Two Lysine Residues in the Bacterial Luciferase Mobile Loop Stabilize Reaction Intermediates

Bacterial luciferase catalyzes the reaction of FMNH2, O2, and a long chain aliphatic aldehyde, yielding FMN, carboxylic acid, and blue-green light. The most conserved contiguous region of the primary sequence corresponds to a crystallographically disordered loop adjacent to the active center (Fisher, A. J., Raushel, F. M., Baldwin, T. O., and Rayment, I. (1995) Biochemistry 34, 6581–6586; Fisher, A. J., Thompson, T. B., Thoden, J. B., Baldwin, T. O., and Rayment, I. (1996) J. Biol. Chem. 271, 21956–21968). Deletion of the mobile loop does not alter the chemistry of the reaction but decreases the total quantum yield of bioluminescence by 2 orders of magnitude (Sparks, J. M., and Baldwin, T. O. (2001) Biochemistry 40, 15436–15443). In this study, we attempt to localize the loss of activity observed in the loop deletion mutant to individual residues in the mobile loop. Using alanine mutagenesis, the effects of substitution at 15 of the 29 mobile loop residues were examined. Nine of the point mutants had reduced activity in vivo. Two mutations, K283A and K286A, resulted in a loss in quantum yield comparable with that of the loop deletion mutant. The bioluminescence emission spectrum of both mutants was normal, and both yielded the carboxylic acid chemical product at the same efficiency as the wild-type enzyme. Substitution of Lys283 with alanine resulted in destabilization of intermediate II, whereas mutation of Lys286 had an increase in exposure of reaction intermediates to a dynamic quencher. Based on a model of the enzyme-reduced flavin complex, the two critical lysine residues are adjacent to the quininoidal edge of the isoalloxazine.

The initial step in the luciferase reaction mechanism is the reversible binding of FMNH 2 ( Fig. 1 and Ref. 1). Enzyme-bound FMNH 2 subsequently reacts with molecular oxygen yielding the 4a,5-dihydro-4a-hydroperoxyflavin or intermediate II (1,2). Intermediate II is unstable at room temperature but is sufficiently stable at low temperatures to allow for chromatographic purification in the presence of long chain alcohols (3,4). Intermediate II reacts with a long chain aliphatic aldehyde to form intermediate IIA (1,5). The decomposition of intermediate IIA results in the formation of water, FMN, carboxylic acid, and a photon of blue-green light (1,5).
Luciferase is a heterodimer composed of two homologous subunits designated ␣ and ␤ (6). Although the ␤ subunit is required for activity, the catalytic site resides exclusively on the ␣ subunit (7,8). The majority of mutations known to result in kinetic defects reside within a single solvent-exposed cleft on the ␣ subunit (7,9,10). Near this cavity is the most stringently conserved region of the enzyme, spanning residues 262-290 on the ␣ subunit (Fig. 2a). This portion of the enzyme corresponds to a protease-labile mobile loop disordered in the two reported substrate-free luciferase structures (10 -12). The location of the active center has recently been experimentally demonstrated (13). The asymmetric unit of crystals of the luciferase-flavin complex contained ␣ subunits from two non-symmetry-related heterodimers. Neither ␣ subunit contained interpretable electron density for a small segment between residues 283 and 290 (13) (Fig. 2b). Unlike the high resolution structure (10), the sequence including residues 262-283 was observed. Despite this progress in structural characterization of the mobile loop, relatively little progress has been made in assigning functional roles to residues in the mobile loop.
At least two nonexclusive roles are conceivable for the bacterial luciferase mobile loop. First, it has been suggested that the mobile loop might serve to protect reaction intermediates from quenching by solvent (14) Second, the mobile loop might afford a docking surface allowing transient complex formation between enzyme and a flavin oxidoreductase that supplies the enzyme with FMNH 2 . The former hypothesis is attractive because the mobile loops of other triosephosphate isomerase barrel proteins have been shown to serve this role (15). The latter hypothesis is attractive because the sequence of the loop region is conserved across luminous bacteria, and complex formation would require structural specificity. The experiments reported here were undertaken to investigate these nonexclusive possibilities.
