Cellular Electrical Impedance as a Method to Decipher CCR7 Signalling and Biased Agonism

The human C-C chemokine receptor type 7 (CCR7) has two endogenous ligands, C-C chemokine ligand 19 (CCL19) and CCL21, displaying biased agonism reflected by a pronounced difference in the level of β-arrestin recruitment. Detecting this preferential activation generally requires the use of separate, pathway-specific label-based assays. In this study, we evaluated an alternative methodology to study CCR7 signalling. Cellular electrical impedance (CEI) is a label-free technology which yields a readout that reflects an integrated cellular response to ligand stimulation. CCR7-expressing HEK293 cells were stimulated with CCL19 or CCL21, which induced distinct impedance profiles with an apparent bias during the desensitisation phase of the response. This discrepancy was mainly modulated by differential β-arrestin recruitment, which shaped the impedance profile but did not seem to contribute to it directly. Pathway deconvolution revealed that Gαi-mediated signalling contributed most to the impedance profile, but Gαq- and Gα12/13-mediated pathways were also involved. To corroborate these results, label-based pathway-specific assays were performed. While CCL19 more potently induced β-arrestin2 recruitment and receptor internalisation than CCL21, both chemokines showed a similar level of Gαi protein activation. Altogether, these findings indicate that CEI is a powerful method to analyse receptor signalling and biased agonism.


Introduction
Chemokine receptors are members of the rhodopsin-like class A G protein-coupled receptors (GPCRs). They regulate immune cell activation and migration, and play a fundamental role in tissue development and organisation [1][2][3]. These functions are elicited by receptor interaction with a specific subset of chemotactic cytokines, namely chemokines. Chemokine-activated receptors, with the exception of atypical chemokine receptors (ACKRs), initiate downstream intracellular signalling mainly through activation of heterotrimeric Gαiβγ proteins that are sensitive to pertussis toxin (PTX) [2][3][4]. Receptor activation leads to G protein-coupled receptor kinase (GRK)-mediated phosphorylation of the receptor's C-terminus, which enables the recruitment of β-arrestins. These multifunctional proteins initiate short-term receptor desensitisation and internalisation, and regulate downstream signalling processes. ACKRs (i.e., ACKR1-4) are devoid of functional G protein coupling but preserved the ability to recruit β-arrestins upon receptor stimulation [5].
Within the human chemokine signalling system, consisting of about 20 chemokine receptors and 50 chemokines, substantial promiscuity exists. Many chemokine receptors interact with more than one chemokine and often a given chemokine can stimulate multiple receptors. Although this was initially seen as signalling redundancy, it is now appreciated that, for several members of the chemokine receptor family, ligand bias occurs naturally,

Results
We investigated the capability of CEI to measure human CCR7 agonist-induced signalling. To this end, HEK293 cells stably expressing CCR7 were seeded in E-plates and an impedance profile was acquired following treatment with either CCL19 or CCL21 ( Figure 1A). Stimulation with either chemokines resulted in an immediate, dose-dependent, and multi-facetted CEI response sharing broad pattern similarities but also clear differences ( Figure 1B). Ligand addition resulted in a brief burst of transient negative cell index (CI) changes. Subsequently, after reaching a global minimum, the CI rose sharply to a maximum and decayed until it stabilised above baseline. At higher agonist concentrations, CCL21 displayed a "head-and-shoulder" profile reaching a second maximum shortly after the first, before initiation of the signal decay. This CCL21-induced shoulder appeared to be delayed and more pronounced with increasing ligand concentrations.
