Glycosaminoglycans Regulate CXCR3 Ligands at Distinct Levels: Protection against Processing by Dipeptidyl Peptidase IV/CD26 and Interference with Receptor Signaling

CXC chemokine ligand (CXCL)9, CXCL10 and CXCL11 direct chemotaxis of mainly T cells and NK cells through activation of their common CXC chemokine receptor (CXCR)3. They are inactivated upon NH2-terminal cleavage by dipeptidyl peptidase IV/CD26. In the present study, we found that different glycosaminoglycans (GAGs) protect the CXCR3 ligands against proteolytic processing by CD26 without directly affecting the enzymatic activity of CD26. In addition, GAGs were shown to interfere with chemokine-induced CXCR3 signaling. The observation that heparan sulfate did not, and heparin only moderately, altered CXCL10-induced T cell chemotaxis in vitro may be explained by a combination of protection against proteolytic inactivation and altered receptor interaction as observed in calcium assays. No effect of CD26 inhibition was found on CXCL10-induced chemotaxis in vitro. However, treatment of mice with the CD26 inhibitor sitagliptin resulted in an enhanced CXCL10-induced lymphocyte influx into the joint. This study reveals a dual role for GAGs in modulating the biological activity of CXCR3 ligands. GAGs protect the chemokines from proteolytic cleavage but also directly interfere with chemokine–CXCR3 signaling. These data support the hypothesis that both GAGs and CD26 affect the in vivo chemokine function.


Introduction
The family of chemotactic cytokines or chemokines is constituted by a group of low molecular mass proteins that direct specific leukocyte migration in a time-and site-dependent manner in health and disease [1,2]. Chemokines are not only crucial players in basal innate and adaptive immune mechanisms, but are also implicated in a broad range of additional physiological and Inspired by these drastic effects of GAGs on CD26-mediated truncation of CXCL10, we investigated the effects of GAGs on CD26-mediated processing of the two other IFN-γ-inducible CXCR3 chemokine agonists, i.e., CXCL9 and CXCL11. Accordingly, we optimized the desalting process after the CD26 incubation and found that chemokines could be detected and quantified by mass spectrometry after pre-purification on C18 pipette tips if the GAG concentrations did not exceed 8.8 µg/mL. The three main CXCR3 agonists were incubated with CD26 in the absence or presence of 8.8, 2.64 or 0.88 µg/mL heparan sulfate, heparin, chondroitin sulfate A or chondroitin sulfate C. All these GAGs had a relative molecular mass (Mr) of 40 kDa (data not shown). After an incubation period of 2 h, CD26 activity was stopped through acidification with TFA. Samples were desalted with C18 tips and subjected to mass spectrometry. Without GAGs, almost all CXCL10 and CXCL11 was cleaved by CD26 to CXCL10  and CXCL11 , respectively. In contrast, CXCL9 was processed by the enzyme for only about 20% (data not shown). These findings were in line with previous studies which demonstrated that CXCL10, and especially CXCL11, are highly efficient CD26 substrates, whereas the half-life of CXCL9 upon incubation with the enzyme is remarkably longer [53,62]. The presence of heparin, heparan sulfate, chondroitin sulfate A or chondroitin sulfate C in the incubation mixture protected the three IFN-induced CXCR3 ligands dose-dependently from truncation by CD26 (Figure 2, Figure S1-S3). At 8.8 µg/mL heparin, heparan sulfate, chondroitin sulfate A and C completely prevented processing of CXCL9 and CXCL10 by CD26. A comparable protective effect was obtained with heparan sulfate and chondroitin sulfate C, whereas heparin and chondroitin sulfate A were less efficient in protecting CXCL11 from proteolytic truncation by CD26. Thus, the obtained results indicated a dose-dependent, GAG-mediated chemokine protection against processing by CD26 for all GAGs tested. However, minor differences in efficiency were detected between the different GAGs although they all had a comparable molecular mass. Inspired by these drastic effects of GAGs on CD26-mediated truncation of CXCL10, we investigated the effects of GAGs on CD26-mediated processing of the two other IFN-γ-inducible CXCR3 chemokine agonists, i.e., CXCL9 and CXCL11. Accordingly, we optimized the desalting process after the CD26 incubation and found that chemokines could be detected and quantified by mass spectrometry after pre-purification on C18 pipette tips if the GAG concentrations did not exceed 8.8 µg/mL. The three main CXCR3 agonists were incubated with CD26 in the absence or presence of 8.8, 2.64 or 0.88 µg/mL heparan sulfate, heparin, chondroitin sulfate A or chondroitin sulfate C. All these GAGs had a relative molecular mass (M r ) of 40 kDa (data not shown). After an incubation period of 2 h, CD26 activity was stopped through acidification with TFA. Samples were desalted with C18 tips and subjected to mass spectrometry. Without GAGs, almost all CXCL10 and CXCL11 was cleaved by CD26 to CXCL10 (3-77) and CXCL11 , respectively. In contrast, CXCL9 was processed by the enzyme for only about 20% (data not shown). These findings were in line with previous studies which demonstrated that CXCL10, and especially CXCL11, are highly efficient CD26 substrates, whereas the half-life of CXCL9 upon incubation with the enzyme is remarkably longer [53,62]. The presence of heparin, heparan sulfate, chondroitin sulfate A or chondroitin sulfate C in the incubation mixture protected the three IFN-induced CXCR3 ligands dose-dependently from truncation by CD26 (Figure 2, Figures S1-S3). At 8.8 µg/mL heparin, heparan sulfate, chondroitin sulfate A and C completely prevented processing of CXCL9 and CXCL10 by CD26. A comparable protective effect was obtained with heparan sulfate and chondroitin sulfate C, whereas heparin and chondroitin sulfate A were less efficient in protecting CXCL11 from proteolytic truncation by CD26. Thus, the obtained results indicated a dose-dependent, GAG-mediated chemokine protection against processing by CD26 for all GAGs tested. However, minor differences in efficiency were detected between the different GAGs although they all had a comparable molecular mass.

