*2.3. Non-Peptide Opioids Elicit Different Intracellular Ca2+ Signalling Dynamics*

Time-lapse CLSM imaging of intracellular Ca2+ levels using the cell-permeant Fura Red ratiometric dye (Figure 4A), showed that stimulation with equimolar concentrations of different non-peptide opioids acutely induced different changes in the intracellular

Ca2+ levels in HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato (Figure 4B). In untreated cells, stationary state intracellular Ca2+ levels were observed. Following the addition of 750 nM morphine, the stationary state appeared to have lost its stability and sinusoidal oscillations in Ca2+ levels with smoothly increasing amplitudes and a period of about 5 min, emerged. Treatment with 750 nM codeine also induced oscillations in intracellular Ca2+ levels. However, these oscillations showed features of so-called relaxation oscillations [32], which are characterized by a relatively long relaxation period during which the system remained in a stationary state, alternating with a short period in which the abrupt decrease in fluorescence intensity, i.e., the increase in intracellular Ca2+ level was observed. Treatment with 750 nM fentanyl did not cause any oscillations in intracellular Ca2+ levels, but a four-fold increase in Fura Red fluorescence intensity was noted, indicating that intracellular Ca2+ levels decreased markedly following the addition of fentanyl. Finally, treatment with 750 nM oxycodone induced small-amplitude relaxation oscillations with gradually increasing amplitudes over a period of about 5 min. Of note, while the time series shown in Figure 4 was recorded in individual cells, the dynamic behaviour is representative, as it is most often encountered in the analysed population of cells. Molecules 2022, 27, x FOR PEER REVIEW 7 of 21 treatment with 750 nM oxycodone induced small-amplitude relaxation oscillations with gradually increasing amplitudes over a period of about 5 min. Of note, while the time series shown in Figure 4 was recorded in individual cells, the dynamic behaviour is representative, as it is most often encountered in the analysed population of cells.

Figure 4. Stimulation with non-peptide opioids causes different intracellular Ca2+ signalling dynamics in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. (A) CLSM time-lapse imaging of Ca2+ levels (Fura Red, dark violet) in HEK293 cells expressing MOP-eGFP (green) and 5-HT1A-Tomato (red) after 30 min treatment with 750 nM oxycodone. White arrows indicate oscillatory changes in Fura Red fluorescence intensity, where a transient decrease in fluorescence intensity reflects an increase in the concentration of Ca2+ ions. (B) Fluctuations in Fura Red fluorescence intensity over time following treatment of HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato with equimolar concentrations of different opioids. **Figure 4.** Stimulation with non-peptide opioids causes different intracellular Ca2+ signalling dynamics in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. (**A**) CLSM time-lapse imaging of Ca2+ levels (Fura Red, dark violet) in HEK293 cells expressing MOP-eGFP (green) and 5-HT1A-Tomato (red) after 30 min treatment with 750 nM oxycodone. White arrows indicate oscillatory changes in Fura Red fluorescence intensity, where a transient decrease in fluorescence intensity reflects an increase in the concentration of Ca2+ ions. (**B**) Fluctuations in Fura Red fluorescence intensity over time following treatment of HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato with equimolar concentrations of different opioids.

#### 2.4. Non-Peptide Opioids Differ in the Extent to Which They Activate Major Signal Transduction Pathways *2.4. Non-Peptide Opioids Differ in the Extent to Which They Activate Major Signal Transduction Pathways*

deine, or fentanyl when compared to untreated cells (Figure 5).

In order to assess the downstream effects of non-peptide opioids in HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato, phosphorylation of MAPKs ERK1/2 and p38 was probed because MOP activation was shown to trigger the phosphorylation of both ERK1/2 [33] and p38 [34]. Western blot analysis showed an increase in phosphorylated ERK1/2 and p38 in cells that had been treated with 750 nM of morphine, oxycodone, co-In order to assess the downstream effects of non-peptide opioids in HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato, phosphorylation of MAPKs ERK1/2 and p38 was probed because MOP activation was shown to trigger the phosphorylation of both ERK1/2 [33] and p38 [34]. Western blot analysis showed an increase in phosphorylated ERK1/2 and p38 in cells that had been treated with 750 nM of morphine, oxycodone, codeine, or fentanyl when compared to untreated cells (Figure 5).

Figure 5. Opioids differ in their capacity to activate extracellular signal-regulated kinase (ERK) and p38 mitogen-activated protein kinase (MAPK) signalling pathways in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. Top: HRP chemiluminescence images of western blotting

deine, or fentanyl when compared to untreated cells (Figure 5).

equimolar concentrations of different opioids.

