Heparan Sulfate/Heparin Oligosaccharides Protect Stromal Cell-derived Factor-1 (SDF-1)/CXCL12 against Proteolysis Induced by CD26/Dipeptidyl Peptidase IV*

Stromal cell-derived factor-1 (SDF-1) is a CXC chemokine that is constitutively expressed in most tissues and displayed on the cell surface in association with heparan sulfate (HS). Its numerous biological effects are mediated by a specific G protein-coupled receptor, CXCR4. A number of cells inactivate SDF-1 by specific processing of the N-terminal domain of the chemokine. In particular, CD26/dipeptidyl peptidase IV (DPP IV), a serine protease that co-distributes with CXCR4 at the cell surface, mediates the selective removal of the N-terminal dipeptide of SDF-1. We report here that heparin and HS specifically prevent the processing of SDF-1 by DPP IV expressed by Caco-2 cells. The level of processing increases with the level of differentiation of these cells, which correlates with an increase of DPP IV activity. A mutant SDF-1 that does not interact with HS is readily cleaved by DPP IV, a process that is not inhibited by HS, demonstrating that a productive interaction between HS and SDF-1 is required for the protection to take place. Moreover, we found that protection depends on the degree of polymerization of the HS sulfated S-domains. Finally a structural model of SDF-1, in complex with HS oligosaccharides of defined length, rationalizes the experimental data. The mechanisms by which HS regulates SDF-1 may thus include, in addition to its ability to locally concentrate the chemokine at the cell surface, a control of selective protease cleavage events that directly affect the chemokine activity.

Stromal cell-derived factor-1␣ (SDF-1␣) 1 /CXCL12, a secreted protein of 68 amino acids, is a member of the CXC chemokine subfamily. SDF-1 binds to CXCR4, a G-protein-coupled receptor, and coordinates a diverse array of cellular functions, in particular the control of trafficking and homing of various subsets of hematopoietic cells from the blood stream to specific anatomical sites (1)(2)(3). However, a large body of evidence implicates SDF-1 outside the regulation of leukocyte migration, i.e. SDF-1 and its receptor have a critical role in determining the metastatic destination of tumor cells (4,5), the chemokine competes with the human immunodeficiency virus (HIV) for binding to CXCR4 (6,7), and it also acts as a morphogen during the developmental process (8). Inactivation of either SDF-1 or CXCR4 induces a number of defects including severe abnormalities in the immune, circulatory, and central nervous systems (9 -11). Finally SDF-1 is up-regulated in several diseases such as rheumatoid arthritis, atherosclerosis, autoimmune disease, or allergic disorder (12). Consistent with this broad range of activities, CXCR4 is widely expressed in a large variety of cells and tissues. SDF-1 is also constitutively expressed by a large number of cells (13)(14)(15)(16), and it has been proposed that proteolytic degradation of the chemokine could serve important regulatory function. Structure-function analysis of SDF-1 has identified the N-terminal region as important for CXCR4 binding and activation (17). The N-terminal domain of the chemokine SDF-1 was shown to be efficiently processed by a number of proteases, including matrix metalloproteinase (18), cathepsin G (19), elastase (20), and dipeptidyl peptidase IV (DPP IV)/CD26 (21)(22)(23). Interestingly, among these enzymes, CD26, a cell surface serine protease, co-distributes and co-immunoprecipitates with CXCR4, suggesting a functional relationship with SDF-1 (24). CD26 removes the N-terminal dipeptides from a number of proteins having either a Pro or an Ala residue in the penultimate position (25), a characteristic shared by several chemokines (26,27). The processing of SDF-1 by CD26, giving rise to SDF-1 , results in a loss of lymphocyte chemotactic activity and signaling through CXCR4. However, SDF-1 (3-68) still interacts with its receptor and is able to desensitize for Ca 2ϩ responses toward intact SDF-1, although with a considerably lower efficiency (17,22).
