RAMP3 determines rapid recycling of atypical chemokine receptor-3 for guided angiogenesis

Significance

G-protein–coupled receptors (GPCRs) exist within multi-protein complexes on the surface of cells in order to respond to a wide variety of extracellular stimuli such as neurotransmitters, migratory cues, hormones, light, and odors. In this study, we discover and characterize an expanded repertoire of GPCRs that interact with receptor-activity–modifying proteins (RAMPs)—a class of proteins that can modulate the type and consequences of extracellular signals to GPCRs. Specifically, we find that RAMP interaction with chemokine GPCRs is essential for enabling these receptors to bind and degrade extracellular migratory cues and thereby establish gradients for directed cellular migration. In the absence of these critical proteins, the process of blood vessel sprouting within the postnatal retina is dysfunctional.

Abstract

Receptor-activity–modifying proteins (RAMPs) are single transmembrane-spanning proteins which serve as molecular chaperones and allosteric modulators of G-protein–coupled receptors (GPCRs) and their signaling pathways. Although RAMPs have been previously studied in the context of their effects on Family B GPCRs, the coevolution of RAMPs with many GPCR families suggests an expanded repertoire of potential interactions. Using bioluminescence resonance energy transfer-based and cell-surface expression approaches, we comprehensively screen for RAMP interactions within the chemokine receptor family and identify robust interactions between RAMPs and nearly all chemokine receptors. Most notably, we identify robust RAMP interaction with atypical chemokine receptors (ACKRs), which function to establish chemotactic gradients for directed cell migration. Specifically, RAMP3 association with atypical chemokine receptor 3 (ACKR3) diminishes adrenomedullin (AM) ligand availability without changing G-protein coupling. Instead, RAMP3 is required for the rapid recycling of ACKR3 to the plasma membrane through Rab4-positive vesicles following either AM or SDF-1/CXCL12 binding, thereby enabling formation of dynamic spatiotemporal chemotactic gradients. Consequently, genetic deletion of either ACKR3 or RAMP3 in mice abolishes directed cell migration of retinal angiogenesis. Thus, RAMP association with chemokine receptor family members represents a molecular interaction to control receptor signaling and trafficking properties.

Receptor-activity–modifying proteins (RAMPs) are a family of single-pass transmembrane proteins with 3 mammalian members: RAMP1, RAMP2, and RAMP3, each encoded by a distinct and corresponding gene (1–3). The RAMPs were first identified by virtue of their requirement for promoting the forward translocation of the calcitonin-receptor–like receptor (CLR) from the cytoplasm to the plasma membrane (3). Moreover, their association with this receptor imparts ligand-binding specificity, such that a RAMP1-CLR oligomer preferentially binds the neuropeptide, calcitonin-gene–related peptide (CGRP), while the RAMP2-CLR or RAMP3-CLR oligomers display higher affinity for a related vasodilatory peptide, adrenomedullin (AM). Recent cryogenic electron microscopy resolution of the molecular interaction between RAMP1 and CLR exemplifies how RAMPs impart ligand-binding specificity (4), which has been exploited in the design of the Food and Drug Administration-approved antibody-based therapy against a G-protein–coupled receptor (GPCR) for the treatment of migraine (5). Subsequent studies have further defined critical roles for RAMPs to impart biased downstream signaling and intracellular trafficking dynamics to numerous GPCRs. However, these pleiotropic effects of RAMPs have mostly been studied in the context of a few receptors in the Family B/Secretin family, including the calcitonin receptor (CTR), secretin receptor, glucagon (6, 7), CRF₁ (8, 9), and VPAC_1/2 (8) and the Family A receptors CaSR (10) and GPR30/GPER1 (11). The strong coevolution of RAMPs with most GPCR families suggests that they may have expanded interacting partners (12, 13).

