Embryonic stem (ES) cells can be cultured in conditions that either maintain pluripotency or allow differentiation to the three embryonic germ layers. Heparan sulfate (HS), a highly polymorphic glycosaminoglycan, is a critical cell surface coreceptor in embryogenesis, and in this paper we describe its structural transition from an unusually low-sulfated variant in ES cells to a more highly sulfated form in fluorescence-activated cell sorting-purified neural progenitor cells. The characteristic domain structure of HS was retained during this transformation. However, qualitative variations in surface sulfation patterns between ES and differentiated cells were revealed using HS epitope-specific antibodies and the HS-binding growth factor fibroblast growth factor 2 (FGF-2). Expression profiles of the HS modification enzymes indicated that both “early” (N-sulfotransferases) and “late” (6O- and 3O-sulfotransferases) sulfotransferases contributed to the alterations in sulfation patterning. An HS-null ES line was used to demonstrate the necessity for HS in neural differentiation. HS is a coreceptor for many of the protein effectors implicated in pluripotency and differentiation (e.g., members of the FGF family, bone morphogenic proteins, and fibronectin). We suggest that the stage-specific activities of these proteins are finely regulated by dynamic changes in sulfation motifs in HS chains.
Disclosure of potential conflicts of interest is found at the end of this article.
Heparan sulfate proteoglycans (HSPGs) are ubiquitous residents of most organs and tissues. They consist of long glycosaminoglycan (GAG) chains covalently attached to core proteins via a short linker, and their presence at the cell surface and extracellular matrix is critical for their varied roles in cell growth, morphogenesis, and cell-cell/cell-matrix interactions [1, 2]. These functions are in general mediated by the heparan sulfate (HS)-GAG side chains. For some factors, HSPGs are cell surface coreceptors (e.g., fibroblast growth factor [FGF]/FGF receptor or hepatocyte growth factor/met systems ), whereas for others (e.g., hedgehog and Dpp), the high abundance of cell-surface HSPGs, combined with low-affinity interactions, makes them excellent candidates for modulating morphogen gradients and controlling delivery to signaling receptors [3, 4].
HS is synthesized as a nonsulfated precursor of alternating N-acetylglucosamine (GlcNAc) and glucuronic acid (GlcA) sugars. Polymerizing chains undergo a series of sequential stepwise modifications, beginning with de-N-acetylation and re-N-sulfation of GlcNAc to generate N-sulfoglucosamine (GlcNS), followed by epimerization of GlcA to iduronic acid (IdoA) and O-sulfation at C-2 of the uronic acids and then at C-6 (and more rarely C-3) of the glucosamine residues. These modifications are incomplete and interdependent, resulting in a mature HS chain composed of relatively short but hypervariable domains of high sulfation (S domains) flanked by regions of intermediate sulfation (N-acetylated [NA]/N-sulfated [NS] domains or transition zones) distributed at regular intervals within an N-acetylated backbone. These modifications are catalyzed by a bank of enzymes, many of which are members of multigene families that have highly conserved tissue-specific expression patterns during development and in adult life .
Many critical pathways for early development are shared between species and are dependent on factors whose activity is influenced by interactions with HSPGs [4, –6]. Mice mutant for various components of the HS biosynthetic pathway have a range of phenotypes, ranging from early embryonic lethality for embryos deficient either in HS or in polymer sulfation [7, –9] to more tissue-specific defects (e.g., loss of kidneys in mice lacking 2-O-sulfation  or IdoA [11, 12] and abnormal mast cell production in mice with reduced N-sulfation of heparin ). These mammalian studies, complemented by those in Drosophila, Caenorhabditis elegans, and zebrafish have implicated HS in many critical signaling pathways in early development, including those mediated by members of the FGF, hedgehog, bone morphogenic protein, and Wnt families .
Understanding how the expression of biosynthetic enzymes translates into specific sulfated motifs in HS is an area of considerable interest in developmental biology. We have investigated this relationship using embryonic stem (ES) cells as an in vitro model of early embryonic development. We have taken advantage of a published neural differentiation protocol that uses a Sox1-enhanced green fluorescent protein (EGFP) reporter cell line, 46C , in culture medium that suppresses alternative fates. Sox1 is the earliest known specific marker of neuroectodermal precursors that distinguishes these proliferating precursor cells from more terminally differentiated neurons and glia in vitro. Early neural development is known to involve a number of HS-dependent factors, in particular members of the FGF family, which are essential for early proliferative and specification events [15, , –18]. The commitment of pluripotent ES cells to Sox1-EGFP-expressing neural progenitor cells (NPCs) is dependent on autocrine production of FGF4 . This pathway therefore provides an excellent system to study the relationship among HS structure, enzyme expression patterns, and HS function.