Among the molecules that also regulate chemokines, glycosaminoglycans (GAGs), in particular heparan sulfate (HS), play an important role. Different studies suggest that GAGs immobilize and enhance local concentrations of chemokines, promoting their oligomerization and facilitating their presentation to the receptors (28,29), and in vivo data suggest that, within tissues, SDF-1 is sequestered by HS (30,31). This oligosaccharide, which is found ubiquitously at the cell surface and in the extracellular matrix, consists of sulfated glucosamine and iduronic residues clustered in a series of domains of variable lengths and sulfation patterns separated by regions of low sulfation and relatively uniform structure (32). SDF-1 binds with high affinity (K d of 30 nM) to the sulfated regions of HS (called S-domains), and we have shown that amino acids Lys 24 and Lys 27 have a dominant role in the binding. In addition, residues Lys 1 , Arg 41 , and Lys 43 participate in the recog-nition process but are not strictly required for the interaction to take place (33). Given the substrate specificity of DPP IV and the possible involvement of the N-terminal residue (Lys 1 ) of SDF-1 in the complex with HS, the present study investigates whether heparin or HS could interfere with the processing of the chemokine by CD26. For that purpose, human colon adenocarcinoma epithelial cells (Caco-2), which express distinct levels of CD26/DPP IV depending on their degree of differentiation (34), were used. We first highlighted the enzymatic activity of CD26 on solubilized Caco-2 cell membranes and used this material as a source of CD26 to show that heparin-and HS-derived oligosaccharides inhibit CD26-induced cleavage of SDF-1, depending on their degree of polymerization.
Cell Culture-Caco-2 cells were obtained from the European Collection of Cell Cultures. They were maintained in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, 1% non-essential amino acids, and 2 mM Glutamax. Cells were grown at 37°C under a 5% CO 2 atmosphere. Caco-2 cells were routinely passaged every 6 days and kept 12 days post-confluency to induce differentiation as described (34). The medium was changed regularly during cell differentiation.
Preparation of Cellular Membranes and Solubilization-Caco-2 cells were detached with EDTA (Versene) and incubated for 20 min in hypotonic buffer (10 mM Tris, pH 7.6, containing protease inhibitors) and then disrupted by Dounce homogenization. The homogenate was centrifuged at 3000 ϫ g for 10 min. The supernatant was collected and further centrifuged at 220,000 ϫ g for 30 min. The cell membranecontaining pellet was then solubilized in 100 mM Hepes buffer (pH 7.6) and 1% Triton X-100 and cleared by centrifugation at 220,000 ϫ g for 30 min. The supernatant was aliquoted and used as a source of CD26/DPP IV activity.
Assay for DPP IV Activity on Non-differentiated and Differentiated Caco-2 Cells-DPP IV activity was assayed using Gly-Pro-pNA as a chromogenic substrate. Caco-2 cells were seeded into 96-well flat-bottomed plates (2 ϫ 10 5 cells per well) in 200 l of Eagle's minimum essential medium with 10% fetal calf serum until confluence or 12 days post-confluence. Cells were incubated at 37°C, for different times, with 0 -4 mM Gly-Pro-pNA in 120 mM NaCl, 5 mM KCl, 120 mM MgSO 4 , 8 mM glucose, and 10 mg/ml bovine serum albumin. Enzymatic activity was determined by monitoring the release of pNA in the supernatant with spectrophotometry at 405 nm. Cell-free and substrate-free blanks were run in parallel. Inhibition of Gly-Pro-pNA cleavage was determined in presence of specific inhibitor, diprotin A, preincubated for 15 min at 37°C prior to the addition of the substrate. The assay was also performed using either cellular membranes or solubilized membranes from non-differentiated and differentiated Caco-2 cells. Tests were run in triplicate. To assess a possible effect of GAGs on DPP IV activity, solubilized membranes were preincubated with different concentrations of heparin or heparan sulfate and then incubated in the presence of the chromogenic substrate (2 mM).
SDF-1 Synthesis-Biotinylated SDF-1 was synthesized by the Merrifield solid phase method on a fully automated peptide synthesizer using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry as described previously (35). Selective biotinylation at the C-terminal position was achieved by incorporating a lysine residue (Lys 68 ) bearing a 4,4-dimethyl-2,6-dioxocyclohex-1-ylidene)ethyl (Dde) protective group on the side chain. Coupling of biotin was performed on the peptide resin after Dde deprotection. A mutant SDF-1 (substitution of Lys 24 and Lys 27 by Ser 24 and Ser 27 ) was synthesized with the same method. Purity and molecular weights of wild type or mutant SDF-1 were analyzed by high performance liquid chromatography and ion spray mass spectrometry (33,35).
