Mechanical and chemical activation of GPR68 probed with a genetically encoded fluorescent reporter

G-protein-coupled receptors (GPCRs) constitute the largest known family of membrane receptors, comprising at least 831 human homologs organized into six functional classes (A to F). They play essential roles in a wide range of biological functions spanning all major physiological systems, including olfaction, energy homeostasis and blood pressure regulation. They also control embryonic development and tissue remodeling in adults. The biological significance of GPCRs is underscored by the fact that ∼13% of all known human GPCRs represent the primary targets of ∼34% of all pharmaceutical interventions approved by the US Food and Drug Administration (Hauser et al., 2018).

GPCRs possess a conserved structure encompassing seven transmembrane helices and switch between resting and active conformations depending on the presence of specific physicochemical stimuli. In addition to recognizing a vast repertoire of small molecules, such as odorants, hormones, cytokines and neurotransmitters, some GPCRs sense physicochemical signals, including photons (Filipek et al., 2003), ions (Strasser et al., 2015), membrane depolarizations (Barchad-Avitzur et al., 2016; Ben-Chaim et al., 2006; Birk et al., 2015; Rinne et al., 2015; Vickery et al., 2016) and mechanical forces (Chachisvilis et al., 2006; Storch et al., 2012; Wei et al., 2018; Xu et al., 2018). Once activated, GPCRs physically interact with heterotrimeric G-proteins (Gα, Gβ and Gγ), promoting the exchange of guanosine diphosphate (GDP) for guanosine triphosphate (GTP) on the Gα subunit. This process, called G-protein engagement, enables the GTP-bound Gα subunit and the Gβγ complex to dissociate from their receptor and activate downstream cellular effectors.

GPCRs often recognize more than one stimulus and interact with one or more Gα proteins among 18 known homologs, enabling them to finely tune downstream biological responses to a complex stimulus landscape (Syrovatkina et al., 2016). One example of a GPCR sensing multiple stimuli and triggering pleiotropic G-protein signaling is GPR68, a class-A GPCR first identified in an ovarian cancer cell line and hence initially named ovarian cancer G-protein-coupled receptor 1 (OGR1) (Xu and Casey, 1996). GPR68 is expressed in a surprisingly large number of tissues (Regard et al., 2008; Xu et al., 2018) and is often upregulated in many types of cancers (Wiley et al., 2019; Xu et al., 2018). Although sphingophosphorylcholine lipids were proposed to act as endogenous GPR68 ligands (Mogi et al., 2005; Xu et al., 2000), it is now well established that GPR68 is physiologically activated by extracellular protons (Ludwig et al., 2003), a property shared with only three other GPCRs to date (GPR4, GPR65 and GPR132). As reported for several GPCRs, GPR68 is also activated by endogenous mechanical stimuli, such as fluid shear stress (FSS) and membrane stretch (Erdogmus et al., 2019; Wei et al., 2018; Xu et al., 2018). Importantly, this mechanosensitivity seemingly enables GPR68 to mediate flow-induced dilation in small-diameter arteries (Xu et al., 2018). GPR68 can engage Gα_q/11, which increases the cytosolic concentration of Ca²⁺ ions {[Ca²⁺]_cyt} through phospholipase C-β (PLC-β; also known as PLCB1), as well as Gα_s, which increases the production of cyclic adenosine monophosphate (cAMP) through adenylate cyclase activation.

The synthetic GPR68 agonist Ogerin increases pH-dependent cAMP production by GPR68 but reduces pH-dependent Ca²⁺ signals, suggesting that Ogerin acts as a biased positive allosteric modulator of GPR68 (Huang et al., 2015). Interestingly, Ogerin suppresses recall in fear conditioning in wild-type but not Gpr68^−/− mice, suggesting a role for GPR68 in learning and memory (Huang et al., 2015). Hence, although the contribution of GPR68 to vascular physiology has been relatively well established, its role in other organs remains unclear.