GAGs Did Not Inhibit the Enzymatic Activity of Soluble CD26 Directly
The exact enzymatic activity of the purified natural human CD26 sample was determined to be 4.6 ± 0.6 U/L (mean ± SEM, n = 3) using a chromogenic assay with Gly-Pro-p-nitroanilide (Gly-Pro-pNA) as the substrate. No pNA release was observed upon incubation of the substrate with the highest concentration of heparin DP30 in the absence of CD26. To investigate a direct effect of GAGs on the activity of the enzyme, the release of pNA was detected when Gly-Pro-pNA and CD26 were incubated in the absence or presence of 10 or 100 µg/mL heparin DP30. The CD26 activities in conditions with and without GAG were highly similar (Table 1). Thus, no evidence was found for GAGs to inhibit the proteolytic activity of CD26 directly, which was in line with a former study that reported that heparan sulfate did not inhibit the enzymatic activity of CD26 [46].

GAGs Interfered with Chemokine Signaling through CXCR3
Chemokine-induced CXCR3-signaling is associated with the release of intracellular calcium from the endoplasmic reticulum. Consequently, we reasoned that measuring the [Ca 2+ ] i after stimulation of cells with CXCL9, CXCL10 or CXCL11 with or without GAGs, would provide us with new insights on the effect of GAGs on the chemokine-induced G protein-dependent signaling through CXCR3. To this end, Chinese Hamster Ovarian (CHO) cells, stably transfected with CXCR3A and loaded with the calcium-binding fluorescent dye Fura-2, were stimulated with final concentrations of 3 ng/mL to 1 µg/mL CXCL9, CXCL10 or CXCL11. The corresponding chemokine-induced calcium responses were calculated using the Grynkiewicz equation. For CXCL10 and CXCL11, a concentration of 3 ng/mL resulted in an increase of the [Ca 2+ ] i with 106.8 nM (n = 39) and 304.5 nM (n = 18), respectively, and 3 ng/mL CXCL10 or CXCL11 was selected for further experiments in combination with GAGs. Cells were treated with CXCL10 or CXCL11 with or without 0.04 µg/mL, 2 µg/mL or 10 µg/mL GAG. Representative experiments are shown in Figure 3. The observed calcium responses were calculated as percentages of the corresponding reference values in the absence of GAGs. A dose-dependent negative correlation was found between the GAG concentration and the ability of CXCL10 and CXCL11 to evoke an intracellular calcium release through CXCR3 ( Figure 4A,B). Heparin molecules with different length were tested in combination with CXCL10 and the longer heparin molecules were more potent inhibitors of the calcium response compared to the shorter DP8 form. For the less potent CXCL9, a concentration of 1 µg/mL was selected, resulting in an increase of the [Ca 2+ ] i with 598.1 nM (n = 4). Heparan sulfate also dose dependently inhibited the calcium response induced by this weaker CXCR3 ligand ( Figure 4C). It remains to be elucidated whether the effect of GAGs on calcium signaling is due to direct binding of GAGs to chemokines, CXCR3 or both. In addition, it cannot be excluded that

