A Native Threonine Coordinates Ordered Water to Tune Light-Oxygen-Voltage (LOV) Domain Photocycle Kinetics and Osmotic Stress Signaling in Trichoderma reesei ENVOY*

Light-oxygen-voltage (LOV) domain-containing proteins function as small light-activated modules capable of imparting blue light control of biological processes. Their small modular nature has made them model proteins for allosteric signal transduction and optogenetic devices. Despite intense research, key aspects of their signal transduction mechanisms and photochemistry remain poorly understood. In particular, ordered water has been identified as a possible key mediator of photocycle kinetics, despite the lack of ordered water in the LOV active site. Herein, we use recent crystal structures of a fungal LOV protein ENVOY to interrogate the role of Thr101 in recruiting water to the flavin active site where it can function as an intrinsic base to accelerate photocycle kinetics. Kinetic and molecular dynamic simulations confirm a role in solvent recruitment to the active site and identify structural changes that correlate with solvent recruitment. In vivo analysis of T101I indicates a direct role of the Thr101 position in mediating adaptation to osmotic stress, thereby verifying biological relevance of ordered water in LOV signaling. The combined studies identify position 101 as a mediator of both allostery and photocycle catalysis that can impact organism physiology.

Light-oxygen-voltage (LOV) domain-containing proteins function as small light-activated modules capable of imparting blue light control of biological processes. Their small modular nature has made them model proteins for allosteric signal transduction and optogenetic devices. Despite intense research, key aspects of their signal transduction mechanisms and photochemistry remain poorly understood. In particular, ordered water has been identified as a possible key mediator of photocycle kinetics, despite the lack of ordered water in the LOV active site. Herein, we use recent crystal structures of a fungal LOV protein ENVOY to interrogate the role of Thr 101 in recruiting water to the flavin active site where it can function as an intrinsic base to accelerate photocycle kinetics. Kinetic and molecular dynamic simulations confirm a role in solvent recruitment to the active site and identify structural changes that correlate with solvent recruitment. In vivo analysis of T101I indicates a direct role of the Thr 101 position in mediating adaptation to osmotic stress, thereby verifying biological relevance of ordered water in LOV signaling. The combined studies identify position 101 as a mediator of both allostery and photocycle catalysis that can impact organism physiology.
Light-oxygen-voltage (LOV) 3 domain-containing photoreceptors are widely distributed in nature, where they couple blue light absorption to regulation of a diverse array of signal transduction pathways (1). In general, LOV proteins can be divided into two subclasses: 1) short LOV proteins (sLOV) that exist as the isolated LOV domain with short ancillary N-or C-terminal caps (2)(3)(4) and 2) modular LOV proteins that couple blue light activation to allosteric regulation of effector domains (1). Although the modular LOV proteins are present in plants, bacteria, and all fungi, the sLOV variety are predominantly found in bacteria and only some fungi (4). Currently, LOV proteins have received widespread attention because of their important roles in regulation of circadian function (3,5,6), growth and development (7,8), stress responses (9 -11), and adaptation to and regulation of pathogenicity (12). In addition, their wide ranging utility has led to the development of optogenetic tools that harness their modular design (13,14). Despite substantial research, key questions involving LOV photocycles remain.
LOV domain chemistry is characterized by blue light-induced formation of a covalent adduct between a bound flavin cofactor (FMN, FAD, or riboflavin) and a conserved Cys residue in a GXNCRFLQ motif. Concomitant with adduct formation is protonation of the N5 position of the isoalloxazine ring. Current models indicate that signal transduction is coupled to N5 protonation via allosteric regulation of N-or C-terminal effector elements remote from the flavin active site (3,15,16). The covalent adduct is defined by a broad UV-visible absorption band centered at ϳ390 nm (LOV 390 ). Upon return to the dark, the adduct state spontaneously decays to an oxidized flavin (LOV 450 ) on a timescale of seconds to days (17,18). Currently the biological role of the wide range in photocycle lifetimes is unknown; however, several studies have suggested that the range facilitates adaptation to changing levels of light intensity (17,19,20). For these reasons, chemical tuning of the LOV photocycle lifetime through understanding of the adduct decay mechanism has been attempted in several systems (2,17,18,(21)(22)(23)(24).
