Identification of sites in domain I of perlecan that regulate heparan sulfate synthesis.

Perlecan is primarily a heparan sulfate containing proteoglycan found in all basement membranes. Rotary shadowed images of perlecan show it to contain three glycosaminoglycan (GAG) side chains extending from one end of its core protein. Domain I is at the N terminus of perlecan and contains three closely spaced Ser-Gly-Asp sequences that may serve in GAG attachment. We evaluated the serines in these three sequences for GAG attachment by preparing a cDNA construct encoding for the N-terminal half (domains I, II, and III) of perlecan and then a series of constructs containing deletions and mutations within domain I of the domain I/II/III construct, expressing these constructs in COS-7 cells, and then analyzing the recombinant product for GAG side chains and GAG type. The results showed that all three serine residues in the Ser-Gly-Asp sequences in domain I can accept both chondroitin and heparan sulfate side chains but that a cluster of acidic residues N-terminal to these sequences is the primary determinant responsible for targeting these sites for heparan sulfate. Furthermore, there are two elements that can enhance heparan sulfate synthesis at a targeted site: 1) the presence of a the SEA module in the C-terminal region of domain I and 2) the presence of multiple acceptors in close proximity. These results indicate that the proportion of heparan and chondroitin sulfate at any one site in domain I of perlecan is regulated by multiple factors.

Perlecan is a proteoglycan that is present in all basement membranes (1,2). The amino acid sequence of its core protein was initially deduced from cDNA clones to murine perlecan (3). It consists of five distinct domains with homology to agrin, low density lipoprotein receptor, the ␣1 chain of laminin, and N-CAM (3)(4)(5). The murine perlecan originally cloned has a ϳ400-kDa core protein (3). Alternate splicing of perlecan has been shown to occur in human (6,7), mouse (8), and nematode (9).
Perlecan was initially characterized in the Engelbreth-Holm-Swarm tumor as exclusively containing heparan sulfate side chains (10), but recent studies indicate that a portion of the perlecan produced by a cell line derived from this tumor is made as a chondroitin sulfate/heparan sulfate hybrid (11). Although heparan and chondroitin sulfate are structurally and functionally different, they are both O-linked via a galactosegalactose-xylose linkage region to serine residues in the core protein of the proteoglycan (12)(13)(14). Not all serine residues in the core protein, however, are substituted with a glycosaminoglycan (GAG) 1 chain. Testing synthetic polypeptides for the ability to accept xylose, the first reside in the linkage region added to the core protein, indicates that serine residues followed by glycine residues serve best as acceptors for xylosyltransferase (15) and that this acceptor activity could be enhanced by the presence of acidic residues preceding the Ser-Gly sequence (16). Analysis of sequence in the GAG attachment regions of aggrecan, however, showed that Ser-Gly sequences followed immediately by an acid residue occurred more frequently and it is likely these could also accept xylose (17).
Recent attempts to identify core protein features that designate a Ser-Gly site for heparan sulfate rather than chondroitin sulfate priming have utilized expressing vector systems in eukaryotic cells. Short sequences of cDNA derived primarily from syndecan and betaglycan were ligated to a cDNA for protein A, the recombinant product expressed in Chinese hamster ovary cells, and the product analyzed for chondroitin and heparan sulfate content (18,19). The results show that either deleting a tryptophan immediately following a Ser-Gly, moving a cluster of acid residues (normally 7, 8, and 10 residues C-terminal to the serine) closer to the serine or moving two adjacent Ser-Gly sequences apart will reduce heparan sulfate synthesis.
The serine residues that accept GAG chains and the core protein features that designate GAG type have not yet been identified in perlecan. Rotary shadowed images of perlecan show it to contain up to three GAG side chains extended from one end of its core protein (20,21), which would correspond to domain I or V. In both human and murine perlecan, domain I has three closely spaced Ser-Gly sequences, but they are not adjacent to each other and they are followed by aspartic acid, not tryptophan. Perlecan does have a cluster of acidic amino acids 7, 8, 9, and 10 residues away from the serine in the first Ser-Gly-Asp sequence, but the cluster is N-terminal to the serine, not C-terminal. Recent studies show expression of a cDNA construct consisting of domain I (22) as well as expression of a cDNA construct coding for protein A fused to a 25amino acid sequence containing the three Ser-Gly-Asp sequences from domain I (19), produced recombinant products bearing both heparan and chondroitin sulfate side chains. Consequently, in the present study, we examined the serines each of these sequences in domain I for GAG priming activity and identified the features in domain I that designate the GAG type primed at these serines by expressing constructs containing domain I in eukaryotic cells and then analyzing the recombinant product for the type of GAG side chains.

