Synthesis

To replace the succinyl moiety in the macrocycle, azobenzenes 6–8 and arylazopyrazoles 12 and 13 were prepared according to Scheme 1 providing a free and a protected carboxylic acid and an individual substitution pattern. Therefore, the nitroso compounds 2, 3, 9 and 10 were synthesized from the corresponding amines in an oxidation reaction with oxone by a literature known reaction.¹⁰ The Mills reaction that yielded the azobenzenes 6–8 was conducted in acetic acid over four days. For the preparation of the arylazopyrazoles 12 and 13, the Mills reaction was performed under alkaline conditions in dichloromethane and triethylamine. The unnatural photoswitchable amino acid 16 was also synthesized in a Mills reaction. Starting from the amino-phenylalanine 15 and nitrosobenzene 14 the photoswitchable derivative of Fmoc-protected phenylalanine 16 was synthesized¹¹ (Scheme 1).

The N^ω-carbamoylated arginine building blocks 17 and 18 (Figure 2), containing a dimethylene (17) or a tetramethylene (18) spacer, were used in solid phase peptide synthesis (SPPS) to incorporate the respective N^ω-carbamoylated arginine in position 3 in the hexapeptides.

The original linker length of four carbons in 1 was shortened to two carbons for the incorporation of 6–8, 12 and 13 to keep the ring size closer to the parental compound 1. Compounds 17 and 18 were synthesized using previously reported procedures (for detailed conditions see Scheme S1, Supporting Information).¹²

The linear peptides 19–26, precursors for the cyclic peptides 27–34, were synthesized by Fmoc strategy SPPS using a Sieber amide resin (Scheme 2). 2-(1H-Benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU)/hydroxyl-benzotriazole (HOBt)/N,N-diisopropylethylamine (DIPEA) were used as coupling reagents for the natural amino acids as well as for the unnatural amino acids 16–18. For the photoswitches 6–8, 12 and 13, benzotriazol-1-yloxytripyrrolidino-phosphonium hexafluorophosphate (PyBOP)/HOBt/DIPEA were used as coupling reagents. The natural amino acids were used in a “double coupling”, i.e. every coupling step was performed twice in 5-fold excess for 45 min at 35 °C. A “single coupling”, using 3-fold excess overnight at 35 °C, was applied for the unnatural amino acids 16–18 and photoswitches 6–8, 12 and 13. After the final Fmoc deprotection, either of the photoswitches 6–8, 12 or 13 was coupled to the peptide. For peptides 24–26, containing the photoswitch in the amino acid side chain, the resin was treated with succinic anhydride. Cleavage from the resin was achieved using a mixture of trifluoro acetic acid (TFA)/CH₂Cl₂ (1 : 3), followed by side chain deprotection using TFA/CH₂Cl₂ (1 : 1) and purification by preparative high performance liquid chromatography (HPLC) yielding the linear peptides 19–26. The cyclization of the peptides was conducted in solution not exceeding a peptide concentration of 5 mM. The amide bond was formed between the primary amine of the N^ω-carbamoylated arginine residue and the carboxylic acid of the N-terminally attached photoswitchable moiety (19–23) or the N-terminal succinyl group (24–26). PyBOP/HOBt/DIPEA was used as coupling reagent and the reaction mixture was stirred at room temperature for 24 h. Purification of the final peptides was performed using preparative HPLC.

Photophysical Characterization

The photophysical properties of all cyclic photoswitchable peptides and the linear peptides 20 and 25 were investigated in aqueous buffer to mimic physiological conditions. All compounds were switched to their cis-isomers by irradiation at a wavelength of 340 nm. Whereas compounds 20, 25, 27–29 and 32–34, containing an azobenzene, were switched back to the trans- isomer at a wavelength of 455 nm, arylazopyrazoles 30 and 31 were switched back to trans by using light of a longer wavelength (528 nm) due to a slightly red-shifted absorption spectrum. Switching back and forth over several cycles showed that the compounds exhibit high fatigue resistance. No decomposition or side reactions seem to take place upon irradiation (Figure 3).

