST8Sia-IV and its non-autopolysialylated mutants mut2.3 and L151A were co-expressed in COS-1 cells with Myc-tagged NCAM or NRP-2. D, Fc-tagged NCAM and SynCAM 1 were co-expressed separately with V5-tagged ST8Sia-II or its non-autopolysialylated mutant (ST8Sia-II mut2.4.5) in COS-1 cells. For all panels, protein polysialylation and relative protein expression levels were assessed by immunoblotting with either the anti-polySia 12F8 antibody (A-C) or the anti-polySia 735 antibody (D) and appropriate anti-tag antibodies as described under "Experimental procedures." All experiments were performed at least three times, but quantitation was not performed because under the conditions tested there was no polysialylation (A-C) or the absence of a specific polysialylated form (D).

Requirements for polysialyltransferase substrate recognition
impacted by polyST autopolysialylation, the domain adjacent and N-terminal to the polysialylated Ig5 domain is Ig4 and it has a more acidic pI of 5.2 (45) than the analogous NRP-2 domain.
Another attractive possibility is that the polySia chains on the polySTs may stabilize or promote the elongation of polySia chains on the NRP-2 or SynCAM 1 substrates via direct carbohydrate-carbohydrate interactions. Although it might seem counterintuitive for two highly negatively charged chains to interact, using atomic force microscopy, Finne and co-workers (46) demonstrated that oligomers of polySia with 12 or more sialic acid units can assemble into filament networks. It is interesting to note that both NRP-2 and SynCAM 1 appear to engage fewer ST8Sia-IV and ST8Sia-II PBR residues than NCAM, and this observation begs the question of whether for these substrates an interaction with the polyST polySia chains provides that additional and, in the case of NRP-2, necessary stability required to maintain contact and achieve optimal polysialylation.
In summary, we have identified basic amino acids in the PBR sequences of ST8Sia-II and ST8Sia-IV that are required for substrate polysialylation. We note that these requirements vary for each substrate but do include common residues. Of the PBR residues required for substrate polysialylation, some are involved exclusively in substrate recognition. These include ST8Sia-IV Arg 93 (NCAM), ST8Sia-II Arg 97 , Lys 102 , and Lys 108 (NCAM), and Lys 102 (SynCAM 1). Other PBR residues play roles in both substrate recognition and autopolysialylation, a process that our data suggest is a true self-polysialylation and does not involve interaction between two enzyme monomers. These residues may serve two functions. First as contact points in the protein-protein interaction between substrate and polyST, and second as part of a basic surface on the polyST that engages growing polySia chains to stabilize them and promote elongation. This latter function would be shared by both substrate polysialylation and polyST autopolysialylation. Residues in this category include ST8Sia-IV Arg 82 (NCAM and NRP-2), ST8Sia-II Lys 114 and Lys 118 (NCAM), and ST8Sia-II Lys 114 (SynCAM 1). For SynCAM 1 polysialylation, Lys 118 may be solely involved in polySia chain elongation. Finally, we found for the first time that NRP-2 polysialylation by ST8Sia-IV requires that this polyST be autopolysialylated and that the autopolysialylation of ST8Sia-II promotes the polymerization of longer chains on SynCAM 1. From this, we can conclude that the impact of replacing PBR residues that are key for enzyme autopolysialylation (ST8Sia-IV Arg 82 and Lys 99 ; ST8Sia-IV Lys 114 and Lys 118 ) on NRP-2 or SynCAM 1 polysialylation may at least in part reflect a loss of enzyme autopolysialylation that in turn compromises these substrates' polysialylation. The one case where a loss of polyST autopolysialylation may be the predominant factor in a loss of substrate polysialylation is the ST8Sia-IV K99A mutant, which dramatically decreases enzyme autopolysialylation to 24% that of the wild-type enzyme, and NRP-2 polysialylation to 12% that of wild-type enzyme, without substantially impacting NCAM polysialylation.
Our data and that of others (36 -38, 43) allow us to propose a general model for substrate polysialylation. We predict that polysialylation is a two-step process. The first step consists of an initial protein-protein interaction between polyST and substrate that involves specific residues in the PBR of the polyST and a complementary acidic region of the enzyme. The second step occurs when the polySia chain is long enough to prohibit the protein-protein interaction and may shift the engagement to a protein-glycan interaction or a glycan-glycan interaction between polyST and substrate. For example, in the case of NCAM polysialylation by ST8Sia-II, Arg 97 , Lys 108 , Lys 114 , and Lys 118 may all play roles in the initial protein-protein interaction, whereas Lys 114 and Lys 118 also take part in the second step to engage the growing polySia chain on NCAM. This step is also predicted to include basic residues in the PSTD region of the enzyme (38,43). In the case of NRP-2 polysialylation by ST8Sia-IV, Arg 82 is essential for the polyST-substrate protein-protein interaction, and although Arg 82 and Lys 99 may be involved in promoting polySia chain elongation on this substrate as part of a basic surface on the polyST, we propose that the polySia chains on the enzyme itself may directly impact continued substrate engagement and polySia chain polymerization. Future work will be directed at testing these models.

