Isolation of bacterial luciferases by affinity chromatography on 2,2-diphenylpropylamine-Sepharose: phosphate-mediated binding to an immobilized substrate analogue

A covalently immobilized form of an inhibitor of bacterial luciferase, 2,2-diphenylpropylamine (D phi PA), was an effective affinity resin for purifying this enzyme from several distinct bacterial species. The inhibitor is competitive with the luciferase aldehyde substrate but enhances binding of the flavin substrate FMNH2 (reduced riboflavin 5'-phosphate); comparable binding interactions occur with luciferase, the immobilized inhibitor D phi PA-Sepharose, and the substrates [Holzman, T. F., & Baldwin, T. O. (1981) Biochemistry 20, 5524-5528]. The effect of FMNH2 on the binding of luciferase to D phi PA-Sepharose was mimicked by inorganic phosphate; the luciferase-phosphate complex had a greater affinity for D phi PA-Sepharose than did luciferase. This observation led to the development of a method using D phi PA-Sepharose to purify bacterial luciferase. When crude enzyme in a high-phosphate buffer was applied to a column of the affinity matrix, the luciferase activity was removed from solution. After the column was washed with the same buffer to remove unbound protein, the luciferase was eluted with a non-phosphate cationic buffer. The affinity column has proven useful for rapid purification of luciferase in much greater yield than has been previously possible with standard ion-exchange techniques. This approach has allowed one-step purification of luciferases from ammonium sulfate precipitates of Vibrio harveyi, Vibrio fischeri, and Photobacterium phosphoreum. The dissociation constants in 0.10 M phosphate for the affinity ligand: luciferase complexes were 0.49 micro M, 0.28 micro M, and 0.15 micro M, respectively, for the three species. The dissociation constant for the V. harveyi mutant AK-6, which has normal aldehyde binding but greatly reduced affinity for FMNH2, was 0.30 micro M, while that for the V. harveyi mutant AK-20, which has greatly reduced affinity for aldehyde but a slightly increased affinity for FMNH2, was 1.2 microM. Preliminary experiments indicated that the yellow fluorescence protein (YFP) that participates, through energy transfer, in bioluminescent emission in V. fischeri strain Y-1 could be separated from the luciferase in this strain by chromatography on the affinity matrix, whereas other methods of separating luciferase and YFP have had limited success because of the binding of YFP to luciferase.     

Effects of mutations of the alpha His45 residue of Vibrio harveyi luciferase on the yield and reactivity of the flavin peroxide intermediate

This work was undertaken to investigate the functional consequences of mutations of the essential alpha His45 residue of Vibrio harveyi luciferase, especially with respect to the yield and reactivity of the flavin 4a-hydroperoxide intermediate II. A total of 14 luciferase variants, each with a different single-residue replacement for the alpha His45, were examined. These variants showed changes, mostly slight, in their light decay rates of the nonturnover luminescence reaction and in their Km values for decanal and reduced riboflavin 5'-phosphate (FMNH2). All alpha His45 mutants, however, showed markedly reduced bioluminescence activities, the magnitude of the reduction ranging from about 300-fold to 6 orders of magnitude. Remarkably, a good correlation was obtained for the wild-type luciferase, 12 alpha His45-mutated luciferases, and six additional variants with mutations of other alpha-subunit histidine residues between the degrees of luminescence activity reduction and the dark decay rates of intermediate II. Such a correlation further indicates that the activation of the O-O bond fission is an important function of the flavin 4a-hydroperoxide intermediate II. Both alpha H45G and alpha H45W were found to bind near-stoichiometric amounts of FMNH2. Moreover, each variant catalyzed the oxidation of bound FMNH2 by two mechanisms, with a minor pathway leading to the formation of a luminescence-active intermediate II and a major dark pathway not involving any detectable flavin 4a-hydroperoxide species. This latter pathway mimics that in the normal catalysis by flavooxidases, and its elicitation in luciferase was demonstrated for the first time by single-residue mutations.     

Polycistronic mRNAs code for polypeptides of the Vibrio harveyi luminescence system.