Furthermore, all photochromic peptides showed high photostationary states (PSS) when switched to the cis-isomer (≥90 %). Switching back to the trans-isomer worked best for the arylazopyrazoles 30 and 31 with around 90 % PSS. Peptides 20, 25, 27–29 and 32–34 containing an azobenzene moiety could be switched back to the trans-isomer with 74–88 % PSS.

Thermal half-lives (t_1/2), describing how fast the thermal back-isomerization to the trans-isomer occurs, ranged between 7.5 and 52 days, which ensures that during the time required for the biological assays back-isomerization is not significant. Table 1 summarizes the photophysical properties of the photoswitchable peptides.

Ligand Binding and Functional Characterization

In the following paragraphs the term “E-isomer” refers to the ratio of isomers after irradiating the compound in thermal equilibrium with 455 nm and 528 nm, respectively. The term “Z-isomer” refers to the ratio of isomers after irradiating the compound in thermal equilibrium with 340 nm. To ensure the optimized ratio for each individual isomer in binding studies and functional assays, the compounds were always pre-illuminated with the corresponding wavelength. Assays were performed immediately afterwards in the dark.

The photochromic peptides 20, 25 and 27–34 were investigated with respect to their Y₄R binding affinity by radioligand competition binding yielding pK_i values (Table 2). The incorporation of a photoswitchable moiety was well tolerated and binding affinities were still in the sub- to low-nanomolar range with 25 and 34 being the only exceptions. It was observed that for peptides 27–31, with the photoswitch integrated in the cyclic part, the Z-isomers exhibited higher Y₄R affinities (lower K_i values) compared to their E-isomers. Z-27 and Z-30 showed the highest affinities, similar to the endogenous ligand human pancreatic polypeptide (hPP) and to 1. Compound 28 displayed the biggest difference between the E- and Z-isomer with an almost 10-fold difference in binding affinity (Figure 4A). In contrast, the E- and Z-isomer of its precursor 20 exhibited identical Y₄R affinities. The isomers of 27, 30 and 31 exhibited seven- to eight-fold differences in Y₄R binding. Selected compounds were also subjected to radioligand competition binding studies for the NPY receptor subtypes Y₁, Y₂ and Y₅ (for details see Supporting Information Table S1) revealing a clear selectivity for the Y₄R.

Moreover, the ligands were investigated in a miniG_si protein recruitment assay to study their capability of activating the Y₄R. A split-luciferase-based miniG_si protein sensor was recently reported as a tool to characterize ligands for histamine receptors.¹³ The NanoLuciferase (NLuc) used for the histamine sensor shows a maximum emission around 470 nm. As the light emitted by NLuc may induce cis- to trans-isomerization under assay conditions for some of the ligands studied here, we developed a Y₄R split-luciferase miniG_si recruitment assay based on the click beetle red luciferase (CBR) with an emission maximum around 615 nm. Due to its broad emission spectrum, partial isomerization of the photochromic Y₄-peptides caused by CBR emission has to be considered. In order to keep the influence of CBR-induced photo-isomerization as low as possible, dose-response curves were established based on the maximum of the signal time courses instead of using the area under the curve (AUC). The former were observed approximately 4 to 7 minutes after agonist addition, whereas the AUC integration is typically extended beyond the signal maximum and thereby increases the impact of CBR-induced back-isomerization (see Figure S55, Supporting Information). For comparison, compound 28 was additionally investigated in a NLuc-based Y₄R miniG_si recruitment assay, developed within the scope of this study. Whereas E-28 exhibited a 5-fold lower potency than Z-28 in the CBR-based miniG_si assay, the difference was less pronounced (2.5-fold) in the NLuc assay (Figure S53) indicating that NLuc-emitted light (λ_max≈470 nm) may cause, to some extent, a back-isomerization from Z to E. All compounds were found to behave as partial agonists of the Y₄R in the CBR-based miniG_si recruitment assay, like the parental compound 1.⁹ The arylazopyrazoles 30 (ΔE_max=38 %) and 31 (ΔE_max=27 %) and the peptide with the photoswitchable amino acid sidechain in position 4 (33, ΔE_max=39 %) revealed the largest differences in efficacy (Table 2). Consistent with the results found in the binding studies, Z-isomers of 27–31 were more potent than their correspondent E-isomers. In contrast, peptide 33 displayed a higher potency in its E-configuration. Whereas peptides 27–31 exhibited slightly lower potencies than the endogenous ligand hPP, 33 and 34 showed a drastically reduced potency. The differences in EC₅₀ values between E- and Z-isomer were less pronounced than the differences in K_i values (Table 2). With a 5-fold difference in EC₅₀ values, peptide 28 displayed the most pronounced change in potency for its two isomers, followed by 31 with an approximately 4-fold difference (34 is not considered here because of the low potency and efficacy). The discrepancies between K_i values and EC₅₀ values may be due to the absence or presence of sodium ions in the buffer composition used for binding and functional assays, respectively, as has been discussed previously.¹⁴ In addition to miniG_si recruitment assays, the E- and Z-isomer of 28 were also studied in a Ca²⁺ aequorin assay, measuring intracellular calcium mobilization upon Y₄R activation. In this functional assay Z-28 also proved to be more potent than E-28 by a factor of 3. However, compared to the data from the Y₄R CBR miniG_si recruitment assay, the efficacies were considerably lower (cf. Figure S54, Supporting Information). Likewise, the potencies were slightly lower, which is most likely due to the strong non-equilibrium character (rapid cellular response) of the Ca²⁺ assay.

