CXCR3 (GPR9/CD183) is a receptor for four CXC chemokines: monokine induced by gamma interferon (MIG), interferon gamma-induced protein 10 (IP-10), interferon-inducible T-cell alpha chemoattractant (I-TAC), and platelet factor 4 (PF-4). The expression levels of MIG, IP-10, and I-TAC are usually moderated under homeostatic conditions but upregulated in inflammation. As their name implies, they can be induced by interferon gamma (IFN-γ) in monocytes, endothelial cells, fibroblasts, and some cancer cells (Tokunaga et al., 2018). IP-10 also is induced by IFN-α, IFN-β, and weakly induced by tumor necrosis factor α (TNF-α). I-TAC is induced only by IFN-γ and IFN-β (Groom and Luster, 2011; Ohmori et al., 1993). PF-4, an α-granule CXC-chemokine, is released by activated platelets during acute vascular injury or chronic disease (Files et al., 1981; Kasper et al., 2007). PF-4 is involved in regulating some immune cells such as basophils, T cells, natural killer (NK) cells, monocytes, and endothelial cells (Fleischer et al., 2002; Martí et al., 2002; Mueller et al., 2008; Schenk et al., 2002). CXCR3 is expressed in a variety of cells including CD4⁺ and CD8⁺ T cells, monocytes, NK cells, dendritic cells, and some cancer cells (Groom and Luster, 2011; Loetscher et al., 1996; Thomas et al., 2003; Yamamoto et al., 2000). CXCR3 and its ligands have important roles in inflammatory conditions by leukocyte trafficking, especially CD4⁺ and CD8⁺ T cell (Farber, 1997). As regard to cancer, CXCR3 and its ligands are upregulated in many primary and metastatic tumors of colon, prostate, stomach, ovary, tongue, and skin (Kawada et al., 2004; Li et al., 2018; Windmüller et al., 2017; Wu et al., 2012; Yang et al., 2016; Zipin-Roitman et al., 2007). The chemokine-CXCR3 axis has been considered as a promising target for cancer treatment along with other immunotherapies (Reynders et al., 2019; Tokunaga et al., 2018).

Many chemokine receptors belonging to the rhodopsin family of G protein-coupled receptors (GPCRs) are translated from a single exon. Some of these gene transcripts proceed through alternative splicing pathways to produce chemokine receptors containing different amino acids at the N-terminal region. Humans have three splicing variants of CXCR3: CXCR3A, CXCR3B, and CXCR3Alt. CXCR3A contains 368 amino acids, and was first identified as a selective functional receptor for IP-10 and MIG (Loetscher et al., 1996). I-TAC was identified as another CXCR3 ligand with significantly higher affinity than IP-10 and MIG (Cole et al., 1998). The open reading frame of CXCR3A is encoded in two exons: the second exon encodes the main receptor fragment, whereas the first exon only encodes four amino acids. CXCR3B contains 415 amino acids (aa) encoded in a single exon, which is relatively longer than the second exon of CXCR3A. The CXCR3A and CXCR3B isoforms share the same 364 aa sequence but have different N-terminal amino acid sequences: CXCR3A contains 4 N-terminal residues, whereas CXCR3B contains 51 N-terminal residues (Lasagni et al., 2003). CXCR3Alt is generated by posttranscriptional exon skipping, resulting in the truncation of 159 aa and the addition of 59 aa by frame shift. This deletion/replacement alters the C-terminal region after the fourth transmembrane (TM) domain, but the functional consequences have not been determined (Ehlert et al., 2004).

CXCR3A is representative of CXCR3 activity in studies of chemokine-stimulated cellular responses. After chemokine binding, CXCR3A activates Gαi/o, leading to downstream events such as intracellular Ca²⁺ flux, ERK1/2 phosphorylation, and β-arrestin recruitment (Berchiche and Sakmar, 2016; Cole et al., 1998; Loetscher et al., 1996; Smith et al., 2017). CXCR3A activation can induce cellular chemotaxis and proliferation (Li et al., 2019; Yang et al., 2016). By contrast, CXCR3B may exert different functions, although it mediates signaling events through activation by high concentrations of I-TAC. CXCR3A is upregulated during cancer progression, whereas CXCR3B is downregulated in cancers such as ovarian and prostate tumors. Downregulation of CXCR3A and upregulation of CXCR3B is negatively correlated to proliferation and invasion of prostate cancer cells (Wu et al., 2012). Upregulation of CXCR3B is considered as a favorable prognostic marker in patients with gastric cancer (Yang et al., 2016). CXCR3B may suppress cell growth and angiogenesis, and may not be involved in cell chemotaxis, although this is debated in the literature (Lasagni et al., 2003). CXCR3Alt unexpectedly displayed pertussis toxin-sensitive chemotactic activity toward I-TAC despite its structural defects, possibly due to its intact N-terminus; however, this result was not repeated in other studies (Ehlert et al., 2004). Physiological functions of CXCR3Alt remain unknown with very few studies.

Investigations of receptor interaction with different ligands and subsequent signaling activation and cellular responses have primarily focused on CXCR3A, although there are a few studies on the splicing variants. Therefore, a comprehensive study of the interactive relationships between CXCR3 variants and their four ligands (MIG, IP-10, I-TAC, and PF-4) is crucial to determine the pathophysiological relevance of the receptors in immune responses and diseases including cancers. Here, we investigated the expression profiles of CXCR3 variants and their distinctive and interactive roles in cellular responses to the cognate chemokines (e.g., intracellular signaling and chemotactic cell migration) using a molecular interaction-based detection method called NanoBiT technology.

