Eur. J. Biochem. 210,711 -719 (1992) 0 FEBS 1992

Fluorescence study of the ligand stereospecificity for binding to lumazine protein John LE.E', Bruce G. GIBSON1, Dennis J. O'KANE', Angelika KOHNLE' and Adelbert BACHER'

' Department of Biochemistry, University of Georgia, Athens, USA Lehrstuhl fur Organische Chemie und Biochemie, Technische Universitiit Munchen, Federal Republic of Germany

(Received May 27/September 14, 1992) - EJB 92 0750

6,7-Dimethyllumazine derivatives, substituted at the 8-position with aldityls or monohydroxyalkyl groups, have been examined for their binding ability to lumazine apo-protein from two strains of Photobacterium phosphoreum using fluorescence dynamics techniques. On the protein the lumazine has a nearly monoexponential decay of fluorescence with lifetime 13.8 ns (20°C). In free solution the lifetime is 9.6 ns. The concentration of free and bound lumazine in an equilibrium mixture can be recovered readily by analysis of the fluorescence decay. Only the aldityl derivatives D-xylityl and 3'- deoxy-D-ribityl, having stereoconfigurations at the 2' and 4' positions identical to the natural ligand, 8-(1'-~-ribityl), show comparable dissociation constants (0.3 FM, 20 "C, pH 7.0). D-Erythrityl and L- arabityl have dissociation constants of 1-2 pM. All other ligands show no interaction at all or have dissociation constants in the range 6 - 80 pM, which can still be determined semi-quantitatively using the fluorescence decay technique. In the case of these very weakly bound ligands, unambiguous detection of bound ligand can be shown by a long correlation time (23 ns, 2 "C) for the fluorescence anisotropy decay. Examination of the bound D-xylityl compound's fluorescence anisotropy decay at high time resolution (< 100 ps) shows rigid association, i.e. no mobility independent of the macromolecule. All bound ligands appear to be similarly positioned in the binding site. The influence of the stereoconfiguration at the 8-position found for lumazine protein parallels that previously observed for the enzyme riboflavin synthase, where the lumazines are substrates or inhibitors. This is consistent with the finding of significant sequence similarity between these proteins. The binding rigidity may have implications for the mechanism of the enzyme.

6,7-Dimethyl-8-(l'-~-ribityl)lumazine is the biosynthetic precursor to riboflavin (see [l, 21 for recent reviews). However this lumazine derivative has also been discovered to perform quite a distinct function as the light emitter of the biolumin- escence in bacteria belonging to the genus Photobacterium (see [3] for a recent review). The lumazine occurs in these bioluminescent bacteria bound to a 20-kDa protein named lumazine protein. It was of much interest therefore to find considerable sequence similarity in the primary structures of lumazine protein and of riboflavin synthase, indicating a close evolutionary relationship among these proteins [4, 51.

Riboflavin synthase catalyzes the conversion of two mol- ecules of 6,7-dimethyl-8-ribityllumazine to yield riboflavin and 5-amino-6-ribitylamino-2,4-( 1 H,3H)pyrimidinedione. Li- gand binding studies have indicated that each protein subunit of the riboflavin synthase trimer has two substrate binding sites in line with the dismutation mechanism [6]. Moreover, it was shown that the amino acid sequence shows strong simi- larity between the N-terminal and the C-terminal halves [7]. This suggested that the peptide forms two homologous do- mains, each of which provides the binding site for one of the substrate molecules, and implies that the catalytic site is located at the interface between the two domains. Internal sequence similarity is also found in lumazine protein [4]. How- ever, in contrast to riboflavin synthase, the monomeric

Correspondence to J. Lee, Department of Biochemistry, University

Enzyme. Riboflavin synthase (EC 2.5.1.9). of Georgia, Athens, GA 30602, USA

lumazine protein binds only one lumazine molecule/protein molecule [8].

This present study concerns the effect of alteration of the 8-substituted group of the lumazine on its binding to the lumazine apo-protein. The results are found to parallel the binding properties of riboflavin synthase, where the interac- tion is reflected by effects on K, or by the inhibition constant,

A further outcome of this work is a demonstration of the utility of fluorescence intensity and fluorescence anisotropy decay techniques in detecting the binding and in recovering the binding parameters. These techniques are applicable be- cause all these lumazine derivatives are highly fluorescent, both free and protein-bound, and have long fluorescence life- times. From the measurement of anisotropy decay the bound ligands appear to be quite immobilized, with no free move- ment independent of the protein's rotational diffusion [12,13].

Ki [9, 101.

EXPERIMENTAL PROCEDURES Two different strains, A13 and P, of the bioluminescent

bacterium Photobacterium phosphoreum, were used as the source of the lumazine protein because of availability but also to provide sample variability. Lumazine protein was purified to homogeneity from both strains by literature methods [14]. The lumazine ligand and the protein are associated in a 1 : 1 ratio so the concentration of lumazine protein was determined by the absorption coefficient of the bound ligand, E~~~ nm = 10300 M- ' cm-l [15].

71 2

The apo-proteins from both types were prepared by modi- fication of the method previously published involving repeated filtration in the presence of either 2 M urea (50 mM sodium phosphate pH = 7.0) [8] or at pH 5.8. Some aggregation oc- curs and the aggregated protein is removed by centrifugation. Some of the preparations could not be reconstituted and this was not apparent in our earlier study where fresh preparations were used [16]. The problem may be due to the long storage time of the proteins. The reconstitutability or ‘competence’ for rebinding was in the same range, 40-70%, for either preparation method. The competence was stable over about a day but all preparations were used within a few hours.

