Traffic control: Mechanisms of ligand-specific internalization and intracellular distribution of CCR5

Abstract

CC chemokine receptor (CCR) 5 promotes inflammatory responses by driving cell migration and scavenging chemokine. A CCR5 inhibitor Maraviroc has been approved for blocking HIV entry; however, inhibitors for the treatment of other diseases have had limited success, likely because of the complexity of CCR5 pharmacology and biology. CCR5 is activated by natural and engineered chemokines that elicit distinct signaling and trafficking responses, including receptor sequestration inside the cell. Intracellular sequestration may be therapeutically exploitable as a strategy for receptor inhibition, but the mechanisms by which different ligands promote receptor intracellular retention versus presence on the cell membrane are poorly understood. In this study, we systematically compared the time-dependent trafficking behavior of CCR5 following stimulation with its endogenous agonist, CCL5, and 2 CCL5 variants that promote CCR5 intracellular retention. Using a broad panel of pharmacologic assays, fluorescence microscopy, and live cell ascorbic acid peroxidase proximity labeling proteomics, we identified distinct ligand-dependent CCR5 trafficking patterns with temporal and spatial resolution. All 3 chemokines internalize CCR5 via β-arrestin–dependent, clathrin-mediated endocytosis but to different extents, with different kinetics and varying dependencies on G protein-coupled receptor kinase subtypes. The agonists differ in their ability to target the receptor to lysosomes for degradation, as well as to the Golgi compartment and the trans-Golgi network, and these trafficking patterns translate into distinct levels of ligand scavenging. The results provide insight into the cellular mechanisms behind CCR5 intracellular sequestration and suggest how trafficking can be exploited for the development of functional antagonists of CCR5.

Significance Statement

CC chemokine receptor (CCR) 5 plays a crucial role in the immune system and is important in numerous physiological and pathological processes such as inflammation, cancer, and transmission of HIV. It responds to different ligands with distinct signaling and trafficking behaviors; notably, some ligands induce retention of the receptor inside the cell. This study reveals the cellular basis for receptor sequestration that can be exploited as a therapeutic strategy for inhibiting CCR5 function.

1. Introduction

CC chemokine receptor (CCR) 5 is a G protein-coupled receptor (GPCR) that mediates migration and activation of immune cells in response to chemokines CCL3, CCL4, and CCL5 (Flanagan, 2014). Because of its role in facilitating entry of the human immunodeficiency virus (HIV) into leukocytes during viral transmission, CCR5 was intensively studied as a therapeutic target for the prevention and treatment of HIV (Murphy, 2001; Brelot and Chakrabarti, 2018), resulting in FDA approval of the small molecule CCR5 antagonist Selzentry (Maraviroc) (Sax, 2007). It has also been investigated because of its role in inflammation associated with Alzheimer disease (Joy et al, 2019), nonalcoholic steatohepatitis (Lefere et al, 2020), multiple sclerosis (Boven et al, 2000), atherosclerosis (Zernecke et al, 2006), inflammatory bowel disease (Ajuebor et al, 2001), and multiple cancers (de Oliveira et al, 2014; Vangelista and Vento, 2018). Like other chemokine receptors, agonist-mediated activation of CCR5 results in coupling to the cAMP-inhibitory (G_i) class and phospholipase activating (G_q) class of heterotrimeric G proteins (Zhao et al, 1998). Stimulation of PI3K (Aramori et al, 1997), MAPK (Christmann et al, 2011), and JAK/STAT (Wong et al, 2001) signaling pathways has also been reported. Some ligands promote phosphorylation of the receptor through G protein-coupled receptor kinases (GRKs) with subsequent engagement of β-arrestins, which regulate signaling through receptor desensitization and internalization (Aramori et al, 1997; Mack et al, 1998; Oppermann, 2004; Hüttenrauch et al, 2005). However, experiments with N-terminally modified CCL5 variants demonstrated that the exact pharmacological responses of CCR5, including its postactivation intracellular trafficking behavior, can markedly differ depending on the ligand. For example, while WT CCL5 promotes efficient receptor recycling back to the cell surface following initial internalization (Mack et al, 1998), the superagonist CCL5 variant AOP-CCL5 induces more rapid and efficient receptor internalization and retention in endosomes, making it a particularly potent inhibitor of HIV entry into cells (Simmons et al, 1997).

Insights into the mechanisms of such ligand-dependent CCR5 signaling and trafficking have been reported (Mack et al, 1998; Gaertner et al, 2008; Lorenzen et al, 2018; Martins et al, 2020), but the molecular picture is far from complete. Because of the therapeutic potential for “functional antagonism” (Simmons et al, 1997; LaMontagne et al, 2006) that hinges on CCR5 internalization and sequestration, we were particularly interested in understanding the mechanisms by which some ligands promote intracellular retention of the receptor. We chose to investigate 3 chemokine variants that differ only in the first few N-terminal amino acid residues yet induce distinct signaling and trafficking patterns of CCR5: WT CCL5 (CCL5); the superagonist [6P4]CCL5 (6P4); and a partial and possibly G_i-biased agonist [5P14]CCL5 (5P14) (Gaertner et al, 2008). To obtain insight into the ability of these ligands to regulate the kinetics and trafficking of CCR5 into different intracellular compartments, a variety of cell-based assays were used: bioluminescence resonance energy transfer (BRET) of protein interactions and localization, flow cytometry–based receptor internalization and recycling, fluorescence microscopy visualization of receptor subcellular localization, and receptor degradation. As receptor trafficking is determined by distinct ligand-dependent interactions with intracellular proteins, we also used ascorbate peroxidase (APEX2)–catalyzed proximity biotinylation coupled with mass spectrometry (MS) to reveal directly and indirectly interacting proteins with a high degree of spatial (<20.0 nm) and temporal resolution. Based on patterns of interactors, the APEX2 experiments revealed the time-dependent location of CCR5 in the cell (CCR5 neighborhoods) and both guided and complemented the pharmacologic assays in understanding mechanisms responsible for its ligand-dependent sequestration. In addition to promoting cell migration, CCR5 also scavenges chemokines (Turner et al, 2012), which is hypothesized to be critical for maintaining appropriate extracellular chemokine levels, thereby preventing receptor downregulation, as well as resolution of inflammatory processes (Cardona et al, 2008; Gilliland et al, 2013; Shroka et al, 2023). Because scavenging depends on internalization of the receptor and transport of chemokine for degradation in lysosomes, followed by recycling of the receptor back to the cell surface to repeat the process (Gilliland et al, 2013; Shroka et al, 2023), we investigated the ability of CCR5 to scavenge the 3 ligands and how scavenging efficiency correlated with receptor trafficking profiles.

Overall, the results demonstrate how different ligands can markedly alter the endocytic fate and function of CCR5 through internalization, recycling, degradation, and subcellular compartment trapping. The identification of such mechanisms may prove useful for therapies based on exploiting receptor sequestration. These mechanisms may also have broad implications for the regulation of GPCRs by their natural ligands.

2. Materials and methods

2.1. Chemokine production and use in assays

WT CCL5, 5P14, and 6P4 were produced as described in the Supplemental Methods. For most experiments, 100 nM chemokine was used as the saturating concentration based on previous studies (Martins et al, 2020).

2.2. DNA constructs

The CCR5 in pcDNA3.1 Hygro(+) was created by PCR amplification of the coding region of N-terminal Flag-tagged CCR5 with subsequent subcloning of the product into pcDNA3.1 Hygro(+) under the human cytomegalovirus promoter with the following primers: 5'-CCGGTACCGACTATCAGGTC-3' and 5'-AAGCGGCCGCCTTAAGTCCCACACTGATTTCCTG-3'.

The APEX2-catalyzed proximity biotinylation proteomics experiments require fusion of ascorbate peroxidase to the C-terminus of CCR5, which enables biotinylation of directly and indirectly interacting proteins within ∼20 nm (Lobingier et al, 2017; Paek et al, 2017; Polacco et al, 2022). To generate the fusion construct, the APEX2 sequence was PCR amplified from pcDNA3-CYTO-APEX2 vector (kindly gifted by Dr Mark von Zastrow, University of California, San Francisco) (Lobingier et al, 2017) and subcloned into a Flag-CCR5-pcDNA3.1 Hygro+ vector at the C-terminus of CCR5, with a linker (amino acid sequence: AAASGS) separating the 2 proteins. The following primers were used:

• •5′-ACAAGGACGATGATGACAAACC-3′,
• •5′- GGTACCGGGTTTGTCATCATC-3′,
• •5′- CGAGCAGGAAATCAGTGTGGGACTGTCGGGCTCGGGAAAGTC-3′,
• •5′- CAGTCCCACACTGATTTCCTG-3′,
• •5′- TAATGACTCGAGTCTAGAGGGCC-3′, and
• •5′- CGGGCCCTCTAGACTCGAGTCATTACGAGCCCGAGGCATC-3′.

The spatial references pcDNA3-PM-APEX2, pcDNA3-CYTO-APEX2, and pcDNA3-ENDO-APEX2 were also kindly gifted by Dr Mark von Zastrow (Lobingier et al, 2017). The pcDNA3.1(+) constructs of β-arrestin1-RlucII, β-arrestin2-RlucII, rGFP-CAAX, and rGFP-Rab4 were kindly gifted by Dr Michel Bouvier (Université de Montréal, Canada) (Namkung et al, 2016). The rGFP-Rab7 construct was generated by PCR amplification of the Rab7 coding sequence of EGFP-Rab7A, a gift from Qing Zhong (UC Berkeley, California), and inserted in-frame into the XbaI and PmeI sites of rGFP-Rab4 as previously reported (Shroka et al, 2023). The CCR5-RlucII construct was created by PCR amplification of the coding region of CCR5; products were subcloned in-frame at the N-terminus of the RlucII sequence into the pcDNA3.1 RlucII vector with the following primers: 5′-CCGGTACCGACTATCAGGTC-3′ and 5′-CGCGGATCCCCTGGTTCCAGTCCCACACTGATTTCCTG-3′.

