The proton and metal binding sites responsible for the pH-dependent green-red bioluminescence color tuning in firefly luciferases

Firefly luciferases produce yellow-green light under physiological and alkaline conditions, however at acidic pH, higher temperatures or in the presence of heavy metals the color changes to red, a property called pH-sensitivity. Despite many decades of studies, the proton and metal binding sites responsible for pH-sensitivity remain enigmatic. Previously we suggested that the salt bridge E311/R337 keeps a closed conformation of the luciferin phenolate binding site. Here we further investigated the effect of this salt bridge and mutations of the neighbor residues H310 and E/N354, on metal and pH-sensitivity of firefly luciferases emitting distinct bioluminescence colors (Cratomorphus distinctus: 548 nm; Macrolampis sp2: 569 nm). The substitutions of H310 and E/N354 modulate metal sensitivity, whereas the carboxylate of E311 may work as the catalytic base essential for green bioluminescence and pH-sensitivity. Modeling studies showed that H310, E311 and E354 side-chains coordinate Zinc, constituting the metal binding site and the pH-sensor. Electrostatic potential and pKa calculations suggest that the external couple H310/E354 is affected by pH, whereas E311/R337 make a stabilized internal pair which retains excited oxyluciferin ejected proton near its phenolate group into a high energy state, promoting yellow-green bioluminescence. Protonation or metal binding weaken these electrostatic gates and their ability to retain the excited oxyluciferin released proton near its phenolate, promoting red light emission.


Firefly luciferases bioluminescence spectra display different pH-sensitivities. When compar-
ing the bioluminescence spectra of four firefly luciferases displaying different colors and pH-sensitivities ( Fig. 2: Amydetes vivianii, 538 nm; Cratomorphus distinctus, 548 nm; Photinus pyralis, 558 nm, and Macrolampis, 564 nm), the bioluminescence spectrum of Amydetes vivianii luciferase was the most blue-shifted with the lowest amount of red light at pH 8.0, and is the less sensitive to pH, whereas that of Macrolampis sp2 luciferase has the broadest one with largest amount of red light even at pH 8, with Photinus pyralis and Cratomorphus disctinctus luciferases displaying intermediate values (Fig. 2). According to the solvent effect on the fluorescence spectra of luciferin and derivatives 16,17 , the results indicate that the active site of Amydetes luciferase is the most hydrophobic for both green and red emitters at pH 8 and 6. Recent quantum yield measurements by Ando et al. 56 showed that the pH affects only the quantum yield of green emission, whereas the quantum yield of red emission is almost unaffected. Thus, we can assume that in other firefly luciferases too, only the amount of green emitter changes at different pHs. The results reinforce our previous proposal that in firefly luciferases the emission spectra are determined mainly by the ratio of green and red emitters 52 , although both green and red emitters may also experience slightly different environments in each luciferase, influencing the overall bioluminescence spectrum.
Firefly luciferases display different metal sensitivities. Considering that the structural targets of sensitivity to pH, metal ions and temperature must be the same, and that firefly luciferases display different degrees of pH-sensitivity, we compared the effect of Zn 2+ , as a representative metal íon, on the bioluminescence spectra of the above firefly luciferases and their mutants. Similarly to the pH sensitivity, quantum yield measurements showed that the binding of these metal ions to P. pyralis luciferase induce decrease of the quantum yields of the green emission component, resulting in the remaining red emission spectra 57 . C. distinctus luciferase, similarly to Photinus pyralis luciferase, showed a larger red shift than Macrolampis sp2 luciferase in the presence of Zn 2+ (Fig. 3). The luciferase of Amydetes firefly was much less to this metal (results not shown). Therefore, among these luciferases, that of C. distinctus was the most sensitive to these metal ions.