1A). Stimulation with either chemokines resulted in an immediate, dose-dependent, and multi-facetted CEI response sharing broad pattern similarities but also clear differences ( Figure 1B). Ligand addition resulted in a brief burst of transient negative cell index (CI) changes. Subsequently, after reaching a global minimum, the CI rose sharply to a maximum and decayed until it stabilised above baseline. At higher agonist concentrations, CCL21 displayed a "head-and-shoulder" profile reaching a second maximum shortly after the first, before initiation of the signal decay. This CCL21-induced shoulder appeared to be delayed and more pronounced with increasing ligand concentrations. Figure 1. The CCR7 impedance profile. (A) Experimental overview of an impedance profile measurement. Cells were seeded and grown for 20 to 24 h. Thereafter, the growth medium was exchanged for serum-free medium and 4 h later cells were stimulated with agonist. (B) CEI measurements in HEK293 cells stably expressing CCR7 following stimulation with a CCL19 or CCL21 dilution series. Data are represented as the mean (line) and SD (shaded region) of three independent experiments with two technical replicates each. (C) Representation of relevant impedance profile features. The impedance profile is divided into three bins from which the minimal and maximal cell index (CI), and area under the curve (AUC) can be determined. (D-G) Dose-response curves of (D) minimal CI[0-3], (E) the maximal CI [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20], (F) the decay rate and (G) the AUC of the final bin (AUC ). Data are represented as the mean and SD of three independent experiments with two technical Figure 1. The CCR7 impedance profile. (A) Experimental overview of an impedance profile measurement. Cells were seeded and grown for 20 to 24 h. Thereafter, the growth medium was exchanged for serum-free medium and 4 h later cells were stimulated with agonist. (B) CEI measurements in HEK293 cells stably expressing CCR7 following stimulation with a CCL19 or CCL21 dilution series. Data are represented as the mean (line) and SD (shaded region) of three independent experiments with two technical replicates each. (C) Representation of relevant impedance profile features. The impedance profile is divided into three bins from which the minimal and maximal cell index (CI), and area under the curve (AUC) can be determined. (D-G) Dose-response curves of (D) minimal CI [0][1][2][3] , (E) the maximal CI [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] , (F) the decay rate and (G) the AUC of the final bin (AUC  ). Data are represented as the mean and SD of three independent experiments with two technical replicates each. Curves were fitted to a three-parametric non-linear regression model except for (F) which was fit to a four-parametric model. HEK293 cells not expressing CCR7 did not respond to CCL19 addition, which validated that the observed response was indeed CCR7-specific. However, CCL21 did display a small increase in CI over time in these cells (Supplementary Figure S1).
To validate the mechanisms that could explain the similarities and differences in the impedance profile induced by CCL19 and CCL21, we investigated CCR7-mediated signalling and function by employing various label-based assays. To minimize system bias, all these experiments were performed in the same cellular background used for the CEI experiments. First, as a proxy for Gαi activation, agonist-dependent inhibition of forskolininduced cyclic adenosine monophosphate (cAMP) production was measured in HEK293 cells stably expressing CCR7 transiently transfected with a cAMP GloSensor plasmid. Both CCL19 and CCL21 induced robust Gαi activation with similar efficacy and potency ( Figure 4A; Table 2). As expected, performing the assay with CCR7-expressing Gαi-KO cells completely abolished the agonist-induced inhibition ( Figure 5A). Additionally, we assessed the direct activation of different Gαi-subunits using a nanoBRET-based biosensor approach. To this end, we fused Nluc to Gαi-1, Gαi-2 and Gαi-3 at position 91, in accordance with the recent literature [25][26][27], and transiently co-expressed each of them with Gγ2 N-terminally fused with LLS-mKate2 in stable CCR7-expressing HEK293 cells. In this system, CCL19 and CCL21 induced robust G protein activation of all tested Gαi isoforms with similar efficacy and potency (Figure 4D-F; Table 2). in accordance with the recent literature [25][26][27], and transiently co-expressed each of them with Gγ2 N-terminally fused with LLS-mKate2 in stable CCR7-expressing HEK293 cells.