To create the FLAG-SNAP-CCR5 construct, the coding sequence of a pcDNA3.1(+) CCR5 vector was PCR amplified and inserted into a pRK5–FLAG-ST-CXCR4 plasmid (Levoye et al, 2015) with the following primers: 5′- CGCACGCGTGACTATCAGGTCAGCTCCC-3′ and 5′- ACGGGCCCTCATTACAGTCCCACACTGATTTCC-3′, containing the mGlu5 receptor signal peptide that promotes proper receptor trafficking to the cell surface (kind gift of Dr Angélique Levoye, University of Paris, France).

2.3. Cell lines

All KO and parental (wild type) control HEK293 cells were a kind gift of Dr Asuka Inoue (Tohoku University, Japan). A dual β-arrestin1 and β-arrestin2 knockout (β-arrestin1/2 KO) was prepared by CRISPR/Cas9 targeting of ARRB1 and ARRB2 as described previously (O’Hayre et al, 2017). Similarly, CRISPR/Cas9 was used to generate GRK2/3 KO, GRK5/6 KO, and GRK2/3/5/6 KO in the HEK293A cell line (Pandey et al, 2021). Stable chemokine receptor–expressing cells (WT CCR5 and CCR5-APEX) and spatial reference-expressing cells were generated by transfecting HEK293T with expression vectors using the TransIT-LT1 transfection reagent (Mirus Bio, MIR2305). Stable transfectants were subsequently selected using 100 μg/mL hygromycin B (Life Technologies; 10687010) for chemokine receptor–expressing cells or 500 μg/mL G418 for spatial reference cells. Receptor surface expression was confirmed by flow cytometry (Guava EasyCyte 8HT; Luminex) with an anti–hCCR5-allophycocyanin–conjugated antibody (eBioscience; 501123098). Cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with GlutaMax (Gibco), 10% FBS, and selecting reagent, and grown at 37 °C with 5% CO₂.

2.4. APEX proximity labeling and streptavidin pull down

CCR5-APEX stable cells were cultured in 15-cm dishes in DMEM with GlutaMax (Gibco) and 10% FBS until confluent. The cells were then preincubated with 500 μM biotinyl tyramide (Sigma; SML2135) for 30 minutes at 37 °C and then treated with 100 nM corresponding CCL5 ligands for the indicated periods of time. Labeling was initiated with 1 mM H₂O₂ for 45 seconds and then quenched with ice-cold quenching buffer composed of 1 mM CaCl₂, 10 mM sodium ascorbate, 1 mM Trolox and 1 mM sodium azide in PBS. The cells were harvested and lysed in RIPA buffer containing 10 mM sodium ascorbate, 1 mM Trolox, 1 mM sodium azide, 1 mM DTT, protease inhibitor (Sigma; 5056489001) and Halt phosphatase inhibitor (Thermo Fisher Scientific; PI78420). After clarifying the lysates by centrifugation at 17,000g, at 4 °C for 15 minutes, the lysates were incubated with streptavidin magnetic beads (Thermo Fisher Scientific; 88817) with rotation at 4 °C overnight. The beads were then washed 3 times with buffer containing 4 M urea, 0.5% DDM (w/v), and 100 mM sodium phosphate, pH 8; followed by 3 washes with 4 M urea, 100 mM sodium phosphate, pH 8 buffer; and finally additionally washed with 50 mM HEPES, pH 8.5. The beads were resuspended in 50 mM HEPES pH 8.5 and stored at −80 °C before further processing.

2.5. Quantitative multiplex proteomics

Quantitative multiplex proteomics include sample preparation and peptide generation for biotinylated proteins, tandem mass tag (TMT) labeling of the subsequent biotinylated peptides, MS-based proteomic analysis of the labeled peptides, and finally bioinformatics analysis of the results. Please see Supplemental Methods for details.

2.6. BRET assay

HEK293 cells were seeded into 6-well plates (750,000 cells per well) and transfected the next day using Mirus TransIT-LT1 transfection reagent at ∼70% confluency. Cells were transfected with a BRET donor (32.5 ng of Receptor-RlucII or 19.8 ng of β-arrestin1-RlucII or β-arrestin2-RlucII per well) along with 130 ng of BRET acceptor per well (eg, rGFP-CAAX, rGFP-Rab4, rGFP-Rab11, or rGFP-Rab7). All assays were performed ∼30 hours after transfection, using previously described methods (Shroka et al, 2023) with minor modifications. Specifically, transfected cells were detached with 0.25 mL/well Accutase (Invitrogen; 00-4555-56) for 3 minutes at 25 °C, lifted from the 6-well plates by pipetting, and seeded into a 96-well white microplate (Tecan; 30122300) at 100,000 cells per well in prewarmed Tyrode buffer (140 mM NaCl, 2.7 mM KCl, 1 mM CaCl₂, 12 mM NaHCO₃, 5.6 mM D-glucose, 0.5 mM MgCl₂, 0.37 mM NaH₂PO₄, 25 mM HEPES, pH 7.4). A cell permeable RlucII substrate, Prolume Purple (NanoLight Technologies) was added to the wells at a final concentration of 5 μM, ∼3 to 6 minutes before BRET measurements. Three baseline BRET measurements were performed approximately 1 minute apart, followed by the addition of the indicated concentrations of chemokines. Subsequent BRET measurements were taken approximately 1 minute apart for 50 minutes or the indicated times.

For evaluation of receptor recycling, cells were transfected and seeded as described earlier. Following 30 minutes of chemokine stimulation and BRET measurements approximately 2 minutes apart, inhibitor Maraviroc (Sigma; PZ0002) was added to the cells at a final concentration of 5 μM, and subsequent BRET measurements were taken for an additional 40 minutes.

All BRET measurements were read using a VictorX Light luminescence reader (Perkin Elmer) or Spark microplate reader (Tecan). Values are presented as the fold change over mock-treated.

2.7. Flow cytometry–based internalization

Two different internalization assays were used to evaluate the absolute amount of internalized receptor (prelabel) and receptors that remain on the surface after internalization and recycling (postlabel). The difference in the surface level of the receptor between the prelabel and the postlabel assay indicates the extent of receptor recycling. In the prelabel assay, cell surface receptors are labeled with a primary antibody before allowing them to internalize. Then, after a defined period of incubation at 37 °C, receptors remaining on the cell surface are quantified with a secondary antibody. Neither the receptors that internalized nor those that reappear on the cell surface via recycling are detected by the secondary antibody. For the prelabel assay, HEK293 cells stably expressing CCR5 were labeled with 2 μg/mL rabbit anti-Flag tag antibody (F7425; Sigma) in FACS buffer (PBS with 0.5% bovine serum albumin [BSA]) for 30 minutes on ice and protected from light. Unbound antibody was removed by repeated washes with FACS buffer. Cells were then resuspended in Assay Buffer (DMEM, 0.5% BSA) containing 100 nM CCL5 ligands (CCL5, 6P4, or 5P14) or without ligands, and then either held at 4 °C (which prevents receptor trafficking and internalization) or transferred to 37 °C (which allows for normal receptor trafficking) and incubated for 1 hour. After incubation, cells were transferred to wet ice and the remaining surface receptor was labeled with 1:100 antirabbit antibody conjugated to PE (F0110; R&D Systems) in FACS buffer for 30 minutes on ice and protected from light. Following 3 washes with FACS buffer, CCR5 expression was assessed by flow cytometry using a Guava EasyCyte 8HT flow cytometer (Luminex). Data were analyzed with FlowJo software. The geometric mean fluorescent intensity of analyzed cells was used to quantify surface expression of CCR5 and compared with noninternalized control to determine relative percent of receptor remaining at the surface.

For the postlabel assay, all steps were similar to the prelabel assay except that the primary antibody staining was performed after the ligand incubation, thereby allowing receptors to internalize for a given period and then to recycle back to the cell surface. As noted earlier, this protocol ensures the detection of all cell surface receptors, including those that remain on the surface after internalization and those that emerge on the surface as a result of recycling.

2.8. Receptor degradation

HEK293 cells were seeded into 10-cm dishes and transfected at 70%–80% confluency with FLAG-CCR5 using Mirus TransIT-Lt1 transfection reagent. After 24 hours, the cells were re-plated into a 6-well plate, 3.5 × 10⁵/mL cells per well in 2.5 mL DMEM media. After 48 hours, cells were pretreated with 50 μg/mL cycloheximide for 15 minutes at 37 °C and incubated in the same media with or without 100 nM chemokine ligands for 4 hours at 37 °C. Following incubation, cells were placed on ice, rinsed with PBS, and lysed in RIPA lysis buffer (Thermo Fisher Scientific) containing cOmplete Mini EDTA-free Protease Inhibitor Cocktail (Roche). Cell lysates were collected and rotated end-over-end for 1 hour at 4 °C. Total protein concentrations were determined by Pierce Rapid Gold BCA assay (Thermo Fisher Scientific). Equivalent amounts of lysates were resolved by SDS-PAGE and transferred to a nitrocellulose membrane (Bio-Rad). Membranes were blocked with 5% nonfat dry milk (Bio-Rad) diluted in wash buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20) and subsequently washed in the same buffer. Blots were incubated overnight at 4 °C with 1:1000 anti-FLAG rabbit antibody (F4725; Millipore Sigma) and 1:1000 antiactin mouse antibody (A3853; Millipore Sigma) diluted in wash buffer containing 5% BSA. Membranes were washed and probed with 1:10,000 of the corresponding secondary IRDye 800CW antibody and 680RD antibody (LI-COR Biosciences) diluted in a wash buffer containing 5% BSA at room temperature for 1 hour. After final washes, membranes were imaged on a Chemidoc MP imaging system (Bio-Rad). Densitometry was performed using Chemidoc Image Lab software (Bio-Rad).