We have investigated HS from 46C Sox1-EGFP reporter ES cells and compared this with HS from fluorescence-activated cell sorting (FACS)-sorted Sox1-EGFP expressing NPCs. By using a recently developed large-scale reverse transcription-polymerase chain reaction (RT-PCR) screen, we have directly correlated the transcription profile of the HS biosynthetic enzymes and proteoglycan (PG) core proteins with the structure of HS as revealed by biochemical analysis and surface expression of a distinctive repertoire of sulfated motifs recognized by a panel of single chain variable fragment (ScFv) antibodies . Evidence is also presented to show that HS is essential for ES neural differentiation in vitro.
Materials and Methods
Maintenance of ES Lines
The 46C ES line was generated as described [14, 20], and the Ext1−/− line was generously provided by Prof. Dan Wells. ES cells were maintained in Knockout Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Paisley, U.K., http://www.invitrogen.com) supplemented with 10% (vol/vol) fetal bovine serum (HyClone) or 10% (vol/vol) Knockout Serum Replacement (Invitrogen), 1% (vol/vol) Non Essential Amino Acids (NEAA), 1% (vol/vol) l-glutamine, 0.1% (vol/vol) β-mercaptoethanol (β-ME) (all from Gibco), and leukemia inhibitory factor (1,000 units/ml ESGRO; Chemicon, Chandlers Ford, U.K., http://www.chemicon.com) at 37°C/ 5% CO2. Cells were grown on tissue culture plastic coated with 0.1% gelatin and passaged every 2–3 days. For experiments with added heparin, Innohep (Leo Laboratories, U.K., http://www.leo-pharma.com) was partially digested with heparinase I, and degree of depolymerization >24 (>dp24) oligosaccharides were pooled and used at 1 μg/ml.
46C Neural Differentiation Assay
Cells were plated in N2B27 medium (a 1:1 mixture of DMEM/Ham's F-12 [Gibco, Paisley, U.K., http://www.invitrogen.com] supplemented with modified N2 and neurobasal medium supplemented with B27 [all from Gibco] plus bovine serum albumin [BSA] fraction V [50 μg/ml] and l-glutamine [500 μM]), on gelatin-coated tissue culture plastic at 1 × 104 cells per cm2. Medium was replaced every 2 days. For monitoring green fluorescent protein (GFP), cells were trypsinized, resuspended in 0.1% formaldehyde/phosphate-buffered saline (PBS), and analyzed on a Becton Dickinson FACScan (Franklin Lakes, NJ, http://www.bd.com).
Fluorescent Sorting of ES Cells
46C ES cells were differentiated in N2B27 medium for 7 days and dissociated with Dissociation Buffer (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), and Sox1-positive NPCs were separated from the negative population by GFP fluorescence in a FACSVantage sorter (Becton Dickinson). Cells were pelleted, and total RNA was extracted with the RNeasy Mini Kit (Qiagen, Hilden, Germany, http://www1.qiagen.com).
Two micrograms of total RNA was used for cDNA synthesis with Superscript reverse transcriptase using a random hexamer primer according to the manufacturer's instructions (Invitrogen). cDNA generated from 20 ng of total RNA was used in each reaction in to amplify 200–500-base pair intron-spanning fragments (when possible) of the targeted genes. HotstarTaq (Qiagen) was used under the following conditions: 95° for 15 minutes followed by 35 or 40 cycles of 95° for 30 seconds, 55° for 30 seconds, and 72° for 1 minute. Products were resolved on 2% agarose gels and scored as present or absent. Conclusions are drawn from duplicate RT-PCRs from at least two independent RT reactions. Genes that showed potential transcription regulation by this semiquantitative method were subsequently analyzed by quantitative PCR, as described below. The semiquantitative reverse transcription-PCR primer list is given in supplemental online data.
