Heparan Sulfate Structure in Mice with Genetically Modified Heparan Sulfate Production*

Using a high throughput heparan sulfate (HS) isolation and characterization protocol, we have analyzed HS structure in several tissues from mice/mouse embryos deficient in HS biosynthesis enzymes (N-deacetylase/N-sulfotransferase (NDST)-1, NDST-2, and C5-epimerase, respectively) and in mice lacking syndecan-1. The results have given us new information regarding HS biosynthesis with implications on the role of HS in embryonic development. Our main conclusions are as follows. 1) The HS content, disaccharide composition, and the overall degree of N- and O-sulfation as well as domain organization are characteristic for each individual mouse tissue. 2) Removal of a key biosynthesis enzyme (NDST-1 or C5-epimerase) results in similar structural alterations in all of the tissues analyzed. 3) Essentially no variation in HS tissue structure is detected when individuals of the same genotype are compared. 4) NDST-2, although generally expressed, does not contribute significantly to tissue-specific HS structures. 5) No change in HS structure could be detected in syndecan-1-deficient mice.

Heparan sulfate (HS) 1 biosynthesis is a complex process that results in structures important for a large number of biological processes (1,2). These processes include physiological as well as pathological situations where HS acts in morphogen binding, enzyme activation, or cellular signaling, all of which are essential for the development and maintenance of an organism (3,4).
HS is characterized by a long linear backbone structure consisting of alternating glucuronic acid and N-acetylglucosamine residues. During biosynthesis, this backbone is modified by N-deacetylation/N-sulfation of the glucosamine unit, C5-epimerization of the glucuronic acid residue to iduronic acid, and O-sulfation of both monosaccharide units. Not all of the sites are modified, and a large number of different sequences are produced that provide tailored binding sites for different proteins (2). Glucosaminyl N-deacetylase/N-sulfotransferase (NDST) is the first enzyme engaged in the modification of the sugar backbone and is believed to create substrates for further enzymatic activity along the polymer (1). NDST removes the acetyl group from the glucosamine residue and replaces it with a sulfate group. By action of this enzyme, a domain structure is imprinted on the chain with stretches of N-sulfated disaccharides (NS domains) interrupted by stretches of N-acetylated disaccharides (NA domains). At the border of these two domain types, short alternative N-sulfated/ N-acetylated areas are found (NA/NS domains) (5,6). Four isoforms of NDST have been identified in mammalian organisms with NDST-1 and NDST-2 being the most widely expressed (7). The roles of the different isoforms in HS/heparin biosynthesis are only partly known. The drastically different effects of targeted mutations in NDST-1 and NDST-2 suggest critical differences (8 -11). The NDST-2 knock-out mouse is healthy and fertile, but the connective tissue type mast cells are reduced in number and contain no sulfated heparin (8). A lack of the NDST-1 isoform results in neonatal lethality where the newborn pups die in a condition resembling respiratory distress syndrome (10,11). These mice also display skull and skeletal defects (7,10).
Stretches of N-sulfated disaccharides in NS domains are the substrates for the next enzyme in the biosynthetic cascade, the glucuronyl C5-epimerase (Epi), an enzyme that is encoded by a single gene. Epimerization provides the substrate for the 2-Osulfotransferase. In addition, uronic acid residues give more flexibility to the chain, which may favor protein binding (12). Epimerase deficiency (Hsepi Ϫ/Ϫ ) in mice results in kidney agenesis, skeletal defects, and a lung phenotype, and the embryos die neonatally (13).