The concentration of the apoprotein preparations was de- termined by the fluorescence intensity at 340 nm on excitation at 280 nm, using the native protein as a standard. It has been shown that this protein fluorescence intensity, arising from the single tryptophan residue, is unchanged on removal of the ligand [8].

All the lumazine derivatives were synthesized by published procedures [17, 181. All other chemicals used were of the best commercial grades.

Steady-state anisotropy

The anisotropy measurements were made with an SLM Instruments Inc. (Urbana, IL) model 8000 spectrofluorimeter fitted with Glan-Thompson prisms in the T-configuration [19, 261. A slit width of 4 nm was used and the sample was held in a thermostatted 3-ml quartz cuvette. The total fluorescence emission viewed on the reference arm of the T passed first through an interference filter of maximum transmission at 488 nm, and a color filter (Corning 3-74) to cut down on scattered excitation without introducing fluorescence impurity from the glass itself (e.g. as does the type 3-73).

The excitation and emission wavelengths for all measure- ments of samples were 412 and 480 nm, corresponding to the isosbestic and uncorrected normalized isoemissive points of the free and bound ligands [ 11,201. The samples were exposed to the excitation radiation only for the duration of the measurement (about 3 min) and this avoided photochemical damage. The anisotropy was determined in the usual way from the ratio of the signals in the two arms of the emission T on alternate orientation of the excitation polarizer vertically and horizontally. These orientations were calibrated by ob- taining values for polarization, p , close to the theoretical with a very dilute sample of glycogen (p z 1, for emission wave- length = excitation wavelength = 412 nm) and for the 340 + 440 nm fluorescence of quinine sulfate (p z 0).

For steady-state fluorescence spectra, the excitation polarizer was oriented vertically and the emission mono- chromator polarizer at 55” from the vertical, the ‘magic-angle’ position. This arrangement eliminates the effect of polariza- tion distortion of the spectra. Only the signal from the emission monochromator side of the T was collected. Absorp- tion spectra were measured with a diode array spectrometer (Hewlett-Packard 8452A) controlled by SPECOS software [211.

Fluorescence dynamics

The fluorescence dynamics measurements utilized a laser system and detection optics described fully elsewhere [22]. The sample was contained in a 3-ml thcrmostatted cuvette and the path lengths for excitation and emission were 10 mm. Two sides of the cuvette were blackened to eliminate reflected light

Table 1. Fluorescence decay parameters for 6,7-dimethyl-8-substituted lumazines. Values were measured at 20°C except where noted. Exci- tation at 380 nm; emission at 480 nm; pH =7.0. Fluorescence inten- sity fitted to: Z = a l . exp( - t / ~ ~ ) + uZ . exp( - t / ~ ~ ) , with a1 + a2 = 1.0.

[Pi] 8-Substituent a1 T i 7 2

M ns

0.05 D-Ribityl D-Ribityl (2°C) 3’-Deoxy-u-ribityl 3’-Deoxy-u-ribityl(2 “C) Methyl

0.3 D-Ribityl L-Ribityl

L-Xylityl D- Arabit yl L-Lyxityl 3'-H ydroxypropyl 4-Hydroxybutyl 5’-Hydroxypentyl 6-H ydroxyhexyl D-Ribityl (2°C)

3’-Deoxy-~-ribityl 3’-Deoxy-~-ribityl (2°C)

D-Xylityl

0.35 D-Ribityl

0.92 10.3 5.5 0.94 11.2 3.8 0.86 10.3 6.9 0.94 11.0 4.4 1 .oo 11.6

0.90 9.55 5.0 0.87 9.2 3.8 0.92 9.5 4.4 0.91 9.5 3.6 0.90 9.7 5.1 0.85 9.1 5.1 0.63 10.1 0.07 0.72 10.3 0.13 0.70 10.8 0.1 0.75 10.7 0.13 0.95 11.2 3.3

0.89 8.4 4.6 0.91 8.1 3.8 0.92 9.5 2.0

pulses. The electronic system has been up-dated by using an MCA board (Tennelec-Nucleus, Oak Ridge, TN; PCA-I1 board) installed in a 486/33 computer for the data collection.

The data analysis software (Globals Unlimited, Labora- tory for Fluorescence Dynamics, University of Illinois, Ur- bana IL) was also installed on the same computer. When a number of data sets could be combined, by virtue of having the same lifetimes for example, advantage was taken of global analysis techniques to improve the statistical significance of the results. When parallel and perpendicular data sets were taken, the simultaneous method of analysis was used in ad- dition to the global combination [23]. The derivations of equations for the recovery of binding parameters from both steady-state and dynamic fluorescence data are described in the Appendix.

RESULTS

The lumazine derivatives used in this work all have the same absorption spectra, with a maximum in aqueous solution (pH 7.0, 20°C) at 408 nm [9, 241. Their fluorescence spectra are indistinguishable from the D-ribityl derivative [ 121 (495-nm maximum) except for the 2’-hydroxyethyl derivative (498-nm maximum) and are therefore not shown. Table 1 shows that the compounds also display very similar, though distinguishable, dynamic fluorescence properties, with a major decay component of lifetime, z1 around 10 ns. This decay time is shortened at the higher phosphate concentrations used in some of the binding studies. The decay time is lengthened by about 1 ns going from 20°C to 2°C. Also the hydroxyalkyl derivatives have a longer decay time by about 1 ns than the aldityl derivatives, accompanied by a lower amplitude (a1). In the protein mixtures to be studied, the appearance of this lifetime around 10 ns will be used to identify the contribution of the free ligand.

