Exploring Photoswitchable Binding Interactions with Small‐Molecule‐ and Peptide‐Based Inhibitors of Trypsin

The ability to photochemically activate a drug, both when and where needed, requires optimisation of the difference in biological activity between each isomeric state. As a step to this goal, we report small‐molecule‐ and peptide‐based inhibitors of the same protease—trypsin—to better understand how photoswitchable drugs interact with their biological target. The best peptidic inhibitor displayed a more than fivefold difference in inhibitory activity between isomeric states, whereas the best small‐molecule inhibitor only showed a 3.4‐fold difference. Docking and molecular modelling suggest this result is due to a large change in 3D structure in the key binding residues of the peptidic inhibitor upon isomerisation; this is not observed for the small‐molecule inhibitor. Hence, we demonstrate that significant structural changes in critical binding motifs upon irradiation are essential for maximising the difference in biological activity between isomeric states. This is an important consideration in the design of future photoswitchable drugs for clinical applications.


Introduction
[3][4] This can be achieved photochemically using an approach referred to as photopharmacology. [2]Here a molecular photoswitch is incorporated into the structure of a drug, the geometry of which can be reversibly controlled on irradiation with light of a defined wavelength.The associated change in structure on irradiation with light must influence the molecule's interaction with its target, to modulate biological activity. [5,6]Such photoswitchable drugs have been reported to target ion channels, [7] receptors, [8] and enzymes, [9] for treatment of diseases such as cancers [10,11] and bacterial infections. [12]espite reports of numerous such photoswitchable drugs, [6,13] and also novel ways to deliver the required light source where and when required, [13,14] most reported exam-ples only show a modest two-to tenfold difference in biological activity between isomers. [6]More recent studies have reported compounds with a > 20-fold difference, [15] however still only a few reported photoswitchable drugs have been trialled in animal studies [13] and there are even fewer examples of human clinical trials. [16]It remains unclear how large the difference in activity needs to be between isomeric states at the cellular level to produce a significant and beneficial difference in physiological effect in humans.However, maximising this difference-with the "on" state showing strong potency and the "off" state showing minimal activity-remains of crucial importance to this field of research and future clinical applications of photoswitchable drugs.
A better fundamental understanding of how photoswitchable drugs interact with their biological target is needed to optimise this fold-difference.In this paper, we explore this concept with the design and testing of a series of smallmolecule inhibitors and peptide-based inhibitors that target the same enzyme.Trypsin was selected for this study, as its structure is well-studied and characterised by crystallography. [17][21][22] This enabled us to study the structural interaction between our inhibitors and the known 3D structure of trypsin, and also aided the design of compounds which feature modifications on reported inhibitors as discussed below.The inclusion of both small-molecule and peptidic inhibitors of the same protease allowed the effect of size, 3D structure, mechanism and binding interactions on photoswitchable inhibitory activity to be systematically investigated.
We chose spiropyran and azobenzene as photoswitches in this study (Figure 1).The spiropyran component of the small molecule inhibitors 1, 2, 5 and 6 incurs a large change in structure and polarity upon switching, from the uncharged spiropyran (SP) state to the zwitterionic merocyanine (MC) state after irradiation with UV light. [23]Azobenzene was chosen as the switch for the remaining small molecule inhibitors 3, 4 and 7, and the peptidic inhibitors p1-p6, as its switching between states is known to be particularly efficient and quantifiable, [24] whilst affecting a significant change in geometry and polarity suitable for influencing binding to trypsin.
The design of small molecules 1-7 and peptides p1-p6 was based on well-studied and well-characterised inhibitors of trypsin reported in the literature. [19,20,25]These known inhibitors display potent anti-trypsin activities, have published co-crystal structures, and are also amenable to synthetic modification and therefore the inclusion of a photoswitchable group.The small-molecule series is described below, followed by the peptidic inhibitors.