MATERIALS AND METHODS
Preparation of cDNA Constructs-cDNA constructs encoding for perlecan were prepared from cDNA clones 16 and 54 (3) and a cDNA clone encoding domain III (23) in pBluescript II SK ϩ (Stratagene) by using restriction enzyme sites in the multicloning region of the vector and the inserts. HindIII linkers (New England Biolabs) were used to add HindIII sites at different locations but in the same reading frame to provide for interchangeability of some constructs and facilitate construction. Short cDNA segments (35-65 base pairs), used to mutagenize specific amino acids in perlecan's sequence, were prepared by synthesizing (Oligos, Etc.) upper and lower strands that, when annealed together, would produce a cDNA encoding for the required amino acid sequence and containing the desired overhand at the ends for in-frame ligation into unique restriction sites in the insert. NotI and XbaI sites at the 5Ј and 3Ј ends, respectively, of construct were used for ligation into the multicloning site of the pRc/CMV expression vector.
All constructs contained 31 bases of untranslated sequence 5Ј to the methionine start codon of perlecan's signal peptide to initiate protein synthesis and sequence encoding perlecan's signal peptide to target the product for secretion. Designating the start methionine in the signal peptide as amino acid 1, the domain I construct (D I) encoded amino acids 1-195, the domain I/II construct (D II/III) encoded amino acids 1-484, and the domain I/II/III construct (D I/II/III) encoded amino acids 1-1680 ( Fig. 1). A previously (23) prepared 140 base pairs of cDNA encoding perlecan's signal peptide followed by the first 14 amino acids of domain I and containing a HindIII site at its 3Ј end for in-frame ligation was used to add a signal peptide to domain II. The domain II/III construct encoded for amino acids 1-35 and 195-1680 with the addition of four amino acids (Asp-Pro-Ser-Leu) between amino acids 35 and 195 because of the construction of the HindIII sites. A number of constructs were made with a deletion or with a deletion and mutations in domain I but containing all of domains II and III. The ALT 1 construct encoded amino acids 1-84 and 195-1680 with the addition of Ser-Leu between amino acids 88 and 195. This construct deleted 110 amino acids from the C terminus of domain I but retained the signal peptide and 63 amino acids of N-terminal sequence of domain I that contain the three Ser-Gly-Asp sequences (Fig. 1). The ALT 2 construct was identical to ALT 1, except that the serines in the three Ser-Gly-Asp sequences at amino acids 65, 71, and 76 were mutated to threonine (Fig. 1). The SER 1, SER 2, and SER 3 constructs were also identical to the ALT 1 construct except that the serines at residues 71 and 76 were mutated to threonine for SER 1, the serines at residues 65 and 76 were mutated to threonine for SER 2, and the serines at residues 65 and 71 were mutated to threonine for SER 3. The ALT 3 construct encoded amino acids 1-35 and 52-1680 with Asp-Pro-Ser-Leu between amino acids 35 and 52. This construct deleted 16 amino acids near the N terminus of domain I but retained sequence in domain I that contained the three Ser-Gly-Asp sequences and the cluster of six acidic amino acids Nterminal to the SGD sequences (Fig. 1). The ALT 4 construct was identical to the ALT 3 construct except that the six acidic amino acids at residues 55-58, 62, and 63 were mutated to polar amino acids (Fig.  1). The ALT 5, 6, and 7 constructs were also identical to the ALT 3 construct except that they contained either 2, 4, or 6 additional amino acids (selected from residues 58 -63 of ALT 4) inserted between amino acids 63 and 64 of the ALT 3 construct (Fig. 1).