Syndecan-1 knock-out mice (Sdc-1 Ϫ/Ϫ ) (14) develop normally and are fertile yet seem to develop enhanced leukocyte adhesion under inflammatory conditions as induced by stimulation of endothelium with tumor necrosis factor-␣ (15). In addition, Sdc-1 Ϫ/Ϫ mice are characterized by markedly delayed skin and cornea wound healing (16) as well as resistance to Wnt-1mediated mammary gland tumorigenesis (14) and certain microbial lung infections (17). The expression of Sdc-1 is temporally restricted and highly regulated during embryonic * This work was supported by the Swedish Research Council (Grants 32X-15023 and 521-2002-3413), the programme "Glycoconjugates in Biological Systems" and Grant A303:156e sponsored by the Swedish Foundation for Strategic Research, Polysackaridforskning AB, Gustaf V:s 80-årsfond, Swedish Cancer Society (4708-B02-01XAA), and the "Forschungsverbund Nordrhein-Westfalen-Schweden" (to M. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Most of the information regarding HS structure originates from studies of HS from a few larger organs from different species and individuals (22)(23)(24) or from cell lines (25). Structural information has also been provided from immunohistochemical staining of tissue sections with HS monoclonal antibodies (26) or phage display antibodies (27). Yet very little is known regarding the quantity and quality of HS found in different organs of a single animal or how changes in the biosynthetic machinery affect different tissues. The increasing number of genetically modified mice produced with defects in HS biosynthesis and structure (28) has increased the demand for sensitive and high throughput methods for HS-structural analyses. To study the impact of HS-structural changes on the phenotype of genetically altered organisms, we have developed suitable methods to characterize HS from different tissues. We analyzed HS from minute tissue samples of wild type animals and animals deficient in either the biosynthesis enzymes (NDST-1, NDST-2, and Epi) or a core protein (Sdc-1). To our surprise, we found an astonishing stability of tissue-specific HS production in control as well as in mutant mice.
Large Scale Isolation of Glycosaminoglycans (GAGs)-Five grams of swine kidney powder were homogenized in cold acetone and left for 24 h. The solvent was removed by filtration over Whatman No. 3MM paper, and the pellet was suspended in a minimal volume of water. Methanol and chloroform were added while stirring to a final ratio of 3:8:4 (H 2 O:CH 3 OH:CHCl 3 v/v) and left overnight at room temperature. The solvent mixture was removed by filtration as above, and the filtrate was washed twice with ethanol and dried at room temperature. The defatted tissue powder was suspended in Pronase digestion buffer (0.1 M Tris/HCl, pH 8.0, 2 mM CaCl 2 , 3% ethanol) to a concentration of 100 mg/ml, and Pronase was added (5 mg/g tissue). After incubation for 24 h at 55°C, another aliquot of Pronase was added and the incubation continued for 24 h. The incubation was terminated by heating the sample for 15 min at 96°C, and the suspension was centrifuged for 15 min at 2000 ϫ g. Before adding benzonase (125 units), the supernatant was adjusted to a final 2 mM MgCl 2 . After the addition of the enzyme, the sample was incubated for 24 h at 37°C. The supernatant was centrifuged as above and loaded onto a column of 5 ml of DEAE-Sephacel equilibrated in 50 mM Tris/HCl, pH 7.5, 0.2 M NaCl. The column was washed with one column volume of equilibration buffer and three volumes of 50 mM NaAc, pH 4.5, 0.2 M NaCl before a gradient of NaCl from 0.2 to 2 M was applied in the same buffer (a 15 volume total). Fractions were collected and tested for uronic acid content, and the positive fractions were pooled, dialyzed against H 2 O, and dried.
Purification of HS from GAGs-Purified GAGs (100 g of uronic acid) were incubated in 100 l of 40 mM Tris-Ac, pH 8.0, containing 100 milliunits of chondroitinase ABC overnight at 37°C. The digest was loaded onto a 2-ml DEAE-Sephacel column equilibrated in 50 mM Tris/ HCl, pH 8.0, 0.1 M NaCl. The column was successively washed with five column volumes of equilibration buffer, five volumes of wash buffer (50 mM NaAc, pH 4, 0.1 M NaCl), and two volumes of 0.2 M NH 4 HCO 3 before elution with five volumes of 2 M NH 4 HCO 3 .
