Site-Specific Photocleavage of Hen Egg Lysozyme and Bovine Serum Albumin

Proc. Natl. Acad. Sci. USA Vol. 95, pp. 10361–10366, September 1998 Chemistry

Photochemical protease: Site-specific photocleavage of hen egg lysozyme and bovine serum albumin

CHALLA V. KUMAR†‡,APINYA BURANAPRAPUK†,GREGORY J. OPITECK§,MARY B. MOYER§,STEFFEN JOCKUSCH¶, AND NICHOLAS J. TURRO¶

†Department of Chemistry, University of Connecticut, Storrs, CT 06269-3060; §Glaxo Wellcome, Inc., Analytical Chemistry, Research Triangle Park, NC 27709; and ¶Department of Chemistry, Columbia University, New York, NY 10027

Contributed by Nicholas J. Turro, July 9, 1998

ABSTRACT Site-specific photocleavage of hen egg lysozyme and bovine serum albumin (BSA) by N-(l-phenylalanine)-4-(1-pyrene)butyramide (Py-Phe) is reported. Py-Phe binds to lysozyme and BSA with binding constants 2.2 ؎ 0.3 ؋ M؊1 and 6.5 ؎ 0.4 ؋ 107 M؊1, respectively. Photocleavage 105 of lysozyme and BSA was achieved with high specificity when a mixture of protein, Py-Phe, and an electron acceptor, FIG. 1. Structure of protein probe with a specific recognition cobalt(III) hexammine (CoHA), was irradiated at 344 nm. element. The pyrenyl moiety can be photoactivated in the presence of suitable electron acceptors. Quantum yields of photocleavage of lysozyme and BSA were 0.26 and 0.0021, respectively. No protein cleavage was ob- controlling the wavelength of excitation, specific chromophores served in the absence of Py-Phe, CoHA, or light. N-terminal can be selectively activated to high energies, thereby minimizing sequencing of the protein fragments indicated a single cleav- side reactions. age site of lysozyme between Trp-108 and Val-109, whereas the Vanadate, for example, has been used as a photoprobe to cleavage of BSA was found to be between Leu-346 and Arg-347. investigate phosphate-binding sites of proteins (16). Vanadate Laser flash photolysis studies of a mixture of protein, Py-Phe, competes for phosphate-binding sites of proteins (17), and it and CoHA showed a strong transient with absorption centered Ϸ induces photocleavage of proteins with high preference for at 460 nm, corresponding to pyrene cation radical. Quench- serine residues (17–19). Applicability of vanadate is limited to ing of the singlet excited state of Py-Phe by CoHA followed by phosphate-binding proteins (18) and, hence, developing alter- the reaction of the resulting pyrenyl cation radical with the natives to vanadate is challenging. protein backbone may be responsible for the protein cleavage. Two approaches for the design of chemical proteases can be The high specificity of photocleavage may be valuable in distinguished: (i) affinity-based design (1, 5) and (ii) covalent targeting specific sites of proteins with small molecules. linking of redox-active metal complexes to specific residues of the protein, followed by activation of the metal complex to Reagents that can cleave the protein backbone with a high achieve protein cleavage (2–4, 11–14). Coupling chiral recog- specificity (chemical proteases) can be useful as biochemical tools nition elements with a suitable chromophore has been our (1–6). Structure–activity studies of proteins, investigation of approach to design photochemical proteases. N-(l-Phenylala- protein structural domains, and studies of the proximity of nine)-4-(1-pyrene)butyramide (Py-Phe; see Fig. 1), for exam- specific residues in a given protein are some applications of ple, induces site-specific photocleavage of hen egg lysozyme chemical proteases. Rapid proliferation of chemical nucleases for (Lyso), bovine serum albumin (BSA), and carboxypeptidase A DNA cleavage studies suggests how small-molecule-based chem- (CpA), under mild conditions (15). Advantages of using the ical reagents can be of use as tools in molecular biology (7–10). 4-(1-pyrene)butyroyl (Py) chromophore in the design are as Chemical proteases, analogously, could be useful for the manip- follows. (i) Py has high extinction coefficients, and hence, low ulation of proteins (1–6, 11–14). Determining the exposure of concentrations of the probe can capture a significant fraction selected residues of membrane-bound proteins to the aqueous of the incident light. (ii) Py has a long-lived (Ϸ110-ns), environment, design of new therapeutic agents, and converting high-energy (Ϸ70 kcal͞mol) singlet excited state that can drive large proteins into smaller fragments that are more amenable to a number of photochemical reactions. Short-lived excited sequencing are additional applications of chemical proteases. states are more likely to waste excitation by rapidly decaying Chemical proteases can be activated by light, and novel pho- to ground state without a fruitful chemical event. (iii)Pyhas tochemical proteases (15) are described here that have (i) high been used to probe the microenvironments of micelles, pro- affinities for selected sites of proteins, (ii) strong absorption bands teins, DNA, and other organized media (20). (iv) Considerable iii in the near visible region, ( ) long-lived singlet excited states that mechanistic information regarding Py excited states and re- are convenient to initiate photoreactions, and (iv) a chromophore ϩ ϩ lated intermediates (such as Py . , PyH , and Py.) is available. that has been used to probe microenvironments. Use of light as (v) Py-Phe shows strong induced circular dichroism bands in a reagent for protein cleavage has distinct advantages. Photore- the 300- to 400-nm region, and these spectra are convenient to actions can be started and sustained with light, providing a sharp monitor the interaction of Py-Phe with proteins. control for initiation and termination of the reaction. Visible light Lyso, BSA, and CpA have been chosen as model proteins for is perhaps the least toxic reagent for environmentally benign photocleavage studies. Lyso is a small (Ϸ14-kDa) water- chemical approaches. Photoreactions are amenable for time- soluble enzyme whose activity can be readily monitored. The resolved mechanistic studies to delineate the elementary steps. By x-ray crystal structure of Lyso and its substrate-binding site are