Small molecules
The structures of the target inhibitors are shown in Figure 2. Compounds 1 and 3 are based on traditional small-molecule protease inhibitors, [26] with a methyl ester electrophile, an arginine residue at the pseudo P1 position, and the photoswitch bound via an amide bond. [26]Trypsin's S1 pocket contains a negatively charged aspartic acid residue (Asp189) and preferentially accommodates positively charged residues, specifically arginine. [18]Compounds 2 and 4 have the photoswitch bound via an amide linkage to a benzamidine unit.Benzamidine is well studied in the literature as a simple trypsin inhibitor, known to bind in the S1 pocket of trypsin with a K i value of 19 μM. [19]he inclusion of benzamidine was probed further in the design of 5 and 7, where the benzamidine unit is incorporated directly onto the photoswitch, as part of the phenyl rings of the azobenzene or spiropyran moiety.It is noteworthy that the structure of 5 in its MC form very closely mimics the most potent reported trypsin inhibitor in the literature (15566, Figure 2) [20] which has a reported K i value of 13 nM.Derivative 5 was therefore predicted to be the most active of the series, with the MC form of each spiropyran compound predicted to be the more potent isomer.Finally, 6 was designed to be structurally similar to 5, but with the benzamidine unit instead incorporated onto the phenyl ring on the chromene side of spiropyran, to examine the effect of inserting the biologically active motif at the opposite end of the photoswitch.Both 5 and 6 lack a methyl substituent on the indoline nitrogen, a common feature in spiropyran-based photoswitchable molecules. [23]This deviation was made to better mimic the structure of 15566, where crystal structures show the NÀ H on the indole of 15566 forms a hydrogen bond to Ser195 on trypsin which is important for binding. [20]he syntheses of compounds 1-7 are outlined in Schemes 1, 2, 3 and 4 and described further in the Experimental Section.Compounds purified by RP-HPLC were isolated in diminished yield.
Compound 8 was formed by esterifying Boc-l-arginine hydrochloride hydrate with thionyl chloride in methanol.Spiropyran 10 was synthesised by reacting 9 [26] with 2hydroxy-5-nitrobenzaldehyde and TEA in anhydrous ethanol at reflux.Spiropyran 10 was then coupled to 8, using HCTU and DIPEA in DMF, to give 1 (13 % yield).Derivative 10 was also coupled to 4-aminomethyl benzamidine dihydrochloride, using HATU and DIPEA in DMF, to give 2 (10 %).Azobenzene 11 was prepared by reacting 4-aminobenzoic acid with nitrosobenzene in AcOH at 40 °C.The product of this reaction was then coupled to 8 (using the conditions described above for SP 1) to give 3 (1 %) and 4-aminomethyl benzamidine  Only one photochemical isomer, as defined in Figure 1, is shown.Inset: Structures of literature trypsin inhibitors benzamidine and 15566 [19,20] and MC isomer of 5.
Following literature procedure, [27] a solution of 4-aminobenzamidine dihydrochloride in 6 M aqueous HCl was reacted with sodium nitrite in water, followed by tin(II) chloride, to give hydrazine 12. Indoline 13 was formed on reaction of 12 with 3-methyl-2-butanone and sulphuric acid in ethanol.The amidine was Boc-protected, using 1 M aqueous NaOH, THF, H 2 O and Boc 2 O, to give 14.SP 15 was then formed by reacting 14 with 2-hydroxy-5-nitrobenzaldehyde and TEA in anhydrous ethanol at reflux, in 36 % yield.Multiple conditions were trialled to achieve successful Bocdeprotection of 15, with 25 % TFA in DCM for 10 min proving optimal.Purification of the crude material, by RP-HPLC, gave SP 5 in 25 % yield.Imidate ester 16 was formed in a Pinner reaction; [28] 3-formyl-4-hydroxybenzonitrile in EtOAc was added to a solution of acetyl chloride in ethanol, at À 78 °C, and the mixture left in the fridge for 5 days, to give 16.This was then reacted with a solution of ammonia in methanol to give amidine 17, which was subsequently treated with 2,3,3trimethylindolenine and TEA in ethanol.The product SP 6 was isolated in 4 % yield after purification by RP-HPLC.Azobenzene 7 was formed, in one step, by reacting 4aminobenzamidine dihydrochloride with nitrosobenzene in AcOH at reflux (1 %), as depicted in Scheme 4.
Compounds 1-7 were separately irradiated for 15 min (352 nm) in DMSO and Tris buffer, and the resulting photostationary states analysed by absorbance and fluorescence.NMR spectra were also collected in CD 3 OD before and after irradiation for 1 h at 352 nm.Absorbance and fluorescence measurements provide qualitative information on switching, while integration of NMR signals was used to quantify the ratios of isomers.These data are presented in full in Figures S1-S4 in the Supporting Information, and are sum-marised in Table 1, for the thermally adapted state (TAS, before irradiation) and photostationary state (PSS, after irradiation).
NMR data-supported by absorbance and fluorescence data-show the TAS and PSS distributions of isomers varied greatly between compounds (Table 1).In particular, azobenzenes 3, 4 and 7 switched more efficiently than spiropyrans 1, 2, 5 and 6.Compound 4, with its benzamidine group bound to the photoswitchable azobenzene unit via an amide bond, showed the greatest difference in isomeric composition between the two states with 89 % trans-4 in the TAS and 82 % cis-4 in the PSS.
Each separate isomeric state (TAS and PSS) of compounds 1-7 was assessed for anti-trypsin activity, using bovine trypsin and Boc-Phe-Ser-Arg-AMC as the fluorogenic substrate.IC 50 values were calculated from a plot of measured initial reaction velocity relative to the positive control against the logarithm of inhibitor concentration (Figure S8).These IC 50 values were then used to estimate K i values of competitive inhibition, [29] as described in the Experimental Section and summarised in Table 2.