The chemokine receptor subfamily is composed of 24 different GPCRs that each bind a diverse array of chemokine and peptide ligands, leading to a wide variety of signaling events and physiological functions, ranging from innate immunity, developmental chemotaxis, and HIV infection to cancer metastasis (14–16). The redundancy of ligands binding to multiple chemokine receptors underscores the importance, and also the complexity, of chemokine signaling. As one example, the establishment of extracellular chemokine gradients for directed cell migration relies on the concerted, spatiotemporal functions of both typical signaling chemokine receptors with atypical chemokine receptors (ACKRs), sometimes referred to as “decoy” receptors (17). Unlike typical chemokine receptors which elicit a downstream signaling pathway in response to ligand, ACKRs do not signal through G proteins but rather bind and degrade ligands rapidly from the extracellular milieu. Considering this high level of complexity in ligand binding and signaling, we hypothesized that RAMPs may interact with chemokine receptors to modulate their complex signaling, trafficking, and physiological properties. Here, we developed a multimodal screening platform which identified robust interactions between RAMPs and the family of chemokine receptors. We further found that RAMPs play essential roles in determining the internalization, trafficking, and recycling of ACKR3, ultimately influencing guided angiogenesis in the postnatal retina.

Results

Identification of Chemokine Receptor: RAMP Interactions Using Bioluminescence Resonance Energy Transfer and Fluorescence-Activated Cell Sorting.

We established a bioluminescence resonance energy transfer (BRET)-based screening platform to identify RAMP-GPCR protein interactions by transiently coexpressing a constant amount of GPCR-rLuc protein with increasing amounts of RAMP-YFP protein in HEK293T cells. The canonical RAMP-interacting receptor CLR was used as a positive control and the noninteracting beta-2 adrenergic receptor (β₂ADR) served as a negative control in order to empirically establish a systematic and multivariate process for identifying and scoring RAMP-GPCR interactions. To evaluate each interaction, the first major discriminator applied was a threshold of B_max > 0.100 (Fig. 1 and SI Appendix, Table S1). All potential interactions failing to reach this minimum signal level were deemed negative, as indicated by salmon shading. Next, we applied a best-fit comparison for linearity versus hyperbolic curve fitting for all curves with a B_max > 0.100. Interactions with a linearity R² greater than the hyperbolic R² were considered poor, indicated by yellow shading. Finally, the remaining curves which satisfied the hyperbolic curve fit test were further divided based on their BRET₅₀ values. Hyperbolic curves with a BRET₅₀ > 10 were considered good interactions and colored blue, while curves with BRET₅₀ < 10 were considered the strongest possible candidates and shaded green. All chemokine receptors exhibited interactions with at least one RAMP. RAMP3 interactions were the most frequent and strongest (24/24 receptors), followed by RAMP2 (20/24) and RAMP1 (15/24) (Fig. 1).

Fig. 1.

BRET screening of the chemokine receptor family reveals heterodimers with RAMPs. ΔBRET was determined using HEK293T cells for each receptor: RAMP pair and plotted as a function of the total fluorescence/total luminescence ratio. Curves were calculated using one site-binding (hyperbola) and representative saturation isotherms are displayed for each receptor. (A) A systematic, multicomponent approach was used to score interactions. First, all interactions failing to reach a requirement of B_max > 0.1 were deemed negative (salmon). Next, a comparison of fits between hyperbolic and linear models was used where Linear R² > Hyperbolic R² was deemed a poor interaction (yellow). Finally, the remaining interactions were deemed good or strong based on the BRET₅₀ values—BRET₅₀ > 10 (blue) or BRET₅₀ < 10 (green), respectively. (B) Curves are representative of n = 3 to 4 independent experiments for each RAMP-receptor interaction, with average data reported in SI Appendix, Table S1.