Materials

PCR primers and related reagents were obtained from Cosmo Genetech (Korea). Cell culture media (DMEM, RPMI-1640, and Opti-MEM) were from Invitrogen (USA) and WELGENE (Korea). All chemokines were obtained from PeproTech (USA). The pcDNA3.1 vector was purchased from Invitrogen. The SRE-luciferase (SRE-Luc) vector was acquired from Stratagene (USA). Anti-HA antibody (Cat. No. H3663) was obtained from Sigma-Aldrich (USA); anti-pERK (Thr202/Tyr204) antibody (Cat. No. 4370) was obtained from Cell Signaling Technology (USA); and anti-ERK antibody (Cat. No. sc-514302) was obtained from Santa Cruz Biotechnology (USA). Anti-rabbit and anti-mouse secondary antibodies were obtained from SeraCare (USA). The NanoBiT starter kit, pBiT3.1 plasmid, and all reagents for the NanoBiT assay were obtained from Promega (USA). Other reagents and chemicals were purchased from Sigma-Aldrich unless otherwise stated.

Cell culture

HEK293, HeLa, A549, Jurkat, and U937 cells were obtained from the American Type Culture Collection (ATCC, USA). HEK293 and HeLa cells were maintained in DMEM supplemented with 10% fetal bovine serum (FBS), 100 IU/ml penicillin G, and 100 μg/ml streptomycin (Invitrogen). A549, Jurkat, and U937 cells were maintained in RPMI-1640 supplemented with 10% FBS, 100 IU/ml penicillin G, and 100 μg/ml streptomycin. Cells were cultured at 37°C in a 5% CO₂ incubator. HEK293 cells expressing Gαqi chimera were developed by transfection and antibiotic selection.

Plasmid construction

Human cytomegalovirus (CMV) promoter sequence in pcDNA3.1 was replaced with the promoters of Ubiquitin C (UbiC) or Herpes simplex virus thymidine kinase type 1 (HSV-TK) to develop a regulated expression system. The transcription activities of each promoter were confirmed in previous study (Nguyen et al., 2020a). CXCR3A, CXCR3B, CXCR3Alt genes were inserted into these vectors. The Nano-Luciferase gene fragments in the NanoBiT vectors (Promega) were inserted into a multicloning site in the UbiC promoter-vector. All genes for the NanoBiT assay were constructed as N-terminal or C-terminal tagged forms in the vector containing the UbiC promoter. Mini-G protein constructs were developed to clarify specific interactions between CXCR3 and Gα subunits as described in previous reports (Nehmé et al., 2017; Wan et al., 2018).

RT-PCR

Total RNA was isolated from cells using TRIzol according to the manufacturer’s instructions (Invitrogen). Next, cDNA was synthesized from 2 μg of RNA using M-MLV reverse transcriptase (Promega). The following primer pairs were used to amplify the CXCR3 variant-specific fragments of 448 bp, 706 bp, and 454 bp: human CXCR3A (forward 5′-TGAGGTGAGTGACCACCAAG-3′, reverse 5′-GTTCAGGTAGCGGTCAAAGC-3′); human CXCR3B (forward 5′-CACAGGTGAGTGACCACCAAG-3′, reverse 5′-ATGTGGGCATAGCAGTAGGC-3′); and human CXCR3Alt (forward 5′-ACAGCCTCCTCTTTCTGCTG-3′, reverse 5′-GACCCCTGTGGGAAGTTGTA-3′), respectively. The PCR reaction was performed on a thermal cycler (Thermo Fisher Scientific, USA) using the following conditions: 95°C for 5 min; 35 cycles (95°C for 30 s, 60°C for 30 s, 72°C for 50 s); and 72°C for 10 min. PCR products were separated by 1.5% agarose gel electrophoresis.

HiBiT assay for membrane expression

The Nano-Glo HiBiT was applied to quantify the expression of N-terminal HiBiT-tagged receptors in the cell membrane. HEK293 cells (2 × 10⁴ cells/100 μl) were seeded in a white 96-well plate (Costar, USA). The next day, cells were transfected with a mixture of HiBiT-receptor constructs (0.5, 5, or 50 ng DNA) and 0.1 μl of Lipofectamine 2000 (Invitrogen). Twenty-four hours after transfection, the medium was replaced with serum-free medium, and incubated for 10 min at room temperature. Then, 100 μl of Nano-Glo HiBiT Lytic Reagent Mix (1 μl of LgBiT protein + 2 μl of Nano-Glo HiBiT substrate + 97 μl of Nano-Glo HiBiT Lytic Buffer) was added to each well. The reactions were equilibrated for 4 min at room temperature, and then luminescence was measured using a luminometer (BioTek, USA).

NanoBiT complementation assay for protein interactions

CXCR3 receptor interactions with either β-arrestin or mini-G protein (an engineered Gα subunit) were observed using the NanoLuc Binary Technology (NanoBiT). HEK293 cells (2 × 10⁴ cells/well) were seeded into 96-well plates. The next day, 50 ng of receptor plasmids (fused with LgBiT or SmBiT at the C-terminus) and 50 ng of β-arrestin or mini-G protein plasmids (fused with SmBiT or LgBiT) were mixed with 0.2 μl Lipofectamine 2000 and added to each well. This reaction mix also was used for all other two-gene combination assays. For receptor coexpression assays, 40 ng of the tested CXCR3 receptor and 40 ng of β-arrestin NanoBiT constructs along with 80 ng of another CXCR3 receptor under different promoters were mixed with 0.3 μl Lipofectamine 2000 and added to each well. At 24 h after transfection, the medium was replaced with 100 μl of Opti-MEM. Then, 25 μl of Nano-Glo Live Cell Reagent (furimazine) was added to each well, and baseline luminescence was measured for the first 10 min. Finally, cells were stimulated with 10 μl of ligand at a concentration of 100 ng/ml, and the change in luminescence was measured for 1 h.