Preparation of Heparin-and Heparan Sulfate-derived Oligosaccharides-Heparin and heparan sulfate oligosaccharides were prepared as described previously (33). Briefly, porcine mucosal heparin (10 g) or heparan sulfate (8 g) were depolymerized with heparinase I (8 milliunits/ml) in 5 mM Tris, 2 mM CaCl 2 , 50 mM NaCl, pH 7.5, and 0.1 mg/ml bovine serum albumin for 54 h at 25°C or with heparitinase (25 milliunits/ml) in 5 mM Tris, 2 mM CaCl 2 , 50 mM NaCl, pH 7.5, and 0.1 mg/ml bovine serum albumin, respectively, for 72 h at 30°C. Enzyme digestion was stopped by heating the reaction mixture at 100°C for 5 min. Digestion products were then size-separated using a Bio-Gel P-10 column (150 ϫ 4.4), equilibrated with 0.25 M NaCl, and run at 1 ml/min. Eluted material was detected by absorbency at 232 nm. Graded series of size-uniform oligosaccharides resolved from disaccharide (dp2) to octadecasaccharide (dp18) were obtained for heparin and from dp2 to dp16 for heparan sulfate. Top fractions of each peak were pooled, dialyzed against distilled water, and quantified by colorimetric assay or weighted.
SDF-1 Processing by Solubilized Membranes Containing CD26 -Biotinylated SDF-1 (600 nM) was incubated with Triton X-100 Caco-2 cell-solubilized membranes (1 mg/ml of total protein) for 1 h at 37°C in 100 mM Hepes buffer, 120 mM NaCl, 5 mM KCl, 1.2 mM MgSO 4 , and 8 mM glucose, pH 7.6. The reaction was stopped by the addition of SDS-PAGE sample buffer and heating at 100°C for few minutes. In some cases, the DPP IV specific inhibitor, diprotin A (340 g/ml), was incubated with solubilized membranes for 15 min at 37°C prior to the addition of biotinylated SDF-1. To investigate the effect of GAGs, biotinylated SDF-1 was preincubated with different concentrations of heparin, heparan sulfate, or size-defined oligosaccharides (dp8 to dp16) prior to incubation with solubilized membranes.
SDF-1 Cleavage Analysis-SDF-1 processing was determined by SDS-PAGE and Western blotting with a monoclonal antibody (mAb K15C) that selectively recognizes the N terminus of the chemokine and with horseradish peroxidase-conjugated ExtrAvidin that detects the biotin label at the C terminus.
The reaction mixture was resolved by SDS-PAGE (14% polyacrylamide) followed by electroblotting onto a polyvinylidene difluoride membrane. The membrane was blocked in Tris-buffered saline containing 5% milk overnight at 4°C and then incubated with mAb K15C (2 g/ml) for 1 h in Tris-buffered saline containing 0.05% Tween 20 and 5% dry milk. Detection was performed using a horseradish peroxidaseconjugated anti-mouse antibody and enhanced chemiluminescence. To detect the biotin label of SDF-1, the membrane was incubated overnight in stripping buffer (20 mM glycine, pH 2, 100 mM 2-mercaptoethanol, and 0.05% Tween 20), washed, and reprobed with horseradish peroxidase-ExtrAvidin. Signals were quantified by densitometry.
Molecular Modeling-The coordinates of the crystal structure of the native SDF-1 dimer (36) were taken from the Protein Data Bank (37) (PDB code 1QG7). The amino acids missing on the N-terminal moiety of chain B (Pro 2 to Leu 5 ) were added using structural homology with chain A. Lys 1 was added on both chains, and the missing side chains of a few amino acids were modeled. The geometry of all side chains was optimized using the Tripos force field (38) in the SYBYL molecular modeling package (Tripos, St. Louis, MO). Hydrogen atoms were added and optimized, and partial atomic charges were derived using the Pullman procedure (39).