Construction of ST8Sia-IV-Myc and ST8Sia-II-Myc PBR mutants
The ST8Sia-II and ST8Sia-IV cDNAs were cloned into the EcoRV and XbaI sites of a previously digested pcDNA3.1/ Myc-HisB mammalian expression vector, containing a C-terminal Myc epitope tag and a stop codon prior to the His 6 tag. The resulting plasmids containing the DNA sequences for ST8Sia-II-Myc or ST8Sia-IV-Myc served as the templates for the creation of all PBR mutants, in which each arginine or lysine residue within the PBR was individually replaced with alanine. All ST8Sia-IV PBR mutants were created as described previously by Foley et al. (32). All ST8Sia-II PBR mutants were created using the QuikChange TM site-directed mutagenesis kit and Pfu DNA polymerase according to manufacturer's instructions using the oligonucleotide primers listed in Table 5. To confirm the presence of the desired ST8Sia-II PBR mutations, the resultant DNAs were sequenced using the DNA Sequencing Facility at the Research Resources Center at the University of Illinois at Chicago.

Construction of ST8Sia-II-Myc H331K and ST8Sia-IV-Myc H346K PBR mutants
Catalytically inactive, full-length ST8Sia-IV-Myc H331K, ST8Sia-II-Myc H346K, and associated PBR mutants were created using the oligonucleotide primers listed in Table 5. Wildtype ST8Sia-II and ST8Sia-IV, as well as the catalytically active ST8Sia-II and ST8Sia-IV PBR mutants above, served as the templates for the introduction of either the H331K (ST8Sia-IV and ST8Sia-IV PBR mutants) or H346K (ST8Sia-II and ST8Sia-II PBR mutants) point mutations. All H331K and H346K mutations were created and confirmed as described above.

Construction of the pcDNA4-SynCAM 1-Fc construct
Extracellular portion of SynCAM 1 was amplified with primers that included a 15-bp overhang on each side with sequence corresponding to the pcDNA4 vector on the N terminus and Fc fragment on the C terminus (Table 5). pcDNA4-Fc vector was amplified separately using primers indicated in Table 5. The PCR-amplified products were mixed, and homologous recombination was carried out to create the pcDNA4-SynCAM 1-Fc construct using Clontech In-Fusion HD cloning kit, according to the manufacturer's protocol.

Immunofluorescence analysis of the intracellular localization of polyST mutants in COS-1 cells
COS-1 cells maintained in Dulbecco's modified Eagle's medium (DMEM), 10% fetal bovine serum (FBS) were plated onto 12-mm glass coverslips and grown in a 37°C, 5% CO 2 cell incubator until 50 -70% confluent. Cells on each coverslip were transfected with 0.5 g of Myc-tagged ST8Sia-II, ST8Sia-IV, ST8Sia-IV H331K, and ST8Sia-II H346K or associated PBR mutant cDNA in 300 l of Opti-MEM I supplemented with 3 l of Lipofectin transfection reagent. Cells were incubated with the transfection mixture for 6 h at 37°C, followed by the addition of 1 ml of DMEM, 10% FBS, and allowed to grow in the ST8Sia-IV and its PBR mutants 5Ј-CCAATGCAAGCCCTAAGAGAATGCCATTAG-3Ј 5Ј-CTAATGGCATTCTCTTAGGGCTTGCATTGG-3Ј 10 H346K ST8Sia-II and its PBR mutants 5Ј-CAGGCCAGCCCGAAGACCATGCCCTTG-3Ј 5Ј-CAAGGGCATGGTCTTCGGGCTGGCCTG-3Ј 11 Amplification of pcDNA4-Fc vector for in-fusion pcDNA4-NRP-2-Fc 5Ј-ACAGGTAAGTGGAGGGAGGGTG-3Ј 5Ј-GCCAGCTTGGGTCTCCCTATAG-3Ј 12 Amplification of ectodomains to make SynCAM-Fc construct Requirements for polysialyltransferase substrate recognition incubator for 18 h. Coverslips were washed three times with phosphate-buffered saline (PBS) then permeabilized and fixed with Ϫ20°C methanol to view internal structures. The cells were then blocked at room temperature in immunofluorescence blocking buffer (5% normal goat serum in PBS) for 1 h, followed by incubation at room temperature for 2 h with a rabbit anti-Myc epitope tag antibody and a mouse anti-GM130 antibody diluted 1:250 in immunofluorescence blocking buffer. After incubation with the primary antibodies, the cells were washed twice with PBS, incubated for 1 h with both a FITCconjugated goat anti-rabbit IgG secondary antibody and a TRITC-conjugated goat anti-mouse antibody each diluted 1:100 in blocking buffer, washed four more times with PBS, and then treated for 5 min with 300 nM DAPI diluted in PBS. Following a final wash with PBS, the coverslips were mounted on glass microscope slides using 20 l of mounting medium (15% Vinol 205 polyvinyl alcohol (w/v), 33% glycerol (v/v), 0.1% azide, pH 8.5), and the cells were visualized using a Zeiss Axiovert 200 M inverted confocal microscope using a ϫ40 water immersion objective.