Table 2. Fluorescence and anisotropy decay for native and reconstituted lumazine proteins. Values were measured at 2 0 T , 0.3 M Pi, except where noted; lumazine proteins: native 20 pM, apo-protein 10 pM + ligand 7 pM. Excitation at 380 nm, emission at 480 nm. For equations see Table 1 legend and the Appendix. Lum = 6,7- dimethyllumazine; nd = not determined.

Protein Condition Fluorescence Anisotropy type

ai T I T 2 ro 4 ns

A13 Native 0.99 13.7 3 Native, 0.05 M P, 0.97 13.8 3 Apo + D-ribityl-Lum 0.99 13.8 0.4 Apo + D-xylityl-Lum 0.97 13.9 4 Apo + ~-ribityl-Lum, 0.05 M P, 0.96 14.1 5 Native, 2 "C 0 94 14.5 0.8 Native, 0.25 M P,, 2°C 0.94 14.7 3 Apo + D-ribityl-Lum, 2°C 0.99 14.4 0.3 Apo + D-ribityl-Lum, 2°C 0.92 14.2 0.6 Apo + D-xylityl-Lum, 2 "C 0.92 14.0 2

P Native 0.87 13.9 5.6 Native, 2°C 0.94 14.6 1.3 Apo + D-xylityl-Lum, 2°C 0.86 13.4 1.3 Apo + D-ribityl-Lum, 2 "C 0.82 13.8 5.8

deoxyribityl-Lum, 2°C 0.86 13.3 5.7 ApO + D-3'-

ns ~

0.321 14.4 0.33 14.2 0.30 12.1 0.325 13.2

0.35 13.9 0.345 25.3 0.35 24.2

0.347 24.5

0.34 25.7

0.34 24.6

0.35 13.0 0.30 21.9

0.325 24.7

nd nd

nd nd

A second lifetime component, z2, is seen for all derivatives except the 8-methyl. The presence of multi-exponential fluo- rescence decay in small molecules is often attributed to the population of different conformational states and this possi- bility is supported by the single lifetime observed for the methyl derivative. Consistent with this idea, for the D-ribityl and D-xylityl derivatives the amplitude of the second lifetime is much reduced in the protein-bound state (Table 2).

Table 2 gives the fluorescence results for the two types of lumazine proteins, A1 3 and P, together with the apo-proteins rebound either to the authentic lumazine or to the two other derivatives that appear to bind equally well. In all cases there appears a dominant fluorescence lifetime, zl, around 14 ns. This value is not changed by increase in Pi concentration but increases slightly (14.5 ns) when the temperature is lowered to 2°C. There is no significant difference in z1 between the two types of lumazine protein or between the various bound ligands. This 14-11s lifetime therefore can be used to identify the presence of the bound ligand in equilibrium mixtures. The rotational correlation time, 4, has the values calculated from the Stokes-Einstein equation for the rotational diffusion of a 22-kDa hydrated spherical molecule, i.e. around 23 ns at 2°C and 13 ns at 20°C [13]. This implies that these bound ligands have no motion independent of the rotational diffusion of the macromolecule. The only property that distinguishes the two types of lumazine protein used in these experiments is that, for the native proteins, the 4 value for the P type is consistently about 10% less than for the A13 type.

Fig. 1 shows the fluorescence decay of a dilute solution of lumazine protein. The fluorescence contains contributions

713

1.0, 4.22 1

2 104 206 307 403 511 channel number (100 ps/ch)

Fig. 1. Fluorescence decay of a dilute solution of A13 lumazine protein (0.2 gM, 0.35 M Pi, 20OC). The full line is the function: Z (counts) = 0.666 exp(-z/13.8) + 1.37 exp(-ir/8.4) + 0.41 exp(-t/2.0). The values 23.8 ns and 8.4 ns are fixed corresponding to the decay times for free and bound lumazine (Tables 1 and 2). The fit to the data is excellent as judged by the reduced 1' = 1.02, the randomness of the residuals and autocorrelation function (inset). The triangular line at about channel 25 is the laser pulse widened to about 150 ps half-width by the instrumental response. Excitation 380 nm, emission 480 nm.

from both the free and bound ligands. Therefore the decay is fitted by a three-exponential function with two lifetimes z1 = 13.8 and z2 = 8.4 ns, constrained to the values in Tables 2 and 1 respectively, expected for the bound and free lumazine derivative under the experimental conditions used. It is seen that the fitted function describes the data well as judged by the random distribution of the residuals around zero, the flatness of the autocorrelation function, and that the reduced x2 is close to unity. An unconstrained fit with three com- ponents does not give a statistically better fit (x' = 1.00). A two-exponential fit, however, can be rejected on the basis of the residuals, the autocorrelation function and x 2 = 1.25. The third component, z3 in Fig. 1, contributes only about 5% to the total fluorescence intensity, the product of z and a. This corresponds to the average fluorescence expected of the minor component, labelled z2 in Tables 1 and 2, which is required to fit the decay of the free or bound ligand.

The dissociation of lumazine protein on dilution was fol- lowed by the analysis of fluorescence decay curves as in Fig. 1 and by measuring steady-state anisotropy. The results are summarized in Table 3. Under these solution conditions both types of fully bound lumazine protein have the same steady- state anisotropy, ( r ) = 0.192. Following dilution, the ani- sotropy decreases as expected due to the averaging in of the free ligand. Equilibrium appears to be established within 1 h at 2'C so measurements are not started until this time. Also at 2' C, the average anisotropy remains the same for several days. At 20"C, some aggregation is observed after 1 day. Also, the anisotropy measured before and after the fluorescence decay measurement, does not indicate any photochemical changes from the laser irradiation.