The starting conformations of oligosaccharides with alternating 2-Nsulfated, 6-O-sulfated ␣-D-Glc and 2-O-sulfated ␣-L-iduronic acid monosaccharides were taken from our previous docking study (33). Atom types and partial charges were defined according to the Pérez-Imberty-Meyer (PIM) energy parameters for carbohydrates (40) to be used within the Tripos force field. Different lengths of oligosaccharides were considered with either 8 or 12 monosaccharides and different conformations of the Lys 1 -Pro 2 N-terminal moiety. For each complex, energy minimizations were performed with constraining only the backbone of amino acids 9 -67. A distance-dependent dielectric constant was used in the calculations. Energy minimizations were carried out using the Powell procedure (41) until a gradient deviation of 0.5 kcal⅐mol Ϫ1 ⅐Å Ϫ1 was attained.

The N-terminal Domain of SDF-1 Is Processed by Dipeptidyl
Peptidase IV Activity Solubilized from Caco-2 Cells Membranes-To assess DPP IV activity in Caco-2 cells (an enterocyte-like colon cell line), we measured the production of pNA, which resulted from the cleavage of the chromogenic substrate Gly-Pro-pNA, by using freshly prepared solubilized membranes. Solubilized membranes from differentiated cells (i.e. cells maintained in a post-confluent state for 12 days in culture) displayed higher level of DPP IV activity than non-differentiated cells (Fig. 1A), which is consistent with the increase in CD26 gene expression in cells that display enterocytic differentiation (34). The addition of diprotin A, a DPP IV inhibitor, to the solubilized membranes from differentiated cells reduced the level of pNA by 85-90% (Fig. 1B), demonstrating the specificity of the assay (EDTA and Pefabloc have no effect; data not shown). The addition of heparin or HS (up to 0.5 mg/ml) to the reaction mixture did not change the amount of pNA produced (Fig. 1B), indicating that DPP IV activity was not affected by the presence of GAGs. To further demonstrate that heparin and HS did not interfere with DPP IV activity, a time course was also performed (Fig. 1B, inset) and showed that none of the GAGs either reduced or enhanced the degradation of the substrate and, thus, did not modify DPP IV activity.
SDF-1 N-terminal domain processing was then investigated following exposure of the chemokine to DPP IV activity solubilized from the membrane of either non-differentiated or differentiated Caco-2 cells. For that purpose, SDF-1 carrying a biotin group on the C-terminal residue (Lys 68 ) was incubated for 1 h with the solubilized membranes. The reaction mixture was then resolved by SDS-PAGE, and N-terminal processing of the chemokine was determined by Western blot analysis using a monoclonal antibody, K15C, the epitope of which critically contains the SDF-1 residues Lys 1 and Pro 2 (35). Incubation of the chemokine with differentiated cell-solubilized membranes led to a complete loss of the K15C immunoreactivity, indicating an efficient processing of the SDF-1 N-terminal domain (Fig. 1C). Incubation of the chemokine with non-differentiated cell-solubilized membranes (which contained a lower level of DPP IV activity) led to a partial loss of the K15C immunoreactivity, whereas the addition to the reaction mixture of diprotin A, a specific DPP IV inhibitor, restored the K15C reactivity (Fig.  1C). To ensure that each lane contained an equivalent amount of SDF-1, the Western blot membrane was stripped, and a second detection for the biotin label was performed with horseradish peroxidase-conjugated ExtrAvidin.