DNA coding for the alpha and beta subunits of Vibrio harveyi luciferase, the luxA and luxB genes, and the adjoining chromosomal regions on both sides of these genes (total of 18 kilobase pairs) was cloned into Escherichia coli. Using labeled DNA coding for the alpha subunit as a hybridization probe, we identified a set of polycistronic mRNAs (2.6, 4, 7, and 8 kilobases) by Northern blotting; the most prominent of these was the one 4 kilobases long. This set of mRNAs was induced during the development of bioluminescence in V. harveyi. Furthermore, the same set of mRNAs was synthesized in E. coli by a recombinant plasmid that contained a 12-kilobase pair length of V. harveyi DNA and expressed the genes for the luciferase subunits. A cloned DNA segment corresponding to the major 4-kilobase mRNA coded for the alpha and beta subunits of luciferase, as well as a 32,000-dalton protein upstream from these genes that could be specifically modified by acyl-coenzyme A and is a component of the bioluminescence system. V. harveyi mRNA that was hybridized to and released from cloned DNA encompassing the luxA and luxB genes was translated in vitro. Luciferase alpha and beta subunits and the 32,000-dalton polypeptide were detected among the products, along with 42,000- and 55,000-dalton polypeptides, which are encoded downstream from the lux genes and are thought to be involved in luminescence.     

Activity coupling and complex formation between bacterial luciferase and flavin reductases.

Luminous bacteria contain several species of flavin reductases, which catalyze the reduction of FMN using NADH and/or NADPH as a reductant. The reduced FMN (i.e. FMNH(2)) so generated is utilized along with a long-chain aliphatic aldehyde and molecular oxygen by luciferase as substrates for the bioluminescence reaction. In this report, the general properties of luciferases and reductases from luminous bacteria are briefly summarized. Earlier and more recent studies demonstrating the direct transfer of FMNH(2) from reductases to luciferase are surveyed. Using reductases and luciferases from Vibrio harveyi and Vibrio fischeri, two mechanisms were uncovered for the direct transfer of reduced flavin cofactor and reduced flavin product of reductase to luciferase. A complex of an NADPH-specific reductase (FRP(Vh)) and luciferase from V. harveyi has been detected in vitro and in vivo. Both constituent enzymes in such a complex are catalytically active. The reduction of FRP(Vh)-bound FMN cofactor by NADPH is reversible, allowing the cellular contents of NADP(+) and NADPH as a factor for the regulation of the production of FMNH(2) by FRP(Vh) for luciferase bioluminescence. Other regulations of the activity coupling between reductase and luciferase are also discussed.     

Construction of cloning vectors using the Vibrio harveyi luminescence genes luxA and luxB as markers

A synthetic oligodeoxyribonucleotide harboring four new restriction sites was inserted into the luxB gene of Vibrio harveyi. This insertion did not disrupt the reading frame. An active beta-subunit was synthesized since a plasmid with both the luxA and mutated luxB genes conferred upon Escherichia coli the bacterial luciferase (Lux) phenotype in the presence of an aldehyde. Ligation of a piece of foreign DNA at these new cloning sites in the vector extinguish the Lux phenotype of the transformed bacteria. Therefore, the plasmid was used as a cloning vector, and recombinant DNA-containing bacteria were detected by the loss of bioluminescence. To create more versatile plasmids, the intergenic region of phage f1 was inserted outside of the lux genes. The selection by loss of bioluminescence presents several advantages over the white/blue selection of the lacZ gene on indicator plates.     

Expression of the Photorhabdus luminescens lux genes (luxA, B, C, D, and E) in Saccharomyces cerevisiae

The luxA, B, C, D, and E genes from Photorhabdus luminescens were cloned and functionally expressed in Saccharomyces cerevisiae to construct a bacterial lux-based yeast bioreporter capable of autonomous bioluminescence emission. The bioreporter was engineered using a series of pBEVY yeast expression vectors that allowed for bi-directional constitutive or inducible expression of the individual luxA, B, C, and E genes. The luxD gene, encoding the acyl-ACP transferase that ultimately supplies the requisite aldehyde substrate for the bioluminescent reaction, was fused to a yeast internal ribosomal entry site (IRES) sequence to ensure high bi-cistronic expression. Although self-generation of bioluminescence was achieved by the bioreporter, the signal was relatively weak and decayed rapidly. To overcome this instability, a flavin oxidoreductase gene (frp) from Vibrio harveyi was co-expressed to provide sufficient concentrations of the FMNH(2) co-factor required for the bioluminescent reaction. Expression of frp with the lux genes not only stabilized but also enhanced bioluminescence to levels approaching 9.0x10(5) times above background.     

Relationship between the conserved alpha subunit arginine 107 and effects of phosphate on the activity and stability of Vibrio harveyi luciferase.

