GPCR-selective effects of endocytosis on cellular signaling through the cAMP/PKA cascade

Abstract

Many G-protein–coupled receptors (GPCRs) use endocytosis to promote gene transcription by prolonging signaling through the Gs-coupled cAMP/cAMP-dependent PKA cascade. However, not all GPCRs efficiently internalize after agonist-induced activation, and, among those that do, considerable differences have been observed in the ability of different GPCRs to stimulate endosomal cAMP production in different cell types. We asked if endocytosis distinguishes the signaling profiles of GPCRs that are naturally coexpressed in the same cells, focusing on three Gs-coupled GPCRs endogenous to human embryonic kidney–derived (293) cells: the vasoactive intestinal peptide receptor-1 (VIPR1/VPAC1) and β2-adrenergic receptor (β₂AR/ADRB2), which both rapidly internalize after activation, and the adenosine-2B receptor (A_2BR/ADORA2B) which we show here does not. For VIPR1, endocytosis significantly prolongs both the global cAMP elevation and cytoplasmic PKA activity increase. For β2AR, endocytosis has little effect on the global cAMP elevation but, nonetheless, it significantly prolongs the cytoplasmic PKA activity increase. A_2BR differs still further, with endocytosis having little effect on signal duration measured at either intermediate step. We then show that further downstream steps in the cascade, nuclear PKA activation and transcriptional induction, are significantly endocytosis dependent when stimulated through VIPR1 and β2AR but endocytosis independent through A_2BR. We conclude that endocytosis indeed distinguishes the signaling profiles of endogenously coexpressed GPCRs. We propose that quantitative differences in GPCR internalization and activation in endosomes program in cells a GPCR-selective, spatiotemporal “cAMP code” that is spatially “decoded” by proximity to local cytoplasmic PKA stores and then temporally interpreted by the nucleus.

Cells sense their chemical environment through membrane-embedded receptors that link the binding of a cognate agonist ligand to specific regulatory effects on cellular biochemistry and physiology. G-protein–coupled receptors (GPCRs), nature's largest receptor family, signal by allosterically coupling to heterotrimeric G protein transducers that cluster into four main classes, each differing in downstream effects and in biochemical selectivity for coupling among GPCR paralogs (1). Mammalian cells characteristically coexpress many more GPCR paralogs than G protein classes (e.g., (2, 3)), however, and can discern a remarkably large diversity of chemical cues (4, 5, 6, 7). Thus, there must exist additional mechanisms for achieving selectivity in cellular GPCR signaling.

Achieving signaling selectivity is a fundamental problem in cell biology that is not unique to GPCRs; nevertheless, it is of particular importance to signaling by GPCRs because of the large number of different GPCR paralogs that mammalian cells characteristically coexpress (2, 3). Studies across multiple systems and pathways have converged on two basic cellular strategies for conferring additional selectivity on receptor-mediated signaling. First, selectivity can be achieved temporally, such as by generating the same signal transiently from one receptor and persistently from another (8). Second, selectivity can be achieved spatially, such as by segregating key pathway components into multiprotein complexes or in laterally separated domains of the plasma membrane (9, 10).

Accumulating evidence indicates that ligand-activated GPCRs can initiate G protein signaling from endomembranes, as well as from the plasma membrane, suggesting an added dimension of signal specification (11, 12, 13). A clear example is signaling from endosomes, which has been shown to promote downstream transcriptional induction by producing a prolonged, GPCR-initiated activation of the Gs-coupled cAMP/cAMP-dependent PKA cascade (14, 15, 16). However, GPCRs have long been known to differ considerably in their ability to internalize after agonist-induced activation (17, 18, 19). Furthermore, among those GPCRs that efficiently internalize after activation, the observed effects on signaling differ. In some studies, endocytosis has been observed to substantially prolong the global cAMP production, whereas, in others, it appears to selectively facilitate PKA activation downstream of measured changes in global cAMP (11, 20, 21). Such distinctions have been recognized largely through comparisons across different studies and often utilizing individual GPCR types. Thus, it is not known if they result from true receptor-selective differences in signaling by endocytosis or reflect other variables between studies, such as cell background examined or experimental conditions (11, 13). Does endocytosis differentiate the functional signaling profiles of distinct GPCR paralogs when coexpressed in the same cells, at endogenous levels, and under comparable conditions of experimental activation?

We investigated this question in the present study by systematically profiling cAMP/PKA signaling produced by selective activation of three different Gs-coupled GPCR paralogs that are endogenously expressed in human embryonic kidney 293 (HEK293) cells, using simple bath agonist application and then examining the effects of endocytic inhibition on each profile. Our results support the hypothesis that endocytosis indeed enables cells to generate distinct signaling profiles through the conserved cAMP/PKA cascade. We propose a model in which endocytosis, by sculpting GPCR-selective differences in the location and duration of cellular cAMP production, generates a spatiotemporal “cAMP code” that is subsequently interpreted or “decoded” in both dimensions by intracellular PKA stores.