The TAS and PSS of compounds 1-7 all show reasonable inhibition of bovine trypsin, with low-/sub-micromolar activities.With the exception of 6 and 7, each compound is more potent than the known inhibitor benzamidine (K i = 19 μM) in both isomeric states. [19]Spiropyran 2 proved to be the most potent inhibitor of the series in both the TAS and PSS, however the observed activity difference between states is only 1.2-fold.This is likely due to modest switching of this  compound upon irradiation, as revealed by the absorbance and fluorescence data (Figure S1).The PSS for both spiropyrans 5 and 6 show decreased activity compared to the respective TAS.These results are consistent with observations made from the earlier NMR spectra (Table 1), where the MC isomer decreases in concentration after UV irradiation.We expected the MC isomers to be more potent than the SP isomers for the spiropyran compounds; therefore, the TAS for spiropyrans 5 and 6 showed better anti-trypsin activity than the PSS due to a higher concentration of MC isomer.Azobenzene 4 showed the largest difference in activity between the two isomeric states (3.4-fold)-consistent with it showing the largest difference in TAS/PSS compositions-but weakened anti-trypsin activity compared to spiropyran 2.
With the exception of 6, spiropyrans 1, 2, and 5 all displayed improved anti-trypsin activity compared to the azobenzenes 3, 4, and 7, likely due to structural similarity of the MC isomers with known trypsin inhibitor 15566. [20]owever, the azobenzene compounds provided a larger difference in activity between isomeric states-up to 3.4-fold, compared to maximum 2.2-fold for the spiropyran inhibitors.This is likely due to the larger differences in isomeric composition between the two states for the azobenzenes.Therefore, the azobenzene small molecules are better photoswitchable inhibitors of trypsin than the spiropyran small molecules.
Azobenzene 4, with the largest difference in inhibitory activity between its two isomeric states (3.4-fold), was docked into the bovine trypsin active site to investigate the likely binding conformations of the TAS and PSS [30] (Figure 3).Trans-4 was used to represent the trans-enriched TAS, and cis-4 the cis-enriched PSS.ICM software was used to dock both isomers of 4 onto bovine trypsin, co-crystallised with benzamidine (PDB ID: 3ATL [30] ).3ATL shows benzamidine binds in the S1 pocket of trypsin, with hydrogen bonds between the amidine group and Asp189 as expected. [18,30]igure 3 shows the benzamidine motif of 4 also binds into the S1 pocket (close to Asp189), however the azobenzene moiety does not appear to fit inside this pocket and is as such aligned outside it.Hence, the azobenzene group is unlikely to significantly impact the binding affinity, which is known to depend heavily on interactions inside the S1 pocket. [18]The benzamidine motif, which is bound inside this pocket, is thus likely to be the main contributor toward the activity of the inhibitor.The positioning of the benzamidine group inside the S1 pocket is very similar in both trans-and cis-4.
Therefore, we postulate that the observed modest 3.4fold difference in affinity between the two isomeric states of 4 is likely due to the switchable azobenzene moiety neither being directly involved in binding, nor changing the confirmation of the benzamidine group which is responsible for the anti-trypsin activity.To manifest a larger difference in biological activity between isomeric states, significant structural changes upon irradiation are required in the moieties essential to binding.As such, we believed that a smallmolecule-based approach to inhibitors (as in 1-7) would not exploit the small size of the active site (S1 pocket) in trypsin. [30]e thus considered an alternative approach to photoswitchable inhibitors in order to optimise photoswitching and differences in activity.Peptidic-based inhibitors were next investigated as described in the following section.

Peptides
Peptides p1-p6 (Figure 4) were designed based on inhibitor p-SFTI [25] (Figure 4, inset), a reported cyclic peptidic trypsin inhibitor, that itself is based off sunflower trypsin inhibitor 1. [25] Analogues of p-SFTI1 containing a photoswitchable diarylethene moiety have recently been reported with effective modulation of biological activity between isomers, [31] demonstrating the suitability of this peptide for photoswitchable studies.p-SFTI has a reported K i value of 103 nM-its activity is due to its rigid secondary structure, due to a well-defined β-hairpin loop. [25]Literature suggests that a cis-azobenzene mimics a β-turn in peptides. [32]Therefore, we postulated that replacing the l-Pro5-d-Pro6 β-turn motif of peptide p-SFTI with an azobenzene unit could enable photocontrol of trypsin activity.The cis isomer of the photoswitch should position the peptide into the active, βhairpin structure, while the trans isomer should position the Figure 3. Docked conformations of trans-4 (left, orange) and cis-4 (right, purple) in bovine trypsin (PBD ID: 3ATL [30] ).
peptide into an unfolded structure that doesn't bind as well to trypsin.