Transduction Pathways

Figure 5. Opioids differ in their capacity to activate extracellular signal-regulated kinase (ERK) and p38 mitogen-activated protein kinase (MAPK) signalling pathways in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. Top: HRP chemiluminescence images of western blotting **Figure 5.** Opioids differ in their capacity to activate extracellular signal-regulated kinase (ERK) and p38 mitogen-activated protein kinase (MAPK) signalling pathways in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. Top: HRP chemiluminescence images of western blotting membranes showing different levels p-ERK1/2 (**A**) and p-p38 (**B**) following 18 h treatment using equimolar concentrations of different opioids. Bottom: Protein level of phosphorylated-Erk1/2 relative to β-actin (**A**) and the protein level of phosphorylated-p38 relative to β-actin (**B**), following 18 h treatment with 750 nM of morphine (red), codeine (green), oxycodone (blue) or fentanyl (magenta), as compared to untreated cells (grey). Statistical analysis: Paired *t*-test. Experiments were carried out in triplicate (see Supplementary Materials, Section S4. Western blotting (Figure S5)).

treatment with 750 nM oxycodone induced small-amplitude relaxation oscillations with gradually increasing amplitudes over a period of about 5 min. Of note, while the time series shown in Figure 4 was recorded in individual cells, the dynamic behaviour is rep-

Figure 4. Stimulation with non-peptide opioids causes different intracellular Ca2+ signalling dynamics in HEK293 cells stably expressing MOP-eGFP and 5-HT1A-Tomato. (A) CLSM time-lapse imaging of Ca2+ levels (Fura Red, dark violet) in HEK293 cells expressing MOP-eGFP (green) and 5-HT1A-Tomato (red) after 30 min treatment with 750 nM oxycodone. White arrows indicate oscillatory changes in Fura Red fluorescence intensity, where a transient decrease in fluorescence intensity reflects an increase in the concentration of Ca2+ ions. (B) Fluctuations in Fura Red fluorescence intensity over time following treatment of HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato with

In order to assess the downstream effects of non-peptide opioids in HEK293 cells expressing MOP-eGFP and 5-HT1A-Tomato, phosphorylation of MAPKs ERK1/2 and p38 was probed because MOP activation was shown to trigger the phosphorylation of both ERK1/2 [33] and p38 [34]. Western blot analysis showed an increase in phosphorylated ERK1/2 and p38 in cells that had been treated with 750 nM of morphine, oxycodone, co-

2.4. Non-Peptide Opioids Differ in the Extent to Which They Activate Major Signal

resentative, as it is most often encountered in the analysed population of cells.

Fentanyl elicited the strongest ERK1/2 activation (mean = 1.156, SD = 0.183, *<sup>p</sup>* = 8.52 <sup>×</sup> <sup>10</sup>−<sup>4</sup> ), unlike oxycodone (mean = 0.506, SD = 0.139, N.S). In contrast, oxycodone elicited the strongest p38 activation (mean = 1.441, SD = 0.517, *p* = 0.025), while the effects of fentanyl, morphine, and codeine were similar. Interestingly, LC-MS/MS metabolite analysis indicated that these effects are likely attributed to the primary non-peptide opioid compounds in their own right, as there were no common opioid metabolites detected either in the cell culture medium or in the cell lysate (Supplementary Materials, Section S6, Table S2).

#### **3. Discussion**

Advanced fluorescence microscopy-based techniques allow us to quantitatively characterize molecular interactions in live cells and bring about new understanding of dynamical processes that underlie complex biological functions [35–37]. They also enable us to test with unprecedented precision new mechanistic hypotheses. In this study, FCCS, a quantitative time-resolved analytical method with single-molecule sensitivity, was used to examine in live cells the hypothesis that prolonged exposure to non-peptide opioids promotes heterodimer formation between MOP and 5-HT1A. This hypothesis, derived from preclinical [6–8] and clinical studies [11,12,20], further asserts that altered cellular signalling due to receptor heterodimer formation may contribute to neuroplastic changes that, eventually, lead to sensitization of pronociceptive pathways at the organism level.

To test the initial statement in this hypothesis, FCCS was used to quantitatively characterize in live cells interactions between MOP and 5-HT1A receptors and the effects of some of the most commonly used non-peptide opioid drugs: morphine, oxycodone, codeine and fentanyl, on the extent of these interactions. The CLSM imaging, biochemical assays and LC-MS/MS were used to assess the downstream consequences of these interactions. The most important results are summarized in Table 1 and discussed below.

We found that MOP and 5-HT1A receptors associate in the plasma membrane (Figures 1B, 2A and 3A), building heterodimer complexes characterized by an apparent dissociation constant, K app d = (440 ± 70) nM. This result, obtained nondestructively in live cells, confirms the findings by Cussac et al. who have shown using co-immunoprecipitation

and Bioluminescence Resonance Energy Transfer (BRET) that functional MOP and 5-HT1A heterodimers are formed in overexpressing cells [24]. We have verified this finding in cells expressing physiologically relevant levels of the investigated receptors and determined the apparent dissociation constants for MOP–5-HT1A heterodimers in live cells, K app d = (440 ± 70) nM. Moreover, in agreement with the results obtained by Cussac et al. [24], we have also observed that DAMGO induces prominent MOP internalization but not the internalization of MOP–5-HT1A heterodimer complexes, whereas the non-peptide opioids did not cause internalization, neither of individual receptors, nor of heterodimer MOP–5-HT1A complexes (Figure 2).

