Differential Effects of Posttranslational Modifications of CXCL8/Interleukin-8 on CXCR1 and CXCR2 Internalization and Signaling Properties

CXCL8 or interleukin (IL)-8 directs neutrophil migration and activation through interaction with CXCR1 and CXCR2 that belong to the family of G protein-coupled receptors (GPCRs). Naturally occurring posttranslational modifications of the NH2-terminal region of CXCL8 affect its biological activities, but the underlying molecular mechanisms are only partially understood. Here, we studied the implications of site-specific citrullination and truncation for the signaling potency of CXCL8. Native CXCL8(1-77), citrullinated [Cit5]CXCL8(1-77) and the major natural isoform CXCL8(6-77) were chemically synthesized and tested in internalization assays using human neutrophils. Citrullinated and truncated isoforms showed a moderately enhanced capacity to induce internalization of CXCR1 and CXCR2. Moreover, CXCL8-mediated activation of Gαi-dependent signaling through CXCR1 and CXCR2 was increased upon modification to [Cit5]CXCL8(1-77) or CXCL8(6-77). All CXCL8 variants promoted recruitment of β-arrestins 1 and 2 to CXCR1 and CXCR2. Compared to CXCL8(1-77), CXCL8(6-77) showed an enhanced potency to recruit β-arrestin 2 to both receptors, while for [Cit5]CXCL8(1-77) only the capacity to induce β-arrestin 2 recruitment to CXCR2 was increased. Both modifications had no biasing effect, i.e., did not alter the preference of CXCL8 to activate either Gαi-protein or β-arrestin-dependent signaling through its receptors. Our results support the concept that specific chemokine activities are fine-tuned by posttranslational modifications.


Introduction
Following entrance of an infectious agent or upon tissue injury, the generation of an appropriate and strictly controlled immune response is essential. Equipped with oxidative and non-oxidative defense strategies, neutrophil granulocytes play a central role in early immune events [1]. Their capacity to release reactive oxygen species, neutrophil extracellular traps and degradative enzymes, and their phagocytic abilities, imply that the recruitment and activation of these innate leukocytes require tight regulation. In addition to highly potent but less neutrophil-specific chemotactic factors, such as of the total CXCL8 yield [45]. In addition to truncated CXCL8 variants and an elongated isoform, a natural site-specifically citrullinated CXCL8 isoform, from now on called [Cit5]CXCL8 , was identified [45]. [Cit5]CXCL8(1-77) derives from peptidylarginine deiminase-mediated conversion of the positively charged Arg residue at position 5 in the NH 2 -terminal sequence of CXCL8 to an uncharged citrulline (Cit) residue [45,46]. Although citrullination hardly affects the molecular mass of the chemokine (plus 1 Da), it significantly affects its biological activity. Aspects of the biological characteristics of CXCL8  and [Cit5]CXCL8  have been clarified to a certain extent [45,46,50,[60][61][62], but several questions remain unanswered. Hence, we aimed to further unravel the potencies of these natural CXCL8 isoforms on the two high-affinity CXCL8 receptors CXCR1 and CXCR2.
In the present study, we aimed to detangle the complex effects of posttranslational modifications of the NH 2 -terminal region of CXCL8 on several aspects of the CXCR1-and CXCR2-signaling axes. We compared the full length, unmodified CXCL8  with the naturally occurring modified isoforms CXCL8  and [Cit5]CXCL8 , not only to define the implications of specific proteolysis and citrullination of CXCL8 on Gα i protein activation through CXCR1 and CXCR2, but also to investigate the consequences of these modifications on receptor coupling to β-arrestins and on receptor internalization. We found that the two modifications modulate the potency of CXCL8 on its receptors, thereby adding another level of complexity to the CXCL8-CXCR1 and CXCL8-CXCR2 signaling loops.