2.9. Confocal microscopy

HEK293 cells expressing Flag-SNAP-CCR5 were seeded onto 10.0-mm glass-bottom dishes (FluoroDish; FD3510, WPI) previously coated with 5 μg/mL fibronectin. The next day, cells were labeled with 5 μM cell-impermeable SNAP-Surface Alexa Fluor 488 (New England Biolabs) in complete media (DMEM + 10% FBS) on ice in the dark. After 1 hour, the cells were washed with ice-cold complete media and fixed with 4% paraformaldehyde after being incubated at 37 °C with 100 nM of different agonists for 1 hour. Control cells were fixed directly after labeling. Fixed cells were permeabilized using PBS containing 0.1% saponin for 30 minutes and blocked using PBS with 1% BSA and 1% goat serum for 1 hour before labeling with 1:100 dilution of rabbit polyclonal anti–TGN-46 antibody (Novus Biologicals; NBP1-49643). After primary antibody labeling, cells were washed using PBS containing 0.1% saponin and incubated using goat antirabbit IgG Alexa Fluor Plus 647 (1:1000 dilution) for 1 hour. Cells were washed and incubated briefly with Hoechst 33342 before imaging using an Eclipse Ti2-E (Nikon) equipped with a CSU-X1 (Yokogawa) spinning disk field scanning confocal system. Image analysis was performed using ImageJ (NIH) software, and colocalization was measured using the JaCoP plugin.

2.10. Scavenging

Non–receptor-expressing or stable CCR5-expressing HEK293 cells and corresponding β-arrestins knockout (KO) cells were seeded in triplicate in 96-well dishes at 50,000 cells/well in DMEM/10%FBS media and allowed to adhere for ∼8 hours. Subsequently, media was replaced with DMEM/10% FBS media containing 5 nM CCL5 ligands (CCL5, 6P4, or 5P14) and 10 μg/mL heparin and incubated for ∼16 hours. Remaining chemokine levels in the supernatants of cultured cells were measured in triplicate using the commercially available colorimetric LEGEND MAX human CCL5 ELISA kit (BioLegend; 440807) according to the manufacturer’s instructions and read with a SpectraMax M5 plate reader (Molecular Devices). Remaining levels of exogenous chemokine are reported as the percentage of levels in supernatants of the corresponding non–receptor-expressing cells.

2.11. Structure modeling

Three-dimensional models of complexes between C-terminally Ser-phosphorylated CCR5 with 5P14 and dimeric 14-3-3 protein zeta/delta (1433ζ) were built from amino acid sequences of CCR5 (aa 1–352), 5P14 (aa 1–69), and 2 copies of 1433ζ (aa 1–245) using AlphaFold3 (Abramson et al, 2024) through the AlphaFold Server (https://golgi.sandbox.google.com/). A single phosphoserine modification was applied on S336, S337, S342, or S349; 5 models were built per complex. Models were evaluated by the predicted per-residue local distance difference test (pLDDT) scores and inspected visually. The model with the highest pLDDT score is shown in Figure 8.

Fig. 8.

The 1433ζ protein is a potential determinant of unique trafficking of 5P14-stimulated CCR5. (A) Identification of proteins whose quant profiles are consistent with 5P14-specific early recruitment to CCR5. Each point in the scatter plot is a protein from the APEX dataset whose quant profile features an increase at both 3 and 10 minutes after 5P14 addition. The x-axis shows the P values of the difference between the response profiles of these proteins to 5P14 vs WT CCL5. The y-axis shows the P values of the difference between the response profiles of the same proteins to 6P4 vs WT CCL5. The highlighted top left quadrant contains proteins whose 5P14 response profiles are different from their response to WT CCL5 (P < .05), but their response to 6P4 is no different from WT CCL5 (P > .25); the identified response profiles of these proteins are shown in Supplemental Figure 8A. (B) The 1433ζ biotinylation temporal profile showed a 0- to 10-minute increase exclusively in response to 5P14. (C, D) A 3D model of 5P14-bound CCR5 (purple and black, respectively) in complex with 1433ζ (gray). (C) The overall view of the complex and (D) a close-up of the interaction interface. The C-terminal tail of CCR5 binds in the amphipathic groove of 1433ζ, with CCR5 pSer337 (shown in black sticks) forming 6 hydrogen bonds with Arg56, Arg127 (both known pSer coordination hubs) (Petosa et al, 1998), and Tyr128 of 1433ζ. Structures were predicted using the 1433ζ homodimer; the second subunit is not shown (see Materials and methods). Model coordinates are provided as Supplemental Material 4.

2.12. Statistical analysis

All raw data points were presented from 3 independent experiments conducted separately (biological replicates) along with their corresponding mean. Within each independent experiment, there were 3 technical replicates for each condition conducted in parallel, except for the APEX proximity labeling and degradation assays which had only biological replicates. For pairwise comparisons, data were analyzed using GraphPad Prism with statistically significant differences (P < .05) using unpaired t test with Welch correction. For BRET internalization, β-arrestin recruitment, and recycling, the area under the curve (AUC) was determined using GraphPad Prism and statistical significance for three-way comparisons was determined using one-way ANOVA with Tukey multiple comparisons test. For microscopy experiments, more than 200 Z-stack images containing >30 cells per condition were analyzed for Pearson correlation coefficient with statistically significant differences (P < .0001) determined using the Mann-Whitney test.

3. Results

3.1. Time-resolved ligand-dependent trafficking of CCR5 revealed by APEX2 proximity labeling proteomics

As an unbiased strategy for investigating the ligand-specific trafficking of CCR5, we used APEX2 proximity labeling proteomics (Lobingier et al, 2017; Paek et al, 2017). This technique can reveal time-dependent protein networks in proximity to CCR5 following stimulation with CCL5, 5P14, and 6P4, which in turn provides time-resolved information about receptor location in the cell. Labeling experiments were conducted in HEK293T cells stably expressing CCR5 C-terminally tagged with APEX2 peroxidase (CCR5-APEX), which biotinylates direct and indirectly interacting proteins within ∼20 nm. Because C-terminal modifications can alter signaling and trafficking of GPCRs (Lobingier et al, 2017), we first confirmed that WT CCR5 and CCR5-APEX are similar in their ability to trigger G protein activation and β-arrestin2 recruitment following chemokine stimulation (Supplemental Fig. 1, A–C). This was done using BRET-based assays measuring G protein subunit dissociation and receptor-β-arrestin2 association, respectively. Importantly, CCR5-APEX internalized at a level comparable with WT CCR5 after chemokine treatment (Supplemental Fig. 1D). We also demonstrated that the 3 different chemokines under study (CCL5, 6P4, and 5P14) induce different signaling and trafficking phenotypes of both WT CCR5 and CCR5-APEX (Supplemental Fig. 1). Consistent with previous reports that 6P4 is a superagonist, it induced greater G_i protein activation, β-arrestin2 recruitment and receptor internalization than WT CCL5 at the same concentration (100 nM). By contrast, 5P14 treatment resulted in Gα_i/Gβγ dissociation comparable with WT CCL5, but it induced the recruitment of only ∼20% of the β-arrestin2 recruited to the receptor by WT CCL5 (Supplemental Fig. 1C). Overall, the results suggest that despite the C-terminal tag, CCR5-APEX faithfully represents WT CCR5, including its differential responses to the 3 ligands, and is therefore suitable for the proximity labeling experiments.

The ligand treatment scheme and the strategy for proximity labeling coupled with TMT MS (Thompson et al, 2003) are illustrated in Supplemental Fig. 2. Briefly, the CCR5-APEX stably transfected cell line was preincubated with biotin-phenol and then treated with different CCL5 ligands for various periods. The labeling reaction was initiated with H₂O₂ and quenched shortly thereafter. The biotinylated proteins near CCR5-APEX were then pulled down from the cell lysates, trypsinized, and subjected to TMT-labeling MS analysis (see Materials and methods).

In total, 21,550 individual peptides, representing 3187 proteins, were identified from the entire proteomic dataset (see Materials and methods; Supplemental Fig. 3). Of these proteins, 428 showed statistically significant variation in their TMT-quantified abundances (for brevity, referred to as quants thereinafter), across all ligand-time conditions (Fig. 1A; Supplemental Material 1). Variations in the observed protein quants are indicative of changes in levels of biotinylation, as a result of altered proximity of the proteins to CCR5-APEX2 at different time points after agonist addition. Among the statistically significantly varied proteins, a strong overrepresentation of Gene Ontology (Ashburner et al, 2000; Paulo et al, 2016) terms associated with cellular trafficking, signaling, and cytoskeleton and cell-cell junction regulation was observed (Fig. 1B).

Fig. 1.

Time-resolved ligand-dependent trafficking of CCR5 revealed by APEX proximity labeling proteomics. (A) Heatmap of time-dependent CCR5-APEX labeling of proteins that statistically significantly varied across all ligand conditions, as evaluated by two-way moderated ANOVA with Benjamini-Hochberg (BH) adjustment for multiple comparisons (Ritchie et al, 2015) (see Materials and methods). Protein quants from all ligand conditions (different chemokines at 3, 10, 30, or 60 minutes after stimulation) were normalized to buffer control without stimulation (0 minutes). An increase in quants is indicated by varying shades of red corresponding to the level of increase. Decreases in quants are colored in blue with varying shades indicating different levels of decrease. (B) Top 20 overrepresented Gene Ontology (GO) Biological Process (BP) terms among 428 proteins labeled by CCR5-APEX that statistically significantly varied across all ligand conditions. (C) Centers of representative clusters of proteins labeled by CCR5-APEX: response profiles to CCL5, 6P4, and 5P14 are shown in red, blue, and purple, respectively. Dots represent aggregated peptide centers from all proteins in the cluster that share the same overall trend in labeling change over time and the line connects the quant distribution center of all members at each time point. GO overrepresentation analysis was performed on each cluster and the most overrepresented cellular component (CC) term associated with the corresponding cluster is shown on the right.