Impedance-Based Analysis of Y₄R Activation

Both isomers of compounds 28 and 30 were analysed by time-resolved impedance measurements with respect to the response of adherent CHO cells overexpressing the Y₄ receptor. These two compounds belong either to the azobenzene (28) or arylazo-pyrazole (30) family of photoswitches used in this study. In these experiments the cells were grown to confluence in 96-well plates with integrated, planar gold-film electrodes in every well. The impedance magnitude |Z| was recorded at an AC frequency of 12 kHz using non-invasive voltage amplitudes (40 mV) to induce a weak AC current in the μA range. As has been shown previously, activation of GPCR signalling cascades are readily monitored by the impedance time course that mirrors the associated changes in cell morphology.^4d-5 Figure 5 summarizes a typical experiment using E-30, Z-30 (1 nM), the endogenous ligand hPP (100 nM) and a vehicle control (DMSO, 0.01 % v/v). The impedance magnitudes of the individual electrodes have been zeroed to the impedance of the cell-covered electrodes immediately before adding the agonists.

In these experiments 30 was irradiated with light of 528 nm or 340 nm to yield E-30 or Z-30, respectively, prior to cell exposure. Saturating the receptor with the endogenous ligand hPP (100 nM) induced an impedance change of approximately 650 Ω within 20 min and it remained stable within the observation time of 80 min. The corresponding DMSO control (0.01 % v/v) increased the impedance just by 20 Ω along the entire observation time. The two isomers Z-30 and E-30 showed distinctly different impedance responses when applied in 1 nM concentrations. Whereas Z-30 induced an impedance increase of approximately 400 Ω with a similar time course as hPP, E-30 was not significantly different from the DMSO control. For a more quantitative analysis we used the area under the curve (AUC) for each impedance time course in the time interval from 20 min to 40 min using the DMSO control as the lower border of integration (see orange area in Figure 5). AUC has proven to yield a more robust readout than the maximum impedance change in particular for non-monotonic time courses.⁶ Therefore, AUC analysis was applied to experiments conducted with increasing concentrations of both isomers of compounds 28 and 30 to reveal their individual dose-response relationships as summarized in Figure 6.