Effect of Citrullination and Truncation on CXCL8-Induced CXCR1 and CXCR2 Internalization
CXCL8 is the prototype human CXCR1 and CXCR2 agonist and mediates neutrophil chemotaxis and activation upon interaction with these receptors. However, CXCL8-mediated receptor internalization, a phenomenon thought to play a major role in correctly ceasing chemotaxis, receptor re-sensitization and turnover, also occurs. We used a flow cytometry-based approach to compare the membrane expression of CXCR1 and CXCR2 on CXCL8-stimulated neutrophils with the receptor expression on control cells stimulated with buffer. We first established that native CXCL8(1-77) significantly induced internalization of naturally expressed CXCR1 and CXCR2 on freshly purified human neutrophils from 10 nM onwards. To explore the potential effects of posttranslational modifications on the capacity of CXCL8 to induce internalization, cells were exposed to 0.1 to 100 nM CXCL8(1-77), [Cit5]CXCL8  or CXCL8 . All CXCL8 isoforms clearly induced internalization of CXCR1 and CXCR2 on neutrophils in a dose-dependent manner. Moreover, CXCR1 and CXCR2 were slightly more efficiently internalized on neutrophils that were stimulated with [Cit5]CXCL8  or CXCL8(6-77) compared to cells stimulated with an equal dose of authentic CXCL8(1-77) ( Figure 1). However, the effects of site-specific citrullination and truncation on CXCL8-induced receptor internalization seemed only moderate and significant differences were only observed for some doses tested. of CXCR1 and CXCR2 on neutrophils in a dose-dependent manner. Moreover, CXCR1 and CXCR2 were slightly more efficiently internalized on neutrophils that were stimulated with [Cit5]CXCL8  or CXCL8(6-77) compared to cells stimulated with an equal dose of authentic CXCL8(1-77) (Figure 1). However, the effects of site-specific citrullination and truncation on CXCL8induced receptor internalization seemed only moderate and significant differences were only observed for some doses tested.

Effect of Citrullination and Truncation on CXCL8-Induced Gα i -Dependent Signaling
CXCR1 and CXCR2 are GPCRs that upon agonist stimulation mainly couple to Gα i proteins, resulting in a decrease of endogenous cAMP levels. To define the capacity of CXCL8 isoforms to induce Gα i -signaling, CXCR1-and CXCR2-transfected HEK-293T cells were stimulated with 0.1 to 100 nM CXCL8 in the presence of forskolin. Forskolin enhances endogenous cAMP concentrations, which were measured with the Alphascreen technology [63]. Regarding CXCR1-mediated signaling, we found that the potency of [Cit5]CXCL8(1-77) and CXCL8  to activate G protein-dependent signaling was significantly enhanced, as indicated by significantly lower intracellular cAMP levels upon G protein stimulation with these isoforms as compared to native CXCL8(1-77) (Figure 2A,B). Specific EC50 values of CXCL8 forms are reported in Figure 2C. Analogously, both modified CXCL8 forms were significantly stronger inducers of G protein-dependent signaling through CXCR2 than native CXCL8(1-77) ( Figure 2D-F). In general, these results showed that the potency of site-specifically truncated and citrullinated CXCL8 forms was at least ten-fold higher on both receptors.
we found that the potency of [Cit5]CXCL8(1-77) and CXCL8  to activate G protein-dependent signaling was significantly enhanced, as indicated by significantly lower intracellular cAMP levels upon G protein stimulation with these isoforms as compared to native CXCL8(1-77) (Figure 2A,B). Specific EC50 values of CXCL8 forms are reported in Figure 2C. Analogously, both modified CXCL8 forms were significantly stronger inducers of G protein-dependent signaling through CXCR2 than native CXCL8(1-77) ( Figure 2D-F). In general, these results showed that the potency of sitespecifically truncated and citrullinated CXCL8 forms was at least ten-fold higher on both receptors.  In dose-response plots, results are represented as percentages of activity in cAMP reduction over values obtained with authentic CXCL8(1-77) ± SEM. Mean LogEC50 ± SEM were obtained by nonlinear regression curve fitting in the cAMP assays. Results were statically compared by one-way Anova with Tukey multiple comparison. * p < 0.05, ** p < 0.01.