A variant has been constructed in which the entire mobile loop was deleted (14) This deletion results in loss of ϳ8% of the luxA gene. Although the tertiary structure, the yield of the carboxylic acid, and the affinity for aldehyde, FMN, and FMNH 2 were relatively unaltered, the total quantum yield was reduced by 2 orders of magnitude. It was suggested that the cause of this reduction was an inability to shield reaction intermediates from bulk solvent (14) This hypothesis was formulated based on the observation that many triosephosphate isomerase barrel enzymes utilize substrate sensitive mobile loop movements for analogous lid gating mechanisms (15,16). To determine whether the enzymatic properties of the loop deletion mutant were due to the loss of specific residues, we targeted positions throughout the mobile loop for mutagenesis in an attempt to disrupt the "seal" of the active center provided by the mobile loop.
An additional focus of the present study was to define the possible role of the mobile loop in facilitating the transient association of accessory proteins. Luciferase requires reduced flavin to catalyze the bioluminescence reaction (17). Maintaining a constant level of reduced flavin may be difficult during aerobic growth, because free reduced flavin readily reacts with molecular oxygen (18). It has been proposed that a complex is formed between luciferase and flavin-oxidoreductase enzymes in which FMNH 2 is transferred directly from the reductase to the active site of luciferase (19,20). The transfer model implies molecular specificity in the transient complex between the two enzymes. Bacterial luciferase is highly active in a variety of recombinant bacteria such as Escherichia coli (21,22). The gene encoding the bioluminescence-supporting oxidoreductase in E. coli, fre, was recently identified and characterized (23) The loop is a promising potential docking platform between luciferase and the oxidoreductase based on both stringent sequence conservation and proximity to the active center ( Fig. 2, a and b, and Ref. 13, 24, and 25). If a given residue within the loop facilitates flavin transfer, bioluminescence may be impaired in assays utilizing enzymatic reduction of FMN but display normal levels of activity when supplied with excess catalytically reduced FMN. Therefore, activity measurements were obtained for point mutants throughout the mobile loop using activity measurements in vivo and by a second method where the flavin is reduced chemically and provided by free diffusion in vitro.

EXPERIMENTAL PROCEDURES
Chemicals-All of the chemicals were obtained from Sigma-Aldrich unless otherwise noted and were of reagent grade or higher. The cells were cultured in standard Luria-Bertani broth or agar supplemented with 100 g/ml ampicillin.
Vector Construction and Mutagenesis-Using a high fidelity polymerase (Pfu Turbo; Stratagene), Vibrio harveyi lux AB was amplified from pJHD500 plasmid using the nucleotide primers 5Ј-gagcccctcgagcgagtgatatttg (sense) and 5Ј-ccatatgaaattcggaaacttccttc (antisense) (26) All DNA mutagenesis and sequencing primers were obtained from IDT DNA. Amplification with this pair of primers led to incorporation of an NdeI site and an XhoI site downstream. After digestion with both restriction endonucleases (New England Biolabs), fragments were gel-purified (Qiaex gel purification system) and ligated into a prepared pET21-b vector (Novagen). The vector resulted in the addition of a series of six histidine residues onto the C terminus of the LuxB gene. Sequencing of the entire insert was used to verify fidelity (Arizona Research Laboratories Sequencing Facility, University of Arizona). The mutants were generated from pZCH2 using sitedirected mutagenesis (27). These clones were verified by local sequencing at the site of mutation.
Luciferase Purification-Luciferase protein was expressed from pZCH2 in a BL21 (DE3) cell line after growth to an A 600 of 0.5 (Stratagene). Expression was initiated by the addition of isopropyl ␤-D-thiogalactopyranoside to 1 mM. Expression was continued for ϳ6.5 h at 25°C with constant agitation. Clarified lysate was applied to a custom nickel affinity column (Amersham Biosciences) and purified to Ͼ90% purity assessed by SDS-PAGE analysis (28). Purified protein was dialyzed extensively into buffer containing 100 mM Na ϩ /K ϩ phosphate and 100 mM NaCl, pH 7.0.
Determination of Protein Concentration-Concentrations of luciferase and the individual ␣ and ␤ subunits were determined using extinction coefficients at 280 nm of 1.136, 1.41, and 0.71 (mg/ml) Ϫ1 cm Ϫ1 , respectively (29).