The publication costs of this article were defrayed in part by page charge Abbreviations: Py, 4-(1-pyrene)butyroyl; Py-Phe, N-(l-phenylalanine)- payment. This article must therefore be hereby marked ‘‘advertisement’’ in 4-(1-pyrene)butyramide; Lyso, hen egg lysozyme; CpA, carboxypep- accordance with 18 U.S.C. §1734 solely to indicate this fact. tidase A; CoHA, cobalt(III) hexammine trichloride. © 1998 by The National Academy of Sciences 0027-8424͞98͞9510361-6$2.00͞0 ‡To whom reprint requests should be addressed. e-mail: cvkumar@ PNAS is available online at www.pnas.org. nucleus.chem.uconn.edu.

10361 Downloaded by guest on September 27, 2021 10362 Chemistry: Kumar et al. Proc. Natl. Acad. Sci. USA 95 (1998)

known (21). BSA is a plasma protein that has many physio- spectrometry on a LaserMat instrument (FinniganMat, San Jose, logical functions. Hydrophobic ligands such as fatty acids, CA) with sinapinic acid (Sigma) as the matrix. steroids, and bilirubin bind to BSA (22). Because of its affinity Quantum Yield Measurements. A ferrioxalate actinometer for a variety of biological ligands, the interactions of BSA with (25) was used to determine light intensities at specific wave- specific ligands have been extensively investigated. Several lengths for quantum yield measurements. The entire experi- C-terminal phenylalanine peptides are known to be substrates ment was carried out in a dark room lit only by a yellow for CpA (23). Phenylalanine, therefore, is chosen as the chiral safelight. The actinometer solution (0.006 M) was prepared by ⅐ recognition element. Site-specific photocleavage of these pro- dissolving potassium ferrioxalate [K3Fe(C2O4)3 3H2O, 0.14 g] teins by Py-Phe and mechanistic details of protein photocleav- in water (45 ml) and sulfuric acid (0.5 M, 5 ml). Actinometer age are presented here. solutions (200 ␮l) were irradiated at 344 nm for various lengths of time (0, 5, 10, and 15 min). Aliquots (150 ␮l) of irradiated MATERIALS AND METHODS actinometer solutions were mixed with excess 1,10-phenanth- ϭ ϭ ϭ ͞ Lyso (Mr 14,300), BSA (Mr 66,267), and CpA (Mr roline (50 mg 50 ml of H2O, total 1 ml), and concentration of 35,250) were purchased from Sigma. Protein solutions were the ferrous ion was estimated from absorbance at 510 nm. prepared by dissolving the appropriate amounts of the protein Light intensities were measured before and after each irradi- in 50 mM Tris⅐HCl, pH 7.0. Py-Phe was synthesized as de- ation and an average of two measurements was recorded scribed previously (15). All probe solutions were prepared (within Ϯ5%). Quantum yields for the photocleavage were fresh, and calibration graphs were constructed by using Beer’s estimated from three separate runs. law. Probe concentrations were restricted to the linear region Laser Flash Photolysis. Light pulses (355 nm, Ϸ8mJper of the graph. The molar extinction coefficient of Py-Phe was pulse, 8 ns) from a Continuum Surelite I Nd-YAG laser were estimated to be 42,600 MϪ1⅐cmϪ1 at 343 nm. used to excite the samples. A pulsed 150-W xenon lamp, Photochemical Protein Cleavage. Photocleavage reactions combined with an ISA (Metuchen, NJ) H10 monochromator, were carried out at room temperature in 50 mM Tris⅐HCl served as the monitoring system. Signals from a Hamamatsu buffer, pH 7.0. The mixtures of protein (15 ␮M) and Py-Phe R928 photomultiplier tube were terminated into 50 ohms and (15 ␮M) in the presence of cobalt(III) hexammine trichloride digitized by using a Tektronix 380 (400 MHz) digitizer. The ␮ [Co(NH3)6Cl3; CoHA] (1 mM), in a total volume of 100 l, entire experiment was controlled with a Macintosh Quadra 650 were irradiated at 344 nm by using a 100-W xenon lamp using LABVIEW 3 software as described earlier (26). Transient attached to a PTI (South Brunswick, NJ) A1010 monochro- absorption spectra were produced by obtaining full kinetic mator. A UV cut-off filter (WG-345; 78% transmission at 344 traces at various wavelengths and by sampling the change in nm) was used to remove stray UV light. Irradiated samples optical density at a fixed delay after the laser pulse. The change were dried for gel electrophoresis experiments. in absorption was then plotted as a function of wavelength at