Effect of Citrullination and Truncation on CXCL8-Induced β-arrestin Recruitment to CXCR1 and CXCR2
Although classical CXCR1-and CXCR2-signaling is Gα i protein-dependent, ligand-stimulation also leads to initiation of signaling pathways that do not involve G proteins, among which the best known is mediated by β-arrestins. The effect of posttranslational modifications on the ability of CXCL8 to recruit β-arrestins 1 and 2 to CXCR1 and CXCR2 was therefore examined. HEK-293T cells were co-transfected with RLuc-tagged CXCR1 or CXCR2 and EYFP-tagged β-arrestin 1 or β-arrestin 2. When in close proximity, the RLuc-tag on the receptor excites the acceptor fluorophore EYFP on the β-arrestins via energy transfer resulting in light emission at 530 nm, thereby allowing investigation of β-arrestin recruitment induced by CXCL8. The obtained results were converted into the ratio of emission at 530 nm/emission at 480 nm [Bioluminescence Resonance Energy Transfer (BRET) ratio], which is related to energy transfer. CXCL8(1-77) as well as [Cit5]CXCL8(1-77) and CXCL8(6-77) induced recruitment of both β-arrestins 1 and 2 to CXCR1 ( Figure 3A,B and Figure 4A,B) and CXCR2 ( Figure 3C,D and Figure 4C,D). Either site-specific citrullination or loss of its five most NH 2 -terminal residues resulted in an increased potency of CXCL8 to induce β-arrestin 2 recruitment via both receptors ( Figure 4) and of β-arrestin 1 recruitment through CXCR1 ( Figure 3A,B). Truncation, but not citrullination, also tended to increase the potency of CXCL8 to induce β-arrestin 1 recruitment to CXCR2 ( Figure 3C,D). It is worth mentioning that detection of β-arrestin 1 association seemed to occur less efficiently. As at present we cannot define whether this was a biological fact or a technical insufficiency, we focused on β-arrestin 2 recruitment. Dose-response experiments showed that citrullination did not significantly alter the potency of CXCL8 to induce β-arrestin 2 to CXCR1 ( Figure 5A). In contrast, the truncated isoform CXCL8(6-77) was almost ten times more potent at inducing β-arrestin 2 recruitment to this receptor ( Figure 5B,C). Regarding CXCR2-mediated signaling, [Cit5]CXCL8(1-77) as well as CXCL8(6-77) were more powerful inducers of β-arrestin 2 recruitment than native CXCL8(1-77) ( Figure 5D-F). occur less efficiently. As at present we cannot define whether this was a biological fact or a technical insufficiency, we focused on β-arrestin 2 recruitment. Dose-response experiments showed that citrullination did not significantly alter the potency of CXCL8 to induce β-arrestin 2 to CXCR1 ( Figure  5A). In contrast, the truncated isoform CXCL8(6-77) was almost ten times more potent at inducing βarrestin 2 recruitment to this receptor ( Figure 5B,C). Regarding CXCR2-mediated signaling, [Cit5]CXCL8(1-77) as well as CXCL8(6-77) were more powerful inducers of β-arrestin 2 recruitment than native CXCL8(1-77) ( Figure 5D-F).