To further investigate the ligand-dependent trafficking patterns of CCR5, we examined the collective spatiotemporal behavior of the labeled proteins. Specifically, proteins were hierarchically clustered based on profiles of their time-dependent biotinylation (shapes of the curves of protein quants vs time) (Fig. 1C), after which overrepresentation of Gene Ontology Cellular Compartment terms was calculated separately for each emerging group of proteins (also known as protein cluster, see Materials and methods). Over 60 clusters were generated, with proteins resident to specific subcellular compartments enriched in each cluster (Fig. 1C; Supplemental Table 1). These protein clusters and their biotinylation quant profiles provide rich information about ligand-dependent CCR5 subcellular trafficking. For example, the first profile shown in Figure 1C, shared by many plasma membrane(PM)–resident proteins, demonstrates that agonist-stimulated CCR5 departs from the PM. The second profile, shared by endosomal proteins, suggests that upon departure from the PM, CCR5 traffics to early endosomes and, with slower kinetics, associates with endolysosomes, thus following the canonical endosomal trafficking pathway. Notably, the 3 ligands induced dramatically different levels and rates of CCR5 internalization, as revealed by the time-dependent quant profiles of PM–resident proteins (Fig. 1C). Compared with WT CCL5, 6P4 triggered a stronger and more rapid decrease in PM protein labeling, suggesting a faster and more pronounced departure of the receptor from the PM in response to this ligand, whereas 5P14 induced a lower level of change that was more gradual (Fig. 1C). Conversely, many early endosomal proteins exhibited a corresponding increase in labeling over time, indicating accumulation of the receptor in the early endosomal compartments after ligand stimulation, with ligand-dependent differences similar to those observed with the PM proteins (Fig. 1C). In addition to these 2 subcellular locations, we also observed marked ligand-dependent differences in labeling patterns for protein clusters associated with the cytoskeleton, late endosomes, lysosomes, endoplasmic reticulum, Golgi, trans-Golgi network (TGN), nucleus, and exocytic vesicles (Fig. 1C; Supplemental Table 1). It is also important to note that our analysis generated not a 1-to-1 but rather a many-to-many correspondence between clusters and subcellular locations: some locations were overrepresented in more than 1 response profile cluster and, conversely, most clusters showed overrepresentation of proteins from more than 1 subcellular location. For example, PM proteins were overrepresented in at least 5 clusters, all exhibiting similar overall trends but with different ligand responses (Supplemental Table 1). Interestingly, among them, clusters with overrepresentation of cytoskeletal proteins showed much less variation among ligand conditions compared with the others (Supplemental Table 1). This demonstrated that our APEX proximity data have sufficient spatial resolution to distinguish distinct regions, substructures, or microdomains of a single subcellular location. Overall, this provided a systematic view of the ligand-dependent subcellular trafficking patterns of CCR5 and guided targeted studies, described in the following sections.

3.2. β-Arrestin–dependent clathrin-mediated endocytosis drives ligand-dependent internalization

Ligand-dependent patterns of CCR5 endocytosis observed by proximity labeling were next validated by a bystander BRET assay measuring proximity between CCR5 C-terminally tagged with Renilla luciferase (CCR5-RlucII) and a PM marker (the polybasic prenylated CAAX box of KRas N-terminally tagged with Renilla green fluorescent protein, rGFP-CAAX), or early endosome marker (the rGFP-tagged FYVE domain of endofin, rGFP-FYVE) (Namkung et al, 2016). Chemokine agonists triggered CCR5 departure from the PM, as demonstrated by a decrease in the BRET between CCR5-RlucII and rGFP-CAAX (Fig. 2A). Accumulation of the receptor in the early endosome was indicated by an increase in the BRET between CCR5-RlucII and rGFP-FYVE (Fig. 2A). The different levels and rates of BRET changes associated with each ligand condition matched the ligand-dependent changes in PM or early endosome protein labeling observed in the APEX data (Fig. 1B; Fig. 2A; Supplemental Fig. 4A).

Fig. 2.

CCR5 internalization by all 3 ligands is dependent on β-arrestin, proceeds through CME and involves receptor translocation to early endosomes. (A) Ligand-dependent temporal profiles of bystander BRET between CCR5-RlucII and rGFP-CAAX (a PM marker) or rGFP-FYVE (an EE marker) in HEK293T cells support CCR5 ligand-dependent internalization patterns revealed by APEX proteomics. BRET signals from ligand-treated samples (100 nM CCL5, 6P4, or 5P14) are subtracted from buffer-treated controls and are presented from 3 independent experiments (dots) with their corresponding mean values (solid lines). Control-subtracted BRET traces from CCL5-treated, 6P4-treated, and 5P14-treated samples are colored in red, blue, and purple respectively; buffer controls are colored in black. Addition of ligands is indicated with a red arrow on the plots. (B) Heatmap of time-dependent CCR5-APEX labeling of proteins associated with clathrin-coated pits and β-arrestin2 (ARRB2, indicated by a red box). The values are protein quants from all ligand conditions (different chemokines at 3, 10, 30, or 60 minutes after stimulation) normalized to buffer control without stimulation (0 minutes). The histogram showed the distribution of the values (how many proteins are associated with a specific value). (C, D). Effect of CME inhibitors Pitstop 2 and Dyngo-4a on ligand-induced internalization of CCR5. (E) BRET between CCR5-RlucII and β-arrestin2-GFP10 in HEK293T cells showing β-arrestin2 recruitment directly to CCR5 after ligand treatment (100 nM CCL5, 6P4, or 5P14). A zoom of the BRET signal trace from the 5P14-treated sample (purple box) is shown on the right. (F) Bystander BRET between RlucII-β-arrestin2 and rGFP-FYVE in CCR5 expressing HEK293T cells showing β-arrestin2 recruitment to early endosomes after ligand stimulation of CCR5 (100 nM CCL5, 6P4, or 5P14). A zoom of the BRET signal trace from the 5P14 treated samples (purple box) is shown on the right. (G) Role of β-arrestin1/2 in ligand-induced internalization of CCR5 revealed by bystander BRET between CCR5-RlucII and rGFP-CAAX in control HEK293 cells or β-arrestin1/2 KO HEK293 cells.

After agonist stimulation, CCR5 is known to interact with β-arrestin1/2 and internalize via clathrin-mediated endocytosis (CME) (Ferain et al, 2011). We speculated that β-arrestin1/2 and CME regulators might also be responsible for the ligand-dependent differences in CCR5 internalization. Consistent with our hypothesis, we found that β-arrestin2 (ARRB2) as well as proteins known to be involved in CME, such as clathrin light chain and heavy chain proteins (CLH1/2), components of AP-2 complex (AP2B1), and Huntingtin-interacting proteins (HIP1 and HIP1R) are hits in our APEX labeling dataset, and the labeling level and kinetics of these proteins matched the ligand-dependent CCR5 internalization level and kinetics (Fig. 2B). Compared with other ligands, 6P4 induced the strongest change in labeling of β-arrestin2 and known CME regulators, including the phosphatidylinositol-binding clathrin assembly protein PICAL (Dreyling et al, 1996) and EGFR substrate EPS15 (Benmerah et al, 1995) as early as 3 minutes after stimulation (Fig. 2B). To further test whether CME is the major pathway for ligand-induced internalization, we used 2 different pharmacologic inhibitors of CME: pitstop2 and Dyngo-4A. The former inhibits interactions between CME regulators and clathrin (Dutta et al, 2012), and the latter inhibits dynamin-dependent membrane fusion (McCluskey et al, 2013). Both inhibitors suppressed CCR5 internalization by all 3 ligands as shown by the CCR5-RLucII/rGFP-CAAX bystander BRET results (Fig. 2C; Supplemental Fig. 4B); they also almost completely abrogated receptor entry into early endosomes (Fig. 2D; Supplemental Fig. 4C). Interestingly, CCR5 internalization by CCL5 and 5P14 was more severely inhibited, whereas its internalization by 6P4 was partially preserved. This may reflect an incomplete blockade of CME by pharmacologic inhibitors.

Because β-arrestins have been shown to play important roles in CCR5 internalization (Hüttenrauch et al, 2005) as well as more generally in internalization of GPCRs (Ferguson et al, 1996; Goodman et al, 1997), we further investigated the interaction of CCR5 with β-arrestin1/2 following the different ligand treatments. As mentioned earlier, 6P4 induced the fastest, strongest, and most persistent labeling of β-arrestin2, while CCL5 induced much lower and more gradual labeling compared with 6P4, but the labeling was persistent, with appreciable signal even at 60 minutes. However, with 5P14 treatment, β-arrestin2 labeling was not only slower, but more transient; it exhibited a peak at 30 minutes but dropped dramatically by 60 minutes (ARRB2 in Fig. 2B). Again, the ligand differences in proximity labeling patterns were corroborated by BRET-based effector membrane translocation assays using 2 different donor-acceptor pairs: (1) CCR5-RlucII and C-terminally GFP10-tagged β-arrestin2 (β-arrestin2-GFP10), which monitors the direct recruitment of β-arrestin2 to the receptor (Paradis et al, 2015); and (2) C-terminally RlucII-tagged β-arrestin2 (β-arrestin2-RlucII) and rGFP-FYVE, which monitors the translocation of β-arrestin2 to the early endosome (Cahill et al, 2017). Recruitment of β-arrestin2 to the receptor or to the early endosome compartment was demonstrated by increases in BRET ratios of these donor-acceptor pairs. Distinct levels and kinetics for each ligand were observed, consistent with the proximity labeling data (Fig. 2E; Supplemental Fig. 4D). Importantly, we demonstrated robust recruitment of β-arrestin2 to the early endosomal compartment with 6P4 and CCL5 treatment; however, with 5P14, such recruitment was virtually undetectable (Fig. 2F). Because both β-arrestin2 recruitment and CCR5 entry to the early endosome occurred without delay and with similar kinetics, it is likely that β-arrestin2 accompanied the receptor from the PM to the early endosome after CCL5 and 6P4 treatment as opposed to direct recruitment of β-arrestin2 to the receptor at the early endosome. This is consistent with the ability of CCL5 and 6P4 to induce more persistent β-arrestin2 interactions with CCR5, whereas 5P14 only induces transient interactions, as reported previously (Bönsch et al, 2015). Because in HEK293T cells, β-arrestin1 is much less abundant than β-arrestin2, our MS experiment did not identify any endogenous β-arrestin1 peptides in CCR5-APEX–labeled samples; however, β-arrestin1 was similar to β-arrestin2 in the corresponding BRET-based assays (Supplemental Fig. 5).