A genetically encoded fluorescent reporter of GRP68 activation would help bridge this gap. GPCR activation is often monitored using fluorescence resonance energy transfer (FRET) or bioluminescence resonance energy transfer (BRET) (Angers et al., 2000; Chachisvilis et al., 2006; Erdogmus et al., 2019). However, FRET necessitates complex measurements to separate donor and acceptor emissions, whereas BRET often requires long integration times and sensitive detectors to capture faint signals. In contrast, recent reporters engineered by fusing GPCRs with a cyclic permuted green fluorescent protein (cpGFP) have enabled robust and rapid in vitro and in vivo detection of GPCR stimuli, including dopamine (Patriarchi et al., 2018; Sun et al., 2018), acetylcholine (Jing et al., 2018), norepinephrine (Feng et al., 2019) and serotonin (Dong et al., 2021), using simple intensity-based fluorescence measurements. Here, we borrowed a similar cpGFP-based engineering approach to create a genetically encodable reporter of GPR68 stimuli. We call it ‘indicator of GPR68 stimuli by flow and low pH’ (iGlow).

We designed iGlow by borrowing a protein engineering design from previously developed GPCR-based fluorescent reporters containing cpGFP. In cpGFP, the N- and C-termini are relocated to a β-sheet, locally disrupting the integrity of the β-barrel, the main function of which is to shield the proteinogenic chromophore from collisions with solvent molecules. Circular permutation thus makes cpGFP fluorescence sensitive to the conformation of its N- and C-termini (Nasu et al., 2021). Large fluorescence changes are hence anticipated to occur when cpGFP is inserted via its N- and C-termini into the third intracellular loop of GPCRs, a region known to undergo large conformational rearrangement upon stimulus-mediated activation (Rasmussen et al., 2011). We generated iGlow by genetically translocating the cpGFP module taken from the voltage-indicator ASAP1 (St-Pierre et al., 2014), which was originally obtained from GCaMP3 (Tian et al., 2009), and inserting it into the third intracellular loop of human GPR68. We then flanked cpGFP with the genetically optimized N-terminal (LSSLI) and C-terminal (NHDQL) linkers from the dopamine sensor dLight1.2, designed by fusing cpGFP into the third intracellular loop of the dopamine receptor DRD1 (Patriarchi et al., 2018) (Fig. 1A; Fig. S1 and Table S1).

To explore the sensitivity of iGlow to fluid flow, a physiological stimulus of GPR68, we first co-transfected HEK293T cells with two plasmids, one encoding iGlow and the other encoding a cytosolic red fluorescent protein, mCherry. Transfected cells were seeded onto microscope-compatible flow chambers and exposed to FSS by circulating Hank's balanced salt solution (HBSS; pH 7.3) using a computer-controlled peristaltic pump, while green and red fluorescence were recorded simultaneously using split imaging optics mounted on an inverted epifluorescence microscope (flow-induced shear stress was calibrated according to the flow chamber manufacturer's instructions, see Fig. S2). Cells were stimulated with a discontinuous FSS stimulation protocol (10 s on, 10 s off) in which the amplitude of FSS pulses was incrementally increased. This protocol produced robust and transient increases in green, but not red, fluorescence intensity, suggesting that the transient change in green emission intensity did not result from imaging artifacts, which would have equally affected both fluorescence channels (Fig. 1B).

In iGlow, as in other engineered GPCR-based reporters, the cpGFP module is located near the lipid bilayer. One cannot thus exclude the possibility that some molecular components embedded or associated with the cell membrane (e.g. lipids, membrane proteins, cytoskeletal elements) could physically collide with cpGFP when the membrane is exposed to flow. Such hypothetical molecular collisions might induce cpGFP barrel distortions, causing changes in fluorescence emission. If that were the case, all or part of the fluorescence signals observed in Fig. 1B could occur independently of stimulus-induced conformational rearrangements of the linkers connecting GPR68 to cpGFP. One way to determine the contribution of linker-independent cpGFP barrel distortions to flow-induced iGlow fluorescence signals is to evaluate flow-induced fluorescence responses of cpGFP inserted into a non-mechanosensitive membrane protein at a position near the membrane, i.e. similar to how cpGFP in iGlow is located near the lipid bilayer. The presence of mechanically evoked fluorescence signals in such a construct would indicate that iGlow might respond to flow in a linker-independent manner. To test this possible scenario, we tested two cpGFP-containing membrane proteins that are not known (or anticipated) to exhibit mechanosensitivity: the voltage indicator ASAP1 (St-Pierre et al., 2014) and Lck-cpGFP, a cpGFP fused to the myristoylated and palmitoylated N-terminal domain of the lymphocyte-specific kinase (Lck) (Shigetomi et al., 2010). In ASAP1, cpGFP is located extracellularly, whereas in Lck-cpGFP, cpGFP is located intracellularly (Fig. 1C). Nevertheless, in both constructs cpGFP is located in the vicinity of the lipid bilayer and thus these constructs would seem valid to test our hypothesis that cpGFP fluorescence may be modulated by membrane components. Upon application of our intermittent flow protocol, no fluorescence change other than normal photobleaching decay occurred, even upon high flow conditions (>10 dyne cm⁻²; 1 dyne=10^–5 N). These results confirm that the fluorescence signals mediated by iGlow are not likely to be contributed by linker-independent barrel distortion of cpGFP and are rather likely produced as a result of conformational changes propagated from the GPR68 core to the cpGFP chromophore in a linker-dependent manner (Fig. 1D,E).