Effect of E354 and H310 mutations on bioluminescence spectra and pH-sensitivity. Considering H310 and E354 are located close to E311 and R337, and that natural substitutions and mutations of E354 were shown to affect bioluminescence spectra and pH-sensitivity, we investigated the effect of mutations by residues which lack or replace the acid-base side chains of E354 and H310 in firefly luciferases pH-sensitivity. All mutants of H310 and E354 (H310A, H310C, N354C, N354H, N354E), independently of being substituted by acid-base or non-acid-base residues, still displayed pH-sensitivity ( Fig. 3; Table 1). The wild-type C. distintus luciferase, was more sensitive to pH than Macrolampis luciferase, which naturally displays the substitution E354N. The most pH-sensitive mutants of Macrolampis firefly luciferase were N354C > N354E > H310C/N354C > N354H > WT.
Whereas the mutant H310A has little effect on the bioluminescence spectrum of Macrolampis sp2 luciferase at pH 8.0 33 , the mutation H310A in Cratomorphus luciferase, resulted in a considerable broadening and red-shifting of the spectrum (results not shown). The absence of effect of H310A in Macrolampis sp2 luciferase, which naturally has the substitution E354N, and the broadening effect of H310A in Cratomorphus luciferase, which has E354, supports the existence of a critical stabilizing interaction between H310 and E354 in Cratomorphus luciferase (and perhaps P. pyralis luciferase) which may stabilize a closed green-emitting conformation, differently from Macrolampis sp2 firefly luciferase which lacks the interaction and produces a broader and red-shifted emission. Furthermore, the mutation of H310R in Macrolampis sp2 luciferase, which inserts a permanent positive charge in this region, was previously shown to result in a very broad and red-shifted spectrum 20 , indicating that a positive charge at this position change the distribution of emission toward the red.
Altogether, these results indicate that although mutations the positions 310 and 354 influence spectral distribution, they are not determinants of pH-sensitivity. showing the residues H310, E311, R337 and E/N354 and interacting residues: (gray shadow) conserved residues of the pH-sensor in firefly luciferases; (yellow shadow) variable residues and (red shadow) unique substitutions affecting pH and metal sensitivity; (Lower panel) three-dimensional homology model of Cratomorphus distinctus firefly luciferase (83% identity with P. pyralis luciferase) showing E311/R337 and H310/E354 salt bridges and the network of polar interactions with the loop between residues 223-235 according to Viviani et al. 52  nescence spectrum and pH-sensitivity of firefly luciferases, we decided to investigate whether this position is also involved with metal sensitivity. The mutant Mac-N354E in Macrolampis firefly luciferase displayed a larger red-shift in presence of Zn 2+ , similar to that observed for the wild-type C. distinctus luciferase that naturally displays the substitution N354E ( Fig. 3; Table 1). The reverse mutant, E354N in C. distinctus luciferase, had opposite effect, undergoing a smaller red shift, similar to that observed for the wild-type Macrolampis sp2 that displays N354 at this position. These results indicate that the position 354 is indeed important for metal sensitivity. Since in Cratomorphus distinctus and Photinus pyralis luciferases E354 is at hydrogen bonding distance to H310, potentially making a salt bridge, and histidines have strong affinity to divalent metals, we also investigated whether the mutation of H310 affects metal sensitivity. The mutant H310A 8 , in which the metal chelating imidazole side-chain is removed, indeed was much less sensitive to metal ions, indicating that H310 is also important to bind metal ions such as Zn 2+ . These results prompted us to investigate the effect of other substitutions of H310 and N354 by residues with metal chelating side-chains such as His, Cys or Glu on the metal-sensitivity of Macrolampis firefly luciferase, resulting in a recently published paper in which we produced mutants with different sensitivities to Zn 2+ , Hg 2+ Cd 2+ and Ni 2+ which could be used to ratiometrically estimate these metal concentrations 8 .
Altogether, the above results clearly indicate that H310 and especially E354 are important sites for metal-sensitivity. However, the fact that the mutation of these residues by residues with non-chelating side-chains does not completely abolish metal-sensitivity, indicate that these residues are not the only targets for metal binding, and that there must be at least another metal binding group in the neighborhood.
Structure of the metal binding site. Thus, in order to see whether the divalent heavy metal ions can bind to H310, E354 and surrounding region, modeling studies of beetle luciferases docked with Zn 2+ were done (Fig. 4). Indeed, Zn +2 complexed with H310 N and carbonyl groups, with the side-chains of N354 in Macrolampis firefly luciferase or E354 in Cratomorphus firefly luciferase, and furthermore with E311 carboxylate, supporting the experimental results, showing that H310, E311 and E/N354 constitute the metal binding site.