The target peptides p1-p6 have the key azobenzene unit incorporated into the peptide backbone, at different locations around the ring.Peptides p1 and p2 have the azobenzene in place of the l-Pro5-d-Pro6 resides of p-STFI, peptides p3 and p4 have the azobenzene replacing the Ile4-Pro5 residues, and p5 has the azobenzene replacing the Ile4d-Pro6 residues.The azobenzene linkages to the rest of the macrocycle were also varied to determine the effect on activity and switching; peptides p1, p3 and p5 all have para, meta linkages to the rest of the macrocycle, while p2 and p4 have meta, meta linkages.Finally, the linear peptide p6 was designed to validate the importance of the macrocycle for activity-this peptide represents the linear version of macrocycle p2.
Peptides p1-p6 (Figure 4) were prepared by solid phase peptide synthesis (SPPS) on 2-chlorotrityl functionalised polystyrene resin, as described in the Experimental Section.Scheme 5 summarises the synthesis of peptide p2 as a representative example, with peptides p1-p6 presenting different amino acid sequences and/or azobenzene substitutions (meta, meta or para, meta).The component azobenzene units of p1-p6 were prepared and incorporated into the peptide sequence as described in literature. [33]SPPS gave the linear sequences of each peptide, which were subsequently cleaved from the resin.Linear peptides p1-p5 were irradiated at 352 nm for 18 h to give the corresponding cis isomers, which brings the termini closer to facilitate cyclisation on subsequent treatment with HATU.Cyclisation was performed overnight under further UV irradiation.The component tBu and Boc protecting groups were removed on treatment with TFA in DCM to give the desired cyclic peptides p1-p5, which were purified by RP-HPLC.Peptide p6 was purified by RP-HPLC in its linear form.
Solutions of peptides p1-p6 were then separately irradiated for 15 min at 352 nm, and analysed using the methodology described above for the small molecules.Each peptide was switched in DMSO and Tris buffer for analysis by absorbance spectroscopy, and in CD 3 OD for NMR analysis.These data are summarised in Table 3, with full data presented in Figures S5-S7.
The absorbance data show efficient switching from the trans to cis isomer upon UV irradiation for each peptide p1-p6, which is consistent in the NMR data.In all peptides, there is � 74 % trans isomer in the TAS, and the azobenzene switches efficiently to � 85 % cis isomer in the PSS.Of all the peptides, p6 gave the highest % trans in the TAS and PSS, which indicates that the linear structure of p6 favours a trans conformation more so than the cyclic structures of p1-p5.There is no clear trend in the effect of a para, meta substitution (of peptides p1, p3 and p5) compared to a meta, meta substitution (p2 and p4) on the PSS or TAS isomeric distributions.
Peptides p1-p6 were next assayed against bovine trypsin, as described earlier for the small molecules.The K i values were calculated for all peptides in both isomeric states, as  [25] S9.Across all peptides, there is no clear trend in potency of inhibitors or fold-difference between states based on meta, meta or meta, para substitution of the azobenzene.Peptide p5 lacked anti-trypsin activity, suggesting inclusion of the azobenzene at the Ile4-d-Pro6 position on the macrocycle is detrimental to inhibition.Linear peptide p6 also lacked activity against trypsin, compared to its cyclic and particularly potent analogue p2.This is consistent with the known importance of the macrocycle in peptidic inhibitors of trypsin. [25]Peptide p2 was the most active inhibitor of the series with a K i = 7.4 μM/2.56 μM in the TAS/PSS; however, p3 showed the largest fold-difference between states (> 5 times).This large activity difference, due to potency of 16.1 μM in the PSS compared to no activity (> 86 μM) in the TAS, indicates that peptide p3 is the best photoswitchable inhibitor from this peptide series.
The more than fivefold activity difference observed for peptide p3 is larger than the 3.4-fold difference for the optimal small molecule 4 described above.This suggests that any structural change resulting from irradiation of peptide p3 has more of an effect on the binding affinity to trypsin compared with the small molecules.