No Biasing Effect of Citrullination or Truncation on CXCL8-Induced Signaling Through CXCR1 and CXCR2
In addition to investigating the direct implications of NH 2 -terminal processing for individual signaling cascades, we wondered whether citrullination or truncation had a biasing effect on CXCL8-induced signaling. Therefore, we examined whether these modifications altered the tendency of CXCL8 to preferentially induce either Gα i -dependent signaling or β-arrestin recruitment. To compare these activities, we relied on the operational model of bias. This standardized model has the advantage of normalizing signals obtained with all isoforms tested, regardless of whether these are full or partial receptor agonists, and allows bias quantification without risks of system interference or observational bias that may manifest due to the experimental set up [64]. Using this approach, we calculated the transduction coefficient or log(τ/KA), with τ incorporating the ligand efficacy, receptor density and taking into account the assay system and KA being the reciprocal of the conditional ligand affinity in the assay system [65][66][67], demonstrate that neither site-specific citrullination nor removal of its five most NH 2 -terminal amino acids turned CXCL8 into a preferential inducer of Gα i -dependent-or β-arrestin 2 pathways for both CXCR1 and CXCR2 ( Figure 6). or observational bias that may manifest due to the experimental set up [64]. Using this approach, we calculated the transduction coefficient or log(τ/KA), with τ incorporating the ligand efficacy, receptor density and taking into account the assay system and KA being the reciprocal of the conditional ligand affinity in the assay system [65][66][67], demonstrate that neither site-specific citrullination nor removal of its five most NH2-terminal amino acids turned CXCL8 into a preferential inducer of Gαidependent-or β-arrestin 2 pathways for both CXCR1 and CXCR2 ( Figure 6).
Stimulation with high chemokine concentrations may elicit receptor internalization rather than initiating Gα i -dependent pathways [20][21][22][23][24][25][26], and prolonged stimulation with the chemokine ligand generally ends up in downregulation of the receptor. Indeed, it has been suggested that during leukocyte migration along a chemokine gradient, receptor internalization adequately stops the cell from migrating when the chemokine concentration is no longer increasing, thereby ceasing chemotaxis. Given the increased overall activity of CXCL8(6-77), one may speculate that the threshold of CXCL8(6-77) to induce receptor internalization is also lower compared to CXCL8(1-77). However, the results of the present study show that loss of its five most NH 2 -terminal amino acids only moderately enhances the potency of CXCL8 to induce internalization of its receptors, implying that a specific modification may have a divergent effect on different downstream pathways.
In the context of GPCR internalization, an important role has been established for β-arrestins. β-arrestins are adaptor proteins that modulate the chemokine-receptor signaling axis through interaction with phosphorylated Ser and Thr residues present in the receptor COOH-terminal tail [74][75][76][77][78][79]. However, increasing evidence suggests that also other receptor regions are essential for manifestation of β-arrestin-mediated processes, including internalization [21]. Of note, experimental evidence suggests that the effects of β-arrestins on the chemokine-receptor axis extend beyond the mere uncoupling of chemokine receptors from conventional signaling pathways and promoting receptor internalization. Specifically, in addition to their role as uncoupling molecules, β-arrestins may function as scaffolding proteins at the leading edge, where they recruit signaling molecules and play a role in actin skeleton regulation [80][81][82][83]. These features imply a complex role for β-arrestins in fine-tuning numerous migration-related intracellular pathways. The highly similar β-arrestin 1 and β-arrestin 2 have 78% identical amino acids and have been mainly compared in in vitro assays. In addition to many overlapping functions, emerging evidence suggests that they each fulfil certain unique roles, a phenomenon that likely depends on the receptor involved [84]. The only chemokine receptors studied in this context are CXCR2 [85], CXCR4 [86] and CXCR7 (also known as Atypical Chemokine Receptor 3) [87]. For CXCR2, an exclusive role for β-arrestin 2, but not β-arrestin 1, was found in upregulating β2-integrin expression, adhesion strengthening and activation of the GTPase Rap 1 [85]. In the present study, we established that CXCL8 is able to induce recruitment of both β-arrestins 1 and 2 to CXCR1 and CXCR2. Focusing on β-arrestin 2 association, we demonstrated that cleavage to CXCL8(6-77) significantly enhanced the potency of CXCL8 to mediate β-arrestin recruitment to both high-affinity CXCL8 receptors. These results may partially explain the slightly enhanced capacity of CXCL8(6-77) to induce internalization of CXCR1 and CXCR2.