It is seen in Table 3 that the equilibrium constants obtained from the anisotropy data, K,, are on average lower than the ones recovered from the fluorescence decay measurements, K,. The fluorescence data only was subjected to Scatchard analysis and gave quite satisfactory plots with very acceptable correlation coefficients ( r ) . On the other hand, the anisotropy data gave results which are more scattered, the correlation

714

Table 3. Analysis of dissociation of native lumazine proteins by fluores- cence decay and steady-state fluorescence anisotropy. All measure- ments were at 20"C, 0.35 M Pi and are given & SD; nd = not deter- mined. Scatchard analysis for type A13: N = 0.8 . f O . 1 ; Kf = 0.11 +0.01 pM, r = -0.98. For type P: N=0.87+0.16; Kf = 0.036 0.007 vM, r = -0.97, where N = number of binding sites, r = correlation coefficient.

Protein Concn Fluorescence Anisotropy type

% / ~ 2 Kf ( r ) Kr

PM A13 0.1

0.14 0.20 0.40 0.40 0.73

P 0.059 0.098 0.141 0.196 0.294

0.75 0.40 0.49 0.75 0.75 0.88 0.52 0.74 0.84 1.065 1.30

FM 0.18 0.155 0.17 0.18 0.18 0.26 0.046 0.045 0.054 0.051 0.056

PM 0.0925 0.10 0.092 0.13 0.107 0.123 0.130 0.110 0.127 0.126 nd nd

0.130 0.017 0.142 0.019 0.139 0.027 0.147 0.025 0.150 0.018

H-+-OH H-$-OH CHIOH CHIOH CHZOH

0 0

HO-$-H HO-C-H CH,OH CHIOH CHZOH

Scheme 1. 6,7-Dimethyllumazine derivatives.

coefficients are -0.85 and -0.7 for the A13 and P types respectively, giving rise to unacceptable standard deviations. The absorption spectra of the apo preparations (not shown) indicate some scattered light in the solution and this would account for the overestimation of anisotropy values.

The P type lumazine protein has a significantly lower dis- sociation constant than A13. For both types the stoichiometry for ligand binding (A') is reasonably close to unity.

A 2 pM lumazine protein solution at 2°C in 50 mM so- dium phosphate pH = 7.0 is fully associated and ( r ) = 0.245. Even at 1 pM, the ligand is almost fully bound, ( r ) = 0.225. If a preparation of apoprotein and the authentic D-ribityl ligand is mixed with each at this same 1 pM concentration, a typical result is that ( r ) = 0.16 after about the first hour, then increases slowly over the following 10 h or more, but only to a value of about 0.19, where it remains constant. This value of the ( r ) below that expected is explained if only a certain fraction of the protein binds the ligand with about the same equilibrium constant as for the native protein. The fraction of competent protein can be determined by the fluorescence de- cay method or from the value of ( r ) of this single mixture, assuming the value of K to be the same as for the native protein. Alternatively the apoprotein can be titrated with the D-ribityl ligand at 2°C and ( r ) measured. The equilibrium constants listed in the following are calculated on the basis of the concentration of competent protein, which is in the range 40 - 70% for the preparations used.

Strongly associated ligands

Table 4 shows the results for the three ligands (1-3; Scheme 1) that show dissociation constants comparable to the native ligand and for the D-erythrityl derivative (6; Scheme 1). Both fluorescence decay and steady-state ani- sotropy results are listed. The apo-protein from type P was prepared by the pH method and type A13 by the urea method. The binding was measured in each group for the same prep- aration batch. The concentration refers to the 'competent binding protein' and for this experiment was 50% of the total protein for the P type, and 40% for the A13 type.

The first set of measurements in the top part of Table 4 for D-ribityl illustrates the slow equilibration for rebinding to the apo-protein. After mixing, the sample was held at 20°C for the time given in the parentheses, except that it was stored at 5 "C overnight. Both the fluorescence decay and anisotropy of the 3 pM D-ribityl sample were measured at the times indi- cated and showed a slow increase in the binding, i.e. an in- crease in ( r ) and in the al/ccz ratio till about 24 h, or 8 - 10 h at 20°C. Therefore all following measurements were made after such an equilibration time. In this set the minimum value of Kf, about 0.2 pM, is taken as the equilibrium value.

For the P-type apo-protein, the D-xylityl and 3'-deoxy-~- ribityl derivatives produce dissociation constants nearly the same as for the natural ligand, D-ribityl, i. e. in the range 0.15 - 0.4 pM (20°C). These values are not significantly different considering the limitations of the methodology. However the 8-D-erythrityl compound clearly has a dissociation constant several times higher. The dissociation constants from the ani- sotropy data were not calculated for the P type because of the interference from light scattering, as already noted. It is also found that the L-arabityl derivative as measured from ani- sotropy also binds moderately well, but this one was not studied systematically and is relegated to Table 5.

For this particular apo-protein preparation from A1 3 type, the dissociation constants for D-Xylityl and 3'-deoxy-~-ribityl are in the same range as for the P type (near 0.3 pM) but the natural ligand shows a larger dissociation constant in this case. However separate determinations with other prep- arations gave values for D-ribityl also in the 0.3 pM range. Again, the D-erythrityl definitely binds less, Kf = 1.6 pM. These relative comparisons can also be seen to hold in the anisotropy data. The uncertainty in estimating the concen- tration of the competently binding apo-protein severely limits the absolute accuracy of the results. Nevertheless these state- ments about relative binding are reliable. We conclude that both types of protein rebind these ligands to the same extent, but with less affinity than measured by dilution assay of the holoprotein, probably due to instability of the apo-protein.