Flow cytometry was performed as an orthogonal assay to verify if BRET-based protein-protein interactions translated to effects on RAMP surface expression upon cotransfection with each receptor (SI Appendix, Fig. S1A). In the absence of a GPCR, FLAG- and HA-RAMP1 and RAMP2 displayed minimal plasma membrane expression, indicative of endogenous intracellular localization. Both FLAG- and HA-RAMP3 exhibited endogenous plasma membrane expression, an effect also observed in previous studies (18). Importantly, RAMP3 plasma membrane expression could be further enhanced by coexpression with CLR or CTR. CCR5, CXCR2, CX3CR1, and CMKLR1 exhibited highly significant RAMP interactions in both BRET-based and cell-surface expression assays. Interestingly, we noted that 3 of the members of the subfamily of confirmed ACKRs (ACKR1–3, but not ACKR4) interacted robustly with RAMP3 and, unlike RAMP1/2, reduced RAMP3 cell-surface expression compared to RAMP3 expressed alone. To confirm that the observed effects on RAMP cell-surface expression were not due to overexpression, FLAG-RAMPs were titrated with a fixed concentration of receptor. We observed a saturable level of cell-surface expression above vector alone for FLAG-RAMP1 with CCR5 and FLAG-RAMP2 with CXCR2, with the plateau corresponding to a 1:1 RAMP:receptor ratio (SI Appendix, Fig. S1B). For FLAG-RAMP3, which is able to traffic to the plasma membrane alone, ACKR3 restricted its cell-surface expression in a saturable manner. ACKR3 has been reported to more broadly localize to early endosomes prior to ligand stimulation (19). Therefore, a possible explanation for the reduced FLAG-RAMP3 plasma membrane localization upon coexpression with ACKR3 is that the ACKR3-RAMP3 complex resides largely intracellularly or, alternatively, is targeted for degradation. This mechanism may also apply to other GPCRs that reduce FLAG-RAMP3 plasma membrane expression. Interestingly, but not unexpectedly, the results from the BRET and fluorescence-activated cell sorting (FACS) experiments reveal some differences in specific RAMP-GPCR interactions. The BRET assay provides a measure of total protein-protein interactions regardless of cellular location, whereas FACS is one of a number of techniques that measure the extent to which GPCRs alter RAMP expression at the cell surface. Thus, a comparative integration of results for these 2 methodologies provides the most informed interpretation for the biological functions of these putative RAMP-GPCR interactions.

ACKR3 Interacts with RAMP2/3 without Affecting G Protein or β-Arrestin Coupling.

We were particularly intrigued by the putative ACKR3-RAMP interaction because, using genetic mouse models, we previously showed that the developmental phenotypes of precocious cardiomyocyte and lymphatic hyperplasia in Ackr3^−/− mice are attributable to gain-of-function in the AM ligand (20), which elicits its signaling through the canonical RAMP-interacting receptor heterodimers CLR-RAMP2 and CLR-RAMP3. Therefore, we hypothesized that ACKR3 may also utilize RAMPs to modulate its ligand-scavenging activities. To address this, we further validated the ACKR3-RAMP3 protein interaction by observing the colocalization of Myc-ACKR3 and HA-RAMP3 at the plasma membrane of nonpermeabilized HEK293T cells by confocal microscopy (Fig. 2A) and within the cytoplasm by proximity ligation assay (Fig. 2B), to an extent similar to that observed for the CLR-RAMP–positive controls. Consistent with the established scavenging properties of ACKRs, we used the BRET-based biosensor EPAC (21) to confirm that RAMP1, -2, or -3 could not induce activation of Gαs (SI Appendix, Fig. S2A) nor of Gαi/o (SI Appendix, Fig. S2B) in response to logarithmic dosage of AM or SDF-1/CXCL12 ligands. Furthermore, we did not detect any effects of RAMPs on the AM- or SDF-1/CXCL12–induced recruitment of β-arrestin-1-YFP or β-arrestin-2-YFP to ACKR3-rLuc, as evidenced by nonlinear regression best-fit curves of BRET activation (SI Appendix, Fig. S2 C and D). Both the AM and SDF-1/CXCL12 ligand caused the recruitment of β-arrestin-1 and β-arrestin-2 to ACKR3, although SDF-1/CXCL12 showed more efficacy at lower doses than AM (SI Appendix, Fig. S2 C and D). Collectively, these data establish RAMPs as interacting partners of ACKR3, but these interactions do not influence the downstream G-protein or β-arrestin responses of ACKR3 to either AM or SDF-1/CXCL12 ligands.