The Arg107 of the alpha subunit is a conserved residue for all known bacterial luciferases. The phosphate moiety of the reduced flavin mononucleotide (FMNH(2)) side chain has been hypothesized to be anchored at this site (A. J. Fisher, F. M. Raushel, T. O. Baldwin, and I. Rayment Biochemistry 34, 6581-6586, 1995). Mutations of alphaArg107 of the Vibrio harveyi luciferase to alanine, serine, and glutamate were carried out to test such a hypothesis. These variants were characterized and compared with the wild-type luciferase with respect to their K(m) for decanal, FMNH(2), and reduced riboflavin in both low- (0.01 or 0.05 M) and high- (0.3 M) phosphate buffers at pH 7.0. Results are consistent with the hypothesized binding of the FMNH(2) phosphate group by alphaArg107. Moreover, the alphaArg107 residue was apparently important in the expression of the luciferase maximal activity and aldehyde binding. Phosphate ion is also known to have other effects on luciferase stability. We compared the three luciferase variants with the native enzyme with respect to the decay rate of the FMN 4a-hydroperoxide intermediate II, and rates of inactivation by trypsin digestion, modification by N-ethylmaleimide, and heat treatment in low- and high-phosphate buffers. On the basis of patterns of the phosphate effects, alphaArg107 appeared to be important to the enhancement of luciferase stability against trypsin proteolysis at high phosphate but was not involved in regulating the intermediate II decay or sensitivity to N-ethylmaleimide modification. Differential effects of mutations on luciferase thermal stability were observed. It is uncertain whether alphaArg107 is involved in the enhanced thermal stability of the native luciferase in high phosphate buffer.     

Inhibition of Vibrio harveyi bioluminescence by cerulenin: in vivo evidence for covalent modification of the reductase enzyme involved in aldehyde synthesis.

Bacterial bioluminescence is very sensitive to cerulenin, a fungal antibiotic which is known to inhibit fatty acid synthesis. When Vibrio harveyi cells pretreated with cerulenin were incubated with [3H]myristic acid in vivo, acylation of the 57-kilodalton reductase subunit of the luminescence-specific fatty acid reductase complex was specifically inhibited. In contrast, in vitro acylation of both the synthetase and transferase subunits, as well as the activities of luciferase, transferase, and aldehyde dehydrogenase, were not adversely affected by cerulenin. Light emission of wild-type V. harveyi was 20-fold less sensitive to cerulenin at low concentrations (10 micrograms/ml) than that of the dark mutant strain M17, which requires exogenous myristic acid for luminescence because of a defective transferase subunit. The sensitivity of myristic acid-stimulated luminescence in the mutant strain M17 exceeded that of phospholipid synthesis from [14C]acetate, whereas uptake and incorporation of exogenous [14C]myristic acid into phospholipids was increased by cerulenin. The reductase subunit could be labeled by incubating M17 cells with [3H]tetrahydrocerulenin; this labeling was prevented by preincubation with either unlabeled cerulenin or myristic acid. Labeling of the reductase subunit with [3H]tetrahydrocerulenin was also noted in an aldehyde-stimulated mutant (A16) but not in wild-type cells or in another aldehyde-stimulated mutant (M42) in which [3H]myristoyl turnover at the reductase subunit was found to be defective. These results indicate that (i) cerulenin specifically and covalently inhibits the reductase component of aldehyde synthesis, (ii) this enzyme is partially protected from cerulenin inhibition in the wild-type strain in vivo, and (iii) two dark mutants which exhibit similar luminescence phenotypes (mutants A16 and M42) are blocked at different stages of fatty acid reduction.     

Differential synthesis of the polypeptides of aldehyde dehydrogenase and NAD(P)H:flavin oxidoreductase in the bioluminescent bacterium Beneckea harveyi.

The proteins of the bioluminescent bacterium Beneckea harveyi have been labelled with [3H]leucine prior to the induction of bioluminescence, and with [14C]leucine during the development of the bioluminescent system. An aliphatic aldehyde dehydrogenase and a NAD(P)H:flavin oxidoreductase, two enzymes that may be directly involved in the metabolism of the substrates (aldehyde, FMNH2) for the luminescent reaction catalyzed by luciferase, were purified and the isotope ratios of their respective polypeptide chains determined after sodium dodecyl sufate gel electrophoresis. A comparison of these isotope ratios to (a) the isotope ratios of the induced polypeptide chains of luciferase, purified in the same experiment, and (b) the average isotope ratio for the proteins synthesized in concert with growth has provided direct evidence that the synthesis of aldehyde dehydrogenase but not NAD(P)H:flavin oxidoreductase is induced during the development of bioluminescence.     

