Effects of cognate, non-cognate and synthetic CXCR4 and ACKR3 ligands on human lung endothelial cell barrier function

Recent evidence suggests that chemokine CXCL12, the cognate agonist of chemokine receptors CXCR4 and ACKR3, reduces thrombin-mediated impairment of endothelial barrier function. A detailed characterization of the effects of CXCL12 on thrombin-mediated human lung endothelial hyperpermeability is lacking and structure-function correlations are not available. Furthermore, effects of other CXCR4/ACKR3 ligands on lung endothelial barrier function are unknown. Thus, we tested the effects of a panel of CXCR4/ACKR3 ligands (CXCL12, CXCL11, ubiquitin, AMD3100, TC14012) and compared the CXCR4/ACKR3 activities of CXCL12 variants (CXCL12α/β, CXCL12(3–68), CXCL121, CXCL122, CXCL12-S-S4V, CXCL12-R47E, CXCL12-K27A/R41A/R47A) with their effects on human lung endothelial barrier function in permeability assays. CXCL12α enhanced human primary pulmonary artery endothelial cell (hPPAEC) barrier function, whereas CXCL11, ubiquitin, AMD3100 and TC14012 were ineffective. Pre-treatment of hPPAEC with CXCL12α and ubiquitin reduced thrombin-mediated hyperpermeability. CXCL12α-treatment of hPPAEC after thrombin exposure reduced barrier function impairment by 70% (EC50 0.05–0.5nM), which could be antagonized with AMD3100; ubiquitin (0.03–3μM) was ineffective. In a human lung microvascular endothelial cell line (HULEC5a), CXCL12α and ubiquitin post-treatment attenuated thrombin-induced hyperpermeability to a similar degree. CXCL12(3–68) was inefficient to activate CXCR4 in Presto-Tango β-arrestin2 recruitment assays; CXCL12-S-S4V, CXCL12-R47E and CXCL12-K27A/R41A/R47A showed significantly reduced potencies to activate CXCR4. While the potencies of all proteins in ACKR3 Presto-Tango assays were comparable, the efficacy of CXCL12(3–68) to activate ACKR3 was significantly reduced. The potencies to attenuate thrombin-mediated hPPAEC barrier function impairment were: CXCL12α/β, CXCL121, CXCL12-K27A/R41A/R47A > CXCL12-S-S4V, CXCL12-R47E > CXCL122 > CXCL12(3–68). Our findings indicate that CXCR4 activation attenuates thrombin-induced lung endothelial barrier function impairment and suggest that protective effects of CXCL12 are dictated by its CXCR4 agonist activity and interactions of distinct protein moieties with heparan sulfate on the endothelial surface. These data may facilitate development of compounds with improved pharmacological properties to attenuate thrombin-induced vascular leakage in the pulmonary circulation.

Introduction Presto-Tango β-arrestin 2 recruitment assay The PRESTO-Tango (parallel receptorome expression and screening via transcriptional output, with transcriptional activation following arrestin translocation) assay was performed as recently described [33]. The Tango plasmids were a gift from Dr. Bryan Roth (all from Addgene). HTLA cells (2.5x10 5 /well) were seeded in a 6-well plate and transfected with 1.5 μg of the Tango plasmids using Lipofectamine 3000 (ThermoScientific). The following day, transfected HTLA cells (1x10 5 cells/well) were plated onto Poly-L-Lysine pre-coated 96-well microplates and allowed to attach to the plate surface for at least 4 hours prior to treatment. Proteins used for treatment were prepared in twice the final concentration in culture media, added at a 1:1 vol/vol ratio and incubated overnight at 37˚C, 5% CO 2 in a humidified environment. The following morning, media was removed from cell culture plates and replaced with a 100 μL 1:5 mixture of Bright-Glo (Promega) and 1x HBSS, 20 mM HEPES solution. Plates were then incubated at room temperature before measuring luminescence on a Biotek Synergy II plate reader.

Data analyses
Data are expressed as mean ± SEM from n independent experiments that were performed on different days. Data were analyzed with unpaired Student's t test, one-or two-way analyses of variance with Bonferroni's multiple comparison post hoc test, as appropriate. Dose-response curves were generated using nonlinear regression analyses. All analyses were performed with the GraphPad-Prism 7 software. A two-tailed P < 0.05 was considered significant.