We next determined the dynamic range of iGlow fluorescence by exposing cells to single FSS pulses ranging from 1.3 dyne cm⁻² to 20.8 dyne cm⁻² (Fig. 1F). We used the amplitude of the fluorescence peaks, calculated as maximal (max) ΔF/F₀, to quantify the amount of iGlow activation. The peak amplitude gradually increases over the tested shear stress range, the largest being a nearly 3-fold increase between 10.4 dyne cm⁻² and 20.8 dyne cm⁻² (∼25–75%), which matches well the dynamic range of GPR68 activation measured with Ca²⁺ imaging (Xu et al., 2018) (Fig. 1G). In addition, the delay between mechanical stimulation and peak fluorescence (time to peak) was negatively correlated with the amplitude of the FSS pulses (Pearson's correlation coefficient=−0.77), with stronger pulses producing shorter time-to-peak values. The relationship between pulse amplitude and time to peak is not linear and was best fitted by an exponential function (R²=0.81, red line in Fig. 1H). Next, we performed a multiple-pulse protocol with 1 min interval between each pulse to measure the repeatability of iGlow signals. Our data indicate that, although subsequent pulses induce subsequent fluorescence peaks, the amplitudes of the subsequent peaks tend to be lower than that of the first one (Fig. 1I).

From our epifluorescence imaging data, we noticed that, in some cells, both baseline iGlow fluorescence (i.e. before stimulation) and FSS-induced signals appear to originate not only from the plasma membrane, but also from within the cell, whereas in other cells, these signals tend to be localized at the cell periphery consistent with cell surface localization (Fig. 1J). We investigated the cellular localization of iGlow with higher spatial resolution using confocal microscopy. To this aim, HEK293T cells were co-transfected with two plasmids, one encoding iGlow and the other encoding a red fluorescent marker of actin (LifeAct-mScarlet), or a red fluorescent protein tagged with a C-terminal CAAX domain (dsRed-CAAX and HcRed-CAAX), which enables trafficking of RAS proteins to cell membranes (Michaelson et al., 2005). Live-cell confocal fluorescence imaging shows that, whereas iGlow baseline fluorescence appears to localize near or at the cell surface, a more diffuse fluorescence can also be seen throughout some cells (Fig. 2).

Next, we investigated the ability of iGlow to activate upon application of the GPR68-selective agonist Ogerin and extracellular acidifications. First, we acutely perfused iGlow-expressing cells with HBSS as a vehicle control, or HBSS containing 10 µM Ogerin or 10 µM non-active Ogerin analog (Huang et al., 2015). Compared to vehicle control, application of Ogerin leads to transient iGlow activation with large max ΔF/F₀ values (Dunn's multiple comparisons test P=0.0067), whereas application of the non-active analog yielded no significant signal (Dunn's multiple comparisons test P>0.9999) (Fig. 3A–D). To assess the optical dynamic range of iGlow, we compared iGlow signals with those produced by cells expressing dLight1.2 and acutely exposed to 10 µM dopamine. As expected, dLight1.2 produces robust responses to dopamine, the maximal amplitudes of which were not significantly different from the responses of iGlow to Ogerin (Dunn's multiple comparisons test P>0.9999). The time-to-peak values of Ogerin-induced iGlow signals were also significantly shorter when stimulating iGlow with 10 µM Ogerin compared to a strong FSS pulse of 20.8 dyne cm⁻² (Mann–Whitney U-test P=0.0307) (Fig. 3E).