**Table 1.** Equimolar concentrations of non-peptide opioids differently affect the apparent dissociation constant of MOP–5-HT1A, fluorophore brightness, ERK1/2 and p38 activation, and Ca2+ levels and signalling dynamics.


Prolonged exposure to all opioids tested facilitated heterodimer formation between MOP and 5-HT1A receptors, albeit to a different extent (Figure 3), differently altered intracellular Ca2+ levels and signalling dynamics (Figure 4) and activated ERK1/2 and p38 signal transduction pathways to a different extent (Figure 5). Fentanyl, the most potent off all non-peptide opioids tested here, exhibited in the concentration range 50–750 nM, a dose-dependent effect on MOP–5-HT1A heterodimer formation (Figure 3A,B) and stabilized significantly the heterodimer complexes, as evident from the five-fold decrease in the apparent dissociation constant from K app d = (440 ± 70) nM in untreated cells, to K app d, Fentanyl = (80 ± 70) nM in cells treated with 750 nM fentanyl. It also elicited the highest activation of the ERK1/2 and a comparably strong activation of the p38 (Figure 5). Finally, fentanyl caused an acute decrease in Ca2+ levels, as evident from the pronounced increase in Fura Red fluorescence (Figure 4, Table 1), which is in line with previously reported findings [38]. In contrast, oxycodone elicited the weakest stabilizing effect on MOP–5-HT1A heterodimer formation, as evident from the two-fold reduction in K app d, Oxycodone = (200 ± 70) nM (Figure 3C) and elicited the highest activation of the p38 (Figure 5B), while causing an insignificant activation of the ERK1/2 (Figure 5A, Table 1). Treatment with 750 nM oxycodone did not significantly affect Ca2+ signalling dynamics, although a small reduction in Ca2+ level and the appearance of small-amplitude relaxation oscillations were noted. The very strong activation of the p38 observed in our study is in line with recent findings in rats showing increased p38 activity during chronic oxycodone exposure [39]. p38 activation may also be relevant for the aversive, addictive effects of oxycodone—p38 activation was shown to underlie opioid reward behaviour in mice [40] and the kappa opioid receptor (KOP)-induced p38 activation has been shown to reinstate drug seeking behavior in mice [41]. Based on this, a recent study argued that the addictive qualities of oxycodone outweighed its benefits as a prescription drug [42].

Morphine and codeine showed significant differences in their potency to stabilize MOP–5-HT1A heterodimer complexes, with codeine eliciting a higher stabilizing effect than morphine, K app d,Morphine = (200 ± 70) nM and K app d, Codeine = (100 ± 70) nM (Figure 3C, Table 1). Codeine also elicited more dramatic effects on intracellular Ca2+ signalling, reducing to a larger extent intracellular Ca2+ levels and causing more dramatic changes in Ca2+ signalling dynamics than morphine (Figure 4B). However, they activated the

ERK1/2 and p38 signalling pathways to a similar extent. The unexpected observation that codeine more strongly stabilized MOP and 5-HT1A heterodimer complexes than morphine is contrary to the general view that codeine is an inactive prodrug with a low affinity for MOP, the effect of which is obtained first after its metabolic conversion to morphine [43–45] and dihydrocodeine-6-O-gluconoride [46,47]. To interrogate this further, an LC-MS/MS analysis was deployed. The LC-MS/MS showed that the concentration of codeine metabolites in the cell culture medium and the cell lysate, if present at all, is below the detection limit of the applied method (Supplementary Materials, Section S6, Table S1). This finding is in line with the fact that HEK293 cells do not express the CYP2D6 gene (https://www.proteinatlas.org/ENSG00000100197-CYP2D6/cell, 3 February 2022), which is crucial for metabolizing codeine [48]. Taken together, our data indicate that codeine is an active compound in its own right. Recent studies showed that codeine has a 6-fold higher permeability and crosses the plasma membrane faster than morphine [49], which could potentially explain the strong response elicited in our cell model. This finding suggests that the pharmacodynamics of codeine is not yet fully elucidated and warrants further studies.