Role of the salt bridge between E311 and R337. Previously, we showed that the salt bridge between E311 and R337 is essential to stabilize the closed hydrophobic conformation in firefly luciferases, and that mutations which removed the charge or decreased the size of the side-chains, resulted in red light emission in firefly  54 . In order to investigate whether the salt bridge between E311 and R337 has just a conformational effect on bioluminescence color, or whether these residues display additional catalytic effects on bioluminescence color, we inverted the identity of these residues by constructing the double mutant E311R/R337E. Such double mutant is expected to keep the salt bridge, and therefore a closed conformation. If the color of bioluminescence was related exclusively to the maintenance of the closed conformation, we would expect that the double mutant would also produce yellow-green light at pH 8.0, such as the wild-type enzyme. However, the double mutant produced weak red bioluminescence (601 nm; Table 1). These results clearly indicate that green emission does not depend exclusively on a closed conformation kept by the salt bridge, but rather on additional specific interactions that such residues may play, especially E311. cate that E311 display a key role in bioluminescence color. The mutants E311A and E311Q, in which the negative charge was removed, besides being the most red-shifted ones, are totally pH-insensitive and completely lost the bioluminescence activity at pH 6.0. The mutant E311R completely lost the activity, indicating that charge reversal is incompatible at this position. Only the mutants E311D and R337K, in which the side chains were shortened, but the charge preserved, still displayed a considerable degree of pH-sensitivity: at pH 6.0 E311D spectrum shifts about 17 nm to the red and R337K about 10 nm (Fig. 5). These results indicate that E311 negative charge is essential for green-yellow bioluminescence and also for pH-sensitivity in firefly luciferases. Furthermore, in contrast to R337 mutants, which independently of charge removal or reversal still produce weak red bioluminescence  Table 1. Summary of the effect of mutations of residues H310, E311, R337 and N/E354 (P. pyralis luciferase sequence numbers) in beetle luciferases bioluminescence spectra. *The relative activities were calculated for each mutant in relation to its wild-type luciferase (100%). **Estimated error in bioluminescence spectrum peak was ±2.5 nm. activity, the charge reversal of E311 (E311R), or removal (E311Q and E311A) under acidic pH, completely abolished the bioluminescent activity. Altogether, these results provide compelling evidences that E311 is the primary base responsible for green bioluminescence, and for proton and metal binding responsible for pH-and metal-sensitivityies, playing also a catalytic role in bioluminescence.
Bioluminescence spectra of mutants with 6′Amino-luciferin. Previously we showed that the 6′-amino-luciferin, which has the 6′ hydroxil group substituted by an amino group, is a pH-insensitive luciferin analog which is useful to probe the oxyluciferin phenolate binding site microenvironment polarity 21 . Therefore, in order to check whether the mutations of H310 and N354 cause substantial polarity changes, we analyzed the bioluminescence spectra of this analog with mutants of these residues. However, the bioluminescence spectra peaks for all the mutants were very close to each other ( Table 1), indicating that the microenvironment polarity is quite similar and is not considerably affected by these mutations.
pKa analysis of the side-chains of pH-sensor. In order to better understand the involvement of the above residues and their mutants in pH-sensitivity, we calculated the pKa of the identified residues in the wild-type luciferases and mutants, and correlated them with the electrostatic potential and bioluminescence spectra at different pHs and in the presence of metals ( Fig. 6; Table 2 Supplemental Materials). Among the wild-type residues, H310 was the one which displayed the largest shift of pKa value in relation to the reported value in water (~7.0), displaying a much more basic value (7.85 for Macrolampis luciferase and 8.95 for Cratomorphus luciferase). The pKa values for E311 and R337 were 4.53 and 13, respectively, close to the expected values in water.