We next perfused iGlow-expressing cells with an acidic physiological solution, decreasing extracellular pH from 7.3 to 6.5, or with a vehicle control maintaining a neutral external pH. iGlow produced robust and transient signals that were larger at pH 6.5 (Mann–Whitney U-test P=0.0034) (Fig. 4A,B). In few cells, iGlow also responded to some degree to the vehicle control. We interpret this response by the fact that iGlow might be partially activated by the shear stress produced by rapidly exchanging solution in the culture vessel. To reduce the amount of baseline activation prior measurement, we next increased the external pH to 8.2 (Xu et al., 2018) and applied a single FSS pulse of 2.6 dyne cm⁻² using a physiological solution at pH 8.2 (n=28), 7.3 (n=24) or 6.5 (n=10) (Fig. 4C). Our data show that flow-evoked iGlow signals tend to be larger at pH 6.5 versus pH 8.2 (Tukey's multiple comparisons test P=0.0155) (Fig. 4D,E). In line with the effect of extracellular acidifications on signal amplitude, the mean time-to-peak value was also significantly shorter at pH 6.5 versus pH 8.2 (Tukey's multiple comparisons test P=0.0261) (Fig. 4F).

To test whether iGlow signals depend on specific interactions between GPR68 and regulatory cytoplasmic proteins, we stimulated iGlow with a 2.6 dyne cm⁻² FSS pulse in cells pre-treated with one of several pharmacological agents. We used GTP-γ-S, a non-hydrolyzable GTP analog that prevents Gα protein association with GPCRs; NF449, a GDP→GTP exchange inhibitor that selectively prevents Gα_s dissociation from its receptor (Hohenegger et al., 1998); and BIM-46187, a non-specific GDP→GTP exchange Gα inhibitor (Ayoub et al., 2009). We also tested CMPD101, an inhibitor of G-protein receptor kinases 2/3 (GRK2/3) (Fig. 5A). Most treatments did not significantly change the mean peak fluorescence or the time-to-peak values (Fig. 5B–D). One exception was GTP-γ-S treatment, which imparted a significant increase in mean max ΔF/F₀ from 66±6% (control) to 100±9% (Dunnett's multiple comparisons test P=0.0044).

Unlike its parent receptor DRD1, dLight1.2 cannot trigger downstream G-protein signaling (Patriarchi et al., 2018). Consistent with this study, we found that dLight1.2 signals were not significantly affected by any of our pharmacological treatments (one-way ANOVA P=0.5441) (Fig. S3). To determine whether iGlow exhibits an analogous loss of downstream signaling due to cpGFP insertion, we stimulated cells with extracellular protons (pH 6.5) and measured Ca²⁺ responses by monitoring fluorescence from the red Ca²⁺ indicator JRGECO1a (Dana et al., 2016) (Fig. 5E). Co-transfecting cells with a WT GPR68-encoding plasmid significantly increased Ca²⁺-dependent fluorescent responses (Dunnett's multiple comparisons test P=0.0248), whereas transecting cells with iGlow did not (Dunnett's multiple comparisons test P=0.9943), suggesting a loss of iGlow-mediated downstream signaling through the Gα_q/11 pathway.

We next sought to determine whether iGlow signals depend on the integrity of the actin cytoskeleton. To this aim, we transfected HEK293T cells with LifeAct-mScarlet to monitor real-time actin disorganization upon treatment with 20 µM cytochalasin D (CD), an inhibitor of actin polymerization. Actin filaments were completely disorganized after 20 min (Fig. 6A). Because we observed visible cell death upon 1 h of 20 µM CD treatment, we monitored iGlow's response to FSS immediately after a 20 min CD incubation. At the shear amplitude of 2.6 dyne cm⁻², iGlow produced fluorescence signals similar to those of untreated cells, even upon increasing CD concentration to 50 µM CD (one-way ANOVA P=0.6825) (Fig. 6B,C). These results show that, at this shear stress amplitude, the integrity of the actin cytoskeleton is not required for shear stress sensing by iGlow.