Peptides p2 and p3 were modelled in the lowest-energy conformations using the NWChem 6.6 package-described further in the Experimental Section-for both trans and cis isomers (representing the TAS and PSS, respectively; Figure 5).These were then compared to the lowest-energy conformation of the known inhibitor p-SFTI (Figure 5) to better understand the trends in K i data.In p-SFTI, the β-turn is induced by the l-Pro-d-Pro residues, [25] and is observed on the right-hand side of the structure in Figure 5.A cisazobenzene moiety in cis-p2 mimics the β-turn in p-SFTI and aligns the other residues of cis-p2 very similarly to p-SFTI.This supports the experimental K i data which shows that p2 is the most potent peptidic inhibitor of the series.In contrast, the trans-azobenzene group in trans-p2 does not appear to mimic a β-turn with the overall macrocyclic structure differing significantly from p-SFTI, resulting in reduced activity compared to cis-p2.Interestingly, despite these structural differences, the fold-difference in activity between the TAS/PSS of p2 is not as large as that observed for peptide p3.In p3, the β-turn (induced by the l-Pro-d-Pro residues [25] ) is retained in both isomers; the azobenzene was placed on the opposite side of the macrocycle to the β-turn to determine its effect at this position.As can be seen in Figure 5, the overall 3D structures of trans-and cis-p3 are very different to one another.Notably, the location of residues Thr1 and Pro6, and the overall rigidity of the macrocycle, varies greatly between the two isomers.This results in a fivefold difference in biological activity, suggesting that for peptides p1-p6, not only is a βturn important for binding to trypsin but so is the orientation of other residues and the overall rigidity of the macrocycle.
The modelling and K i data discussed above suggest that, at least for trypsin, macrocyclic peptides make better photoswitchable inhibitors than small molecules, as they can induce a larger difference in biological activity between isomeric states.The inhibitory activity of peptides p1-p6 is defined by the overall shape of the macrocycle and alignment of multiple residues, compared to inhibition by benzamidine-based small molecules 1-7, which depends solely on interactions inside the P1 pocket.Trans/cis isomerisation in the peptides, as in p3, therefore causes a larger fold-difference in activity, due to more significant effects on 3D structure.

Conclusions
In this work, a series of small-molecule photoswitchable inhibitors (1-7) and a series of cyclic peptidic photoswitchable inhibitors (p1-p6) of the protease trypsin were designed, synthesised and assayed in two isomeric states, the TAS (before irradiation) and PSS (after irradiation).Spiropyran 2 was found to be the most potent small-molecule inhibitor, with a K i of 1.02 μM in the TAS and 0.82 μM in the PSS, while azobenzene 4 had the largest difference (3.4-fold) in antitrypsin activity between isomeric states for small molecules 1-7.However, in-silico docking of 4 with trypsin showed the azobenzene unit was unlikely to be directly involved in binding in the S1 pocket, and hence the small-molecule approach to trypsin inhibitors is unlikely to result in an optimum photoswitchable inhibitor.Peptide p2 showed the most potent activity of the peptide-based inhibitors, with K i = 7.4 μM in the TAS and 2.56 μM in the PSS; however, p3 showed the largest fold-difference between isomeric states (> 5 times).Lowest-energy-confirmation modelling suggests that isomerisation between trans-p3 and cis-p3 induces a very large change in the overall 3D structure of peptide p3.Therefore, the cyclic-peptide approach is more favourable for photoswitchable inhibitors of trypsin, as the large structural change in the macrocycle upon irradiation results in a greater difference in biological activity between isomeric states.
This work demonstrates that many structural factorssuch as the positioning of key binding residues and overall structural rigidity-must be considered when designing a photoswitchable inhibitor in order to optimise the difference in biological activity between isomeric states.This can be achieved by designing the inhibitor to significantly change 3D structure upon irradiation in a way that affects binding (e. g., in a macrocyclic peptide); or by incorporating the most important functional group for binding close to the switch, so that it changes conformation significantly upon isomerisation.Maximising the fold-difference in activity ensures the photoswitchable inhibitor acts as a viable "switch", thereby helping to bring these drugs closer to a future clinical application.

Experimental Section
Unless otherwise stated, all starting materials, enzymes, reagents and solvents were purchased from Merck (NSW, Australia) and were used without further purification.All amino acids, coupling reagents and resins were purchased from Chem-Impex International, Inc (IL, USA).All amino-acids are l-configuration unless otherwise specified. 1H and 13 C NMR spectra were recorded on a Varian 500 MHz/Varian Inova 600 MHz instrument in the indicated solvents.Chemical shifts are reported in ppm (δ).High resolution mass spectra were recorded on an Agilent 6230 TOF LC/MS fitted with an Agilent 1260 Infinity HPLC unit.Where indicated, compounds were analysed and purified by reversedphase (RP-) HPLC, using an Agilent 1260 Infinity LC system equipped with a Phenomenex Luna C-18(2) column (250×4.6 mm) for analytical traces; and a Gilson GX-Prep HPLC system equipped with a Phenomenex Luna C-18(2) column (250×10 mm) or a Phenomenex Phenyl-Hexyl column (250×10 mm) for semi-preparative purification.H 2 O/TFA (100 : 0.1 v/v) and MeCN/TFA (100 : 0.1 v/v) were used as aqueous and organic buffers.All graphs were generated using GraphPad Prism 9 software.A Sankyo Penki F8T5 352 nm 8 W bulb was used for photoswitching experiments.A Synergy H4 Hybrid Multi-Mode Microplate Reader was used with black walled, clear bottomed 96 well plates (Corning, Australia) for photoswitching experiments and trypsin assays.