Results
To confirm that CXCR4 and ACKR3 are expressed in hPPAEC and to assess whether both receptors form heteromeric complexes, we performed PLA to detect individual receptors and receptor-receptor interactions at single molecule resolution. As shown in Fig 1A, we observed positive PLA signals for both receptors individually and for CXCR4:ACKR3 heteromeric complexes. We then tested the effects of a panel of CXCR4 and ACKR3 ligands on hPPAEC monolayer permeability in transwell-permeability assays with FITC-dextran ( Fig 1B). Ubiquitin, a non-cognate CXCR4 agonist that does not bind to ACKR3, CXCL11, an ACKR3 and CXCR3 agonist, TC14012, a synthetic CXCR4 antagonist and ACKR3 agonist, and AMD3100, a CXCR4 antagonist and allosteric ACKR3 agonist, did not affect hPPAEC permeability [26, [38][39][40][41]. In contrast, CXCL12 enhanced hPPAEC barrier function.
To be able to assess the effects of CXCL12 on thrombin-induced impairment of hPPAEC barrier function under optimized conditions, we determined the dose-response characteristics for thrombin in the permeability assay. The effects of thrombin (10-100 nM) on hPPAEC monolayer permeability are shown in Fig 2A. Thrombin dose-and time-dependently induced permeability of the hPPAEC monolayer. The time to reach plateau for the permeability- Cells were then exposed to vehicle or 50 nM of CXCR4/ACKR3 ligands for 10 minutes, as indicated, followed by the addition of FITC-dextran.  (Fig 2B). The thrombin-mediated effects showed a sigmoidal dose-effect relationship at all time points. The EC 50 for thrombin-induced impairment of hPPAEC barrier function was 30±2 nM after 55 min, 33±2 nM after 135 min and 36±2 nM after 255 min. The maximal impairment of endothelial barrier function (100% permeability = measured permeability in the absence of hPPAEC) reached 49±3% at 55 min and 59±4% and 72±4% at 135 min and 255 min, respectively.
We then tested whether pre-treatment with CXCL12 and ubiquitin influences hyper-permeability of hPPAEC induced by a sub-maximal dose of thrombin. hPPAEC were treated for 10 min with either 100 nM of CXCL12 or ubiquitin, followed by the addition of of thrombin ( Fig 3A). Pre-treatment with both CXCR4 agonists significantly attenuated thrombin-induced hyper-permeability of hPPAEC. CXCL12 was more efficacious than ubiquitin in this assay. The effects of CXCL12 and ubiquitin when added after thrombin treatment of hPPAEC are shown in Fig 3B and 3C. CXCL12 significantly attenuated thrombin-induced permeability of hPPAEC and this effect could be antagonized with the CXCR4 antagonist AMD3100. AMD3100 treatment alone did not affect thrombin-mediated hyper-permeability of hPPAEC ( Fig 3B). In contrast to CXCL12, ubiquitin-treatment and ubiquitin plus AMD3100-treatment did not modulate thrombin-induced hyper-permeability when tested in parallel experiments (Fig 3C). To exclude that the dose-effect relationship for ubiquitin is different from the dose-effect relationship for CXCL12, we tested ubiquitin in various concentrations (30 nM-3 μM), including concentrations above the K D -value of ubiquitin for CXCR4 binding [39]. Ubiquitin treatment, however, did not attenuate thrombin-induced hyper-permeability of the hPPAEC monolayer at any tested concentration under these experimental conditions ( Fig 3D).