Western Blotting-A small volume of each culture was withdrawn and normalized for cell density at 600 nm. The cells were collected by centrifugation and lysed in an equal volume of 200 mM Tris-HCl, 200 mM dithiothreitol, 4% SDS, 0.2% bromphenol blue, and 25% glycerol. The samples were then boiled for 5 min and briefly centrifuged. Subsequent electrophoresis was conducted using 12.5% SDS-PAGE gels prior to transfer onto a nitrocellulose membrane (28,30). The membrane was incubated overnight at 4°C in blocking buffer (1% powdered milk with 0.05% Tween 20). The membrane was washed repeatedly in phosphate-buffered saline containing 0.05% Tween 20. Luciferase was detected with a polyclonal anti-serum (1:5000 dilution in blocking buffer) followed by exposure to a Cy dyeconjugated anti-rabbit secondary antibody (Odyssey IRDye). Unbound secondary antibody was removed by repeated washing with phosphate-buffered saline. Fluorescence excitation was carried out using an Odyssey Li-COR IR imager at 680 nm. Emission was detected at 700 nm and automatically corrected for background.
Activity Assays-Specific activities were determined by the flavin injection assay (31). Enzyme was incubated in 1.0 ml of 100 mM Na ϩ /K ϩ phosphate containing 0.5 mg/ml bovine serum albumin and 0.001% aldehyde. The reactions were initiated with the rapid injection of 1.0 ml of 50 M photo-reduced FMNH 2 . Reduced flavin was obtained by photo-reduction using white fluorescent light in the presence of EDTA (14). Peak luminescence was recorded using a custom bench top luminometer (32). Aldehyde dependence was measured using the flavin injection assay varying the aldehyde incubated with enzyme. The rate of luminescence decay (k ϭ 0.69/t1 ⁄ 2 ) was determined from the elapsed time for the light intensity to decay by 50% (t1 ⁄ 2 ) (33). Total quantum yield was determined based on the equation I(t) ϭ I o2 Ϫt/t (1/2) . The yield was calculated based on integration over the first 100 s of the reaction based on observed values of I o and t1 ⁄ 2 . Activity measurements in vivo were made on 1.0-ml aliquots. Light emission was recorded following the rapid injection of 1.0 ml of 0.1% decanal (v/v in water). Peak luminescence was reached ϳ5 s after injection.
Dissociation constants for flavin were determined by the dithionite method as described (34) Intermediate II was formed and assayed as described (35). K d values for aldehyde were determined by the using the standard flavin injection method (31).
Bioluminescent Emission Spectra-Samples contained 1 nm V. harveyi Frp, 2 M FMN, 1 mM NADPH, 80 M n-decyl aldehyde suspension, and 10 M protein solution in 100 mM Na ϩ /K ϩ phosphate, 100 mM NaCl, pH 7.0. The spectra were acquired on Cary Eclipse fluorimeter lacking excitation. The instrument detected relative light intensity as a function of wavelength using a wavelength dispersion for emission band pass fixed at 5 nm. The scan speed was ϳ500 nm/min. The disordered loop is stringently conserved among ␣ subunits but absent from all known ␤ subunits. The BLASTP search algorithm was used to find these sequences, which were subsequently aligned with a BLOSUM 62 scoring function within ClustalX (56,57). b, Residues 262-290 are shaded in either yellow or cyan corresponding to either the FMN-bound or FMN-free ␣ subunit, respectively. c, bioluminescence emission in vivo. Mutant proteins were expressed following supplementation of 10-ml cultures to 1 mM isopropyl ␤-D-thiogalactopyranoside and ϳ6.5 h of expression at 25°C prior to assay using decanal. The values were corrected for cell density and normalized to the bioluminescence of wild-type enzyme. d, qualitative assay of mutant luciferases. Colonies were induced for 6 h at 25°C prior to imaging with a GE Healthcare Typhoon scanner. The aldehyde substrate was provided by vapor diffusion. Wt, wild type. NOVEMBER 20, 2009 • VOLUME 284 • NUMBER 47

JOURNAL OF BIOLOGICAL CHEMISTRY 32829
Collisional Quenching-The experiments were conducted at 20°C essentially as described (36). Decay rates and initial velocities were determined from samples containing ϳ1 M luciferase and an appropriate concentration of KCl to maintain a constant ionic strength of 250 mM.
