Structure–property relationships of photoresponsive inhibitors of the kinesin motor

Recently we demonstrated the photoregulation of the activity of kinesin-1 using an azobenzene-tethered peptide (azo-peptide: Azo-Ile-Pro-Lys-Ala-Ile-Gln-Ala-Ser-His-Gly-Arg-OH). To understand the mechanism behind this photoswitchable inhibition, here we studied the structure-property relationships of a range of azo-peptides through systematic variations in the structures of the peptide and azobenzene units. The vital peptide sequence for kinesin inhibition-mediated through electrostatic, hydrophobic and C-Hπ interactions-was the same as that for the self-inhibition of kinesin. We also identified substituents on the azobenzene capable of enhancing the photoswitchability of inhibition. As a result, we developed a new inhibitor featuring a relatively short peptide unit (-Arg-Ile-Pro-Lys-Ala-Ile-Arg-OH) and an azobenzene unit bearing a para-OMe group. In the trans form of its azobenzene unit, this finely tuned inhibitor stopped the kinesin-driven gliding motility of microtubules completely at a relatively low concentration, yet allowed gliding motility with a relatively high velocity in the cis form obtained after UV irradiation.


Introduction
Kinesin is an ATP-fuelled biomolecular motor that transports intracellular cargoes (e.g., chromosomes, vesicles, organelles) to specific sites by walking processively along microtubule tracks in living cells. [1][2][3][4] Although it has dimensions of only approximately 80 nm, kinesin exhibits excellent force-generating capabilities and converts the chemical energy of ATP into mechanical work with high efficiency. [5][6][7][8] Accordingly, the kinesin-microtubule system has found application in artificial nano-transportation systems as molecular shuttles, where the kinesin motors and microtubule filaments have been suitably functionalized for the transport of various nano cargoes. [9][10][11] Such molecular transport systems can be employed in, for example, sorting, separation, and purification. 6 To enhance the applicability of kinesin as a molecular machine, more complete and precise control over its motor activity will be necessary. Several attempts have already been made in this regard; for example, using a thermo-responsive polymer, 12 electrically switchable polymer surface, 13 caged ATP, 14,15 phototunable non-nucleoside triphosphate energy molecule, 16 caged inhibitory peptides, 17 and enzyme-coated particles that can generate ATP in situ. 18 Nevertheless, each of these approaches has suffered from drawbacks related to either one-way regulation or incomplete reversible regulation. Greene et al. reported the complete inhibition of mutated kinesin in the presence of Zn 2+ ions and subsequent reactivation using a metal chelator. The percentage reactivation decreased significantly, however, after the second cycle as a result of the deactivation of kinesin, through either protein denaturation or irreversible inhibition by the Zn 2+ ions. 19 Complete and reversible regulation over many cycles without any dysfunction would be most desirable.
In living cells, when kinesin is not attached to any cargo, the tail domain bends and binds to the motor domains to avoid the unnecessary consumption of ATP. 17,[20][21][22] Inspired by this self-inhibition mechanism, we derived an inhibitory peptide sequence from the tail domain of kinesin-1 and tethered it to a photoresponsive azobenzene unit at the N-terminus to allow reversible and spatiotemporal regulation of the inhibitory activity through the effects of light irradiation. 23 The large conformational change associated with the photoisomerization of the azobenzene moiety 24-26 led to hugely different affinities for its two different isomeric forms and allowed us to switch the motility on and off completely and reversibly. Nevertheless, we have not previously been in a position to explain precisely the modes of interaction involved in the inhibition by the reverse-ordered peptide; in addition, we had not identified the critical peptide sequence required for inhibition, nor had we examined the effects of substituent groups present on the azobenzene moiety on the inhibitory activity and photoswitchability. For practical applications, we would require more efficient inhibitors that function at lower concentrations with greater photoswitchability. Recently, Kaan et al. reported the X-ray crystal structure of the motor domain dimer of Drosophila melanogaster kinesin-1 (homologue of human kinesin-1) with the tail peptide. 27 They identified the important amino acid residues and their detailed interactions (electrostatic, hydrophobic, C-H⋯π) during the self-inhibition mechanism. Interestingly, our earlier-reported inhibitory azo-peptide having the sequence -Azo-Ile-Pro-Lys-Ala-Ile-Gln-Ala-Ser-His-Gly-Arg-OH comprised the important amino acids of the tail peptide of D. melanogaster kinesin-1, but in the reverse order. This finding prompted us to examine the behavior of related azo-peptides systematically.
In this study, we determined the important amino acid residues responsible for the inhibition, identified through systematic variations in the structure of the peptide unit, and also investigated the interactions of each of these amino acids that were necessary to induce inhibition. The plausible mechanism of inhibition suggests that the interactions between our azo-peptides and the motor domain are similar to those between the tail domain and the motor domain in the selfinhibition mechanism of kinesin. Among the various substituents appended to the azobenzene unit, we found that the OMe group enhanced the switching properties. Combining these results, we rationally designed a new azo-peptide comprising a shorter peptide sequence (-Arg-Ile-Pro-Lys-Ala-Ile-Arg-OH) coupled to an azobenzene unit substituted with an OMe group at the para position. This novel inhibitor stopped the kinesindriven gliding motility of microtubules completely at a lower concentration (750 μM) in the trans form and provided a higher recovery of velocity (86%) in the motility in its cis form after UV irradiation.