Quantitative PCR was performed using SYBR Green master mix with ROX Reference Dye (Invitrogen). Reactions contained 50 ng of cDNA and each primer at 200 nM in 20 μl total volume. Conditions were as follows: 95°C for 10 minutes followed by 50 cycles of 95°C for 15 seconds, 55°C for 30 seconds, and 72°C for 1 minute. Fold modulation was calculated using the comparative cycle threshold (Ct) method from duplicate reactions. Ct is the amplification cycle at which fluorescence exceeds the background by 10-fold. Each additional cycle required to reach Ct reduces relative copy number by one-half. By comparing Ct values of undifferentiated and differentiated cells, fold modulation was determined. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the normalizer, Ct values for GAPDH were unchanged upon differentiation (undifferentiated Ct = 15.24 ± 0.11; differentiated Ct = 15.22 ± 0.08). The primers used for quantitative PCR are described in supplemental online Table 2. The Center for AIDS Research Genomics Core Laboratory at University of California San Diego performed this analysis.
Reactions were set up using SensiMix (dT) (Quantace, Finchley, U.K., http://www.quantace.com) according to the manufacturer's instructions. An ABI Prism 7900 Sequence Detection System (AME Bioscience, Sharnbrook, UK, http://www.amebioscience.com) was used to quantify gene expression. Reactions were performed in triplicate, and data were analyzed using the relative quantification method, normalized to two in-house housekeeping genes, L19 and PUM1, according to instructions. Primers for Oct3/4 were as follows: forward, 5′-GTTGGAGAAGGTGGAACCAA; and reverse, 5′-CTCCTTCTGCAGGGCTTTC. Fluorescent probes provided by the Molecular Biology Core Facility, Paterson Institute for Cancer Research.
Periplasmic fractions of vesicular stomatitis virus (VSV)-tagged AOB408, MPB49V, HS4C3, HS4E4, EV3C3, and RB4EA-12 were prepared as described . Briefly, a phage library was incubated with HS immobilized onto a tube. Nonadherent phages were removed, and phages displaying antibodies specific for HS were recovered and multiplied. This biopanning procedure was repeated several times. Selected phages were then analyzed for anti-HS antibodies by enzyme-linked immunosorbent assay and immunohistochemistry. Cells were harvested using Cell Dissociation Buffer (Gibco) to preserve cell surface architecture, fixed in 1% formaldehyde/PBS for 10 minutes at 4°C, and then incubated for 1 hour at 4°C with ScFv antibody (1:10), washed twice with PBS, followed by 45 minutes with mouse monoclonal anti-VSV glycoprotein (Sigma-Aldrich) (1:1,000), washed twice with PBS, and followed by 45 minutes with phycoerythrin (PE)-conjugated anti-mouse IgG1, all in 0.1% (wt/vol) sodium azide/ 0.2% (wt/vol) BSA. The known specificities of the antibodies are as follows: AOB408, IdoA(2S)-GlcNS(6S); MPB49V, control, non-HS-binding antibody; HS4C3, IdoA(2S)-GlcNS(6S,3S) ; HS4E4, N-sulfation, probably no 6-0 sulfation; EV3C3, needs IdoA; RB4EA-12, GlcA/IdoA-GlcNS(6S).
FGF-2 Binding Assays
Cells were harvested using Cell Dissociation Buffer (Gibco), fixed in 1% formaldehyde/PBS for 10 minutes at 4°C, and incubated with recombinant FGF-2 (10 ng/ml; R&D Systems Inc., Minneapolis, http://www.rndsystems.com) in 0.1% (wt/vol) sodium azide/0.2% (wt/vol) BSA for 2 hours at 4°C. Cells were washed twice with PBS and incubated with rabbit anti-FGF-2 (5 μg/ml; R&D Systems) for 1 hour at 4°C, followed by two additional washes and incubation with anti-rabbit IgG PE (1:100 dilution; Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) for 45 minutes at 4°C. Cells were washed and fluorescence was measured in a Becton Dickinson FACScan. Control experiments included addition of FGF-2 in the presence of 1 μg/ml HS (Celsus, Cincinnati, http://www.heparin.co.uk) and staining of cells with anti-FGF-2 without prior addition of recombinant growth factor. The protocol was modified from Li et al. .
Isolation and Characterization of Metabolically Radiolabeled HS from 46C Cells
HS preparation and analyses were conducted and presented as previously described . A detailed description is given in the supplemental online Methods.