Fig. 2.

Cellular distribution of ACKR3-RAMP3 and translocation to the plasma membrane. (A) HEK293T cells transfected with HA-RAMP alone or with Myc-CLR or Myc-ACKR3 as indicated. Addition of CLR and ACKR3 resulted in an increase in the detection of GPCR-RAMP at the plasma membrane. HA-tagged RAMP2 and RAMP3 exhibited very low levels of GPCR-independent localization. n = 3 for each condition. (B) PLA showed significant increases in signal for CLR with RAMP3 and ACKR3 with RAMP3 when compared with GPCR alone transfected controls. Blue is the nuclear staining and green is beta catenin to stain the membrane. (Right) The number of PLA signals per cell that correspond to the number of associations or heterodimer complexes for GPCR with RAMP3. Each color represents an individual experiment, n = 3, counting 7 to 21 individual cells per n. Statistical significance was determined by an unpaired t test. Error bars represent SEM of the means. (Scale bar, 10 µm.)

ACKR3-RAMP3 Coexpression Scavenges and Attenuates AM Signaling.

We next established cell-based cAMP-EPAC reporter assays to distinguish the cell-intrinsic and cell-autonomous functions of ACKR3 in AM ligand scavenging via activation of the CLR-RAMP3 receptor heterodimer. For example, HEK293T cells transfected with CLR, RAMP3, N-ethylmaleimide-sensitive factor (NSF), and the EPAC reporter responded in a dose-dependent manner to AM stimulation [EC₅₀ = 1.6 nM ± logEC₅₀ = 0.11 (Fig. 3A, purple circles)], and this signaling could be significantly attenuated with a large effect at high AM concentrations, evident by the development of a biphasic dose–response curve fit, by the cotransfection of ACKR3 [EC50_1 = 1 nM ± logEC₅₀_1 = 0.20 and EC50_2 = 3.1 µM ± logEC₅₀_2 = 0.46, (Fig. 3A, orange squares)] This was consistent with our previous results showing reduced pERK activation by AM in the presence of ACKR3 (20). Because we found that RAMPs did not induce ACKR3-mediated cAMP production in response to AM (SI Appendix, Fig. S2 A and B), we reasoned that the reduced potency of AM at the CLR-RAMP3 receptor is caused by reduced bioavailability of AM ligand in the presence of the ACKR3-scavenging receptor. To confirm this, and to provide a model for the cell-autonomous scavenging effects of ACKR3 on the AM ligand, we employed a coculture system in which reporter cells expressing CLR-RAMP3-EPAC were cocultured with cells expressing either ACKR3-RAMP3-NSF or glucagon-like peptide 2 receptor (GLP2R) as a nonscavenging control cell (Fig. 3B). Upon stimulation with AM, the CLR-RAMP3-EPAC reporter cells cocultured with the nonscavenging GLP2R-expressing cells stimulated maximal cAMP production with a potency of 0.52 nM ± logEC₅₀ = 0.05 (black circles). In contrast, when the CLR-RAMP3-EPAC reporter cells were cocultured with ACKR3-RAMP3-NSF–expressing cells, there was a significant loss in potency to 1.3 nM ± logEC₅₀ = 0.08, as demonstrated by the rightward shift in the dose–response curve (red squares). These data demonstrate the ability of ACKR3-RAMP3 to scavenge AM ligand in both cell-intrinsic and cell-autonomous settings, thereby attenuating AM signaling and further validating our previous studies using genetic mouse models.

Fig. 3.