Synthesis of azo-peptides
The general procedure for the synthesis of azo-peptides involved solid phase peptide synthesis using the Fmoc strategy. 23 We formed the peptide chains first and then introduced the azobenzene carboxylic acid derivatives at the N-termini through amide linkages. Fig. 1a and b present the reversible isomerization of the azo-peptide and the list of synthesized azo-peptides, respectively. The previously reported azo-peptide 1 23 was our lead molecule. The azo-peptides 2-6 were designed by removing the N-terminal amino acids one by one to confirm the importance of the -Ile-Pro-Lys-Ala-Ile unit in the inhibition mechanism. We synthesized the azo-peptides 7-11 to clarify the role of the -Ala-Ser-His-Gly-Arg unit: removing the C-terminal amino acids from azo-peptide 1 gave the  azo-peptides 7 and 8 and replacing the Gln residue in the azopeptide 8 with Gly, Lys, and Arg residues gave the azo-peptides 9-11. The azo-peptide 12 was obtained after adding an Arg residue at the N-terminus in the azo-peptide 8-the crystal structure indicated that an Arg residue interacted more strongly with kinesin motor domains than did a Gln residue, as shown later. The azo-peptide 13 was designed, with reference to the symmetrical recognition of the tail peptide by motor domains, by adding an Arg residue at the C-terminus of the peptide unit of the azo-peptide 12. The azo-peptides 14-17 featured different substituents at the para position of the azobenzene unit. Mass spectra (Fig. S1-S16 †) and HPLC chromatograms (  were reversibly isomerized from their trans to cis forms and vice versa through alternating irradiation with UV and visible lights, respectively, for at least 10 cycles of irradiation without fatigue, as evidenced by consistent UV-Vis spectral profiles. When we irradiated the azo-peptide 15 with visible light to achieve trans to cis isomerization, we could not detect the cis-rich state in its UV-Vis spectrum because the thermal relaxation of the cis isomer was faster than the acquisition speed of the spectrophotometer. We examined the thermal stabilities of the cis isomers of 14 and 16 by analyzing their absorption spectra at 25°C in the dark (Fig. S19 †). After 1 h, only around 1% of the trans forms had been recovered through thermal relaxation of the cis isomers, suggesting that thermal relaxation of the cis isomers would be negligible over the duration of a motility assay, which is usually complete within 1 h. We used HPLC to measure the conversion ratios for the trans to cis and cis to trans transformations upon irradiation with light ( Fig. S20 † displays the chromatograms). Table 1 lists the isomer ratios in the UV and visible photostationary states (PSSs) of azo-peptides having unsubstituted and OMe-and COOH-substituted azobenzene chromophores (11,14, and 16, respectively). states of the azo-peptides at various concentrations in the presence of 1 mM ATP. In the trans state of the azo-peptide 1, the gliding velocity decreased sharply at a concentration of 1.5 mM. At a concentration of 2.5 mM, there was no motility in the pre-irradiated state; in the cis-rich state obtained after irradiation with UV light, the microtubules moved with a velocity that was 64% of that in the control experiment (i.e., the velocity measured in the absence of the inhibitor). At this concentration, after irradiation with visible light, the gliding velocity decreased again, but did not reach zero because the concentration of the trans isomer in the visible PSS was insufficient to induce complete inhibition [only ca. 80% of the trans form was obtained in the visible PSS (Table 1)]. When we further increased the concentration of 1 to 3 mM, no motility was evident in either the pre-irradiated state or in the visible PSS. Thus, at a concentration of 3 mM, the motility could be switched completely and reversibly between the ON and OFF states with a 62% recovery of velocity in the motility ON state (Fig. 2a). We calculated the IC 50 values for ready comparison of the affinities of the trans and cis isomers for kinesin. For the trans isomer of 1, the IC 50 was 1.2 mM; for the cis isomer, it exceeded 3.5 mM. The azo-peptides 2, 4-7, and 9 were insoluble in the buffer; thus, we did not test them in motility assays. Table 2 summarizes the behavior of the synthesized azo-peptides in the in vitro motility assays.