In this system, CCL19 and CCL21 induced robust G protein activation of all tested Gαi isoforms with similar efficacy and potency (Figure 4D-F; Table 2).  To monitor CCR7-β-arrestin2 recruitment, HEK293 cells transiently co-transfected with a NanoBiT complementation system were used, in which CCR7 was C-terminally fused with LgBiT and β-arrestin2 was N-terminally fused with SmBiT. We found that CCL19 was significantly more potent than CCL21 with a 1.17-log potency difference ( Figure 4B; Table 2). Although CCL21-mediated β-arrestin2 recruitment was not saturated at the highest concentration, a theoretical E max value similar to CCL19 could be predicted. When cells were treated with the specific GRK2/3 blocker CMPD101, there was minimal β-arrestin2 recruitment inhibition for both chemokines ( Figure 5B). Furthermore, PTX had no substantial effect on β-arrestin2 recruitment, suggesting that G protein activation is not essential for the recruitment of β-arrestin2 ( Figure 5B). Lastly, since β-arrestins are pivotal in regulating GPCR internalisation, we questioned whether the observed differences in β-arrestin2 recruitment would perpetuate differential receptor internalisation. CCR7expressing cells were exposed to varying concentrations of chemokines for 30 min at 37 • C, after which cell-surface receptor expression was quantified and compared to an unstimulated control. In line with the results obtained with the β-arrestin2 recruitment assay, both ligands induced a similar level of internalisation, but CCL19 was 13.8-fold more potent than CCL21 ( Figure 4C; Table 2). Table 2. Overview of the potency and efficacy of CCR7 ligands in label-based assays. Data are representative of the mean and SD of three independent experiments. The difference between EC50 and Emax was analysed using an unpaired t-test with Welch's correction. **, and *** represent p < 0.01 and 0.001, respectively. ns indicates no significant difference was detected.   To monitor CCR7-β-arrestin2 recruitment, HEK293 cells transiently co-transfected with a NanoBiT complementation system were used, in which CCR7 was C-terminally fused with LgBiT and β-arrestin2 was N-terminally fused with SmBiT. We found that CCL19 was significantly more potent than CCL21 with a 1.17-log potency difference (Figure 4B; Table 2). Although CCL21-mediated β-arrestin2 recruitment was not saturated at the highest concentration, a theoretical Emax value similar to CCL19 could be predicted. The effect of the Gαi-KO (∆Gαi) was analysed against the wild-type cells using an unpaired t-test with Welch's correction. (B) HEK293 cells were transiently transfected with CCR7-LgBiT and β-arrestin2-SmBiT and cells were treated with PTX (50 ng/mL) or CMPD101 (10 µM) prior to ligand (CCL19 (93 nM) or CCL21 (837 nM)) exposure. Changes in bioluminescence were monitored to assess compound effect on ligand-induced β-arrestin recruitment. A one-way ANOVA followed by a Dunnett multiple comparison was used to assess the effect of PTX and CMPD101 compared to the vehicle control set at 100%. (A,B) Data are represented as the mean and SD of three independent experiments with each two or three technical replicates. * and ** represent p < 0.05 and 0.01, respectively.

Discussion
The human CCR7 receptor was previously shown to be differentially activated by its natural ligands, CCL19 and CCL21. Although some conflicting data exist concerning the level of G protein activation induced by both chemokines, it is well established that CCL19-induced CCR7 activation results in more potent β-arrestin recruitment and receptor internalisation compared to CCL21 [19][20][21][22][23]. In this study, we analysed CCR7 signalling using CEI and demonstrate that this label-free technology holds promise as a methodology to investigate and decipher GPCR signalling. The CEI readout is not a priori focused on one particular signalling pathway but reflects a holistic cellular response combining the cellular effect of multiple individual signalling events. Importantly, the CEI response reflected the ligand bias for CCL19 and CCL21 indicating that CEI can detect differential receptor activation. The biological mechanisms behind this ligand-dependent discrepancy between the impedance profile of CCL19 and CCL21 could be explained by results obtained with more conventional label-based assays, which were also performed in HEK293 cells to minimize system bias across the different experimental readouts. Both CCL19 and CCL21 elicited a similar CEI response during the initial response phase, with the exception of a transient negative dip that was more pronounced when CCR7 was stimulated with high concentrations of CCL19. Ligand-induced impedance profiles following the negative impedance changes were dominated by Gαi signalling since they were largely abolished by PTX pre-treatment and genetic loss of Gαi. In line with this observation, the dosedependent increase in maximal CI [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] values for both ligands corresponds well with the results reported in our cAMP production and Gαi-biosensor assay. Early studies reported similar findings [19,20]. Two more recent studies, however, showed that there was a difference in potency, but not efficacy, between CCL19 and CCL21 concerning cAMP modulation [22,23]. In the latter studies, CCR7 signalling was studied in Chinese hamster ovary cells. It is important to note that the receptor interactome, which hinges on the relative abundance of intracellular interaction partners and modulators, might be different in these cells. Moreover, in contrast to our data, Corbisier et al. reported a difference in potency and efficacy between CCL19 and CCL21 in CCR7-expressing HEK293 cells using both a cAMP modulation assay and a direct G protein biosensor approach [21]. Currently, we cannot explain why their results differ from ours.