Electrophoresis. SDS͞polyacrylamide gel electrophoresis a given delay time subsequent to laser excitation. (SDS͞PAGE) experiments were performed as reported (24). Samples were heated for 3 min in the presence of SDS (7%, RESULTS wt͞vol), glycerol (4%, wt͞vol), Tris (50 mM), 2-mercaptoethanol Site-specific photocleavage of Lyso and BSA by Py-Phe is (2%, vol͞vol), and bromophenol blue (0.01%, wt͞vol), adjusted demonstrated here. Detailed spectroscopic investigations of to pH 6.8 with HCl. Gels (12% polyacrylamide for Lyso and 8% Py-Phe binding to these proteins will be described separately. for BSA) were run by applying 60 V until the dye passed through Protein Cleavage Studies. Irradiation of Py-Phe͞Lyso mixture the stacking gel. The voltage was then increased to 110 V, and at 344 nm results in protein cleavage with high efficiency and electrophoresis was continued for an additional 2.5 hr for Lyso or specificity as monitored in SDS͞PAGE experiments. Photocleav- 1.5 hr for BSA. The gels were stained with Coomassie blue for 4 age of Lyso (35% yield, Fig. 2) yields only two sharp bands of Ϸ Ϸ hr, and destained in acetic acid (10%, vol͞vol) for 4 hr. The gels lower molecular weights ( 3,300 and 11,000) than that of Lyso were scanned with a Hewlett–Packard scanner (ScanJet 4C) and (14,300). If the photocleavage were random, one would have the images were processed by using Adobe PhotoShop͞NIH- observed a smear in these lanes. The molecular weights of the Image software to quantitate band intensities and band positions. photocleavage products were estimated by using the markers in Molecular weight markers were used in each gel for calibration lane 1. Molecular weights of the daughter fragments add up to and the molecular weights of the fragments were estimated from that of the parent protein, suggesting a single cut in the Lyso their mobilities. For quantum yield measurements, various con- backbone. Irradiation of Lyso and CoHA (1 mM) without Py-Phe centrations of Lyso and BSA were electrophoresed and calibra- did not yield any fragmentation (lane 9). The photoreaction tion graphs were constructed to estimate the intensities of the requires an electron acceptor such as CoHA. Irradiation of Lyso protein fragments. without CoHA, in the presence or absence of Py-Phe, did not Peptide Sequencing and Mass Spectrometry. Protein bands result in protein cleavage (lanes 8 and 7, respectively). Control from the gels were electroblotted onto Gore-Tex, manually experiments, thus, establish that Py-Phe, CoHA, and light are excised, and transferred to a sequencing cartridge. Chemical essential for the protein photocleavage. sequencing was performed on an automated protein sequencer (G1005A, Hewlett–Packard). At least five cycles were performed from the N terminus, while a single cycle for the C terminus was attempted on each band. The entire reaction mixture was desalted and concentrated by using HPLC (Ultra-Plus MicroLC system, Micro-Tech Scientific, Sunnyvale, CA). A short reversed-phase column (Hypersil BDS C-18, 1.0 mm ϫ 20 mm, Keystone Scientific, Bellefonte, PA) was FIG. 2. Gel electrophoresis pattern of Lyso under various condi- used with increasing acetonitrile concentration from 5% (in tions. Lane 1, molecular weight markers; lane 2, Lyso (15 ␮M); lane ␮ ␮ aqueous 0.1% trifluoroacetic acid) to 75% (vol͞vol). The peak 3, Lyso (15 M), Py-Phe (15 M), and CoHA (1 mM), dark control; representing the unreacted protein as well as any fragments, as lanes 4–6, same as lane 3 but after 5, 10, and 20 min of irradiation at 344 nm, respectively; lane 7, Lyso (15 ␮M) after irradiation for 20 min; detected by UV absorption at 215 nm, was collected and com- lane 8, Lyso (15 ␮M) and Py-Phe (15 ␮M), after irradiation for 20 min; pletely desolvated under vacuum. The sample was resuspended in lane 9, Lyso (15 ␮M) and CoHA (1 mM) after irradiation for 20 min. 5% acetic acid and analyzed for intact molecular weight by using The samples were run top to bottom and the new bands in lanes 4, 5, matrix-assisted laser desorption ionization͞time-of-flight mass and 6 correspond to the lower molecular weight fragments. Downloaded by guest on September 27, 2021 Chemistry: Kumar et al. Proc. Natl. Acad. Sci. USA 95 (1998) 10363