Table 5 collects the dissociation constants for these strongly associated ligands together with all the other lumazine derivatives studied. For the stronger ligands at 2"C,

71 5

Table 4. Equilibrium dissociation constants for the binding of the strong binding group of 8-aldityl-6,7-dimethyllumazine derivatives to apo- lumazine protein. In 0.35 M P,; Kf is from the fluorescence decay at 20°C; K, is from the steady-state fluorescence anisotropy; nd = not determined. Numbers in parentheses are time (h) after mixing; the sample was stored at 5°C overnight and then equilibrated for 2-5 h at 20°C before remeasuring. All the A13 measurements were made more than 24 h after mixing to ensure equilibration.

Apo-protein Aldityl a1b2 Kf <r> (20°C) K,

type concn group concn (time)

PM FM (h) FM FM

1.85 3.0 (3) 0.41 0.67 0.121 P 1.85 D-ribityl 3.0 (1) 0.40 1.36 0.118

1.85 3.0 (22) 0.71 0.35 0.145 1.85 3.0 (24) 0.74 0.29 0.147 1.85 3.0 (45) 0.78 0.24 0.151

1.85 D-xylityl 1.9 (24) 0.89 0.6 0.158 1.85 1.9 (44) 0.80 0.6 0.176 0.65 0.61 (20) 0.80 0.3 0.148

1.85 0.71 (20) 1.04 0.9 0.176 0.9 0.38 (20) 1.24 0.4 0.157 1.85 D-erythrityl 0.96 (24) 0.61 1.6 0.140 1.85 0.96 (40) 0.63 1.6 0.134

0.37 0.3 (24) 0.38 0.15 0.102

1.85 3’-deoxy-~-ribityl 0.77 (10) 0.94 1.1 0.211

A1 3 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5

D-ribityl 0.3 0.24 0.24 0.18 0.12

D-xylit yl 0.57 0.35 0.23 0.115

3’-deoxy-~-ribityl 0.475 0.38 0.285 0.19

D-erythrit yl 0.51 0.36 0.205 0.15

0.59 0.35 0.30 nd nd 0.60 0.97 0.70 0.71 0.68 0.73 0.72 0.78 0.18 0.18 0.18 nd

0.62 0.80 0.97

0.26 0.21 0.37 0.42 0.26 0.28 0.33 0.34 1.5 1.7 1.6

( r ) (2°C) 0.186 0.184 nd 0.197 0.2025 0.200 0.21 3 0.219 0.238 0.194 0.196 0.204 0.206 0.1 15 0.124 0.132 0.140

0.1 1 0.13

0.10 0.13 0.12 0.04 0.04 0.02 0.05 0.06 0.07 0.09 0.39 0.40 0.43 0.40

only the anisotropy-determined K, is listed because, at the required low concentrations for measurement (below 0.1 pM), the fluorescence count rate becomes too small to produce reliable results.

Weakly associated ligands

For some of the compounds in Table 5, no bound ligand can be detected even at 2°C. A typical experiment involved incubating 5 - 10 pM apo-protein with lesser amounts of the ligand under consideration. The equilibrium constants for these compounds in Table 5 were derived from measurement of the fluorescence decay, fixing the free and bound lifetimes according to Tables 1 and 2. However the separation of the two lifetime values is such that, if the amplitude (a) of the 14-11s component is less than 10% of the sum of amplitudes, the recovery of the freelbound concentration ratio becomes very uncertain. In other words, a three-component fit to the decay does not have statistical justification over the two- component one typical of the free ligand (Table 1). Therefore the weaker dissociation constants listed in Table 5 must be taken as very approximate. On mixing some of these ligands

with the apo-protein, the ( r ) values indicated the presence of bound form but this might be an artifact due to scattered light interference from the protein, a problem apparent on comparing these results with a preliminary report [16]. A direct method of detecting the genuinely bound state is to observe the decay of fluorescence anisotropy. Scattered light, both from the Rayleigh and Raman effects, can only contribute during the time of the excitation pulse. Following the pulse, anisotropy that is seen to decay, especially with a 4 corre- sponding to the protein’s rotation, must be attributable to bound ligand (the apo-protein itself contributes no fluores- cence).

Fig. 2 shows the anisotropy decay for the 2’-deoxy-~- ribityl derivative mixed with apo-protein. The presence of a long correlation time clearly reveals that this compound has a propensity to bind, even though it lacks the 2’-OH group, which appears to be essential on the basis of the data in Table 4. For the D-lyXityl derivative the fluorescence decay alone does not reveal the contribution of any bound ligand, i.e. no 14.7-11s component, but a small amount of slowly decaying anisotropy is still detectable (result not shown). It was estimated that about 10% of the ligand is bound in this

71 6

. .

Table 5. Interaction of 8-substituted 6,7-dimethyllumazine derivatives with the lumazine apo-protein compared to riboflavin synthase [lo]. The lumazine apo-protein is A13 and is in 0.35 M Pi; none = no interac- tion detected; nd = not determined.

