Recombinant adenoviruses encoding human transferrin receptor (TfnR), beta 2 adrenergic receptor (β2AR), CXCR4, CCR2, or CCR5 were generated by placing them using overlap PCR into a tetracycline transactivator (tTA)-regulated adenoviral vector. The acceptor peptide (AP) was fused at the N terminus to each of the chemokine receptors and β2AR, and at the C terminus of the TfnR. A separate CXCR4 expressing adenovirus was generated with the biotin ligase BirA fused to the N terminus. All constructs containing AP also has an HA tag preceding it, and the BirA-CXCR4 construct has a FLAG tag preceding the BirA.

Retinal pigment epithelial wild type cells (RPE WT, ATCC CRL-2302, gift from Dr. Sandra Schmid, UTSW) were maintained in F-12/DMEM (Gibco) and supplemented with 10% fetal bovine serum (Sigma), 100 units/ml penicillin and streptomycin, and 20 mM HEPES. When seeded for experiments, cells were cultured in DMEM supplemented with 10% charcoal/dextran treated fetal bovine serum (Thermo Scientific), 100 units/ml penicillin and streptomycin, and 20 mM HEPES. Cells were co-infected for 18 hours with tTA adenovirus and adenovirus containing a tetracycline-regulatable promoter (Tet-off) and encoding a protein of interest. In this current study, no tetracycline was added.

RPE WT were cultured and infected as described above. After 18 hr infection, cells were incubated with a biotinylation media consisting of DMEM, 100 μM biotin, 1 mM ATP, and 5 mM MgCl₂ for a designated time period (i.e. 5, 10, 15, or 30 minutes) at 37 degrees. The cells were then rinsed with PBS, detached using citric saline buffer, pelleted and resuspended in 1% BSA/PBS solution. The remaining steps were all performed on ice. After blocking in 1% BSA/PBS solution for 15 minutes, the cells were dual labeled in 1% BSA/PBS solution with a primary anti-HA antibody (Roche) for 30 minutes followed by secondary anti-mouse Alexa Fluor 647 and streptavidin-phycoerythrin (PE) (Invitrogen) for 30 minutes. Cells were rinsed using PBS and fixed in 4% paraformaldehyde. Following fixation, cells were rinsed with PBS, pelleted, and resuspended in 0.5% BSA/PBS. Permeabilized samples were first fixed in 4% paraformaldehyde then incubated with PBS containing 0.1% Triton X-100 for 15 minutes at room temperature before proceeding with the labeling steps as previously described.

Flow cytometry was performed on a Beckman-Coulter Cyan or an Invitrogen Attune and accompanying Summit software or Attune Cytometer Software. PE was excited by the 488 nm laser and the emission was collected using the 575/24 filter. The Alexa Fluor 647 was excited by the 635 nm laser and the emission was collected using the 665/20 filter. Forward scatter, side scatter, and emission were collected for a minimum of 10,000 cells for each sample. Gating on the forward versus side scatter plot eliminated debris and doublet cells. Positive fluorescence was determined by gating compared to a negative control. The percent positive and median were multiplied to calculate signal intensity.

RPE WT were seeded in two rows of a 96-well plate at a density of 20,000 cells/well; one row was designated for labeling of surface receptors, the other row for labeling total receptors after permeabilization. After 18 h of adenoviral infection, the cells were rinsed with PBS and fixed with 4% paraformaldehyde. All wells were washed three times with PBS. Additionally, the sample wells to be permeabilized were washed four times with PBS containing 0.1% Triton X-100 for 5 minutes with moderate shaking. Cells in all wells were incubated in a 1% BSA/PBS blocking solution at room temperature for 15 minutes. All cells were incubated with primary mouse anti-HA antibody (1∶5000, Roche) in 1% BSA/PBS for 1 hour at room temperature. The 1% BSA/PBS labeling solution for the permeabilized sample wells contained 0.1% Tween 20. After washing the plate three times with PBS, all cells were incubated with a secondary labeling solution containing goat anti-mouse IRDye-800CW antibody (Li-Cor) and TO-PRO-3 (Invitrogen) nuclear stain for 1 hour at room temperature. Triplicate samples receiving no labeling solution serve as background subtraction. Following this, the plate was washed five times with PBS containing 0.1% Tween 20 for 5 minutes at room temperature with gentle shaking and then three more times with PBS. Wash solution was completely removed by inversion then the plate was centrifuged at 300 g for 15 seconds. The plate was scanned using a Li-Cor Odyssey Sa Infrared Imager at 200 μm resolution, 3 mm offset, and intensity 7 for channels 700 and 800. Analysis was performed using Li-Cor Image Studio software. Anti-HA IRDye-800 signal was calculated after cell number normalization using the TO-PRO-3 signal collected in channel 700.