As expected, the mutation of these residues affected the pKa of nearby side-chains. Only R337 pKa (13.0) was insensitive to mutations of nearby residues. Among the side-chains that had their pKa values affected by nearby mutations, again the imidazolium side-chain of H310 was the most sensitive site, changing several pH units upon mutations of E311 and R337, whereas the pKa of E311 carboxylate had its pKa only slightly changed upon the mutations of H310 and N354, but much less upon mutation of R337. The largest effects on the pKa of H310 were observed upon the mutations E311A and E311Q which removed the negative charge and decreased the value more than 3 pH units, below 5.4, whereas the mutation E311D, which preserved the negative charge, but with shorter side chain, also considerably lowered the pKa of H310. Altogether, these results are fully consistent with the need of the negative charge of E311 carboxylate to counterbalance the positive charge of R337 guanidinium group. Upon the removal of the negative charge of the E311 carboxylate by the mutations E311A or E311Q, the H310 imidazolium group becomes much more acidic, apparently to loose the extra positive charge in the neighborhood of the unshielded R337 guanidinium ion. Mutation of R337K, which preserved the positive charge but shortened the side chain, also had considerable lowering effect of H310 pKa. In agreement with the above results, charge reversal upon the mutation R337E resulted in a considerable increase of pKa of H310, which is consistent with the need of an additional neighbor positive charge in H310 to counterbalance the extra negative charge of E337. Altogether the results clearly indicate that all these side-chains are close enough to influence each other, especially E311 and R337 strongly influencing H310. . Furthermore, taking the total potential of four residues and the difference of these potentials at pH 6 and 8, unveil a clear relationship between increasing potential difference and a larger spectral shift, reinforcing the role of these residues on pH sensitivity. To find the main potential components involved in the spectral change, the electrostatic potential for each candidate residue was combined in all possible pairs for the three wild-type pH-sensitive proteins of this study (Amydetes vivianii, Cratomorphus distinctus, Macrolampis sp 2 firefly luciferases) and comparatively with the Phrixotrix hirtus railroadworm luciferase, which is an extreme model of pH-insensitive luciferase which naturally emits red light. At pH 8.0, of eleven evaluated regression models (Spectrum function as a function of energy), one model showed the best fitting: energy as the sum of electrostatic potential of the residue pair 310/354 (R2: 0.99) (Fig. 6_psloB). The green emitting luciferases of Amydetes and Cratomorphus display a stabilized electrostatic pair E311/R337, as indicated by favorable micro-environment (lowest potential value, about −900 kJ/mol). Here, the pair H310/E354 acts as strong acid-base pair, ultimately receiving the proton of phenolic hydroxyl group of oxyluciferin. In the more red-shifted Macrolampis luciferase, there is just one strong base (E312; equivalent to position 311 in Photinus pyralis luciferase) which makes an electrostatic pair with R338 (equivalent position to 337). Absence of the second base at position 355 (equivalent to position 354), increased the microenvironment potential, decreasing the strength to retain the proton ejected from excited oxyluciferin phenolate in the neighborhood. The red-emitting luciferase of Phrixotrix does not display the electrostatic pair (308/334 corresponding to 311/337) due the natural substitution of arginine by leucine (L334 corresponding to R337), and the entrance of this cavity (T307 and N351 corresponding to the couple H310/E354) do not have bases to assist residue E308 (equivalent to position 311) to trap oxyluciferin phenolate released proton. Not surprisingly, this luciferase displays the worst microenvironment potential (~−500 kJ/mol), with insensitivity to pH changes. Furthermore, although the mutation L334R slightly blue-shifts the spectrum of Phrixotrix luciferase, its microenvironment potential remained almost unaffected (−503 kJ/mol). Taken together, these results indicate that measurements of the microenvironment potential of pair 310/354 can provide a good estimate about pH sensitivity.

Discussion
The structural identity of the pH-and metal-sensing moiety, as well as the underlying mechanisms of pH-sensitivity and bioluminescence colors in firefly luciferases, have remained elusive for decades. Three mechanisms to explain pH-sensitivity have been proposed: (1) presence of basic residues near oxyluciferin C 5 9 , promoting tautomerization, or phenolic hydroxyl group 12 which could be protonated at lower pH; (2) conformational changes mediated by protonation of specific basic residues important to keep the active-site conformation, leading to entry of water molecules increasing the polarity of the oxyluciferin phenolate binding site 52 ; (3) direct electrostatic interactions with oxyluciferin phenolate 53,57 . These hypotheses are not mutually exclusive.