The ACKR3-RAMP3-NSF complex attenuates adrenomedullin signaling and regulates the ACKR3 receptor’s rapid recycling. (A) Expression of ACKR3 in the same HEK293T cells as the CLR-RAMP3 heterodimer resulted in a marked decrease in cAMP EPAC biosensor signal, indicating reduced potency and a resulting biphasic dose–response (EC₅₀ for −ACKR3 = 1.6 nM, ± logEC₅₀ = 0.11 to +ACKR3 EC₅₀_1 = 1 nM, ± logEC₅₀_1 = 0.20 and EC50_2 = 3.1 µM, ± logEC₅₀_2 = 0.46 with Hill slope for −ACKR3 = 0.58 ± 0.08 and nHill slope for +ACKR3 nH1 = −0.92 ± 0.40 and nH2 = −1.17 ± 0.99). (B) Coculturing the CLR-RAMP3-EPAC reporter cells with HEK293T cells expressing ACKR3-RAMP3-NSF resulted in a loss of AM potency as represented by the rightward shift in the EC₅₀ from 0.52 nM ± logEC₅₀ = 0.05 to 1.3 nM ± logEC₅₀ = 0.08 (Hill slope for +GLP2R = 0.86 ± 0.08, Hill slope for +ACKR3 = 0.50 ± 0.05). For both A and B, curves and statistical significance were determined by nonlinear regression with a comparison of fits (F-test); n = 6 in duplicate. (C) Immunofluorescence confocal microscopy shows ACKR3-RAMP3 colocalized at the plasma membrane. HEK293T cells were not treated with ligand and fixed, treated with AM for 1 h and fixed, or treated with AM for 1 h and washed/allowed to recover for 4 h and fixed. After ligand stimulation, ACKR3 and RAMP3 internalized and showed colocalization with NSF intracellularly. ACKR3 in the presence of RAMP3 and NSF (Left column) resulted in the ACKR3-RAMP3 complex localizing to the plasma membrane after the 4-h recovery phase. ACKR3 in the absence of NSF (Middle column) or RAMP3 (Right column) did not recycle to the plasma membrane after removal of ligand and the 4-h recovery. Images are representative of 3 independent experiments. (Scale bar, 10 µm.)

ACKR3 Rapid Recycling and Lysosomal Trafficking Are Dependent on RAMP3 and NSF.

An inherent characteristic for establishing and maintaining chemotactic gradients for guided cell migration within discrete spatiotemporal boundaries is the rapid and dynamic depletion of extracellular ligands from the nonmigrating region (22). Elegant zebrafish studies focused on primordial cell migration in response to SDF-1/CXCL12 gradients have implicated an important function for ACKR3 in this regard (23, 24). However, the molecular partners that enable ACKR3 to rapidly and cell autonomously scavenge ligands from the extracellular compartment remain unknown (25). RAMP3, by virtue of its C-terminal PSD-95/Discs-large/ZO-1 homology (PDZ)-recognition motif associates with NSF. This has previously been shown to facilitate the rapid recycling and resensitization of CLR to the plasma membrane following ligand-dependent internalization (26). Using confocal imaging, we confirmed these original findings for CLR following a 4-h recovery after removal of AM ligand (SI Appendix, Figs. S3 and S4). Similarly, we found that HEK293T cells transfected with ACKR3, RAMP3, and NSF displayed rapid internalization of ACKR3 following 1 h of AM ligand stimulation, with subsequent recycling of ACKR3 to the plasma membrane after 4 h of ligand removal and cycloheximide treatment (Fig. 3 C, Left column, and SI Appendix, Fig. S5). In contrast, in the absence of either RAMP3 or NSF, although ACKR3 ligand-stimulated internalization occurred at 1 h, the receptor failed to recycle to the plasma membrane and remained within intracellular vesicles after 4 h of recovery (Fig. 3 C, Middle and Right columns, and SI Appendix, Fig. S5). These data demonstrate the requirement of RAMP3-NSF for the rapid recycling of the ACKR3 to the plasma membrane where it can evoke its ligand-scavenging activities.