RPE WT were cultured on coverslips, infected with adenoviruses for 18 hours and incubated in biotinylation media for 15 minutes at 37 degrees. Coverslips were washed with PBS and fixed in 4% paraformaldehyde for 20 minutes. Cells were labeled with streptavidin conjugated Alexa Fluor 568 (Invitrogen) for 1 hour. Fixed samples were imaged on a fluorescence microscope (model Eclipse Ti; Nikon) using a 60x objective with a sCMOS camera (model C11446 Orca Flash 4.0; Hamamatsu) and equipped with motorized excitation and emission filter wheels (Sutter Instrument Co.). Image acquisition was performed using Micro-Manager. Image processing was performed using ImageJ.

RPE WT were cultured in a 6-well plate, infected with adenoviruses for 18 hr, and incubated in biotinylation media for 15 minutes at 37 degrees. Cells were washed with PBS and lysis buffer was applied for 10 minutes on ice. A cell scraper was used to dislodge cells and the lysates were transferred to microtubes on ice. After cell debris was spun down using a microcentrifuge, the lysate supernatants were combined with SDS loading buffer, treated with 1 mM DTT for 10 minutes at room temperature, and then loaded in duplicate into a 10% SDS-PAGE gel for electrophoresis at 90 V for 2 hours. The protein bands were transferred to a nitrocellulose membrane and blocked with Odyssey Blocking Buffer for 1 hour (Li-Cor). One set of samples was probed for protein expression with primary mouse anti-HA antibody (1∶5000, Roche) and rabbit anti-actin (1∶3000, Pierce) followed by secondary labeling solution containing goat anti-mouse IRDye-800CW antibody and anti-rabbit IRDye-680RD (Li-Cor). The second set of samples was probed for biotinylation with Streptavidin Alexa Fluor 750 (1∶10,000). The membranes were scanned using a Li-Cor Odyssey Sa Infrared Imager at 200 μm resolution, 3 mm offset, and intensity 7 for channels 700 and 800. For Figure 4C, band intensity analysis was quantified using Li-Cor Image Studio software. Streptavidin (SA) signal was calculated such that the normalized SA signal was equal to the western blot SA band intensity divided by the surface receptor percentage found by the on cell western multiplied by the western blot HA band intensity.

Analysis of variance (ANOVA) tests for fixed factors were performed followed by Tukey multiple comparison procedures to detect differences between group streptavidin signal means. When homogeneity of variance was not met, a Kruskal Wallis test was performed according to the method described by Zar followed by a nonparametric Tukey-type multiple comparison procedure. There were equal numbers of samples in each group. Statistical significance was considered for α = 0.05. For the Kruskal Wallis and Tukey multiple comparison tests, H_c and q values greater than the critical value for the appropriate degrees of freedom and α = 0.05 are considered significant.

To demonstrate the concept of using proximity biotinylation for receptor dimerization, several receptors fusions were created in an adenovirus vector (Figure 1). The use of adenovirus vectors offers the advantage that infection efficiency is high. For biotinylation, we have adopted the use of BirA, a biotin ligase from E. coli that recognizes and biotinylates the lysine residue on a specific acceptor peptide (AP) sequence (GLNDIFEAQKIE) [14], [15]. The BirA and AP were fused to the extracellular domains of our receptors with short flexible linkers (GSGSTSGSGK) inserted to ensure high probability of enzyme-substrate interaction (Figure 1A and B).

To measure receptor dimerization, it is important to use cells that express low endogenous levels of receptors of interests. Using flow cytometry and a cell line that expresses CXCR4 or CCR5 as positive controls, we found that retinal pigment epithelial (RPE) cells contain undetectable levels of CXCR4, CCR2, and CCR5 and are deemed suitable for the current study (Figure S1). The fetal bovine serum (FBS) that is used for routine tissue culture contains greater than 20 nM of biotin, according to the manufacturer's specification. Although this is a low level of biotin, the presence of 10% FBS during the virus infection steps to express the receptor constructs was sufficient to biotinylate the receptors without exogenously added biotin. To circumvent this problem, charcoal/dextran treated FBS that contains less than 2 nM of biotin was used, and this yielded no detectable biotinylation without addition of exogenous biotin (Figure S2). Charcoal/dextran treated FBS was used in all experiments unless otherwise noted.