The presence of bases, which has been originally suggested for the tautomerization hypothesis, are also required to accept the released proton from excited oxyluciferin phenol group. According to the hypothesis by Hirano et al. 12 , a base close to the phenolic hydroxyl group of oxyluciferin may serve to receive the proton, and the proximity of the conjugate acid to the excited oxyluciferin phenolate, may increase the degree of covalent bond character of the O···H bond between the phenolate group and the hydrogen of the conjugate acid. The compromise between the strength of the base and the polarity of the environment, may affect the degree of covalent character of the O···H bond and therefore modulating bioluminescence spectrum. Protonation of such bases at low pH may therefore affect bioluminescence spectra. However, any protic group, including water and main-chain amide groups, may serve as weak temporary bases to accept the proton ejected from the strongly acidic phenolic hydroxyl group of excited oxyluciferin (pKa ~ 1.0). The main-chain amide group of S/C/T314 (Photinus pyralis sequence number) indeed have already been shown to be important as putative assisting base near luciferin phenol group 44 .
pH-sensitivity was also explained on the basis of protonation of basic residues involved in keeping a closed active site conformation, promoting a conformational change to a more open and polar conformation, in the benzothiazolyl part of oxyluciferin binding site 53 . Nakatsu et al. 39 , showed that L. cruciata luciferase displays a close active site conformation with the critical luciferyl-adenylate intermediate analog, DLSA, and an open one with the products AMP and oxyluciferin. Later studies with 6′-aminoluciferin analogs supported the existence of two binding modes for oxyluciferin in firefly luciferases active site, an apolar (N) responsible for green light emission and another polar (P) responsible for red emission 21,22 . The active site in firefly luciferases may undergo conformational changes between these two extremes, and the light emitting step may take place somewhere between these two conformations. Whether such polar and apolar binding modes correspond to distinct conformations or sites remain to be elucidated.
Finally, electrostatic effects have been also claimed to play important roles in emission spectra modulation 53,57-59 . The presence of positive charges near oxyluciferin phenolate were suggested to blue shift the emission spectrum 53,57 . In this regard, the absence of an arginine at position 334 in the red emitting luciferase of Phrixotrix railroadworm luciferase was shown to be responsible for the very red-shifted spectrum of this enzyme, and inclusion of an arginine upon the mutation L334R (corresponding to R337 in P. pyralis luciferase) blue-shifted the emission 55 . In the pH-sensitive luciferases and their mutants, although the calculation of the overall microenvironment charge, considering the four residues H310, E311, R337 and E354, showed a trend for blue-shifted emission in electrostatically more negative environments and red-shifted emission for more positive environments (Table 2), the mutant R337E, in which there is a substantial increase of the overall negative charge character of the environment, also resulted in red light.
Here we show for the first time that the pH-sensing moiety and metal binding site responsible for pH-sensitivity involve the residues H310, E311, R337 and E(N)354. These four residues, with their acid-base side-chains are involved in two main electrostatic interactions that keep a closed and tight active site conformation favorable for green light emission: E311/R337 and H310/R354. The residues E311 and R337 make a more stable and internal salt bridge closer to oxyluciferin phenolate, whereas H310 and E354 are located in a more external and polar part of this cavity, making a second potential gate which is more susceptible to external pH and temperature changes.  The divalent metal ions such as Zn 2+ , Cu 2+ , Hg 2+ , Pb 2+ , Cd 2+ clearly fit in this amphyphilic cavity, coordinating especially with E311 and E354 carboxylates, which are located close to each other (3.5 Å), and also with H310 imidazolium in most firefly luciferases (Fig. 4). The imidazole and carboxylate side chains of residues at positions 310 and 354 are responsible for modulating the sensitivity of Macrolampis and Cratomorphus firefly luciferases to metal ions such as nickel and zinc 8 . The interaction of the divalent metal ions with these glutamates in most firefly luciferases disrupt the salt bridges between E311/R337 and H310/E354, unshielding the positive charges of H310 and especially of R337, polarizing the environment (Fig. 7). Such excess of positive charges increases the polarity of the oxyluciferin phenolate binding pocket by creating a hydration shell around R337, resulting in red light emission. Noteworthy, the mutation H310R in Macrolampis luciferase which naturally lacks E354, partially simulates the effect of the metal, by inserting an additional positive charge in the neighborhood, and thus broadening and red-shifting the spectra.
