Selective and Wash-Resistant Fluorescent Dihydrocodeinone-Derivatives Allow Single-Molecule Imaging of mu-Opioid Receptor Dimerisation.

mu-Opioid receptors (mu-ORs) play a critical role in the modulation of pain and mediate the effects of the most powerful analgesic drugs. Despite extensive efforts, it remains insufficiently understood how mu-ORs produce specific effects in living cells. We developed new fluorescent ligands based on the mu-OR antagonist E-p-nitrocinnamoylamino-dihydrocodeinone (CACO), that display high affinity, long residence time and pronounced selectivity. Using these ligands, we achieved single-molecule imaging of m-ORs on the surface of living cells at physiological expression levels. Our results reveal a high heterogeneity in the diffusion of mu-ORs, with a relevant immobile fraction. Using a pair of fluorescent ligands of different color, we provide evidence that mu-ORs interact with each other to form short-lived homodimers on the plasma membrane. This approach provides a new strategy to investigate mu-OR pharmacology at single-molecule level.


Introduction
Opioid receptors (ORs) belong to the family Ao f rhodopsin-like Gp rotein-coupled receptors and occur in three major subtypes, m, d and k. [1] They are predominantly found in the central and peripheral nervous system, where they modulate transmission and perception of pain. Importantly, m-ORs mediate most therapeutic effects of opioids, which are the most powerful, but also most addictive analgesics known to date. [2] In contrast, k-a nd d-OR activation causes weaker analgesia, but has been associated with less side effects. [3] Thee ffects of opioids on neurons are mainly mediated by activation of heterotrimeric G i proteins, which inhibit cAMP production, while opening Gp roteincoupled inward rectifying potassium channels (GIRK) and closing voltage-dependent N-type Ca 2+ channels via G bg subunits. [2,3] Because of their importance as drug targets, m-ORs have been intensively investigated both, in vitro and in vivo. [1][2][3][4] A major breakthrough in the field has been the determination of high-resolution three-dimensional structures of the m-OR in complex with an antagonist and, later on, with an agonist and Gp rotein mimetic nanobodies,a sw ell as most recently by cryo-electron microscopy. [5] These studies have provided important insights into the mechanisms of ligand binding and receptor activation, which might pave the way to the rational design of new analgesics with improved efficacy and less side effects,such as addiction. Interestingly,inthe above studies m-ORs were found to crystallize as dimers. [5a,b] Largely based on previous experiments applying resonance energy transfer methods,t hese findings support the hypothesis that m-ORs might form homodimers. [1,4b, 6] However,the nanoscale organization and dynamics of m-ORs on the surface of living cells remain largely unknown, mostly due to technical limitations of conventional methods.T hese typically include the requirement of cell disruption, insufficient spatiotemporal resolution and/or averaging over thousands or even millions of receptors. [7] In an attempt to overcome such limitations,weand others have developed innovative methods based on single-molecule microscopy,w hich allow imaging of individual receptors and other molecules on the surface of living cells with atemporal resolution of approximately 28.4 ms and as patial resolution of approximately 20 nm, which is at least 10 times below the best theoretical resolution of conventional fluorescence microscopy. [8] This approach can provide ahighly quantitative characterization of dynamic events,s uch as protein-protein interactions among membrane receptors, [9] even when involving only as mall fraction of the investigated molecules. [9a,c, 10] Here,w ep resent the synthesis and single-molecule microscopy application of new fluorescent, selective m-OR ligands to study unmodified receptors in living cells at physiological expression levels.

Results and Discussion
Thes tructure of the ligands consists of ap harmacologically active compound, afluorophore,and alinker ( Figure 1).
All three units have been chosen with regard to their individual and specific characteristics to achieve high affinity and selectivity as well as optimal optical properties for singlemolecule fluorescence microscopy,f or example,h igh signalto-noise ratios and low blinking and bleaching rates.