Several lines of reasoning have led to a general mechanism of adduct decay. First, solvent isotope effect experiments indicate that a single proton abstraction event is rate-limiting (17,18). Second, adduct decay can be catalyzed by the presence of small molecule bases such as imidazole (25). Third, residue substitutions at regions that regulate solvent access to the flavin active site have a substantial effect on LOV photocycle lifetimes (18,26). Combined, these experiments implicate N5 deprotonation as the rate-determining step in adduct decay. Consistent with such a model, mutation of residues that regulate accessibility of small molecules to the N5 position or that tune hydrogen bonding characteristics affect kinetics of LOV proteins (17,18,21,22,24,26,27). Importantly, the natural base responsible for N5 deprotonation remains to be determined; however, several possibilities have been suggested that involve elements in the immediate vicinity of the active site flavin (18,26).
In LOV proteins most residues near the N5 position are hydrophobic, with the exception of a conserved Gln residue important for signal transduction (28). That has led to two proposed models regulating N5 deprotonation: 1) Gln facilitated transfer of a proton from the N5 position to the active site cysteine (18). Consistent with the conserved Gln acting as a proton transfer agent, Gln 3 Leu and Gln 3 Asn mutations have a large effect on adduct decay kinetics (29,30). 2) The involvement of ordered water molecules within the flavin binding pocket (26). Crystal structures partially support the latter mechanism, where LOV proteins conserve an ordered solvent channel leading to the active site Cys residue (3,18). However, the water molecules do not penetrate deep enough to directly affect the N5 proton, leading to ambiguity as to their role in adduct decay. Recent FTIR studies indicate that these ordered water molecules play a key role in regulating LOV lifetime (26); however, whether the effect is direct or indirect remains to be determined. Further, analysis of a bacterial sLOV protein (McLOVn) indicates that introduction of a Thr residue in the vicinity of the N5 position can abrogate base catalysis in some systems (31).
Herein, we present a multidisciplinary study of the sLOV protein ENVOY (ENV1) from the filamentous fungus Trichoderma reesei. Recent crystal structures indicate that ENV1 may function as a model system for the exploration of the effect of ordered water on LOV kinetics (11). Specifically, similar to McLOVn, a hydrophilic residue (Thr 101 ) in the vicinity of the N5 position provides a hydrogen bond donor/acceptor that may enable solvent recruitment near N5 (31). Kinetic studies reveal that T101I variants demonstrate a 250-fold slower photocycle (100-fold slower in McLOVn T27I). Further, Thr 101 directly regulates base catalysis and solvent access to the N5 position, but in a manner opposite to that observed in McLOVn. In ENV1, kinetic studies and computational analysis of ordered water indicate that Thr 101 affects both solvent access to the active site and ordering of water molecules adjacent to the N5 position. In vivo analysis verifies that Thr 101 plays a fundamental role in ENV1-mediated adaptation to osmotic stress. In these regards, ENV1 validates a role of ordered water in functioning as an intrinsic base to tune LOV chemistry and function.

Results
Recent studies of a Neurospora crassa Vivid (VVD) homologue from T. reesei (ENV1) revealed divergent signaling mechanisms within closely related fungal LOV domains (11,32,33). Initial photochemical characterization confirmed LOV type chemistry, with a time constant for adduct scission of ϳ1,500 s (11). Such kinetics are 15-fold faster than its homologue, VVD, despite high sequence conservation (11,18). To examine the origin of the altered rate of adduct decay, we conducted a comprehensive study of the ENV1 photocycle and corresponding kinetics.
Full-length (FL; positions  and an N-terminally truncated construct ENV-64 were cloned, expressed, and purified to homogeneity. Using both constructs allows confirmation that photocycle properties are conserved following truncation and allows direct computational studies of the photocycle using the ENV-64 crystal structure (Protein Data Bank code 4WUJ). As reported previously (11), both constructs purify with a bound FMN cofactor and demonstrate spectra consistent with blue light-induced formation of a covalent cysteinyl-flavin C4a adduct (Fig. 1A). Interestingly, depending on purification conditions, the FMN cofactor purifies either as a mixture of oxidized FMN and reduced neutral semiquinone (in the presence of reduced glutathione) or as oxidized FMN (no glutathione) (Fig. 1, A and B). Such behavior suggests that the oxidation potential of the FMN cofactor may reside within physiologically relevant ranges as has been observed previously in LOV protein variants (18,34) and has recently been shown of being competent for signal transduction (35). Thus, ENV1 may employ the semiquinone in dark state signaling. Notably, after cycling through one photocycle, we have so far been unable to chemically re-reduce the FMN cofactor, suggesting that oxidation of an intrinsic element may abolish subsequent chemical reduction.