Moreover, FCCS analysis revealed that all non-opioid peptides tested nearly doubled eGFP brightness, while Tomato brightness was not affected to the same extent and treatment with oxycodone and codeine did not alter Tomato brightness (Table 1). The following processes: homodimerization of MOP and, to a lesser extent, of 5-HT1A; homo- and heterooligomerization of MOP and 5-HT1A; and changes in fluorescence lifetime of eGFP and Tomato due to intracellular environment changes caused by signal transduction, can independently or jointly increase the brightness of eGFP/Tomato. Further studies are, however, needed to distinguish the contribution of these possible mechanisms. Most notably, stringent number and brightness analysis and fluorescence lifetime measurements would be needed to discern the contribution of one effect from the other. Having said this, we point out that changes in brightness consistent with the presence of higher order oligomers were not observed.

The possibility to quantitatively characterize MOP/5-HT1A interactions in live cells is a significant achievement of great general interest—the stability of G protein-coupled receptor (GPCR) homo/heterodimer complexes is measured in live cells for a handful of GPCRs, see for example [34–36,50], even though it is well recognized that numerous GPCRs form homo- and heterodimers and that these interactions are important targets for drug development [51]. However, some limitations of our approach are inevitably present and warrant further discussion. Most notably, FCS/FCCS cannot detect endogenous nonfluorescent receptors, receptor constructs with irreversibly photobleached fluorophores or with fluorophores residing for various reasons in dark states. This affects the actual value of the apparent dissociation constants. However, in the context of our study, this limitation does not affect the conclusions of our work, since relative differences are analysed.

Another limitation of our study is that the work was performed using transfected cells that express the proteins of interest through powerful promoters, which may lead to artefacts due to over-expression. To mitigate this risk, we have generated stably transformed cell lines—it is commonly known that stably transformed cells do not yield as high expression as transiently transfected cells. Besides, we have selected for our analysis cells expressing low levels of MOP-eGFP and 5-HT1A-Tomato—the average number of molecules in the OVE of NMOP = (27 ± 6) and N5-HT1A = (25 ± 3) corresponds to a surface density of about (130 <sup>±</sup> 10) molecules/µm<sup>2</sup> . For comparison, many studies of GPCRs class A show that the average surface density of endogenous GPCRs is typically low, <5 molecules/µm<sup>2</sup> , in healthy but increase severalfold in disease states—a recent study showed endogenous MOP levels of 4 molecules/µm<sup>2</sup> [52]. However, as cautioned by the authors, one needs to bear in mind that this value may be underestimated due to low antibody binding efficiency—theoretical studies show that at GPCRs surface densities < 5 molecules/µm<sup>2</sup> the receptors may be too far apart from each other to allow for the efficient build-up of Gβγ to concentrations needed to modulate the activity of other intracellular proteins and show

that G protein signalling occurs within nanodomains where the local density of GPCRs is easily > 50 molecules/µm<sup>2</sup> [53].

Furthermore, an important limitation of our study is that the effects of one concentration of non-peptide opioids is tested. A dose-response analysis is needed to characterize cellular responses to varied amounts of the selected non-peptide opioids and stringent control experiments are needed to examine to what extent the observed effects are mediated via the monomeric fraction of the receptor pool and what the contribution of the receptor heterodimer is. In the light of our work, it is important to point out that while the affinity of the tested non-peptide opioids for binding to MOP is in the range 1–750 nM [54] and to 5-HT1A in the 2–20 µM range [55], pharmacologically relevant concentrations of non-peptide opioids are considerably higher [56]. For example, in opioid-naive postoperative patients, an analgesic effect of fentanyl is achieved at the lowest blood plasma fentanyl concentration levels of about 1.8–4.4 nM (0.6–1.5 ng/mL) [57]. However, much higher concentrations were measured in cancer patients treated for pain; on average 530 nM (178 ng/mL) [56,58]. The concentrations used in our study are therefore in the pharmacologically relevant range. Moreover, in this study we chose to study equimolar concentrations of opioids, rather than equipotent concentrations. Although there are several conversion tables for opioid potency, they are perceived as unreliable [59]. The few studies that have addressed opioid equianalgesic dose/potency ratios are heterogeneous with respect to size, subjects, specific aims, settings, and study method [60]. Thus, Rennick et al. have concluded from their findings that there is no true universal way to accurately perform equianalgesic conversions for opioids [60]. Given that the aim of our work was to assess the effect of non-peptide opioids on the extent of MOP-eGFP and 5-HT1A-Tomato association in live cells, the study design where cells with similar receptor surface density levels are used and the effects of equimolar concentrations of non-peptide opioids are compared is correct. It may, however, be interesting to examine in the future the effect of equipotent concentrations of non-peptide opioids, determined with regard to a quantifiable effect, such as the ability to alter intracellular Ca2+, ERK1/2 or p38 phosphorylation levels.