Heparin Prevents DPP IV-induced SDF-1 Processing-To determine the effect of heparin on SDF-1 processing, the chemokine was preincubated with a range of heparin concentration and then challenged with DPP IV solubilized from differentiated Caco-2 cell membranes. The K15C immunoreactivity that was lost in the absence of heparin was recovered in the presence of heparin and increased as a function of heparin concentration (Fig. 2, A and B). Because the presence of heparin did not inhibit DPP IV activity (see Fig. 1B), this finding indicates that heparin protected SDF-1 from processing. Quantification of the chemiluminescent signals by densitometry showed that maximum protection of the chemokine was obtained with a 2-fold molar excess of heparin (15 kDa) over SDF-1 ( Fig. 2A). Low molecular mass heparin (6 kDa) was similarly analyzed. A 5-fold molar excess was required to get maximum protection (Fig. 2B). However, it is worth noting that this material contains a proportion of molecules with molecular masses below 3 kDa (not shown) that are inactive regarding SDF-1 protection (see below). We next compared the N-terminal processing of the wild type and a mutant SDF-1 induced by DPP IV. The SDF-1 mutant used here was generated by the combined substitution of the amino acids Lys 24 and Lys 27 by Ser residues and has a strongly impaired heparin binding capacity (33). Our results showed that both the wild type and the mutated chemokines were almost completely processed. However, although a 5-fold molar excess of heparin significantly inhibited the cleavage of the wild type SDF-1, heparin was ineffective in protecting the mutant chemokine from DPP IV activity (Fig. 2C). Together, these data showed that heparin efficiently protected SDF-1 from proc-FIG. 1. DPP IV activity on solubilized Caco-2 cell membranes and SDF-1 processing. Enzymatic activity was measured on solubilized Caco-2 cell membranes by monitoring the release of pNA (absorbance at 405 nm) using the DPP IV chromogenic substrate Gly-Pro-pNA. A, membranes isolated from non-differentiated (circle) or differentiated (square) cells were solubilized with 1% Triton X-100 and incubated with different concentrations of Gly-Pro-pNA at 37°C for 30 min, after which the amount of pNA was measured. B, DPP IV activity from differentiated Caco-2 cell membranes was measured using 2 mM Gly-Pro-pNA as the substrate during a 1-h period without (0) or with 340 g/ml diprotin A (DA), 500 g/ml heparin (HP) or 500 g/ml heparan sulfate (HS). A solubilized membrane-free reaction (S) was run in parallel. The inset shows the time course of the reaction without (circle) or with 500 g/ml heparin (square) or 500 g/ml heparan sulfate (triangle). Assays were performed in triplicate. C, C-terminal biotinylated SDF-1 (600 nM) was incubated with either non-differentiated (ND) or differentiated (D) Caco-2 cell solubilized membranes for 1 h at 37°C. In some cases the DPP IV inhibitor (340 g/ml) was added to the solubilized membrane mixture (DA). Samples were separated by SDS-PAGE on a 14% polyacrylamide gel followed by electroblotting on polyvinylidene difluoride membrane. N-terminal detection (N-ter) was performed with mAb K15C (2 g/ml). The polyvinylidene difluoride membrane was incubated in stripping buffer (as described under "Experimental Procedures") and reprobed with horseradish peroxidase-ExtrAvidin to detect the biotin label at the C-terminus of the chemokine (C-ter). essing by DPP IV activity and that a complex between the GAG and the chemokine was required for such protection to take place.
Effect of GAG-derived Oligosaccharides on SDF-1 Processing by DPP IV-We then analyzed the ability of heparin or heparan sulfate-derived oligosaccharides to prevent or reduce the SDF-1 processing. For that purpose, biotinylated SDF-1 was preincubated with oligosaccharides with increased degrees of polymerization (octasacharides to hexadecasaccharides; dp8 to dp16), each at 10 M, then submitted to solubilized membrane from differentiated Caco-2 cells for 1 h at 37°C. As above, the intact N-terminal domain of the chemokine was revealed with mAb K15C after Western blotting of the reaction mixture. The results showed that heparin-derived oligosaccharides dp12 to dp16 bound to SDF-1 strongly reduced the SDF-1 N-terminal truncation, whereas dp8 and dp10 failed to prevent SDF-1 processing (Fig. 3A). Experiments performed with heparan sulfate-derived oligosaccharides yielded similar results (Fig. 3B). To rule out the possibility that small oligosaccharides could be active but differ in their concentration dependence, similar experiments were performed with dp8 in the concentration range 0 -40 M. We found that SDF-1 was fully processed even at the highest dp8 