Results

β2-adrenergic receptor, vasoactive intestinal peptide receptor, and adenosine-2B receptor are natively coexpressed but differ in their agonist-induced trafficking

As a first step toward testing the hypothesis that endocytosis enables cells to generate different native GPCR signaling profiles, we sought to identify Gs-coupled receptors that are endogenously coexpressed in our HEK293 cell model and that can be selectively activated using well-characterized agonists (Fig. 1A). Previous studies have reported Gs-coupled signaling in HEK293 cells stimulated by the β-adrenergic catecholamine agonist isoproterenol (Iso), polypeptide hormone agonist vasoactive intestinal peptide (VIP), and adenosine analog 5′-N-ethylcarboxamidoadenosine (NECA) (15, 22, 23). We previously showed that vasoactive intestinal peptide receptor-1 (VIPR1) (VPAC1) is the predominant GPCR mediating VIP-stimulated cAMP production in our (American Type Culture Collection derived) HEK293 cell line (24). Using subtype-selective antagonists (CGP 20712 for β1AR and ICI 118551 for β2-adrenergic receptor [β2AR]) (25, 26), we identified β2AR (ADRB2) as the predominant GPCR mediating the Iso-induced cAMP elevation in this same cell isolate (Fig. S1). Adenosine-2B receptor (A_2BR) (ADORA2B) was previously shown to be the major GPCR activated by NECA in HEK293 cells (23). We verified this in our cell line by CRISPR knockout (Fig. S2, A and B) and by demonstrating little effect on the cellular cAMP elevation stimulated by CGS-21680, a potent and selective agonist for the adenosine 2A receptor (A_2AR/ADORA2A; Fig. S2C). Together, these results indicate that VIP signals mainly through VIPR1, Iso through β2AR, and NECA through A_2BR in our cell model.

Figure 1.

GPCR-selective differences in agonist-induced internalization.A, the three GPCRs examined in the present study and agonists used to selectively activate them at native levels in HEK293 cells. B, differences in agonist-induced internalization measured among the selected GPCRs after recombinant expression of N-terminally HaloTagged receptors. Cells were incubated in the presence of each agonist (1 μM Iso, 500 nM VIP, or 20 μM NECA) for the selected time interval and surface labeled with a membrane-impermeant HaloTag dye (JF635i). Changes in surface-labeled receptor fluorescence were determined by fluorescence flow cytometry, and the percent of internalization was calculated as the agonist-induced loss of surface receptor labeling, normalized to the total labeling detected on cells not exposed to agonist. N = 3 independent experiments, and error bars represent SD. GPCR, G-protein–coupled receptor; HEK293, human embryonic kidney 293 cell line; Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

We next asked if the selected GPCRs differ in their ability to internalize after activation. To do so, we used a flow cytometry–based internalization assay utilizing a cell-impermeant HaloTag dye (JF₆₃₅i-HTL) (27) that selectively labels N-terminally HaloTagged receptors when present in the plasma membrane. Consistent with previous reports (22, 24, 28), we observed rapid, agonist-induced internalization of both the tagged human VIPR1 and β2AR. In contrast, we were unable to detect any regulated internalization of the human A_2BR over the same 30-min time course (Fig. 1B). Similar results were obtained using an analogous assay that achieves surface labeling with a monoclonal antibody (Fig. S3A). We were initially surprised by the lack of detectable A_2BR internalization because a tagged rat A_2BR was previously shown to rapidly internalize under similar conditions (29). However, this internalization process required phosphorylation of a specific serine residue (S329) present in a distal C-terminal tail of the rodent A_2BR (30) that is not conserved in human A_2BR (Fig. S3B). Thus, we speculate that the presently observed lack of rapid, agonist-induced internalization of the human A_2BR reflects a species-specific difference between human and rodent orthologs.

In light of the presently observed differences in GPCR trafficking properties, we anticipated that β2AR and VIPR1 are activated after agonist application in endosomes as well as the plasma membrane, as described previously (24, 31), and that A_2BR is activated primarily in the plasma membrane. To test this, we used an mVenus-labeled mini-Gs (mG_s) construct as a conformational biosensor of agonist-induced GPCR activation (31, 32, 33). For all three GPCRs, we observed agonist-dependent recruitment of the labeled mG_s to the cell periphery, indicating that all of them undergo conformational activation in the plasma membrane as expected (Fig. S4). Even though mG_s can inhibit GPCR internalization when expressed at high levels (34), in our experiments, we were able to detect agonist-induced accumulation of FLAG antibody–labeled β2AR and VIPR1 in puncta representing endocytic vesicles. These puncta also accumulated mVenus-mGs, indicating the presence of conformationally activated β2AR and VIPR1 in endosomes as well as in the plasma membrane (24, 31). In contrast, we did not observe any visible recruitment of the labeled mG_s biosensor to intracellular puncta in cells overexpressing FLAG-tagged A_2BR, either in the presence or the absence of NECA. We did observe some intracellular puncta containing FLAG-tagged A_2BR under both conditions, suggesting a low level of constitutive (agonist-independent) internalization of the human A_2BR, but these puncta did not detectably colocalize with labeled mG_s. Together, these results confirm that VIPR1 and β2AR are activated by cognate agonists both in endosomes and the plasma membrane, whereas activation of A_2BR is largely restricted to the plasma membrane.