Thep harmacologically active moiety (parent ligand) is based upon E-p-nitrocinnamoylamino-dihydrocodeinone (CACO), which had been described to be ap otent and selective m-OR ligand, and to retain such properties upon conjugation with the small organic fluorophore BODIPY (EC 50 = 24.4 nm, d/k > 1000 nm)t hrough as mall alkylene linker. [11] Theu nlabeled ligand has been described to act as short-term agonist and long-term antagonist with an IC 50 value of 0.46 AE 0.003 nm for m-OR (d = 4.2 AE 1.3 nm, k = 19 AE 2.8 nm). A K d of 0.52 AE 0.14 nm was measured by means of saturation binding experiments with [ 3 H]DAMGO. [12] The nitrocinnamoyl part of the active compound contains aM i-chael acceptor,w hich potentially forms covalent bonds with nucleophilic side-chains of amino acids in the m-OR binding pocket. This has been shown in the m-OR crystal structure for b-funaltrexamine. [5a] Thep entylene linker connects the pharmacologically active moiety and the fluorophore.I no ur probe design, we incorporated at etraglycine into the linker,i nspired by previous studies based on GPCR-imaging with fluorescent ligands,w hich yielded improved results. [13] Cyanine 3a nd 5 (Cy3/5) were chosen as fluorophores because of their advantageous optical properties which make them particularly suited for single-molecule microscopy,i ncluding emission in the red and near infrared regions of the spectrum, respectively,a swell as high absorption coefficients and quantum yields. [14] Thes ynthesis shown in Scheme 1s tarted with ah etero-Diels-Alder reaction of thebaine 1 and the intermediately oxidized N-hydroxycarbamate 11,p repared from the respective chloroformiate 10. [15] Ther esulting cycloadduct 2 was hydrogenated to give dihydrocodeinone 3. [16] An itrocinnamoyl moiety was introduced by coupling the C14-amino group of 3 with the respective activated acid, yielding compound 4. [17] Subsequent reductive amination of the C-6keto group led to compound 5. [11] In apreliminary approach, we directly used the NHS-activated dyes to couple them to the newly introduced amino-group.H owever,t hese probes had shown high background and ap oor signal-to-noise ratio in single-molecule microscopy experiments (see Supporting Information). While an alkylene chain might be sufficient to bridge pharmacophore and fluorophore and define the distance between the bulky residues,t he tetraglycine moiety additionally increases the polarity of the compound and prevents sticking to the plasma membrane ( Figure 1). This led to am ajor increase in selectivity and reduction of background. Thenewly introduced amino group of 5 was coupled to N-Cbz-protected tetraglycine,y ielding compound 6. [17] After deprotection of Cbz with hydrobromic acid, the free amine 7 was coupled to the NHS-activated dyes Cy3 and Cy5 to give the desired fluorescent ligands 8 and 9 (Scheme 1).
Saturation binding curves obtained for compound 9 showed its selectivity for m-OR ( Figure 2A)inahomogenous time-resolved FRET (HTRF) assay with HEK293 cells,which gave optimal results.N os ignificant binding was observed to the other OR-subtypes ( Figure 2A). These findings are in good agreement with previous binding studies on the parent compound CACO. [12] We then tested compounds 8 (Cy3 conjugate) and 9 (Cy5 conjugate) against wild-type m-OR expressed in CHO cells, which adhere very well to glass-coverslips,r esulting in ap articularly flat plasma membrane.C ompounds 8 and 9 bound to cell-surface receptors were selectively imaged by total internal reflection fluorescence microscopy (TIRF), which illuminates only approximately 200 nm at the interface between the coverslip and the cells ( Figure 2B). CHO cells devoid of m-OR expression were used as control for unspecific binding.The results confirmed ahighly specific binding of the compounds to m-ORs,w ith negligible unspecific binding to the cells or the coverslip ( Figure 2B). This approach also allowed us to obtain concentration-binding relationships for both compounds.I nt hese experiments,w eincubated CHO cells with increasing concentrations of either compound 8 or 9 for 20 min and imaged them by TIRF microscopy.T he mean intensities of at least 40 cells per condition were averaged for each concentration of ligand, which allowed us to estimate their affinities (K D )( Figure 2C). A K D value of 87 AE 49 nm for compound 8 was reached, whereas compound 9 showed at hree-fold lower affinity for m-ORs with an estimated K D value of 295 AE 141 nm.The initial affinity of CACO for the m-OR was reported to be 0.52 AE 0.14 nm. [12] Intrinsic activity of compound 8 was determined in an inositol mono phosphateaccumulation assay for G-protein mediated signaling.I nt his assay,t he compound acted as ap artial agonist (EC 50 = 190 nm, E max = 57 %ofmaximal response to morphine), while it was inactive in the b-arrestin-2 recruitment assay (cf. Supporting Information). Although chemical modifications often lead to changes in the pharmacological profiles of small molecules,b oth compounds retain high affinity towards m-OR. A5 0% wash resistance of CACO has