Agonist responses diverged significantly after reaching the maximal CI [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] . In contrast to CCL19, CCL21 displayed a head-and-shoulder profile, which was more pronounced at higher concentrations, and a slower signal decay was observed. The faster decay observed for CCL19 correlated well with the increased level of β-arrestin recruitment and CCR7 internalisation observed in the label-based assays. Moreover, when CCL19-induced impedance profiles were recorded in β-arrestin1/2-KO HEK293 cells the signal decay was significantly slower. As a result, in this cellular background, the CCL19 profile was more reminiscent of the CCL21 profile. In line with these findings, Watts et al. showed that CXCR3 ligands less effective in recruiting β-arrestin also displayed a prolonged CEI response as well as a more pronounced shoulder [28]. In contrast, when a synthetic βarrestin superagonist (VUF10661) was tested, no shoulder was visible and the return to baseline occurred more quickly. In addition, stimulation of CXCR7, an ACKR known to recruit β-arrestins, but devoid of G protein activation, did not generate detectable CEI responses [29]. Altogether, it appears that β-arrestins alter G protein-mediated impedance profiles, but based on current data, cannot induce impedance changes independently of G protein activation. Hence, analysis of the impedance profiles can enable the quantification of differential G protein and β-arrestin activation by ligands acting on the same receptor. CCL21 strongly interacts with GAGs expressed on the cell surface via its C-terminal extension [30]. It is likely that this interaction is at the root of the slight time-dependent increase in CI seen when challenging wild-type CCR7-negative HEK293 cells with high CCL21 concentrations. This CCR7-independent GAG interaction may counteract a stronger signal decay that would be expected due to the increased β-arrestin recruitment and internalisation occurring at higher CCL21 concentrations. Furthermore, it was recently postulated that the CCL21-GAG interaction might result in the formation of a local reservoir from which CCL21 is released over time, coinciding with persistent but weaker CCL21-mediated signalling [17].
Promiscuous G protein couplings to GPCRs are not uncommon. We found that, although Gαi signalling was the dominant contributor to the CEI responses, Gαq and Gα12/13 also influenced the CEI profile, though to a lesser extent. ROCK kinases, downstream of Gα12/13, are involved in regulating cell morphology and CCR7-mediated migration [31]. Genetic loss of Gα12/13, as well as pharmacological ROCK1/2 inhibition by Y-27632 resulted in the abrogation of the negative transient phase. Furthermore, retraction of the trailing edge, which is attributed to Gα12/13-mediated RhoA activation, seems to require PLCβ-induced calcium mobilisation [32]. A recent study implied the need for a functional and activated Gαq protein to induce calcium release via the Gαi-Gβγ-PLCβmediated pathway [33]. Surprisingly, loss of Gαq, but not its pharmacological inhibition, resulted in an exclusively positive CEI response. It is possible that, despite inhibition, Gαq still retains its functionality in Gαi-mediated calcium release, in contrast to Gαq KO, where Gαq is completely absent.
The presented study describes CCR7 signalling using CEI and label-based assays. We showed that the CCL19 impedance profile is biased towards β-arrestin recruitment compared to CCL21. Similarly, in our label-based assays, CCL19 and CCL21 induced Gαi activation with equal potency and efficacy, but CCL19 was significantly more potent at recruiting β-arrestin2 and inducing internalisation. Furthermore, we demonstrated that CEI is a valuable addition to the GPCR research repertoire. Not only can it discriminate between differential receptor activation, but in concert with pharmacological modulation and KO, CEI provides an opportunity to study receptor signalling from a top-down perspective, allowing identification of individual components that contribute to the overall receptor signalling profile. Since CEI operates without labels, it is also applicable to investigate signalling in primary cells, which is much harder to realise using label-based assays. If throughput increases in the future, we believe that CEI can support initial screenings for biased ligands, specifically due to the integrated nature of the assay. Cellular transfections were performed in suspension per the manufacturer's protocol. Briefly, 1.5 × 10 5 cells per mL were transfected with 0.5 µg plasmid DNA per mL using a 3:1 FuGENE ® HD Transfection Reagent (#E2311, Promega, Madison, MD, USA) to DNA ratio. The FuGENE ® HD Transfection Reagent/DNA mixture contained 20 ng/µL DNA and was incubated for 10 min at ambient temperature before adding it to the cell suspension. When a co-transfection of two plasmids was carried out, half the DNA concentration was used per plasmid.
The transfection setup for nanoBRET-based G protein activation assay differed slightly and is explained below.