Decanoic Acid Production-Carboxylic acid analysis was performed as described with minor modifications (14). Gas chromatography was used to quantify the yield of decanoate following a single enzymatic turnover. 1.0 ml of photo-reduced FMNH 2 was injected into 3.0-ml samples containing 50 M protein and ϳ10 M sonicated decanal (31). 1.0-ml aliquots were withdrawn and supplemented with 100 l of a 200 g/ml dodecanoic acid solution as internal standard. The sample was acidified with 50 l of 4 M H 2 SO 4 . The acid was removed with the addition of 1.0 ml of CH 2 Cl 2 . This mixture was vortexed for 1 min vigorously prior to removal of the organic phase. The acid was then derivitized with the addition of 0.4 ml of N-methyl-Ntrifluoroacetamide and 10 min of incubation at 37°C. 1 l of each sample was loaded onto a 5988A gas chromatography/ mass spectrometry. It was found that the decanoic acid had a retention time of 3.61 min, and the dodecanoic acid standard had a retention time of 9.29 min. The total area under each curve was integrated for both peaks. The values for the decanoic acid yield are given relative to the quantity of the dodecanoic acid internal standard.

RESULTS
Point mutants were generated using site-directed mutagenesis at 15 of the 29 mobile loop residues between positions 262 and 291. This collection represents all of the charged residues in the mobile loop plus the single mobile loop tryptophan residing at position 277. Activity measurements, normalized for cell density, were determined in vivo 6.5 h post-induction (Fig. 2, c  and d). Nine mutants had reduced levels of activity in vivo. To exclude the possibility that reduced activity was due to changes in protein synthesis and/or folding, the mutants were analyzed by Western blots (data not shown). The level of expression for all of the mutants was the same as for the wild-type control. To determine relative activity using a nonenzymatic method for providing reduced flavin, mutants with decreased activity in vivo were purified and subjected to detailed kinetic analysis ( Table 1). All of the mutants had specific activities in vitro within a factor of 2 of the activity obtained in vivo.
The turnover number of luciferases from different bioluminescent bacterial species varies depending on the aldehyde chain length, whereas the quantum yield is independent of chain length (5,37). Despite the characterization of multiple mutations that alter the aldehyde chain length dependence, the molecular basis of the aldehyde chain length dependence of bioluminescence decay remains unclear (9,38). The possible role of residues within the mobile loop in the aldehyde alkyl chain length preference was investigated. The bioluminescence decay rates using octanal, decanal, and dodecanal for the V. harveyi luciferase have been referred to as "slow, fast, and slow", respectively (Table 1 and Ref. 39). Three mutants, K283A, K286A, and R291A all show considerable increases in the decay rate of bioluminescence using the octanal substrate, relative to the wild-type enzyme. These mutants show a slower decay rate of bioluminescence than wild type using the decanal substrate. The majority of the mutants had only minor changes in the decay rate of bioluminescence when assayed with dodecanal, whereas there were substantial changes for decays with octanal and decanal.
Many of the mutations known to induce shifts in the wavelength of peak light emission reside in or near the active center (7,9,33,40). The emission spectrum of each mutant was recorded using a coupled assay system containing purified V. harveyi Frp oxidoreductase supplied with NADPH (23). Based on comparison with wild-type enzyme, none of the alanine mutants appear to cause significant alterations in the color of light emission (Table 1). This result is in agreement with the lack of color change in the whole loop deletion mutant (14).
The location of the aldehyde binding site has not been determined. To examine whether Lys 283 and/or Lys 286 interact with the aldehyde substrate, K m values for decanal were determined for each mutant. The initial velocity was recorded at increasing aldehyde concentrations using the flavin injection technique a Specific activities of purified protein were determined by the standard FMNH 2 injection assay (31). The values listed in the table were normalized to the wild-type enzyme.