Comparisons of the rate of adduct decay in the presence (reduced) and absence (oxidized) of glutathione demonstrates similar rates of adduct decay under all conditions and constructs tested. However, the rate of oxidation of the semiquinone displays distinctly different kinetic properties. Kinetic traces obtained at 450 and 478 nm from light-activated oxidized ENV-64 and FL ENV1 reveal time constants of 1,500 and 1,600 s, respectively ( Fig. 1C and Table 1). In contrast, kinetic traces of reduced samples are biexponential. The minor component (30%) demonstrates a time constant comparable with oxidized samples (1,100 s), whereas the major component (70%) has a time constant of 180 s (Fig. 1D). Analysis of the kinetic trace at 622 nm, which isolates the spectral signature of the neutral semiquinone, reveals monoexponential decay with a time constant of 180 s (Fig. 1D). We thus assign the major component as resulting from FMN oxidation and the minor component resulting from adduct decay. We conclude that the rate of oxidation proceeds ϳ10-fold faster than adduct decay pathways in ENV1. Similar mixtures of oxidized and reduced semiquinones were observed in slow cycling variants of N. crassa VVD, where the kinetics of adduct decay and oxidation seemed to occur on similar time scales (18,34).
Sensitivity of WT ENV1 to reducing conditions and a 15-fold faster rate of adduct decay demonstrate key photocycle differences compared with its homologue VVD. To better understand factors affecting the ENV1 photocycle, we probed the reaction landscape using Eyring and Arrhenius analyses. Kinetic studies revealed energies of activation comparable with other circadian clock photoreceptors but divergent compared with typical LOV proteins. Specifically, both FL and ENV-64 demonstrated lower than expected energies of activation of 67 and 70 kJ/mol, respectively (Fig. 2). Arrhenius parameters for most LOV proteins, of similar photocycle lifetimes, have reported activation barriers on the order of 80 -100 kJ/mol (17,22). Notably, studies of other circadian clock photoreceptors indicate atypically low energy barriers that Eyring analysis reveals are compensated by large unfavorable entropies of activation (17). Similarly, Eyring analysis of ENV1 indicates that both FL and ENV-64 exhibited low enthalpies (⌬H ‡ ) of activation (68 and 72 kJ/mol, respectively) that are compensated by large unfavorable entropies (⌬S ‡ ) (Ϫ75 and Ϫ63 J mol Ϫ1 K Ϫ1 , respectively) (Fig. 2B). Studies of other LOV proteins with entropy-compensated photocycles revealed alterations in Hbonding to the active site flavin (17,22). These alterations in H-bonding often affect or are affected by solvent accessibility to the active site (17,22,24). Examination of solvent isotope effects in FL ENV1 is consistent with proton abstraction being rate-limiting (solvent isotope effect ϭ 4.1); thus solvent likely plays a key role in regulating ENV1 kinetics. To better examine the factors altering the ENV1 reaction landscape, we examined solvent accessibility in ENV1.
Base catalysis studies were performed by the addition of varying concentrations of the small base imidazole. Imidazole has been reported as an efficient enhancer of dark state recovery in LOV proteins (25). Although no direct conclusions can be drawn concerning the recovery mechanism, the sensitivity of LOV kinetics to changes in imidazole concentration can indicate alteration of solvent site accessibility (10,17). Base catalysis of FL ENV1 and ENV-64 revealed a modest effect of imidazole on ENV1 kinetics (Fig. 3). Specifically, both FL and ENV-64 exhibit accessibility factors of ϳ1,500, which is comparable with fast cycling bacterial LOV proteins (17) but moderate compared with the highly sensitive truncated LOVK (10).
Importantly, ENV1 is ϳ15-fold more sensitive to imidazole than its homologue VVD (accessibility factor ϭ 94). Thus, the difference in photocycle kinetics in ENV1 and its homologue VVD may be explained by the difference in solvent accessibility of these proteins.