Repeatability of gliding motility photoregulation
The azo-peptide 17 induced complete inhibition at a submillimolar concentration (750 μM) with 86% recovery of the gliding velocity in the motility ON state. With this azo-peptide, we performed an experiment including four cycles of alternating irradiation with UV and visible light for 5 and 10 s, respectively. Fig. 3 displays the gliding velocities of the microtubules calculated after each irradiation with UV and visible light. Prior to irradiation, all of the microtubules were in the OFF state. After irradiation with UV light, they were moving with an average velocity of 0.43 μm s −1 . After subsequent irradiation with visible light, the gliding velocity of the microtubules again reached zero. No significant decrease in the recovery velocity in the motility ON state was evident after the fourth irradiation cycle. The motility can be switched ON and OFF completely for at least 10 cycles in the in situ photoregulation experiment shown below (the data is not included here).

In situ photoregulation of gliding motility
Using the azo-peptide 17 at a concentration of 750 μM in BRB-80 buffer in the presence of 1 mM ATP, we performed a gliding motility experiment in which we recorded a video for 6 min (Movie in the ESI †) while performing multiple cycles of irradiation with UV and visible light. The ON and OFF switching of motility was achieved through alternating irradiation at 365 nm for 1 s and at 510 nm for 3 s, respectively. Prior to irradiation with UV light, we observed a complete OFF state for the microtubules. After 1 min, when the flow cell had been irradiated with light at 365 nm for 1 s, all of the microtubules were in motion, with the elimination of their gliding motion after irradiation for 3 s with light at 510 nm. Fig. 4 highlights the motion of two selected microtubules that did not change positions for 60 s prior to irradiation, but did change their positions after irradiation with UV light and then, again, remained motionless after irradiation with visible light. For self-inhibition, the tail domain interacts with the kinesin motor domains through noncovalent interactions. X-ray crystallography of D. melanogaster kinesin-1 with the tail peptide (-940 Ala-Gln-Ile-Ala-Lys-Pro-Ile-Arg 948 Ser-) revealed that the tail peptide is almost symmetrical with respect to 944 Lys of the tripeptide Ile-Ala-Lys core. This Ile-Ala-Lys core is highly conserved in kinesin-1s and regarded as both a necessary and sufficient condition for inducing inhibition. 20,21,28,29 The center residue 944 Lys in the tail peptide interacts electrostatically with residue 185 Asp of the motor domain. The two isoleucine residues in the tail peptide interact with the two hydrophobic pockets composed of residues 123 Phe, 126 Ile, and 135 Phe of both motor domains. A C-H⋯π interaction exists between residue 945 Pro of the tail and residue 179 Phe of the motor domain (Fig. 5b and c). Human kinesin-1 is highly conserved with D. melanogaster kinesin-1 (Fig. S21 †). The critical interaction sites mentioned above are well preserved, although only residue 185 Asp in D. melanogaster kinesin-1 is replaced by 178 Glu in human kinesin-1. Because of the high similarity, we can apply almost the same mechanism to the tail-mediated inhibition of human kinesin-1. Interestingly, in the crystal structure, the tail peptide occupied the interface composed of the symmetrical dimer of motor domains [30][31][32] in two different modes. One is N→C (N terminus to C terminus), while the other is the opposite sequence: that is, the C←N peptide. The overlay of the N→C and C←N tail peptides in Fig. 5(a) reveals little difference in their structures. Both of the peptides are acceptable by the motor domain dimer, indicating that the main chain of the tail peptide does not have a great influence on the binding of the tail peptide. The previously reported azopeptide 1 (Azo-Ile-Pro-Lys-Ala-Ile-Gln-Ala-Ser-His-Gly-Arg-OH) has the reverse sequence of the natural tail domain (-914 Arg-Gly-His-Ser-Ala-Gln-Ile-Ala-Lys-Pro 925 Ile-). Thus, inhibition by the azo-peptide 1 presumably occurred through the same mechanism as that of the natural tail peptide. The azo-peptide 1 contains all of the important amino acid residues in the sequence -Ile-Pro-Lys-Ala-Ile-Gln-, with the photoresponsive azobenzene moiety placed