Distinct profiles of endogenous β2AR, VIPR1, and A_2BR signaling

We next profiled signaling through the canonical cAMP/PKA cascade when stimulated by each GPCR, using real-time biosensors of cAMP (cADDis) and PKA activity (ExRai-AKAR2) (35). Importantly, all signaling experiments were performed without overexpression of tagged receptors to measure effects only of the endogenously expressed receptor complement. We used a bath agonist application protocol to simplify comparisons across receptors and measured responses over a range of agonist concentrations (Fig. S5A). We focus here on concentrations that stimulated a maximal response across all assays (Fig. S5B, Table S1): 100 nM for Iso and VIP and 100 μM for NECA.

Intracellular cAMP elevation

Bath application of each agonist produced a similar maximum level of global cAMP increase detected by cADDis (Fig. S6A). However, the kinetic profile of the cAMP elevation produced by each agonist was distinct (Fig. 2A). The Iso response peaked and decreased gradually over the 30-min time course. The VIP response peaked and decayed more rapidly and then reached a second peak followed by a plateau as described previously (24). The NECA response differed still further, remaining prolonged at a high level throughout the measured time course of agonist exposure. Confirming pronounced temporal differences in the cAMP responses produced by each agonist, we quantified decay constants and found them to differ by ∼30-fold (Table S2). We also calculated a “sustained activity metric,” which compares the response measured after 30 min with that measured at the peak, as described previously (36). Using this metric, and fully consistent with the qualitative descriptions summarized above, we verified that NECA stimulated the most prolonged cAMP response, VIP stimulated the most transient response, and Iso stimulated a transient response but with distinguishable kinetics from that stimulated by VIP (Fig. S6B).

Figure 2.

GPCR-selective differences in cAMP and PKA dynamics. Fluorescence time courses of biosensors for total intracellular cAMP (A), total cellular PKA activity (B), and nuclear PKA activity (C) upon stimulation of HEK293 cells with 100 nM Iso (green circles), 100 nM VIP (blue triangles), or 100 μM NECA (pink squares). N = 3 independent experiments, and error bars represent SD. GPCR, G-protein–coupled receptor; HEK293, human embryonic kidney 293 cell line; Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

Intracellular PKA activity

Each native GPCR also produced a distinct temporal profile of cellular PKA activation (Fig. 2B), and the profile produced by each receptor resembled the global cAMP elevation measured using cADDis, as described above. This close temporal correspondence was verified quantitatively by the similar metrics of peak and sustained activity for each agonist between the assays (Fig. S6, A and B).

Nuclear PKA activity

To probe PKA activity in the nucleus, we expressed a version of the PKA biosensor fused to a tandem nuclear localization signal (ExRai-AKAR2-2xNLS) (37). Each agonist produced a different profile of nuclear PKA activity elevation (Fig. 2C). However, in contrast to the global PKA activity increase that broadly mirrored the global cAMP elevation, the kinetic profile of the nuclear PKA activity increase differed markedly for all agonists. Specifically, the nuclear PKA activity was generally delayed and more long-lasting (Fig. S6C). This finding is consistent with previous results (38, 39, 40, 41), and, in particular, the noted resemblance of nuclear PKA activity to a temporally integrated or “low-pass filtered” version of the cytoplasmic PKA activity (38).

Effects of endocytosis on the endogenous GPCR signaling profiles

We next asked how endocytosis impacts each of the measured signaling profiles. As VIPR1 and β2AR rapidly internalize after agonist-induced activation, but the human A_2BR does not, we anticipated that endocytosis might differentiate the native signaling profiles of these GPCRs. We measured agonist effects on global cAMP and nuclear PKA activity as described above and further refined our assay of cytoplasmic PKA activity by fusing a nuclear export signal to the biosensor (ExRai-AKAR2-NES) to drive its active exclusion from the nucleus (37). We then assessed the effect of inhibiting dynamin-dependent endocytosis on each signaling profile, based on previous knowledge that agonist-induced endocytosis of both VIPR1 and β2AR is dynamin dependent (24, 42).