To investigate the importance of β-arrestin1/2 to ligand-induced internalization of CCR5, we next assessed CCR5 departure from the PM (via CCR5-RluII/rGFP-CAAX bystander BRET) in a β-arrestin1/2 CRISPR KO HEK293 cell line. Strikingly, knocking out β-arrestin1/2 almost completely eliminated ligand-dependent CCR5 internalization, even with 6P4 treatment (Fig. 2G; Supplemental Fig. 4E). Altogether, these data indicate that CCR5 internalization by all 3 ligands is β-arrestin dependent and the internalization is primarily clathrin mediated.

3.3. CCR5 phosphorylation by different GRK subtypes affects ligand-dependent internalization through β-arrestin1/2

It is well established that phosphorylation of GPCRs by GRKs is a prerequisite for β-arrestin recruitment and β-arrestin–dependent internalization (Yang et al, 2017). We therefore investigated whether CCR5 phosphorylation by different GRK subtypes impacts ligand-induced internalization and explains at least in part the ligand-dependent differences observed in the β-arrestin2 interactions with CCR5. To this end, we conducted CCR5-RlucII and rGFP-CAAX BRET experiments in 3 HEK293 cell lines in which different combinations of GRKs were knocked out (ΔGRK2/3, ΔGRK5/6, and ΔGRK2/3/5/6). In the ΔGRK2/3 HEK293, CCL5-mediated internalization of CCR5 was almost completely inhibited, whereas knock out of GRK5/6 (ΔGRK5/6) had no effect (Fig. 3A; Supplemental Fig. 6A), indicating the dominance of the Gβγ-dependent GRK. The signal of 5P14-induced internalization was small in parental HEK293 cells so the differences among GRK KO cell lines were harder to interpret (Fig. 3A; Supplemental Fig. 6A). Surprisingly, neither GRK2/3 nor GRK5/6 KO had a large impact on internalization induced by 6P4 (Fig. 3A; Supplemental Fig. 6A). Contrary to the diverse effects of double-GRK KOs, the elimination of all 4 GRKs (ΔGRK2/3/5/6 HEK293) completely abrogated CCR5 internalization by all 3 ligands (Fig. 3A; Supplemental Fig. 6A), consistent with its dependence on receptor interaction with β-arrestin1/2. Furthermore, the complete elimination of 6P4-induced receptor internalization in the ΔGRK2/3/5/6 cells but the lack of an effect in the ΔGRK2/3 and ΔGRK5/6 cells indicates that with 6P4-stimulated receptor, there is compensation by the remaining GRKs in these latter 2 KO cell lines.

Fig. 3.

CCR5 phosphorylation by different GRK subtypes affect ligand-dependent internalization through β-arrestins recruitment. (A, B) BRET between CCR5-RlucII and rGFP-CAAX (PM marker, A) or β-arrestin2-GFP10 (B) in parental HEK293A cells, ΔGRK2/3, ΔGRK5/6, or ΔGRK2/3/5/6 KO HEK293A cells. BRET signals from ligand-treated groups are subtracted from buffer-treated controls and are presented from 3 independent experiments (dots) with their corresponding mean values (solid lines). Control-subtracted BRET traces of different cellular backgrounds from the same ligand treatment condition (100 nM CCL5, 6P4, or 5P14) are plotted together to better demonstrate the differences in CCR5 PM departure (A) or β-arrestin2 recruitment (B) in the GRK KO lines.

To examine the dependence of the CCR5 interaction with β-arrestin2 on the different GRKs, we directly measured β-arrestin2–GFP10 association with CCR5-RlucII by BRET in the 3 GRK subtype KO cell lines. With CCL5 stimulation, β-arrestin2 recruitment to CCR5 was severely impacted in ΔGRK2/3 but not in the ΔGRK5/6 cells (Fig. 3B; Supplemental Fig. 6B). Similar to internalization, with 5P14 treatment, β-arrestin2 recruitment to CCR5 was at a much lower level even in the parental HEK293 cells, making comparisons in the GRK KO inconclusive (Fig. 3B; Supplemental Fig. 6B). With 6P4 stimulation, β-arrestin2 recruitment was inhibited in the ΔGRK2/3 but not in the ΔGRK5/6 cells, although the defect in ΔGRK2/3 cells was not as severe as for the CCL5-stimulated receptor, and it did not negatively impact receptor internalization (Fig. 3B; Supplemental Fig. 6B). This again suggests that GRK2/3 are the dominant kinases but that GRK5/6 can compensate in the absence of GRK2/3. As expected, agonist-induced β-arrestin2 recruitment to CCR5 was completely inhibited in the ΔGRK2/3/5/6 cells regardless of the agonist variant used (Fig. 3B; Supplemental Fig. 6B). Overall, the patterns of ligand-dependent β-arrestin2 recruitment to CCR5 tracked well with its ligand-dependent internalization (Fig. 3B). Together these data suggest that the different ligands induce different levels and/or patterns of CCR5 phosphorylation by GRKs, which in turn affect its ability to recruit β-arrestin1/2 and internalize.

3.4. CCL5 ligands differentially affect receptor turnover at the cell surface but do not impair recycling pathways

Some of the ligand-dependent and time-dependent proximity biotinylation profile clusters in our data show overrepresentation of proteins from subcellular locations other than PM or early endosomes (Fig. 1C; Supplemental Table 1). This indicates that the ligands differentially regulate not only CCR5 internalization but also other aspects of CCR5 trafficking inside the cell. Consistent with this, previous research demonstrated prolonged intracellular receptor retention following treatment with 6P4 and 5P14 but not CCL5 (Gaertner et al, 2008), indicating potential effects on recycling. To further probe the mechanisms, we first assessed the constitutive and ligand-induced internalization of CCR5 via prelabel versus postlabel flow cytometry (see Materials and methods). In the prelabel assay, cell surface receptors are labeled with a primary antibody before allowing them to internalize; then, after a defined period of incubation at 37 °C, receptors remaining on the cell surface are quantified with a secondary antibody. Neither the receptors that internalized nor those that reappear on the cell surface via recycling are detected by the secondary antibody. By contrast, in the postlabel assay, both the primary and the secondary antibody are applied after allowing receptors to internalize for a given period at 37 °C. This protocol ensures the detection of all cell surface receptors, including those that remain on the surface after internalization and those that emerge on the surface as a result of recycling. Thus, the difference in the surface level of receptor between the prelabel and the postlabel assay indicates the extent of receptor recycling.

According to the prelabel experiment, 1 hour treatment with CCL5 and 5P14 caused internalization of approximately 65% and 60% of CCR5, respectively (Fig. 4A). Using 6P4, almost 90% of the receptor was internalized (Fig. 4A). Note that CCR5 also constitutively internalizes as indicated by the similar level of internalization of the receptor under buffer only conditions (indicated as mock in Fig. 4A) and induced by WT CCL5. As indicated by the differences in the fluorescent staining between the prelabeled and postlabeled cells, the extent of CCR5 recycling to the surface was approximately 25% for CCL5, but only 10% in the 6P4-treated samples and 6% in the 5P14-treated samples after 30 minutes (Fig. 4A). As additional evidence, we evaluated CCR5-RlucII reassociation with the PM (via the CCR5-RlucII/rGFP-CAAX BRET assay). In these experiments, cells were treated with the CCR5 antagonist Maraviroc after 30 minutes of agonist-induced internalization to competitively inhibit receptor interaction with chemokine and thus prevent further chemokine-induced internalization (Shroka et al, 2023). Interestingly, 6P4-treated CCR5 returned to the surface at a higher level compared with WT CCL5 (Fig. 4B). However, when normalized to the total amount of internalized CCR5, the percentage of the recycling receptors was significantly lower (P = .0014) following 6P4 treatment (Fig. 4C). Consistent with the results from the prelabel and postlabel experiments, 5P14-treated receptor showed much slower internalization kinetics and reduced recycling to the cell surface compared with the other 2 ligands (Fig. 4, B and C). These results are consistent with previous research that showed more retention of CCR5 inside the cells when stimulated by 6P4 and 5P14 (Gaertner et al, 2008).

Fig. 4.