Despite their importance for metal sensitivity, neither H310 nor E354 can be considered the only proton and metal binding sites, since H310 and E354 are not strictly conserved in other firefly luciferases which are also pH-sensitive, and their mutations by non-acid-base and residues do not abolish either pH-and metal-sensitivity. The residue H310 is substituted by threonine or valine in some firefly and other beetle luciferases. In Macrolampis firefly luciferase the lack of E354 negative charge upon the natural substitution E354N is responsible for the observed weaker metal sensitivity and for the naturally increased proportion of red light observed in its bioluminescence spectrum in relation to the close C. distinctus and P. pyralis firefly luciferases. Therefore, other groups could be important for metal binding, including the main chain amide bond of C/S/T314 which has been already shown to be located near oxyluciferin phenolate and to be important for bioluminescence color modulation 44 .
The internal salt bridge between E311 and R337 was previously shown to play a critical role in bioluminescence color, by keeping a closed hydrophobic conformation favorable for green light emission 55 . Although we already suggested that these residues could play additional specific functions besides the salt bridge, the effective contributions of the salt bridge in keeping a closed conformation, or specific interactions of E311 and R337 in bioluminescence color determination, remained to be elucidated. Here we showed that the exchange of these residues in the double mutant E311R/R337E, with the aim to keep the salt bridge and a closed conformation, resulted in a red emitting mutant, indicating that green light emission does not depends exclusively on a closed active site conformation, but also on more specific effects of E311 and R337.
The results shown in this manuscript bring compelling evidences that E311 carboxylate is the main base responsible for green light emission and pH-sensitivity, and the third metal binding site in firefly luciferases. The acid-base properties of E311 carboxylate, the residual pH-sensitivity upon its substitution by the conservative aspartate (E311D), the complete abolishment of pH-sensitivity upon substitution by other residues that lack negative charge (E311A/Q), provide compelling evidences that E311 carboxylate, most likely through a water molecule, acts as a main base in bioluminescence color determination. Although, the estimated pKa value of E311 carboxylate in this environment (~4.5) is apparently lower than the expected value for a critical base mediating the pH effect between 6.0-8.0, the close proximity to oxyluciferin phenol which acts as a strong acid upon chemiexcitation, may affect the pKa of E311 increasing the sensitivity of this group to external pH-changes. Furthermore, the complete abolishment of bioluminescence activity upon charge reversal in the mutant E311R, and in the negative charge lacking mutants E311A and E311Q at pH 6.0, indicate the unforeseen possibility of a more integral role of E311 carboxylate in bioluminescence catalysis. All beetle luciferases have E311, and in most of them the negative charge of E311 carboxylate is counterbalanced by the positive charge of guanidinium group of residue R337 (only Phrixotrix luciferases do not). According to CTIL (charger transfer induced luminescence) mechanism, during the chemiexcitation step, the phenolic hydroxyl group of the dioxetanone intermediate must be necessarily deprotonated to give phenolate, in order to induce intramolecular charge transfer, promoting its decomposition into carbon dioxide and the excited singlet oxyluciferin which decays with emission of light 8 . The carboxylate of E311, assisted by a water molecule, may play the essential role of catalytic base for proton abstraction of the phenolic hydroxyl group of the dioxetanone intermediate, generating excited oxyluciferin phenolate during the chemiexcitation step. Because excited oxyluciferin phenol is a much stronger acid (pKa < 1.0) than E311 carboxylate (pKa ~ 4.5), it is likely that E311 carboxylate will work as a base retaining the proton. Under such circumstances, E311 carboxylate will be protonated and R337 guanidinium ion will in turn counter stabilize the negative charge of oxyluciferin phenolate. In contrast, in the electropositive microenvironment of the mutants E311R, or E311A/Q at pH 6.0, the phenol proton abstraction necessary for the generation of the critical phenolate ion during the dioxetanone intermediate activation could be halted, severely impacting bioluminescence activity.