Effect of endocytosis on global cAMP elevation

Prolonged genetic inhibition of endocytosis, imposed by 24-h expression of a dominant-negative mutant version of dynamin (mCherry-DynK44E), suppressed the maximum cAMP elevation produced by all the GPCRs tested, and it also reduced the later phase of cAMP elevation mediated specifically by VIPR1 (Figs. S7A, S8–S10). We interpret the receptor-nonspecific component as a distinct, long-term adaptation because it was not observed when endocytic inhibition was imposed by a 10-min preincubation of cells with Dyngo4a, an acute chemical inhibitor of dynamin activity (Fig. S7B). The receptor-selective effect on the kinetics of the cAMP increase, in contrast, was observed using either method of endocytic inhibition. Indeed, when kinetic profiles measured in mCherry-DynK44E–expressing cells were normalized to their individual maxima, rather than to the maximum measured in control cells not expressing mutant dynamin, the receptor-nonspecific component was no longer evident, and the receptor-selective kinetic effect was clear and indistinguishable between methods (Fig. S7C). Dyngo4a produced significant cytotoxicity after prolonged (more than a few hours) drug exposure, as was necessary for assessing effects of endocytic inhibition on downstream transcriptional control (see below). We interpret this as an off-target effect of prolonged Dyngo4a exposure because endocytic inhibition by mCherry-DynK44E for even longer (≳24 h) was well tolerated. For these reasons, and to maintain consistency in the endocytic inhibition method across signaling assays, in subsequent experiments, we focused on mCherry-DynK44E and individually normalized the data.

Endocytic inhibition had little effect on the overall dynamics of the global cAMP elevation elicited by Iso or NECA, as noted above, but it specifically reduced the later phase of cAMP elevation elicited by VIP to a level indistinguishable from vehicle treatment (Figs. 3A, S8). To examine in more detail the effect of endocytic inhibition on the dynamics of global cAMP, we calculated a difference curve between the individually normalized control and DynK44E time courses (Fig. 3B). This verified (24) an effect of endocytosis in prolonging the cAMP elevation stimulated by VIP, without detectably impacting the kinetics of global cAMP elevation stimulated by Iso or NECA. This is consistent with previous results indicating that β2AR stimulates cAMP production from endosomes relatively weakly (21) and with the present evidence that human A_2BR internalizes poorly over the measured time course of signaling. We further verified a significant effect of endocytosis on the VIP-stimulated response, with only a trend or no effect on that stimulated by Iso or NECA, both by area under the curve integration (Fig. 3C) and using the sustained activity metric (Fig. 3D).

Figure 3.

Effect of endocytic inhibition on global cAMP dynamics.A, fluorescence time course of a total intracellular cAMP biosensor (cADDis) in control cells expressing mCherry (control, colored shapes) or endocytosis-inhibited cells expressing mCherry-DynK44E (DynK44E, open shapes), after bath application of 100 nM Iso (green circles), 100 nM VIP (blue triangles), or 100 μM NECA (pink squares). Data are normalized as displayed in Figure S7C. B, difference curve, generated by subtracting the endocytosis-inhibited curve from the control curve for each agonist to further extract effects of endocytosis on dynamics. C, area under the curve (AUC) integration of the difference curve, to assess the overall contribution of endocytosis to prolonging the cAMP elevation. Statistical significance was determined using an ordinary one-way ANOVA with Tukey's multiple comparisons test. D, a “sustained activity metric,” as described and cited in the text, is defined as the fluorescence elevation measured after 30 min of continuous agonist exposure divided by that measured at the cAMP peak. This provides a simple metric describing the ability of each agonist to produce a prolonged cellular response. Significance was determined using an ordinary two-way ANOVA with Sidak's multiple comparisons test. For all panels, N = 3 independent experiments, and error bars represent SD. Statistically significant differences are indicated (ns = not significant, ∗p < 0.05, ∗∗∗p < 0.001). Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

Effect on cytoplasmic PKA activity

Endocytic inhibition markedly reduced the later phase of the cytoplasmic PKA activity elevation produced by both VIP and Iso, but it had little or no effect on the activity increase produced by NECA (Figs. 4A, S9). These results largely track the observed effects of endocytosis on the global cAMP elevation for VIP (strong effect) and NECA (no detectable effect). However, the two assays diverged strikingly for Iso. Here, endocytic inhibition suppressed the later phase of PKA activity elevation induced by Iso while having little measurable effect on the cAMP kinetics (compare Figs. 4 and 5). This selective effect of endocytic inhibition on the later phase of the Iso-induced PKA activation was clearly evident in the calculated difference curve (Fig. 4B) and verified by area under the curve analysis (Fig. 4C) as well as the sustained activity metric (Fig. 4D). These results are consistent with previous studies indicating that endocytosis prolongs β2AR-mediated stimulation of the cAMP/PKA cascade primarily by increasing the efficiency of PKA activation downstream of cAMP and without producing sufficient cAMP from endosomes to substantially increment the global cytoplasmic cAMP increase (15, 21, 37, 43).

Figure 4.