To examine the origin of the variation in solvent accessibility, we examined fungal sequences for residue substitutions in the vicinity of the flavin cofactor. Alignment of ENV1 homologues from different genera and phylogeny revealed that the threonine at position 101 is conserved in the fungi Trichoderma, Cordyceps, and Beauveria, as well as an isolated member of the bacteria Methylocystis (Fig. 4). Residue substitutions at this site are known to have large effects on LOV kinetics in VVD and other proteins (18,36). Specifically, an I74V variant of VVD accelerates the photocycle 24-fold; however, it has no effect on solvent accessibility (18). To the best of our knowledge, I74T or related substitutions have not been studied in VVD; however, similar variants have been shown to substantially affect photocycle kinetics in a bacterial sLOV protein McLOVn (31). Further, Thr 101 constitutes a casein kinase II phosphorylation site (TLCD) in T. reesei that is not present in other species (Fig. 4). Casein kinases are well known to be involved in light response and circadian rhythms (37) in N. crassa. In Trichoderma spp., the group of CKII kinases shows an expansion compared with N. crassa (38). N. crassa PRD-3 was found to be sensitive to sodium chloride, which suggests relevance for dealing with osmotic stress. Hence, the unique presence of a Thr residue at position 101 in fungal and some bacterial LOV proteins may reflect evolutionary adaptation to allow integration of solventdependent response pathways. Thus, to examine whether the

Ordered Water Tunes LOV Chemistry and Function
Thr at position 101 may affect solvent accessibility in ENV1, we studied a T101I variant in FL ENV1 to extract relevant photocycle parameters. In turn we validated a role of Thr 101 in solvent recruitment using molecular dynamic simulations and verified a direct role in sensing osmotic stress in vivo.
Spectroscopic investigation of T101I indicated several differences compared with WT. First, the alternative pathway involving the reduced neutral semiquinone was abolished and kinetics were characterized as monoexponential. Consistent with our predictions that Thr 101 dictates accelerated photocycle  kinetics via solvent recruitment, introduction of a T101I variant resulted in a 250-fold decrease in the rate of adduct decay. Base catalysis studies of the T101I variant confirms that the large effect on the kinetics of adduct scission correlates with a large effect on solvent accessibility. Specifically, the T101I variant is 100-fold less sensitive to imidazole (accessibility factor ϭ 14). Thus, the presence of a Thr residue in the LOV active site confers additional sensitivity to imidazole. Notably, the 250-fold change in photocycle kinetics and only a 100-fold effect on solvent accessibility indicates that other factors may be involved in tuning LOV photochemistry in these systems (Figs. 2 and 3). Moreover, in contrast to McLOVn, where a Thr abrogates imidazole catalysis, in ENV1 a Thr enhances a base catalyzed mechanism (31). Combined, the data indicates that the site adjacent to N5 is involved in the LOV mechanism but acts in concert with additional sites.
Examination of the ENV1 crystal structure reveals several possible conclusions that may have implications for LOV kinetics and adduct scission mechanism throughout the LOV family. Previous studies indicate a possible role of ordered solvent in adduct decay mechanisms (10,18,26); however, a precise locale of action has remained elusive. Given that I74V variants in VVD have no effect on imidazole access but introduction of a corresponding Thr to the ENV1 active site imparts a large effect on LOV kinetics, we propose that hydrophilic residues at this position may facilitate recruitment of solvent to the active site (Figs. 3D and 5, A-D). Consistent with such a mechanism, a Thr at position 101 allows sufficient space to order a water molecule in H-bonding contact to Thr 101 , the flavin N5 position and Gln 204 , which have all been implicated in the rate-limiting steps of adduct decay, as well as signal transduction (18,21,22,25). To probe such a mechanism further, we conducted computational FIGURE 4. Evolutionary relationships of fungal ENV1 homologues. Sequences were retrieved from the JGI Mycocosm database, and species names are given along with the protein ID as provided in the respective JGI genome database. The bar on the right side represents the amino acid present at the position corresponding to Thr 101 in T. reesei ENV1. Phylogenetic analysis was performed as described previously (11) using the minimum evolution algorithm of MEGA4 with 500 bootstrap cycles. JULY 8, 2016 • VOLUME 291 • NUMBER 28 studies to probe the reaction landscape and corresponding ability to recruit solvent to the active site to increase the rate of adduct decay.