Effect of Soluble GAGs on CXCL10-Mediated CD26-Positive T cell Chemotaxis In Vitro
Activated T cells express membrane-bound CD26. Thus, we wondered whether the observed GAG-mediated protection of CXCL10 against inactivation by CD26 was reflected in an increased CXCL10-induced T cell chemotaxis. To this end, we first confirmed CD26 expression and activity on

Effect of Soluble GAGs on CXCL10-Mediated CD26-Positive T Cell Chemotaxis In Vitro
Activated T cells express membrane-bound CD26. Thus, we wondered whether the observed GAG-mediated protection of CXCL10 against inactivation by CD26 was reflected in an increased CXCL10-induced T cell chemotaxis. To this end, we first confirmed CD26 expression and activity on cultured T cells. Flow cytometry was performed and revealed that, as expected, the majority of cultured cells expressed CD26 in addition to CD3 and CXCR3 (data not shown). Different cell concentrations were incubated with the CD26 substrate Gly-Pro-pNA and the enzymatic activity was determined. As expected, an increased enzymatic activity was found with increasing concentrations of T cells (data not shown). Purified soluble natural seminal fluid-derived CD26 was used as a positive control.
Subsequently, in vitro Boyden chamber assays were conducted to evaluate the effects of GAGs on CXCL10-mediated chemotaxis of T cells. We investigated the migratory response of T cells towards 3 to 300 ng/mL CXCL10 with or without 0.04 or 4 µg/mL soluble heparin or heparan sulfate ( Figure 5). Chemotactic indices were calculated by dividing the number of migrated cells to chemokine by the number of cells that migrated in response to buffer. As expected, a dose-dependent CXCL10-mediated T cell chemotaxis was observed with 10 ng/mL to 300 ng/mL CXCL10. In contrast to our expectations, presence of soluble heparan sulfate did not significantly affect CXCL10-induced T cell chemotaxis in vitro ( Figure 5A). Heparin (4 µg/mL) significantly inhibited T cell chemotaxis to 300 ng/mL CXCL10 and a trend towards reduced chemotaxis was found for other CXCL10 concentrations ( Figure 5B). It remains to be determined whether the lack of inhibition with heparan sulfate was due to in vitro counteraction between the two observed GAG-mediated phenomena, i.e. heparan sulfate-mediated protection of CXCL10 against processing by CD26 on the one hand, and the negative effect of heparan sulfate on the CXCR3-CXCL10 dialog on the other hand.
T cell chemotaxis in vitro ( Figure 5A). Heparin (4 µg/mL) significantly inhibited T cell chemotaxis to 300 ng/mL CXCL10 and a trend towards reduced chemotaxis was found for other CXCL10 concentrations ( Figure 5B). It remains to be determined whether the lack of inhibition with heparan sulfate was due to in vitro counteraction between the two observed GAG-mediated phenomena, i.e. heparan sulfate-mediated protection of CXCL10 against processing by CD26 on the one hand, and the negative effect of heparan sulfate on the CXCR3-CXCL10 dialog on the other hand.

Inhibition of Membrane-Bound CD26 Did Not Affect CXCL10-Mediated T cell Chemotaxis In Vitro
To directly investigate the effects of specific inhibition of CD26-mediated cleavage of CXCL10 on CXCL10-directed T cell migration, chemotaxis assays were performed in the presence of the specific CD26 inhibitor sitagliptin. We first conducted a CD26 activity test with T cells that were treated with 0 to 2 mM sitagliptin and demonstrated that sitagliptin dose-dependently decreased the CD26 activity of cultured T cells ( Figure 6). A concentration of 200 µM was selected for use in in vitro chemotaxis experiments. No significantly increased CXCL10-mediated T cell chemotaxis was observed when cells were treated with sitagliptin, which was in contrast with our expectations and with former in vivo studies [52,63] (Figure 7). However, in vivo, soluble CD26 is present on capillary endothelial cells and in body fluids including plasma, whereas, in our in vitro chemotaxis experiments, CXCL10 was only confronted with membrane-bound CD26 on the T cells. Therefore, it could be speculated that CXCL10 in in vitro assays was already bound to CXCR3 prior to potential truncation by CD26. Additionally, it cannot be excluded that, under the usedin vitro conditions, the amount of membrane-bound CD26 was too low to be able to detect a significant influence of its inhibition by sitagliptin. heparin (B) on the in vitro migratory response of PHA-activated T cells towards10-300 ng/mL CXCL10 was investigated using the 48 well Boyden chamber chemotaxis assay. Values represent the mean chemotactic index (±SEM) (n = 7). Statistical analysis was performed using the Mann-Whitney U test. * = p < 0.05; compared to stimulation with CXCL10 without GAG (black, ).