Miniscale Isolation of GAGs from Tissues and Cells-Mouse tissues were isolated and dried by lyophilization. Dry tissue (Յ50 mg) was dissolved in 0.5 ml of Pronase buffer (50 mM Tris/HCl, pH 8, 1 mM CaCl 2 , 1% Triton X-100) containing 0.8 mg/ml Pronase and incubated for 24 h at 55°C with end-over-end mixing. A second aliquot of Pronase was added after 20 h. After heat inactivation of the Pronase, the buffer was adjusted to 2 mM MgCl 2 and benzonase (12 milliunits) was added. The sample was incubated for 2 h at 37°C, heat inactivated, adjusted to a final sodium chloride concentration of 0.1 M, and centrifuged at 13,000 ϫ g for 10 min. For purification of GAGs from tissue digests, 0.3-ml columns of DEAE were prepared in Bio-Rad 10-ml columns. The columns were primed by washing with one column volume of elution buffer (2 M NH 4 HCO 3 ) and three column volumes of loading buffer, pH 8 (50 mM Tris/HCl, pH 8, 0.1 M NaCl, 0.1% Triton X-100). The supernatants of the tissue extracts were applied, and the columns were washed successively with six volumes of loading buffer, six volumes of wash buffer, pH 4 (50 mM NaAc, pH 4, 0.1 M NaCl, 0.1% Triton X-100), one volume of H 2 O, and five volumes of 0.2 M NH 4 HCO 3 . The elution of GAGs was achieved with six volumes of 2 M NH 4 HCO 3 . The eluate was collected in an Eppendorf tube and repetitively dried by a SpeedVac until the pH reached ϳ7. For the isolation of HS, the GAG pool was digested with 50 milliunits of chondroitinase ABC in a final volume of 50 l of 40 mM Tris-Ac buffer, pH 8.0. The CS digestion was allowed to proceed for Ն3 h at 37°C. To stop the reaction, 50 l of H 2 O was added and boiled. From the CS digest, 10 l was removed for analysis of the CS content by reversed-phase ion pairing-high performance liquid chromatography (RPIP-HPLC) (see below). The remainder of the chondroitinase-digested GAGs (i.e. HS) was cleaned on a new DEAE-column performed in the same way as described for the GAG isolation. Under these conditions of elution, a quantitative recovery of all types of HS was achieved. After freeze-drying, the purified HS was dissolved in 100 l of H 2 O and the pH was checked. Two equal aliquots were dried and prepared for heparitinase digestion. One of the aliquots was treated with 0.4 milliunits each of heparitinases I, II, and III in 15 l of heparitinase buffer (5 mM Hepes, pH 7.0, 50 mM NaCl, 1 mM CaCl 2 , 0.7 mg/ml bovine serum albumin) and incubated for 16 h at 37°C. The other aliquot was incubated under the same conditions without enzymes. The reaction was stopped by heat inactivation. All of the samples were dried and resuspended in 45 l of H 2 O for injection on the HPLC.
Production of Oligosaccharides from GAGs and HS-GAGs and HS isolated from swine kidney were cleaved by deamination at pH 1.5 into disaccharides, tetrasaccharides, and larger oligosaccharides according to standard procedures (38) and subfractionated on Sephadex G-15 as described above. Individual pools were analyzed by RPIP-HPLC as described below.
Deamination of HS Chains and Tissue Samples-For fingerprinting different tissue and cell GAGs, purified GAG or HS (0.5-10 g) was dried down in Eppendorf tubes. 10 l of freshly prepared deamination reagent (50 mM NaNO 2 in H 2 O adjusted to pH 1.5) was added to the samples. Samples were incubated for 10 min at room temperature and evaporated. The dry cleavage products were redissolved in 100 l of H 2 O and analyzed directly by RPIP-HPLC without further purification. As control, the non-cleaved sample was mixed with the reagent salts of an aliquot of evaporated deamination reagent dissolved in 100 l of H 2 O. Under the conditions chosen, quantitative cleavage of HS chains at available N-sulfated glucosamine residues was achieved in a window up to 5 g of starting HS (data not shown).
Analysis of GAGs-GAGs were quantified by colorimetric determination of hexuronic acid using the meta-hydroxy-diphenyl method (39) with GlcA as a standard. A factor of 3 was arbitrarily employed to convert values to saccharide mass. Reducing disaccharides were quantified by the methylbenzothiazolinone hydrazone hydrochloride assay using GlcN as a standard (40). The analysis of the 3 H-labeled anhydromannitol disaccharide species was performed by strong anion exchange (SAX-HPLC) on a Partisil-10 SAX column (41)  In B, the standards were the deamination-created disaccharides: 1, GlcA/IdoA-anMan; 2 IdoA2S-anMan; 3, IdoA-anMan6S; 4, GlcA-anMan6S; 5, GlcA2S-anMan; and 6 IdoA2S-anMan6S. The elution position of isolated 4-mers and Ͼ4-mers, respectively, is indicated with a bar. Contaminating CS disaccharides (disulfated ⌬CS disaccharides of type ⌬HexA-GalNAc4.6S, ⌬HexA2S-GalNAc4S, and ⌬HexA2S-GalNAc6S, respectively) in the HS-preparation are indicated by ⌬CSdi and provide an internal standard when comparing deaminative with enzymatic cleavage products. 11.5 min, 56 -106 mM in 1.5 min, and 106 mM for 6 min) followed by a column wash with 120 mM NaCl for 1 min essentially as described by Staatz et al. (42). A flow rate of 1.1 ml/min was kept, and post-column detection was performed by the addition of 2-cyanoacetamide (0.25%) in NaOH (0.5%) at a flow rate of 0.35 ml/min as described previously (42). Signals were quantified against known amounts of standard disaccharides analyzed in parallel runs.