For both compounds 28 and 30, we found the Z-isomer to be more potent than the E-isomer, which is in line with the radioligand competition binding and the miniG_si recruitment assay (Table 2). pEC₅₀ values as determined from impedance-based cell monitoring are summarized in Table 3. At saturating concentrations both compounds showed the same intrinsic activity as the endogenous ligand hPP. This phenomenon is often observed for integrative assays capturing the very end of the signalling cascade.¹⁵

K_i values of 28 and 30 determined from radioligand competition binding differed by a factor of approximately 10 between the isomers (E/Z). In miniG_si recruitment, this difference was less pronounced. The factor between EC₅₀ values of the isomers was just 5 for 28 and 2 for 30. In impedance measurements 30 was found to show the bigger difference in potency between E and Z-isomers with an approximately 6-fold difference, whereas the factor for 28 was only 2.5. It is important to recognize that miniG_si-recruitment assays report on signalling events that are localized directly at the receptor (proximal) while label-free wholistic measurements like impedance analysis integrate over the entire cell body and the entire signalling cascade (distal) similar to experiments performed with entire organs.¹⁶ This integrative character together with the inherent amplification along the signalling cascade is also responsible for the fact that impedance measurements do not reveal any significant difference in efficacy for the two isomers relative to hPP in contrast to miniG_si recruitment.^15b In impedance-based assays, both isomers of 28 and 30 are considered full agonists. Another result is noteworthy: in impedance-based analysis both isomers of 30 are significantly more potent than hPP. The other two assays do not report a similar behaviour. A potential explanation is functional selectivity (biased agonism).¹⁷ The concept of functional selectivity proposes that different ligands are capable of triggering different signaling cascades to different degrees upon binding to the same receptor. Whereas competition binding and G-protein recruitment assays report on molecular changes close to the receptor, impedance measurements are very distal integrating over all signaling cascades that might be involved as discussed above. As such, impedance measurements might be affected by functional selectivity whereas the other two readouts are not.

The observed impedance increase may report on two potential changes in cell morphology that are indistinguishable from the current data: (i) Strengthening of cell-cell interactions which leads to a reduced width of the intercellular cleft; (ii) enhancement of cell-matrix interactions which reduces the width of the narrow electrolyte-filled channel between lower cell membrane and electrode surface; or (iii) both. In either case, the pathways for ionic current flow are reduced which translates into an increase in impedance. Recording the impedance at a set of distinct frequencies instead of just one would allow for a clear assignment of the morphology changes¹⁸ but was not in the scope of this study.

Real-Time Monitoring of Light-Controlled Y₄R Agonism

From the dose-response studies (Figure 6) the optimum concentrations for 28 (50 nM) and 30 (1 nM) were selected to perform in situ switching from one isomer to the other while the impedance is continuously monitored. Figure 7 summarizes the outcome of these in situ switching experiments for all four compounds: (A) Z-28, (B) E-28, (C) Z-30 and (D) E-30. The reader may take a look at the graphical abstract for a visualization of the experiment. For easy comparison, each graph contains the time courses of impedance for 100 nM hPP (solid black line) and the 0.01 % (v/v) DMSO vehicle control (dashed black line), which define the upper and lower limit of impedance values in this assay. Time course data for the pure isomers are given in red (Z) or blue (E). The green curve represents the isomer that was switched in situ by irradiation with the proper wavelength at the time indicated by the vertical dotted lines. When Z-28 (Figure 7A) was added to the CHO cells at time zero, impedance increased to about 80 % of the values recorded for 100 nM hPP in line with the dose-response data. Irradiating the system with 455 nm after 20 min of agonist exposure induced a sharp decrease of impedance and stabilization at values recorded for E-28 at this time point. Impedance did not change for almost 50 min indicating that the cells were equilibrated with these conditions. Subsequent irradiation with 340 nm almost 80 min after initial ligand exposure led to a pronounced increase again that does not quite reach the impedance levels of pure Z-28 but stabilized again. Repeating this sequence of light-induced isomerization yielded similar cell responses with some signs of fatigue within the reversibility of the signalling cascades or cell physiology in general as the signal changes became smaller. We take the fact that the impedance of all cell populations slightly drifted to lower values along the experiment as an indicator, that the prolonged stimulation in a buffer with a minimum of nutrients tires the cells out. In the same line, we have previously observed reduced cell responses to GPCR stimulation when we applied a cumulative dosing protocol during which the cells were sequentially exposed to increasing agonist concentrations.^4e According to the in vitro characterization of the ligands, the observed phenomenon is not due to a limited reversibility of photo-induced isomerization but we cannot exclude changes in photostability in the binding pocket of the receptor. Figure 7B summarizes the opposite experiment. This time E-28 (green curve) was applied to the CHO cells before the system was irradiated with 340 nm to induce isomerization. Under these conditions impedance did not increase to values recorded for the Z-isomer at the same time but showed a minor rise that stabilized close to the values of the pure E-isomer. After the opposite in situ switching using 455 nm, the impedance dropped significantly as expected when going from Z to E and stabilized at a new steady state. When the switching cycle is repeated, the impedance increased and decreased as expected from dose-response data even though the absolute values showed some offset relative to the isomers that were exposed to light prior to the experiment. Figures 7C and D summarize similar in situ irradiation experiments for compound 30 demonstrating the reversible switching between two distinct agonistic activities in this assay monitoring the integrated cell response.