Detection of intracellular calcium increase

Changes in intracellular calcium were detected using a new NanoBiT-based method (Nguyen et al., 2020b). HEK293 or HEK293-Gαqi cells were seeded in a 96-well plate. The next day, 30 ng of CXCR3 receptor plasmids and 30 ng of each construct of calmodulin-SmBiT and LgBiT-MYLK2S were mixed with 0.2 μl Lipofectamine 2000 (Invitrogen) and added to each well. For receptor cotransfection, 30 ng of the CXCR3 receptor plasmids and 50 ng of the other plasmids containing different promoter-driven receptors were cotransfected with 30 ng of calmodulin-SmBiT and 30 ng of LgBiT-MYLK2S constructs. These four plasmids were mixed with 0.3 μl Lipofectamine 2000 and added to each well. At 24 h post transfection, the medium was replaced with 100 μl of Opti-MEM. Cells were stabilized for 10 min at room temperature before measuring the luminescence. Then, 25 μl of Nano-Glo Live Cell Reagent was added to each well, and basal luminescence was measured for the first 10 min for signal stabilization. Finally, cells were stimulated with 10 μl of ligand, and the change in luminescence was measured for 30 min.

Reporter gene assay

HEK293-Gαqi cells in 48-well plates were transfected with the following mixture: 75 ng of pcDNA3.1/CXCR3A, pcDNA3.1/CXCR3B, or pcDNA3.1/CXCR3Alt; 75 ng of SRE-Luc reporter gene plasmid; and 0.3 μl of Lipofectamine 2000. Next day, after aspiration of culture media cells were kept in serum-free DMEM overnight, treated with 100 ng/ml ligands for 6 h, and lysed with 100 μl of lysis buffer. The luciferase activity of cell extracts was measured with luminometer.

Western blotting and coimmunoprecipitation

Cells were lysed in lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 10 mM KCl, 1% Triton X-100, 10 mM NaF) with protease inhibitor cocktail (Roche, USA). Cell extracts were separated by 10% SDS-PAGE gel. Then, proteins were transferred to nitrocellulose membranes. The nitrocellulose membrane was blocked with 5% skimmed milk, and then probed with appropriate antibodies. Finally, the HRP-conjugated secondary antibody signal was developed using an enhanced chemiluminescence kit (Thermo Fisher Scientific).

For coimmunoprecipitation, HEK293 cells in 60-mm dishes were transfected with the HA-tagged CXCR3A plasmid and each of the FLAG-tagged CXCR3 variant plasmids. At 36 h after transfection, cells were washed with cold phosphate-buffered saline (PBS) and lysed with 1 ml lysis buffer for 30 min. The lysate was centrifuged at 15,000 rpm for 15 min at 4°C. The supernatant was incubated with anti-FLAG antibody-conjugated beads (#A2220; Sigma-Aldrich) at 4°C for 2 h. The beads were washed with lysis buffer, and the precipitates were separated in a 10% SDS-PAGE gel. The coprecipitated proteins were detected by immunoblotting with anti-HA antibodies.

Chemotaxis assay

Chemotactic migration of Jurkat cells harboring CXCR3 variants was assessed using 24-well transwell plates with 5-μm pore size (Corning, USA). The upper compartment of the transwell was loaded with 100 μl of 2 × 10⁶ cells/ml in RPMI medium containing 0.1% bovine serum albumin (BSA), whereas the lower well was loaded with 650 μl of 0.1% BSA containing 100 ng/ml chemokine. The plate was incubated at 37°C in 5% CO₂ for 3 h. The upper chamber was then removed, and the cells that migrated into the bottom well were counted using a hemocytometer. All experiments were performed three times in duplicate.

Statistical analysis

All statistical analyses were performed using a Prism 9 software (GraphPad, USA). Statistical differences among experimental groups were analyzed by unpaired Student’s t-tests or one-way ANOVA. Group means were further analyzed using Tukey’s multiple comparison test. P values less than 0.05 were considered statistically significant. Data were presented as mean ± SD, and all experiments were performed in triplicate unless otherwise indicated.

Expression of CXCR3 variants in human cell lines

Many reports described CXCR3 expression in various cells, although its functional roles were determined primarily in immune cells, especially T lymphocytes (Loetscher et al., 1996; Thomas et al., 2003; Yamamoto et al., 2000). Alternative splicing may generate four CXCR3 variants including CXCR3Alt-B form (Fig. 1A): the gene structures were identified based on cDNA information in NCBI GenBank (https://www.ncbi.nlm.nih.gov/nucleotide/), and the gene sequences were retrieved from the Ensembl genome browser database (https://asia.ensembl.org/). CXCR3A and CXCR3B share most residues except for some amino acids of the N-terminus. However, the CXCR3Alt forms are spliced in the middle of the main exon, resulting in the deletion of TM domains 5, 6, and 7, and substitution of the C-terminal sequence with other amino acids. The CXCR3 variant mRNAs were detected in HEK293, A549, Jurkat, and U937 cells, suggesting that all CXCR3 variants may be expressed in these cells (Fig. 1B).