Transfection for Western Blot Analysis-COS-7 cells cultured in Dulbecco's modified Eagle's medium containing high glucose, 10% fetal bovine serum, and 1.85 g of sodium bicarbonate/liter were used for transfection. Constructs in the pRc/CMV vector were transfected into COS-7 cells using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions, except that the cells were confluent at the time of transfection. COS-7 cells in 35-mm dishes were transfected with 2 g of DNA/12 l of LipofectAMINE/1.4 ml of serum-free medium for 6 h. The transfection solution was removed, and cells were cultured overnight in regular medium. The transfected cells were rinsed twice in serum-free medium and then cultured for 48 h in 2 ml of serum-free medium. The serum-free medium was removed from the dish, any debris removed by low speed centrifugation, and then dialyzed against distilled water (DW) overnight. The medium was lyophilized and reconstituted in 50 l of DW. 10-l aliquots, with and without digestion by chondroitinase ABC and heparatinase combined, were subjected to SDS-polyacrylamide gel electrophoresis in 7.5% acrylamide gels and transferred to nitrocellulose for Western blot analysis using rabbit antiserum against perlecan (24). Western blots were done from two separate transfections for each construct. The results were similar, and only one set is shown. In some cases RNA was isolated from transfected cells for analysis by Northern blot (23). The D I/II/III construct encoded for domains I, II, and III. The D II/III construct encoded for domains II and III. The ALT 1 construct encoded for the N-terminal one third of domain I, which contains the three serines (asterisks) at residues 65, 71, and 76 used in GAG attachment and all of domains II and III. The ALT 2 construct encoded for the same sequence as the ALT 1 construct except the serines at residues 65, 71, and 76 were mutated to threonines (underlined). The SER constructs also encoded for the same sequence as the ALT 1 construct except that in SER 1 the serines at 71 and 76 were mutated to threonines (underlined), in SER 2 the serines at 65 and 76 were mutated to threonines (underlined), and in SER 3 the serines at 65 and 71 were mutated to threonines (underlined). The ALT 3 construct encoded for all but 16 amino acids in the N-terminal region of domain I and all of domains II and III. The ALT 4 construct encoded for the same sequence as ALT 3 except that the acidic amino acids at positions 55-58, 62, and 63 were mutated to asparagine and glutamine (underlined). ALT 5 construct is the same as ALT 3 except for the addition of two asparagines between residues 63 and 64 in ALT 3. ALT 6 construct is the same as ALT 3 except for the addition of four amino acids (leucine-alanine-asparagineasparagine) between residues 63 and 64 in ALT 3. ALT 7 construct is the same as ALT 3 except for the addition of 6 amino acids (asparagineleucine-leucine-alanine-asparagine-asparagine) between residues 63 and 64 in ALT 3. The added amino acids are in bold type.
Transfection for 35 SO 4 Radiolabeling and GAG Characterization-COS-7 cells in T-75 flasks were transfected with 20 g of DNA/120 l of LipofectAMINE/14 ml of serum-free medium for 6 h. After overnight culture in regular medium, the cells are cultured for 40 h in 12 ml of regular medium containing 600 Ci of 35 SO 4 to radiolabel the GAG side chains on the recombinant product. The radiolabeling medium was removed from the flask and any debris removed by low speed centrifugation. The radiolabeled recombinant protein was captured by overnight mixing at 4°C with 125 l of packed protein G-Sepharose beads (Pharmacia Biotech Inc.) that had been previously absorbed with 100 l of rabbit antiserum to perlecan (25). The material bound to the beads was eluted with 2.0 ml of buffered (0.05 M Tris, pH 6.8) 4 M guanidine HCl at 50°C for 1 h, and any remaining unincorporated 35 SO 4 was removed by chromatography on PD-10 columns (Pharmacia) in buffered 4 M guanidine HCl. The macromolecular material from the PD-10 column was then fractionated on Superose 6 in buffered 4 M guanidine. Aliquots of each tube were taken for measurement of radioactivity by liquid scintillation counting. Each construct was used in three separate transfections and 35 SO 4 radiolabeling. The results were similar, and the Superose 6 profile was only one transfection for each construct shown.
Fractions from Superose 6 chromatography containing radioactivity were pooled, dialyzed against DW, lyophilized, and reconstituted in ϳ1 ml of DW. GAG side chains were released by treatment with 1 M sodium borohydride, 0.05 M NaOH at 45°C for 40 h. The proportion of heparan and chondroitin sulfate was determined by digestion with nitrous acid or chondroitinase ABC, followed by precipitation of the macromolecular material (undigested) with 2 volumes of 95% ethanol containing 1% potassium acetate and measuring digested material in the supernatant by liquid scintillation counting. The proportion was also determined by chromatography of undigested and chondroitinase-digested material on Superose 6 and calculating the proportion of radioactivity shifting to low molecular weight (digested).