Inhibition of Membrane-Bound CD26 Did Not Affect CXCL10-Mediated T Cell Chemotaxis In Vitro
To directly investigate the effects of specific inhibition of CD26-mediated cleavage of CXCL10 on CXCL10-directed T cell migration, chemotaxis assays were performed in the presence of the specific CD26 inhibitor sitagliptin. We first conducted a CD26 activity test with T cells that were treated with 0 to 2 mM sitagliptin and demonstrated that sitagliptin dose-dependently decreased the CD26 activity of cultured T cells ( Figure 6). A concentration of 200 µM was selected for use in in vitro chemotaxis experiments. No significantly increased CXCL10-mediated T cell chemotaxis was observed when cells were treated with sitagliptin, which was in contrast with our expectations and with former in vivo studies [52,63] (Figure 7). However, in vivo, soluble CD26 is present on capillary endothelial cells and in body fluids including plasma, whereas, in our in vitro chemotaxis experiments, CXCL10 was only confronted with membrane-bound CD26 on the T cells. Therefore, it could be speculated that CXCL10 in in vitro assays was already bound to CXCR3 prior to potential truncation by CD26. Additionally, it cannot be excluded that, under the usedin vitro conditions, the amount of membrane-bound CD26 was too low to be able to detect a significant influence of its inhibition by sitagliptin.

Inhibition of CD26 Significantly Increased CXCL10-Induced Lymphocyte Influx into the Joint In Vivo
Chemokine injection into the tibiofemoral articulation was used as an experimental model to study the effect of specific CD26 inhibition in vivo, since this model has the advantage of very low basal leukocyte counts upon vehicle injection (Figure 8). Moreover, CXCL10 is typically a chemokine found in high concentrations in synovial fluids from inflammatory joints of patients with septic, rheumatoid or psoriatic arthritis [22,23]. To investigate the in vivo effect of CD26 inhibition on CXCL10-induced lymphocyte recruitment into the joint, CXCL10 was injected into the tibiofemoral articulation of sitagliptin-treated and untreated mice.

Inhibition of CD26 Significantly Increased CXCL10-Induced Lymphocyte Influx into the Joint In Vivo
Chemokine injection into the tibiofemoral articulation was used as an experimental model to study the effect of specific CD26 inhibition in vivo, since this model has the advantage of very low basal leukocyte counts upon vehicle injection ( Figure 8). Moreover, CXCL10 is typically a chemokine found in high concentrations in synovial fluids from inflammatory joints of patients with septic, rheumatoid or psoriatic arthritis [22,23]. To investigate the in vivo effect of CD26 inhibition on CXCL10-induced lymphocyte recruitment into the joint, CXCL10 was injected into the tibiofemoral articulation of sitagliptin-treated and untreated mice.
No lymphocyte infiltration was seen in mice that did receive vehicle injection in the joint. Injection of CXCL10 resulted in a lymphocyte recruitment in both untreated and sitagliptin treated mice ( Figure 8). However, the number of lymphocytes in joints of mice that had received sitagliptin was significantly higher than the lymphocyte counts in mice that were not treated with the CD26 inhibitor. These results provide direct in vivo evidence that CD26 inhibition protects CXCL10 against cleavage, which is reflected into an enhanced lymphocyte extravasation into the joint. Moreover, these findings are in line with previous studies that reported that sitagliptin treatment in mice is translated into increased CXCL10-mediated lymphocyte infiltration into tumor tissue [63] and increased lymphocyte recruitment to intraperitoneally injected CXCL10 [52]. . Sitagliptin-treatment of mice enhances CXCL10-mediated lymphocyte extravasation into the joint. The effect of CD26 inhibition on CXCL10-induced lymphocyte extravasation was determined in vivo. Vehicle or CXCL10 were injected into the tibiofemoral articulation of sitagliptintreated C57BL/6 mice. Total leukocyte numbers migrated into the joint were determined 3 h post injection. Percentages of lymphocytes were differentially counted on May-Grünwald-Giemsa stained cytospins and were used to calculate total lymphocyte numbers. Each symbol represents an individual mouse (n ≥ 8). Horizontal lines indicate the median number of lymphocytes for each treatment group. Statistical analysis was performed using the Mann-Whitney U test. * p < 0.05, *** p < 0.001.