FIG. 3. Gel electrophoresis pattern of BSA with four different FIG. 4. Percent photoproduct yields of BSA and Lyso vs. time of cobalt(III) complexes. Lane 1, molecular weight standards, as marked irradiation. The yields of the photoproducts were obtained from Ϫ (ϫ10 3); lane 2, BSA (15 ␮M), Py-Phe (15 ␮M), and CoHA (1 mM), quantitated intensities of the product bands on the gel. dark control; lane 3, BSA (15 ␮M), Py-Phe (15 ␮M), and tris(bipyridine) cobalt(III) chloride (1 mM), after irradiation for 60 min; lane 4, most likely because the expected Trp residue was modified by the BSA (15 ␮M), Py-Phe (15 ␮M), and Co(III) sepulchrate trichloride (1 ␮ photocleavage chemistry, and thus not amenable to sequencing. mM), after irradiation for 60 min; lane 5, BSA (15 M), Py-Phe (15 The cleavage site was determined from the known sequence of ␮M), and chloropentammine cobalt(III) chloride (1 mM), after irradiation for 60 min; lane 6, same as lane 2, but after irradiation for 60 Lyso and from the N- and C-terminal sequence data to be min. The irradiation was performed at 344 nm. between Trp-108 and Val-109 (Fig. 5). Mass spectrometry of the desalted reaction mixture from Lyso BSA (15 ␮M) was also successfully cleaved by Py-Phe (15 ␮M), showed the presence of three species, peaks at mass to charge resulting in two fragments of molecular weights 40,800 and 28,400 ratio (m͞z) of 14,550, 11,460, and 3,057. A standard Lyso (Sigma) (344-nm irradiation, 1 mM CoHA, Fig. 3). Electron acceptors sample showed a protonated mass of 14,314 Da. The highest m͞z such as chloropentammine cobalt(III) chloride, tris(bipyridine) peak (14,550) from the reaction mixture was determined to be 3ϩ cobalt(III) chloride [Co(bpy)3 ], and cobalt(III) sepulchrate tri- Lyso that had been modified, presumably by Py-Phe. Therefore, chloride all resulted in the same cleavage fragments as CoHA a mass increase of 236 Da is indicated (Py-Phe has a mass of 435 (Fig. 3) although with various efficiencies. No photocleavage was Da). The experimental error under these conditions is about observed in the absence of light (lane 2) or the electron acceptor. Ϯ1.0%, or a mass of Ϯ15 Da at this m͞z. Addition of pyrenyl Thus, the cleavage most likely occurs at the Py-Phe binding site, moiety to this fragment (201 Da) can partly explain the increased irrespective of the Co(III) species. molecular mass. The average theoretical masses of the unmodi- Yields of the photoproducts increase steadily with irradiation fied 3-kDa and 11-kDa fragments were 2,489 Da and 11,840 Da, time (Fig. 4). The yields of Lyso cleavage products increased respectively. These also appeared at higher masses with differ- rapidly with time and saturated at Ϸ35%, whereas the yields of ences of 568 Da and 380 Da, respectively. BSA products increased linearly at short times. Irradiations for Chemical sequencing of the BSA fragments was similarly longer than 20 min increased the product yields only marginally, performed on the electroblotted and excised bands. The N- with maximum conversions of 35% for Lyso and 21% for BSA. terminal sequence of the Ϸ28-kDa band was found to be (Arg)- The photocleavage of Lyso was found to be more rapid, with Leu-Ala-Lys-Glu. This is a sequence internal to BSA, and Arg better yields than those of BSA. Perhaps the flexible structure of was not directly observed, but was shifted to lower retention time, Lyso plays an important role in exposing the bound pyrenyl indicating that it had been modified. The C-terminal sequencing moiety to CoHA to initiate the photoreaction. Surface binding of of the 28-kDa band showed an Ala residue, consistent with the the probe to Lyso would also permit facile access of the pyrenyl known C terminus of BSA. The N-terminal sequencing of the chromophore to CoHA and can result in higher yields. Ϸ40-kDa band yielded a sequence of Asp-Thr-His-Lys-Ser, as Absorption and CD spectra of Lyso or BSA (200- to 300-nm expected from the known N terminus of BSA. The C-terminal region) showed no evidence of protein structural changes in sequencing of this fragment did not produce an interpretable the presence of CoHA (1 mM). Irradiation of heat-denatured signal, again indicating modification of the residue adjacent to the BSA in the presence of Py-Phe and CoHA did not result in any cleavage site. From the preceding data and the known sequence photoproducts, suggesting the importance of the tertiary struc- of BSA, the cleavage site was determined to be between Leu-346 ture of the native protein for the photoreaction. and Arg-347 (Fig. 5). Enzyme Activity Studies. Activity of Lyso was followed as a The mass spectra of BSA photoproducts showed peaks at m͞z function of the progress of the photoreaction to gain further values of 66,890, 39,930, 33,420, and 26,860. The mass spectrom- insight. Glycol chitin was used as a substrate for Lyso (27), and eter had been calibrated immediately prior to this experiment Lyso activity was unchanged as the photocleavage progressed, with unmodified BSA, the singly charged protein at m͞z 66,430 indicating no deactivation of the enzyme prior to or after chain and the doubly charged species at 33,220. Therefore, the un- scission. Fragmented Lyso is nearly as active as the parent, or cleaved BSA in the reaction mixture had been modified in some the fragments reassociate to form the active enzyme. Cleavage way to increase its mass by 460 Da. The Ϸ28-kDa band seen in of Lyso with retention of the key residues required for activity the gel and chemically sequenced was most likely the 26,860 m͞z is indicated by these studies. peak, a deviation from its average theoretical mass by 100 Da. The Chemical Sequencing Studies. Sequencing of the photoprod- peak in the mass spectrum at m͞z 33420 resulted from the doubly ucts was carried out to determine the cleavage site. The N- charged intact protein. The remaining peak in the mass spectrum terminal sequence of the Ϸ3-kDa fragment from Lyso was found at m͞z 39,930 then must correspond to the Ϸ40-kDa fragment, to be Val-Ala-(Trp)-Arg. Interestingly, the tryptophan residue and it deviates from its average theoretical mass by 240 Da. These was not directly observed, but appeared as a peak whose retention deviations are within Ϯ1% of instrumental error. The chemical time on the reverse-phase column had decreased, indicating that sequencing and mass spectral data, thus, confirmed the high the photocleavage process in some way modified this residue, specificity of the photocleavage. Sequencing data also revealed reducing its hydrophobicity. The C-terminal sequencing of the modification of a few residues during the photoreaction. Despite 3-kDa fragment yielded a single Leu residue, consistent with the these modified residues, six of eight termini were sequenced known C terminus of Lyso. The Ϸ11-kDa fragment was observed successfully. to have an N-terminal sequence of Lys-Val-Phe-Gly, consistent Flash Photolysis Studies. The protein photocleavage was in- with the N-terminal sequence of Lyso. The C-terminal sequenc- vestigated in flash photolysis experiments to gain insight into the ing of the Ϸ11-kDa fragment did not yield an interpretable signal, photocleavage mechanism. Excitation of Py-Phe (30 ␮M), Lyso Downloaded by guest on September 27, 2021 10364 Chemistry: Kumar et al. Proc. Natl. Acad. Sci. USA 95 (1998)