been reported in competition binding experiments with DAMGO,p robably due to the capability of the 14b-pnitrocinnamoylamino-side chain of CACO to bind covalently to the m-OR. [12] Fluorescent probes for single-molecule microscopy not only need to possess high labeling efficiency,but ideally,a lso long residence time on the receptor.T hus, intrigued by the possibility that CACO may bind covalently to m-OR, we investigated if compounds 8 and 9 also retained ah igh wash resistance.F or this purpose, we performed washing experiments in CHO cells transiently treated with either compound 8 or 9 for 20 min and imaged by TIRF microscopy (Figure 3). During washing,w e observed aslow decrease of fluorescence intensity,u ntil it reached ap lateau at approximately 33 %a nd 53 % of the initial values for compound 8 and 9,r espectively.I mportantly,t hese results indicate that both fluorescent ligands exhibit al ong residence time on m-OR, with af raction of virtually non-dissociating receptors. Next, we tested the applicability of the new fluorescent ligands for single-molecule fluorescence microscopy.B oth compounds showed excellent photophysical properties,giving rise to easily detectable fluorescent spots in TIRF microscopy. Compound 8 showed slower photobleaching in comparison to compound 9,c onsistent with the known photophysical properties of Cy5 and Cy3, respectively. [18] Therefore,t he Cy3 ligand was used for subsequent single-color experiments. CHO cells transiently transfected to express wild-type m-OR at low physiological densities (approximately 0.8 receptors per mm 2 )w ere incubated with as aturating concentration of compound 8 (1 mm)and imaged by fast TIRF microscopy.The transfected cells were easily distinguishable from the background, confirming ah ighly specific binding ( Figure 4A). Individual m-ORs carrying afluorescent ligand were detected and tracked using an automated algorithm. [9a,c, 19] At imeaveraged mean square displacement (TAMSD) analysis was performed, which allowed investigating the diffusion of the receptors within the plasma membrane (cf.S upporting Information). This analysis revealed ah igh heterogeneity in the mobility of m-ORs on the plasma membrane.Individual m-OR particles were classified into four categories based on the type and speed of their motion. [9c] Apercentage of 22 AE 2%of the receptors were virtually immobile,l ikely due to trapping in small nanodomains or binding to immobile membrane structures.S ub-diffusive motion was observed for 34 AE 1%, meaning their motion was hindered by either crowding or interactions with their environment. This phenomenon has been previously described for other membrane receptors and can arise from different factors,s uch as transient trapping in nanodomains. [9c,20] Another 34 AE 1% of the receptors showed normal diffusion, that is,B rownian motion. Super-diffusion, that is, directional motion, was observed for the remaining 10 AE 1% of particles.
These results are in agreement with previous findings for other prototypical GPCRs like the b 2 -o rt he a 2A -adrenergic receptor. [9c] We then explored the applicability of compounds 8 and 9 in single-molecule microscopy experiments to investigate the dimerization of m-OR at low/physiological expression levels in living cells.F or this purpose,C HO cells were transiently transfected with wild-type m-OR, yielding at otal receptor density of around 1.7 particles per mm 2 .T he cells were then simultaneously labeled with amixture of compounds 8 and 9 (1 mm and 0.5 mm,r espectively) to label as many receptors as possible (approximately 80 %) with either compound, while keeping unspecific binding to the glass-coverslip low.T hen, imaging by fast two-color TIRF microscopy was carried out. Thec o-localization between receptors labeled with compound 8 and 9 was analyzed by automated computer algorithms as previously described. [9c] In order to correct for the presence of random co-localizations and to estimate the duration of m-OR interactions,w ea pplied ap reviously developed method based on deconvolution of the co-localization times. [9c] Fort his purpose,w ea dditionally performed the same experiment using m-ORs labeled with compound 8 and CD86, amonomeric control protein not interacting with m-OR, labeled with Alexa Fluor 647 via aSNAP-tag fused at its N-terminus ( Figure S2 in the Supporting Information). [9a] This served as control for the co-localizations expected in the absence of interactions.T he deconvolution analysis revealed  that more than 95 %o fm-ORs were diffusing on the plasma membrane as monomers.H owever, it also revealed as mall but consistent fraction of receptors that apparently underwent transient interactions lasting approximately 1-2 seconds. At the low receptor densities analyzed, this fraction of m-ORs that were interacting at any given time was approximately 4-5%.Although this represents only afraction of the receptors, this value is remarkably similar to the one observed between active receptors and Gproteins, [9c] afundamental interaction in GPCR signaling,s uggesting that