Acute incubation with micromolar concentrations of the small spider toxin GsMTx4 inhibits activation of certain mechanosensitive ion channels by membrane stretch, FSS and mechanical indentation (Alcaino et al., 2017; Bae et al., 2011; Jetta et al., 2019; Li et al., 2019; Suchyna et al., 2004). In addition, GsMTx4 also reduces the activity of the membrane motor prestin (Fang and Iwasa, 2006), an essential component of the cochlear amplifier of outer hair cells (Dallos and Fakler, 2002; Zheng et al., 2000). Given the apparent large inhibition spectrum of GsMTx4 on mechanotransduction membrane proteins, we wondered whether GsMTx4 could also inhibit iGlow. We first performed a positive control experiment by measuring Ca²⁺ entry mediated by the GsMTx4-sensitive mechanosensitive PIEZO1 channel in response to a 2.6 dyne cm⁻² pulse in the presence or absence of 2.5 µM GsMTx4 (Bae et al., 2011). We monitored intracellular free Ca²⁺ ions by co-transfecting PIEZO1-deficient cells (Dubin et al., 2017) with a mouse Piezo1 plasmid and a plasmid encoding the fluorescent Ca²⁺ indicator GCaMP6f (Chen et al., 2013). Our data show that this toxin concentration was able to reduce GCaMP6f fluorescence response (max ΔF/F₀) from +75±5% to +16±2%, a nearly 5-fold reduction (Student's t-test P=9.7×10⁻¹⁸) (Fig. 6D,E). In contrast, the same treatment did not significantly affect the amplitude of iGlow signals induced by a 2.6 dyne cm⁻² pulse (Student's t-test P=0.9116) (Fig. 6F,G).

Class-A GPCRs harbor a structurally conserved amphipathic helical motif located immediately after the seventh transmembrane segment, called helix 8. A recent study showed that deletion of helix 8 abolished mechanical, but not ligand-mediated, activation in the histamine receptor H1R (Erdogmus et al., 2019). Furthermore, transplantation of H1R helix 8 into a mechanoinsensitive GPCR was sufficient to confer mechanosensitivity to the chimeric receptor (Erdogmus et al., 2019). The online tool NetWheels indicates that GPR68 also contains an amphipathic helical motif resembling the helix 8 of H1R (Fig. S4). We introduced a non-sense codon (TGA) to eliminate this motif and the remainder of the C-terminal region from iGlow (Fig. S1) and tested the sensitivity of the deletion mutant, H8Del, to a single FSS pulse of 2.6 dyne cm⁻². At this shear stress amplitude, the mean peak amplitude produced by H8Del was not statistically different from those produced by the full-length iGlow (Student's t-test P=0.3933) (Fig. 6H,I). In addition, the time to peak was similar in both cases (Student's t-test P=0.8808) (Fig. 6J). These results show that helix 8 is not required for shear flow activation by iGlow. Because the C-terminal deletion may have eliminated regulatory sites involved in GPCR desensitization, H8Del could enable repeated stimulations with less signal loss than iGlow. When H8Del was stimulated with a 1.7 dyne cm⁻² pulse every minute, the peak amplitude remained relatively constant for the first four peaks, with less signal loss than iGlow (Fig. 6K,L). To determine whether H8Del senses shear stress when expressed in other cells, we transfected this construct into HEK293T cells and Chinese hamster ovary (CHO-K1) cells and stimulated these cells with a 2.6 dyne cm⁻² pulse of 10 s duration. The mean response of H8Del was not significantly different in both cell types (Student's t-test P=0.2519) (Fig. S5).

We next transiently expressed iGlow in vivo under the control of an astrocyte-specific promoter by stereotaxically injecting adeno-associated viruses (AAVs) into the hippocampal CA3 region of mice. The choice of this injection site was motivated by several studies indicating endogenous expression of GPR68 in the brain (Xu et al., 2018), including the hippocampus (Regard et al., 2008), whereas the choice of the promoter was motivated by the expression of GPR68 in reactive astrocytes (Schneider et al., 2012). Using confocal microscopy on acute hippocampal slices, robust green fluorescence was visible in the CA3 region, whereas adjacent brain regions had minimal fluorescence, suggesting that iGlow could, in the future, be expressed in a spatially defined manner for in vivo detection of GPR68 stimuli (Fig. S6).