Endogenously expressed GPCRs are generally difficult to detect by western blotting because of the quality of the antibodies or the TM protein expression level. Therefore, CXCR3 variant expression was investigated using C-terminal HA- or FLAG-tagged forms. When the cell lysates in 1.5× sample buffer were subjected to western blotting with anti-HA or anti-FLAG antibodies, all forms were detected as multiple bands, and most CXCR3A and CXCR3B were detected at the interface between the stacking gel and the running gel, suggesting that these protein aggregates cannot penetrate the running gel. These results were clearly observed in boiled samples and were not resolved by the addition of 6 M urea, which denatures proteins. The band intensities indicated that the CXCR3A abundance was higher than that of CXCR3B. CXCR3A and CXCR3Alt was detected as multiple bands (Fig. 1C). These results indicate that the long N-terminal region of CXCR3B affects the expression efficiency. Western blotting examined the CXCR3 variant expression levels, but not the functional expression of each variant. We investigated the membrane expression of the proteins by conducting HiBiT assays in cells expressing the receptors tagged with HiBiT at the N-terminal. The extracellular luciferase activities of HiBiT-receptors and LgBiT in the assay reagents increased with the amount of plasmid. The luciferase activities in HiBiT-CXCR3A-expressing cells were approximately 6-fold higher than those in HiBiT-CXCR3B-expressing cells, suggesting that the N-terminal of CXCR3B negatively affected plasma membrane localization of the receptor. HiBiT-CXCR3Alt was not efficiently expressed at the cell surface, although it had high expression levels, since it does not have functional structure of GPCR (Fig. 1D). To further examine the cellular expression patterns of the variants, C-terminal enhanced green fluorescent protein (EGFP)-tagged CXCR3 variants were expressed in HEK293 cells. Since most EGFP signals were detected in cytosol, it was hard to distinguish membrane expression of the receptors. However, I-TAC induced clustering of EGFP signals in CXCR3A-expressing cells, indicating that CXCR3A were activated and internalized by the ligand. The fluorescence signals were not changed by the ligand in cells expressing the other variants. CXCR3B-EGFP signals in cellular imaging and western blotting with anti-EGFP antibodies were relatively weak compared to the other variants, which is consistent to the data in Fig. 1C (Supplementary Fig. S1). These results suggest that different N-terminal residues may affect the expression and localization of the two intact 7-TM variants (CXCR3A and CXCR3B), although it is not excluded the possibility that signaling efficiency as well as membrane localization of CXCR3B were influenced by expression level.

CXCR3 variants differentially mediate extracellular signal-regulated kinase (ERK) phosphorylation

ERK is one of the main cellular effectors that are activated by GPCRs. Chemokine-induced ERK phosphorylation was comparatively analyzed in HEK293 cells expressing exogenous CXCR3 variants. None of the chemokines induced ERK phosphorylation in HEK293 cells indicating that endogenous CXCR3 proteins are scarcely expressed regardless of mRNA expression of all three variants. I-TAC, IP-10, and MIG induced different levels of CXCR3A-mediated ERK phosphorylation, whereas PF-4 did not. All four chemokines bind CXCR3, but the resulting phosphorylation pattern may delineate differing affinities between the CXCR3 receptor and each of the chemokines. PF-4 does not appear to activate CXCR3, although it apparently binds the receptor. I-TAC and IP-10 induced CXCR3B-mediated ERK phosphorylation, whereas MIG and PF-4 did not. CXCR3Alt did not mediate ERK phosphorylation, suggesting that it is a nonfunctional signal mediator (Fig. 2A). ERK phosphorylation induced by I-TAC and IP-10 was observed until 15 min after stimulation and disappeared by 30 min after stimulation in cells expressing CXCR3A. Relatively weak MIG-induced ERK phosphorylation was detected only at 5 min and 10 min after stimulation (Fig. 2B). ERK phosphorylation mediated by CXCR3B after I-TAC stimulation was relatively weak. The other chemokines induced negligible ERK phosphorylation mediated by CXCR3B (Fig. 2C). These combined results indicate that CXCR3A is most efficient receptor mediating ERK phosphorylation. Further, CXCR3A is differentially activated by the chemokines with the following efficacies: I-TAC > IP-10 >> MIG.

Analysis of ligand-dependent interactions of CXCR3 variants and Gα subunits

Chemokine receptors are postulated to elicit cellular responses via activation of Gαi/o and/or Gα12/13 or Gαq. Therefore, we investigated these molecular interactions using NanoBiT technology with mini-G proteins, which are engineered Gα proteins that are useful for biophysical analysis of GPCR activation (Nehmé et al., 2017; Wan et al., 2018). Mini-G proteins were developed as NanoBiT constructs. Analyses with different combinations of LgBiT or SmBiT-tagged proteins suggested that N-terminal–tagged LgBiT constructs of each mini-G protein were likely to efficiently bind CXCR3-SmBiT. CXCR3A increased the luminescence signal by binding to Gsi43 (engineered form of Gαi/o) in response to I-TAC. CXCR3A also increased the luminescence signal by binding to Gsq70, an engineered form of Gαq (Fig. 3A). No interaction of CXCR3A with Gαs was observed in this study. Previous studies reported that chemokine receptors induce some signals by interacting with Gα12/13 and Gα16 (Rosenkilde et al., 2004; Yagi et al., 2011). These interactions were not observed using NanoBiT technology, possibly due to the lack of well-engineered mini-G proteins of Gα12 and Gα16 (Nehmé et al., 2017; Wan et al., 2018).

Next, the interaction of CXCR3A with Gα proteins was examined in the presence of different chemokines using NanoBiT constructs. IP-10 and I-TAC induced the interaction of CXCR3A with Gsq70, whereas MIG and PF-4 did not (Fig. 3B). The interaction of CXCR3A with Gsi43 was observed in the presence of I-TAC and IP-10 with different efficacies (I-TAC >> IP-10) (Fig. 3C). By contrast, the luminescence from the interaction of CXCR3B and any engineered mini-G protein was not detected (Fig. 3D), suggesting that membrane expression of CXCR3B-SmBiT may be not sufficient to induce detectable luminescence signals with mini-G protein after chemokine stimulation.