Effect of endocytic inhibition on cytoplasmic PKA dynamics.A, fluorescence time course of a cytoplasmic PKA activity biosensor (ExRai-AKAR2-NES) in control cells expressing mCherry (control, colored shapes) or endocytosis-inhibited cells expressing mCherry-DynK44E (DynK44E, open shapes), with 100 nM Iso (green circles), 100 nM VIP (blue triangles), or 100 μM NECA (pink squares) bath applied as indicated. B, the difference curve generated by subtracting the endocytosis-inhibited curve from the control curve measured using each agonist. C, area under the curve (AUC) integration of the difference curve to assess the contribution of endocytosis in prolonging the elevation of cytoplasmic PKA activity. Statistical significance was determined using an ordinary one-way ANOVA with Tukey's multiple comparisons test. D, the “sustained activity metric,” defined as the fluorescence elevation measured after 30 min of continuous agonist exposure divided by that measured at the peak of measured cytoplasmic PKA activity. Significance was determined using an ordinary two-way ANOVA with Sidak's multiple comparisons test. For all panels, N = 3 independent experiments, and error bars represent SD. Statistically significant differences are indicated (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001). Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

Figure 5.

Effect of endocytic inhibition on nuclear PKA dynamics.A, time course of nuclear PKA activity detected by ExRai-AKAR2-2xNLS in control cells expressing mCherry (control, colored shapes) and in endocytosis-inhibited cells expressing mCherry-DynK44E (DynK44E, open shapes), after bath application of 100 nM Iso (green circles), 100 nM VIP (blue triangles), or 100 μM NECA (pink squares). B, difference curves generated by subtracting the endocytosis-inhibited time course from the control time course for each agonist. C, area under the curve (AUC) integration of the difference curve, to assess the overall contribution of endocytosis to the nuclear PKA activity increase. Statistical significance was assessed using an ordinary one-way ANOVA with Tukey's multiple comparisons test. D, a “sustained activity metric,” defined as the fluorescence elevation measured after 30 min of continuous agonist exposure divided by the maximum. Significance was assessed using an ordinary two-way ANOVA with Sidak's multiple comparisons test. For all panels, N = 3 independent experiments, and error bars represent SD. Statistically significant differences are indicated (ns, not significant, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001). Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

Effect on nuclear PKA activity

Significant differences were also observed in the effects of endocytic inhibition on the nuclear PKA activity increase. Here, endocytic inhibition markedly reduced the ability of Iso and VIP to produce a prolonged increase of nuclear PKA activity; in contrast, the time course of the PKA activity elevation produced by NECA was unchanged (Figs. 5A, S10). This receptor-selective difference was evident also in the calculated difference curve (Fig. 5, B and C) and sustained activity metric (Fig. 5D).

Relationship to downstream control of cAMP-dependent transcription

We assessed transcriptional control by measuring stimulation of cAMP-dependent response element (CRE)–driven luciferase reporter expression. We measured induction after activation of each native GPCR type for 5 h by bath application of Iso, VIP, or NECA and then normalized each response to that produced by direct activation of cellular adenylyl cyclases for the same period with forskolin (10 μM). All agonists tested produced a robust, concentration-dependent transcriptional response using this assay (Fig. 6A), and, remarkably, NECA produced the strongest response among the three agonists (Fig. 6, A and B).

Figure 6.

Relationship to downstream control of cAMP-dependent transcription.A, concentration-dependent induction of CRE-dependent luciferase reporter gene expression induced by bath application of Iso (green circles), NECA (pink squares), or VIP (blue triangles) for 5 h. Transcriptional induction was assessed by the change in luminescence signal produced by each agonist normalized to that stimulated by forskolin (normalized ΔLum). A three-parameter concentration–response fit is shown, and estimated EC₅₀ values are listed in Table S1. B, comparison of maximal responses for each agonist determined from the concentration–response analysis. Statistical significance of the indicated differences was assessed using an ordinary one-way ANOVA with Tukey's multiple comparisons test. C, transcriptional reporter assays in control cells expressing mCherry (control, colored shapes) or endocytosis-inhibited cells expressing mCherry-DynK44E (DynK44E, open shapes). Induction was measured after bath application of Iso (green), VIP (blue), or NECA (pink) for 5 h, and forskolin-normalized ΔLum values are shown. Three-parameter dose–response curve fits are shown, and EC₅₀ values are listed in Table S1. Significance was determined using a two-way repeated-measures ANOVA with Sidak's multiple comparisons test. For all panels, N = 4 independent experiments, and error bars represent SD. Statistically significant differences are indicated (ns, not significant, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001). Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide.

We then assessed the effect of endocytic inhibition imposed by DynK44E (Fig. 6C). Endocytic inhibition significantly suppressed the maximal amount of transcriptional induction produced by Iso and VIP (∼40% and ∼60% inhibition, respectively), without detectably changing its concentration dependence (Table S1). In contrast, and under the same conditions, endocytic inhibition had very little effect on the transcriptional response elicited by NECA, reducing the measured luciferase activity by ≤15% and producing a statistically significant suppression at only one concentration. These results indicate that both β2AR and VIPR1 strongly depend on endocytosis for driving transcriptional control, but A_2BR does not, despite A_2BR stimulating transcription robustly.