Furthermore, the treatment time length is an important variable. We have chosen 18 h, taking into consideration the cell doubling time, which is under the condition of our experiments ~36 h for HEK293 cells and ~72 h for PC12 cells. In this way, the cells were exposed to treatment for a considerable fraction, 0.25–0.50, of their cycle time and the effect of the number of divisions during the course of an assay is small [61]. In future studies it may, however, be interesting to examine the effect of treatment time length on MOP-eGFP and 5-HT1A-Tomato association in live cells in order to understand the relevance of this phenomenon in acute vs chronic treatment with non-peptide opioids.

Finally, we have not used in our study antagonists of MOP and 5-HT1A receptors to block effects mediated via monomeric receptors. Consequently, we cannot discern to what extent Ca2+ level and dynamics, and ERK1/2 or p38 phosphorylation levels change via heterodimer-mediated pathways and whether these effects can also be blocked by the selective antagonists of MOP and 5-HT1A.

#### **4. Materials and Methods**

#### *4.1. Cell Culture and Transfection*

Human embryonic kidney (HEK293) and rat phaeochromocytoma (PC12) cell lines (American Type Culture Collection (ATCC)), were used because they are capable of posttranslational folding and modifications required to express MOP and 5-HT1A [62–64]. The experiments were first performed in HEK293 cells, and key findings were validated in PC12 cells. The HEK293 and PC12 cells were stably transformed to simultaneously express the human MOP receptor genetically fused at the C-terminus with the enhanced Green Fluorescent Protein (MOP-eGFP) and the human 5-HT1A receptor genetically fused at the C-terminus with the Tomato Red Fluorescent Protein (5-HT1A-Tomato). Both constructs were cloned into the pBudCE4.1 vector (Thermo Fisher, Munich, Germany), with the MOP-eGFP gene being expressed under the control of the hEF-1 promoter (KpnXho) and

5-HT1A-Tomato gene under the control of the CMV promoter (HindXba). For details, see Supplementary Materials (Section S1, Figure S1A).

For cultivation, untransformed and stably transformed HEK293 and PC12 cells were cultured in collagen coated T25 flasks (Sarsted) at 37 ◦C in a humidified atmosphere containing 5% CO2. HEK293 cells were cultured in DMEM medium supplemented with 10% Fetal Bovine Serum (FBS), 100 U/mL penicillin and 100 µg/mL streptomycin (PenStrep). For PC12 cells, the RPMI 1640 medium supplemented with 10% Horse Serum (HS) and 5% FBS, 100 U/mL penicillin and 100 µg/mL streptomycin was used. All cell culture reagents were from Invitrogen, Stockholm, Sweden.

For generating the stably transformed cell lines, the HEK293 and PC12 cells were grown to 70% confluence in 8-well chambered cover slides (Nalge Nunc International, Rochester, NY, USA) and transfected using Lipofectamine 2000 (Invitrogen), following the transfection protocol provided by the manufacturer. Stably expressing cell lines were isolated through selection using culture medium supplemented with phleomycin D1 antibiotic (0.4 mg/mL, Thermo Fisher). Positive and negative control cells were cultured and transfected in the same way. For details, see Supplementary Materials (Section S1, Figure S1A,B). The functionality of MOP-eGFP and 5-HT1A-Tomato receptors was validated by assessing how treatment with the selected agonists morphine, serotonin or buspirone or their combination: morphine and serotonin, or morphine and buspirone, affects phosphorylation of Erk1/2 and p38 MAPKs. The data show that all tested compounds and their combination increase the protein levels of p-ERK1/2 and p-p38 as compared to their levels in untreated cells (Section S1, Figure S1C).

#### *4.2. Confocal Laser Scanning Microscopy (CLSM) Imaging and Fluorescence Correlation and Cross-Correlation Spectroscopy (FCS/FCCS)*

The CLSM imaging and FCS/FCCS were performed using an individually modified ConfoCor 3 system (Carl Zeiss, Jena, Germany), as previously described [65,66]. Briefly, the system comprises an inverted microscope for transmitted light and epifluorescence (Axiovert 200 M); a VIS-laser module housing the Ar/ArKr (458, 477, 488 and 514 nm), HeNe 543 nm and HeNe 633 nm lasers; a scanning module LSM 510 META modified to enable imaging using silicon avalanche photodiodes (SPCM-AQR-1X, PerkinElmer, Waltham, MA, USA) in order to allow studies of cells expressing low levels of the proteins of interest; and an FCS/FCCS module with two detection channels. The C-Apochromat 40×/1.2 W UV-VIS-IR objective was used throughout. A stage incubator consisting of a heated microscope stage (Heating insert P), incubator box (Incubator S), atmospherecontroller (CTI-Controller 3700) and a temperature regulator (Temp control 37-2 digital), was used to maintain the cells at 37.0 ◦C and supply them with heated humidified air containing 5.0% CO2. The temperature and CO<sup>2</sup> levels were continuously monitored and regulated via a digital feedback control algorithm, allowing temperature control within ±0.2 ◦C and atmosphere control within ±0.1% CO2.