Substitution Apo-protein at Riboflavin synthase at

26c 2c 37 "C

Kf K, or Kf K,,, or K i

D-Ribityl D-Xylityl 3'-Deoxy-~-ribityl 5'-Deoxyribi tyl L-Threityl D-Erythrit yl L- Arabityl

1.-Lyxityl L-Ribityl L-Xyli tyl 6'-H ydroxyhexyl S'-Hydroxypentyl 4-Hydroxybutyl

3'-Hydroxypropyl 2'-Deoxyribityl D-Arabityl D-Glycerityl n-Lyxityl 1.-Glyceri tyl (S)-2'-Hydroxypropyl (R)-2'-Hydroxypropyl 2'-H ydroxyethyl

P M

0.8 0.8 0.3

~

nd nd

1.6 1

20 30 80 18 6 6

none none none none none none none none none

0.12 0.06 0.07 nd

nd 0.4 0.1

6 12 8

20 6

19

5 16 20 50 70 none none none nd

10 10 0.8

40 190 none none

none nd nd nd nd nd

nd none none nd none nd nonea nonea none

a Measured for the (RS) 3'-hydroxypropyl derivative

0.3

x Q 0

v)

K

L c 0 0.2 .-

a

0.1

U 0 100 200 300 400 500

Channels (1 05 ps / ch) Fig. 2. Decay of fluorescence anisotropy of the 2'-deoxy-~-rihityl- lumazine derivative showing the bound ligand with 4 = 23 ns. Excitation 380 nm, emission 480 nm, 2 ° C 0.05 M Pi.

mixture, corresponding to a dissociation constant not greater than 100 pM.

Binding rigidity

Fig. 3 shows the decay of the fluorescence anisotropy of the D-xylityl derivative under conditions where it must be

almost fully bound to the lumazine apo-protein. The measure- ment was made with high time resolution to emphasise that, even within the first z 100 ps (the time width of the instrumen- tal response function, dotted curve), the decay is the same as at longer times, with q5 = 23 ns. There is no evidence of a second more rapid component of decay of the bound ligand. The very rapid oscillation at the time of the excitation is due to the small amount of the free ligand which should be in equilibrium. Exactly the same picture is obtained for the native A13 lumazine protein (published elsewhere [25]) and for the P type, which is therefore not shown. The beginning of the excitation pulse is at about 0.5 ns so that any signal before that is due to random fluctuation.

For a control, Fig. 4 shows a 2: 1 mixture of native lumazine protein and the free D-ribityl derivative. An uncon- strained but associated analysis (see Appendix) of the decay data linking five experiments at high and low time resolution, recovers parameters corresponding quite closely to those ex- pected for the bound (1) and free (2) ligands: z1 = 14.8 ns associated with Cpl = 22 ns, z2 = 8.9 ns associated with +z = 100 ps, al /a2 = 2.1, x 2 = 1.41. In this mixture the oscillatory behavior is not apparent because its appearance depends on the relative contribution of the species to the fluorescence signal [26].

DISCUSSION

Table 5 also lists the enzyme interaction constants of yeast riboflavin synthase for comparison with the dissociation con- stants of the lumazine derivatives with the lumazine apo- protein. The substituents are divided into three groups on the basis of the strength of interaction with the apo-protein or the eniyme. In the first group are those compounds that have an interaction with either protein comparable to the native chromophore. It is seen from Scheme 1 (1-4, 6) that the strongly interacting substituents all have the same configu- ration as the natural ligand at the 2'-position. This includes

71 7 0.4

0.3

x Q 0

v)

L

5 0.2 .- 2

0.1

0 . ' ' ' ' ' I ' ' ' ' ' . ' I ' . 0 100 200 300 400 500

Channels (1 2.5 ps / ch)

Fig. 4. Decay of fluorescence anisotropy of native lumazine protein (A13, 40 pM) and 20 pM added D-ribityl derivative. The plot shows the rapid loss of anisotropy at early times due to the rotational diffusion of the unbound ligand. Excitation 380 nm, emission 480 nm, 2"C, 0.05 M Pi with 5 mM 2-mercaptoethanol, pH = 7.0.

the D-erythrityl (6) and L-arabityl (4) derivatives which were not reported to interact with the enzyme [lo]. Elimination of the 3'-hydroxyl (3) does not affect the apo-protein binding and this derivative is also a strong competitive inhibitor of the enzyme.

The derivatives in the second group starting with L-lyxityl (5), show significantly weaker dissociation constants (Kf). It is also apparent from the L-lyxityl case that, although the 2' orientation is favorable for binding, other factors must be involved. The longer-chain alcohol derivatives also bind weakly to the lumazine protein but this appears to be less specific than for the aldityl side chains. Although there is rather large uncertainty (> 50%) for the Kf values in this group, it is concluded that the alcohol derivatives have only a small temperature coefficient, implying that the interaction is largely entropic.

The last group is for those showing no interaction at 20 "C, with a weak effect observed for some at 2°C only. Some of the compounds falling in these last two groups were also reported in the riboflavin synthase study [lo]. These had no interaction with the enzyme, showing the advantage of the fluorescence method for detecting these weak protein interac- tions.

The stereospecificity of the 8-substituted position of the lumazine derivative for binding to the lumazine apo-protein parallels that found for the substrate and inhibitor specificities of riboflavin synthase [9, 101. In the enzyme studies it was proposed that the stereostructure of both the 2'- and 4 - hydroxy groups in the 8-aldityl substituent on the lumazine provides the critical factor required for interaction [lo]. For lumazine apo-protein however, it appears that only the 2'- position is critical for binding. Both D-erythrityl, without chirality at the 4'-position, and L-arabityl, with the L-configu- ration, both bind well, albeit less tightly than the D-ribityl.

These results imply a certain relationship between the structure of the binding site in the enzyme and lumazine pro- tein. The sequence similarity between the enzyme from Bacil- lus subtilis and lumazine protein has already given rise to this

same idea, but the yeast enzyme sequence, which is expected to be similar, is not available for direct comparison.