Cell lines differentiated in six-well plates were fixed in 4% paraformaldehyde for 10 minutes prior to staining. Fixed cells were blocked by incubation with 1% goat serum, 1 mg/ml BSA, 0.1% Triton X-100 in PBS for 30 minutes. Cells were incubated with primary antibody (Oct4, 1:100; β3 tubulin, 1:200) for 1 hour followed by five washes with PBS. Secondary antibodies (AlexaFluor 488- or 546-tagged goat anti-mouse, anti-rabbit, or anti-rat, at 1:1,000) were then incubated for 1 hour. Following an additional five washes, DAPI was added to the plates for 5 minutes, followed by a final five washes. Plates were mounted in Pro-Long antifade mountant (Molecular Probes Inc., Eugene, OR, http://www.probes.invitrogen.com) and viewed using an Olympus BX51 inverted fluorescence microscope (Olympus, Tokyo, http://www.olympus-global.com). Images were processed using Adobe Photoshop, version 6.0 (Adobe Systems Inc., San Jose, CA, http://www.adobe.com).
Neural Differentiation of ES Cells in a Monolayer System
The Sox-1-EGFP ES cell line (46C) has previously been used to demonstrate the use of a monolayer differentiation protocol for in vitro formation of NPCs from ES cells . We confirmed the reported Sox1 expression pattern during differentiation of 46C ES cells following replating of cells into N2B27 media. A GFP-positive subpopulation was observed by flow cytometry after 72 hours and reached a peak of ∼55% by days 4–5, after which they declined, falling to ∼20% by day 10 as cells committed to neuronal and glial lineages. The use of FACS to separate Sox1-EGFP-expressing NPCs from the Sox1-negative cells was vital for these studies, enabling us to prepare both RNA and HS from the same discrete populations. Cells were separated on the basis of GFP expression 7 days after plating in N2B27 and gated to ensure clearly defined populations. For generation of radiolabeled HS, 3H glucosamine and 35S sulfate (50 μCi/ml) were introduced into the media from day 5.
Differential Changes in Gene Expression
We compared the transcription profiles of ES cell and day 7 Sox1-EGFP-expressing NPC populations. Total RNA isolated from the cell populations was screened using RT-PCR to assay for the presence of message from a wide range of genes encompassing all GAG synthetic enzymes and proteoglycan core proteins (Table 1). Interestingly, the initial screen of ES cells demonstrated that message was present for a wide variety of PG core proteins, including many that have been shown to have tightly regulated spatiotemporal distribution in vivo. These include HSPG syndecan 3 and chondroitin sulfate proteoglycans (CSPGs) testican  and leprecan . However, the CSPG NG2, often used as a neural-specific marker , was found only in the day 7 cells (both GFP-positive and GFP-negative). Similarly, the majority of HS biosynthetic enzyme isoforms were found to be expressed in ES cells. However, potential changes in expression levels of certain HS sulfotransferases with neural differentiation were revealed; these enzymes were quantitatively analyzed by real-time PCR. All enzymes analyzed by this method exhibited a decrease in Ct value in the NPCs from ES cells (supplemental online Fig. 1); for clarity, this is presented here as fold increase in mRNA levels of the enzymes in NPCs compared with ES cells (Fig. 1A). The most significant increases were seen in the more restricted isoforms NDST4, 3OST-3A, and 3OST-5 (9,000-, 27.67-, and 23.92-fold increases, respectively). These enzymes were virtually absent from ES cells. Message levels for NDST3, 6OST-2, and 6OST-3 were increased by 18-, 2.7-, and 8-fold, respectively. These increases in enzyme transcription levels may result in altered HS structural patterns in NPCs differentiated from ES cells.
Table Table 1.. The panel of genes included in the semiquantitative reverse transcription-polymerase chain reaction screen
Changes in Antibody-Binding Properties of Sox1-EGFP-Expressing NPCs Reflect a Shift in HS Sulfation Patterning
We wanted to investigate how potential changes in HS structure might influence the HS-mediated ligand-binding properties of ES cells as they undergo differentiation. For this, we exploited a panel of phage-display antibodies that have previously been used in studies of tissue-specific HS expression . These antibodies recognize patterns of sulfation in the HS chain, the content of which is well-defined for some antibodies and less so for others (described in Materials and Methods). We used flow cytometry to quantify the binding of these antibodies to ES cells (Fig. 1B–1G, black traces) and Sox1-EGFP-expressing NPCs (Fig. 1B–1G, green traces) derived after 7 days in N2B27 medium. MPB49V was used as a non-HS-binding control (Fig. 1C). Antibodies AOB408 and HS4E4 (Fig. 1B, 1E) showed a broad distribution, with little change between ES cells and Sox1-EGFP-expressing NPCs. Antibodies HS4C3 and EV3C3 (Fig. 1D, 1F) displayed modest increases in binding to Sox1-EGFP-expressing NPCs compared with the parental ES cells. Strikingly, however, antibody RB4EA-12, which has a specific requirement for 6S as part of its binding epitope, demonstrated a dramatic increase in epitope expression in Sox1-EGFP-expressing NPCs, across the whole population of cells, clearly distinct from the ES cell population (Fig. 1G). Used as a panel in this way, the antibodies allow the generation of a characteristic “fingerprint” of binding that we found to be reproducible and that indicates that HS ligands will have differential binding to the two cell populations.