As observed in hPPAEC, thrombin also dose-and time-dependently induced permeability in the human lung microvascular endothelial cell line HULEC-5a (Fig 4A and 4B). When compared with hPPAEC, potency and efficacy of thrombin to induce permeability were reduced in HULEC-5a. The EC 50 for thrombin-induced impairment of HULEC-5a barrier function was 64±7 nM after 55 min, 64±6 nM after 135 min and 57±5 nM after 255 min. Addition of CXCL12 and ubiquitin after treatment of HULEC-5a cells with a sub-maximal dose of HULEC-5a were grown to a confluent monolayer on collagen-coated permeable membranes and then exposed to different concentration of thrombin for 10 min, followed by the addition of FITC-dextran. Endothelial permeability was assessed by measuring the amount of FITC-dextran that permeated through the cell monolayer. No cells: 100% permeability. RFU: Relative fluorescence units. N = 3 in quadruplicate. (B) Dose-response curves for thrombin-induced permeability, data from A. 100% permeability = permeability in the absence of HULEC-5a. Open squares: Permeability at t = 55 min. Light grey squares: Permeability at t = 135 min. Dark grey squares: Permeability at t = 255 min. (C/D) HULEC-5a were grown to a confluent monolayer on collagen-coated permeable membranes and then exposed to 50 nM of thrombin or vehicle. After 10 min, thrombin-exposed cells were treated with vehicle, CXCL12 (50 nM) (C) or ubiquitin (50 nM) (D), followed by the addition of FITC-dextran. The experimental conditions are indicated. Endothelial permeability was assessed by measuring the amount of FITC-dextran that permeated through the cell monolayer. RFU: Relative fluorescence units. N = 3 in quadruplicate. *: p<0.05 vs. thrombin/vehicle (2-way ANOVA/Bonferroni's multiple comparison post hoc test).
Next, we utilized the Presto-Tango β-arrestin 2 recruitment assay to assess CXCR4 and ACKR3 agonist activities of the natural splice variants CXCL12α and CXCL12β, of truncated CXCL12 (3-68) and the engineered constitutively monomeric (CXCL12 1 ) and dimeric (CXCL12 2 ) CXCL12 variants. To confirm the dimeric and monomeric structure of the CXCL12 1 and CXCL12 2 variants, we performed polyacrylamide gel electrophoresis (PAGE) under non-reducing and reducing conditions (Fig 5). Consistent with the mono-and dimeric nature of the CXCL12 variants [27,29], the migration position of CXCL12 2 was close to 20 kDa under non-reducing conditions, whereas CXCL12 and CXCL12 1 migrated to a position corresponding to a lower molecular mass. It should be noted that a faint band migrating at the position of CXCL12 2 was visible in non-reducing SDS-PAGE with CXCL12 (loaded at 4 μM), which is consistent with its dimerization K d of 140 μM [42]. Under reducing conditions all three proteins showed an identical migration position, corresponding to the monomeric molecular mass of approximately 8 kDa.
In addition, we tested CXCL12 S-S4V, a protease resistant mutant, CXCL12 K27A/R41A/ R47A, which shows significantly reduced heparan sulfate proteoglycan binding properties, and CXCL12 R47E, which activates CXCR4 with reduced potency, as compared with CXCL12α. The dose-response curves are shown in Fig 6 and Table 1 summarizes the corresponding EC 50 concentrations and top plateau values (efficacy) for each protein.
In contrast, all proteins induced β-arrestin 2 recruitment to ACKR3 with an EC 50 in the low nM range (p>0.05 for all vs. CXCL12α). While the efficacy for β-arrestin 2 recruitment to ACKR3 was significantly reduced for CXCL12 , the efficacies of all other proteins for βarrestin 2 recruitment to ACKR3 were comparable.

Discussion
In the present study, we evaluated the effects of CXCR4 and ACKR3 ligands on the barrier function of human lung endothelial cells. CXCL12 has previously been described to enhance transendothelial electrical resistance, a surrogate marker of endothelial barrier function, of bovine aortic, human pulmonary artery and umbilical vein endothelial cells [22]. Furthermore, pre-treatment of bovine aortic endothelial cells with CXCL12 has been reported to attenuate thrombin-induced FITC-dextran transfer in transwell permeability assays. Likewise, co-treatment of human microvascular endothelial cells with CXCL12 or CTCE-0214, a synthetic CXCL12 analogue, plus thrombin attenuated the reduction of transendothelial resistance that was detectable with thrombin alone [15,22]. Our observations from the present study are in CXCR4/ACKR3 activation and lung endothelial barrier function agreement with previous reports and now provide direct evidence that CXCL12 enhances barrier function of hPPAEC in the absence of permeability-inducing agents. In addition, we demonstrate that pre-treatment of hPPAEC with CXCL12 and with the non-cognate CXCR4 agonist ubiquitin, which does not bind to ACKR3 [38], attenuates thrombin-mediated hPPAEC barrier function impairment. These findings provide a possible mechanism underlying lung protective effects of intravenous CXCL12 pre-treatment in an oleate-induced lung injury model in rabbits and of ubiquitin pre-treatment in an endotoxic shock model in pigs [17,20].