At alkaline pH, the electrostatic couple E311/R337, reinforced by the external H310/E354 salt bridge, may work as a gate which keeps a closed conformation, as well as an electrostatic cushion that keeps the oxyluciferin ejected proton near its phenolate into a high energy state (Fig. 7). Acidic pH, the presence of heavy metals or higher temperatures weaken such electrostatic gates, decreasing the force to retain the oxyluciferin excited state released proton near its phenolate group. The few surrounding organized active site water molecules (Fig. 7) may escape as hydronium ions and be exchanged with water, polarizing and increasing the mobility of the microenvironment. Altogether these factors will decrease the energy of excited state resulting in red light emission. Similarly, in the absence of the external salt bridge between H310 and E354 in some beetle luciferase and mutants, the more internal salt bridge between E311 and R337 could be also weakened and the environment relaxed, increasing the proportion of red light. The lack of these salt bridges is observed in Macrolampis firefly luciferase which has E354 substituted by asparagine and produces a broader bioluminescence spectrum, and in a more extreme case, Phrixotrix hirtus railroadworm red emitting luciferase, which also displays asparagine at the corresponding position 351 and, and additionally lacks arginine at position 334 (L334 corresponding to R337 of P. pyralis luciferase), missing both salt bridges corresponding to E311/337 and H310/E354.

Material and Methods
Plasmids and beetle luciferases cDNAs. The cDNAs for Macrolampis sp2 luciferase was originally inserted in pPro vector (Invitrogen) and then was subcloned into NdeI site of pCold vector (Takara, Japan), the cDNA of Cratomorphus distincus luciferase in pBluescript vector 32 , the cDNA of Amydetes vivianii luciferase was cloned in pSport vector (Invitrogen) 35 .
Site-directed mutagenesis. The mutants Mac H310A, H310C, N354E, N354C, N354H, E311A, E311Q, E311D were previously constructed 8 by site-directed mutagenesis using an Agilent site-mutagenesis kit (Catalog 200518). The following primers were designed for generating new mutations: (E311R) CT AAT T TG CAC AGA ATT GCT TCT GG; (R337E) CCA GGT ATA GAA CAA GGA TAT GGG C. To prepare the double mutant Mac-E311R/R337E we performed mutagenesis using the mutant E311R and the primers Mac-R337E. The plasmids containing the luciferase cDNAs were amplified using Pfu turbo polymerase and 2 complementary primers containing the desired mutation, using a thermal cycler (1 cycle 95 °C; 25 cycles 95 °C, 30 s; 55 °C, 1 min and 68 °C 7 min). After amplification, mutated plasmids containing staggered nicks were generated. The products were treated with Dpn I in order to digest non-mutated parental plasmids, and used directly to transform E. coli XL1-Blue cells.