Vibrio harveyi flavin reductase--luciferase fusion protein mimics a single-component bifunctional monooxygenase

Vibrio harveyi luciferase and flavin reductase FRP are, together, a two-component monooxygenase couple. The reduced flavin mononucleotide (FMNH2) generated by FRP must be supplied, through either free diffusion or direct transfer, to luciferase as a substrate. In contrast, single-component bifunctional monooxygenases each contains a bound flavin cofactor and does not require any flavin addition to facilitate catalysis. In this study, we generated and characterized a novel fusion enzyme, FRP-alphabeta, in which FRP was fused to the luciferase alpha subunit. Both FRP and luciferase within FRP-alphabeta were catalytically active. Kinetic properties characteristic of a direct transfer of FMNH2 cofactor from FRP to luciferase in a FRP:luciferase noncovalent complex were retained by FRP-alphabeta. At submicromolar levels, FRP-alphabeta was significantly more active than an equal molar mixture of FRP and luciferase in coupled bioluminescence without FMN addition. Importantly, FRP-alphabeta gave a higher total quantum output without than with exogenously added FMN. Moreover, effects of increasing concentrations of oxygen on light intensity were investigated using sub-micromolar enzymes, and results indicated that the bioluminescence produced by FRP-alphabeta without added flavin was derived from direct transfer of reduced flavin whereas bioluminescence from a mixture of FRP and luciferase with or without exogenously added flavin relied on free-diffusing reduced flavin. Therefore, the overall catalytic reaction of FRP-alphabeta without any FMN addition closely mimics that of a single-component bifunctional monooxygenase. This fusion enzyme approach could be useful to other two-component monooxygenases in enhancing the enzyme efficiencies under conditions hindering reduced flavin delivery. Other potential utilities of this approach are discussed.     

Effects of 3' end deletions from the Vibrio harveyi luxB gene on luciferase subunit folding and enzyme assembly: generation of temperature-sensitive polypeptide folding mutants

Ten recombinant plasmids have been constructed by deletion of specific regions from the plasmid pTB7 that carries the luxA and luxB genes, encoding the alpha and beta subunits of luciferase from Vibrio harveyi, such that luciferases with normal alpha subunits and variant beta subunits were produced in Escherichia coli cells carrying the recombinant plasmids. The original plasmid, which conferred bioluminescence (upon addition of exogenous aldehyde substrate) on E. coli carrying it, was constructed by insertion of a 4.0-kb HindIII fragment of V. harveyi DNA into the HindIII site of plasmid pBR322 [Baldwin, T.O., Berends, T., Bunch, T. A., Holzman, T. F., Rausch, S. K., Shamansky, L., Treat, M. L., & Ziegler, M. M. (1984) Biochemistry 23, 3663-3667]. Deletion mutants in the 3' region of luxB were divided into three groups: (A) those with deletions in the 3' untranslated region that left the coding sequences intact, (B) those that left the 3' untranslated sequences intact but deleted short stretches of the 3' coding region of the beta subunit, and (C) those for which the 3' deletions extended from the untranslated region into the coding sequences. Analysis of the expression of luciferase from these variant plasmids has demonstrated two points concerning the synthesis of luciferase subunits and the assembly of those subunits into active luciferase in E. coli. First, deletion of DNA sequences 3' to the translational open reading frame of the beta subunit that contain a potential stem and loop structure resulted in dramatic reduction in the level of accumulation of active luciferase in cells carrying the variant plasmids, even though the luxAB coding regions remained intact.     

Characterization of the aldehyde binding site of bacterial luciferase by photoaffinity labeling

A photoaffinity probe 1-diazo-2-oxoundecane has been synthesized and used to examine the aldehyde-binding site of the nonidentical dimeric luciferase (alpha beta) from Vibrio harveyi cells. In the dark, the probe competes against aldehyde in binding to luciferase. Irradiation of luciferase and the probe at 254 nm resulted in primarily specific labeling of both alpha and beta subunits with concomitant enzyme inactivation, but significant (congruent to 40%) nonspecific labeling of mainly the beta subunit also occurred. The addition of decanal to protect the active center reduced the rate of inactivation. When 2-mercaptoethanol was included to quench the nonspecific labeling, the amounts of probe incorporated into alpha and beta correlated stoichiometrically with the quantities of enzyme photoinactivated. On the basis of these findings, we postulate that the aldehyde binding site is at or near the subunit interface of luciferase.     