Cellular Electrical Impedance Assay
The xCELLigence Real-Time Cell Analyzer (RTCA) DP instrument (Agilent, Santa Clara, CA, USA) was used to measure changes in cellular impedance following ligand stimulation. Briefly, RTCA E-plate VIEW 16 plates with embedded golden electrodes (#300600880, Agilent, Santa Clara, CA, USA) were coated with 10 µg/mL fibronectin (#F2006, Merck, Darmstadt, Germany) for 30 min and air-dried for one hour. A mandatory reference measurement was performed with 50 µL of growth medium per well to establish background CI values for each well. Thereafter, HEK293 cells were seeded at a density of 30,000 cells/well in a final volume of 100 µL. E-plates were placed at room temperature for 15 min and then transferred to the xCELLigence RTCA instrument, located in an incubator at 37 • C and 5% CO 2 . Cellular growth was monitored overnight every 20 min until a steady state was reached after 20-24 h. Following overnight incubation, E-plates were washed with 100 µL serum-free DMEM, 100 µL serum-free DMEM was added to each well and cell stabilisation was measured each minute for 4 h. Before ligand addition, a short normalisation measurement consisting of 5 total measurements, one every 5 s, was performed. Thereafter, 25 µL ligand at a 5× concentration was added and receptor stimulation was measured every 20 s for 4 h. When investigating the effect of compounds (CMPD101, YM-254890 and Y-27632) on receptor stimulation, cells were washed and 80 µL serum-free DMEM was added. Forty minutes before the ligand addition, 20 µL of 5× concentrated compound was added. For PTX, which was incubated overnight, 25 µL of 5× concentrated compound was added two hours after seeding the cells.

β-Arrestin Recruitment Assay
β-arrestin recruitment was monitored using the NanoBiT PPI system (Promega). CCR7 and β-arrestin2 were C-terminally fused with LgBiT and N-terminally fused with SmBiT, respectively, and cloned into a pEF1α-IRES vector. HEK293 cells were transiently co-transfected in suspension with pEF1α-IRES CCR7-LgBiT(C) and pEF1α-IRES SmBiT-ARRβ2(N), as previously described. Transfected cells were seeded at a density of 1.5 × 10 4 cells/well in white, clear flat-bottom 96-well plates coated with 100 µg/mL poly-D-lysin and incubated for 48 h at 37 • C and 5% CO 2 . Next, cells were washed with an assay buffer (HBSS, 20 mM HEPES buffer (#15630-080, Thermo Fisher Scientific, Waltham, MA, USA), 0.5% FBS, pH 7.4) and incubated with 100 µL of a 1:100 Nano-Glo ® Live Cell Substrate (#N2012, Promega, Madison, MD, USA) working solution for 40 min at 37 • C and 5% CO 2 . The plate was transferred to the FLIPR Tetra and baseline luminescence was measured for 30 s every 5 s. Thereafter, 25 µL of 5× ligand was added automatically to the cell plate by the FLIPR Tetra and changes in bioluminescence were monitored in real time for 40 min every 5 s. When required, 20 µL of 5× compound was added to the working solution Nano-Glo Live Cell Substrate. In the case of PTX, 25 µL of 5× concentrated compound was added after 24 h of incubation.

Data Analysis
Raw relative light units (RLU) (cAMP modulation assay and β-arrestin recruitment assay), raw CI values (CEI assay) or a BRET ratio (G protein activation assay) were used as a starting point for data manipulations. BRET ratios were first calculated by dividing acceptor RLU by donor RLU values. All data were normalised to the baseline before ligand addition to reduce inter-well variation, an approach commonly used for kinetic fluorescent and bioluminescent readouts. This baseline was defined as the mean of a 5-point run-in time before ligand addition. Normalisation was performed by dividing all timepoints following ligand addition by the initial baseline. The technical replicates of these normalised readouts were averaged and then background-corrected by subtracting the values of their respective vehicle controls at each timepoint yielding a normalised background-corrected measurement. Dose-response curves were fitted to three parameterslog(agonist) vs. response model in GraphPad V9.3.1 (GraphPad Software, San Diego, CA, USA) unless otherwise stated. Decay rates were fitted from the global maximum with a one-phase decay model in R version 4.0.5 using the SSasymp function from the stats (version 3.6.2) R package. Statistical analysis was performed as described in figure or table legends.