CCL5 ligands differentially affect receptor turnover at the cell surface but do not impair recycling pathways. (A) Internalization of HEK293 cells stably expressing WT CCR5 after ligand stimulation (100 nM CCL5, 6P4, or 5P14) were assessed by prelabel or postlabel flow cytometry (see Materials and methods). On the right, the remaining CCR5 surface level difference between prelabel and postlabel assays is presented for each ligand condition. Data are presented as a fraction of surface receptor remaining compared with noninternalized control at 4 °C. (B) Bystander BRET between CCR5-RlucII and rGFP-CAAX in cells stimulated with the indicated agonists (red arrow) for 30 minutes, followed by the addition of 5 μM Maraviroc (black arrow). BRET signals from ligand-treated samples are subtracted from buffer-treated controls and are presented from 3 independent experiments (dots) with their corresponding mean values (solid lines). Control-subtracted BRET traces of either inhibitor-treated or buffer-treated samples from the same ligand treatment condition are plotted together to demonstrate the differences in the BRET signal after antagonist addition. (C) AUC was calculated for BRET traces in (B) with the baseline set to 1. The differences between the AUC of the buffer-treated and the Maraviroc-treated samples is plotted for each ligand (left) and the normalization of the difference against the AUC of the buffer control for the corresponding ligand is plotted together for comparison (right). The calculated total area values were presented along with standard errors in the bar graphs. P values were calculated using one-way ANOVA with Tukey multiple comparisons test. (D) Heatmap of time-dependent CCR5-APEX labeling of proteins associated with recycling pathways. The values are protein quants from all ligand conditions (different chemokines at 3, 10, 30, or 60 minutes after stimulation) normalized to buffer control without stimulation (0 minutes). The histogram showed the distribution of the values (how many proteins are associated with a specific value). (E) Bystander BRET between CCR5-RlucII and rGFP-Rab4 (a fast-recycling endosome marker) in HEK293T cells treated with 100 nM CCL5, 6P4, or 5P14. AUC quantification is presented on the right. All data are presented from 3 independent experiments (dots) along with their corresponding mean (solid lines). P values were calculated with one-way ANOVA with Tukey multiple comparisons test. (F) Bystander BRET between CCR5-RlucII and rGFP-FYVE in cells stimulated with the indicated agonists (red arrow) for 30 minutes, following the addition of 5 μM CCR5 inhibitor Maraviroc (black arrow). BRET signals from ligand-treated samples are subtracted from buffer-treated controls and are presented from 3 independent experiments (dots) with their corresponding mean values (solid lines). Control-subtracted BRET traces of either inhibitor-treated or buffer-treated samples from the same ligand treatment condition are plotted together to better demonstrate the differences in BRET signal after antagonist addition. (G) AUC was calculated for BRET traces in (F) with the baseline set to 1. The differences between the AUC of the buffer-treated and the Maraviroc-treated samples are plotted for each ligand (left) and the normalization of the difference against the AUC of the buffer control for the corresponding ligand is plotted together for comparison (right). The calculated total area values were presented along with standard errors in the bar graphs. P values were calculated using one-way ANOVA with Tukey multiple comparisons test.

Because 6P4 and 5P14 cause CCR5 to be retained inside the cells, we wondered whether they negatively impact receptor association with components of the recycling pathway, compared with CCL5. However, according to the APEX proximity labeling data, not only 6P4 and 5P14 induced labeling of those proteins, but also they did so at a comparable or higher level than CCL5 (Fig. 4D). This was especially the case for Rab4, Rab14, and SNX27, which are involved in sorting and recycling of early endosomes (Fig. 4D). A BRET-based assessment of CCR5-RlucII proximity to C-terminally rGFP-tagged Rab4 (Namkung et al, 2016) similarly showed higher recruitment of CCR5 to Rab4-positive recycling endosomes when treated by 6P4 compared with 5P14 and CCL5, which were similar to each other (Fig. 4E). This suggests that 6P4 and 5P14 do not impair the trafficking of CCR5 into receptor recycling pathways; it is also unlikely that they cause CCR5 to be trapped in the early endosomes. Indeed, BRET assays using CCR5-RlucII and rGFP-FYVE showed receptor exiting from early endosomes after Maraviroc treatment with all 3 chemokines, albeit to a lesser extent after 6P4 or 5P14 treatment (Fig. 4, F and G). This suggests that other mechanisms, described further, may contribute to the intracellular retention of both 6P4-treated and 5P14-treated CCR5.

3.5. Compared with CCL5, 6P4 and 5P14 induce greater degradation of CCR5 through the endolysosomal degradation pathway

Our CCR5-APEX proximity labeling data revealed 3 different response profile clusters that feature overrepresentation of proteins located at late endosomes and lysosomes and higher labeling following 6P4 and 5P14 treatment compared with CCL5 (Fig. 1C; Supplemental Table 1). Notable examples within these clusters are late endosomal protein Rab7, lysosomal protein LTOR1, and components of the endosomal sorting complex required for transport (ESCRT) pathway such as STAM1/2, HGS, TOM1, TOLIP, and Endofin (ZFY16), all of which exhibited higher labeling after 10 minutes of 6P4 or 5P14 stimulation compared with CCL5 (Fig. 5A). A bystander BRET-based assessment of CCR5-RlucII proximity to C-terminally rGFP-tagged Rab7 suggested that CCR5 does enter Rab7-positive late endosomes following treatment with 6P4 and 5P14, but to a lesser extent when treated with CCL5 (Fig. 5B). Thus, we hypothesized that 6P4 and 5P14 might induce more CCR5 degradation through the endolysosomal degradation pathway. To test this, we performed degradation assays using HEK293 cells expressing Flag-tagged CCR5 and indeed observed significantly more receptor degradation following treatment with 6P4 (P = .0042) and 5P14 (P = .0005) than CCL5 (Fig. 5C). In support of our hypothesis, this difference was eliminated when the cells were pretreated with bafilomycin, inhibiting lysosome function by targeting the V-ATPase and preventing lysosomal acidification (Mauvezin and Neufeld, 2015) (Fig. 5C). No ligand-mediated difference in receptor degradation was observed using the β-arrestin1/2 KO and ΔGRK2/3/5/6 cells (Fig. 5D), presumably due to impaired ligand-induced internalization of CCR5.

Fig. 5.

6P4 and 5P14 induce more CCR5 degradation through the lysosomal degradation pathway than CCL5. (A) Heatmap of time-dependent CCR5-APEX labeling of proteins associated with late endosomes, lysosomes and endolysosomal degradation pathways. The values are protein quants from all ligand conditions (different chemokines at 3, 10, 30, or 60 minutes after stimulation) normalized to buffer control without stimulation (0 minutes). The histogram showed the distribution of the values (how many proteins are associated with a specific value). (B) Bystander BRET between CCR5-RlucII and rGFP-Rab7, a late endosome marker, in HEK293T cells stimulated with 100 nM of CCL5, 6P4, or 5P14. AUC quantification is presented on the right. (C) Western blot detection of total Flag-CCR5 in HEK293T cells pretreated or untreated with 1 μM bafilomycin (12 hours), pretreated with 50 μg/mL cycloheximide (15 minutes), and stimulated or untreated with 100 nM CCL5, 6P4, or 5P14 for 4 hours at 37 °C in the presence of cycloheximide. Endogenous α-actin was used as a loading control. Receptor degradation was quantified, and the data (mean ± SE) shown are expressed as the fraction of receptor remaining compared with untreated control cells as determined from 3 independent experiments (lower panel). (D) Degradation assays were performed as in (C) but using β-arrestin1/2 KO and GRK2/3/5/6 KO HEK293 cells and no bafilomycin treatment. All raw data points from 3 independent experiments were presented (colored dots) along with the corresponding mean value (solid lines). P values for AUC quantifications were calculated with one-way ANOVA with Tukey multiple comparisons test, and P values for the degradation assay were calculated with unpaired t test with Welch correction.

3.6. 6P4 directs CCR5 to the Golgi and TGN more than CCL5 and 5P14

Several APEX proximity biotinylation profile clusters featured overrepresentation of proteins from the Golgi apparatus and the TGN, all of which exhibited higher labeling following 6P4 treatment compared with the other ligands (Fig. 1C; Supplemental Table 1). One of the clusters showed a marked increase in labeling of Golgi/TGN resident oxysterol-binding proteins and members of the Golgin family (GOGA2-4 and GOGB1) and did so only after 6P4 stimulation for 30–60 minutes (Fig. 1B; Fig. 6A). This suggests that besides directing CCR5 to late endosomes and lysosomes, 6P4 also directs more receptor to the Golgi apparatus and TGN compared with the other 2 ligands. To further validate these APEX proximity labeling observations, we assessed SNAP-CCR5 colocalization with a TGN marker (TGN integral membrane protein 2, TGOLN2, or TGN46) by confocal microscopy in HEK293T cells. The microscopy images showed a strong colocalization of CCR5 with the TGN marker only after 6P4 treatment, as indicated by a significantly higher (P < .0001) Pearson correlation coefficient value compared with other conditions, whereas CCR5 that internalized after 5P14 and CCL5 treatment did not substantially accumulate in the TGN (Fig. 6, B and C). Taken together, it appears that 6P4 sequesters more CCR5 inside the cells for a prolonged period compared with other CCL5 ligands by mechanisms involving degradation and/or by directing CCR5 to the Golgi and TGN, whereas 5P14 retains CCR5 by receptor degradation and/or by inducing slower receptor internalization and recycling.

Fig. 6.

6P4 directs CCR5 to Golgi and TGN more than CCL5 and 5P14. (A) Heatmap of time-dependent CCR5-APEX labeling of proteins associated with the Golgi apparatus and TGN. The values are protein quants from all ligand conditions (different chemokines at 3, 10, 30, or 60 minutes after stimulation) normalized to buffer control without stimulation (0 minutes). The histogram showed the distribution of the values (how many proteins are associated with a specific value). (B) Representative microscopy images showing HEK293 cells expressing SNAP-CCR5 labeled in green (first panel) with TGN markers labeled in magenta (second panel) and nucleus labeled in blue (third panel) after 1 hour treatment with either buffer or 100 nM CCL5, 6P4, or 5P14. The merge images from 3 channels are presented in the fourth panel. Images are representatives of 3 independent experiments, where >30 cells imaged for each condition. (C) Pearson correlation coefficient (PCC) was calculated for CCR5 and TGN. More than 200 Z-stack images containing >30 cells per condition were analyzed. Each dot represents the PCC obtained from each Z-stack image. ∗∗∗∗P < .0001 determined by the Mann-Whitney test.