A construct encoding for domain I was ligated in-frame to maltosebinding protein in the pMAL C2 vector (New England Biolabs) and expressed according to the manufacturer's instructions. Four constructs (domain I/II/III, domain I/II, domain II/III, and domain I) were also transfected into HT 1080 cells, and stably transfected cell lines were selected and analyzed for recombinant product.

RESULTS
The constructs were transfected into COS-7 cells and the media examined for expression of recombinant protein by SDSpolyacrylamide gel electrophoresis followed by Western blot using antiserum to native perlecan. The products of the domain I construct and the domain I/II construct could not be consistently or clearly detected in either COS-7 cells or HT 1080 cells,, although mRNA for these constructs could be detected in Northern blots of RNA from transiently transfected cells (data not shown). A derivative of the domain I construct did, however, express in prokaryotic cells as a fusion protein with maltose-binding protein, and the antiserum to perlecan readily reacted with recombinant product (data not shown). This indicates that the inability to detect the recombinant product of the domain I and domain I/II constructs in eukaryotic cells was not due to the lack of antibodies to these domains in the antiserum to native perlecan. These constructs were considered to be poorly expressed in eukaryotic cells, and their products were not further characterized. COS-7 cells transfected with the domain I/II/III construct did produce a immunoreactive recombinant product that appeared as a sharp band (arrowhead) just above the 213-kDa marker and a prominent broad band of higher molecular mass (Fig. 2,  lane 1). Digestion with heparatinase and chondroitinase prior to electrophoresis shifted the migration position of the band broad to that of the sharp band just above the 213-kDa marker (Fig. 2, lane 2). This shift to a lower molecular mass upon digestion indicates the broad band product of the domain I/II/ III construct is a proteoglycan. The sharp band (arrowhead, lane 1) did not shift upon digestion (lane 2). This indicates that a portion of the product of the domain I/II/III construct lacks GAG side chains. Digestion also revealed a sharp band or doublet of high molecular mass (vertical bar, lane 2). This band is also produced by digestion of medium from cells transfected with the pRc/CMV vector lacking any insert (Fig. 2, lane 5) and is likely the native perlecan core protein (estimated at 400 kDa) produced by the COS-7 cells. The product of the domain II/III construct was detected as a band at 213 kDa (Fig. 2, lane 3) that did not change in appearance when digested with heparatinase and chondroitinase (Fig. 2, lane 4). HT 1080 cells permanently transfected with the domain II/III construct produced a recombinant product that behaved similarly (data not shown). This indicates that product of the domain II/III construct lacks GAG side chains. These data indicate domain I can accept GAG side chains.
The presence or absence of GAG side chains on recombinant products was also determined by biosynthetically radiolabeling transfected COS-7 cells with 35 SO 4 , immunoprecipitating the recombinant protein from the medium with antiserum to perlecan, and then chromatographing the solubilized material on a column of Superose 6. Chromatography showed the product of domain I/II/III construct was radiolabeled with 35 SO 4 (Fig. 3A). Cells transfected with the domain II/III construct, however, produced the same levels of immunoprecipitated 35 SO 4 -radiolabeled product as cells transfected with the pRc/CMV vector lacking an insert (Fig. 3A). This confirms the Western blot data obtained for these constructs (Fig. 2). Characterization of the GAG type on the 35 SO 4 -radiolabeled products of the domain I/II/III construct showed it contained 73-81% heparan sulfate (Table I). HT 1080 cells permanently transfected with the domain I/II/III construct produced a recombinant product containing 71-80% heparan sulfate.