RESULTS
Heparan Sulfate Content and Structure Is Distinct in Different Mouse Tissues-GAGs were purified from mouse tissues by a miniscale protocol adapted for handling a larger number of samples with identical results as by the use of traditional isolation methods (data not shown). The samples were analyzed with a method modified from a protocol originally used for structural analysis of GAGs from Drosophila tissues (43). In this method, isolated HS is cleaved with heparitinases I-III to generate disaccharides containing a 4,5-unsaturated hexuronic acid (⌬HexA) at the non-reducing end (see Fig. 1) that are analyzed by RPIP-HPLC. Also CS can be studied by analyzing the chondroitinase digest. As little as 5 mg of dry weight of any HS-producing mouse tissue give enough material for the analysis of quantitative and qualitative aspects of HS composition. The amounts of HS present in different tissues of an adult mouse (C57BL strain) varied considerably, i.e. kidney and lungs with 600 -800 ng of HS disaccharides/mg of dry tissue being the most HS-rich tissues, whereas skeletal muscle, liver, skin, or brain contained Ͻ150 ng/mg dry weight (Fig. 2).
The HS disaccharide composition (Fig. 3A) as well as the overall degree of N-and O-sulfation (Fig. 3B) was characteristic for each individual tissue. The most remarkable pattern relative to a more "average" HS composition was seen in lung tissue with an extremely high overall 2-O-sulfation level due to the presence of monosulfated ⌬HexA2S-GlcNAc species (Fig.  3A). HS from spleen, kidney, and liver on the other hand was relatively rich in 6-O-sulfate groups. In contrast to all of the other tissues examined, skin samples from different mice contained HS with a somewhat variable structure.
HS Fingerprints Are Tissue-specific-By enzymatic digestion of HS followed by disaccharide compositional analysis (Fig. 1A), only part of the information concerning the HS structure was obtained. The epimerization state of the different disaccharide species as well as the domain arrangement of the chain cannot be analyzed by this method. To overcome this limit, we adapted the low pH deamination method (44) to miniscale followed by direct analysis using RPIP-HPLC (Fig. 1B). By deamination at pH 1.5, HS chains are cleaved at N-sulfated glucosamine residues. The fragments produced contain NS disaccharides, NA/NS tetrasaccharides, and hexasaccharide and longer saccharides of the general structure [NA] n -NS (Fig. 1). Each of the fragments contains an anhydromannose at the reducing end, which allows for postcolumn tagging. Disaccharide species were identified by co-elution with known standard disaccharides (Fig. 1B), whereas tet- rasaccharides and larger fragments were compared with the elution products of size-separated HS cleavage products purified by gel chromatography. Although the signals for the disaccharides behaved in a linear fashion over a broad range (0 -1000 pmol), we did not attempt to quantify the tissue patterns due to partial overlap of some species. Instead, we used the possibility to gain a tissue-specific fingerprint by looking for differences in (a) the amounts and relative proportions of N-sulfated disaccharides from NS domains and (b) the domain pattern manifested in the proportions of larger fragments (Fig. 4, A-D). A remarkable con-stancy of such tissue fingerprints between different (inbred) individuals was found for HS from liver, kidney, lung, bladder (Fig.  4, A-D), intestine, and testis (data not shown). The relative proportions of different disaccharides and