4-(Phenyldiazenyl)benzimidamide (7):
4-Aminobenzamidine • 2HCl (100 mg, 0.48 mmol) and nitrosobenzene (77 mg, 0.72 mmol) were suspended in AcOH (5 mL) and the mixture heated at reflux for 24 h.The mixture was diluted with water (10 mL), basified to pH 14 with 4 M NaOH and extracted with EtOAc (×3).The aqueous layer was filtered to remove the solid residue and then re-extracted with EtOAc (×2).The organic extracts were combined and washed with NaHCO 3 followed by water and brine, then dried over MgSO 4 , filtered, and concentrated under reduced pressure.The crude product was purified by RP-HPLC (20-100 % MeCN/H2O) to afford 7 as a pink solid (1 mg, 1 %; 6 : 1 ratio of E : Z isomers). 1 12): 4-Aminobenzamidine dihydrochloride (200 mg, 0.96 mmol) was suspended in 6 M HCl (3 mL) and the mixture cooled in an ice-water bath.A solution of sodium nitrite (73 mg, 1.06 mmol) in water (1 mL) was added dropwise to the reaction mixture with vigorous stirring.The mixture was stirred in the ice-water bath for a further 30 min.Tin(II) chloride (395 mg, 1.92 mmol) was added in portions to the reaction mixture, which was stirred for 20 min and allowed to warm to RT.The solvent was removed under reduced pressure and the residue triturated with ethanol.The solid was collected by suction filtration and washed with absolute ethanol to afford 12 as an off-white solid (215 mg, quant.). 1

3',3'-Dimethylspiro[chromene-2,2'
-indoline]-6-carboximidamide (6): Acetyl chloride (4.8 mL, 68 mmol) was added dropwise to anhydrous ethanol (4.4 mL, 75 mmol) cooled to À 78 °C under N 2 and the solution stirred for 5 min.A suspension of 3-formyl-4hydroxybenzonitrile (100 mg, 0.68 mmol) in EtOAc (5 mL) was then added to the reaction mixture under nitrogen at À 78 °C, which was stirred vigorously for a further 10 min.The mixture was warmed to 5 °C and left at 5 °C for 5 days, after which it was warmed to RT and opened to atmosphere.EtOAc (15 mL) was added to the mixture, and the precipitate was collected by suction filtration.The solid was washed with EtOAc (30 mL) and air dried to afford 16 as a pale pink solid (88 mg).The crude product was used in the following reaction without further purification.
7 M ammonia in MeOH (2 mL, 14 mmol) was added dropwise to a vigorously stirring solution of 16 (60 mg, 0.26 mmol) in MeOH (1 mL).The mixture was stirred for 5 days, and then the solvent was removed under reduced pressure to afford crude 17 as a white powder (30 mg) which was used in the following reaction without further purification.
To a solution of 17 (14 mg, 0.09 mmol) in anhydrous ethanol (1 mL) was added 2,3,3-trimethylindolenine (14 μL, 0.09 mmol) and TEA (44 μL, 0.09 mmol).The mixture was heated at reflux under N 2 for 18 h.The solvent was removed in vacuo and the crude product purified by RP-HPLC (20-70 % MeCN/H2O) to afford 6 as a yellow solid (1 mg, 4 %; 2.5 : 1 ratio of SP:MC isomers).General synthesis of linear peptides by SPPS: 2-Chlorotrityl chloride functionalised polystyrene resin (Chem Impex Intl.Ltd., 1.4 mmol g À 1 , 1.0 g) was allowed to swell for 15 min in DCM (15 mL).Fmoc-Ser(tBu)-OH (1.0 mmol) was dissolved in DCM (10 mL) and added to the swollen resin, followed by DIPEA (5.0 mmol) and the mixture agitated for 2 h.MeOH (5 mL) was then added to the mixture which was agitated for a further 10 min.The solution was drained from the resin, and the resin was subsequently washed with DCM (5×10 mL) and diethyl ether (5×10 mL) and left to air dry overnight and the loading of the Fmoc-Ser(tBu)-OH determined by a weight increase.