FIG. 5. Cleavage patterns for Lyso and BSA. The cleavage of Lyso was found to be between Trp-108 and Val-109, and the cleavage of BSA was observed between Leu-346 and Arg-347.

(30 ␮M), and CoHA (3 mM) by 355-nm Nd-YAG laser pulses traces observed with Py-Phe, BSA, and CoHA at the correspond- (full width at half maximum, 8 ns) resulted in a strong transient ing peak maxima (Fig. 7 C and D) differed from each other, absorption spectrum with maxima at 390, 425, 470, and 510 nm indicating the formation of several transients. (Fig. 6A). Similarly, excitation of Py-Phe (30 ␮M), BSA (30 ␮M), and CoHA (3 mM) resulted in transient absorption bands at 405, DISCUSSION 420, 460, 490, and 510 nm (Fig. 6B). The transient signals Design and synthesis of chemical proteases is challenging. Use of observed in case of Lyso share similarity with those observed for chemical proteases complements protein structural methods such BSA. as x-ray diffraction, NMR, light scattering, and gel electrophore- Quenching of the Py-Phe excited state (Py-Phe*) by the sis. Developing new and highly selective chemical proteases (as proteins resulted in specific transient signals (in the absence of biochemical tools) could also contribute to our basic understand- CoHA). A weak transient spectrum with maximum at 450–470 ing of how small molecules may interact with proteins. In this nm was observed when a mixture of Py-Phe (30 ␮M) and Lyso context, chromophores linked with specific recognition elements (30 ␮M) was excited with laser pulses (Fig. 6C). In contrast, a were synthesized and tested in our laboratory for their interaction strong transient absorption spectrum with a maximum at Ϸ420 with proteins (15, 29, 30). These probes show high affinities for nm was observed with Py-Phe (30 ␮M) and BSA (30 ␮M) (Fig. various proteins (29, 30), and length of the linker separating the 6D). Theϩ 420-nm transient was later identified as pyrene triplet chromophore from the recognition element plays an important (3Py). Py . was generated independently by quenching Py-Phe* role in the binding behavior (29). with CoHA (3 mM). This resulted in a transient with a strong A molecular probe such as Py-Phe, consisting of hydrophobic absorption at 460 nm (Fig. 6B, lower spectrum). and hydrophilic groups, is likely to bind such that the hydrophobic The kinetics of various transients produced were monitored at Py chromophore is away from the protein surface͞solvent, while their respective maxima. The 460-nm transient produced by the the polar COOH group is interacting with polar residues at the Ϸ quenching of Py-Phe* by CoHA decayed with a lifetime of 50 surface (29–31). Simultaneous binding to sites that can accom- ␮ s (Fig. 7A) and roughly corresponds to the recovery of ground modate both these structural features is expected to predominate. Ϸ ␮ state bleaching monitored at 340 nmϩ ( 63 s, Fig. 7B). The Binding of the probe will be most favored at a site where the array 460-nm transient was assigned to Py . on the basis of literature of microscopic environments of the binding site is complemen- ␭ reports ( max at 457 nm) (28) and additional evidence (discussed tary to that of the probe. The binding free energy for such a ␮ below). The kinetic traces observed for Py-Phe (30 M) in the multisite binding model is expected to be lower than for binding ␮ presence of Lyso (30 M) and CoHA (3 mM) at 390, 425, 470, of Py-Phe to only hydrophobic or to only polar functions of the and 510 nm all decayed with lifetimes similar to the lifetime of the protein. Detailed spectroscopic studies indicate binding of Py-Phe Py-Phe cation radical (data not shown). The transient decay to Lyso at the protein surface, whereas binding to BSA may occur