larger fragments were tissue-specific. All of the disaccharide species originate from contiguous N-sulfated domains, and by comparing the relative proportions (Fig. 4), differences in composition of these domains could be monitored. In general, the level of GlcA/IdoA-anMan6S reflecting HexA-GlcNS6S in the native chain (peak 3 ϩ 4 in Fig.  4) constituted a minor but stable proportion in all of the tissues , lung (C), and bladder (D) of an adult wild type and a NDST-2 Ϫ/Ϫ mouse was deaminated at pH 1.5, and the resulting fragments were analyzed by RPIP-HPLC as described under "Experimental Procedures" and in Fig. 1. The curves of wild type (gray line) and NDST-2 Ϫ/Ϫ HS (thin black line) are superimposed to emphasize the close similarity in structure. The boxed area of the respective chromatograms is enlarged (AЈ, BЈ, CЈ, and DЈ). The standards are as described in Fig. 1. The same results were obtained for another pair of animals. examined, whereas the levels of GlcA/IdoA2S-anMan (from HexA2S-GlcNS) (peak 2 ϩ 5 in Fig. 4) and IdoA2S-anMan6S (from IdoA2S-GlcNS6S) (peak 6 in Fig. 4) were variable. In liver and kidney, for example, IdoA2S-GlcNS6S was the dominating species, whereas HexA2S-GlcNS dominated in lung and bladder (Fig. 4, A-D). More 2-O-sulfated disaccharides were found in lung HS than in HS from other tissues, and lung was the only tissue where appreciable amounts of 2-O-sulfated GlcA-containing disaccharides could be detected (peak 5 in Fig. 4C). The fingerprint from regions with spaced N-sulfates, i.e. fragments larger than disaccharides (Fig. 4, AЈ-DЈ) revealed that also regions outside the NS domains expressed significant structural variation. Whereas liver and kidney again were rather similar (Fig. 4, AЈ-BЈ), lung and bladder (Fig. 4, CЈ and DЈ) manifested distinctly different patterns.
Removal of Sdc-1 Core Protein Does Not Affect HS Structure-The total level of expression as well as the relative proportion of different proteoglycan core proteins vary between tissues and could tentatively contribute to the tissue-specific variation in HS structure. By analyzing HS composition in tissues from Sdc-1 Ϫ/Ϫ mice, using the RPIP-HPLC enzyme method, we wanted to see whether HS biosynthesis had been disturbed in such a way that the tissue-specific structures had been changed. Sdc-1 expression has been reported to be high in lung airway epithelia and skin (18,19), and these tissues have been associated with the only phenotypes of Sdc-1 Ϫ/Ϫ mice. Yet the removal of Sdc-1 from these tissues did not affect the composition of HS to any extent (Table I).
HS Structures Are Not Altered in NDST-2-deficient Mice-Despite ubiquitous expression of NDST-2, only a restricted mast cell phenotype is seen in NDST-2-deficient mice (8). In addition, a detailed characterization of liver HS in these mice could not demonstrate any alteration in HS structure compared with HS of control mice. 2 To screen for differences in HS structure in other tissues, compositional disaccharide analysis after heparitinase digestion was performed on liver, kidney, lung, cerebellum, cerebrum, intestine, bladder, testis, and skin (data not shown). No differences between HS from control and NDST-2 Ϫ/Ϫ tissues were found. In addition, the HS fingerprint of liver, lung, kidney skin, intestine, testis, and bladder did not reveal any appreciable difference in the GlcA/IdoA content or in the domain arrangement ( Fig. 4 and data not shown), suggesting no or minimal participation of NDST-2 in determination of tissue-specific HS structures in these tissues.