Although pre-and co-treatment experiments provide information on possible preventive properties of CXCR4 agonists, such experiments are unable to address therapeutic potential. Thus, we performed post-treatment experiments and detected that activation of CXCR4 after thrombin-exposure of hPPAEC and HULEC5a cells attenuates thrombin-mediated impairment of lung endothelial barrier function. These findings support the concept that CXCR4 agonists have therapeutic potential to limit thrombin-mediated pulmonary vascular leakage, which likely contributed to lung protective effects of CXCR4 agonists that have been observed in various models when administered after the insult [15,17,18,21,43].
In contrast to CXCL12, the non-cognate CXCR4 agonist ubiquitin did not enhance hPPAEC barrier function in the absence of thrombin. As compared with CXCL12, ubiquitin was less efficacious to reduce thrombin-mediated barrier function impairment in pre-treatment experiments with hPPAEC, showed similar efficacy to protect barrier function after thrombin exposure of HUELC5a and failed to protect barrier function after thrombin exposure of hPPAEC. These findings could be explained by ubiquitin's lower affinity for and weaker agonist activity at CXCR4, as compared with CXCL12 [38,39,[44][45][46].
Recently, we provided evidence that ubiquitin functions as a biased CXCR4 agonist, which does not recruit β-arrestin 2 to CXCR4 [47]. Thus, it appears also possible that the differences between CXCL12 and ubiquitin that we observed in the present study reflect differences in functional outcomes of balanced and biased CXCR4 signaling in lung endothelial cells.
Because none of the ACKR3 agonists affected hPPAEC barrier function and AMD3100 abolished protective effects of CXCL12 on thrombin-mediated barrier function impairment, activation of ACKR3 alone appears not to contribute to the observed effects.
Previously, CXCR4 has been shown to form heteromeric complexes with ACKR3 in expression systems and in human vascular smooth muscle cells [34-36, 48, 49]. Our present finding that PLA signals for CXCR4 and ACKR3 interactions are also detectable in hPPAEC suggests the existence of such endogenous receptor heteromers in the lung endothelium. Thus, another explanation for the observed differences between CXCL12 and ubiquitin could be that simultaneous activation of CXCR4 and ACKR3 within the heteromeric complex is more efficacious to reduce thrombin-mediated endothelial barrier impairment than activation of CXCR4 alone. To address this possibility, detailed mechanistic studies to elucidate the roles of the CXCR4: ACKR3 heteromer will be required in the future. Such experiments, however, are beyond the scope of the present study.
Among the CXCL12 variants that we tested, only CXCL12 (3-68) lacked relevant CXCR4 activity, showed significantly reduced efficacy to activate ACKR3 in Presto-Tango assays and did not attenuate thrombin-induced hPPAEC barrier function impairment. This loss of function is consistent with the loss of function of N-terminal truncated CXCL12 that has been reported previously in other assay systems [23-25].
As expected, both natural CXCL12 splice variants, CXCL12α and CXCL12α, showed comparable properties in Presto-Tango and permeability assays [26]. CXCL12 exists as a monomer at low concentrations and forms dimers at high concentrations or when bound to heparan sulfate on the endothelial surface [31, 50,51]. Consistent with previous reports, the constitutive monomeric CXCL12 variant (CXCL12 1 ) showed a behavior similar to wild type proteins in CXCR4/ACKR3 β-arrestin 2 recruitment assays and in permeability assays [27,28].
Despite activities of the disulfide-locked dimeric CXCL12 variant CXCL12 2 in β-arrestin 2 recruitment assays for CXCR4 and ACKR3 that were comparable with CXCL12α/β, CXCL12 2 showed significantly reduced potency to attenuate thrombin-induced permeability of hPPEAC. The previous finding that CXCL12 2 binds to ACKR3 with very low affinity is not contradictive to our findings in ACKR3 β-arrestin 2 recruitment assays because maximal biological responses of other GPCRs have been observed at ligand occupancies of only a small fraction of receptors [28, 52, 53] and a large receptor reserve is likely in expression systems, such as the Presto-Tango assay. The effects of CXC12 2 in CXCR4 β-arrestin 2 recruitment assays that we observed using the Presto-Tango cell system, however, are conflicting with previous measurements in intermolecular bioluminescence resonance energy transfer (BRET) assays [28]. The Presto-Tango assay utilizes a transcriptional read-out that is measured several hours after the actual signaling event. Thus, it appears possible that few β-arrestin recruitment events upon ligand binding, which may not generate a significant intermolecular BRET signal, can lead to transcription of luciferase in the Presto-Tango system. Furthermore, as compared to previous intermolecular BRET assays in which cells were exposed to CXCL12 2 for 30 min [28], cells were exposed to CXCL12 2 in our Presto-Tango assays for longer time periods, which may contribute to the observed effects. Irrespective of this discrepancy, the low potency of CXCL12 2 to inhibit thrombin-mediated barrier function impairment in the present study in combination with the previously described lack of chemotactic activity of CXCL12 2 [28,29] demonstrate that this variant does not induce the complete spectrum of biological effects that are mediated via CXCR4 and/or ACKR3 upon activation with the wild type proteins and the constitutively monomeric variant.