FIG. 6. Transient optical absorption spectra of aqueous Py-Phe solutions (30 ␮M) recorded at 2–17 ␮s after the laser pulse. In FIG. 7. Kinetic traces after laser flash photolysis of aqueous Py-Phe addition, the solutions contained: Lyso (30 ␮M) and CoHA (3 mM) solutions (30 ␮M) recorded at different wavelengths in solutions (A); BSA (30 ␮M) and CoHA (3 mM) (upper spectrum), or CoHA (3 containing CoHA (3 mM) without the protein (A and B)orCoHA(3 mM) (lower spectrum) (B); Lyso (30 ␮M) (C); or BSA (30 ␮M) (D). mM) and BSA (30 ␮M) (C and D). Downloaded by guest on September 27, 2021 Chemistry: Kumar et al. Proc. Natl. Acad. Sci. USA 95 (1998) 10365

at an interior site with binding constants in the range of 106 to 107 states of CoHA are not accessible by excitation at 344 nm, and MϪ1 (15). Current sequencing data support such a binding model. pyrenyl fluorescence is quenched by CoHA at diffusion- Photocleavage. Although the pyrenyl chromophore or its controlled rates. We propose that only excitation of the pyrenyl singlet excited state (S1) do not cleave proteins, irradiation of probe results in the photoreaction. This proposal was tested by probe–protein complexes in the presence of an electron ac- plotting photoproduct yields as a function of irradiation wave- ceptor such as CoHA results in protein cleavage. The number length (action spectrum, Fig. 9). The plot resembles the absorp- of products, product molecular weights, and sequencing data tion spectrum of Py-Phe and is distinguishable from the broad indicate a single cut in the protein backbone. The photoprod- weak absorption bands of CoHA. Therefore, light absorption by uct formation with Lyso was rapid and reached a maximum of Py-Phe is essential for the reaction. These results are consistent Ϸ35%, whereas in the case of BSA the photoproduct yield with photocleavage studies using various Co(III) complexes, reached a maximum of 21% at longer times (Ͼ20 min). These where the same product formation was indicated, regardless of results clearly indicate the high efficiency of the Lyso reaction the Co(III) complex used.ϩ compared with that of BSA. The quantum yield of photocleav- Direct evidence for Py . intermediacy was obtained from flash age, accordingly, for Lyso was found to be 0.26, in contrast with photolysis studies, which yielded the following observations. (i) 0.0021 for the BSA reaction. Fluorescence quenching studies The rapid formation of the cation radical is consistent with a fast ϭ quenching of the Py-Phe singlet excited state by CoHA (kq indicate greater accessibility of Py-Phe bound to Lyso com- Ϫ Ϫ ϩ pared with Py-Phe bound to BSA, consistent with the above 1.8 ϫ 1010 M 1⅐s 1). (ii) The 460-nm transient assigned to Py . quantum yield measurements. was not quenched by oxygen, which is typical for cation radicals Characterization of the Cleavage Site. Successful use of the (32). (iii) The absorption spectrum of pyrene cation radical was probes for protein cleavage (as reagents in molecular biology) reported to have a strong, sharp maximum around 445 nm at 77 demands that the sequence at which protein cleavage occurs be K in argon matrix (33). The small differences in the peak positions identified and that the fragments be amenable to sequencing. observed here could be due to the substituent, temperature, and ␧ Analysis of at least five N-terminal cycles and one C-terminal matrix effects. (iv) The extinction coefficient ( 460) of the 460-nm ϫ 4 cycle of chemical sequencing indicated the cleavage sites to be transient estimated from the ground state bleaching is 9.3 10 Ϫ1⅐ Ϫ1 at residues 108–109 for Lyso and at residues 346–347 for BSA. M cm , comparable to the reported extinction coefficient of ϫ 4 Ϫ1⅐ Ϫ1 Of the eight peptide termini, six have been successfully se- pyrene cation radical (9.6 10 M cm ) (34). ͞ quenced. The newly generated N-terminal sequences were Laser flash photolysis of Py-Phe CoHA systems in the successfully sequenced, whereas the newly produced C-termini presence of Lyso or BSA (Fig. 6 A and B, respectively) indicates could not be sequenced. Oxidative cleavage of proteins, in the absorption maximaϩ corresponding to that of the pyrene . some cases, may block the N or C termini for sequencing (11, cation radical. Py , thus, is produced under the photocleavage 12). Present data, thus, indicate the viability of photofragments conditions. The cation radical may be a key intermediate in the photocleavage mechanism. for sequencing, albeit with some limitations. ϩ . Molecular modeling studies using the known crystal structure To test the intermediacy of Py in the protein photocleavage, of Lyso indicate that the cleavage site of Lyso is at the hydro- the effect of electron donors such as triethylamine and ethanolamine on the protein cleavage yields were evaluated. The long phobic cleft, which is surrounded by tryptophan residues (Trp-62, ϩ. Ϸ ␮ -63, -111, and -123). Arginine residues (Arg-112 and -114) are also lifetime of Py ( 50 s, Fig. 7A) makes it possible to intercept located near the cleavage site of Lyso. Lyso activity studies this intermediate without affecting the much shorter-lived singlet excited state. Ethanolamine, for example, quenches the photo- indicate that the photocleaved Lyso is active, again consistent ϭ Ϫ1 cleavage yield of BSA (Ksv 60 M ) without significantly with the sequencing data. Although the crystal structure of BSA ϭ Ϫ1 ␶ ϭ is unknown, the observed cleavage site is at the hydrophobic quenching the Py-Phe singlet excited state (Ksv 0.06 M , f 110 ns). Therefore, interception of cation radical by ethanolamine binding site of BSA in subdomain IIA. ϩ. Only two fragments are produced for each protein with a high is indicated. Analogously, kq for quenching of Py by ethanolamine estimated from the flash photolysis experiments was 1.3 ϫ specificity. The MS data suggest that Py-Phe or its derivatives are Ϫ Ϫ Ϫ 106 M 1⅐s 1 with an estimated K (K ϭ k ␶) value of 60 M 1. covalently attached to the protein at specific sites. Sequencing ϩ sv sv q Therefore, Py . is a key intermediate in the photocleavage. Singlet data confirm the modification of nucleophilic residues such as oxygen is not involved in these reactions, as degassing of the Trp and Arg near the cleavage site. The presence of Arg residues reaction mixture did not alter the site of photocleavage or reduce at the cleavage sites suggests salt bridge formation between the the photoproduct formation. Addition of azide (2 mM), a singlet carboxyl group of the Py-Phe with Arg side chains. The methyl oxygen quencher, had no effect on the photocleavage yield or ester of Py-Phe (not capable of salt bridge formation with Arg side specificity. chains) does not cleave either Lyso or BSA. These observations Flash photolysis studies of protein–CoHA solutions (30 ␮M are consistent with anchoring of Py-Phe at specific sites on protein, 3 mM CoHA) without Py-Phe did not produce the protein, perhaps by means of salt bridge formation with Arg transient absorptionϩ at 460 nm, consistent with the assignment of residues (Fig. 8). In addition to salt bridge formation, hydropho- 460 nm band to Py . . Transient absorption spectra of radicals bic interactions of the pyrenyl group with Trp and Phe residues may be significant. Mechanistic Studies. The photoreaction requires both Py-Phe and Co(III) species. Since Co(III) complexes have a weak broad absorption in the 300- to 400-nm region, the reaction could be due to light absorption by Co(III) or Py-Phe. The charge transfer