Ordered Water Tunes LOV Chemistry and Function
Based on our kinetic analysis of the ENV1 photocycle, we hypothesized that the addition of a large, hydrophobic side chain in the flavin-binding pocket disrupts recruitment of ordered water. In turn, the reduction in area available for water to reside creates a higher energy barrier for water to cross, reducing the velocity of dark state recovery. Although it has been previously noted that water diffuses into the active site (39), no energetics associated with this process have been measured. To test our hypothesis, we used molecular dynamics simulations to calculate the free energy change for water binding in the flavin HN5 region (protonated N5). Ten starting positions within the first and second coordination spheres about the flavin were randomly selected (Fig. 5B). In addition, we correlate recovery kinetics with the stability of water within the flavinbinding pocket.
Comparison of WT ENV-64 and the T101I variant indicate distinct differences in the energetics of active site water that have implications for both water occupancy and possible signal transduction mechanisms in LOV proteins. In both proteins, HN5 dictates the affinity of water for the flavin active site, where water occupancy is most dense between 200 and 300 pm from the HN5 position (Fig. 5, E-H). Despite similarities in the location and duration of active site water occupancy, WT and T101I show different energies of water binding (Fig. 5, E and F). Specifically, WT-ENV1 demonstrates a lower binding energy compared with T101I (11.5 and 15.5 kJ/mol, respectively). The energy difference of 4.0 kJ/mol corresponds to the lack of a hydrogen bond to water from Ile 101 .
Further examination of molecular dynamics simulations provides insight into the coupling of ordered water in LOV active sites and conformational changes in the adjacent Thr 101 and Gln 204 sites. Currently, no structures of LOV proteins exist with an ordered solvent molecule adjacent to N5. The lack of water likely results from a dynamic competition between a buried conformation of Gln 204 and solvent, where the relative populations vary based on the residue identity at position 101. Instead, ordered water in crystal structures is confined to a solvent channel adjacent to the flavin ribityl moiety (Fig. 3D). MD simulations indicate that water accesses the flavin active site through two primary modes. The predominant access point is via a channel above the active Cys, opposite the ribityl group, between the D␣ and E␣ helix loops. In addition, water occasionally enters from a channel near the ENV1 C terminus. In both cases, solvent access to the active site is coupled to rotation of Thr 101 and movements of Gln 204 that have implications for signal transduction and activation of adduct scission (Fig. 5,  A-C).
Water recruitment by Thr 101 results in rotation of the hydroxyl group to H-bond to the bound water. H-bond formation in turn results in local ordering of Thr 101 . In contrast, a T101I variant is ordered regardless of water occupancy ( Table  2). Further upon binding, the HN5/Thr 101 -cordinated water recruits additional waters to the active site. Thus, Thr 101 stabilizes ordered water to the active site and lowers the energy barrier for additional dynamic water near the flavin active site. Consistent with their dynamic nature, these additional waters are not fixed in one place but rather readily diffuse into and out of the flavin-binding pocket. These dynamic waters in turn are coupled to the residue identity at position 101 and conformational changes in Gln 204 .
Rotamer analysis of Gln 204 indicates that adduct formation and water recruitment to HN5 is coupled to rotation of Gln 204 , from a buried position (H-bonding to N5), to an exposed position (H-bonding to O4) (Fig. 5). Such conformational changes have been observed in other MD simulations and have been implicated as a possible signal transduction mechanism (39). However, crystal structures and MD simulations indicate that signal transduction may rather involve only rotation of Gln 204 to H-bond to the newly protonated HN5 (3,34,40). Interestingly, an examination of HN5-water and HN5-Gln distances reveals a single site of water near HN5 that is stabilized by local ordering of Thr 101 (Fig. 5, G and H). In contrast, the lack of an H-bond in T101I allows release of water from HN5 to a position near O4 and Gln 204 (Fig. 5D). In this manner, the HN5, Gln 204 , Thr 101 locus is implicated in both local structural ordering (near position 101) conformational changes to the C terminus (Gln 204 ) and activation of HN5 deprotonation through water recruitment to the flavin active site and stabilization adjacent to HN5.