In these experiments the recorded impedance showed some drifting to lower values along the observation time as well. Since (i) the experiments were performed in serum-free buffer with a minimum of nutrients and (ii) impedance measurements are rather sensitive to morphological changes coupled to energy metabolism, we assign the impedance drift to assay conditions and length of experiment. In control experiments we carefully ensured that there neither is an impact of the light itself (i) on resting cells (ii) nor on cells stimulated with a non-photochromic ligand (iii) nor on the electrodes that may lead to misinterpretations. We therefore performed in situ irradiation with the intensities and wavelengths as used for photoswitching with confluent CHO-Y₄R cells on the electrodes, but in absence of the photochromic ligands (Figure S59A, Supporting Information) or in presence of the ligands but without cells on the electrodes (Figure S59B, Supporting Information). In a third control experiment cells were stimulated with the non-photochromic endogenous ligand hPP and irradiated with light of the different wavelengths (Figure S59C, Supporting Information). In either case, impedance recordings did not show any significant change upon irradiation with light of wavelength 340 nm, 455 nm or 528 nm. Neither the cells without photochromic ligands nor the ligands without cells induced any measurable impedance shift upon light irradiation supporting our understanding that the individual interactions of the photochromic ligands with Y₄R are responsible for the distinct impedance profiles.

A physiological interpretation of the impedance data during the switching cycles requires a closer look into the details of the Y₄R-dependent signal transduction. The canonical coupling of the Y₄R is reported to rely on G_α,i signalling and was verified by the miniG_si recruitment assay. Once G_α,i is released from the heterotrimeric G-protein, it binds to adenylate cyclase (AC) and inhibits cAMP formation. At the same time, cAMP is constantly degraded by cAMP-dependent phosphodiesterase (PDE). In resting cells constitutive activity of AC and PDE establish a dynamic equilibrium of cAMP production and degradation yielding a steady state concentration of cAMP. This energy consuming process is seemingly useless but allows very quick changes of cAMP concentrations in response to external triggers. Since cAMP-levels are low (<50 nM)¹⁹ in resting cells and get even lower from G_α,i activity, it is generally difficult to monitor Y₄R activation by reading cAMP concentrations or downstream signalling events. Accordingly, it is common practice to pre-stimulate the cells by micromolar concentrations of forskolin, a receptor-independent, membrane permeable activator of AC isolated from plants. Forskolin increases intracellular cAMP concentrations to a level that facilitates experimental analysis of signalling mechanisms that lead to its reduction.²⁰ Thus, a similar protocol was applied in impedance-based assays as performed in this study. Prior to adding the ligands, the cells were pre-stimulated by 0.4 μM forskolin. The subsequent forskolin-induced cAMP net production led to a significant decrease of the cell impedance by 400 Ω from values recorded for the resting cells. After the cells had equilibrated to these conditions within 30 min, the different ligands were applied and monitored for their impact on cell impedance (Figure 6, 7). Since forskolin concentrations were kept constant in all buffers and all phases of the experiment, release of G_α,i from the complete G-protein after receptor activation led to a competitive regulation of adenylate cyclase. Whereas forskolin—constantly present in all cell culture fluids—led to an activation of AC, G_α,i reduced its activity in a concentration-dependent manner. So, when the Y₄ receptor and the associated signalling cascade were fully activated by the endogenous ligand hPP or high concentrations of the synthetic ligands, cAMP is efficiently reduced by over-compensating the forskolin activity bringing impedance back up to values of the resting cells or higher. Any partial activation of the receptor led to a reduced release of G_α,i activity and thus, to a competitive regulation of AC between forskolin and G_α,i that may be simply controlled by their individual concentrations and activities. If this interpretation applies, the in situ switching between the isomers and the associated change in receptor activation led to a fine-tuning of AC activity and all processes dependent thereof explaining the reversibility of the cell response during switching. For the CHO cells studied here, it seems that impedance measurements mirror the time-dependent cAMP concentrations inside the cell most likely via a cAMP-regulated protein or protein network that is involved in cell shape control. The response time of impedance readings to forskolin or agonist addition is in the order of 10–20 minutes. Accordingly, the cell response as reported by impedance is not dependent on changes in gene expression as the latter would require more time. The cell response is more likely determined by changes in functional activity of existing proteins.