FIG. 8. Binding modes. (A) Anchoring of the hydrophilic part of FIG. 9. Action spectrum obtained from the photocleavage of BSA the probe at the protein surface. (B) Binding of the probe at the (15 ␮M) with Py-Phe (15 ␮M) and CoHA (1 mM) at various interface. The hydrophobic moiety of the probe is buried in a nonpolar wavelengths. The plot of percent yield of the photoproduct formation region of the protein, while the hydrophilic function of the probe as a function of wavelength of irradiation results in a pyrene-like interacts with charged residues at the protein surface. absorption spectrum. Downloaded by guest on September 27, 2021 10366 Chemistry: Kumar et al. Proc. Natl. Acad. Sci. USA 95 (1998)

ϩ FIG. 10. Possible mechanism for the photocleavage reaction. The initial step is hydrogen atom abstraction at the ␣-carbon by Py . , followed by the addition of O2 to form hydroperoxyl radical. The cleavage of the peptide backbone may then occur after mild hydrolysis. ⅐ ⅐ ⅐ derived from aromatic amino acid side chains (Trp ,Tyr, His as 1. Rana, T. M. & Meares, C. F. (1991) J. Am. Chem. Soc. 113, 1859–1861. well as the corresponding radical ions) may be responsible for the 2. Rana, T. M. & Meares, C. F. (1991) Proc. Natl. Acad. Sci. USA 88, 10578–10582. additional peaks observed in the transient spectra. The absorp- ⅐ ⅐ 3. Rana, T. M. & Meares, C. F. (1990) J. Am. Chem. Soc. 112, 2457–2458. tion maxima of Tyr and Trp are reported to be at 405 and 510 4. Ghaim, J. B., Greiner, D. P., Meares, C. F. & Gennis, R. B. (1995) nm, respectively (28, 32–35). Oxidative quenching of Py-Phe* by Biochemistry 34, 11311–11315. . 5. Ettner, N., Ellestad, G. A. & Hillen, W. (1993) J. Am. Chem. Soc. 115, disulfide links may result in RSSR (absorption maximum at 400 2546–2548. nm) (35). The relative yields of the reactive intermediates derived 6. Ettner, N., Metzger, J. W., Lederer, T., Hulmes, J. D., Kisker, C., Hinrichs, from various amino acid side chains are expected to depend upon W., Ellestad, G. & Hillen, W. (1995) Biochemistry 34, 22–31. their reactivities, number of these residues present at the probe 7. Krotz, A. H., Kuo, L. Y., Shields, T. P. & Barton, J. K. (1993) J. Am. Chem. Soc. 115, 3877–3882. binding site, and their proximity to Py-Phe. 8. Gupta, N., Grover, N., Neyhart, G. A., Singh, P. & Thorp, H. H. (1993) The assignment of all the observed transients is beyondϩ this Inorg. Chem. 32, 310–316. report, but the data clearly indicate the formation of Py . as a 9. Sitlani, A., Long, E. C., Pyle, A. M. & Barton, J. K. (1992) J. Am. Chem. key intermediate. The 420-nm transient observed with BSA͞ Soc. 114, 2303–2312. 10. Mack, D. P. & Dervan, P. B. (1992) Biochemistry 31, 9399–9405. Py-Phe mixture, in the absence of CoHA, (Fig. 6D) decayed 11. Schepartz, A. & Cuenoud, B. (1990) J. Am. Chem. Soc. 112, 3247–3249. with a lifetime of Ϸ18 ␮s (air-saturated solutions) character- 12. Cuenoud, B., Tarasow, T. M. & Schepartz, A. (1992) Tetrahedron Lett. 33, istic of pyrene triplet. This transient is quenched by oxygen 895–898. ϭ ϫ 8 Ϫ1⅐ Ϫ1 13. Hoyer, D., Cho, H. & Schultz, P. G. (1990) J. Am. Chem. Soc. 112, with kq 1.3 10 M s , consistent with the value obtained 3249–3250. for the oxygen quenching of pyrene triplet state in BSA 14. Ermacora, M. R., Delfino, J. M., Cuenoud, B., Schepartz, A. & Fox, R. O. Ϫ Ϫ ␭ ϭ ϭ ϫ 8 1⅐ 1 (1992) Proc. Natl. Acad. Sci. USA 89, 6383–6387. solution ( max (triplet) 415 nm;ϩkq 1.0 10 M s ) (36). Low diffusional mobility of Py . compared with that of the 15. Kumar, C. V. & Buranaprapuk, A. (1997) Angew. Chem. Int. Ed. Engl. 36, 2085–2087. hydroxyl radical can localize the reactive intermediate. High 