Effects of CXCR3 variants on chemokine-stimulated Ca²⁺ influx

Chemokines bind chemokine receptors, which interact with Gαi/o family proteins, thereby activating PLC-β via Gβγ released from the Gαi/o, leading to the production of inositol 1,4,5-trisphosphate (IP3) and diacylglycerol and subsequent intracellular calcium increase (Neptune and Bourne, 1997). CXCR3A appears to interact with Gαq in a ligand-dependent manner (Figs. 3A and 3B). This suggests that Gαi/o and Gαq may be involved in CXCR3A-mediated intracellular Ca²⁺ flux. We adapted NanoBiT technology to detect intracellular calcium signals by tagging calmodulin (CM) with SmBiT of Nluc at the C-terminal and tagging the calmodulin-binding motif of myosin light chain kinase 2 (MYLK2S) with LgBiT of Nluc at the N-terminal (Nguyen et al., 2020b). In this system, ligand-receptor interaction increases cytosolic Ca²⁺ leading to calmodulin-Ca²⁺ complex binding to MYLK2S, which can induce the interaction of LgBiT and SmBiT to generate luciferase activity. HEK293 cells expressing CXCR3A were treated with each ligand, and then real-time luminescence signals were measured. The results showed that Ca²⁺ responses were observed in cells treated with I-TAC and IP-10 (Fig. 4A). I-TAC–stimulated Ca²⁺ influx was completely abolished by pretreatment with 100 ng/ml pertussis toxin (Fig. 4B), suggesting that the Ca²⁺ signals were induced by Gβγ released from Gαi/o, but not by Gαq. The observed Ca²⁺ signals were not sufficiently strong to differentiate the responses induced by different ligands. Thus, HEK293-Gαqi cells were used to investigate the Ca²⁺ response, because the chimera could feasibly detect chemokine receptor activation by converting the Gαi/o signal to the Gαq signal (Wang et al., 2009). CXCR3A increased the Ca²⁺ response to I-TAC and IP-10 with differing efficiencies, whereas CXCR3B only responded to I-TAC. Neither I-TAC nor IP-10 induced the Ca²⁺ response in cells expressing CXCR3Alt (Fig. 4C).

The CXCR3A dose-response to I-TAC was much higher than that of CXCR3B (Fig. 4D, Supplementary Fig. S2). IP-10 and MIG also increased Ca²⁺ responses in a dose-dependent manner in cells expressing CXCR3A, but not in those expressing CXCR3B (Figs. 4C and 4D, Supplementary Figs. S3 and S4). Regarding to CXCR3A, three chemokines displayed the following order of potency for CXCR3A: I-TAC (EC50 = 3.82 nM) > IP-10 (EC50 = 20.2 nM) > MIG (EC50 = 56.4 nM) (Fig. 4D). This suggests that I-TAC acts as a full agonist for CXCR3A, whereas IP-10 and MIG may act as partial agonists. Significantly lower potency (EC50 = 16.6 nM) and efficacy were observed in response to I-TAC in cells expressing CXCR3B (Fig. 4D). No Ca²⁺ response to PF-4 was observed in cells expressing any CXCR3 variant (Figs. 4A and 4C, Supplementary Fig. S5). Expression of CXCR3Alt did not induce Ca²⁺ response to any tested chemokine at even high concentrations (Fig. 3B, data not shown).

Cytosolic Ca²⁺ increase leads to protein kinase C-dependent ERK activation, which results in SRE-dependent gene expression (Cheng et al., 2010). To confirm the downstream signaling of Gαqi activation, the SRE-Luc reporter gene assay was conducted in HEK293-Gαqi cells cotransfected with SRE-Luc and each of the CXCR3 variants. I-TAC stimulation significantly increased luciferase activity through CXCR3A, whereas IP-10 induced only slightly increased luciferase activity (Fig. 4E). Neither MIG nor PF-4 stimulated CXCR3A-mediated luciferase activity, although MIG induced a weak Ca²⁺ signal. There was no increase in SRE-Luc activity in response to any tested chemokine in cells expressing CXCR3B or CXCR3Alt (Fig. 4E). This may be due to the short duration of the Ca²⁺ transient induced by the chemokine-receptor interaction, or due to the sensitivity of the reporter gene assay.

Chemokines differentially stimulate β-arrestin recruitment to CXCR3 variants

GPCR activation induces Gβγ release from active Gα, leading to G protein-coupled receptor kinase (GRK) recruitment to the GPCR; subsequent receptor phosphorylation enables interaction with β-arrestin, which downregulates the signaling event or the induction of β-arrestin–mediated signaling (Jean-Charles et al., 2017). To investigate chemokine-mediated β-arrestin recruitment to CXCR3 variants, cells coexpressing NanoBiT constructs of β-arrestin1 and each CXCR3 variant construct were treated with the I-TAC, IP-10, MIG, and PF-4 chemokines. The results showed that I-TAC and IP-10 increased cellular luminescence through β-arrestin1 interactions with CXCR3A, but MIG and PF-4 did not (Supplementary Fig. S6, upper graphs). CXCR3B generated weak luminescence signals only in the presence of I-TAC (Supplementary Fig. S6, middle graphs). By contrast, no luminescence signal was detected after chemokine treatment of cells expressing CXCR3Alt (Supplementary Fig. S6, lower graphs). β-arrestin2 recruitment to CXCR3 was similar to that of β-arrestin1, except that the luminescence signals resulting from CXCR3A and β-arrestin2 interaction were weaker than those observed with β-arrestin1, even with MIG treatment (Supplementary Fig. S7). These combined results indicate that chemokine-dependent β-arrestin recruitment to CXCR3A or CXCR3B displays similar patterns as those of ERK phosphorylation and Ca²⁺ increase.