sufficiently closely nearby to facilitate a huge difference in affinity between the two isomeric forms of the azo-peptide. A comparison of the inhibitory behavior of the azo-peptides 1 and 3 suggests that the amino acid residues Ile and Pro adjacent to the highly conserved Lys-Ala-Ile core (--Lys-Ala-Ile-Gln-) are important to achieve inhibition, because the inhibition property decreased significantly by removing these residues from the sequence ( Fig. 2a and b). The insolubility of the azo-peptide 7 and the inefficiency of the azo-peptide 8 at inducing inhibition suggested the need to modify the natural sequence. The azo-peptide 9, with a Gly residue at the C-terminus, rather than a Gln residue, was insoluble, whereas the azopeptides 10 and 11, featuring basic amino acid residues (Lys and Arg, respectively) at the C-terminus, rather than the Gln residue, provided complete inhibition at concentrations of 3.5 and 3.0 mM, respectively ( Fig. 2d and e). These observations suggest that the key residue in human kinesin motor domains is the acidic 170 Glu residue, as displayed in Fig. 5d and S21b. † Because the Arg residue is relatively basic, its interaction with residue 170 Glu was stronger than that of a Lys residue. Enhanced inhibition occurred with the azo-peptide 12, which displayed complete inhibition at a relatively low concentration (2.5 mM, Fig. 2f ) compared with that of the azo-peptide 8, which lacks the Arg residue at the N-terminus (Fig. 2c); thus, to improve affinity, an additional Arg residue at the N-terminus was quite effective. As a result, the azo-peptide 13, for which the peptide unit -Arg-Ile-Pro-Lys-Ala-Ile-Arg is symmetrical with respect to the central Lys residue, exhibited inhibition behavior stronger than that of the azo-peptide 12, completely inhibiting the kinesin motor at a concentration of 1.5 mM (Fig. 2g    and f ). Fig. 5d displays a plausible mode of interaction experienced by the azo-peptide 13.
The inhibition properties were enhanced, relative to those of the parent azo-peptide 11, after substitution with OMe and NMe 2 groups: the azo-peptides 14 and 15 induced complete inhibition at concentrations of 2 and 1.25 mM, respectively ( Fig. 2h and i). This behavior represents the relatively greater affinities of these two azo-peptides towards the kinesin. By substituting with an OMe or NMe 2 group, we added an additional moiety that could function as a proton acceptor for hydrogen bonding with the kinesin pocket. The recovery of velocity of the azo-peptide 14, with the OMe substituent, was higher (ca. 80%) in the motility ON state than that of the azopeptide 11 (recovery of velocity: 62%). This behavior was due to the higher percentage of the cis form (95.5%) in the UV PSS for the compound having the OMe group at the para position of the azobenzene unit than that for the unsubstituted-azobenzene azo-peptide (87.4% cis form in the UV PSS; Table 1). In the case of the azo-peptide 15, bearing an NMe 2 group, we could not switch the system into the motility ON state because of the rapid thermal relaxation of the cis isomer and the insufficient intensity of the irradiated light. 33 The azo-peptide 16, substituted with a COOH group, did not display any inhibition effect, even though it featured the amino acid sequence responsible for the inhibitory activity (Fig. 2j), indicating that the interaction site in kinesin does not favor negatively charged substituents on the azobenzene unit.
Finally, from our study of the structure-property relationships, we rationally designed the azo-peptide 17 to feature the peptide unit -Arg-Ile-Pro-Lys-Ala-Ile-Arg-OH and a para-OMe group on the azobenzene unit. We anticipated that this structure would fulfill the plausible mechanism of azo-peptide inhibition displayed in Fig. 5d. Indeed, the azo-peptide 17 induced complete inhibition of kinesin at a concentration of 750 μMin other words, it had four times the affinity of the azo-peptide 1. In the motility ON state, 17 attained a relatively high recovery of velocity (86%); its percentage recovery remained unchanged even after the fourth cycle of motility switching (Fig. 3). Thus, through in situ photoregulation experiments, we have demonstrated the rapid, complete, reversible, and spatiotemporal regulation of the gliding motility of kinesin over several cycles with constant activity (Fig. 4).