The citrullinated isoform [Cit5]CXCL8(1-77) results from peptidylarginine deiminase-mediated, site-specific deimination of Arg at position 5 in the NH 2 -terminal domain of CXCL8 to Cit, and was found in cell culture supernatant from stimulated leukocytes [45]. Citrullination negatively affects the glycosaminoglycan-binding properties of CXCL8 and moderately reduces its Ca 2+ signaling capacity through CXCR2, but slightly enhances the ability of the chemokine to upregulate integrin expression on neutrophils [45,46]. Additionally, the potency of CXCL8 to evoke ERK phosphorylation is impaired upon citrullination. Strikingly, the presence of the Cit at position 5 protects CXCL8 from potentiation by thrombin-and plasmin-mediated cleavage to CXCL8 . Regarding the in vivo consequences of citrullination for the activity of CXCL8, results differ depending on the location of CXCL8 in the body. Compared to native CXCL8, the citrullinated isoform induced less neutrophil extravasation upon intraperitoneal (i.p.) injection in mice [45]. However, following intravenous injection in rabbits, citrullinated CXCL8 was a more potent neutrophil-chemoattractant than authentic CXCL8 [46]. Thus, the used experimental system and potentially the difference between either location in blood vessels or in subendothelial tissues seems of influence for the final outcome. Although concentrations of natural [Cit5]CXCL8(1-77) are probably rather low, scientific evidence suggests that protein citrullination is a more common phenomenon in an inflamed environment, such as rheumatoid arthritis [88,89]. It remains to be determined whether [Cit5]CXCL8  or other citrullinated chemokines are biomarkers for certain inflammatory diseases. Despite the fact that [Cit5]CXCL8(1-77) was originally reported to induce Ca 2+ mobilization and ERK phosphorylation through CXCR2 on transfected HEK-293 cells less potently, we found that the citrullinated isoform is a more potent inhibitor of cAMP formation through CXCR1 and CXCR2 than native CXCL8 . We also demonstrated that the potency of [Cit5]CXCL8(1-77) to induce internalization of CXCR1 and CXCR2 is moderately, enhanced compared to authentic CXCL8 . Additionally, we found that its capacity to mediate recruitment of β-arrestin 2 to CXCR2, but not CXCR1, was significantly increased. Overall, these data underscore the complex effects of chemokine citrullination in general and suggest that, at least for CXCL8, different effects on specific biological activities may be observed in different experimental models.
Biased signaling, i.e., the concept that a specific receptor selectively activates one out of multiple possible signaling cascades or that a modification of the ligand leads to preference for a specific receptor, has been clearly shown for GPCRs in general and also specifically in the chemokine field. This aspect is believed to contribute to specificity in the chemokine system, as both receptor-and ligand-defined bias, as well as cell-or tissue-dependent bias, may manifest [29,90,91]. Therefore, we not only evaluated the effect of site-specific citrullination or truncation of CXCL8 in individual signaling pathways, but also examined whether these modifications turned CXCL8 into a potential biased ligand. However, although citrullination and truncation affected the potency of CXCL8 in Gα i and in β-arrestin assays, these modifications did not modify the preference of CXCL8 for specific downstream signaling pathways.
To conclude, our results support the notion that NH 2 -terminal processing affects the activity of CXCL8. We demonstrated that site-specific citrullination or removal of its five most NH 2 -terminal residues significantly modulates the signaling properties of CXCL8. Both modifications increase the potency of CXCL8 to inhibit adenylyl cyclase through CXCR1 and CXCR2. The effects of truncation and citrullination on its capacity to induce internalization of CXCR1 and CXCR2 were only moderate, though significant for most doses tested. Moreover, compared to authentic CXCL8, CXCL8(6-77) showed a significantly increased ability to recruit β-arrestin 2 to its receptors. Citrullination to [Cit5]CXCL8(1-77) only significantly affected the potency of CXCL8 to induce β-arrestin 2 recruitment to CXCR2. Given the potential drastic effects of chemokine modification in general, we predict that it will be necessary to unravel the abundance and role of specific chemokine isoforms during physiological and pathophysiological conditions in humans to obtain a profound understanding of the role of chemokines in health and disease. In favor of this hypothesis, former research already demonstrated that intact CXCL8(1-77) is usually most abundant in cell culture supernatant from endothelial cells and fibroblasts, whereas stimulated leukocytes preferably process the authentic molecule into truncated and more active isoforms [55][56][57][58][59]. This may imply that chemokine processing becomes more important in inflammatory conditions, where the numbers of circulating leukocytes and release of chemokine-modifying enzymes can dramatically increase. A major future challenge will be the development of technical equipment that allows simultaneous detection and quantification of multiple chemokine isoforms in human samples.