NDST-1 and Epi Take Part in HS Biosynthesis in All of the Tissues Examined-
The low penetrance of some phenotypes reported in NDST-1 Ϫ/Ϫ and Hsepi Ϫ/Ϫ animals might indicate that HS biosynthesis is more variable in these mice. If so, HS structures with a binding capacity for certain growth factors or morphogens could be lost in some but not all of the embryos. However, a striking similarity in the structure of HS isolated from different individuals was evident in fingerprints of HS from embryos of the same genotype as shown for wild type (Fig.  5A) and Hsepi Ϫ/Ϫ (Fig. 5B). This was also true for NDST-1 Ϫ/Ϫ embryos (data not shown). The HS fingerprint of the NDST-1 Ϫ/Ϫ embryos confirmed the general reduction of N-sulfated and thus cleavable disaccharides, the 2-O-sulfated disaccharides (i.e. IdoA2S-anMan and IdoA2S-anMan6S) being most reduced (data not shown). The finding that some tissues have severe phenotypes while others develop normally in NDST-1 Ϫ/Ϫ and Hsepi Ϫ/Ϫ mice (10, 13) might indicate that the effect of the missing enzyme on HS structure varies accordingly. Looking at individual NDST-1 Ϫ/Ϫ tissues with the RPIP-HPLC enzyme method, N-sulfation was reduced to a similar degree in all of the tissues analyzed in favor of more N-acetylation (Table II) such as muscle and skin (data not shown). At the same time, the overall sulfation level also decreased due to an overall reduction in 2-O-and 6-O-sulfation. Yet the Osulfation was affected differently on the disaccharide level. Whereas all of the species of 2-O-sulfated disaccharides were reduced, the overall 6-O-sulfation was mainly reduced through the reduction of the trisulfated disaccharide ⌬HexA2S-GlcNS6S (Fig. 6A). The other 6-O-sulfated disaccharide species, ⌬HexA-GlcNAc6S and ⌬HexA-GlcNS6S, were essentially unchanged. Most amazingly, the HS structural alterations in all of the tissues followed a similar trend, i.e. independent of their tissue-specific "ground state" of total composition in the control, the changes in individual disaccharide species were similar for each of these tissues (Table II). The most profound difference between these tissues was the more distinct reduction in ⌬HexA2S-GlcNS6S in liver compared with the other two tissues.
In the Hsepi Ϫ/Ϫ E18.5 embryo, a total abolishment of epimerization has been reported (13). These results were confirmed with the RPIP-HPLC methods where iduronic acid a Disaccharides, all containing a 4,5-unsaturated hexuronic acid (HexA) unit, were generated by digestion with a mixture of heparitinases and analyzed by RPIP-HPLC as described under "Experimental Procedures." Values are given in mol % of total disaccharides as calculated from peak areas relative to known amounts of standard disaccharides. The values are the mean Ϯ S.D. of n determinations.
b The degree and type of sulfation are indicated per 100 disaccharides calculated from disaccharides species in footnote a .

DISCUSSION
In both Drosophila and mice, HS has been shown to be of crucial importance for embryonic as well as post-embryonic development (28,43,45). Yet very little is known regarding the regulation of HS production. Our goal was to take advantage of a number of mouse models with targeted mutations in genes coding for different components of the HS-proteoglycan biosynthesis apparatus to learn more about the general features of HS biosynthesis. HS composition has so far been determined either on material from metabolically labeled cell lines (46,47) or tissue samples collected from different individuals and processed in large scale in a relatively tedious purification and analysis process (23,24). With our improved isolation method, we succeeded to analyze several tens of samples in parallel without requirement to further purify or radiolabel tiny amounts of material. This enabled us to look into the structural details of HS, even from minor tissues from genetically modified animals. As has been indicated earlier for Drosophila (43), bovine (24), and human tissue samples (23), remarkable tissuespecific HS levels, composition, and domain organization were manifest, pointing to a stringent control of the biosynthetic production line. Combining the results of enzymatic and chemical cleavage, we were furthermore able to appreciate differences in not only the total composition and N-sulfated domains but also the alternating (NA/NS domains) in the very same screens. As alternating domains constitute an appreciable part of each chain and as these domains show a considerable potential for protein binding (5,6), the variation in tetrasaccharide fingerprints may contribute different protein binding preferences. The resolution of our combined analyses allowed us to reveal the similarity of composition and domain structure of liver and kidney HS, for example, compared with the unique features of lung HS structures (Fig. 3, 4).
What happens with HS structure when one of the biosynthesis enzymes normally present is deleted? Our original hypothesis was that we would see different responses in different tissues, tentatively depending on the level of expression of the particular enzyme in the normal state. Therefore, it was surprising that very similar changes were seen in HS structure in all of the tissues analyzed in the knock-out mice (Table II and Fig. 6). This was true both for the NDST-1 Ϫ/Ϫ and the Hsepi Ϫ/Ϫ mice, suggesting that the HS biosynthesis machinery has the same basic organization in all of the tissues and that the effect of the mutation on HS structure in these animals is general. Thus, HS-binding growth factors and morphogens must have some tolerance for changes in HS structure. As many of the tissues in NDST-1 Ϫ/Ϫ , Hsepi Ϫ/Ϫ , and 2-O-sulfotransferase negative (Hs2st Ϫ/Ϫ ) animals apparently develop normally (10,13,48). What is then the point of tissue-specific HS structures? One possibility is that small differences in HS structures are less critical during organogenesis but may be important during adult life. Alternatively, other factors (e.g. maternal) could compensate during the critical period of embryogenesis.