Following this same experimental paradigm, we were curious to address whether RAMP3-NSF expression could influence the fate of ACKR3 to different endosomal-sorting pathways. To address this, we tracked the ligand-activated internalization of tagged ACKR3 to fluorescently labeled, Rab-positive endosomes in the presence or absence of RAMP3-NSF. As shown in Fig. 4 A, Left columns, and SI Appendix, Fig. S6, ACKR3 rapidly internalized to Rab4-positive vesicles following 1 h of treatment of AM ligand and was subsequently recycled to the plasma membrane after 4 h of recovery in the presence of RAMP3-NSF. Conversely, in the absence of RAMP3-NSF, AM-stimulated ACKR3 internalized to non–Rab4-positive vesicles and never returned to the cell surface (Fig. 4 A, Right columns, and SI Appendix, Fig. S6). Moreover, when cells were transfected with a GFP-Rab7a lysosomal marker, AM-activated ACKR3 was spared from the lysosomal degradation pathway and resensitized to the plasma membrane in the presence of RAMP3-NSF (Fig. 4 B, Left columns, and SI Appendix, Fig. S7). In the absence of RAMP3-NSF, however, ACKR3 was predominately sorted toward the lysosomal pathway, where its localization within the lumen of Rab7a-positive lysosomes and its absence from the plasma membrane was particularly evident 4 h after AM stimulation (Fig. 4 B, Right columns, and SI Appendix, Fig. S7). Importantly, we found the same effects of RAMP3-NSF on ACKR3 sorting when the receptor was activated by SDF-1/CXCL12 (SI Appendix, Figs. S8–S10), indicating a ligand-unbiased requirement of RAMP3-NSF for the sorting of ACKR3 to rapidly recycling Rab4-positive endosomes, while being spared from lysosomal degradation.

Fig. 4.

RAMP3 rescues internalized ACKR3 from Rab7a-positive late endosomes, diverting the receptor to Rab4-positive rapid recycling vesicles resulting in resensitization after AM treatment. (A) HEK293T cells transfected with RFP-Rab4 and ACKR3 ±RAMP3-NSF were not treated with AM, treated with 100 nM AM for 1 h and fixed, or treated and allowed to recover for 4 h. After 1 h of AM treatment, ACKR3 +RAMP3-NSF are internalized and show colocalization with Rab4 intracellularly (Middle row, first column and Inset). In the ACKR3 –RAMP3-NSF condition, ACKR3 is internalized but colocalization with Rab4 does not occur (Middle row, third column and Inset). After a 4-h recovery time post AM treatment, in the ACKR3 +RAMP3-NSF cells, ACKR3 and RAMP3 show distribution at the plasma membrane of the cell, demonstrating recycling of the receptor complex (Bottom row, first column and Inset). In the ACKR3 –RAMP3-NSF cells, after the 4-h recovery, ACKR3 is not present at the plasma membrane, and the receptor remained intracellular (Bottom row, third column and Inset). (B) Conversely, HEK293T cells transfected with GFP-Rab7a and ACKR3 ±RAMP3-NSF were treated with AM as stated above. After 1 h of AM treatment, ACKR3 +RAMP3-NSF are internalized and show limited/no colocalization with Rab7a intracellularly (Middle row, first column and Inset). However, in the ACKR3 –RAMP3-NSF condition, ACKR3 is internalized and shows robust localization in the lumen of Rab7a-postive vesicles (Middle row, third column and Inset). After a 4-h recovery time post AM treatment, in the ACKR3 +RAMP3-NSF cells, ACKR3 and RAMP3 show distribution at the plasma membrane of the cell, demonstrating recycling of the receptor complex (Bottom row, first column and Inset). In the ACKR3 –RAMP3-NSF cells, after the 4-h recovery, ACKR3 is not present at the plasma membrane but is detected in the lumen of Rab7a-late endosomes (Bottom row, third column and Inset). Images are representative of 3 independent experiments. (Scale bar, 10 µm.)

