C-terminal truncation of CXCL10 attenuates inflammatory activity but retains angiostatic properties

Interferon-γ-inducible protein of 10 kDa (IP-10/CXCL10) is a dual-function CXC chemokine that coordinates chemotaxis of activated T cells and natural killer (NK) cells via interaction with its G protein-coupled receptor (GPCR), CXC chemokine receptor 3 (CXCR3). As a consequence of natural posttranslational modifications, human CXCL10 exhibits a high degree of structural and functional heterogeneity. However, the biological effect of natural posttranslational processing of CXCL10 at the carboxy (C)-terminus has remained partially elusive. The truncated CXCL10 proteoform CXCL10(1-73), lacking the four endmost C-terminal amino acids, was previously identified in human cell culture supernatant. To further explore the functioning of CXCL10(1-73), we optimized its production in this study through Fmoc-based solid phase peptide synthesis (SPPS) and propose an SPPS strategy to efficiently generate human CXCL10 proteoforms. Compared to intact CXCL10(1-77), CXCL10(1-73) had diminished affinity for glycosaminoglycans including heparin, heparan sulfate and chondroitin sulfate A. Moreover, CXCL10(1-73) exhibited an attenuated capacity to induce CXCR3A-mediated signaling, as evidenced in calcium mobilization assays and through quantification of phosphorylated extracellular signal-regulated kinase-1/2 (ERK1/2) and protein kinase B/Akt. Furthermore, CXCL10(1-73) incited reduced primary human T lymphocyte chemotaxis in vitro and evoked less peritoneal ingress of CXCR3+ T lymphocytes in mice receiving intraperitoneal chemokine injections. In contrast, loss of the four endmost C-terminal residues did not affect the inhibitory properties of CXCL10 on spontaneous and/or FGF-2-induced migration, proliferation, wound healing, phosphorylation of ERK1/2, and sprouting of human microvascular endothelial cells. Thus, C-terminally truncated CXCL10 has attenuated inflammatory properties, but preserved anti-angiogenic capacity. GRAPHICAL ABSTRACT

successfully modeled via NMR spectroscopy and crystallography [45,46]. Therefore, prediction of the precise 48 location and proximity of the C-terminal amino acids relative to other residues in the peptide backbone of CXCL10 49 is still speculative, making structure-activity predictions challenging. C-terminal residues of mCXCL10 have been 50 investigated to a more elaborate extent and were implicated in GAG and receptor binding [47][48][49][50].

52
The aim of the present study was to evaluate the effects of the naturally occurring C-terminal truncation of CXCL10 53 on the functional properties of this human chemokine. We introduced a strategy for Fmoc-based solid phase 54 peptide synthesis (SPPS) of posttranslationally modified CXCL10 proteoforms with the example of CXCL10(1-73). We 55 utilized synthetic CXCL10(1-73) to examine the effects of the C-terminal truncation. We discovered that the C-56 terminal truncation of CXCL10 attenuated the interaction with GAGs, the signaling properties through CXCR3A, 57 and the ability to attract T lymphocytes in vitro and in vivo. However, the angiostatic properties of CXCL10, 58 including the inhibition of migration, proliferation, wound healing, phosphorylation of extracellular signal-59 regulated kinase-1/2 (ERK1/2), and sprouting of endothelial cells, were not affected by the C-terminal processing. 60

75
in the peptide backbone of CXCL10. In addition, 1,1,3,3-tetramethyluronium hexafluorophosphate (HCTU) and 4-76 methylmorpholine (NMM) were used as alternative high quality coupling reagents [55][56][57]. Despite these 77 modifications in the SPPS, a very low amount of successfully generated CXCL10(1-73) was present upon analysis of 78 the synthetic material. One major and highly abundant contaminant was detected, i.e. N-terminally shortened 79 acetylated CXCL10(31-73) (theoretical Mr 4870.78, experimental Mr 4868.86) (Fig. S1B). The acetylation clearly 80 points towards a synthesis artefact, resulting from impaired peptide amide bond formation between Ile 30 and 81 Pro 31 . Coupling of amino acids to C-terminally resin-attached Pro residues is often more challenging given the 82 reduced reactivity of the secondary amine located in the proline oxazolidine ring structure. The formation of this 83 shortened peptide was circumvented via the selective incorporation of a specific dipeptide building block at Ile 30 -

114
In calcium assays, CXCR3A-transfected CHO cells were used to determine whether CXCL10(1-73) had similar potency 115 as CXCL10(1-77) to induce mobilization of intracellular calcium. Only highly elevated concentrations of CXCL10(1- 73) 116 (i.e., 135 nM and 270 nM) were able to induce a moderate increase in intracellular calcium concentrations, thereby 117 reaching comparable calcium levels as upon stimulation with 1 nM CXCL10(1-77) (Fig. 4A). At 3 nM of CXCL10(1-77), 118 calcium mobilization was even significantly higher compared to 270 nM of CXCL10(1-73). We also observed that the 119 time between administration of the stimulus and the initiation of the calcium increase was prolonged for CXCL10(1-120 73) independent of the administrated dose (Fig. 4B). The markedly limited capacity of CXCL10(1-73) compared to 121 CXCL10(1-77) to mobilize intracellular calcium sparks the notion of potential CXCR3 desensitization by this C-122 terminally shortened CXCL10 proteoform at inactive concentrations. Indeed, when 100 seconds prior to a stimulus