To check the surface expression of our chemokine receptor constructs, we developed an on-cell western (OCW) protocol for 96 well plate to be used with an infrared imaging system. Cells were fixed and labeled with anti-HA either with or without permeabilization to determine total and surfaced expressed receptors, respectively. Fluorescence signals were scanned and normalized to a nuclear staining dye to account to cell density variability between wells. Between 35–40% of the chemokine receptors were expressed on the cell surface (Figure S3). To measure receptor dimerization, biotin and ATP were added to cells infected with pairs of CXCR4-BirA and receptor-AP. Cells expressing CXCR4-BirA and receptor-AP were verified by Western blot (Figure 1C). To detect biotinylation, Alexa Fluor 568 labeled streptavidin (AF568 SA) was used to label biotinylated receptors (Figure 1D). In cases where no virus was added or only CXCR4-BirA virus was used, cells did not have specific staining. As a negative control, we have used transferrin receptor-AP (TfnR-AP) that is not expected to dimerize with CXCR4. Indeed, TfnR was not biotinylated in cells expressing CXCR4-BirA/TfnR-AP. For CXCR4-AP, CCR2-AP, and CCR5-AP following biotinylation, immunofluorescence experiments showed that these constructs were biotinylated. In addition, we also examined possible heterodimerization between CXCR4 and a non-chemokine GPCR, β2 adrenergic receptor (β2AR). Although not widely studied, activated CXCR4 has been shown to physically interact with β2AR and regulate its downstream signaling pathway in rat cardiac myocytes [16]. This interaction has been shown previously using co-immunoprecipitation, confocal microscopy, and BRET and we have confirmed this interaction using our proximity biotinylation assay (Figure S4).

Next, different parameters that could affect the biotinylation were systematically explored. First, the amount of biotinylation should vary with the amount of biotinylation time of the expressed receptor-AP. Using flow cytometry to detect surface labeled biotinylation with streptavidin-phycoerythrin (SA-PE), we observed the degree of labeling depends on biotinylation time since biotinylated receptors would remain biotinylated even in the case that the dimerized receptors would dissociate. Indeed, increasing biotinylation time increased the amount of biotinylated CXCR4 (Figure 2A and 2B). It is unlikely that this increase in biotinylation is due to the catalysis time for AP biotinylation since k_cat is ∼0.5 min⁻¹ [15]. Instead, this is likely to reflect the continuous association and dissociation of receptors in monomer-dimer dynamic equilibrium [17]. It is interesting that the biotinylation increased linearly with time and does not seem to reach saturation after 30 minutes, and this will be further discussed later. Akin to changing the acceptor/donor ratio in BRET or FRET applications, we used different amount of adenoviruses to alter expression level. We observed increased labeling of biotinylated CXCR4 with increasing dosage of CXCR4-AP adenoviruses (Figure 2C and 2D). While there was an increase in the degree of biotinylation as CXCR4-AP was increased, the dynamic range was quite small (8 fold increase in virus amount resulted in ∼50% increase in biotinylation). From our result that biotinylation increased with time and that we held our biotinylation to 15 minutes, the degree of labeling was limited by time, such that under the same BirA expression, increasing CXCR4-AP had a smaller effect. Although small, the degree of biotinylation depended on the expression of GPCR-AP and since there was a variable expression among a cell population, we performed dual label flow cytometry experiment using SA-PE and surface immunolabeling of HA-tagged CXCR4-AP. As we expected and confirming earlier results, cells expressing more CXCR4-AP were more biotinylated across different amount of virus added (Figure 2E).