It has been established that the PDZ recognition sequence in the C′-terminal tail of RAMP3 plays a critical role in the recycling of the canonical CLR-RAMP3-NSF complex (26). To test the hypothesis that the RAMP3 type I PDZ recognition motif is a molecular determinate of NSF-ACKR3 receptor trafficking, we generated a RAMP3ΔPDZ expression vector. The deletion of the PDZ motif (-DTLL) disrupted the ligand-stimulated, selective endosomal sorting of the ACKR3-RAMP3 receptor complex (SI Appendix, Fig. S11, for AM ligand, and SI Appendix, Fig. S12, for CXCL12 ligand) irrespective of ligand specificity. Interestingly, we also noted that the receptor complex instead localized to Rab11 slow-recycling endosomes, which was previously not detected with wild-type (WT) RAMP3.

Loss of Guided Cell Migration in Ackr3^−/− and Ramp3^−/− Mice during Retinal Angiogenesis.

To determine whether RAMP3-mediated fating of ACKR3 to the rapid recycling endosomal pathway could impact the scavenging properties of the receptor in a physiological context, we turned to postnatal retinal angiogenesis as a model system of guided cell migration (27). In this context, angiogenic cues, like SDF-1/CXCL12 and AM, are enriched within peripheral astrocytes and serve as chemotactic gradients for guided angiogenesis of retinal vasculature by stimulating tip cells and filopodia (28) Fig. 5 A, D, and G). Previous studies have defined the presence of ACKR3 within trailing arterioles and, concomitantly, its notable absence from leading endothelial tip cells within the developing retina, thereby establishing its spatial positioning within the retina to maintain angiogenic gradients (28). Predictably, a 50% reduction in the expression of the scavenging receptor in Ackr3^+/− animals resulted in a significant reduction in the number of endothelial tip cells within the retinas of postnatal day 3 mice compared to control littermates (Fig. 5 B and E). Although the vast majority of Ackr3^−/− mice die at postnatal day 1, we were fortunate to obtain and characterize a single surviving animal which displayed a profound reduction in tip-cell number (Fig. 5E) and effacement of the retinal angiogenic front. Importantly, we observed the same attenuation of guided angiogenesis in Ramp3^−/− animals (29) compared to control littermates (Fig. 5 C and F). Additionally, we observed a significant decrease in tip-cell filopodia in Ramp3^−/− animals (WT vs. Ramp3^−/−; 20.7 ± 1.34 vs. 16.9 ± 0.90, P = 0.023) and a trend toward decreased tip-cell filopodia in Ackr3^+/− animals (WT vs. Ackr3^+/−; 24.4 ± 1.45 vs. 22.5 ± 1.14, P = 0.333). These findings support an essential physiological function for AM-gradient guided cell migration through the scavenging activities of ACKR3 and RAMP3 (Fig. 5G).

Fig. 5.

Vascular development is disrupted upon genetic deletion of ACKR3 or RAMP3 in vivo through a RAMP3-mediated rapid recycling mechanism. (A–C) Representative images of whole-mount immunofluorescence staining of postnatal day 3 retinal vasculature using isolectin B4 in control, AM overexpressing (Adm^hi/hi), ACKR3 heterozygous (Ackr3^+/−), ACKR3 knockout (Ackr3^−/−), and RAMP3 knockout (Ramp3^−/−) mice. Statistics were evaluated using n = 3 to 7 mice. (Scale bars, 200 μM.) (D–F) Analysis and quantification of the retinal tip cells detected at the vascular periphery in control and gene targeted mice. Colored dots represent quantitation from different retinal quadrants of individual animals. Statistical significance was determined by an unpaired, 2-tailed t test with an n between 3 and 7 mice, as indicated in the figure. Error bars represent SEM of the means. (G) Cartoon depicting gradient-guided migration of vascular endothelial tip cells in response to normal chemotactic gradients (Top) established by ACKR3/RAMP3 (yellow/green) and abnormal gradients (Bottom) in the absence of ACKR3/RAMP3 ligand scavenging. (H) Model of RAMP3-mediated endosomal sorting of ACKR3. (I) Ligand binding of ACKR3 leads to endocytosis of receptor. (II) ACKR3 is internalized to early endosomes, where it is colocalized with RAMP3 and NSF. (III) The PDZ-binding motifs of RAMP3 and NSF are required for the Rab4-positive rapid recycling endosome-dependent and RAMP3-facilitated recycling of internalized ACKR3 to the cell surface via a putative interaction with a PDZ-domain–containing protein. In the absence of RAMP3, ACKR3 is localized to Rab7a-positive late endosomes which controls the trafficking between late endosomes and lysosomes. (IV) ACKR3 scavenging is enabled by the rapid resensitization of ACKR3 to the cell surface by RAMP3.