HS biosynthesis seems to be an extraordinary stable process, because no individual variation in HS structures was seen between individuals of the same genotype ( Fig. 5 and Table II). Thus, the variation in penetrance of different abnormalities observed among NDST-1 Ϫ/Ϫ (10) and Hsepi Ϫ/Ϫ (13) animals, respectively, is unlikely to be due to differences in HS structures. Perhaps the abnormal HS structures lead to a decreased robustness of developmental processes such that timing, for example, becomes more critical. Taken together, our results support a model where most enzymes normally are expressed in excess and the substrate specificity of the enzymes (maybe together with so far unidentified tissue factors) determines the outcome of the biosynthesis. This model goes well with the fact that, so far, no direct coupling has been seen between the transcript level of a certain biosynthesis enzyme and its impact on HS structure. No difference in structure as compared with the control is seen in HS produced by mice heterozygous for the different biosynthesis enzymes, NDST-1, Epi, or 2-O-sulfotransferase (10,13,49). In addition, we know from studies of liver transcript levels in NDST-1 and NDST-2 knock-out animals, that no transcriptional compensation occurs when NDST-1 or NDST-2 is knocked out. 2 As reported in this paper, NDST-2 does not significantly, if at all, contribute to the variation of tissue-specific HS structures in mice. Because different sulfation patterns are seen after overexpression of NDST-1 and NDST-2 (50), these results indicate that NDST-2, if translated (see Ref. 51), does not take part in HS biosynthesis in most tissues. However, NDST-2 is the only expressed NDST isoform in liver of NDST-1 Ϫ/Ϫ animals where HS is still N-sulfated to almost half of its normal level, indicating that NDST-2 functions in HS biosynthesis when NDST-1 is absent. 2 As NDST-1 and NDST-2 are the only generally expressed NDST isoforms, this may apply to HS biosynthesis in all of the tissues examined in NDST-1 Ϫ/Ϫ mice. Thus, NDST-2, although most probably widely expressed in an active form, appears redundant in other cells than mast cells.
The absence of either NDST-1 or Epi results in drastic effects on the sulfation pattern in all of the tissues studied. These findings extend data described earlier on individual tissues and whole embryos (13,52). Previously, HS structure in NDST-1 Ϫ/Ϫ mice has been studied in embryonic fibroblasts and in more detail in E18.5 liver (10). 2 The main characteristics of NDST-1 Ϫ/Ϫ E18.5 liver HS were a reduced N-sulfation and a decrease in 2-O-sulfated disaccharides, whereas the total level of 6-O-sulfation was less affected. 2 In whole Hsepi Ϫ/Ϫ embryos, the HS analyzed contained longer N-sulfated domains and a dramatic increase in GlcA-GlcN6S content was observed (13). Here we show that this effect is general (Table II) and that the amount of ⌬HexA-GlcNAc6S disaccharides is reduced in favor of its N-sulfated analogue, ⌬HexA-GlcNS6S. This effect is possibly explained by a reduction in the number of NA/NS domains where HexA-GlcNAc6S is common (53). An apparent exception from the more general structural changes seen in HS as a consequence of NDST-1 or Epi deficiency is given by the 2-O-sulfated disaccharides of liver. The reduction of ⌬HexA2SGlcNS in both mutant mice strains is smaller in liver compared with that seen in lung and brain, whereas the reduction in the trisulfated disaccharide ⌬HexA2S-GlcNS6S is greater (Fig. 6). However, the different contents of these disaccharides in HS of the wild type tissues may explain the differences. Wild type liver has a higher content of the trisulfated disaccharide than lung and brain, whereas the opposite is true for ⌬Hex2SGlcNS (Table II).