Changes in HS-Mediated FGF2 Binding During Neural Differentiation
To demonstrate the evolving ability of cell surface HS to interact with signaling molecules as the ES cells differentiate, we used recombinant FGF-2 to probe for HS-mediated FGF-2 cell surface binding. FGF-2 was allowed to bind to HS on ES cells and on cells undergoing differentiation in N2B27 medium and then visualized with anti-FGF2 followed by incubation with PE-conjugated secondary antibody for detection by FACS. The proportion of HS able to bind FGF-2 could then be compared between cell populations. There was no change in the amount of FGF-2 bound to HS on ES cells and on cells at 24 and 48 hours of N2B27 differentiation (Fig. 1H). However, concomitant with Sox1-EGFP expression (detected at 72 hours), cells displayed a notable difference in their ability to bind FGF-2 (Fig. 1I). In all cases, this binding could be competed by the coincubation of Celsus HS with recombinant FGF-2 (Fig. 1J). These data suggest that the changes in sulfation patterning detected with the antibodies have functional consequences in terms of the binding of growth factors and potential modulation of signaling.
Structural Characterization of HS
Preparation of HS.
Metabolically 3H/35S-radiolabeled HS was isolated from 46C ES cells and from the Sox1-EGFP-expressing NPCs and Sox1-EGFP-negative cells taken from day 7 of the N2B27 assay, separated by FACS as described above. Pronase-digested PG preparations were isolated from culture media and a cell extract. Following digestion with Chondroitinase ABC, HS chains were separated by chromatography on a CL-6B column.
Probing the Level and Distribution of N-Sulfated Residues: Low pH Nitrous Acid Depolymerization.
Medium- and cell-derived HS chains from day 0 and day 7 cell populations were cleaved at GlcNS residues by low-pH nitrous acid . Separation of the oligosaccharides on a BioGel P10 (Bio-Rad, Hemel Hempsted, U.K., http://www.biorad.com) column allows an analysis of the proportion of GlcNS-containing disaccharides within a chain and provides information on domain organization. Regions of contiguous N-sulfated disaccharides (S domains) will be depolymerized to disaccharides (dp2), with regions of alternating GlcNS and GlcNAc residues (NA/NS domains) generating tetrasaccharides (dp4). The distribution of sized saccharides is summarized in Table 2. The profiles obtained indicate a typical domain structure for all HS chains analyzed. However, HS from ES cells was found to be one of the most poorly N-sulfated HS species characterized to date, with only 30%–33% of the chain composed of N-sulfated disaccharides (compare with ). The proportion of GlcNS residues increased significantly with neural differentiation (Table 2), particularly for contiguous GlcNS residues (i.e., those within S domains).
Table Table 2.. Proportion and distribution of N-sulfated residues in heparan sulfate purified from day 0 and day 7 preparations, determined by low-pH nitrous acid depolymerization and fractionation of oligosaccharides by P-10 chromatography
This analysis demonstrated a trend of increased N-sulfation with differentiation and, in addition, highlighted increases in 6-O-sulfation and, to a lesser extent, 2-O-sulfation in all fractions analyzed (Fig. 2C). These increases were manifest in all the sulfated disaccharides except the trisulfated disaccharide UA(2S)-GlcNS(6S), which was similar (at ∼2.5%) for all the cell fractions (Fig. 2A, 2B), in good agreement with the binding data for antibody AOB408, which requires the UA(2S)-GlcNS(6S) motif for recognition. The pattern of sulfation within the cell-derived material was that day 7 Sox1-EGFP-negative cells contained HS, which was typically less sulfated than the Sox1-EGFP-positive population but more sulfated (2-O- and 6-O- in addition to N-sulfation) than ES cell HS.
K5 Lyase Digestion.