The CLSM images were acquired in a sequential, i.e., dual track mode, one channel at a time. The eGFP fluorescence was excited using the 488 nm line of the Ar/ArKr laser. A band pass 505–530 nm emission filter was used to spectrally narrow the emitted fluorescence. Tomato fluorescence was excited using the 543 nm HeNe laser, and a long pass 580 nm emission filter was used to collect the emitted fluorescence. Incident and emitted light were separated using the main dichroic beam splitter HFT 488/543/633. The eGFP and Tomato fluorescence were separated using a secondary dichroic beam splitter NFT 545 (Figure 1A). Images were acquired without averaging, using a pixel dwell time of 51.2 µs and a 512 × 512 pixels format (Figure 1B).

The optical setup for FCCS was the same as for CLSM imaging described above. Fluorescence intensity fluctuations were recorded at the apical plasma membrane of live cells identified by an axial fluorescence intensity scan (Figure 1C). Time series were collected in an array of 10 consecutive measurements, each measurement lasting 20 s (Figure 1D).

#### *4.3. Brief Background on FCS/FCCS*

Fluorescence cross-correlation spectroscopy (FCCS) is a dual color variant of fluorescence correlation spectroscopy (FCS). The FCS measures with sub-microsecond temporal resolution spontaneous fluctuations in fluorescence intensity around a steady state to extract quantitative information about the concentration and diffusion/size of fluorescent molecules [67–69]. The FCS is well suited for biological applications, as it is non-destructive and allows quantitative measurements to be performed in sub-cellular compartments [70]. The fluctuations in fluorescence intensity are recorded in a very small observation volume element (OVE) that is typically about VOVE = 0.1–2 fL ((0.1–2) <sup>×</sup> <sup>10</sup>−<sup>15</sup> L). The OVE is generated by tightly focusing the incident laser light into the sample using a high numerical aperture objective. Fluorescence is collected through the same objective, and the volume from which fluorescence is being collected is reduced by placing a pinhole in the optically conjugate plane in front of the detector [69,71]. The spontaneous diffusion of fluorescent molecules in and out of the OVE gives rise to fluctuations in fluorescence intensity. The size and volume of the OVE is specific for each instrument and is determined in calibration experiments using a reference molecule with a known diffusion coefficient, such as Rhodamine 6G (Rh6G). Using a standard 10 nM Rh6G solution, and following the procedure described in detail in [71], the OVE volume in our system was determined to be VOVE = 0.2 fL [68]. For a quick estimate of the concentration, note that for a 10 nM solution and VOVE = 0.17 fL, the average number of molecules in the OVE (N) is N = 1 [71,72].

The FCS/FCCS works best at low, sub-micromolar concentrations [71], where the signal from a bright fluorescent molecule generates a substantial increase in fluorescence intensity that is well above the background signal from the surrounding molecules. From the recorded fluorescence intensity fluctuations, which are generated by the translational diffusion of fluorescent molecules, one can extract the average number of molecules in the OVE (N), i.e., their concentration and their average translational diffusion time (τD), which is defined by the diffusion coefficient (D), i.e., the size of the molecule (Stokes-Einstein equation). To extract this information from fluorescence intensity fluctuation analysis, the most commonly employed method, which is also used here, is temporal autocorrelation analysis. In temporal autocorrelation analysis, the signal is compared to a copy of itself delayed for a certain lag time (τ) using the autocorrelation function:

$$\mathbf{G}(\mathbf{r}) = 1 + \frac{\langle \delta \mathbf{F}(\mathbf{t}) \cdot \delta \mathbf{F}(\mathbf{t} + \mathbf{r}) \rangle}{\langle \mathbf{F}(\mathbf{t}) \rangle^2} \tag{1}$$

In Equation (1), chevron brackets denote average values of the analyzed variables over time, and fluorescence intensity fluctuation (δ(F(t)) is the deviation of the fluorescence intensity at time t (F(t)) from the mean fluorescence intensity (hF(t)i), δF(t) = F(t) − hF(t)i. Accordingly, δF(t + τ) = F(t + τ) − hF(t)i. When the fluctuations are not random, temporal autocorrelation analysis yields a temporal autocorrelation curve (tACC). The tACC is characterized by a maximal limiting value of G(τ) as τ → 0 and decreases to the value G(τ) = 1 at long lag times, indicating that correlation between the fluorescence intensities is being lost (Figure 1E, green and red). If there is only one process that gives rise to fluorescence intensity fluctuations, the tACC shows one inflection point, that is, one characteristic decay time. If there are more processes giving rise to fluorescence intensity fluctuations, which occur at different time scales, the tACC assumes a more complex shape with more than one characteristic decay time (Figure 1E, green and red). The zero-lag amplitude of the tACC, (G<sup>0</sup> = G(0) − 1) provides information about the concentration of fluorescent molecules as it equals the inverse average number of molecules in the OVE (1/N). Thus, the amplitude of the tACC decreases when N increases. The characteristic decay time of the tACC gives information about the rates at which processes that give rise to the fluorescence intensity fluctuations occur. When fluorescence intensity fluctuations are generated by molecular diffusion, the characteristic decay time of the tACC reflects the average time it takes for a molecule to cross through the OVE by translational diffusion.