Using fusions with luxAB from Vibrio harveyi MAV to quantify induction and catabolite repression of the xyl operon in Staphylococcus carnosus TM300

Abstract The luxA,B genes from the Gram-negative marine bacterium Vibrio harveyi MAV were used in Staphylococcus carnosus TM300 as a reporter system for regulated expression of xylose utilization. The luciferase genes were fused to the xyl operon from Staphylococcus xylosus C2a. Expression of bioluminescence was induced through addition of xylose and repressed in the presence of glucose. A method to quantitate bioluminescence directly from the culture is described.     

Nucleotide sequence, expression, and properties of luciferase coded by lux genes from a terrestrial bacterium

The lux genes required for expression of luminescence have been cloned from a terrestrial bacterium, Xenorhabdus luminescens, and the nucleotide sequences of the luxA and luxB genes coding for the alpha and beta subunits of luciferase determined. The lux gene organization was closely related to that of marine bacteria from the Vibrio genus with the luxD gene being located immediately upstream and the luxE downstream of the luciferase genes, luxAB. A high degree of homology (85% identity) was found between the amino acid sequences of the alpha subunits of X. luminescens luciferase and the luciferase from a marine bacterium, Vibrio harveyi, whereas the beta subunits of the two luciferases had only 60% identity in amino acid sequence. The similarity in the sequences of the alpha subunits of the two luciferases was also reflected in the substrate specificities and turnover rates with different fatty aldehydes supporting the proposal that the alpha subunit almost exclusively controls these properties. The luciferase from X. luminescens was shown to have a remarkably high thermal stability being stable at 45 degrees C (t 1/2 greater than 3 h) whereas V. harveyi luciferase was rapidly inactivated at this temperature (t 1/2 = 5 min). These results indicate that the X. luminescens lux system may be the bacterial bioluminescent system of choice for application in coupled luminescent assays and expression of lux genes in eukaryotic systems at higher temperatures.     

Recovery of components of fluorescence spectra of mixtures by intensity- and anisotropy decay-associated analysis: the bacterial luciferase intermediates

Reaction of FMNH2 and O2 with bacterial luciferase followed by blue light irradiation results in a product previously claimed to have the same fluorescence spectral distribution as the bioluminescence. Preparations of this "high fluorescence" intermediate, however, contain two fluorescent components, one from the intermediate and the other its breakdown product, FMN. Since the intermediate has a fluorescence lifetime of around 10 ns and a rotational correlation time in the range of 100 ns, compared to 5.0 and 0.15 ns, respectively, for the FMN, the two components can be successfully resolved from the total fluorescence by an anisotropy decay- and fluorescence decay-associated analysis employing simultaneous global computational methods. The fluorescence spectra of the intermediates from two types of luciferase were analyzed in this way; one luciferase was from Vibrio harveyi and the other was from an unusual type of V. fischeri that had an in vivo bioluminescence maximum at 505 nm, a wavelength almost 20 nm longer than that of the V. harveyi bioluminescence. For V. harveyi the true fluorescence of the intermediate is distinct from the bioluminescence, being found at a wavelength about 10 nm longer. For the type of V. fischeri examined, any difference in the two spectra is less certain. A control experiment with the dye 8-amino-1- naphthalenesulfonate bound to BSA and mixed with FMN recovered the original spectrum of the bound dye accurately.     

Cloning and expression of the Photobacterium phosphoreum luminescence system demonstrates a unique lux gene organization

The organization of the lux structural genes (A-E) in Photobacterium phosphoreum has been determined and a new gene designated as luxF discovered. The P. phosphoreum luminescence system was cloned into Escherichia coli using a pBR322 vector and identified by cross-hybridization with Vibrio fischeri lux DNA. The lux genes were located by specific expression of P. phosphoreum DNA fragments in the T7-phage polymerase/promoter system in E. coli and identification of the labeled polypeptide products. The luxA and luxB gene products (luciferase subunits) were shown to catalyze light emission in the presence of FMNH2, O2, and aldehyde. The luxC, luxD, and luxE gene products (fatty acid reductase subunits) responsible for aldehyde biosynthesis could be specifically acylated with 3H-labeled fatty acids. The order of the lux genes in P. phosphoreum was found to be luxCDABFE with luxF coding for a new polypeptide of 26 kDa. The presence of a new gene in the P. phosphoreum luminescence system between luxB and luxE as compared to the organization of the lux structural gene in V. fischeri and Vibrio harveyi (luxCDABE) demonstrates that the luminescent systems in the marine bacteria have significantly diverged. The discovery of the luxF gene provides the basis for elucidating the role of its gene product in the expression of luminescence in different marine bacteria.     