This study introduces iGlow, a genetically encoded fluorescent sensor of GPR68 activation. iGlow responds to all currently known activators of GPR68, i.e. FSS, the synthetic agonist Ogerin and extracellular acidifications. More importantly, iGlow does not respond to an inactive Ogerin analog (Huang et al., 2015). Given the high structural similarity between these compounds (the sole difference being the orientation of the benzyl alcohol relative to the triazin ring), this selectivity suggests that iGlow retains the endogenous structure of the GPR68 receptor and the ability of GPR68 to functionally discriminate two structurally similar ligands. In contrast to dLight1.2, the cellular localization of iGlow appears not strictly confined to the plasma membrane. We do not know whether this reflects endogenous trafficking of GPR68 or partial degradation or misfolding due to the presence of cpGFP in iGlow. Alkalinization has been shown to induce internalization of GPR68 in leukocytes (Tan et al., 2018). Such physiological regulation of GPR68 location might, at least in part, contribute to the formation of an intracellular pool of iGlow. Future studies will be needed to investigate this possibility.

Another unexpected property of iGlow is the relatively large cell-to-cell fluctuations of the overall fluorescence time course in response to the same flow stimulus. These fluctuations may reflect the intrinsic heterogeneity of morphologies and mechanical properties of cultured cells, which may lead to heterogeneous flow-induced mechanical stress and thus heterogeneous iGlow activation profiles. Despite these fluctuations, the peak amplitude of iGlow fluorescence signals was, on average, larger and occurred quicker when the amplitude of tested physiological stimuli (shear stress and external protons) was larger.

A potential advantage of iGlow is its insensitivity to pharmacological modulation of G-protein signaling, disruption of actin cytoskeleton and incubation with GsMTx4, a toxin proposed to inhibit certain mechanosensitive ion channels by partitioning into the lipid bilayer, effectively ‘buffering’ the effect of shear stress and/or membrane stretch (Gnanasambandam et al., 2017). As in dLight1.2, the insertion of the cpGFP module near the G-protein binding site apparently uncouples iGlow from G-protein signaling, allowing the reporter to function as an ‘observer’ that does not perturb endogenous signaling. However, the lack of signal elimination by CD and GsMTx4 treatments is more puzzling because numerous mechanosensitive ion channels show at least partial reduction of mechanosensitivity in response to these treatments (Alcaino et al., 2017; Bae et al., 2011; Gottlieb et al., 2012; Hurst et al., 2009; Jia et al., 2016; Kamaraju et al., 2010; Nishizawa and Nishizawa, 2007; Ostrow et al., 2003; Shen et al., 2015).

To date, mechanosensitivity has been reported in at least one class-B GPCR (parathyroid hormone type 1 receptor) (Zhang et al., 2009) and many class-A subfamilies including A3 (bradykinin receptor B2, apelin receptor and angiotensin II type 1 receptor) (Chachisvilis et al., 2006; Kwon et al., 2016; Mederos y Schnitzler et al., 2008), A6 (vasopressin receptor 1A) (Mederos y Schnitzler et al., 2008), A13 (sphingosine receptor 1) (Jung et al., 2012), A15 (GPR68) (Wei et al., 2018; Xu et al., 2018), A17 (dopamine receptor DRD5) (Abdul-Majeed and Nauli, 2011) and A18 (muscarinic receptor M5R and histamine receptor H1R) (Erdogmus et al., 2019; Mederos y Schnitzler et al., 2008). Helix 8 is both necessary and sufficient to confer mechanosensitivity in certain class-A GPCRs such as H1R (Erdogmus et al., 2019). However, helix 8 is not necessary for flow-induced iGlow activation. In addition, although mechanical activation is independent from ligand-mediated activation in some class-A GPCRs (Erdogmus et al., 2019), the deletion of five histidine residues (H17, H20, H84, H169 and H269) abrogates both pH sensing and flow sensing in GPR68, suggesting that the binding of protons is required for GPR68 mechanosensitivity (Ludwig et al., 2003; Xu et al., 2018). Hence, GPR68 and iGlow may sense both stimuli using a non-canonical helix 8-independent pathway. Further investigations will be necessary to identify the underlying mechanisms of mechanosensitivity.

To conclude, iGlow probes GPR68 activation by endogenous and exogenous stimuli and should be useful to determine the biological roles of GPR68 in vascular and non-vascular physiology, such as hippocampal plasticity.