Discussion
Binding of chemokines to GAGs is essential to generate chemotactic gradients in vivo [32][33][34][35][36][37][38][39]. Moreover, it has been evidenced that the impact of the interaction with GAGs extends beyond the mere facilitation of chemokine retention on the endothelium, thereby locally concentrating chemokines on the cell surface [44]. Interaction with GAGs was found to mediate chemokine oligomerization and to provoke protection against proteolysis by specific enzymes. For example, CXCL12 was protected by heparan sulfate and heparin oligosaccharides against cleavage by CD26 [46]. CXCL12 or stromal derived factor 1 (SDF-1) is an ELR negative CXC chemokine that induces chemotaxis of lymphocytes and CD34 positive progenitor cells through activation of CXCR4. Via interaction with the same receptor, CXCL12 holds anti-HIV properties [64][65][66][67]. The chemotactic, CXCR4 dependent signaling and antiviral effects of CXCL12 are lost upon truncation by CD26 [68,69]. Percentages of lymphocytes were differentially counted on May-Grünwald-Giemsa stained cytospins and were used to calculate total lymphocyte numbers. Each symbol represents an individual mouse (n ≥ 8). Horizontal lines indicate the median number of lymphocytes for each treatment group. Statistical analysis was performed using the Mann-Whitney U test. * p < 0.05, *** p < 0.001.