Site-directed mutagenesis of bacterial luciferase: Analysis of the ‘essential’ thiol

It has been appreciated for many years that the luciferase from the luminous marine bacterium Vibrio harveyi has a highly reactive cysteinyl residue which is protected from alkylation by binding of flavin. Alkylation of the reactive thiol, which resides in a hydrophobic pocket, leads to inactivation of the enzyme. To determine conclusively whether the reactive thiol is required for the catalytic mechanism, we have constructed a mutant by oligonucleotide directed site-specific mutagenesis in which the reactive cysteinyl residue, which resides at position 106 of the α subunit, has been replaced with a seryl residue. The resulting α106Ser luciferase retains full activity in the bioluminescence reaction, although the mutant enzyme has a ca 100-fold increase in the FMNH₂ dissociation constant. The α106Ser luciferase is still inactivated by N-ethylmaleimide, albeit at about 1/10 the rate of the wild-type (α106Cys) enzyme, demonstrating the existence of a second, less reactive, cysteinyl residue that was obscured in the wild-type enzyme by the highly reactive cysteinyl residue at position α106. An α106Ala variant luciferase was also active, but the α106Val mutant enzyme was about 50-fold less active than the wild type. All three variants (Ser, Ala and Val) appeared to have somewhat reduced affinities for the aldehyde substrate, the valine mutant being the most affected. It is interesting to note that the α106 mutant luciferases are much less subject to aldehyde substrate inhibition than is the wild-type V. harveyi luciferase, suggesting that the molecular mechanism of aldehyde substrate inhibition involves the Cys at α106.     

Cloning and nucleotide sequences of lux genes and characterization of luciferase of Xenorhabdus luminescens from a human wound.

Xenorhabdus luminescens HW is the only known luminous bacterium isolated from a human (wound) source. A recombinant plasmid was constructed that contained the X. luminescens HW luxA and luxB genes, encoding the luciferase alpha and beta subunits, respectively, as well as luxC, luxD, and a portion of luxE. The nucleotide sequences of these lux genes, organized in the order luxCDABE, were determined, and overexpression of the cloned luciferase genes was achieved in Escherichia coli host cells. The cloned luciferase was indistinguishable from the wild-type enzyme in its in vitro bioluminescence kinetic properties. Contrary to an earlier report, our findings indicate that neither the specific activity nor the size of the alpha (362 amino acid residues, Mr 41,389) and beta (324 amino acid residues, Mr 37,112) subunits of the X. luminescens HW luciferase was unusual among known luminous bacterial systems. Significant sequence homologies of the alpha and beta subunits of the X. luminescens HW luciferase with those of other luminous bacteria were observed. However, the X. luminescens HW luciferase was unusual in the high stability of the 4a-hydroperoxyflavin intermediate and its sensitivity to aldehyde substrate inhibition.     

Growth, luminescence, respiration, and the ATP pool during autoinduction in Beneckea harveyi.

The bacterial bioluminescence system is unusual because it is self-induced. In the late logarithmic phase of growth, upon the accumulation of an autoinducer, the synthesis of the components of the system is initiated. We were interested in determining what effect this burst of synthesis and activity has on cellular energy metabolism. The ATP pool of the luminous bacterium Beneckea harveyi was found to dip 10- to 20-fold during the luminescence period, while the respiration per unit cell mass (optical density) increased but by much less. The dip in the ATP pool did not occur in four different types of dark mutants, including one that was temperature conditional and another that was conditional upon added cyclic AMP for luminescence. However, it is neither the synthesis nor the activity of luciferase that is responsible for the ATP dip; the dip does not occur in certain dark "aldehyde" mutants which nevertheless synthesize normal levels of luciferase, whereas it does occur at 36 degrees C in a temperature-sensitive luciferase mutant which forms normal levels of inactive luciferase. Results with other aldehyde mutants implicate the pathway involved in the synthesis of the aldehyde factor with the ATP dip.