Luciferase expression and purification. For luciferase expression, transformed E. coli BL21-DE3 cells were grown in 500-1000 mL of LB medium at 37 °C up to OD 600 = 0.4, and then induced at 18 °C with 0.4 mM IPTG during 18 h. Cells were harvested by centrifugation at 2,500 g for 15 min and resuspended in extraction buffer consisting of 0.10 M sodium phosphate buffer, 1 mM EDTA, 1 mM DTT and 1% Triton X-100, 10% glycerol and protease inhibitor cocktail (Roche), pH 8.0, lysed by ultrasonication and centrifuged at 15,000 g for 15 min at 4 °C. The N-terminal histidine-tagged Macrolampis sp2 luciferase was further purified by agarose-Nickel affinity chromatography followed by dialysis and anion-exchange chromatography, according to established procedures 54 . The concentrations of purified luciferases were between 0.5 and 1 mg/mL, and the estimated purity, according to SDS-PAGE gels were about 90%. Bioluminescence spectra. Bioluminescence spectra reported here were recorded in ATTO Lumispectra spectroluminometer (Tokyo, Japan) with cooled CCD camera whereas the previously reported spectra used the same equipment or a Hitachi F4500 spectrofluorometer. The bioluminescence spectra reported here using the CCD-based spectroluminometer are red-shifted in relation to previously reported values measured with a spectrofluorometer. Such differences are likely to be caused by distortions associated to the longer scanning time and photosensitivity correction in the older spectrofluorometer in relation to the new equipment. For the in vitro bioluminescence recorded using the spectroluminometer, 5.0 μL of luciferases were mixed with 90 μL of Homology Modelling. Homology-based models of Phrixotrix hirtus, Pyrearinus termitilluminans, Amydetes vivianii and Macrolampis sp2. luciferases were constructed using as template the three-dimensional structure of Luciola cruciata luciferase in the presence of DLSA (PDB code 2D1S) and of oxyluciferin and AMP (PDB code -2D1R) as previously described 55 , Modeller v9.9 was used to align the sequences (using align2d function) and to construct 200 three dimensional models of each sequence Visualization and analyses of the best model of each luciferase were performed using PyMol version 1.4.1 60 . pKa estimation and system neutralization. Initially, side-chain conformations for pH-sensor residues were optimized using SCWALL method of Optimize function in YASARA 61 . To achieve theoretical pKa values for each residues of studied luciferase we used the routine "Experiment Neutralization" of software YASARA 62 . Initially, the hydrogen bonding network was optimized 61 . The method to predict pKa is based on a empirical function that uses hydrogen bonds, Coulomb potential and surface accessible area parameters weighted by experimental data. Electrostatic terms are calculated using Ewald summation 63,64 using force field AMBER03. pKa values for Asp, Glu, His and Lys residues were predicted on the pH 6.0 and 8.0. During model construction, protonation states were assigned using these rule: Protonate Asp and Glu if the predicted pKa is higher than the pH; Protonate His if the predicted pKa is higher than the pH and His is not a hydrogen bond acceptor or deprotonate otherwise. Cys is deprotonated if the selected pH is higher than 8.7. Deprotonate Lys if the predicted pKa is lower than the pH. After that, simulation box was filled with water molecules, and ions are placed at the positions to neutralize highest and lowest electrostatic potential residues. A short MD simulation was ran for the solvent, and water molecules are subsequently deleted until the water density has reached 1.0 g/ml.

Potential Energy calculation.
To calculate the potential energy of the selected residues using parameters from force field Amber03 and cutoff distance of 8.0 A. Energies were calculated for wild-type luciferases of Amydetes, Cratomorphus, Macrolampis and Phrixotrix. Statistical software R 65,66 was used to create a script that processed energy data for each pH-sensor residue. To check the combination of residues that better explained the spectrum shift, the energies of four residues were combined in pairs and trios, creating eleven microenvironment putative potentials. Using linear model that relate emission wavelength of a given luciferase against each microenvironment potential, it was defined the best combination using the explained variance statistic (R2). The results are plotted using ggplot2 library 67 .

Concluding Remarks
The proton and metal binding site responsible for pH-and metal-sensitive bioluminescence color change in firefly luciferases involves mainly the side-chains of H310, E311 and E354 near the oxyluciferin phenolate binding site. These residues are involved in two critical salt bridges (E311/R337 and H310/E354) which keep oxyluciferin phenolate binding site electrostatically closed, promoting green light emission. The carboxylate of E311 is the main base responsible for green light emission and pH-sensitivity, playing also a critical role as a water mediated catalytic for bioluminescence activity base, by assisting proton abstraction of the dioxetanone intermediate phenol group during the chemiexcitation step. Under physiological and alkaline pH conditions, the stable salt bridge between the residues E311 and R337, aided by the more external salt bridge between E354 and H310 (in many firefly luciferases), keep a closed hydrophobic active site environment retaining the excited oxyluciferin released proton near its phenolate group into a high energy state, promoting green light emission. At lower pH or upon metal binding to H310, E311 and E354, the salt bridge between E311 carboxylate and R337 guanidinium is disrupted, reducing the ability of E311 to retain the excited oxyluciferin released proton near its phenolate group, polarizing the environment and thereof resulting in red light emission.