Conclusions
Through the systematic design and analysis of the activities of several azo-peptides, we have identified the amino acid residues responsible for the inhibition of kinesin motility and have developed a plausible mechanism for inducing the inhibition. The interactions observed in the mechanism of inhibition induced by the azo-peptides are similar to those in the natural tail-mediated self-inhibition of kinesin. Among the different substituents tested on the azobenzene unit, the OMe group enhanced the inhibition properties and provided a large phototunable change in activity. Taking all of our findings together, we designed a new photoresponsive inhibitor 17 to feature a shorter sequence (-Arg-Ile-Pro-Lys-Ala-Ile-Arg-OH) and an azobenzene moiety substituted with a para-OMe group. This azo-peptide stopped the kinesin-propelled gliding moti-lity of microtubules completely in the trans form of the azobenzene unit at a concentration of 750 μM, while in the cis form of the azobenzene unit it allowed the gliding motility of microtubules with 86% recovery of velocity with respect to the velocity observed in the non-inhibited state of the kinesin.

Chemicals
All chemical and biochemical reagents were purchased from commercial sources (Tokyo Chemical Industry; Watanabe Chemical Industries; Wako Pure Chemical Industries; Dojindo Molecular Technologies) and used without purification.

Instrumentation
A Burrell Wrist Action Shaker (model 75) was used for the azopeptide synthesis. Purification and analysis of the azo-peptides were performed using a Shimadzu reversed-phase (RP) HPLC system. An EYELA FDU-2200 lyophilization system was used for freeze-drying. Electrospray ionization time-of-flight mass spectrometry (ESI-TOF MS) was performed using a JMS-T100CS instrument (JEOL) operated in the positive-ion mode. UV-Vis absorption spectra were recorded using an Agilent 8453 single-beam spectrophotometer and a Hitachi U-3100 absorption spectrophotometer. A mercury lamp

In vitro motility experiments
Gliding motility assays were performed at 25°C using the standard protocol described previously, 23 but with BRB-80 [ pH 6.9; piperazine-1,4-bis(2-ethanesulfonic acid) (PIPES), 80 mM; MgCl 2 , 2 mM; O,O-bis(2-aminoethyl)ethyleneglycol-N,N,N′,N′tetraacetic acid (EGTA), 1 mM] as the assay buffer. For the azopeptides 1-13, light at wavelengths of 365 and 436 nm (Hg lamp) was irradiated for 20 and 40 s, respectively. For the azo-peptide 16, light at wavelengths of 365 and 436 nm (Hg lamp) was irradiated for 40 and 20 s, respectively. The intensities of the 365 and 436 nm light (Hg lamp) were approximately 9.1 and 26.5 mW cm −2 , respectively. For the azo-peptides 14 and 17, light at wavelengths of 365 and 510 nm (LED), with intensities of approximately 1 and 0.46 W cm −2 , respectively, was irradiated for 7 and 15 s, respectively. A video was recorded after photoirradiating the flow cell to reach the respective PSSs. Videos were analyzed using ImageJ software. The gliding velocity was determined from the average of 10 measurements.
In situ photoregulation of the gliding velocity of microtubules A flow cell was prepared containing kinesin, microtubules, oxygen scavengers, ATP (1.0 mM), and the azo-peptide 17 (750 μM). Repeated photoswitching was performed through irradiation with UV light (365 nm, LED) for 1 s and with visible light (510 nm, LED) for 3 s. A video was recorded continuously for 6 min.