For those compounds that are strongly bound, the changes in stereochemistry of the 8-substituent do not produce dis- tinguishable fluorescence decay times (a measure of transition probabilities), rotational correlation times, or differences in the effects on these parameters of temperature and solution ionic strength. Previous studies have demonstrated that the binding site is highly exposed to solvent and induces high polarity in the ring system, suggested to arise from specific H- bond interactions [12, 27, 281. This means that the variations in structure available in the group in Table 4 do not allow the ring system to be 'locked in' in any way other than the same position as the natural ligand. Within the rather large uncer- tainty of the thermodynamic measurements, Table 4 also indi- cates that the same free energy of the interaction pertains. Although the dissociation constants for the apoproteins in Table 4 are several times larger than for the native protein (Table 3), especially in the case of the P type, again this may not be significant in view of the uncertainty involved in the estimate of the competence of the apoprotein preparation.

On this same point, Fig. 3 shows that in the binding site, the D-xylityl derivative is quite immobile, as has also been found for the native D-ribityl itself. However even the weakly binding ligands must also be immobilized probably in the same binding site, as seen in Fig. 2 by the appearance of the protein's rotational correlation time reflected in the ani- sotropy decay of the bound 2'-deoxy-~-ribityl derivative.

Immobilization of the lumazines in the binding sites of riboflavin synthase might also be expected based on the dem- onstrated relatedness to lumazine protein. An accurate po- sitioning of the two substrates may be an important factor in the enzyme's catalytic mechanism.

This work demonstrates the power of fluorescence decay analysis as a direct, unambiguous, and potentially rapid tech- nique for studying protein - ligand interactions. One reason for success in this present case is that the ligands are highly fluorescent and have separable fluorescence lifetimes, both free and bound. The powerful data analysis methods also allow many experiments to be combined with an improvement in statistics. However one drawback here is that to obtain measurable concentrations of both free and bound species, the sample concentrations need to be in the range of the equilibrium constant, and for those tightly bound species these dilute solutions entail long counting times, usually many hours. However, the most troublesome factor is the one gener- ally encountered in work with apo-proteins; estimation of the rebinding competence of the apo-protein preparations [29]. The slow equilibration time combined with loss of combining activity of the apo-protein further complicate the issue.

Steady-state anisotropy has been used before in studies of this protein and also in many other investigations. Light scattering, especially Raman, is very difficult to allow for in this technique and particularly for the very dilute solutions that have to be employed, leading to estimates of the dis- sociation constants that are too low. The employment of steady-state fluorescence anisotropy (or polarization) for pro- tein studies therefore must be made with due caution.

APPENDIX

Fluorescence dynamics

Two types of measurement of the emission decay were performed depending on whether rotational information was

71 8

desired or not. In the first type the total fluorescence was obtained by using magic angle conditions and in the second case vertical and horizontal data sets were collected. The data collection was continued until about 2 x lo4 counts were in the peak channel, then the data transferred to the computer. Several experiments made with different conditions, e. g. time/ channel, were combined for final analysis. The analytical tech- nique first entailed rigorous deconvolution of the data sets from the laser pulse data set followed by fitting to the model equation. For the total fluorescence this was the sums of exponentials :

where z is the fluorescence liftime, CI is the amplitude with i = 2 or 3 in the present work.

The deconvolved vertical, Zv(t), and horizontal, zh(f), inten- sities were simultaneously fitted to:

S(t) = Zpi . exp( - t / z i ) (1)

31, = S(t) . [l + 2r(t)] (2)

3g . = s(f) ’ [I -r(t)]. (3) The g-factor is determined experimentally but will be set equal to unity here to simplify the equations.

Then S(t) = I, + 2 . Zh (4)

r(t) = D(t)/S(t) (5)

and the time-dependent anisotropy is

where D(t) = Zv-Zh.

A monoexponential decay model is adequate for most of the experiments here :

However a more complex equation is used in one case (Fig. 4) the ‘associated’ analysis, where there are two rotating species, with each + belonging to a specific z:

r( t ) = /3 . exp( - ti+). (7)

R(t) . S(t) = r l ( t ) . cxl . exp( - t / z l ) + r2 ( t ) . cx2 . exp( - j i z z ) .

(8)

Analysis of equilibria by fluorescence decay

If the fluorophore can be considered to have a simple monoexponential decay with lifetime z, then the total photons emitted, S, is the integral of S(t) in Eqn (1), i.e.

(9) S=CC.z

= u . Q F (10) where a is the amount of exciting light absorbed, and QF is the fluorescence quantum yield.

From Beer’s Law and the usual simplifying approxi- mations that can be applied for low concentrations :

(1 1) S K 8 . c . QF where E is the absorption coefficient and c the molar concen- tration. For simplification, setting the proportionalities con- stant to unity, we can write

For a solution consisting of two fluorophores, such as the free and bound lumazine derivative under consideration in this present work. the assumption of exponential decay for each species is applicable and a further simplification can be made since it is observed that

(12) c = CI . Z / ( E ’ QF) .