Specific activity values for wild-type enzyme using octanal, decanal, or dodecanal were 3.8, 14, and 7.6 ϫ 10 12 ⅐Q⅐s Ϫ1 ⅐mg Ϫ1 respectively. The error was Ͻ20% of the reported values. b The activity was measured in vivo using only the decanal substrate. The error was Ͻ25% of the reported values. c The decay rate constants are in units of s Ϫ1 . Half-lives were determined based on the time for bioluminescence to decay from 80% of the maximum light intensity to 40%. The error in the decay rate constants was Ͻ50% of the reported values. d Total quantum yield was determined using the initial velocity and decay constant from Table 1 summed over the first 100 s of exponential decay. The values listed in the table were normalized to the wild-type enzyme. These values were nearly identical for wild-type enzyme using the three different aldehyde substrates. The error was Ͻ50% of the reported values. e Bioluminescence spectra were recorded in vitro using the Frp coupled assay. The error in emission maxima was Ͻ5 nm from the reported values. (31). A version of the Michaelis-Menten equation was fit to the data to include the effects of substrate inhibition at high concentrations (14) Wild-type enzyme had a K m of 8 M, whereas K286A and K283A bind decanal with Michaelis constants of 10 and 10.5 M, respectively. Therefore, neither of these mutations appears to significantly impede binding of decanal.
In the structure of the luciferase-FMN complex, the segment of the mobile loop containing Lys 286 and Lys 283 is near the flavin phosphate group. Therefore, the dissociation constant (K d ) for FMNH 2 was determined for K283A and K286A. In this assay, the concentration of FMNH 2 was varied, and the reaction was initiated with the rapid injection of air-equilibrated aldehyde (34). The initial velocity was determined at different concentrations of reduced flavin and used to perform a nonlinear least squares fitting analysis (14). The K d we obtained for wildtype enzyme was 0.9 M, in good agreement with the micromolar value obtained in prior measurements (14). K283A and K286A bind reduced flavin more weakly, with dissociation equilibrium constants of 1.1 and 6.8 M, respectively.
To determine whether the low levels of light emitted by K283A and/or K286A were related to stability of intermediate II, the decay rate was determined using a modified version of the double injection assay (4). In this assay, photo-reduced flavin is rapidly injected into air equilibrated buffer containing luciferase. In the absence of aldehyde, intermediate II decays, yielding hydrogen peroxide and oxidized FMN. Following a secondary injection of aldehyde, the peak light intensity is directly proportional to the remaining quantity of enzymebound intermediate II. The decay process obeys a pseudo-first order kinetic profile. The half-life at 0°C for wild-type enzyme was 53.8 min compared with 23.7 and 5.9 min for the K286A and K283A mutants, respectively ( Fig. 3 and Ref. 8). Thus, the stability of intermediate II upon substitution of lysine for ala-nine at position 283 leads to nearly a 10-fold reduction in the kinetic stability of intermediate II.
Potassium iodide is a dynamic quenching reagent routinely used to probe singlet excited states (41). Collisional quenching is thought to occur because of intersystem crossing to a triplet excited state caused by spin orbit coupling of the singlet excited state and the halide (42). Iodide quenches both the long-lived intermediate responsible for the production of the excited state emitter and the excited state emitter itself, thus decreasing initial light intensity (V o ) and increasing the rate constant for decay of bioluminescence (k) (36) The Stern-Volmer plot primarily indicates the accessibility of the excited state emitter to the quenching agent. Stern-Volmer plots can be either linear or curved. The latter can be indicative of both quenching and a binding event (data not shown). We observed a curved plot indicative of a binding event. Iodide is also believed to quench the intermediate(s) that generate the excited state emitter (36). The decay rate of bioluminescence was significantly increased in the presence of iodide for K283A but only slightly for K286A (Fig. 4). Our interpretation of this result is that the intermediate responsible for populating the excited state emitter (i.e. either the peroxyhemiacetal or a dioxirane) is more solvent-exposed upon substitution of lysine at position 283 with alanine.
To establish whether the observed reduction in luminescence corresponds to a reduction in the yield of the carboxylic acid product, the acid product was quantified using gas chromatography/mass spectrometry as previously described (14). Because of spontaneous oxidation of the decanal to decanoic acid in solution, the basal quantity of acid production was determined in addition to each of the samples. The amount of acid produced in the absence of enzyme was 0.971 g/ml. Based on comparison with the internal standard, using a single turnover assay in the absence of bovine serum albumin, wild type, K283A, and K283A produced 1.5, 1.54, and 1.7 g/ml concentrations of decanoic acid, respectively. Therefore, the yield of the acid product is not significantly altered upon mutation to alanine at either position in the mobile loop. Intermediate II was formed at 0°C by rapidly injecting photo-reduced FMNH2 into 1.0 ml of 100 mM Na ϩ /K ϩ phosphate and 100 mM NaCl, pH 7.0, containing 0.1% dodecanol and 800 g of luciferase. 50-l aliquots were withdrawn as a function of time and in 1.0 ml of buffer containing 100 mM Na ϩ /K ϩ phosphate and 100 mM NaCl, pH 7.0, at 23°C. Light emission was detected following the injection of 1.0 ml of buffer containing 0.05% decanal. In the luciferase-FMN complex, two basic side chains reside within an appropriate distance and orientation to coordinate the phosphate group of the flavin, Arg 125 and Arg 107 (13). Mutation of Arg 107 to glutamate effectively inactivates luciferase and greatly reduces the affinity of the enzyme for substrate (43,44). To examine whether Lys 283 or Lys 286 directly contacts the phosphate group of the flavin, both positions were mutated to glutamate. Both mutants appeared to have wild-type levels of activity in vivo, in stark contrast to the corresponding alanine mutants (data not shown).