Discussion

The present study addressed the question of whether endocytosis differentiates the functional signaling profiles of natively coexpressed GPCR paralogs. Our results indicate that this is indeed the case, and they provide insight into how signaling selectivity is achieved. Specifically, we show that both β2AR and VIPR1 use agonist-induced endocytosis to prolong activation of the cAMP/PKA cascade, but A_2BR does not. We further show that endocytosis prolongs cAMP/PKA cascade activation mediated by β2AR and VIPR1 differently, as indicated by endocytic inhibition suppressing both the late-phase VIPR1-mediated elevation of global cAMP and PKA cytoplasmic activity but, for β2AR, selectively suppressing the elevation of PKA activity with little effect on cAMP. This distinction is consistent with previous studies of each GPCR individually (21, 37, 43). Importantly, we show here that these fundamental differences in the signaling consequences of endocytosis coexist in precisely the same cell background, can be observed under matched experimental conditions, and impact the ability of natively coexpressed GPCR paralogs to drive downstream transcriptional control through the cAMP/PKA cascade.

Our results suggest a simple cellular scheme for how GPCR-selective differences in cAMP/PKA signaling are determined (Fig. 7). A_2BR stimulates prolonged cAMP production from the plasma membrane (top panels) but little or none from endosomes (bottom panels). VIPR1 stimulates transient cAMP production from the plasma membrane, followed by prolonged production from endosomes. β2AR also stimulates two signaling phases, like VIPR1, but it stimulates cAMP production from endosomes less strongly. We propose that these differences define a receptor-selective spatiotemporal cAMP “code,” defined by quantitative differences in the amount and duration of cAMP production from the plasma membrane and endosomes, that is “decoded” downstream by cellular PKA stores. Prolonged cAMP production stimulated by A_2BR from the plasma membrane drives transcription through a global effect. VIPR1, in contrast, relies on cAMP production from endosomes to produce sufficiently prolonged cascade activation for robust transcriptional induction. β2AR signals similarly to VIPR1, except that it stimulates cAMP production from endosomes less strongly. The relatively weak ability of β2AR to stimulate cAMP production from endosomes is consistent with previous studies indicating that GPCR–Gs coupling on endosomes is limited by receptor residence time in the endosome membrane before partitioning into exit domains that mediate receptor recycling and are presumably signaling inert (44, 45). Furthermore, the ability of a relatively small amount of cAMP production stimulated by β2AR from endosomes to prolong the global PKA elevation is consistent with the close proximity of endosomes containing β2AR and adenylyl cyclase to concentrated internal PKA stores (21, 37, 43).

Figure 7.

Proposed GPCR-selective schemes for cellular signaling through the cAMP/PKA cascade. The present data suggest that NECA stimulates, through native A_2BR, prolonged cAMP production from the plasma membrane (PM) but almost none from endosomes, resulting in a long-lasting elevation of global cAMP that drives downstream transcriptional control robustly. VIP stimulates, through native VIPR1, transient cAMP production from the PM, and a prolonged second phase of cAMP production from endosomes. Iso stimulates, through native β2AR, transient cAMP production from the PM and prolonged production from endosomes, albeit less strongly than VIP because of active β2ARs iteratively cycling and having relatively short residence time in endosomes (as discussed and cited in the text). The relatively small amount of Iso/β2AR-stimulated endosomal cAMP production has little impact on global cAMP, but it enhances downstream PKA activation and subsequent transcription because of the close physical proximity of endosomes to concentrated internal PKA stores (also as discussed and cited in the text). A_2BR, adenosine-2B receptor; β2AR, β2-adrenergic receptor; GPCR, G-protein–coupled receptor; Iso, isoproterenol; NECA, 5′-N-ethylcarboxamidoadenosine; VIP, vasoactive intestinal peptide; VIPR1, vasoactive intestinal peptide receptor-1.

The scheme that we propose comports with the general understanding that cellular signaling is spatiotemporally determined—using amplitude, duration, and location as distinct and relevant variables for determining the downstream response (8). We believe that the present results provide a significant step toward extending this understanding to cAMP–PKA signaling from endosomes, but the present study has limitations, and much more remains to be learned. For example, we focused here on long-term bath application of agonists because it is simple to implement and interpret across agonists and GPCRs. However, the physiological relevance of this application condition is presently not clear. Related to this, the functional significance of endocytosis prolonging the β2AR-induced PKA activity elevation, relative to the upstream cAMP increase, remains to be investigated in a physiological system. Another limitation is that we did not consider the effects of local cAMP buffering and hydrolysis, which are well known to impact cAMP/PKA generally (9, 10) and signaling from endosomes in particular (15), nor did we assess the effects of differences in native expression levels among the GPCRs tested. We note that A_2BR was shown previously to be natively expressed at a significantly higher level (∼12,000 receptors/cell) (46) than either VIPR1 or β2AR (each ∼900 receptor/cell) (46, 47). Thus, it remains to be determined whether the ability of the human A_2BR to robustly drive transcriptional control in the absence of detectable agonist-induced internalization requires it to be endogenously expressed at a relatively high level. A related caveat is that we have not yet investigated whether the ability of endosomal β2AR pools to prolong global PKA activation without detectably incrementing global cAMP is specific to this GPCR (such as if β2AR transits a subset of endosomal membranes in particularly close proximity to internal PKA stores), or if VIPR1 could do so as well if its adenylyl cyclase–stimulating activity in endosomes were reduced to a similar degree. Indeed, the present results are limited to the effects of an experimental manipulation that inhibits endocytosis quite broadly. We anticipate that, in future studies, more selective manipulations of GPCR location, activity, and residence time in defined membrane locations can be implemented to more fully elaborate the spatiotemporal signal encoding–decoding scheme that we first reveal here.