K5 lyase sites reside exclusively in N-acetylated regions, and this leads to excision of composite regions of sulfation . Oligosaccharides resulting from digestion were separated by fractionation on Bio-Gel P-10. K5 lyase digestion of day 0 (medium-derived) HS generated a characteristic distribution of oligosaccharides indicating a low-sulfated HS type (Fig. 3A) with a relatively large contribution from di- and tetrasaccharides. The difference in the profile of oligosaccharides generated by K5 lyase digestion of the day 7 (medium-derived) HS was dramatic, with considerably more material eluting in VO of the column and proportionally fewer disaccharides (Fig. 3B). However, when the products of digestion were compared using CL-6B fractionation (Fig. 4B), which allows superior resolution of high molecular weight (MW) fragments excluded from P-10, little difference was observed between their elution positions. Therefore, the K5 lyase data, together with the low-pH nitrous acid data, suggest that there is a clear increase in the level of sulfation of the composite domains (S-domains plus transition zones) in HS produced as ES cells differentiate to NPCs but that the overall organization within the chains does not significantly alter.
Analysis of Sulfated Domain and IdoA(2S) Distribution Using Heparinase III and Heparinase I Digestion.
To investigate S-domain structure of the HS chains, medium- and cell-derived material was digested with heparinase III, which cleaves the chain mainly at GlcNAc/NS(+/−6S)-IdoA/GlcA residues, its action blocked by the presence of IdoA(2S). Digested material was separated on a BioGel P10 column. In good correlation with the low-pH nitrous acid and K5 lyase profiles, the day 0 preparations (Fig. 3C, 3E) produced profiles typical of a low-sulfated HS type. Heparinase III predominantly generated disaccharides, with few longer oligosaccharides, indicating a low number of 2-O-sulfate containing sequences in the chain. The HS from day 7 medium (Fig. 3D) and Sox1-EGFP-positive FACS-sorted cell extract (Fig. 3F) produced very similar profiles, with only minor differences in oligosaccharide distribution, indicating that levels of 2-O-sulfation were likely to be similar between preparations.
Heparinase I has a more restricted site of action within the HS chain, cleaving only at GlcNS/NAc(+/−6S)-IdoA(2S) residues. It therefore cleaves predominantly within the highly sulfated heparin-like S-domains and has been used in combination with K5 lyase digestion to position the transition zone sequences within the polymer . HS chains from day 0 and day 7 preparations were separated by fractionation on CL-6B before and after digestion with heparinase I (Fig. 4C). By comparing the shift in the profiles before and after digestion (compare Fig. 4C with 4A, undigested material), it can be seen that the heparinase-oligosaccharides from the day 7 material represent a considerably shorter population than those from the day 0 material. Although there is partial overlap in the profiles, it is clear that day 7 HS yields heparinase I-resistant fragments that are skewed to the lower end of the molecular size range, ∼5.5 kDa compared with ∼8 kDa for the day 0 HS (calculation of molecular mass is given in supplemental online Methods). Interestingly, the resistant fragments resulting from either K5 lyase digestion or heparinase I digestion of the day 7 material are of very similar sizes, suggesting equidistant but offset positioning of these enzyme cleavage sites within a mature, sulfated chain as depicted in the model (Fig. 4). The smaller size of the heparinase I fragments in day 7 Sox1-positive NPCs reflects the increase in IdoA(2S) residues in the differentiated population.
Cell Surface HS Is Required for Differentiation of ES Cells into β3-Tubulin-Positive Neuronal Cells
EXT1 forms half of the HS polymerase complex; along with EXT2, it catalyzes addition of repeating GlcA-GlcNAc residues to the nascent HS chain. The EXT1−/− mouse is devoid of HS, and embryos die prior to gastrulation . An ES cell line generated from these embryos was used to detail the requirement for cell-surface HS during N2B27-mediated neural differentiation. The HS-deficient ES cells could be maintained in normal ES culture conditions with no apparent loss of pluripotency, as assessed by morphology and Oct4 (a marker of pluripotency in ES cells) expression (data not shown). By day 4 of the N2B27 assay, wild-type 46C cells were found by real-time PCR to have significantly downregulated expression of Oct4 compared with day 0 and to have further decreased expression by day 8 (Fig. 5A). EXT1−/− cells, however, showed only a minor reduction in Oct4 expression by day 4, with little further change by day 8. Interestingly, addition of soluble heparin (a highly sulfated variant of HS) at 1 μg/ml to the culture medium of EXT1−/− cells resulted in partial restoration of the wild-type phenotype, with levels of Oct4 expression at day 8 below those of 46C. In agreement with this finding, staining of 46C and EXT1−/− cells during differentiation demonstrated a clear difference in the nuclear levels of Oct4 present in day 8 cultures (Fig. 5B, iv and v). In the absence of a Sox1-GFP reporter in the EXT1−/− cell line, cultures were stained for neural marker β3-tubulin to investigate the success of neural differentiation. Although by day 8, 46C cells expressed high levels of β3-tubulin, no positive staining was seen within the EXT1−/− cultures (Fig. 5B, i and ii). Again, soluble heparin was found to partially restore the wild-type phenotype to EXT1−/− cells, with some areas of β3-tublin staining now apparent (Fig. 5B, iii). However, β3-tubulin expression in EXT1−/− cultures treated with exogenous heparin never attained the levels observed in 46C cultures (Fig. 5B, compare i and iii). Taken together, these data suggest that endogenously produced HS is necessary for differentiation into neural cells and that in its absence cells retain pluripotency under conditions that would otherwise promote their differentiation.