4.1.Chemokines
Most conventional purification techniques, including reverse phase-high performance liquid chromatography and ion exchange chromatography, do not allow separation of individual CXCL8 isoforms from biological samples. CXCL8(1-77), [Cit5]CXCL8(1-77) and CXCL8(6-77) were therefore chemically synthesized as previously described [60,92]. Briefly, full length native CXCL8 of 77 amino acids and the natural-occurring isoforms [Cit5]CXCL8(1-77) and CXCL8(6-77) were chemically synthesized with Fmoc chemistry using an Activo P11 solid-phase peptide synthesizer (Activotec, Cambridge, UK). The synthesized chemokines were purified to homogeneity by reverse phase-high performance liquid chromatography on a Source 5-RPC column (GE Healthcare, Uppsala, Sweden) and elution of the synthesized proteins was performed with an acetonitrile gradient in 0.1% (v/v) trifluoroacetic acid, with 0.7% of the column effluent being directly injected into a Bruker Amazon SL electrospray-ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany). Online mass spectrometry was used to select fractions containing homogeneous CXCL8 isoforms. These were pooled, evaporated and diluted in ultrapure water. The purified CXCL8 isoforms were folded into their correct configuration as described [92] and analyzed by mass spectrometry. The identity of the chemokines was also confirmed with automated NH 2 -terminal sequencing based on the principle of Edman degradation (Procise 491 cLC sequencer, Applied Biosystems, Foster City, CA, USA). Lastly, specific CXCL8 ELISAs, sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and bicinchoninic acid protein assays (Pierce, Woodland Hills, CA, USA) were conducted to examine chemokine concentrations and purity [92,93].

β-arrestin Recruitment Assay
The chemokine-induced recruitment of β-arrestins to CXCR1 and CXCR2 was examined using the Bioluminescence Resonance Energy Transfer 1 (BRET1) technique [63]. Transfected cells (vide supra) were washed with PBS and harvested using 0.02% (w/v) EDTA (Lonza). The amount of protein was determined with a Bio-Rad DC protein assay kit (BioRad, Hercules, CA, USA) and cell pellets were resuspended at final concentrations of 1 mg/mL in PBS enriched with 0.1% (w/v) glucose to seed 80 µg of cells per well in an opaque white well/black frame 96-well plate (Perkin Elmer, Waltham, MA, USA). For kinetic BRET measurements 5 µM coelenterazine H was added and after incubation in the dark for 8 min at RT, PBS or chemokine variants at final concentrations of 100 nM were added and light emission was immediately measured at 460-500 (RLuc) and 510-550 (EYFP) sequentially for each well using a Synergy H4 Hybrid reader (Biotek, Bad Friedrichshall, Germany). To perform dose-response experiments, 80 µg of cells were seeded per well in an opaque white well/black frame 96-well plate. Different concentrations of CXCL8 forms were added and the plate was incubated at 37 • C for 15 min before the addition of 5 µM coelenterazine H and further 8 min incubation at RT, before light emission measurement. BRET ratios were calculated dividing luminescence values obtained with the EYFP filter by the values obtained using the RLuc filter, and the ligand effect on BRET was obtained subtracting BRET values obtained by cells stimulated with PBS from BRET values obtained by cells stimulated with the different CXCL8 variants.