Discussion
Binding of chemokines to GAGs is essential to generate chemotactic gradients in vivo [32][33][34][35][36][37][38][39]. Moreover, it has been evidenced that the impact of the interaction with GAGs extends beyond the mere facilitation of chemokine retention on the endothelium, thereby locally concentrating chemokines on the cell surface [44]. Interaction with GAGs was found to mediate chemokine oligomerization and to provoke protection against proteolysis by specific enzymes. For example, CXCL12 was protected by heparan sulfate and heparin oligosaccharides against cleavage by CD26 [46]. CXCL12 or stromal derived factor 1 (SDF-1) is an ELR negative CXC chemokine that induces chemotaxis of lymphocytes and CD34 positive progenitor cells through activation of CXCR4. Via interaction with the same receptor, CXCL12 holds anti-HIV properties [64][65][66][67]. The chemotactic, CXCR4 dependent signaling and antiviral effects of CXCL12 are lost upon truncation by CD26 [68,69]. In the present study, we demonstrated that GAGs also protect CXCL9, CXCL10 and CXCL11 against proteolytic processing by CD26. Of note, the study by Sadir et al. [46] relied on colon carcinoma cells as source of CD26 activity whereas in our experiments, natural soluble CD26, isolated from human seminal fluid and purified to homogeneity, was used.
Another study reported that murine CCL11, also named eotaxin, is no longer processed by plasmin, elastase and cathepsin G upon interaction with immobilized heparin [47]. Here, the authors found that heparin directly inhibits the enzymatic activity of elastase and cathepsin G but not plasmin. In the present study, no evidence was found for GAGs to directly block the enzymatic activity of CD26. These results are in line with the aforementioned CXCL12 study were the investigators showed that the enzymatic activity of CD26 was not blocked by heparan sulfate [46]. We therefore suppose that the observed role for GAGs in offering the CXCR3 ligands protection against processing by CD26 is rooted at the level of chemokine-GAG interactions, thereby providing steric hindrance at the level of the NH 2 -terminal domain of the chemokine. For CXCL12 it was shown that the protection of the NH 2 -terminus depends on GAG-induced oligomerization of the chemokine rather than straight interaction with the NH 2 -terminal Lys residue [70]. It remains to be elucidated whether the protective effect of GAG interaction on CXCR3 ligands is a result of GAG-mediated chemokine oligomerization or direct steric hindrance at their NH 2 -terminus. Nota bene, GAG-mediated chemokine oligomerization seems an intriguing mechanism on its own, but the molecular details remain largely unknown. Indeed, although chemokine monomers seem responsible for GPCR activation, GAG-mediated induction or stabilization of chemokine oligomers is a prerequisite for the in vivo activity of certain chemokines including CXCL10 [33,36,71]. In return, oligomerization may enhance chemokine affinity for GAGs, which is also dependent on specific GAG density [72]. Oligomeric forms were described for CXCL9 and CXCL10 at physiological concentrations, whereas the oligomeric state of CXCL11 is less understood [36,[73][74][75]. However, a key role for oligomerization of several chemokines including CXCL11 was recently demonstrated in facilitating reorganization and bridging of GAG chains, thereby conferring another level of complexity to the chemokine-GAG dialog [72].
The CXCR3 agonists, specifically CXCL9, CXCL10 and CXCL11, all contain a proline residue in the penultimate NH 2 -terminal sequence and are biologically relevant CD26 substrates. Upon incubation of 5 µM chemokine with a normal plasma concentration of 25 U/L CD26, the half-lives of CXCL10 and CXCL11 were previously demonstrated to be, respectively, not more than 4 and 2 min, whereas CD26-mediated cleavage of CXCL9 was found to occur less efficiently (half-life of 24 min under the same experimental conditions) [62]. Following cleavage by CD26, the three CXCR3 agonists are biologically inactive in chemotaxis and CXCR3-dependent signaling assays and show a decreased effect on T cells. Moreover, CXCL10(3-77) and CXCL11  act as CXCR3 antagonists [53]. The CD26-truncated isoform CXCL10(3-77) was originally isolated from natural sources including conditioned medium from MG-63 osteosarcoma cells and fibroblasts [22,58]. Studies with CXCL10 −/− mice established a crucial role for CXCL10 in the generation of effector T cells and in T cell trafficking in general [76]. This important physiological role of CXCL10, but also CXCL9 and CXCL11, combined with the fact that CXCL9(3-103), CXCL10(3-77) and CXCL11  are inactive, supports the idea that CD26-mediated proteolysis of the CXCR3 agonists may have major biological consequences. Results of multiple studies provided evidence in favor of this hypothesis. In mice, truncation of CXCL10 by CD26 reduces the infiltration of T cells into tumor tissue [63] and towards intraperitoneally injected CXCL10 [52] and consequently impairs the natural antitumor immunity Administration of the CD26 inhibitor sitagliptin significantly improved the natural antitumor immunity of these mice and their response to existing immunotherapies. In human, the biologically inactive isoform CXCL10(3-77) was found in plasma from patients that suffer from chronical hepatitis C viral infections [77]. Moreover, two prospective studies recently confirmed the CD26-mediated processing of human CXCL10 in vivo, and provided direct evidence in favor of CD26 inhibition to preserve intact CXCL10 [78]. Although human CXCL11 activates CXCR3 with higher potency and is even more efficiently processed by CD26 than CXCL10, the CD26-CXCL11 axis has been studied to a lesser extend in an in vivo context. It is likely to assume that this can be at least partially explained by the fact that murine CXCL11, in contrast to its human counterpart, contains a methionine residue in its penultimate position and is therefore no substrate for CD26. Interestingly, natural human CXCL11(3-73) was previously isolated from IFN-γ stimulated keratinocytes [59][60][61]. In addition, murine CXCL9 is no CD26 substrate due to the leucine residue that occupies the penultimate position in its NH 2 -terminal amino acid sequence. However, in view of the efficient cleavage of all three human CXCR3 agonists, our observed GAG-mediated protection may be highly significant depending on the local conditions in specific human tissues. Indeed injection of CXCL10 in joints still induces limited lymphocyte migration (Figure 8), whereas injection of CXCL10 in the peritoneum did not attract lymphocytes [52].
In the present study, we found that soluble GAGs significantly reduce the calcium mobilizing capacities of CXCL9, CXCL10 and CXCL11 through CXCR3. Given the general idea that the NH 2 -terminal chemokine domain facilitates receptor interaction whereas the COOH-region is considered the major GAG-binding domain, this observation may seem somewhat unexpected. However, it was previously suggested that both interaction domains are not necessarily restricted to, respectively, the NH 2 -and COOH-terminus. The relevance of this hypothesis for CXCL10 for example, is supported by the fact that mutation of amino acids 20 to 24, 46 and 47 impairs both the heparin affinity of the chemokine and its binding and signalization through CXCR3 [79]. Furthermore, citrullination of the most NH 2 -terminal arginine in CXCL10 or CXCL11 reduced their interaction with heparin [80]. At this point it remains to be elucidated whether the negative effect of GAGs on the potency of CXCR3 agonists to induce intracellular calcium release results from competition between chemokines and GAGs for CXCR3 binding or is rooted at the level of signal transduction, where GAGs potentially exert an inhibitory effect either directly or indirectly. Noteworthy, former studies reported that soluble GAGs also significantly reduce the CXCL8, CCL2, CCL3 and CCL5 mediated intracellular calcium mobilization [81]. Specifically, the authors showed that GAGs, in a competitive fashion, bind to chemokine receptors CXCR1, CXCR2 and CCR1. To our notice, no calcium mobilization studies with GAGs, CXCR3 and its chemokine ligands have been conducted before, but we speculate that GAGs may also interact with CXCR3.
Membrane-bound CD26 is a lymphocyte surface marker that is expressed by activated T cells [82]. In our study, we confirmed that cultured T cells were indeed characterized by CD26 expression and we demonstrated that the expressed enzyme was enzymatically active. Combined with the fact that CXCL10(3-77) is biologically inactive and our observation that GAGs protect CXCL10 against cleavage by CD26, we reasoned that in the presence of GAGs, intact CXCL10 would be preserved, thus resulting in enhanced CXCL10-mediated T cell chemotaxis compared to a condition without GAG. However, we found no significant differences in CXCL10-mediated T cell chemotaxis in vitro in the presence of heparan sulfate and a moderately reduced migration in the presence of heparin. Possibly, this could be explained by the negative effect of GAGs on the CXCL10-CXCR3 dialog. Indeed, it appears that the reduced CXCL10(3-77)-CXCR3 interaction overwrites the inhibitory effect of GAGs on CD26-mediated truncation of intact CXCL10 towards inactive CXCL10 . To investigate the effect of specific inhibition of CD26-mediated cleavage on CXCL10-induced T cell chemotaxis in vitro, we evaluated the effect of administration of the CD26 inhibitor sitagliptin. In contrast to our expectations, no significant increase in CXCL10-directed T cell migration was found in vitro when cells were treated with sitagliptin. However, it was previously demonstrated that sitagliptin treatment in mice resulted in enhanced infiltration of T cells into the peritoneum or in tumor tissue in vivo [52,63]. Moreover, in the present study, we showed that intra articular injection of CXCL10 in sitagliptin-treated mice significantly enhances lymphocyte recruitment to the joint compared to mice that did not receive the CD26 inhibitor. These observations are in line with the idea that specific CD26 inhibition protects CXCL10 from inactivation in vivo. Indeed, in contrast to what is the case in our in vitro experiments, soluble CD26 and membrane-bound CD26 on non-lymphoid cells such as certain endothelial cells, fibroblasts and epithelial cells is present in vivo in addition to T cell-associated membrane-bound CD26 [83]. Therefore, in in vitro chemotaxis experiments, CXCL10 was not hindered by soluble CD26, but could only be inactivated by CD26 when it directly contacted the membrane-bound enzyme. A possible explanation for the lack of a significant effect of sitagliptin-mediated CD26 inhibition on CXCL10-induced chemotaxis in vitro, consequently, could be that CXCL10 was already bound to its receptor prior to interaction with membrane-bound CD26 on T cells. Furthermore, the amount of T cells in our experiments was fixed whereas during inflammation in vivo, for example, the number of activated T cells and consequently the availability of membrane-bound CD26, may drastically increase. Additionally, although we confirmed the CD26 expression and activity on cultured T cells, the CD26 activity on these cells was rather low. Thus, one could speculate that the amount of cell-bound CD26 activity in our in vitro tests was too low.