Photoswitching Occurs inside Receptor Binding Pocket

In the experiments described in preceding sections, the ligands were added to the cells at time zero and were continuously present throughout the experiment. The question arose whether isomerization of the ligands upon irradiation occurs (i) in the bulk phase with subsequent competition for the receptor binding site between the bound ligand and its free isomer in solution or (ii) whether isomerization occurs in the binding pocket of the receptor without any ligand exchange. The average residence time for similar, radiolabeled Y₄R ligands was found to be rather dependent on the experimental conditions and ranged between 5 and 30 min^14a so that either mechanism is compatible with the impedance time courses. To gain more insight, we revisited the two isomers Z-28 and E-28 but adapted the experimental workflow by including a washing step after the ligands were incubated with the cells for 20 min (Figure 8).

The time interval needed for washing is indicated by the orange box. The colour code of the impedance time courses is as before. The endogenous ligand hPP (100 nM) is represented by (i) a black solid line when the liquid was not exchanged or (ii) an orange solid line, when the liquid with the free ligand was replaced by buffer. The vehicle control (DMSO, 0.01 % v/v, no washing) is indicated by a black dashed line. The impedance response for Z-28 is shown in red, the one for E-28 in blue. Both isomers were applied in 50 nM concentrations as before. The green curve corresponds to Z-28 in Figure 8A and E-28 in 8B. Only the cell population represented by the green curve was exposed to light (A: Z-28, 455 nm; B: E-28, 340 nm) at the time indicated by vertical dotted lines. As suggested by the red and blue curves in Figures 8A and B, washing led to a significant decrease of the impedance presumably due to a loss of a fraction of ligands from the receptor binding pockets when the bulk concentration was experimentally set to zero by washing. The mechanical load on the cells imposed by liquid handling may also contribute to the observed impedance decrease. In situ switching of the remaining Z-28 to the less potent E-28 after 40 min led to a significant and step-like reduction of impedance which is in line with the dose-response relationship of both ligands (Figure 8A). Re-isomerization to Z-28 was induced by irradiation with 340 nm after 70 min of experimental time. It was associated with an increase in impedance. However, values did not recover to those of the Z-isomer that has not been switched at all. When the corresponding experiment was performed with E-28 (Figure 8B), the outcome was similar. Initial isomerization of E-28 to Z-28 increased impedance almost to the values of the unperturbed Z-isomer. Re-isomerization by irradiation at 455 nm reduced impedance back to values of the unperturbed E-isomer. It is noteworthy that in these experiments no ligand was present in the bulk phase. Any dissociation from the binding pocket led to a practically infinite dilution so that any re-association of a formerly bound ligand back into the binding pocket was elusive. Conducting this washing/in situ photoswitching experiment with compounds E-30 and Z-30 returned very similar impedance time courses (data not shown). Taken together, these experiments demonstrate that in situ switching of the signalling cascade is possible even after unbound ligands have been washed out. Accordingly, the data provides evidence that isomerization is likely to occur with the ligand bound in the binding site of the receptor.