16. Crans, D. C., Sudhakar, K. & Zamborelli, T. (1992) Biochemistry 31, affinity of Py-Phe for selected sites on proteins may be important 6812–6821. in restricting the reagent to selected sites and result in high 17. Cremo, C. R., Loo, J. A., Edmonds, C. G. & Hatlelid, K. M. (1992) specificity. Two distinct contributions to the specificity of the Biochemistry 31, 491–497. ϩ. 18. Rehder, D. (1991) Angew. Chem. Int. Ed. Engl. 30, 148–167. photocleavage can be considered. (i) Upon photoexcitation, Py 19. Correia, J. J., Lipscomb, L. D., Dabrowiak, J. C., Isern, N. & Zubieta, J. reacts with specific residues and causes cleavage at these sites (1994) Arch. Biochem. Biophys. 309, 94–104. (reactivity model). (ii) The probe binds to a unique site in the 20. Kumar, C. V. (1991) in Photoprocesses in Organized Media, ed. Ramamur- thy, V. (VCH, New York), pp. 783–816. protein matrix and reacts with only the neighboring side chains, 21. Blake, C. C. F., Koenig, D. F., Mair, G. A., North, A. C. T., Phillips, D. C. resulting in a high specificity (recognition model). A combination & Sarma, V. R. (1965) Nature (London) 206, 757–763. of recognition and reactivity is proposed to be responsible for the 22. Jones, M. N., Skinner, H. A. & Tipping, E. (1975) Biochem. J. 147, 229–234. observed high specificity. This model is consistent with the lack 23. Garnier, F., Youssoufi, H. K., Srivastava, P. & Yassar, A. (1994) J. Am. Chem. Soc. 116, 8813–8814. of a consensus sequence for photocleavage. On the other hand, 24. Scha¨ggerH. & von Jagow, G. (1987) Anal. Biochem. 166, 368–379. salt bridge formation with Arg side chains may play a significant 25. Hatchard C. G. & Parker, C. A. (1956) Proc. R. Acad. Sci. London A 235, role in cleavage site recognition. Photocleavage may proceed via 518–536. ⅐ ϩ hydrogen atom (H ) abstraction by Py . from the ␣-C atom of 26. McGarry, P. F., Cheh, J., Ruiz-Silva, B., Hu, S., Wang, J., Nakanishi, K. & Turro, N. J. (1996) J. Phys. Chem. 100, 646–654. selected amino acid residues (Fig. 10). Subsequent addition of O ⅐2 27. Imoto, T. & Yagishita, K. (1971) Agric. Biol. Chem. 35, 1154–1156. to the C-centered radical to form a hyperoxyl radical (OCOOO ) 28. Hsiao, J. S. & Webber, S. E. (1993) J. Phys. Chem. 97, 8289–8295. and peptide cleavage may occur as reported in the literature (37). 29. Kumar C. V. & Tolosa, L. M. (1993) J. Phys. Chem. 97, 13914–13919. Current results clearly indicate high specificity for protein 30. Kumar, C. V. & Tolosa, L. M. (1993) FASEB J. 7, A1131 (abstr.). 31. Kumar, C. V. & Asuncion, E. H. (1993) J. Am. Chem. Soc. 115, 8547–8553. photocleavage. Further studies with probes carrying specific 32. Bohne, C., Abuin, E. B. & Scaiano, J. C. (1990) J. Am. Chem. Soc. 112, recognition elements will help us in the rational design of 4226–4231. chemical reagents to target specific sites of proteins. In addi- 33. Hirata, Y. & Mataga, N. (1985) J. Phys. Chem. 89, 2439–2442. 34. Vala, M., Szczepanski, J., Pauzat, F., Parisel, O., Talbi, D. & Ellinger, Y. tion to biochemical applications, such reagents may be of (1994) J. Phys. Chem. 98, 9187–9196. interest for therapeutic purposes. 35. Prutz, W. A., Siebert, F., Butler, J., Land, E. J., Menez, A. & Montenay- Garestier, T. (1982) Biochim. Biophys. Acta 705, 139–149. C.V.K. and A.B. are grateful for the financial support of the University 36. Cooper, M. & Thomas, J. K. (1977) Radiat. Res. 70, 312–324. of Connecticut Research Foundation. S.J. and N.J.T. thank the National 37. Roberfriod, M. & Calderon, P. B. (1995) Free Radicals and Oxidation Science Foundation (Grant CHE93-13102) for their generous support. Phenomena in Biological Systems (Dekker, New York). Downloaded by guest on September 27, 2021