concentration (data not shown). To rationalize these observations, we built structural models of the chemokines in complex with heparin derived octasaccharides or dodecasaccharides. The approach is similar to the one described previously (29), but we used a new crystal structure of the SDF-1 dimer with a corrected conformation of the N-terminal strand (36). Two conformations were considered for the Lys 1 -Pro 2 moiety. In the first model, which presents lower potential energy than that of the second one (see below), an octasaccharide encompassed just the heparin binding site formed by the dimeric association of the chemokine and left the N-terminal domain uncomplexed (Fig. 4, A and B). In contrast, a dodecasaccharide extends away from the central binding site and permit the terminal sugar residues to interact with the N-terminal lysine residues (Fig. 4, C and D) and, as such, presumably protects the chemokine from DPP IV cleavage. A distinct conformation of the Lys 1 -Pro 2 residues resulted in maximum extension of the N-terminal strands away from the protein center. In such a fully extended model (Fig. 4, E and F),   FIG. 2. Heparin prevents the N-terminal processing of SDF-1  by DPP IV. A and B, C-terminal-biotinylated SDF-1 (600 nM) was incubated with differentiated Caco-2 cell solubilized membranes for 1 h at 37°C in the absence or the presence of a 0.5-, 1-, or 2-fold molar excess of 15-kDa heparin (HP 15 ) (A) and a 1.25-, 2.5-, or 5-fold excess of 6-kDa heparin (HP 6 ) (B). Samples were analyzed by Western blotting with mAb K15C (N-ter) and horseradish peroxidase-ExtrAvidin (C-ter) as in Fig. 1C. Signals revealed with mAb K15C (white bars) and horseradish peroxidase-ExtrAvidin (black bars) were quantified by densitometry. C, the wild type SDF-1 (WT) was treated with DPP IV in the absence (Ϫ) or the presence (ϩ) of either diprotin A (DA) or 6-kDa heparin (HP). A mutated SDF-1 (M) without heparin binding activity was also treated with DPP IV under the same conditions. The integrity of the N-terminal domain was analyzed after Western blotting with the mAb K15C. FIG. 3. Effect of heparin-and heparan-sulfate derived oligosaccharides on SDF-1 processing by DPP IV. C-terminal biotinylated SDF-1 (600 nM) was preincubated or not (0) with size-defined oligosaccharides (from octasaccharide to hexadecasaccharide, dp8 to dp16, each at 10 M) derived from heparin (A) or heparan sulfate (B) and then incubated with solubilized membranes from differentiated Caco-2 cells for 1 h at 37°C. In some cases the cell extract was preincubated with diprotin A (DA). As described in Fig. 1C, the integrity of the N-terminal domain was analyzed after Western blotting with the mAb K15C (N-ter), and the presence of SDF-1 was detected with horseradish peroxidase-ExtrAvidin (C-ter). only oligosaccharides with dp16 or more would interact with the two N-terminal lysine residues (data not shown). DISCUSSION A number of proteases efficiently process the N-terminal domain of SDF-1. These include lymphocyte-associated cathepsin G, (19), matrix metalloproteinases (18), leukocyte elastase (20), and CD26/DPP IV (21)(22)(23), which cleave the Leu 5 -Ser 6 , Ser 4 -Leu 5 , Val 3 -Ser 4 , and Pro 2 -Val 3 bonds, respectively. Consistent with the importance of the first N-terminal residues of SDF-1 for receptor activation, these cleavages convert the chemokine from a potent chemoattractant and HIV antagonist to a molecule devoid of biological activity. Because CD26/DPP IV is present on cells of many lineages but is also in the serum in a catalytically active soluble form (42), SDF-1 inactivation could occur ubiquitously in the body. In this context, we provide here the first evidence that heparin and HS, both known to bind SDF-1 with high affinity (33,35), efficiently inhibit the DPP IV-induced cleavage of the chemokine. It has been observed that GAGs did not prevent SDF-1 processing by a number of other proteases, including several matrix metalloproteinases (18). However, because matrix metalloproteinase, cathepsin G, and elastase all bind to and are regulated by GAGs (43)(44)(45)(46), these results are difficult to interpret in the context of the SDF-1/GAG complex. The present work showed that, in contrast to the above-mentioned proteases, CD26/DPP IV activity was not affected by heparin/HS molecules. This finding thus suggests a specific mechanism by which HS, along with DPP IV, regulates the SDF-1 activity (see below). From a structural point of view, this observation reinforces our previous finding suggesting that in the SDF-1/HS complex the Lys 1 residue is involved and, as such, is presumably non-accessible to the DPP IV enzyme. Furthermore, the data reported here also indicate that heparin and