Domain I/II/III constructs with deletions and mutations of amino acids within domain I were prepared to identify the sites in domain I that are involved in GAG attachment. The first of these constructs (ALT 1) contained a deletion of approximately 2/3 of the C terminus of domain (Fig. 1). Transfection of COS-7 cells with the ALT 1 construct produced an immunoreactive product as a sharp band at the 213-kDa marker and a prominent broad band of higher molecular mass (Fig. 4, lane 1). Digestion with heparatinase and chondroitinase shifted the migration position to coincide with the band at 213 kDa (Fig. 4,  lane 2). This indicates that a substantial amount of the ALT 1 product was produced as a proteoglycan. Another construct (ALT 2) was prepared that was identical to the ALT 1 construct except that the serines at amino acids 65, 71, and 76, which are in the three Ser-Gly-Asp sequences in domain I, were mutated to threonines (Fig. 1). Transfection of COS-7 cells with the ALT 2 construct produced an immunoreactive product as a prominent band at 213 kDa (Fig. 4, lane 3), and digestion with heparatinase and chondroitinase did not change the migration position of the product (Fig. 4, lane 4) but did reveal the high molecular weight core protein the native perlecan produced by the COS cells as seen in lane 5. This indicates that the serines in the three Ser-Gly-Asp sequences in domain I accept GAG chains.
The presence or absence of GAG side chains on the recombinant products of the ALT 1 and ALT 2 constructs was also determined by metabolic radiolabeling with 35 SO 4 , immunoprecipitation, and chromatography on Superose 6. The product of the ALT 1 construct contained incorporated 35 SO 4 , but transfection with the ALT 2 construct resulted in levels of incorporated 35 SO 4 similar to that produced by transfection with the pRc/CMV vector lacking insert (Fig. 3B). These data confirm the conclusions made from the results of the Western blot experiments (Fig. 4) with the ALT 1 and ALT 2 constructs. Characterization of the GAG type on the 35 SO 4 -radiolabeled product of the ALT 1 construct indicates it contained 61-62% heparan sulfate (Table I).
The next set of constructs focused on the serine residues in each of the Ser-Gly-Asp sequences, individually. The SER constructs were identical to the ALT 1 construct except that the serine residues in two of three Ser-Gly-Asp sequences were mutated to threonine in each construct (Fig. 1). The SER 1, SER 2, and SER 3 constructs each retained the serines at residue 65, 71, and 76, respectively, while mutating the other two serine residues to threonines. Transfection of COS-7 cells with these constructs produced an immunoreactive product that appeared as a doublet above the 213-kDa marker with the upper band of the doublet being the more prominent (Fig. 5,  lanes 1, 3, and 5). Digestion of the product of these constructs with heparatinase and chondroitinase ABC shifted the migration positions of the upper band to that of the lower band of the doublet (Fig. 5, lanes 2, 4, and 6). This indicates that a major portion of the product of each of these three constructs was produced as a proteoglycan and that the serines in all three Ser-Gly-Asp sequences accept GAG chains. Chromatography 35 SO 4 -radiolabeled recombinant products produced from the SER 1, SER 2, and SER 3 constructs confirmed their production as a proteoglycan (Fig. 3C). Characterization of the GAG type on these recombinant products showed the SER 1 product to contain 47-59% heparan sulfate, the SER 2 product to contain 32-37% heparan sulfate, and the SER 3 product to contain 23-30% heparan sulfate (Table I).
Another set of constructs focused on sequence in domain I that is N-terminal to the three Ser-Gly-Asp sequences. The ALT 3 construct encoded for all of domains II and III but contained a deletion of 16 amino acids in the N-terminal region of domain I (Fig. 1). The ALT 4 construct was identical to the ALT 3 construct except the six acidic residues at positions 55-58, 62, and 63 were changed to polar amino acids (Fig. 1). Transfection of COS-7 cells with the ALT 3 construct produced an immunoreactive product that appeared as a broad band of 213 kDa and higher molecular mass (Fig. 6, lane 1). The prod- uct of the ALT 4 construct (Fig. 6, lane 3) migrated similarly to that seen for the product of the ALT 3 construct. Digestion of the product of both the ALT 3 and ALT 4 constructs produced an immunoreactive band that was more narrowly focused at 213 kDa (Fig. 6, lanes 2 and 4, respectively). This indicates that a portion of the products of both the ALT 3 and ALT 4 constructs was produced as a proteoglycan. Chromatography of 35 SO 4 -radiolabeled recombinant products produced from ALT 3 and ALT 4 constructs confirmed their production as a proteoglycan (Fig. 3D). Characterization of the GAG type on the ALT 3 product show it to contain 67-71% heparan sulfate, but the recombinant product produced by the ALT 4 construct contained no heparan sulfate and was composed instead of 100% chondroitin sulfate (Table I).