DISCUSSION
The mobile loop from bacterial luciferase was subjected to alanine mutagenesis for two major reasons. First, the region of the protein containing the mobile loop is vital for catalysis in many enzymes with the triosephosphate isomerase barrel fold. The canonical triosephosphate isomerase barrel consists of an 8-fold repeat of parallel ␤-strands forming an inner barrel surrounded by eight ␣-helices on the exterior of the protein (45). The triosephosphate isomerase barrel is present in ϳ10% of known structures (16). The approximate location of the active center near the C-terminal ends of ␤ strands appears to be conserved, suggesting divergent evolution from a common ancestor (15). The major source of variation between triosephosphate isomerase barrel enzymes are variable length loop regions connecting secondary structural elements (16). The archetypical triosephosphate isomerase barrel, triose phosphate isomerase, catalyzes the conversion of the ketose, dihydroxyacetone phosphate, into the aldose, glyceraldehyde-3-phosphate. The enediol intermediate is susceptible to loss of phosphate by ␤-elimination (46). To avoid this side reaction, triosephosphate isomerase uses a mobile loop as a lid to sequester the active site following ligand binding (47) Similar lid gating mechanisms are common, making those with uncommon properties particularly interesting (48) Luciferase is unusual because of its distinct reaction chemistry and an unusually long mobile loop consisting of ϳ30 residues. The first aim of this study was to determine whether any labile intermediates might be protected by the bacterial luciferase mobile loop.
We found that mutation of either of the mobile loop residues Lys 283 or Lys 286 to alanine led to a substantial decrease in the stability of intermediate II. This was more pronounced for the K283A mutant. The kinetic stability of intermediate II is decreased upon substitution at a number of active center residues including His 44 , Asp 113 , Cys 106 , and Val 173 (40, 49 -51). Because of the distance from the isoalloxazine to the mobile loop, the finding that mutations in the mobile loop can alter the stability of intermediate II is surprising but not without precedent. In a prior study, two mutations in the mobile loop (G275P and F261D) were found to result in a decreased stability of intermediate II (52). Unlike the lysine mutants we described, the G275P and F261D were unable to effectively oxidize the aldehyde substrate to form the acid product (52). This finding is interesting given the extreme physical separation between the flavin distal portion of the mobile loop and the experimentally determined active center (13). The G275P and F261D mutations may have altered the dynamics of the loop in such a way that luciferase was unable to effectively adopt a closed conformation of the mobile loop (52).
Based on dynamic quenching analysis, we found that the K286A mutation resulted in an inability to protect the longlived intermediate responsible for the production of the excited state emitter. Following reaction of the aldehyde substrate with intermediate II, it is unclear how the singlet excited state is populated. We favor the formation of a dioxirane intermediate prior to population of the excited state (5,53). According to this mechanism the excited state emitter forms following homolytic cleavage of the peroxide bond. Potentially, there are multiple causes for an increase in the decay rate of bioluminescence in the presence of a dynamic quenching agent, such as a defect in the conversion of the intermediate IIA into the dioxirane. Indeed, the effect could be as simple as solvent-mediated deprotonation of N-5 and elimination of the peroxy group or the peracid. Additional NMR studies utilizing isotopically labeled flavins may allow discrimination between these hypotheses.