146
Given that the C-terminal truncation of CXCL10 significantly attenuated CXCR3 signaling, we further investigated 147 whether CXCL10(1-73) also exhibited reduced T lymphocyte chemotactic capacities. To ascertain adequate 148 responsiveness of T lymphocytes, CXCL12α was included as positive control. PHA-and IL-2-stimulated T 149 lymphocytes are known to express CXCR4 and pronouncedly migrate after exposure to CXCL12α [62]. Most T 8 lymphocytes (median expression 81.6%, median MFI 2303) were positive for CXCR3 ( Fig. 5A-B). Given that in vitro 151 T cell chemotaxis induced by CXCL10 has been shown to occur in the absence of coating [47] and to exclude that 152 distinct binding affinities of the CXCL10 proteoforms to extracellular matrix proteins underly the difference in T 153 lymphocyte migration, we evaluated migration through uncoated membranes. We confirmed that chemotaxis was 154 significantly and dose-dependently increased upon stimulation with CXCL12α relative to spontaneous migration 155 of medium-treated cells (Fig. 5C). Starting from 1 nM, CXCL10(1-77) induced a significant and dose-dependent 156 migration of CXCR3 + T lymphocytes compared to cells exposed to buffer. The migratory response of T lymphocytes 157 towards CXCL10(1-73) was significantly increased compared to buffer only from 10 nM CXCL10(1-73) onwards.

204
Fourth, we examined the ability of CXCL10 proteoforms to blunt the FGF-2-induced ERK signal transduction 205 pathway. CXCL10(1-73) and CXCL10(1-77) significantly diminished FGF-2-induced ERK phosphorylation at 120 nM with 206 no significant differences between the two proteoforms ( Fig. 8A). In addition, we evaluated the effects of both 207 CXCL10 proteoforms in the in vitro spheroid sprouting assay, which is a solid in vitro angiogenesis model that is 208 proximal to the in vivo situation and enables to assess angiogenesis in a 3-dimensional (3D) environment [64,65].

209
Pronounced sprouting of collagen-embedded HMVEC spheroids was observed after treatment with 10 ng/ml FGF-

218
Chemokine injection into the peritoneal cavity of NMRI mice followed by peritoneal lavage was used as an

248
In the present study, we characterized the effects of a synthetic CXCL10 proteoform corresponding to natural C-249 terminally truncated CXCL10(1-73) that was previously identified in human cell culture supernatant [27,29,30]. We terms of their synthesis and purification [70,71]. These proteins tend to form inter-and intra-molecular β-sheet 261 interactions, resulting in on-resin aggregation during peptide synthesis and consequently synthesis failure. To 262 11 overcome this major obstacle, the concomitant use of pseudoproline dipeptides and a hydrophilic polyethylene 263 glycol (PEG) resin was explored. This strategy was previously shown to substantially increase the synthesis yield of 264 the human chemokine RANTES/CCL5(1-68) [72]. In addition, we used a high quality coupling system, i.e., HCTU and 265 NMM [55][56][57], as described in a recently established methodology for Fmoc-based SPPS of mCXCL10(1-77) [57]. The 266 latter authors also utilized a pseudoproline dipeptide at Ala 43

276
A former research effort aiming to study native C-terminally truncated CXCL10(1-73) was made by Hensbergen et 277 al. [29]. They opted for the use of recombinant CXCL10(1-73) with an additional N-terminal methionine (Met-

327
Thus, the C-terminal residues of human CXCL10 may act analogous to mCXCL10 where Lys 71 -Arg 75 are involved in 328 CXCR3A activation resulting in downstream calcium signaling and chemotaxis (vide supra) [47]. Also, similar to 329 mCXCL10 is the emerging notion that binding sites for GAG and CXCR3 are partially overlapping in human CXCL10.

331
Analogous to CD26-mediated N-terminally truncated CXCL10(3-77) [34], we observed that the C-terminal truncation  73) is mediated through the proposed angiostasis-mediating CXCR3B [78], through the unidentified receptor for 336 CXCL10 on endothelial cells [79], through differential PTMs of CXCR3 on endothelial cells compared to leukocytes, 337 or through a biased cell type-dependent CXCR3A or GAG signaling pathway [25,48] (that is not affected by these

339
Tyr residues [76], whereas others have reported that the extended N-terminus of CXCR3B has two additional sites 340 (Tyr 6 and Tyr 40 ) that may be prone to sulfation [80,81]. Moreover, lymphocytes exhibit less heparan sulfate 341 proteoglycans on their surface compared to other leukocytes [82], which may emphasize the importance of 342 sulfation sites on CXCR3A to sequester chemokines [75]. In contrast, endothelial cells have relatively more GAGs 343 expressed on their surface, again pointing towards potentially differential signaling of CXCL10 in HMVECs 344 compared to T cells. Alternatively, CXCL10 may also interfere with binding of growth factors to their receptors by 345 directly binding to the growth factor or impairing the growth factor's ability to dimerize, as shown for CXCL4 [83].

363
In conclusion, our study reveals that the C-terminal residues Lys 74 -Pro 77 of CXCL10 are important for GAG binding,

644
Mann-Whitney U test was performed (* p < 0.05, ** p < 0.01 for comparison to control).      Values represent the mean ± SEM of 3 to 4 independent experiments. Kinetic parameters were determined from the association phase (1 to 120 seconds) and dissociation phase (120 to 300 seconds) of the SPR sensorgrams. The KD was calculated from the ratio of koff over kon (nM) determined by the 1:1 binding model with mass transfer correction. kon association rate constant (M -1 s -1 ); koff dissociation rate constant (s -1 ); KD dissociation equilibrium (affinity) constant. N.D., not determined.