The last set of constructs were identical to the ALT 3 construct except that either 2 amino acids (ALT 5), 4 amino acids (ALT 6), or 6 amino acids (ALT 7) were added between residues 63 and 64 of the ALT 3 construct (Fig. 1) to increase the distance between the six acidic residues and the first serine used in GAG synthesis. Western blot assays of transfected cells were not done with these constructs, but metabolic radiolabeling with 35 SO 4 followed by immunoprecipitation and chromatography on Superose 6 showed the product of all three constructs contained incorporated 35 SO 4 (Fig. 3E). Characterization of the GAG type on these products showed that the ALT 5 construct contained 62-72% heparan sulfate, the ALT 6 product 58 -67% heparan sulfate, and the ALT 7 product 50 -68% heparan sulfate (Table I). DISCUSSION The results of this study show that each of the serines in the three Ser-Gly-Asp sequences in domain I of perlecan accept GAG side chains. Mutating all three serines to threonines abolished the GAG-accepting ability of the recombinant product. Mutating any two of the three serines, however, still allowed the recombinant product to be produced as a proteogly-can containing a single GAG chain of either heparan or chondroitin sulfate. There are no other Ser-Gly-Asp sequences in the domains tested in this report, but there is a Ser-Gly-Glu sequence in the cysteine-rich epidermal growth factor-like repeat region of domain III. The recombinant products of the domain II/III construct (this report) and the domain III construct (23) are not made as proteoglycans. This indicates that the site in domain III does not accept GAG chains.
A detectable portion of the recombinant product derived from all constructs examined did not change migration position upon digestion with heparatinase and chondroitinase ABC in the Western blots, indicating that it lacked or had very short GAG side chains. This is not unprecedented, since a portion of the native perlecan produced by colon carcinoma cells was also shown to lack GAG side chains (26). In the transfected COS-7 cells, however, the level of 35 SO 4 incorporated into the domain I/II/III was calculated to be 45-fold higher than that in native perlecan (Fig. 3A). The GAG synthesis system may be saturated to capacity under these conditions, and, as a result, some recombinant product may pass through the endoplasmic reticulum and Golgi with very little or without posttranslational modification.
The recombinant product of the domain I/II/III construct received the highest (73-81%) heparan sulfate content. Changes in the construct reduced the heparan sulfate content and correspondingly increased the chondroitin sulfate content of the resulting recombinant product. The ALT 1 construct was prepared by deleting 110 amino acids from C-terminal end of domain I (approximately 2/3 of domain I). This region of domain I has recently been shown to contain a SEA module, a certain sequence of ϳ80 amino acids that is also found in agrin, a heparan sulfate proteoglycan with homology to perlecan, as  well as in other proteins (4). The SEA module is associated with regions receiving extensive O-glycosylation, but its function is not known (4). Deleting the SEA module from perlecan reduced the heparan sulfate content of the recombinant product to 61-62%. This could indicate that the SEA module enhances heparan sulfate attachment to the three serine residues. These results suggest that sequence in non-GAG-binding regions of the core protein can influence the utilization of GAG attachment sites. Sequences in the non-GAG-binding regions of syndecan I have also been proposed to influence GAG composition, although they have not been identified (27).
Mutating two of the three GAG attachment serines altered both the elution position and the GAG composition of the recombinant product. The 35 SO 4 -radiolabeled recombinant products from the SER 1, SER 2, and SER 3 constructs eluted one tube later than the 35 SO 4 -radiolabeled recombinant product of the ALT 1 construct (Fig. 3, B and C). This indicates the recombinant products of the SER constructs are smaller than the product of the ALT 1 construct. This smaller size is likely due to the presence of only one GAG chain on the recombinant product of each of the SER constructs compared to the presence of three GAG chains on the recombinant product of the ALT 1 construct. Although the size of the recombinant products of the SER constructs were similar, their GAG compositions were different; the serine at residue 65 received 47-59% heparan sulfate, the serine at residue 71 received 32-37% heparan sulfate, and the serine at residue 76 received 23-30% heparan sulfate. Averaging the GAG composition for these three recombinant products gave a value of only 38% heparan sulfate. This is substantially less than that 61-62% heparan sulfate obtained for the recombinant product of the ALT 1 construct, which has all three serine acceptor sites. This suggests that the three Ser-Gly-Asp sites act synergistically to enhance heparan sulfate addition.