The second question we sought to examine in this work was whether the mobile loop may play a role in complex formation between luciferase and an NAD(P)H-dependent oxidoreductase (30). A complex has been proposed involving luciferase and oxidoreductase in which FMNH 2 is transferred directly from the reductase to the luciferase (19,20,54). Transfer obviates the problem of reaction of reduced flavin with molecular oxygen in solution. Implicit in this model is the requirement of molecular specificity in the transient complex. The features of this complex have yet to be described. We were interested in screening for sites in the mobile loop responsible for providing the oxidoreductase binding interface. The expectation was that mutants defective in flavin transfer because of a docking defect should display high levels of bioluminescence when assayed with an exogenous source of flavin in vitro. Therefore, we compared normalized activity measurements both in vivo and in vitro. We did not observe a significant difference between the two measurements, suggesting that either (i) the oxidoreductase interface requires a substantial number of residues and is insensitive to a single mutation, (ii) we selected the wrong set of residues, or (iii) there is no complex between luciferase and the oxidoreductase.
Concurrent with this work, we investigated the source of reduced flavin during bioluminescence in E. coli (23). In this study, the oxidoreductase enzyme (fre) from a nonbioluminescent bacterium (E. coli) was found to be as competent in providing the reduced flavin substrate in vivo or in vitro as the endogenous enzyme in V. harveyi. If direct flavin transfer occurs between proteins, specificity must exist in the encounter complex between a given reductase and luciferase to ensure proper positioning of the oxidoreductase over the active site. Given the lack of sequence similarity between the oxidoreductase from E. coli (Fre) and the endogenous enzyme in V. harveyi (Frp), it is unlikely that a complex is formed between enzymes. Further, the E. coli reductase did not physically interact with luciferase based on a pulldown assay (23). Therefore, it was not surprising that following mutagenesis of the mobile loop, a large discrepancy in activity caused by defective flavin transfer was not observed.
In addition to examination of the aforementioned aspects of the mobile loop, several observations regarding the structure of this region of the enzyme have been made. Structural data on the mobile loop were unavailable until the recent description of the luciferase-FMN crystal structure (13). The lysine mutants reported here were near or within a disordered region between residues 283 and 290 in this structure. This disordered region of the mobile loop is adjacent to the phosphate group of the flavin. The mobile loop may interact with the anion site directly via ionic interaction or indirectly through contacts to the residues that bind the anion. Direct interaction is conceivable based on the compositional bias toward charged residues in the region. The approximate distance separating the mobile loop from the flavin-binding site is also compatible with a direct interaction (13). The K286A mutant had an increase in the K d for FMNH 2 to 6.8 M. We therefore tested whether either lysine residue directly interacted with the anion-binding site for the 5Ј phosphate group of FMNH 2 . Nonconservative mutation at either position to glutamate resulted in a highly active enzyme. It appears that neither lysine residue is involved in directly binding the anion of the flavin. Thus, we were unable to find clear evidence supporting a direct interaction between either lysine residue and the 5Ј phosphate group of FMNH 2 . Based on the properties of the alanine mutants relative to substitutions with glutamate, it appears that the size of the residue at either location is of greater importance for catalytic function than charge state. Additional mutagenesis experiments probing the relationship between activity and residue size are required to test this model.
To characterize the dynamics of the loop region, a recent investigation simulated the dynamics of the luciferase ␣ subunit using replica exchange molecular dynamics. 2 The authors described a model for the enzyme:FMNH 2 complex by incorporating decades of experimental results with detailed network analysis of a vast collection of loop conformations. At the time of this study, only two point mutants had been described in the loop region (52). To examine the agreement of the present study with this proposed complex, a heat map was devised using the severity of mutation on in vivo activity in relation to the site of the mutation in the bound model (Fig. 5). The two most severe mutants were extremely close to the quinoidal face of the isoalloxazine. This result agrees with our proposal that exclusion of solvent from the active center requires moderately sized residues at both positions 283 and 286.
In summary, we were able to identify residues in the mobile loop involved in excluding solvent following substrate binding. Our results build upon previous work where the mobile loop was removed entirely from bacterial luciferase (14). Deletion of the loop significantly decreases the total quantum yield and causes binding of reduced flavin to be slightly weaker. We were able to produce and characterize two alanine point mutants with qualitatively similar kinetic defects. Neither position is universally conserved across luminous bacterial species. However, the relative size of the residue is conserved. We believe that positions 283 and 286 serve a critical function as components of the active center lid that prevents entry of solvent into the active center.