3.7. CCR5 differentially scavenges CCL5 ligands in a β-arrestin–dependent manner

Previous research has shown that CCR5 scavenges CCL3, CCL4, and CCL5, which is hypothesized to regulate chemokine levels and receptor responsiveness and dampen inflammatory responses induced by chemokines (Ariel et al, 2006; Turner et al, 2012). More recently, it was shown that scavenging can regulate collective cell migration (Alanko et al, 2023). Chemokine scavenging by receptors is largely driven by internalization of the receptor–chemokine complex, intracellular release and degradation of the chemokine in the lysosomes, and recycling of chemokine-free receptors back to the cell surface to pick up more ligand and repeat the process (Cardona et al, 2008; Shroka et al, 2023). Because the 3 ligands induce distinct trafficking behaviors of CCR5, we hypothesized that the scavenging efficiency of CCR5 would be ligand-dependent. To test this, the CCL5 ligands were incubated with the CCR5-expressing HEK293 cells or nonexpressing control cells, and the amount of the ligand remaining in the media after 16 hours was quantified by ELISA. As shown in Figure 7, while CCR5 scavenged approximately 50% of WT CCL5 relative to control, its ability to scavenge the other 2 engineered ligands was greatly impaired. Since we previously demonstrated that β-arrestin1/2 is required for ligand-induced receptor internalization, we performed the same experiment in β-arrestin1/2 knock out HEK293 cells and found that the scavenging ability of CCR5 for all 3 ligands was markedly inhibited (Fig. 7). The inhibition was almost complete for CCL5 and 6P4 in the KO cells, whereas inhibition of 5P14 scavenging was less severe (Fig. 7). This correlates with the apparent strength and persistence of the CCR5 interaction with β-arrestin1/2 following the different ligand treatments. Overall, our data indicate that the ability of CCR5 to scavenge different CCL5 ligands varies in accordance with the trafficking patterns induced by these ligands. This suggests that CCR5 may also differentially scavenge its endogenous ligands (CCL3/4/5), which may impact receptors with shard ligands (CCR1, CCR3).

Fig. 7.

CCR5 differentially scavenges CCL5 variants in a β-arrestin1/2–dependent manner. HEK293 parental and β-arrestin1/2 KO HEK293 cells stably expressing CCR5 and respective nonexpressing cells were cultured in media containing 5 nM CCL5 ligands (CCL5, 6P4, or 5P14) for 16 hours. Remaining ligands were quantified by ELISA by interpolating from CCL5 standards and presented as a percentage of respective non–CCR5-expressing cells. All data are expressed as the mean ± SE of 3 independent experiments. P values were calculated using the unpaired t test with Welch correction.

3.8. 1433ζ is a possible contributor to distinct CCR5 intracellular trafficking induced by engineered chemokines

As demonstrated above, the analysis of APEX biotinylated proteins based on statistical significance and collective behaviors (clustering) successfully recapitulated known global trafficking patterns of agonist-stimulated CCR5 and added important details and distinctions when the receptor is stimulated by WT CCL5 vs its engineered variants. The most pronounced differences were observed with 6P4; the trends with 5P14 were subtle and the molecular associations that dictate its peculiar effects on CCR5 trafficking remained elusive. We thus asked whether our APEX proximity biotinylation dataset could assist in generating testable hypotheses about the roles of CCR5 interactors in 5P14-induced trafficking, by studying proteins that get recruited to CCR5 stimulated by 5P14 but not by the other 2 chemokines. For this, we selected proteins whose response profiles to CCL5 and 6P4 were similar to each other but distinct from 5P14 and whose proximity to CCR5 increased (rather than decreased) at both 3 and 10 minutes following 5P14 stimulation (Fig. 8A; Supplemental Material 2; Supplemental Material 3). Of the 9 proteins that met search criteria (Supplemental Fig. 8A; Supplemental Fig. 9; Supplemental Material 2; Supplemental Material 3), 1433ζ attracted our attention as a possible contributor to the observed unique patterns of 5P14-stimulated CCR5 intracellular trafficking.

In accordance with the search criteria, the proximity of 1433ζ to CCR5-APEX significantly increased (P < .0001) when the receptor was stimulated with 5P14 but not with WT CCL5 or 6P4 (Fig. 8B). It then returned to the baseline 30 minutes after stimulation, also matching the same biotinylation levels as 30 minutes after CCL5 treatment. In contrast, 6P4-triggered biotinylation of 1433ζ decreased for 30 minutes.

1433ζ is one of the 7 known isoforms of 14-3-3 proteins, ubiquitous adapters that bind to at least 1% of the mammalian proteome using short linear interacting motifs sparsely phosphorylated at Ser and Thr residues (van Hemert et al, 2001; Obsilova and Obsil, 2022). They have been reported to interact with GPCRs both basally and in an agonist-regulated manner and to compete with β-arrestins for binding GPCR C-terminal tails (Yuan et al, 2019). We thus hypothesized that the slow and transient 1433ζ recruitment to CCR5 in 5P14-stimulated cells is similarly mediated by the C-terminal tail of CCR5, most likely phosphorylated at Ser337 as suggested by the 14-3-3-Pred server (Madeira et al, 2015).

To test this, we constructed AlphaFold3 (Abramson et al, 2024) models of 1433ζ in complex with CCR5 phosphorylated at one of its known C-terminal phosphorylation sites: Ser336, Ser337, Ser342, and Ser349 (Oppermann et al, 1999). Despite the modest confidence of these predictions (as evidenced by the pLDDT scores), the C-terminal tail of CCR5 was found to bind to 1433ζ in the canonical geometry in all cases. The complex with CCR5 (pS337) (Fig. 8C; Supplemental Material 4) featured the best pLDDT scores and conformational consistency across the AF3-predicted model ensemble; we thus hypothesized that Ser337 phosphorylation may mediate the 5P14-triggered association between CCR5 and 1433Z and is specifically promoted by CCR5 phosphorylation at Ser337 (Fig. 8D; Supplemental Material 4). As shown in Figure 3, CCR5 internalization in response to WT CCL5 is dependent on GRK2/3 but not on GRK5/6, whereas that induced by 5P14 is dependent on GRK5/6 but not on GRK2/3. Therefore, the recruitment of different GRKs to CCR5 may result in distinct patterns of C-terminal tail phosphorylation. In principle, a sparse or single phosphorylation pattern involving Ser337 may promote CCR5 association with 1433ζ while being insufficient for robust recruitment of β-arrestin. Once bound, 1433ζ can mask the C-terminal tail of CCR5 and keep it away from β-arrestin, explaining why 5P14 triggers only minimal β-arrestin association (Fig. 2E; Supplemental Fig. 8B). Bound 1433ζ may also keep the C-terminus of CCR5 away from the recycling machinery (eg, SNX27) (Supplemental Fig. 8C), thereby possibly explaining why 5P14-stimulated CCR5 is preferentially targeted to the late endosomes and lysosomes compared to WT CCL5.

This example demonstrates the multidimensional nature of our APEX proximity biotinylation dataset and its utility for generating testable hypotheses. Future studies with ζ-isoform–specific or broad-spectrum 1344 inhibitors (Stevers et al, 2018) delivered to the cells via, for example, the BioPORTER system will be needed to confirm that 1433ζ does indeed drive the 5P14-specific trafficking of CCR5 compared with the other 2 ligands.

4. Discussion

A major opportunity for developing effective therapeutics against CCR5 is the ability of different ligands to regulate receptor levels on the cell surface. Most efforts to therapeutically target chemokine receptors involve antagonists that block binding of native chemokines to their receptors and the subsequent receptor signaling responses. By preventing stimulation of downstream effectors that facilitate internalization, antagonists typically stabilize receptors on the cell surface and thus make them available for reengagement with endogenous agonists upon antagonist dissociation. However, an alternative approach is to minimize the presence of target receptors on the cell surface to render them less accessible to extracellular ligands via functional antagonism (Simmons et al, 1997; LaMontagne et al, 2006). This strategy was shown to be effective in inhibiting CCR5-mediated HIV entry due to the ability of certain chemokine variants to promote CCR5 sequestration inside the cell (Simmons et al, 1997; Mack et al, 1998; Gaertner et al, 2008). However, the mechanisms by which such ligands retain CCR5 inside the cell are poorly understood.

Our APEX proximity labeling studies revealed that the 3 CCL5 ligands exhibit major differences in receptor internalization, subcellular trafficking, recycling, and degradation. Striking differences were observed not only in the amount of receptor at various subcellular locations but also in the receptor trafficking kinetics. 6P4 induced the fastest internalization of CCR5 and allowed receptors to enter recycling pathways as well as late endosomal-lysosomal compartments for degradation. It also directed the receptor into Golgi and TGN more so than the other ligands and induced more labeling of proteins from all trafficking pathways, likely by overwhelming multiple trafficking pathways with a high level of internalized receptor (Supplemental Fig. 7A). This mode of CCR5 intracellular sequestration is similar to that previously described for a synthetically modified superagonist variant PSC-CCL5 (Escola et al, 2010). By contrast with 6P4 and CCL5, 5P14 induced the slowest receptor internalization and slowest redistribution into other subcellular locations. Nevertheless, it still directed receptors into recycling pathways as well as lysosomal degradation pathways, although it induced less receptor accumulation in the Golgi and TGN, consistent with a previous study (Bönsch et al, 2015). Despite inducing a lower level of internalized receptor compared with WT CCL5, 5P14 exhibited a strong preference for late endosome-lysosomal trafficking (Supplemental Fig. 7A).