Variable sulfation patterns in HS chains are critical for the specific recognition and activation of many of the peptide growth factors and morphogens that control cell growth and differentiation in the developing embryo [2, 31]. Many experimental systems, including use of saccharide libraries  and developmental models , confirm the importance of the correct sulfation pattern for HS function. The structural heterogeneity of HS is tightly regulated, with a recent study of murine HS showing a remarkable, consistent tissue-to-tissue variation in disaccharide composition . In the current study, the novel use of the 46C reporter line allowed isolation of a defined cell population, the Sox1-EGFP-expressing NPCs, that was used to generate both RNA for transcript analysis and metabolically radiolabeled HS. This enabled us to address HS structure at a distinct stage of neural differentiation and to compare it with the parental ES cell population.
The HS isolated from Sox1-EGFP-expressing NPCs was structurally distinct from that of undifferentiated ES cells. Both N- and O-sulfation were increased in the differentiated population, with considerably more material from the GFP-positive population resistant to extensive K5 lyase digestion than that from ES cells. This indicates that the increased N-sulfation is focused into sulfated domains and transition zones rather than occurring sporadically throughout the chain. When compared with HS from a variety of murine tissues, ES cell HS can be seen to be one of the least sulfated species studied to date . The data from our analysis are also in good agreement with the recent disaccharide analysis of HS from 129 and C57 ES cells by Holmborn et al. . The increased sulfation observed in the Sox1-EGFP-positive population, however, makes the HS from these NPCs more similar to HS from tissues such as the cerebellum and cerebrum . In all of the analyses discussed above, material isolated from the GFP-negative sorted cell population was found to exhibit characteristics lying between those of the day 0 ES cell population and those of the GFP-positive sorted cells. This likely reflects the heterogeneous nature of this population, composed of ES cells that have not differentiated (and are therefore Sox1-negative), non-neural differentiated cells, and neural cells that have differentiated beyond the Sox1-expressing stage .
Uniquely, we have directly correlated HS structure with the HS biosynthetic/modification enzyme transcription profile of the ES and neural cell populations. ES cells transcribe most of the HS biosynthetic enzyme isoforms, and the NPC expression profile retains all those enzyme messages that were present in ES cells; novel transcripts include the enzymes absent in ES cells, NDST4, 3OST-3A, and 3OST-5. This semiquantitative analysis allowed the selection of a smaller panel of genes that were analyzed by quantitative PCR. The most striking finding was the increase in all the NDST enzymes, especially NDST4 and NDST3 (Fig. 1A; supplemental online Fig. 1). NDST4 expression is detected only in the brains of adult mice, unlike NDST1 and NDST2, which are found in most cell types . The increase in N-sulfation and enhanced heparinase I susceptibility of the Sox1-EGFP-expressing NPC surface HS compared with ES cells may therefore be at least in part mediated by the activity of NDST4.