Chemokines and CD26
Full length recombinant human chemokines CXCL9, CXCL10 and CXCL11 were purchased from PeproTech (Rocky Hill, NJ, USA). An Activo-P11 automated solid phase peptide synthesizer (Activotec, Cambridge, UK) was used to chemically synthesize CXCL10, based on N-(9-fluorenyl) methoxycarbonyl (Fmoc) chemistry as described previously [84]. To avoid synthesis problems due to its COOH-terminal proline, a H-Pro-2Cl-Trityl resin (Activotec) was used for synthesis of CXCL10. The synthesized chemokine was purified to homogeneity with reverse phase-high performance liquid chromatography using a Source 5-RPC column (4.6 × 150 mm; GE Healthcare, Uppsala, Sweden). An acetonitrile gradient in 0.1% (v/v) TFA was used for elution of the synthesized protein and 0.7% of the effluent was directly injected into an electrospray-ion trap mass spectrometer (Bruker AmaZon SL mass spectrometer; Bruker Daltonics, Bremen, Germany). Fractions containing homogenous CXCL10 were selected, pooled, evaporated and dissolved in ultrapure water. Following folding of purified CXCL10 according to the protocol described by Loos et al. [84], the identity of the chemokine was confirmed by ion trap mass spectrometry and automated NH 2 -terminal sequencing based on the principle of Edman degradation (Procise 491 cLC sequencer, Applied Biosystems, Foster City, CA, USA). In addition, SDS-PAGE and bicinchoninic acid (BCA) protein assays (Pierce, Woodland Hills, CA, USA) were used to determine protein concentrations and purity.
Natural human soluble CD26 was isolated from human seminal fluid and purified to homogeneity by anion exchange and affinity chromatography as described [85].