Despite similar effects on HS biosynthesis in different tissues, the impact of altered HS structures on individual organs is clearly different. Whereas mice deficient in Epi or 2-O-sulfotransferase lack kidneys, most organs develop normally (13,48). In Hsepi Ϫ/Ϫ mice, the level of 2-O-sulfation is reduced, a Disaccharides, all containing a 4,5-unsaturated HexA unit, were generated by digestion with a mixture of heparitinases and analyzed by RPIP-HPLC as described under "Experimental Procedures." Values are given in mol % of total disaccharides as calculated from peak areas relative to known amounts of standard disaccharides. The values are the mean Ϯ S.D. of n determinations.
b The degree and type of sulfation are indicated per 100 disaccharides calculated from disaccharides species in a .
FIG. 6. HS from selected tissues of NDST-1 and Epi-deficient animals analyzed by enzymatic cleavage. The HS composition of liver, brain, and lung from NDST-1 Ϫ/Ϫ (A) and Hsepi Ϫ/Ϫ (B) embryos was compared with the wild type HS (Table II), and the data were plotted as the number of changed disaccharides/100 disaccharides for individual disaccharide species. whereas a compensatory 6-O-sulfation occurs, similar to the situation in 2-O-sulfotransferase-deficient mice (49). These results indicate that 2-O-sulfation is crucial for kidney organogenesis and that the increased 6-O-sulfation, which seems to give at least a reasonable binding capacity for fibroblast growth factors (49), does not allow for kidney development. The significantly reduced amount of sulfated structures in the NDST-1 knock-out animal does not abolish the development of seemingly normal kidneys (10). Thus, whereas kidney organogenesis seems to be more dependent on 2-O-sulfation, the lung development is not, even though lungs contain a high proportion of 2-O-sulfated HS (Fig. 3). Instead, the development of functional lungs seems to critically depend on factor(s) requiring proper epimerization of the HS chains as indicated by lung phenotypes in both NDST-1 Ϫ/Ϫ and Hsepi Ϫ/Ϫ animals. Besides the proper HS structures, the concentration of HS also may influence protein binding capacity and phenotypes.
Although a variety of interesting phenotypes have been described for Sdc-1-deficient mice (14,16,17,21), these phenotypes are only apparent when the mice are challenged. Because Sdc-1 is expressed very early during embryonic development (54) and in a highly regulated manner, it was a surprise that the Sdc-1 Ϫ/Ϫ mice develop normally and are fertile. In our study, no difference in HS structure could be recorded in Sdc-1-deficient mice, even in tissues with high Sdc-1 expression (Table I) (19). Therefore, we may conclude that the removal of Sdc-1 has not induced a change in the overall tissue-specific HS modification pattern. As a result, the mild Sdc-1 Ϫ/Ϫ phenotypes could in part be explained by the presence of other HS proteoglycans being decorated with normal HS structures within the same tissue such that the biological functions of HS independent of core protein identity still can occur (55)(56)(57)(58). On the other hand, the phenotypes observed may illustrate core protein-specific functions (21,59,60). Our results further support in vitro studies where no or minimal differences in HS structure have been observed between different proteoglycans isolated from the same mouse mammary gland epithelial cell (57,58). However, because Sdc-1 may constitute only a minor fraction of the total HS proteoglycan pool in some of the investigated tissues, less dramatic changes in HS structure may have been missed. Experiments with defined cell populations isolated from Sdc-1 Ϫ/Ϫ mice where the relative proportion of Sdc-1 can be compared with other HS proteoglycans could be useful to elucidate how or whether different core proteins influence HS biosynthesis. It could also be studied whether the up-regulation of other core proteins, as suggested for mouse mammary gland epithelial cells (61), leads to altered HS structure.
In conclusion, we have shown that embryo tissues from NDST-1 Ϫ/Ϫ and Hsepi Ϫ/Ϫ mice, which die neonatally, generally express highly abnormal HS structures, whereas tissues from the viable NDST-2 Ϫ/Ϫ and Sdc-1 Ϫ/Ϫ mice express apparently normal HS. It will be interesting to investigate how more subtle manipulations of HS biosynthesis in mice and other model organisms will affect embryonic development. We believe that the RPIP-HPLC methods presented in this paper will be valuable tools in analyzing the HS (and CS) structure in the different model organisms. The methods also work well on cultured cells, making it possible to study different aspects of GAG biosynthesis in vitro.