For dual colour FCCS, two spectrally distinct fluorescent molecules, such as eGFP and Tomato, are used to render the molecules of interest visible. Fluorescence intensity fluctuations are then simultaneously recorded for both fluorophores using overlaying excitation pathways, but separate detector pathways (Figure 1A). The fluorescence intensity fluctuations observed in FCCS (Figure 1D) are subjected to temporal auto- and crosscorrelation analysis. This generates two tACCs, one for each fluorophore (Figure 1E, red and green) and, when the molecules of interest bind, one temporal cross-correlation curve (tCCC; Figure 1E, black), which reflects the population of dually labelled molecules diffusing as one [36,73,74]. As in FCS, the amplitudes of the individual tACCs contain information about the total average number of green- and red-labelled molecules in the OVE, now being the sum of unbound singly labelled molecules and the bound dually labelled complexes. Thus, for the eGFP-labelled MOP receptors, N<sup>g</sup> total = N<sup>g</sup> + Ngr, and for the Tomato-labelled 5-HT1A receptors, N<sup>r</sup> total = N<sup>r</sup> + Ngr (Figure 1E). Only the dually labelled receptor-receptor heterodimer molecules give rise to fluorescence intensity fluctuations in both detectors at the same time, and are thus the only ones to contribute to the tCCC, obtained by calculating the cross-correlation function:

$$\mathbf{G\_{CC}(\tau)} = 1 + \frac{\langle \delta \mathbf{F\_{green}(t)} \cdot \delta \mathbf{F\_{red}(t+\tau)} \rangle}{\langle \mathbf{F\_{green}(t)} \rangle \langle \mathbf{F\_{red}(t)} \rangle} \tag{2}$$

In contrast to the amplitudes of the tACCs (Equation (1)), which are inversely proportional to the average number of molecules in the OVE (see detailed explanation in [71,75], the zero-lag amplitude of the tCCC (Equation (2)) is directly proportional to the number of dually labelled molecules (Ngr) and thus increases as Ngr increases:

$$\mathbf{G\_{CC}}(0) - 1 \propto \frac{\mathbf{N\_{gr}}}{\left(\mathbf{N\_{g}} + \mathbf{N\_{gr}}\right) \cdot \left(\mathbf{N\_{r}} + \mathbf{N\_{gr}}\right)}\tag{3}$$

In order to characterize the degrees of binding between MOP-eGFP and 5-HT1A-Tomato, i.e., to determine the number of the heterodimer receptor complexes, FCCS data are further analysed to obtain a dimensionless value known as the relative cross-correlation amplitude (RCCA) [75]. The RCCA is defined as the limiting value, when the lag time approaches zero (τ → 0), of the cross-correlation curve relative to the autocorrelation curve for a single fluorophore. For example, the number of bound, dually labelled molecules carrying both the green and the red label, relative to the total number of molecules carrying the red label (N<sup>r</sup> total = N<sup>r</sup> + Ngr), equals the amplitude of the cross-correlation curve (GCC(0) − 1) relative to the amplitude of the green autocorrelation curve (GAC,g(0) − 1) [75]:

$$\text{RCCA} = \frac{\text{G}\_{\text{cc}}(0) - 1}{\text{G}\_{\text{AC}, \text{g}}(0) - 1} = \frac{\text{N}\_{\text{gr}}}{\text{N}\_{\text{r}}^{\text{total}}} = \frac{\text{N}\_{\text{gr}}}{\text{N}\_{\text{r}} + \text{N}\_{\text{gr}}} \tag{4}$$

Knowing the concentration of MOP-eGFP and 5-HT1A-Tomato molecules, and the concentration of heterodimer MOP-eGFP–5-HT1A-Tomato complexes, the apparent dissociation constant for the receptor-receptor heterodimer complex can be calculated:

$$\mathbf{K}\_{\rm d}^{\rm app} = \frac{\mathbf{c}\_{\rm free}^{\rm MOP} \cdot \mathbf{c}\_{\rm free}^{5-\rm HT1A}}{\mathbf{c}\_{\rm MOP-5-\rm HT1A}} \tag{5}$$

or, when expressed using the quantities determined by FCCS:

$$\mathbf{K\_{d}^{\rm app}} = \frac{\left(\mathbf{N\_{g}^{\rm total}} - \mathbf{N\_{r}^{\rm total}} \cdot \mathbf{RCCA}\right) \cdot (1 - \mathbf{RCCA})}{\mathbf{RCCA}} \cdot \frac{1}{\mathbf{N\_{A}} \cdot \mathbf{V\_{OVE}}} \tag{6}$$

In Equation (6), N<sup>A</sup> is the Avogadro number. For determining the dynamic range of the RCCA, i.e., the smallest and the largest RCCA values that could be reliably measured, see control experiments described in Supplementary Materials (Section S1. Transfection, positive and negative control cells (Figure S2)). Derivation of Equation (6) is given in Supplementary Materials, Section S2. Calculation of the apparent dissociation constant).