These observations were supported by bystander BRET assays assessing receptor (CCR5-RlucII) association with different endocytic compartments (early endosomes represented by rGFP-FYVE, recycling endosomes by rGFP-Rab4 and late endosomes by rGFP-Rab7). After normalizing the receptor level in each compartment against the level of internalization, it appears that 6P4 does not preferentially direct CCR5 into any 1 compartment: they all showed high receptor accumulation (Supplemental Fig. 7B). On the contrary, 5P14 exhibited significantly higher levels of receptor (P < .0001) in Rab7-labeled late endosomes relative to the amount of the internalized receptor (Supplemental Fig. 7B). In stark contrast, CCL5 drove receptors into recycling pathways but exhibited negligible degradation and trafficking to the Golgi or TGN. Taken together, it appears that 6P4 sequesters CCR5 by rapidly internalizing a large amount of receptor and driving it into the Golgi/TGN and along lysosomal degradation pathways, while 5P14 causes slow receptor internalization and recycling as well as induction of receptor degradation.

The trafficking behavior of CCR5 impacts its ability to scavenge chemokine, which is not surprising given the dependence of scavenging on receptor internalization and recycling. CCR5 scavenged CCL5 more efficiently than 6P4 and 5P14: although 6P4 induced a faster rate and higher degree of CCR5 internalization compared with CCL5, a smaller percentage of receptors returned to the cell surface to continuously bind and internalize ligand as required for scavenging. In the case of 5P14, receptor internalization and recycling were slower than that induced by 6P4 and CCL5, consistent with the decreased scavenging of 5P14. By virtue of completely blocking ligand-dependent CCR5 internalization, β-arrestin1/2 KO fully abrogated scavenging of all ligands by CCR5.

How do these CCL5 ligands trigger the different subcellular trafficking patterns of CCR5? From the proximity labeling data, β-arrestin1/2 emerged as one of the most important master regulators. The level and kinetics of β-arrestin1/2 recruitment to CCR5 following treatment with the different ligands matched the ligand-dependent CCR5 internalization profiles. Knocking out β-arrestin1/2 eliminated ligand-induced internalization for all ligands, also attesting to its dominant role. Finally, numerous CME regulators had similar labeling patterns as β-arrestin1/2, and inhibition of CME by pharmacologic inhibitors significantly decreased ligand-induced internalization (P < .01). Thus, the ligand-dependent differences in receptor internalization seem to be defined by β-arrestin1/2 recruitment and subsequent CME.

GRKs were not observed in the proximity labeling data, consistent with the transient nature of their interactions with GPCRs (He et al, 2017). However, receptor phosphorylation is a prerequisite for β-arrestin1/2 interactions (Martins et al, 2020; Isaikina et al, 2023), and thus, we hypothesized that the level and pattern of GRK phosphorylation may play important roles in ligand-dependent internalization of CCR5. Corroborating this, β-arrestin2 recruitment to CCR5 and CCR5 internalization showed different sensitivities to KO of different GRK subtypes in a ligand-dependent manner. For CCL5 and 6P4, GRK2/3 played more important roles than GRK5/6 in recruiting β-arrestin2 and driving CCR5 internalization. Of the 2 ligands, 6P4 showed the strongest effects (Fig. 3, A and B), likely due to its higher potency and efficacy in inducing G protein activation (Supplemental Fig. 1B), which in turn results in stronger phosphorylation by GRK2/3. By contrast, 5P14 treated CCR5 was more sensitive to GRK5/6, presumably because it has a reduced capacity to activate G proteins compared with the other ligands, rendering it less susceptible to phosphorylation by the G protein–dependent kinases GRK2 and GRK3. These ligand-dependent results likely reflect different levels and patterns of “phosphorylation barcodes” (Liggett, 2011) by the different GRKs, leading to differences in β-arrestin1/2 interactions and arrestin scaffolded proteins that in turn have consequences on internalization and trafficking. In particular, using our APEX dataset, 1433ζ emerged as a protein that might chaperone CCR5 in a 5P14-specific manner as a consequence of a GRK5/6-mediated phosphorylation pattern that is sufficient to promote its interaction with CCR5 but not the interaction of CCR5 with β-arrestin1/2.

The precise influence of β-arrestin1/2 on CCR5 subcellular trafficking is more difficult to parse out. However, 6P4 and CCL5 induced persistent CCR5-β-arrestin2 interactions and recruited β-arrestin2 to early endosomes, whereas 5P14 induced more transient interactions. Similar observations were previously made based on immunoprecipitation experiments and speculated to affect CCR5 endocytic trafficking (Bönsch et al, 2015). The persistently receptor-bound β-arrestins could negatively regulate CCR5 degradation via direct interaction with the ESCRT machinery, as they have been reported to do for CXCR4 (Malik and Marchese, 2010; Alekhina and Marchese, 2016). Indeed, our APEX proximity labeling data did show that both 6P4 and 5P14, but not WT CCL5, triggered strong labeling of ESCRT components such as STAM1/2 and HGS (Fig. 5A). Because 5P14 induced only transient β-arrestin association with CCR5 and did not recruit them to early endosomes, the lack of β-arrestin interactions with ESCRT proteins in endosomal compartments may lead to more receptor degradation through the endolysosomal pathway compared with that by the other ligands (Fig. 5; Supplemental Fig. 7). On the contrary, with stronger and more persistent β-arrestin interactions, 6P4-treated and CCL5-treated CCR5 were found more in TGN and recycling pathways, respectively. This said, the mechanisms that selectively promoted the proximity to ESCRT proteins of 5P14-stimulates and 6P4-stimulated CCR5 remain elusive. In addition, mechanisms independent of β-arrestin2 could also promote 6P4-mediated and 5P14-mediated increases in receptor degradation. We also found increased labeling of components of the alternative ESCRT0 complex such as TOLIP, endofin (ZFY16), and TOM1 after 6P4 or 5P14 treatment, which targets ubiquitinated proteins for lysosomal degradation. It could be that engineered chemokines trigger specific ubiquitination of CCR5 distinct from the native chemokine CCL5. Other mechanisms may involve prolonged complexation of engineered chemokines with CCR5, potentially in the acidic environment in the endolysosomes, strong subtype G protein signaling bias (Lorenzen et al, 2018), or other as yet unknown aspects of their unique pharmacology.

While successfully revealing the ligand-dependent intracellular trafficking patterns of CCR5, our study also reveals limitations of APEX proximity biotinylation. First, APEX preferentially labels proteins that are abundant in the cell; as a result, even after removing the bulk of highest abundance proteins prevalent in proximity biotinylation experiments (Mellacheruvu et al, 2013), the dataset is dominated by trafficking and housekeeping proteins, while low-abundance signaling proteins are not well-represented. The use of data-dependent acquisition in MS further aggravates the problem. Although one might be able to highlight interesting low-abundance associations by focusing only on proteins with a small number of peptides, this approach is hard to formalize and/or automate for the discovery of direct regulators of CCR5 trafficking and signaling. Comparisons with temporal profiles of bystander/proximity signals from APEX-tagged compartment markers may help to better prioritize such proteins (Polacco et al, 2023).

Additionally, we found that aspects of constitutive receptor trafficking are hard to deconvolute from APEX proximity biotinylation data. This is because a receptor that constitutively internalizes and recycles is simultaneously present on all endomembranes and labels their resident proteins in both the basal (no agonist) and the agonist-stimulated conditions. Agonist stimulation triggers a shift in relative abundance of the receptor at different organelles (eg, a decrease of the PM population with simultaneous increase in the endosomes), but this still happens in the background of constitutive presence of the receptor in many compartments throughout the cell. Monitoring time-dependent changes in the labeled proximal proteins provides insight into changes between the basal state and the agonist-stimulated state, but it does not say much about the basal state itself. Agonist-induced redistribution explains perplexing decreases in CCR5 association with selected compartment-resident proteins (eg, GOGA2/5). In other words, APEX as well as many other experimental methods (eg, BRET or immunoflourescence) are limited in describing the constitutive trafficking of the receptor and instead report agonist-induced changes, which gives a false sense of the overall trafficking of a receptor that can be dominated by constitutive processes. The prelabel flow cytometry–based internalization is one of very few methods suitable for characterizing constitutive GPCR trafficking.

A third limitation relates to the ability of APEX to identify “a magic bullet” for the differential ligand-dependent trafficking. Although the data reveal direct and indirect interactors of CCR5 and ideally one would identify specific proteins that regulate the different trafficking patterns, in reality, trafficking of membrane receptors is a complex process not defined by single pathways and intracellular compartments. Furthermore, trafficking involves multiprotein complexes that are involved in many cellular processes; thus, many of the identified proteins can be compensated if knocked out. This was clearly demonstrated by KO of specific GRKs in the case of 6P4-stimulated CCR5. We also attempted knocking out other proteins such as PICAL, SNX3, and STAM1 with limited success. Thus, we used APEX2 to provide unbiased insight into ligand-dependent trafficking patterns that could guide pharmacologic studies. The data guided and reinforced pharmacologically targeted experiments and added substantial time-dependent and spatially detailed information to the mechanisms.

In summary, we demonstrated that ligand-dependent internalization of CCR5 is regulated through interactions with β-arrestin1/2 and CME. As a master regulator of trafficking, β-arrestin1/2 may also contribute to the postinternalization fate of CCR5 stimulated by some ligands, for example, by interfacing with ESCRT and other trafficking machinery. However, with other ligands, different mechanisms may be involved. The ligand binding geometry and residence time on the receptor may alter receptor conformation, G protein activation, phosphorylation levels and patterns, and β-arrestin1/2 recruitment levels, as well as kinetics and persistence of receptor complexes with intracellular effectors. Further insight into the ligand-dependent trafficking mechanisms of CCR5 may therefore emerge from deeper investigations of its interactions with other components of the trafficking machinery. In turn, it may be possible to exploit such knowledge for the development of potent CCR5 sequestering agents in the treatment of disease where blocking CCR5 function is desired.

Conflicts of interest

Tracy Handel is cofounder of Lassogen, Inc, and serves on the Scientific Advisory Boards of Artica, Abilita Bio, Abalone Bio, and Aikium Inc. The terms of these arrangement have been reviewed and approved by the University of California San Diego in accordance with its conflict-of-interest policies. All other authors declare no conflicts of interest.