Sequence in syndecan 1 was also found to have three closely spaced Ser-Gly sites that act synergistically to enhance heparan sulfate synthesis (19). Three sites in syndecan, however, only received 60% heparan sulfate, which is considerably less than the maximum of 81% we achieved with the three sites in perlecan with the domain I/II/III construct (Table I). This may be due to the different spacing of the three acceptor sites in syndecan and perlecan. Alternatively, the higher levels of heparan sulfate priming in the domain I/II/III product may be due to the presence of heparan sulfate enhancing elements, such as the SEA module, in the recombinant product. In support of this, fusing protein A to the 25 amino acid sequence from perlecan containing only the three acceptor sites and the adjacent cluster of acidic amino acids resulted in a recombinant product that received only 64% heparan sulfate (19). These observations support our hypothesis that the SEA module enhances heparan sulfate synthesis. Another laboratory (22), however, expressed a construct containing primarily domain I of perlecan (amino acids 1-198 plus 6 histidine residues) and found the recombinant product contained only 45% heparan sulfate. Since this construct contains the SEA module but lacks domains II and III, it may be that the SEA modules acts in concert with domains II and III to enhance heparan sulfate synthesis, or that the histidine residues added to domain I act to reduce heparan sulfate priming.
The deletion of 16 amino acids from the N-terminal region of domain I in our ALT 3 construct reduced the heparan sulfate content of the recombinant product but only to 67-71%, and this may not be significant. Mutating the aspartic and glutamic acids at residues 55-58, 62, and 63 to asparagines and glutamines (ALT 4 construct) did not alter the GAG accepting capacity of the serine residues in the three Ser-Gly-Asp se-quences, but changed the GAG composition of recombinant product to be 100% chondroitin sulfate. This change from ϳ 30% chondroitin sulfate to 100% chondroitin sulfate occurred in the presence of the SEA module and on multiple acceptors in close proximity. These results indicate that the acidic residues are primary determinant responsible for targeting the serine residues in the three Ser-Gly-Asp sequences for heparan sulfate and that the multiple acceptors and the SEA module act only as enhancing elements. There are no tryptophan residues near acceptor serines in perlecan that could enhance heparan sulfate synthesis as there is in betaglycan (18). The closest tryptophan to the acceptor serines is over 60 residues away, but it is located in the SEA module and that enhances heparan sulfate synthesis.
The acidic amino acids at residues 55-58 are 7, 8, 9, and 10 residues away from the first GAG priming serine (residue 65) in perlecan. Sequence in betaglycan was also shown to contain acidic amino acids 7, 8, and 10 amino acids away from serine used for heparan sulfate priming (18). The acidic residues in betaglycan, however, are C-terminal to the serine, while the acidic residues in perlecan are N-terminal to the serine. Moving the acid residues in betaglycan closer to the serine or deleting the residues reduced heparan sulfate priming (18). In perlecan, there is a correlation between the number of acidic residues, their distance from an attachment site, and the percentage of heparan sulfate synthesized at that site in the SER 1, 2, and 3 constructs. The serine at position 65 received 47-59% heparan sulfate and has acidic residues 7, 8, 9, and 10 residues away. The serine at position 71 received 32-37% heparan sulfate and has acidic residues 8 and 9 residues away. The serine at position 76 received 23-30% heparan sulfate and has one acidic residue 9 residues away (Fig. 1). Moving the acidic residues in perlecan 2, 4, or 6 amino acids away from the first serine (ALT 5, 6, and 7) resulted in only a small reduction in heparan sulfate synthesis (Table I), but this may be due to the fact that the four acidic residues at positions 55-58 were essentially replaced by the two acidic residues at 62 and 63 by the addition of the 6 amino acids in the ALT 7 construct (see Fig. 1). Nevertheless, these results taken together indicate that it is not the direction of the acidic group from the priming site but the number of residues and a minimum distance (ϳ7 residues) that is important in regulating heparan sulfate priming. The maximum distance has yet to be determined, and this may be difficult because protein folding could possibly bring a far away cluster of acid residues closer to the serine.