CD26 Activity Assays
In a flat bottom 96 well plate, (Greiner Bio-One, Kremsmünster, Austria), 5 U/L purified natural human CD26 was incubated with 500 µM of the substrate Gly-Pro-p-nitroanilide (Gly-Pro-pNA; Sigma-Aldrich), with or without GAG, in 0.22 µm-filtered 75 mM Tris-HCl buffer (pH 8.3). Evolution of the optic density (OD) at 405 nm was followed using a spectrophotometer (BioTek PowerWave XS, Winooski, VT, USA) and represented the kinetics of CD26-mediated enzymatic conversion of colorless Gly-Pro-pNA into yellowish pNA. OD measurements were performed at 37 • C every 5 min for 3 h. CD26 activity curves were constructed by plotting OD values against time. Slopes of activity curves reflected the number of converted substrate molecules per min and were used to calculate enzymatic activities in U/L using the Lambert-Beer law. CD26 activity assays were also used to investigate the CD26 activity of PHA-activated T cells. Briefly, 10 4 to 3 × 10 6 T cells per mL in 75 mM Tris-HCl buffer (pH 8.3), with or without 20 µM to 2 mM sitagliptin (Merck Sharpe & Dohme (MSD) Whitehouse Station, NJ, USA), were incubated with 500 µM Gly-Pro pNA and the same protocol was followed.

In Vitro Chemotaxis Assays
In vitro, 48-well Boyden chamber cell migration assays were used to determine chemotaxis of PHA-activated T cells in response to CXCL10 with or without GAG [87]. Briefly, CXCL10 and T cells were diluted in HBSS buffer enriched with 0.5% (v/v) human serum albumin (HSA; Red Cross Blood transfusion center, Leuven, Belgium). Wells in the lower part of the chamber were filled with 10-300 ng/mL CXCL10 with or without 0.04 or 4 µg/mL GAG in a total volume of 30 µL per well. Buffer without chemokine or GAG was used as a negative control. The lower chamber was covered with a 5 µm polycarbonate membrane (Nuclepore Track-Etch Membrane, Whatman, Little Chalfont, UK) that was treated with 20 µg/mL fibronectin (Gibco) in PBS overnight. The upper part of the chamber was filled with 2 × 10 6 T cells per mL in buffer (50 µL per well), with or without 200 µM sitagliptin. Following 3 h of incubation at 37 • C, membranes were fixed and stained (Hemacolor Solution I-III, Merck). For each individual test condition, numbers of migrated T cells were counted microscopically in 10 separate fields. Chemotactic indices were determined by dividing total numbers of T cells that migrated to the sample through the total number of cells that migrated in response to buffer alone.

In Vivo Cell Migration Assay
The effect of CD26 inhibition on lymphocyte attraction was determined after intra-articular (i.a.) injection of human CXCL10 as described by Janssens et al. [88]. Briefly, the drinking water of 8-week-old C57BL/6 mice was complemented with 1.7 mg/mL sitagliptin from 3 days prior to i.a. injection of 1 µg synthetic full length CXCL10 diluted in 0.9% (w/v) NaCl. A Limulus amoebocyte lysate assay (Cambrex) demonstrated that endotoxin levels in injected samples were lower than 0.125 pg LPS per µg chemokine. During injection of 10 µL of the CXCL10 dilution, mice were anaesthetized using 3.75% (w/v) ketamine plus 0.25% (w/v) xylazine in PBS. Mice were sacrificed after 3 h of incubation, and articular cavities were washed with 3% (w/v) BSA in PBS. Total leukocyte numbers were calculated after staining with Turk's solution in a Neubauer chamber, followed by differential counting of samples on cytospins that were stained with May-Grünwald-Giemsa. All in vivo experiments were conducted in the animal research facility of the University of Minas Gerais after approval by the Animal Ethical Committee.

Flow Cytometry
Flow cytometry was used to confirm the CXCR3 and CD26 expression of cultivated T cells. Tubes were filled with 3 × 10 5 cells and centrifuged (5 min, 4 • C, 315 g). Supernatant was removed and cells were suspended in PBS containing Fc block (MACS Miltenyi Biotec, Bergisch Gladbach, Germany) and Aqua Zombie-BV510 (Biolegend, San Diego, CA, USA), according to the recommendations of the companies. Following 15 min of incubation at 20 • C, cells were centrifuged (5 min, 4 • C,