A portion of the FmocAA functionalised resin (0.25 mmol) was pre-swelled in a solution of 1 : 1 DCM/DMF (10 mL) for 15 min, then the solution drained from the resin.The Fmoc protecting group was then deprotected by addition of a 20 % solution of piperidine in DMF with 0.1 M HOBt (5 mL) to the resin, and stirred intermittently for 15 min.The solution was drained from the resin in vacuo, and the resin was then washed with DMF (5×10 mL), DCM (5×10 mL) and DMF (5×10 mL).Amino acid coupling was achieved by addition of a solution of FmocAA (or FmocAzo-OH) (5 equiv.)with 0.5 M solution of HATU in DMF (2.5 mL, 5 equiv.),DIPEA (435 μL, 10 equiv.) and DMF (8 mL).The mixture was added to the resin and left for 2 h, with intermittent stirring.The solution was drained from the resin in vacuo and washed with DMF (5×10 mL), DCM (5×10 mL) and DMF (5×10 mL).In the case of coupling Fmoc-Pro-OH, or coupling the residue subsequent to a proline residue, the coupling was repeated once more ("double coupling"), prior to Fmoc deprotection, to ensure complete acylation.Coupling/ deprotection steps were repeated to elongate the peptide.
The TNBS test for a free amine was employed after coupling and deprotection steps to ensure the completion of the reaction, indicated by colourless resin beads.This was performed via addition of a 20 % v/v solution of 5 % aqueous TNBS in DMF (50 μL), and a 5 % solution of DIPEA in DMF (50 μL) to a small amount of resin (micro spatula-full).The mixture was shaken for 1 minute to allow for reaction.Red beads indicated the presence of free (primary) amine, and colourless beads indicated the absence of free amines.If the coupling or deprotection reaction was not complete, then the step was repeated, including the TNBS test for a free amine.Once the desired sequence was achieved, and the final Fmoc group was deprotected, the resin was washed with DMF (5×10 mL), DCM (5×10 mL) and diethyl ether (5×10 mL).The remaining ether was left to evaporate overnight.The peptide chain was cleaved from the resin by addition of a solution of 1 % TFA in DCM (10 mL) and agitation for 1 h.The mixture was filtered (glass sinter) and the TFA solution retained.Fresh cleavage solution was added to the resin, and the reaction repeated for a further 1 h.The combined solutions were concentrated under nitrogen to approx. 1 mL volume, to which diethyl ether (10 mL) was added and the mixture was kept at 4 °C overnight.The mixture was subsequently centrifuged, and the supernatant decanted from the precipitate, and the precipitate dried under nitrogen.The dried solid was then dissolved in 30 % aqueous MeOH (~5 mg mL À 1 ) and lyophilised to give the crude product as an orange/brown solid.
General synthesis for peptide cyclisation: Crude linear peptide (~2 mg mL À 1 in DMF) was converted to the cis geometry by irradiation for 18 h with UV light (352 nm), in a quartz flask.A solution of HATU (3 equiv.),HOAt (3 equiv.)and in DMF with 1 % DIPEA (final peptide conc. of 1 mg mL À 1 ) was added to the peptide and the resulting mixture stirred under UV irradiation (352 nm) for 18 h or until cyclisation was complete as indicated by HPLC and MS.The mixture was concentrated in vacuo, taken up in DCM (15 mL) and washed successively with 1 M HCl (15 mL), saturated aqueous NaHCO 3 (15 mL) and brine (15 mL).The organic layer was concentrated in vacuo and the residue added to a solution of 1 : 1 TFA/DCM with 2.5 % each TIPS and H 2 O (10 mL) and stirred for 2 h.The mixture was concentrated under a stream of nitrogen to ~1 mL, then the peptide precipitated by addition of diethyl ether (10 mL) and the mixture chilled at 4 °C overnight.The mixture was subsequently centrifuged, and the supernatant decanted from the precipitate, and the precipitate dried under nitrogen.The dried solid was then dissolved in 30 % aqueous MeOH (~5 mg mL À 1 ) and purified by RP-HPLC to yield the cyclic peptides.
In solution photoswitching: A stock solution of each photoswitchable small molecule/peptide was made up at 10 mM in DMSO.These stocks were then diluted to 200 μM in DMSO and Tris buffer + 2 % DMSO in a 96-well, clear bottom plate for switching experiments using a Synergy H4 Microplate Reader.For all compounds an absorption spectrum was taken under visible light ("before irradiation") and after irradiation with UV light (352 nm) for 15 min ("after irradiation").For spiropyrans 1, 2, 5 and 6, a fluorescence emission spectrum was also taken (λ ex = 532 nm) before and after irradiation.These spectra look distinctly different for each isomer (SP/MC and trans/cis) and so provide qualitive information on whether switching between isomers has occurred.Experiments were performed in triplicate.
Furthermore, for all compounds except 1 and 2, NMR spectra were used to quantify the composition of the TAS and PSS.Samples were dissolved in CD 3 OD in a quartz NMR tube, and a spectrum under visible light was taken.Each sample was irradiated with UV light (352 nm) in the quartz tube for 1 h, and a further NMR spectrum taken.Integration of aromatic peaks corresponding to each isomer in each spectrum was used to determine the TAS/PSS ratios.