Discussion

Collectively, these data significantly expand the repertoire of GPCRs that interact with RAMPs and provide at least one example of how a RAMP-GPCR interaction can dictate the physiological functions of guided cell migration by governing receptor endosomal sorting and recycling (Fig. 5H). We focused our efforts on the process of retinal angiogenesis because we, and others, have previously shown that both ACKR3 ligands, SDF-1/CXCL12 and AM, and their respective cognate signaling receptors, CXCR4 and CLR, are critical for driving guided cell migration within this vascular bed (28, 30). However, ACKR3 also plays critical roles in the migration of GABAergic interneurons within the embryonic cerebral cortex (1, 31) in the homing of immune cells to lymphoid tissues and in the migration of cancer cells (32, 33). Thus, it will be of interest to determine whether RAMP3 also dictates the scavenging functions of ACKR3 within these other contexts. If so, the RAMP3-ACKR3 interface should provide a therapeutic target for modulating the actions of the CXCR4/CXCL12 axis in anticancer and antiinflammatory strategies (34).

Indeed, the molecular interface formed between RAMP1 and CLR has recently been exploited in the design of the anti-CGRP migraine drug, erenumab (5, 35). In our current study, we find that 23 of the 24 described chemokine receptors display moderate-to-strong molecular interactions with RAMPs. These findings identify unique and pharmacologically tractable avenues for the modulation of chemokine function in a wide range of physiological processes. Additional studies that employ a wide range of biochemical, pharmacological, and cellular assays to elucidate the effects that each RAMP has on the ligand binding, biased functional selectivity, or trafficking of other chemokine GPCRs should provide valuable future exploration and opportunity.

Materials and Methods

Cell culture, BRET assays, flow cytometry, immunofluorescence confocal microscopy, whole-mount immunohistochemistry, in situ proximity ligation assay (PLA), receptor scavenging, internalization, resensitization, and trafficking assays performed in this study are provided in SI Appendix, Materials and Methods.

Animals.

Adm^+/hi and Ackr3^+/− mice were maintained on a C57BL/6 genetic background, and Ramp3^−/− mice were maintained on a 129/SvEv genetic background. Ackr3^−/− and Adm^hi/hi mice were generated by heterozygous intercrosses, respectively. Ramp3^−/− mice were generated through homozygous crosses of Ramp3^−/− mice, and Ramp3^+/+ isogenic mice were used as controls. A total of 6 mice across 2 Adm genotypes (3 Adm^+/+ and 3 Adm^hi/hi), 11 mice across the 3 Ackr3 genotypes (4 Ackr3^+/+, 6 Ackr3^+/−, and 1 Ackr3^−/−), and 14 mice across 2 Ramp3 genotypes (7 Ramp3^+/+ and 7 Ramp3^−/−) were used in this study. This study was powered to attain statistical significance of P < 0.05 with a 90% probability between Adm^+/+ and Adm^hi/hi, Ackr3^+/+ and Ackr3^+/−, and Ramp3^+/+ and Ramp3^−/− mice. All animal procedures and protocols were approved by the Institutional Animal Care and Use Committee of the University of North Carolina, Chapel Hill.

Data and Materials Availability.

No datasets were generated or analyzed during the current study. All correspondence and material requests should be addressed to K.M.C.