Relevance of different chemokines in stimulating CXCR3-mediated cellular responses

Four chemokines have been reported to directly stimulate CXCR3 activation, which was confirmed by radioisotope-labeled ligands (Clark-Lewis et al., 2003; Heise et al., 2005; Weng et al., 1998). To determine whether the chemokines affect each other regarding receptor binding and cellular responses, HEK293 cells expressing the Ca²⁺ response probes, Gαqi, and CXCR3A were treated with the chemokines. I-TAC–stimulated luminescence signals were abrogated by pretreatment with I-TAC, but not with IP-10, MIG, and PF-4 (Fig. 5A). I-TAC or IP-10 pretreatment led to elimination of IP-10–stimulated luminescence signals. MIG pretreatment slightly downregulated IP-10–stimulated luminescence signals, but PF-4 pretreatment had no effect on luminescence signals (Fig. 5B). These sequential analyses of Ca²⁺ responses indicated that I-TAC was the strongest ligand compared to the other three ligands. These combined results suggest that the chemokines display different binding affinities to CXCR3A.

Complex formation of CXCR3 variants

Many GPCRs are expressed as homodimers or heterodimers on the cell surface. Their constitutive or inducible dimerization is a pivotal mechanism for agonist-induced activation and versatile downstream signaling (Springael et al., 2005). A previous study demonstrated that CXCR3A was expressed as a homodimer (Nguyen et al., 2020a). To investigate complex formation of the CXCR3 variants, different combinations of variants tagged with SmBiT or LgBiT at the C-terminal were coexpressed in HEK293 cells. The resulting luminescence signals indicated that all CXCR3 variants appear to homodimerize or heterodimerize with each other. Relatively weak luminescence signals by CXCR3B NanoBiT constructs may be due to its low expression (Figs. 1C and 6A). Although CXCR3Alt was not efficiently expressed on the cell surface (Fig. 1D), it generated strong luminescence signals by dimerizing with itself and other variants, suggesting that CXCR3 variant dimerization occurs primarily in the cytosol. These CXCR3 variant interactions were confirmed by subjecting cells expressing epitope-tagged variants to immunoprecipitation analysis. CXCR3A-HA was detected in all precipitates of anti-FLAG agaroses for the FLAG-tagged variants (Fig. 6B). The cells were lysed with lysis buffer containing 1% Triton X-100, and the Triton X-100–insoluble fraction was removed by centrifugation; therefore, the bands observed in cell lysates were not as strong as those presented in Fig. 1D.

Complex formation of the CXCR3 variants was further confirmed by examining the effect of each variant on ligand-stimulated β-arrestin recruitment to CXCR3A. Cells expressing both CXCR3A-LgBiT and SmBiT-β-arrestin1 under control of the UbiC promoter with different promoter-driven intact CXCR3 variants were treated with I-TAC (Fig. 6C). In the presence of intact variants under the weakest HSV-TK promoter, I-TAC induced higher luminescence compared to that without intact variants. As the variants easily form functionally active complexes on the cell surface, this suggests that most HSV-TK-driven intact receptors may bind CXCR3A-LgBiT, which increases the absolute number of dimerized receptors (CXCR3A-LgBiT/each intact variant or CXCR3A-LgBiT/CXCR3A-LgBiT) that can interact with SmBiT-β-arrestin1. By contrast, overexpression of intact variants with the CMV promoter reduced the luminescence signals, since the overexpressed variants were dominant over CXCR3A-LgBiT on I-TAC stimulation and blocked the ligand binding of CXCR3A-LgBiT. CXCR3B expression level was comparatively lower than that of CXCR3A; therefore, HSV-TK and UbiC-driven intact CXCR3B enhanced the luminescence signals generated by CXCR3A-LgBiT and SmBiT-β-arrestin1 (Fig. 6C, middle graph). Although CXCR3Alt was not functional as a chemokine receptor and scarcely expressed on the cell surface, low expression under weak promoters enhanced the ligand-dependent interaction of CXCR3A-LgBiT and SmBiT-β-arrestin1 (Fig. 6C, right graph). CXCR3Alt expression may increase the absolute number of dimers (e.g., CXCR3A-LgBiT homodimers or CXCR3A-LgBiT/CXCR3Alt heterodimers) to enhance SmBiT-β-arrestin1 recruitment. Alternatively, CXCR3Alt may affect the expression of CXCR3A at the cell surface. To test this, the HiBiT assay was conducted in cells coexpressing UbiC-driven HiBiT-CXCR3A or HiBiT-CXCR3B with CXCR3Alt driven by different promoters. CXCR3Alt overexpression under the CMV promoter decreased the luminescence signals, whereas HSV-TK–driven weak expression enhanced the luminescence signals, implying that weak expression of CXCR3Alt enhances the cell surface expression of functional CXCR3 variants (Fig. 6D, left and middle graphs). CXCR3B also affected the cell surface expression of HiBiT-CXCR3A (Fig. 6D, right graph). However, the enhanced membrane expression of CXCR3A induced by HSV-TK–driven variants may not be biologically relevant because CXCR3A-mediated responses were downregulated by other variants (see below).

CXCR3Alt and CXCR3B negatively regulate CXCR3A-mediated calcium signaling

CXCR3 variant dimers may mutually affect the cell surface expression, ligand recognition, and signal transduction of other dimers. To examine the effect of CXCR3Alt on Ca²⁺ responses mediated by the intact CXCR3 variant constructs, cells were cotransfected with different promoter-driven CXCR3Alt constructs and UbiC-driven CXCR3A or CXCR3B constructs along with NanoBiT constructs of Ca²⁺ indicators. I-TAC and IP-10 stimulation of cells expressing CXCR3A resulted in the downregulation of luminescence signals depending on CXCR3Alt expression levels (CMV>UbiC>>HSV-TK) (Figs. 7A and 7B), suggesting that CXCR3Alt may affect ligand-stimulated activation of CXCR3A. I-TAC–stimulated luminescence in cells expressing CXCR3B was relatively weak and completely abolished by CXCR3Alt (Fig. 7C), indicating that CXCR3Alt affected the cell surface expression and functional activity of CXCR3B. To determine the effect of CXCR3B on CXCR3A-mediated Ca²⁺ signaling, cells coexpressing UbiC-driven CXCR3A and different promoter-driven CXCR3B were treated with I-TAC or IP-10. The results showed that chemokine-stimulated luminescence signals were weakened by CMV-driven and UbiC-driven CXCR3B, suggesting that CXCR3B negatively regulates CXCR3A signaling, although it mediates the cellular response to I-TAC (Supplementary Fig. S8).