HS-derived oligosaccharides with a length of 12 mers (dp12) or more display protective activity. According to our model, this corresponds to the minimum length that is required to span the entire heparin binding site on the chemokine dimer, including the two N-terminal lysine residues. Oligosaccharides of smaller sizes (dp10 and dp8) that overlap the central heparin binding domain of the chemokine (i.e. Lys 24 , Lys 27 , Arg 41 , and Lys 43 ) but are too short to contact the Nterminal residues did not show significant protective activity. However, because of the flexibility of the N-terminal region of the chemokine (17,36), other conformations cannot be excluded. In a "fully extended" conformation, only oligosaccharides of dp16 or more would contact the two N-terminal residues. It is possible that free SDF-1 samples distinct conformations between the two proposed models. In view of our experimental data, showing that dp12 displays protective activity, it can be hypothesized that dp12 locks the N termini in a specific conformation in which the first lysine residues are in contact with the oligosaccharide (compare Fig. 4, panels C and E, for example). Together, the above data raise the possibility that, depending on the cellular level of expression and the fine structure of HS (i.e. length and sulfation profile of the S-domains), SDF-1 would be protected or not protected against DPP IV processing. It is now clear that, in the body, HS is expressed in a cellspecific manner, giving rise to distinct cell and tissue characteristics (47) that can be further affected during differentiation and development as well as during diseases (48). For example, it has been reported that a specific HS proteoglycan isoform of CD44, up-regulated during B cell activation, was able to interact with growth factors such as the hepatocyte growth factor but not with SDF-1 (49). On the other hand, HS has been shown to bind and concentrate SDF-1 at the surface of a number of cells, including endothelial cells, cells of the bone marrow, and cells of the cerebellar external granular layer, where it induces angiogenesis (31), triggers specific arrest of Kaposi's sarcoma herpesvirus-infected cells (50), enhances the chemokine-induced migration of cells by the formation of haptotactic gradient (51,52), or facilitates persistent occupancy and sustained down-regulation of CXCR4 used by HIV as a co-receptor, thus contributing to the HIV inhibitory effect of SDF-1 (53). Whether HS protects or does not protect SDF-1 against DPP IV in vivo is not known. However, CD26/DPP IV has been identified as the enzyme that processes SDF-1 in SDF-1-spiked blood (54). In vivo data showed that sulfated oligosaccharides can specifically displace a pool of sequestered SDF-1 from its HSPG anchors in the bone marrow, leading to its release into the circulation. Importantly, such an increase led to the chemoattraction to both white blood cells and stem/progenitor cells in the blood stream in mice and nonhuman primates, demonstrating that the oligosaccharide/SDF-1 complex remained functional (30) and, thus, presumably protected against DPP IV activity. In vivo, SDF-1 is constitutively expressed in many tissues, including mucosal epithelial cells from the genital and digestive tracts. Such tissues, the cells of which are well characterized for their high content of HS, are highly relevant for HIV transmission and propagation. HS-dependent immobilization of SDF-1 at these anatomic sites could constitute a barrier that may selectively restrain dissemination of HIV isolates that use CXCR4 as a co-receptor and contribute to the predominant propagation of CCR5 using isolates that is observed during the early phases of the infection (15). It is possible that, in addition to its ability to locally concentrate SDF-1, HS may also play a role in stabilizing and protecting the chemokine against DPP IV activity. In this context, it is worth noting that the epithelial intestinal Caco-2 cells we used in this study undergo differentiation to enterocytic like cells when maintained in a post-confluent state for 1-2 weeks. Such cells, which are important points of entry for pathogens into the body (55), express an increased level of brush border membrane-associated dipeptidyl peptidase IV (34). We indeed found that the differentiation of Caco-2 cells led to an increased ability of its membrane extract to process SDF-1. Finally, it has been reported that the differentiation of Caco-2 cell in culture is also accompanied by selective alterations in the sulfation of HS (56). It remains to be determined if such alterations could affect the binding and protection of SDF-1, allowing the preservation of its anti-HIV activity at these particular sites.