One challenge in dual-color FCCS that is particularly important to consider when fluorescence proteins are being used is the risk of false-positives due to spectral crosstalk between channels, which may lead to overestimation of the cross-correlation amplitude [75]. In order to ascertain that this error is minimized, we have validated the optical setting using control cells—cells expressing eGFP and Tomato were used as negative control (Supplementary Materials, Section S1. Transfection, positive and negative control cells (Figure S2C)) and cells expressing genetically fused eGFP-Tomato were used as positive control (Supplementary Materials, Section S1. Transfection, positive and negative control cells (Figure S2A)). The corresponding tCCCs are shown in Figure 1F. Control experiments showed that the dynamic range in dual-color FCCS differed from the theoretical range, 0 ≤ RCCA ≤ 1, and was determined to be, (0.10 ± 0.07) ≤ RCCA ≤ (0.80 ± 0.08). The RCCA value determined in the negative control experiments, RCCAnc = (0.10 ± 0.07), indicated that only values that are significantly larger then RCCAnc should be considered as a positive indication of binding. Positive control experiments indicated that RCCA values higher than RCCApc = (0.80 ± 0.08) may not be reached for reasons explained in (Supplementary Materials, Section S1. Transfection, positive and negative control cells (Figure S2E,F), and that RCCA values as high as RCCApc indicate that 100% binding between the investigated receptors has been reached.

In order to ascertain that errors due to spectral crosstalk are minimized, the optical setting was further validated using the so-called switching mode. In the switching mode, the sample is alternatingly (every 240 µs) illuminated with one laser at a time to excite one fluorophore only [69,71]. By using the switching mode, we could adjust the optical setting so that the crosstalk is minimal when the non-switching mode is being used, as explained in detail in [71], thus ascertaining that increased RCCA are actually being observed following treatment with non-peptide opioids (Supplementary Materials, Section S3. Relative cross-correlation amplitude (RCCA) increased upon opioid treatment. Verification was done by switching FCCS (Figures S3 and S4). Finally, in order to account for the, while minimized, inevitably present cross-talk-induced cross-correlation, the RCCA was corrected by subtracting the cross-talk-induced cross-correlation from the RCCA and scaled up as follows [36,75,76]:

$$\text{RCCA}\_{\text{corrected}} = \frac{\text{RCCA} - \kappa \cdot f}{1 - \kappa \cdot f} \tag{7}$$

where κ is the so-called bleed-through ratio, i.e., brightness as reflected by the counts per second and per molecule (CPM) of the green dye in the red channel (CPMg/r) when the red fluorophore is not present, divided by its brightness in the green channel CPMg/g, κ = CPMg/r/CPMg/g, and *f* is the count rate (CR) ratio in the green and red channels, *f* = CRg/CRr. For the optical setting used in our studies, κ = 0.1 and *f* ≤ 1.2. Following treatment, a two-fold increase in eGFP brightness was observed, while Tomato brightness remained largely unchanged (Table 1). To account for this, the κ factor was accordingly scaled. Thus 0.1 ≤ κ ≤ 0.2, and the product 0.12 ≤ κ·*f* ≤ 0.24.

#### *4.4. Opioid Treatment*

Cells stably expressing MOP-eGFP and 5-HT1A-Tomato were cultured in 8-well chambered coverslides (Nalge Nunc International, USA) at 37 ◦C in a humidified atmosphere containing 5% CO<sup>2</sup> using phenol red-free media, supplemented in the same way as described above. To ascertain that MOP-eGFP is functional and integrated into cellular physiology, the selective MOP receptor peptide agonist DAMGO ([D-Ala<sup>2</sup> , N-MePhe<sup>4</sup> , Gly-ol<sup>5</sup> ] enkephalin), was used (Figure 2) [77]. For experiments with non-peptide opioids, the cells were incubated for 18 h with different concentrations of fentanyl (50 nM, 500 nM, 750 nM or 1 µM) or morphine (250 nM, 500 nM or 750 nM). Based on the results of these experiments (explained in the Section 2), the 750 nM concentration was selected as suitable

for further studies. Hence, the cells were subsequently treated with 750 nM codeine or oxycodone. Non-peptide opioids and the peptide opioid DAMGO were all purchased from Sigma-Aldrich.