A fragment containing the human GPR68 cDNA was obtained by digesting a pBFRT-GPR68 plasmid (a gift from Drs Mikhail Shapiro, Caltech and Ardèm Patapoutian, Scripps Research) by NdeI and BamHI. The insert was ligated into an in-house pCDNA3.1-Lck-GCaMP6f plasmid linearized by the same enzymes, creating the plasmid pCDNA3.1-GPR68. A cpGFP cassette was amplified by PCR from a pCDNA3.1 plasmid encoding ASAP1 [Addgene #52519, deposited by Dr Michael Lin (St-Pierre et al., 2014)] and inserted into pCDNA3.1-GPR68 using the NEBuilder HiFi DNA Assembly kit (New England Biolabs). The pCNDA3.1-jRGECO1a plasmid was cloned by assembling PCR-amplified fragments from pGP-CMV-NES-jRGECO1a [Addgene #61563, deposited by Dr Douglas Kim (Dana et al., 2016)] and pCDNA3.1. All constructs were confirmed by Sanger sequencing (GENEWIZ). The pLifeAct-mScarlet-N1 plasmid was obtained from Addgene [#85054, deposited by Dr Dorus Gadella (Bindels et al., 2017)]. The DsRed-CAAX and HcRed-CAAX plasmids were a gift from Dr Bradley Andersen (Western University of Health Sciences, Pomona, CA, USA). All molecular biology reagents were purchased from New England Biolabs.

HEK293T and CHO-K1 cells were obtained from the American Type Culture Collection and ΔPZ1 cells were a gift from Ardèm Patapoutian (Scripps Research). Cells were not recently authenticated or tested for contamination. Cells were cultured in standard conditions (37°C, 5% CO₂) in a Dulbecco's modified Eagle medium supplemented with penicillin (100 U ml⁻¹), streptomycin (0.1 mg ml⁻¹), 10% sterile fetal bovine serum, 1× minimum essential medium non-essential amino-acids and without L-glutamine. All cell culture products were purchased from Sigma-Aldrich. Plasmids were transfected in cells (passage number <35) seeded in 96-well plates at ∼50% confluence 2–4 days before the experiment with FuGene6 (Promega) or Lipofectamine 2000 (Thermo Fisher Scientific) and following the manufacturer's instructions. Then, 1–2 days before experiments, cells were gently detached by 5 min incubation with phosphate-buffered saline and re-seeded onto 18 mm round glass coverslips (Warner Instruments) or onto disposable flow chambers (Ibidi µ-slides 0.4 mm height), both coated with Matrigel (Corning). Cells were treated with each drug at 15 min (CMPD101), 20 min (CD and NF449), 30 min (GTP-gamma-S and GsMTx4) or 2 h (BIM-46187) prior to measurement. pH-shear experiments were performed using a starting pH of 8.2, adjusted 15 min prior to measurement. CMPD101 (#5642) and NF-449 (#1391) were purchased from R&D Systems (Biotechne); GTP-gamma-S was purchased from Cytoskeleton (#BS01); dopamine (#H8502) and Gαq inhibitor BIM-46187 (#5332990001) were purchased from Sigma-Aldrich.

Excitation light of desired wavelengths were produced by a light-emitting diode light engine (Spectra X, Lumencor), cleaned through individual single-band excitation filters (Semrock) and sent to the illumination port of an inverted fluorescence microscope (IX73, Olympus) by a liquid guide light. Excitation light was reflected towards a plan super apochromatic 100× oil-immersion objective with a 1.4 numerical aperture (Olympus) using a triple-band dichroic mirror (FF403/497/574, Semrock). Emission light from the sample was filtered through a triple-band emission filter (FF01-433/517/613, Semrock) and sent through beam-splitting optics (W-View Gemini, Hamamatsu). Split and unsplit fluorescence images were collected by a sCMOS camera (Zyla 4.2, ANDOR, Oxford Instruments). Spectral separation by the Gemini was done using flat imaging dichroic mirrors and appropriate emission filters (Semrock). Images were collected by the Solis software (ANDOR, Oxford Instruments) at a rate of 1 frame s⁻¹ through a 10-tap camera link computer interface. Image acquisition and sample illumination were synchronized using TTL triggers digitally generated by the Clampex software (Molecular Devices). To reduce light-induced bleaching, samples were only illuminated for 200 ms, i.e. during frame acquisition (200 ms exposure). To reduce auto-fluorescence, the cell culture medium was replaced with Phenol Red-free HBSS ∼20 min prior to experiments. Owing to the narrow field of view at this magnification, only one to three transfected cells were measured per flow assay.