In agreement with the low potency of CXCL12 R47E to activate Ca 2+ signaling via CXCR4 [29], we observed that this mutant also induces β-arrestin 2 recruitment to CXCR4 and antagonizes thrombin-mediated hyperpermeability of hPPAEC with reduced potency, but retains ACKR3 activity comparable to wild type proteins. Similarly, the protease resistant mutant CXCL12 S-S4V showed reduced CXCR4 activity in Presto-Tango and permeability assays but retained ACKR3 activity. These findings are in agreement with previous effects of this mutant in CXCR4/ACKR3 β-arrestin recruitment and chemotaxis assays [30]. The observations that CXCL12 R47E and CXCL12 S-S4V showed reduced CXCR4 activity but retained ACKR3 activity further supports the assumption that protection from thrombin-mediated hPPAEC barrier impairment is mediated via CXCR4.
CXCL12 is known to bind to heparin oligosaccharides, which promotes dimerization, interferes with CXCL12 binding to CXCR4 and immobilizes CXCL12 on the endothelial surface to establish a concentration gradient required for cell trafficking [31, [54][55][56]. CXCL12 K27A/ R41A/R47A, which binds heparan sulfates with significantly reduced affinity [31], was the only mutant protein that inhibited thrombin-mediated impairment of hPPAEC barrier function with that same potency as wild type proteins and CXCL12 1 . This mutant, however, showed the lowest potency to activate CXCR4 in β-arrestin 2 recruitment assays and retained high potency to activate ACKR3. These data suggest that the K27, R41 and R47 mutations reduced the binding affinity for CXCR4 or the efficacy to induce signaling events at lower concentrations. The high potency of this mutant to reduce thrombin-induced barrier function impairment, however, can be explained by its reduced heparan sulfate binding properties, which reduces the proportion of protein that is immobilized on the surface of hPPAEC and thus, is not available for receptor activation [57]. The latter suggests that CXCL12 binding to heparan sulfate on HTLA cells, which were used in Presto-Tango assays, does not significantly affect CXCR4 binding and signaling. This implies that distinct cell surface heparan sulfate proteoglycan expression patterns between various cell types modulate CXCL12-mediated biological functions.
In conclusion, our findings suggest CXCR4 as a possible drug target to attenuate thrombinmediated impairment of lung endothelial barrier function, demonstrate that stimulation of human lung endothelial cells with cognate and non-cognate CXCR4 agonists results in functional differences and provide initial information on the structure-function relationship for CXCL12-mediated protection from thrombin-induced barrier function impairment in primary human lung endothelial cells. Our findings indicate that the protective effects of CXCL12 are dictated by its CXCR4 agonist activity and by interactions of distinct protein moieties with heparan sulfate proteoglycans on the endothelial cell surface. Interestingly, in disease conditions that are likely associated with thrombin-induced endothelial permeability impairment, such as sepsis or trauma, systemic CXCL12 concentrations have been reported to increase to levels within the range of the EC 50 for CXCL12 to attenuate thrombin-induced barrier function impairment in our permeability assays [58][59][60]. This may suggest that activation of CXCR4 by its endogenous agonists constitutes a protective mechanism to attenuate endothelial barrier function impairment by thrombin in disease conditions and implies that treatment with exogenous CXCR4 agonists augments this protective response. Our findings are expected to facilitate the development of engineered compounds with improved pharmacological properties to attenuate thrombin-induced vascular leakage in the pulmonary circulation, which may have the potential to attenuate development of lung injury and ARDS.