Experimental procedures

Cell culture

HEK293 cells (American Type Culture Collection, CRL1573) were cultured in Dulbecco's modified Eagle's medium (Gibco, 1196511) supplemented with 10% fetal bovine serum (Hyclone, SH30910.03 or R&D Systems, S12495, via UCSF Tissue Culture Facility) and grown at 37 °C with 5% CO₂. Polyclonal Halo-β2AR and inducible monoclonal Halo-VIPR1 stable cell lines have been previously described (19). Polyclonal populations of cells stably expressing Halo-A_2BR were selected with 25 μg/ml zeocin, and polyclonal populations of cells stably expressing FLAG-β2AR or FLAG-A_2BR were selected with 500 μg/ml geneticin.

CRISPR knockout of A_2BR

Cells were electroporated with single guide RNAs (A_2BR: GACACAGGACGCGCTGTACG, NT: GCACUACCAGAGCUAA; Synthego) in complex with Cas9 (UC Berkeley Macrolab) before monoclonal selection. Genetic modifications were verified by Sanger sequencing of PCR amplicons (primers: TGCGTGAGCACCAGCACGAA and GGCAATTTGTTAGTTATCCGCCGC).

Reagents and antibodies

Information about the chemicals and antibodies used in this study can be found in Tables S3 and S4.

DNA constructs and BacMam virus

Details about the plasmids and BacMam viruses used in this study can be found in Table S5. Transfections using Lipofectamine 2000 (ThermoFisher) were carried out 24 to 48 h before experiments according to the manufacturer's protocol. BacMam viruses from pCMV-Dest plasmids were produced according to the manufacturer's protocol (Invitrogen) and added to cells 24 h before experiments.

Fluorescent biosensor assays

Cells treated with cADDis (Montana Molecular, #U0200G), ExRai-AKAR2, ExRai-AKAR2-NES, ExRai-AKAR2-2xNLS, mCherry, and/or mCherry-DynK44E BacMams were plated into 96-well plates (Corning, 3340) coated with poly-l-lysine (MilliporeSigma, P8920). After 24 h, cells were washed twice with assay buffer (20 mM Hepes [pH 7.4], 135 mM NaCl, 5 mM KCl, 0.4 mM MgCl₂, 1.8 mM CaCl₂, and 5 mM d-glucose) before a 10-min incubation in a prewarmed 37 °C plate reader (Spark, Tecan Life Sciences, controlled by SparkControl v3.2). When appropriate, 30 μM Dyngo4a or vehicle was added immediately prior to preincubation. For βAR antagonist experiments, cells were preincubated for 30 min in assay buffer containing antagonist (300 nM GCP 20712, 300 nM ICI 118551, or 10 μM Alp) or vehicle.

For cADDis, fluorescence was read at an excitation wavelength of 500/5 nm and an emission wavelength of 530/10 nm. For ExRai-AKAR2 constructs, fluorescence was read using 400/20 nm and 485/20 nm excitation filters and a 520/10 nm emission filter. mCherry was read using a 560/20 nm excitation filter and a 620/10 nm emission filter. Data were collected every 30 s for 35 min, with vehicle or agonist added at 5 min.

Intracellular cAMP was calculated using raw fluorescence (F), whereas intracellular PKA activity was calculated using the ratio of fluorescence (R = F₄₈₅/F₄₀₀) background corrected by subtracting the average fluorescence of untransduced wells at each timepoint (29). Changes in cAMP (ΔF/F₀) and PKA activity (ΔR/R₀) were calculated as the change in fluorescence (ΔF = F − F₀ or ΔR = R − R₀) normalized to the average of the 5-min baseline (F₀ or R₀). For samples expressing mCherry or mCherry-DynK44E, curves were further normalized to the peak fluorescence change (ΔF/F₀)_max or (ΔR/R₀)_max as noted in the figure legends. Difference curves were calculated by subtracting mCherry-DynK44E curves from mCherry curves. Sustained activity, measured 30 min postagonist addition, was calculated as (ΔF/F₀)₃₀/(ΔF/F₀)_max × 100 or (ΔR/R₀)₃₀/(ΔR/R₀)_max × 100. Integrated signal was calculated as an area under the curve following agonist addition. For dose–response curves, data were fit to a three-parameter dose–response equation in Prism (version 10; GraphPad). When top and bottom plateaus could not be automatically fit, these were constrained to the average maximum signal and zero, respectively.