although involving only as mall fraction of receptors it might nevertheless be biologically relevant. Even though the presence of an even smaller fraction of higher order oligomers cannot be completely ruled out, the intensities of the majority of receptor particles and their bleaching behavior (i.e., number of observed photobleaching steps) were consistent with them being monomers or at most dimers.B yd econvolving the distribution of the co-localization times observed between m-ORs labeled with compounds 8 and 9, and those obtained between m-OR and CD86, we were able to estimate the frequency and duration of the interactions between m-ORs. Ther esulting relaxation plot of m-OR interactions (Figure 5B)i ndicated that m-ORs were dissociating following an exponential decay,w ith an estimated time constant (t)o f 1.797 AE 0.487 s(corresponding to adissociation rate constant, k off ,of0.557 AE 0.207 s À1 ). This value is in good agreement with previous results obtained with prototypical family A GPCRs. [9a-c, 21] This approach also allowed us to calculate the two-dimensional association rate (k on )f or interactions between m-ORs,t hat is,t he formation of dimers,w hich we estimated to be 0.020 AE 0.004 mm 2 molecule À1 s À1 .The estimated k on and k off values give adissociation equilibrium constant (K d )o f2 7.43 AE 11.75 molecules per mm 2 ,a llowing to predict the fraction of m-ORs in monomeric or dimeric state depending on their density on the plasma membrane.B ased on the obtained K d value,w ee stimate that a m-OR density of approximately 27 molecules per mm 2 would be required for 50 %o ft hem to be in dimeric state (approximately 7d imers and 14 monomers per mm 2 ). Since such densities might occur at synapses, [22] it is possible that ar elevant fraction of m-ORs might form transient dimers in vivo. [23] Interestingly,t he majority of receptors seemed to transiently stop diffusing while interacting ( Figure 5A). A possible explanation for this is that the observed interactions might represent receptors being simultaneously recruited to  the same clathrin-coated pit (CCP) before internalization. To investigate this possibility,w er epeated the same experiment in the presence of co-transfected GFP-tagged clathrin (Figure 6). Computational analyses showed that 77 AE 9% of all observed m-OR interactions happen outside of CCPs,w ith only 23 AE 9% occurring within or near CCPs.T hese results suggest that other mechanisms are involved in the observed transient trapping of m-ORs during their interactions.

Conclusion
Whereas GPCRs have long been believed to be solely monomeric receptors,agrowing body of evidence obtained mainly using fluorescence and bioluminescence resonance energy transfer (FRET and BRET) suggests that they might form dimers or even higher order oligomers. [7, 10b,24] However, the occurrence of GPCR dimerization at physiological expression levels and their stability remain-with few notable exceptions such as family CG PCRs-controversial. These debates also extend to m-OR, for which contrasting results have been obtained using methods relying on different expression levels,r anging from physiological levels to overexpression and even purification. Whereas several of the published studies provide evidence in favor of dimerization, [1, 4b,5a, 25] others indicate that m-ORs can function as monomers or the fraction of dimers is negligible. [26] Our data, obtained on wild-type receptors expressed at levels in living cells contribute to resolve this controversy by showing that whereas most m-ORs are monomeric, as mall, though potentially biologically relevant, fraction of receptors undergo transient dimerization. As the receptors appear to be in equilibrium between monomers and dimers,h igher level of dimerization might be achieved at sites of particular high receptor density such as synapses.
In summary,w ed eveloped two novel, fluorescent, subtype-selective ligands for the m-OR with binding affinities in the nanomolar range and the advantageous property of wash resistance.Importantly,weshow that the favorable properties of these ligands make them ideal for single-molecule microscopy.W eu sed this set of ligands to investigate the diffusion behavior of m-ORs as well as their interactions at physiological expression levels.Our results reveal the occurrence of transient receptor interactions,c onsistent with ad ynamic equilibrium between monomers and dimers.A so nly as mall fraction of receptors was found to be interacting with each other in our study,w hich is too small to be detected in ensemble measurements,o ur results may help to reconcile some of the apparent discrepancies between previous studies. Importantly,our results reveal dynamic interactions among m-ORs that occur outside CCPs that would have likely been missed using ensemble methods.This new approach opens up new venues to investigate m-OR biology and pharmacology in living cells with unprecedented spatiotemporal resolution.

Experimental Section
Comprehensive details of the experimental design can be found in the SupportingI nformation.