CXCR3A-mediated cell migration was negatively regulated by CXCR3B and CXCR3Alt

CXCR3 has a pivotal role in mediating chemotactic motility toward chemokines (Cole et al., 1998; Loetscher et al., 1996). The RT-PCR indicated that many cells may express CXCR3 (Fig. 1B). However, adherent cancer cells did not migrate toward I-TAC even after exogenous expression of CXCR3A (data not shown), although CXCR3-dependent migration of the cancer cells were reported (Shin et al., 2010; Yang et al., 2016). Since CXCR3 is responsible for T cell migration (Mueller et al., 2008); therefore, chemotactic migration was investigated in Jurkat cells. A small number of cells migrated toward I-TAC and IP-10, indicating that CXCR3-mediated chemotactic machinery was functional in Jurkat cells. To determine the effects of CXCR3 variants, Jurkat cells infected with lentivirus harboring each CXCR3 variant were tested in migration assays for the four chemokines. Exogenous CXCR3A expression enhanced basal motility, and migrating cells toward the chemokines significantly increased with different efficiencies (I-TAC>>IP-10>MIG). However, chemotactic activity was not observed in cells expressing CXCR3B and CXCR3Alt (Fig. 8A).

Differences in migration activities of CXCR3 variants may be related to their intracellular signaling efficiencies. We investigated this hypothesis by examining I-TAC–stimulated ERK phosphorylation in Jurkat cells with and without exogenous CXCR3 variants. I-TAC–stimulated ERK phosphorylation in parental cells was enhanced by the expression of exogenous CXCR3A but substantially reduced by CXCR3B and CXCR3Alt (Fig. 8B). To examine if CXCR3A-mediated migration was affected by other CXCR3 variants, cells expressing CXCR3A were infected with viruses harboring other variants. The CXCR3A-mediated migration was significantly reduced by coexpression of CXCR3B or CXCR3Alt, suggesting that these two variants negatively regulated CXCR3A-mediated migration activity (Fig. 8C). Taken together, the functional properties of CXCR3 splicing variants and interrelationship of them are summarized in Fig. 8D.

Alternative splicing pathways generate protein variants with different functional properties that may diversify cellular responses to various stimuli using a finite number of genes. Some chemokine receptor genes produce splicing variants containing different amino acid sequences and lengths in the N-terminal region, which is the first binding target of the cognate chemokines. Thus, these variants may have different ligand binding specificities and/or affinities (Szpakowska et al., 2012). Chemokine receptors are localized in the plasma membrane without a signal peptide, which represents a unique sequence among membrane proteins and secretory proteins. Nevertheless, it cannot be excluded that the unique N-terminal sequence may affect the expression and membrane localization of the receptor variants. Western blots of total cell lysates and HiBiT assays indicate that CXCR3B has relatively low expression and membrane localization compared to CXCR3A. The N-terminal sequence of CXCR3B is longer than that of CXCR3A, which may affect receptor expression and functional interaction with cognate ligands. CXCR3A-mediated ERK phosphorylation, Ca²⁺ influx, and β-arrestin recruitment responded to I-TAC, IP-10, and MIG with different efficiencies, but did not respond to PF-4. By contrast, CXCR3B-mediated activities responded to only I-TAC with low efficiency. These results indicate that CXCR3A is a fully functional chemokine receptor through the N-terminal residues.

Alternative splicing in regions other than the N-terminus appear in two chemokine receptors: CCR2 and CXCR3. CCR2 variants display different tissue expression patterns and downstream signals, but function as chemokine receptors with an intact 7-TM domain (Harmon et al., 2010). CXCR3Alt has 4-TM domain and may not function similarly as other CXCR3 variants. CXCR3Alt had high expression, but little CXCR3Alt protein was detected on the cell surface, suggesting that CXCR3Alt is not efficiently translocated to the plasma membrane but may remain in intracellular organelles. Therefore, CXCR3Alt did not by itself affect chemokine-stimulated cellular responses.

GPCR homodimerization or heterodimerization affects receptor maturation, folding, and trafficking to or from the cell surface. Receptor complex formation proceeds in a constitutive or ligand-dependent manner (Harding et al., 2009; Terrillon and Bouvier, 2004). CXCR3 also appears to be constitutively expressed as a homodimer, similarly as CXCR4 and CXCR7 (Nguyen et al., 2020a), which was confirmed by NanoBiT technology and co-immunoprecipitation. Cell surface expression and ligand-dependent β-arrestin recruitment of each variant were influenced by different expression levels of the other variants, which consolidates dimerization of the variants. Interestingly, membrane expression of CXCR3A was enhanced by weakly expressed other variants with HSV-TK. However, cellular responses were not changed by slight enhancement of membrane CXCR3A. By contrast, CXCR3A-mediated Ca²⁺ signals were significantly decreased in the presence of CXCR3Alt when under the control of the same UbiC promoter, although the plasma membrane expression patterns were not affected. This was obvious in I-TAC–stimulated CXCR3B-mediated cellular responses, which were reduced in the presence of CXCR3Alt. The results from exogenous expression studies using different assay systems in heterogenous cells are sometimes inconsistent, which makes it difficult to precisely determine their physiological relevance. However, the results of the present study are consistent with the hypothesis that CXCR3Alt and CXCR3B may act as negative regulators of CXCR3A.