In vitro trypsin assay: Fluorometric trypsin assays were carried out on a Synergy H4 hybrid reader at 25 � 0.4 °C using 96-well, clear bottom plates.Tris buffer was made up containing 100 mM Trizma base with 10 mM CaCl 2 at pH 7.8 and kept at 4 °C until needed, when it was warmed to RT. Trypsin from bovine pancreas was dissolved daily in 1 mM HCl (0.1 mg mL À 1 ) and kept at 0 °C.A 1 μg mL À 1 solution of trypsin was prepared freshly before each experiment by diluting the 0.1 mg mL À 1 solution into Tris buffer.The substrate Boc-Phe-Ser-Arg-AMC (Bachem, USA) was dissolved in DMSO (40 mM) and stored at À 20 °C until needed, when it was diluted to 1 mM in Tris buffer before each experiment.10 mM stock solutions (DMSO) of each inhibitor were diluted in DMSO to the concentration required for addition to the plate.Each inhibitor was tested at 6 different concentrations [I], which were varied based on the predicted activity of each compound (see plots in Figures S13 and S14).
For the assays under visible light (TAS), 85 μL of Tris buffer was added to each well, followed by 2.5 μL of Boc-Phe-Ser-Arg-AMC and 2.5 μL of inhibitor at each concentration.Finally, 10 μL of trypsin was added to each well, making up a total volume of 100 μL (2.5 % DMSO), and each well was mixed, and the plate then read.For the irradiated assays (PSS), inhibitors were first diluted in a 96-well plate, in 50 : 50 DMSO/Tris buffer.This plate was then irradiated under UV light (352 nm), for 30 min for the small-molecule inhibitors, and 1 h for the peptide inhibitors.The rest of the assay was subsequently carried out in dark conditions.82.5 μL Tris buffer, 2.5 μL of substrate, 5 μL inhibitor and 10 μL enzyme were added to the plate, and the wells were then mixed and read.A positive control containing 95 μL Tris buffer plus 2.5 μL substrate and 2.5 μL DMSO was also run for each compound.Fluorescence emission (440 nm) was recorded every 30 s for 8 min, with excitation at 360 nm.Experiments were performed in triplicate.
The rate of enzyme-catalysed hydrolysis of Boc-Phe-Ser-Arg-AMC was calculated for the positive control (no inhibitor) and set as 100 %.The rate of reaction for each inhibitor concentration was then calculated and normalised as a % of this positive control.Log [I] against v [%] was plotted by using GraphPad Prism 9, which generated IC 50 values for each inhibitor.These IC 50 values were then used to calculate K i values for the inhibitors by using Equation (1): [29] Where [S] is 25 μM and K m = 19 μM (determined experimentally).
In silico docking: Docking experiments were performed using ICM software version 3.8-6 (Molsoft L.L.C., San Diego, CA, USA).The protein for docking was retrieved from PSCB protein data bank 3ATL. [30]Then, formal charges were assigned; protonation states of histidines were adjusted, and hydrogens, histidine, glutamine, and asparagine were optimised using the protein preparation procedure implemented in ICM. [35]The original bound ligand and all water molecules were removed from the binding site before docking.The binding site was defined as the cavity delimited by residues with at least one nonhydrogen atom within a 4.0 Å cut-off radius from the ligand 4. The pocket was represented by 0.5 Å grid maps accounting for hydrogen bonding, hydrophobic, van der Waals, and electrostatic interactions.The molecules were flexibly docked into the rigid binding site and scored based on the ICM scoring function.

Molecular modelling:
The lowest-energy structures for the trans and cis isomers of p2 and p3 were determined in the gas phase, using the NWChem 6.6 package [34] with tight convergent criteria and using a hybrid B3LYP functional with 6-31G basis set for all atoms.The structures were visualised using Discovery Studio software (Dassault Systemes BIOVIA, Paris, France).

Scheme 5 .
Scheme 5. Synthesis of peptide p2.The linear peptide was synthesised by SPPS, and subsequently cleaved from the resin, cyclised, deprotected and purified by RP-HPLC.

1, 2, 5 and 6 and azobenzenes 3, 4 and 7. Ratios are
No data for the irradiated PSS of 1 and 2 as the MC isomer is too unstable to be visualised by NMR spectroscopy.

Table 2 .
K i inhibition data for small molecules 1-7 against trypsin protease, in their respective states (TAS and PSS).
summarised in Table4.Plots of log [I] against v [%] are displayed in Figure

Table 4 .
K i inhibition data for peptides p1-p6 against trypsin protease, in their respective states (TAS and PSS).