cAMP Assay
To measure intracellular cAMP, the AlphaScreen cAMP assay (Perkin Elmer) was used according to the manufacturer's instructions. Briefly, CXCR1-and CXCR2-transfected HEK-293T cells (vide supra) were harvested and washed with PBS prior to being resuspended at final concentrations of 10 7 cells per ml in freshly prepared stimulation buffer composed of 1 × 'Hanks' Balanced Salt Solution' (HBSS; Lonza) complemented with 0.1 % (w/v) BSA, 0.5 nM IBMX (Sigma-Aldrich) and 5 mM HEPES (Lonza). Per well, 10 4 cells were plated in a 384-well white opaque plate (Perkin Elmer) with 0.2 units/µL anti-cAMP acceptor beads in 5µL volume, to which 5µL of 10 µM forskolin and serial chemokine dilutions were added. After 30 minutes incubation at RT in the dark, 15µL of lysis buffer containing 1 Unit of streptavidin-labeled donor beads and 1 Unit biotinylated cAMP were added and the plate was incubated at RT in the dark for 1 h. The luminescence of the beads was read on a Synergy H4 plate reader through a 570/100 nm filter after sample excitation with 680/30 nm filtered light.

Internalization Assay
The potency of CXCL8 isoforms to induce internalization of CXCR1 and CXCR2 on neutrophils was investigated using a flow cytometry-based approach, as previously described [94]. Freshly purified human neutrophils were resuspended in RPMI1640 medium (Cambrex Corporation, East Rutherford, NJ, USA) complemented with 0.5% (w/v) human serum albumin (HSA; Red Cross Blood transfusion center, Leuven, Belgium) at final concentrations of 5 × 10 6 cells per ml. Per well, 5 × 10 5 cells were plated in a round bottom 96-well plate, which was placed at 37 • C for 5 min. Cells were stimulated with 0.1 to 100 nM CXCL8 isoforms for 1 h at 37 • C. Stimulation with solely buffer was used as a control. The plate was kept on ice for 10 min, centrifuged (5 min, 315 g, 4 • C), and pellets were diluted in PBS enriched with 2% (v/v) fetal bovine serum (FBS; Invitrogen). Following centrifugation (5 min, 315 g, 4 • C), cells were stained with APC-labeled mouse anti-human CXCR1 and PE-labeled mouse anti-human CXCR2 antibodies (BD Pharmingen). After 30 min incubation in the dark on ice, cells were centrifuged (5 min, 315 g, 4 • C) and washed two times with PBS + 2% (v/v) FBS. Final pellets were resuspended in PBS enriched with 2% (v/v) FBS and 0.4% formaldehyde. Fluorescence intensities were determined with a BD FACSCaliburTM flow cytometer (BD Biosciences) and results were analyzed with compatible software. Relative mean receptor expression levels were calculated as a percentage as 100 × (MFI CXCL8 ) / (MFI Buffer ), with 'MFI CXCL8 ' and 'MFI Buffer ' being the mean fluorescence intensities (MFI) obtained after treatment with CXCL8 or buffer, respectively.

Bias Calculation
To determine whether post-translational modifications of CXCL8 can induce functional selectivity on CXCR1 and CXCR2, we calculated for each isoform its Transduction Coefficient (TC = log(τ/KA)). First, using Graphpad Prism we fitted the Black-Leff operational model to dose-response curves of cAMP reduction and β-arrestin 2 recruitment [63,65]. Native CXCL8(1-77) was considered as reference ligand. Therefore, we calculated for each assay the relative effectiveness of each post-translationally modified CXCL8 variant subtracting from its TC the TC of native CXCL8 (∆log(τ/KA) = log(τ/KA)CXCL8 isoform − log(τ/KA)CXCL8(1-77)). Relative effectiveness of each chemokine isoform was then used to calculate the bias factor (∆∆log(τ/KA) = ∆log(τ/KA)cAMP reduction − ∆log(τ/KA)β-arrestin 2 recruitment) which reflects the ability of each post-translational modification to preferentially activate one single signaling pathway downstream of each receptor.

Statistical Analysis
Wilcoxon matched pairs tests were performed to evaluate whether results of paired groups in internalization assays were significantly different or not. LogEC50 values obtained in cAMP and β-arrestin assays and bias factors were statistically compared using one-way ANOVA tests with Tukey's multiple comparison. A p value of 0.05 or less was considered significant.