Recovery of NCAM, NRP-2, and the polySTs from cell lysates or cell medium
Proteins were immunoprecipitated as described previously (31). For the NCAM, NRP-2, and autopolysialylation experiments, 24 h post-transfection, cells were washed twice with PBS and lysed in 200 l of immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 0.1% SDS). Cell lysates were pre-cleared with 25 l of protein A-Sepharose beads (50% suspension in PBS) for 1 h at 4°C with rotation. V5-tagged NCAM and NRP-2 and Myc-tagged polySTs were immunoprecipitated by incubating cell lysates with 1 l of anti-V5 tag antibody or 1 l of anti-Myc antibody overnight at 4°C with rotation. Samples were then rotated with 25 l of protein A-Sepharose beads for 1 h at 4°C. The beads were washed four times with immunoprecipitation buffer and once with immunoprecipitation buffer containing 1% SDS.
Samples were then resuspended in 25 l of Laemmli sample buffer containing 10% ␤-mercaptoethanol, heated for 8 min at 65°C. To determine the relative NCAM, NRP-2, and polyST protein expression levels, 20 l of cell lysate was removed prior to immunoprecipitation, mixed with 20 l of Laemmli sample buffer containing 10% ␤-mercaptoethanol, and heated for 8 min at 100°C to remove polySia. Fc-tagged NCAM or SynCAM 1 were precipitated from the cell medium using protein A-Sepharose beads, as described previously (31). To determine relative protein expression levels, 25% of these beads were resuspended in Laemmli buffer and heated to 100°C to release proteins and remove polySia, as described previously (31). All protein samples were separated on 4 -15% Mini-PROTEAN TGX gels.

Immunoblot analysis of protein expression and polysialylation
Following gel electrophoresis, the proteins were transferred to nitrocellulose membranes at 100 V for 2 h. Membranes were blocked for 1 h at 4°C in blocking buffer (5% nonfat dry milk in Tris-buffered saline, pH 8.0, 0.1% Tween 20). To detect NCAM, SynCAM 1, or NRP-2 polysialylation, nitrocellulose membranes were incubated overnight at 4°C with 12F8 anti-polySia antibody diluted 1:1000 in 2% nonfat dry milk in Tris-buffered saline, pH 8.0 (1:500 dilution of 12F8 anti-polySia antibody was used to detect SynCAM-Fc polysialylation). Anti-polySia 735 antibody was diluted to 1:2000 in 5% nonfat dry milk in Trisbuffered saline, pH 8.0, to detect NCAM and NRP-2 polysialylation and 1:1000 for SynCAM-Fc polysialylation. Expression of V5-tagged proteins was detected using a 1:10,000 dilution of anti-V5 antibody, and Myc-tagged proteins were detected using a 1:2500 dilution of anti-Myc antibody. Following overnight incubation with primary antibodies, membranes were then incubated for 1 h at 4°C with HRP-conjugated goat antirat IgM (12F8-treated membranes) or HRP-conjugated goat anti-mouse IgG (anti-V5 or anti-Myc or anti-polySia 735 antibody-treated membranes) secondary antibodies diluted 1:5000 in blocking buffer. Fc-tagged proteins were detected with HRPconjugated anti-human IgG (HϩL) diluted 1:2500 in blocking buffer containing high salt (500 mM NaCl, 150 mM Tris-HCl, pH 8.0, and 0.1% Tween 20) for 45 min. Membranes were washed four times before and four times after secondary antibody incubation with Tris-buffered saline, pH 8.0, 0.1% Tween 20 (TBST) for 10 min per wash with agitation. Membranes blotted with anti-human IgG were washed with TBST containing high salt as described previously (31). Immunoblots were developed using the Clarity ECL Western blotting substrate and HyBlot CL autoradiography films. Please note the reactivity of the anti-polySia antibodies 12F8 and 735 has previously been established by ourselves and others by demonstrating that the epitope recognized by these antibodies is destroyed by the polySia-specific bacteriophage enzyme endoneuraminidase N (47).