CXCR3 is responsible for T cell migration and maturation. Thus, receptor-mediated cellular responses were investigated in Jurkat cells derived from T lymphocytes. I-TAC, IP-10, and MIG stimulated the migration of wild-type Jurkat cells and cells expressing exogenous CXCR3A, whereas chemokine-dependent migration was blocked in cells expressing exogenous CXCR3B and CXCR3Alt. Functional differences among the CXCR3 variants were determined by investigating I-TAC-stimulated ERK phosphorylation, which was detected in wild-type cells and further enhanced in the presence of exogenous CXCR3A. By contrast, ERK phosphorylation declined in the presence of CXCR3B and CXCR3Alt. CXCR3B and CXCR3Alt lowered cell migration toward the chemokines compared to CXCR3A alone. This result is consistent with the hypothesis that the molecular activities of CXCR3A were downregulated by complex formation with other variants. The negative role of CXCR3B in opposition to CXCR3A in cancer progression may be ascribed to low ligand-binding affinity and dimerization with CXCR3A (Lasagni et al., 2003; Li et al., 2019; Reynders et al., 2019). The present study suggested that CXCR3Alt may be a potent negative regulator of CXCR3A.

Chemokine receptors induce Ca²⁺ influx by activating Gαi/o- Gβγ or Gα16 in a limited cell population such as immune cells (Kuang et al., 1996; Shi et al., 2007). Our NanoBiT-based Ca²⁺ assay system revealed that I-TAC-stimulated intracellular Ca²⁺ flux was mediated by CXCR3A. The NanoBiT assay using receptor and mini-G protein constructs showed that CXCR3A directly binds Gαq and Gαi/o. As Gα16 is not expressed in HEK293 cells, the observed Ca²⁺ flux may be induced by Gαq and Gβγ released from Gαi/o. CXCR3A-dependent signaling did not dramatically increase Ca²⁺ flux compared to other GPCR activation of Gαq, suggesting that the interaction between CXCR3A and Gαq was relatively weak, although the NanoBiT assay revealed a relatively strong interaction. Pertussis toxin completely inhibited the I-TAC-stimulated Ca²⁺ increase, indicating that Gαq is dispensable for CXCR3A-mediated Ca²⁺ signaling. Further work is needed to define the functional significance of CXCR3A interaction with Gαq, and determine whether it is nonspecific or necessary for receptor regulation or cellular responses.

Four chemokines have been identified as CXCR3 ligands, which may provide functional redundancy in receptor-mediated cellular responses to chemokines (Cole et al., 1998; Loetscher et al., 1996; Mueller et al., 2008). The chemokines bind to the cognate receptors with different affinities (Rajagopalan and Rajarathnam, 2006). Our results confirmed that CXCR3A mediated ERK phosphorylation, Ca²⁺ mobilization, and β-arrestin recruitment with differing efficiencies in response to the I-TAC, IP-10, and MIG chemokines. By contrast, PF-4 did not stimulate any cellular responses in our assay system, although it has been reported as a strong chemoattractant through CXCR3 for T lymphocytes, neutrophils, fibroblasts, and monocytes (Eisman et al., 1990; Mueller et al., 2008). In the present study, Jurkat cells expressing endogenous CXCR3 responded to all tested chemokines but not PF-4. PF-4 has been suggested to have physiological roles in inflammation and wound repair (Bodnar, 2015; Lord et al., 2017), although this cannot be explained by the interaction with CXCR3. As no other chemokine receptors bind PF-4, its physiological roles might be mediated by unknown receptors or specific glycosaminoglycans. Sequential chemokine treatment in CXCR3A-mediated Ca²⁺ signaling verified that CXCR3A had different affinities for the chemokines, and indicated that I-TAC was the strongest ligand. This suggests that different chemokine affinities may generate different cellular responses depending on the ligand binding.

The NanoBiT complementation assay is a powerful technology for mapping the signaling pathways induced by ligand stimulation. We produced specific NanoBiT constructs to investigate the cellular and biochemical reactions triggered by CXCR3 variants and their cognate chemokines. Controlled expression of the NanoBiT constructs of various downstream signaling components as well as the receptors and real-time monitoring of live-cell responses provided powerful tools to characterize the molecular properties of the CXCR3 variants and chemokines under essentially physiological conditions. These integrated analyses combined with other biochemical methods demonstrated that CXCR3A is an authentic chemokine receptor, and its functional activities can be modulated by the CXCR3B and CXCR3Alt variants through the formation of heterodimer complexes. The activation of specific splicing machineries under specific cellular environments may regulate the generation of these variants to modulate CXCR3A-mediated cellular responses. For example, the negative effect of CXCR3B on cancer progression may be attributed to the predominance of CXCR3B over CXCR3A via alternative splicing processes or inactive heterodimer formation. PF-4 did not elicit any observed cellular responses, although there is a report of molecular interaction between CXCR3 and PF-4. Therefore, we recommend that it should be removed from the list of CXCR3 ligands.

This work was supported by a National Research Foundation of Korea (NRF) Grants (2022R1F1A1074216, 2020M3E5D9080165) funded by the Korea government (MSIT).

H.T.N., S.H., and J.-I.H. designed and performed experiments and wrote the manuscript. L.P.N., T.U.N., and H.-K.P. performed experiments and analyzed the data. J.Y.S., C.S.L., B.-J.H., and J.-I.H. supervised this study.

The authors have no potential conflicts of interest to disclose.