The increase in transcription of 6OST-2 and 6OST-3 is in agreement with the increased overall 6S revealed by the disaccharide analysis data. Interestingly, levels of UA(2S)-GlcNS(6S) were unchanged, suggesting some selectivity in positioning of 6S residues by 6OSTs. Message for the cell-surface Sulf enzymes was found in both ES cells and NPCs, and these enzymes may have some influence on final 6-OS distribution of the HS chains. The two 3OST isoforms whose expression was induced with neural differentiation displayed increases of 24–28-fold. The previously published expression pattern of 3OST-5 is particularly striking, with levels in fetal brain greatly in excess of those of all other tissues studied ; 3OST-3a and 3OST-5 are expressed in the adult brain . Although unable to directly quantify levels of 3-sulfation within the isolated HS species, an indirect assay (binding to antithrombin; supplemental online data) together with the increased binding of the ScFv antibody HS4C3, which is able to specifically recognize 3-O-sulfated epitopes , suggested that there are increased levels of 3-O-sulfated disaccharides in NPC HS. Taken together, these examples indicate that a distinctive NPC HS is synthesized by Sox-1-positive cells, and this is likely to be the result of an altered composition of the biosynthetic enzyme complex.
Two complementary FACS-based analyses demonstrated the consequences of the evolving HS structure described in this study for interaction with protein factors. First, we used a panel of ScFv phage display-derived antibodies that recognize subtly different sulfated HS epitopes. Although binding requirements for these antibodies are not fully defined, their utility when used as a panel for FACS is to allow the rapid and nondestructive analysis of changes in cell-surface HS sulfation. The antibody that showed the most dramatic increase in reactivity, RB4EA-12, has a specific requirement for 6S as part of its binding epitope , which correlates well with the increased 6S seen by disaccharide analysis. Interestingly, one of the antibodies for which 6S has been found to be inhibitory (HS4E4) showed no significant change in binding between ES and Sox1-EGFP-expressing NPC populations, whereas the other, HS4C3, which has an additional requirement for 3S and is dependent on the activity of HS3OST-3a and HS3OST-5 , showed increased binding to Sox1-EGFP-expressing NPCs. In addition, we have directly demonstrated that changing the HS patterns results in an altered ability to bind FGF-2 (the archetypal HS-binding growth factor) by flow cytometry. We observed a decrease in FGF-2 binding to HS concomitant with Sox1 upregulation as ES cells commit to NPCs in N2B27 medium. It has previously been shown that specific FGF family members bind to similar, but distinct, patterns within HS chains. FGF-2 binds to S-domains in HS, showing a specific reactivity with sequences of IdoA(2S)-GlcNS repeats (a common motif within HS); however, 6-sulfate groups are not required for HS recognition . We have used FGF-2 as a probe in this instance, to demonstrate that the changes in HS patterning detailed above between ES cells and NPCs are sufficient to alter one example of HS-dependent growth factor binding.
Many factors currently identified as being involved in the maintenance of pluripotency and the early differentiation of ES cells (e.g., BMP4 and FGF4 ) are dependent on HS for their activity. We propose that by expressing a low sulfated form of HS, ES cells may be protected from signaling via many of these factors. The behavior of the HS-deficient EXT1−/− ES cells supports this hypothesis. Cultures of these cells are maintained in a pluripotent state; however, they are unable to undergo neural differentiation (as detected by appearance of β3 tubulin-positive cells) and retain Oct4 expression beyond that observed in wild-type cells. Although soluble heparin can “rescue” differentiation, the deficiency of stage-specific HS motifs probably prevents optimal response of the cells in the N2B27 assay. This offers an interesting opportunity in an in vitro system to use defined HS species and saccharides of known structure to determine the essential features of the HS coreceptor for neural differentiation.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank the University of California San Diego, Center for AIDS Research Genomics Core Laboratory (Director, Dr. Jacques Corbeil; Grant 5P30 AI36214) and the San Diego Veterans Medical Research Foundation for assistance with quantitative PCR. We also acknowledge Prof. Marion Kushe-Gullberg (University of Bergen, Bergen, Norway) for invaluable assistance with the antithrombin-binding assays and the Core Facilities of the Paterson Institute (Manchester, United Kingdom) for help. This work was funded by Cancer Research UK (C.E.J., G.R., J.T.G., and C.L.R.M.), the Medical Research Council (C.E.J.), the British Council UK-Netherlands Partnership Programme in Science (G.t.D. and C.L.R.M.), the Dorothy Hodgkin Postgraduate Award Scheme (A.L.W.), Biotechnology and Biological Sciences Research Council (A.S. and V.W.), and Human Frontier Science Program Organisation (M.S.). M.S. is currently affiliated with the Division of Cell and Developmental Biology, University of Dundee, Dundee, United Kingdom; C.L.R.M. is currently affiliated with the School of Materials, Materials Science Centre, University of Manchester, Manchester, United Kingdom; A.S. is currently affiliated with the Welcome Trust Centre for Stem Cell Research, Cambridge, United Kingdom.