Then (14)

For technical reasons, the dynamic fluorescence measure- ments cannot be made at the isosbestic point but instead at 380 nm where, setting c1 as the bound lumazine

E J E ~ = 1.45. (1 5 )

Analysis of equilibria by steady-state anisotropy

the steady-state fluorescence anisotropy is defined as : For a fluorescent molecule in solution at concentration c,

( r ) = D / S (16)

i. e. the difference between the fluorescence intensity when the polarizer is oriented parallel (FPAR) to the direction of Dolarization of the excitation radiation and perpendicular ( F P E R ) to it.

is the total fluorescence intensity. For simplicity again, the instrumental g-factor allowing for the difference between transmission through the monochromator for polarization direction, is assumed to be unity. In the case of a solution of two fluorescing substances at concentrations c1 and c2, then from Eqn (12):

and S = Q F I . ~1 ‘ CI + Q F 2 ’ E2 ’ ~2

D = F P A R , ~ - F P E R , ~ + FPAR,Z-FPER,Z

(19)

(20)

= D 1 + D 2 (21)

In the present experiments for steady-state anisotropy, the excitation is at an isosbestic point so that

Setting R = Q F ~ / Q F Z

by substituting for S and D, and by rearrangements, an ex- pression for the mole fractionf, of the c1 component can be derived (cf. Eqn 5.58 of Lakowicz [19] which appears to be similar but there is some confusion of symbols):

If the fluorescence quantum yields of the two components are the same, this expression reduces to a simple additivity

This work was supported by the US National Institutes of Health (grant GM28139) and by grants from the Deutsche Furschungsgemein- schaft and the Fonds der Chemischen Industrie.

REFERENCES 1. Bacher, A. (1991) in Chemistry and biochemistry ofj7avuenzymes

(Muller, F . , ed.) vol. I, pp. 215-259, CRC Press, Boca Raton FL.

719

2. Bacher, A. (1991) in Chemistry and biochemistry offlavoenzymes (Muller, F., ed.) vol. I, pp. 293-316, CRC Press, Boca Raton FL.

3. Lee, J., Matheson, I. B. C., Muller, F., O’Kane, D. J., Vervoort, J. & Visser, A. J. W. G. (1991) in Chemistry and biochemistry offlavoenzymes (Muller, F., ed.) vol. 11, pp. 109-151, CRC Press, Boca Raton FL.

4. O’Kane, D. J., Woodward, B., Lee, J . & Prasher, D. C. (1991) Proc. Nut1 Acad. Sci. USA 88, 1100-1104.

5. O’Kane, D. J. & Prasher, D. C. (1992) Mol. Microbiol. 6, 443- 449.

6. Otto, M. K. & Bacher, A. (1981) Eur. J. Biochem. 115, 511 -537. 7. Schott, K., Kellerman, J., Lottspeich, F. & Bacher, A. (1990) J .

8. O’Kane, D. J. & Lee, J. (1985) Biochemistry 24, 1467-1475. 9. Winestock, C . H., Aogaichi, T. & Plaut, G. W. E. (1963) J . Bid.

Chem. 238,2866-2874. 10. Plaut, G. W. E. & Beach, R. L. (1976) in Flavins andflavoproteins

(Singer, T. P., ed.) pp. 737 - 746, Elsevier, Amsterdam. 11. Visser, A. J. W. G. &Lee, J. (1980) Biochemistry 19,4366-4372. 12. Lee, J., O’Kane, D. J. & Visser, A. J. W. G. (1985) Biochemistry

13. Visser, A. J . W. G. & Lee, J. (1982) Biochemistry 21, 2218-2226. 14. O’Kane, D. J. & Lee, J. (1986) Methods Enzyrnol. 133, 149-172. 15. O’Kane, D. J., Karle, V. A. & Lee, J. (1985) Biochemistry 24,

16. O’Kane, D. J., Lee, J., Kohnle, A. & Bacher, A. (1990) in Chemis- try andbiology ofpteridines 1989 (Curtius, H.-Ch., Ghisla, S. & Blau, N., eds) pp. 457-461. Walter de Gruyter Co., Berlin.

Biol. Chem. 265,424 - 428.

24, 1476- 1483.

1461 - 1467.

17. Cresswell, R. M. & Wood, H. C. S. (1960) J . Chem. SOC., 4768. 18. Reuke, B., Korn, S., Eisenreich, W. & Bacher, A. (1992) J .

Bacteriol. 174, 4042 -4049. 19. Lakowicz, J. R. (1983) Principles of fluorescence spectroscopy,

Plenum Press, New York. 20. Lee, J., Carreira, L. A., Gast, R., Irwin, R. M., Koka, P., Small,

E. D. & Visser, A. J. W. G. (1981) in Bioluminescence and chemiluminescence (DeLuca, M. & McElroy, W. D., eds) pp. 103 - 112, Academic Press, New York.

21. Wampler, J. (1991) Anal. Instrum. 19, 203-230. 22. Lee, J., O’Kane, D. J. & Gibson, B. G. (1989) Biophys. Chem. 33,

23. Beechem, J. M., Gratton, E., Ameloot, M., Knutson, J. R. & Brand, L. (1991) in Topics in fluorescence spectroscopy (Lakowicz, J. R., ed.) vol. 2, pp. 241 - 305. Plenum Press, New York.

24. Beach, R. L. & Plaut, G. W. E. (1971) J. Org. Chem. 36, 3937- 3943.

25. Lee, J., Bradley, E. A. & O’Kane, D. J. (1992) Proc. SOC. Photo- Opt. Instrum. Eng. 164, 148 - 158.

26. Lee, J., Wang, Y. & Gibson, B. G. (1990) Anal. Biochem. 185,

27. Kulinski, T., Visser, A. J. W. G., O’Kane, D. J. & Lee, J. (1987) Biochemistry 26, 540- 549.

28. Vervoort, J., O’Kane, D. J., Muller, F., Bacher, A. & Lee, J. (1990) Biochemistry 29, 1823- 1828.

29. Muller, F. &van Berkel, W. J. H. (1991)in Chemistryandbiochem- istry offlavoenzymes (Muller, F., ed.) vol. I, pp. 261 - 274, CRC Press, Boca Raton FL.

99- 111.

220-229.