To evaluate the degree of CXCR4 homodimerization and CXCR4 heterodimerization with CCR2 and CCR5, dual channel flow cytometry measurements were performed for cells expressing pairs of CXCR4-BirA and a GPCR-AP. SA-PE signals were normalized by anti-HA signals to properly account varying GPCR-AP expression between different conditions. Furthermore, we used the product of percent positive and median signal of the positive-gated cells as a metric for comparison. The rational is that percent positive may not reveal the full detail of biotinylation as the intensity distribution of the positive-gated cells is missing from such commonly reported measurements. Similar to the immunofluorescence data, expression with a control receptor TfnR both showed little biotinylation. By comparison, expression of CXCR4-AP, CCR2-AP, or CCR5-AP together with CXCR4-BirA, all showed substantial degree of biotinylation (Figure 3A). Interestingly, our assay revealed CXCR4-CCR5 heterodimers, where there has been inconsistent literature [18]–[20]. To ascertain that the biotinylation of GPCR-AP represents bona fide receptor heterodimerization, we generated CCR2 construct lacking fused AP. We showed that CCR2-no AP could effectively compete with CCR2-AP, leading to a marked decrease in CCR2 biotinylation (Figure 3B). This is further confirmed by Western blot analysis showing the reduction in CCR2 biotinylation in cells expressing both CCR2-AP and CCR2-no AP compared to cells expressing CCR2-AP (Figure 3C). Together, these experiments demonstrated the specificity of the biotinylation towards only AP fused receptors.

To investigate whether CXCR4 agonist C-X-C motif ligand 12 (CXCL12) would affect homodimerization of CXCR4, biotinylation experiments were performed in the presence and absence of 5 nM CXCL12 for different amount of biotinylation time (Figure 3D). We found that surface biotinylated CXCR4 was reduced by the addition of CXCL12 at all the time points. Since CXCL12 is known to induce CXCR4 internalization through clathrin-mediated endocytosis [21], the reduction in surface biotinylation could be in part due to the internalization of biotinylated receptors. Indeed, CXCL12-treated cells had comparable level of total biotinylated receptors measured with permeabilized samples at 37 degrees (Figure 3E), indicating the effect of CXCL12 induced CXCR4 endocytosis. Since endocytosis is heavily temperature dependent [22], we carried out the biotinylation experiment at room temperature (∼20°C) which would significantly reduce endocytosis. Interestingly, in the presence of CXCL12, there was a significant decrease in the biotinylation (p<0.001), suggesting that there might be a second contributor to the reduction in surface biotinylation of CXCR4-AP. We attribute this decrease to the stabilization of CXCR4 dimers which increased dimer lifetime leading to less rounds of biotinylation. These results are summarized in a schematic that illustrates how CXCL12 reduced surface biotinylated CXCR4 at 37 degrees and room temperature (Figure 3F). More specifically, in a permeabilized sample at 37 degrees, there would be little difference in the detected biotinylated CXCR4. However, in a non-permeabilized sample where only surface receptors were measured, increased endocytosis in CXCL12-treated cells would lead to a reduction in biotinylated CXCR4. At room temperature, where membrane trafficking slowed down significantly, the reduction in biotinylation in the presence of CXCL12 supported the idea that CXCR4 dimers were stabilized by CXCL12.

One of the most notable advantages of our biotinylation assay is the ability for multiplex detection for receptor dimerization. Analogous to donor and acceptor in FRET, BirA and AP can be viewed as donor and acceptor, respectively. In FRET, it would be very difficult to resolve FRET signals using a single fluorescence donor and multiple acceptors, though bimolecular fluorescence complementation FRET can circumvent this shortcoming [23]. In our scheme, a single CXCR4-BirA can biotinylate multiple receptors that are fused to AP, in this case, CXCR4-AP, CCR2-AP, and CCR5-AP (Figure 4A). Subsequently, separation and identification of the biotinylated receptors could reveal the differential homodimerization and heterodimerization of CXCR4. Although CXCR4, CCR2, and CCR5 have similar molecular weights (39,745, 41,914 and 40,524 Da respectively), the three biotinylated receptors could be resolved on a fluorescent Western blot (Figure 4B). We were able to deduce the bands by systematically removing one receptor-AP and look for the difference in band patterns. CXCR4-AP, CCR2-AP, and CCR5-AP had similar levels of surface expression relative to the total from our OCW experiments (Figure S3) and this information was used to determine the surface receptor expression for the Western blot data. In RPE cells that expressed all three receptor-AP constructs, we detected the highest proportion of biotinylated CXCR4-CXCR4 followed by CXCR4-CCR5 then CXCR4-CCR2 dimerization but no significant difference between pairs (p = 0.087). When only CXCR4-AP and CCR2-AP were expressed along with CXCR4-BirA, we found more CXCR4 homodimer compared CXCR4-CCR2 heterodimer (p = 0.001) (Figure 4C).