Internalization by flow cytometry

Cells stably expressing N-terminally tagged GPCRs in 12-well plates were treated with Iso (100 nM or 1 μM), 500 nM VIP, or 20 μM NECA for the indicated times. After three washes with cold PBS, cell surface receptors were labeled with 200 nM JF₆₃₅i-HTL (HaloTagged β2AR, VIPR1, or A_2BR, Fig. 1B) or M1 antibody conjugated to Alexa Fluor 647 (FLAG-tagged β2AR or A2BR, Fig. S3A) for 30 min at 4 °C. FLAG-tagged samples were mechanically lifted, whereas HaloTagged samples were washed once with PBS–EDTA (UCSF Cell Culture Facility) before being lifted with TrypLE Express (Thermo Fisher). Surface staining was measured on an Attune NxT Flow Cytometer equipped with a CytKick Autosampler and controlled by Attune Cytometric Software (version 5.3.2415.0) using a 637 nm excitation laser and a 670/14 nm emission filter. Data were analyzed using FlowJo, v.10.10.0 software (BD Life Sciences). Populations were gated for cells expressing receptors, and internalization was calculated as [1 - (F_t/F_t0)] ∗ 100, where F_t is the median fluorescence at time t. Each biological replicate represents the average of three technical replicates.

Luminescence transcriptional reporter assay

Cells were transfected with pGL4.29[luc2P/CRE/Hygro] (Promega, E8471) and replated into white 96-well plates (Corning, 3917) coated with poly-l-lysine (MilliporeSigma, P8920) at a concentration of 50,000 cells/well. If cells were pretreated with mCherry or mCherry-DynK44E, BacMam virus was added upon replating. After overnight incubation, media were replaced before treatment with vehicle (water or dimethyl sulfoxide), Iso, VIP, or NECA at the indicated concentrations. Cells were incubated at 37 °C with 5% CO₂ for 4.75 h, followed by a 15 min incubation at room temperature. After a wash with room temperature assay buffer (20 mM Hepes [pH 7.4], 135 mM NaCl, 5 mM KCl, 0.4 mM MgCl₂, 1.8 mM CaCl₂, and 5 mM d-glucose), cells were treated with 1.6 mM d-luciferin (Gold Biotechnology LUCNA) in assay buffer. Luminescence was read 2 min later by a plate reader (Spark, Tecan Life Sciences, controlled by SparkControl v3.2). Change in luminescence (ΔLum) was calculated as the ratio of the average luminescence of agonist-treated wells divided by the average luminescence of vehicle-treated wells. Within each biological replicate, changes in luminescence were normalized to either a 10 μM forskolin control or each agonist's maximum response in the mCherry control.

Live-cell imaging

Cells were cotransfected with FLAG-tagged GPCRs and venus-mini-G_s and replated into poly-l-lysine (MilliporeSigma, P8920)–coated glass bottom dishes (Cellvis, D35-20-1.5-N). After 48 h, cells were labeled with M1 antibody conjugated to Alexa Fluor 647 for 10 min at 37 °C, washed three times, and imaged live in Dulbecco's modified Eagle's medium without phenol red (Gibco, 31053028) supplemented with 30 mM Hepes (pH 7.4). Confocal microscopy was carried out on a Nikon Ti inverted microscope controlled by NIS Elements HC v.5.20.02 (Nikon) and fitted with a CSU-22 spinning disk unit (Yokogawa), a custom laser launch (100 mW at 405, 488, 561, and 640 nm; Coherent OBIS), a Sutter emission filter wheel, an Apo TIRF 100×/1.49 numerical aperture oil objective (Nikon), and a Photometrics Evolve Delta EMCCD camera. Cells were kept in a temperature- and humidity-controlled chamber (Okolab) at 37 °C. Cells were imaged for 2 min before the addition of agonist and 20 min after agonist. Images were processed and analyzed using Fiji (version 2.3.0) (48) (https://imagej.net/software/fiji).

Data analysis and reproducibility

Statistical analysis and curve fitting were carried out using Prism (v.10) as noted in the figure legends. All data are shown as individual biological replicates or as mean ± standard deviation from at least three biologically independent experiments, unless otherwise noted. For fluorescent biosensor assays, luminescent transcriptional reporter assays, and flow cytometry internalization assays, each biological replicate represents the average of at least two technical replicates.

Data availability

Data and materials are available upon request.

Supporting information

This article contains supporting information (10 figures and 5 tables) and two additional citations (49, 50).

Conflict of interest

M. v. Z. serves on the Scientific Advisory Board of Deep Apple Therapeutics. The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Henrik Dohlman