A MATLAB script (available to download at Open Science Framework, see below) was used to calculate the average fluorescence intensity of each cell of interest at each frame, expressed as percentile above the fluorescence at t=0 (‘deltaF/F’). Prior to analysis, photobleaching was corrected using the exponential fit method, and a ‘mask’ (.jpg format) separating cells of interest from the background was manually drawn in ImageJ. This mask, in addition to a bleach-corrected image sequence (.tif format) corresponding to each frame of the video, is read by the MATLAB script. The output of the script is a text file containing deltaF/F values at each frame for each cell in the mask (bleachcorrected.txt, rows and columns correspond to frame ID and individual cells, respectively), and a figure showing the original image, the mask and deltaF/F traces for each cell (figure.jpg). Backgrounds are either removed manually in ImageJ, or within the MATLAB script by taking the average fluorescence intensity of all non-cell pixels and subtracting this value from the intensity values associated with cells at each frame.

All animal procedures followed the Institutional Animal Care and Use Committee of the Western University of Health Sciences. Brain-optimized AAVs PHP.eB (Chan et al., 2017) (AAV-PHP.eB) harboring iGlow under transcriptional control by the GFAP promoter (GFAP-iGlow-WPRE AAV-PHP.eB) were produced by VectorBuilder. To express iGlow in the CA1 pyramidal fields of the hippocampus, high titers of GFAP-iGlow-WPRE AAV-PHP.eB virus (0.7 µl, 1×10¹³ genome copies per ml) were stereotaxically injected into the hippocampal CA1 sub region of 3-month-old male C57Bl/6N mice through a glass micropipette at four sites at the following coordinates relative to bregma: anterior posterior (AP), −1.8 mm; medial lateral (ML), ±0.8 mm; dorsal ventral (DV), −1.6 mm; or AP, −2.5 mm; ML, ±2 mm; DV, −1.6 mm. After injection, the micropipette was left in place for an additional 5 min to ensure full virus diffusion. After surgery, mice were treated with antibiotics and their health was monitored every day for 2 weeks. For brain slice imaging, mice were euthanized by isoflurane inhalation and their brains dissected and chilled in ice-cold artificial cerebrospinal fluid (aCSF) containing the following: 124 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 24 mM NaHCO₃, 2 mM ascorbic acid, 10 mM glucose, 1.5 mM MgSO₄ and 2.5 mM CaCl₂, bubbled with 95% O₂ and 5% CO₂. Hippocampal slices (0.35 mm thick) were obtained using a McIlwain tissue chopper as previously described (Zhou and Baudry, 2006). After isolation, hippocampal slices were placed in incubation baskets in aCSF saturated with 95% O2–5% CO₂ and incubated for a 1 h recovery period at room temperature. Slices were next transferred to Petri dishes and placed under a confocal microscope (Zeiss LSM 880). Standard GFP settings were used to image iGlow.

FSS stimulation was done by circulating extracellular physiological solutions at various speeds into µ-slide channels (Ibidi) using a Clampex-controlled peristaltic pump (Golander). In order to accurately determine the amplitude of shear stress applied inside the flow chambers, we compared shear stress values determined by multiplying the average flow rate by a coefficient provided by the manufacturer (see Fig. S2).

The number n represents the number of independent cells or cell clusters analyzed. To evaluate pairwise differences between mean data sets, we performed Mann–Whitney U-tests when n≤10 and Student's t-tests when n>10 in both data sets. For comparing means of more than two groups, Dunnett's or Tukey's multiple comparisons tests were performed following one-way ANOVA when n≥10 in each group, whereas Dunn's multiple comparisons tests were performed following one-way Kruskal–Wallis analysis of variance when n<10 in at least one group. All error bars are s.e.m. Statistical tests were performed using OriginPro 2018, GraphPad Prism 8.2 or a Mann–Whitney online calculator.

We thank Ardèm Patapoutian for the gift of human GPR68 cDNA and Bradley Andresen for help with confocal imaging.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.255455