Kinetic and computational studies report a role for Thr 101 in recruiting solvent to the active site, which may in turn alter LOV allostery. To examine whether Thr 101 is important for LOV signaling in T. reesei, we constructed a QM6a⌬env1 strain complemented with the mutated T101I allele. The dramatic decrease of photocycle lifetime in ENV1 compared with VVD should alter gene expression of target genes caused by stabilized ENV1 in the night. Previously, ENV1 was shown to influence a considerable number of metabolic genes and shows a severe growth phenotype on diverse carbon sources (33). Hence we were interested whether the strain bearing the mutated allele (T101I) would show an altered reaction to changing light conditions or light-dark cycles. Therefore we tested growth during 2-h cycles of low light (200 lux) to high light (5,000 lux) with a ramping period of 2 h, similar 3-h cycles without ramping, 12-h cycles of darkness to light, constant low light and constant high light conditions. As nutritional conditions we chose malt extract medium or minimal medium with carboxymethylcellulose, glucose, or glycerol as carbon source.
Predicting effects of the T101I variant necessitates considering both the effect of kinetics (long lifetime) and the effect of solvent induced protein dynamics (allostery). Any effects of the long lifetime should be confined to changes in the light-dark cycle, particularly light intensity. We would expect the T101I to stabilize the light state and lead to higher light state populations at low light intensity. Consistent with light-dependent activity of ENV1, we found that a strain lacking env1 shows differences in hyphal extension rates compared with the wild type under the different light conditions (Fig. 6A), especially upon growth on carboxymethylcellulose and glycerol. Surprisingly, despite this indication that ENV1 might influence tolerance of different light intensities and reaction to changes in light intensities, the T101I mutant behaved like wild type under all conditions. Thus, any biological effects of T101I are unlikely to be relevant for adaptation to changes between light and darkness or different light intensities and are rather confined to effects on protein allostery.
Thr 101 is involved in tolerance of osmotic stress in T. reesei. Because previous analyses indicated an involvement of ENV1 in reaction to stress conditions, we tested the relevance of T101I in this process. Based on the role of Thr 101 in regulating solvent access and dictating solvent-dependent conformational changes, we predicted that the T101I variant may affect osmotic stress responses. Given the close proximity of Thr 101 to Cys 96 , which has been implicated in oxidative stress, we cannot rule out effects on general stress responses. In examining the effect of T101I on oxidative stress, we found no significant differences in hyphal extension in darkness, low light (1,500 lux), or high light (8,000 lux) in the strains bearing the T101I allele upon growth on malt extract medium or minimal medium with either glucose or glycerol as carbon source.
In contrast, examination of osmotic stress under analogous conditions revealed a direct role for Thr 101 in osmotic stress responses. Specifically we observed an increased hyphal extension compared with the wild-type strain in high light upon growth on carboxymethylcellulose (Fig. 6) with all three conditions inducing osmotic stress (1 M NaCl, 1 M sorbitol, or 1 M KCl). No significant changes were observed in low light or darkness on any carbon source. These results are consistent with MD simulations and kinetic studies indicating that the T101I variant disrupts solvent recruitment to the active site. Under osmotic stress conditions solvent occupancy would be diminished in Thr 101 , leading to alteration of the Gln 204 conformation. The occlusion of solvent by T101I would mimic the decreased solvent occupancy predisposing Gln 204 to an "osmotic stress-like" conformation and a constitutive amplified response. These should be exacerbated under high light conditions. Indeed, we see hyperactivity in response to osmotic stress in these variants that are specific to high light intensities. Thus, we conclude that T101I is involved in regulation of tolerance to osmotic stress in high light in a carbon source-dependent man- ner. These results demonstrate that Thr 101 acts in a signaling region to alter integration of environmental stress.

Discussion
Because of the utility of LOV proteins in optogenetic tools, several studies have aimed to improve the fidelity of LOV optogenetic devices by either tuning photocycle parameters or signal transduction pathways (14). These efforts have been hampered by several fundamental elements. First, often it is difficult to decouple effects of solvent, steric constraints and electronic effects on photocycle parameters. These problems are exacerbated by an incomplete understanding of the role of water in mediating adduct scission in LOV proteins. Second, the effect of these photocycle-altering variants have only been studied in a structural context in a few systems (18,23,24,41,42), and the effect on signal transduction in vivo has been poorly explored (20). Further, in recent years the role of solvent-protein dynamics has been highlighted as playing key roles in coupling cofactor chemistry to protein structural changes (43)(44)(45). Thus, the incomplete understanding of the role of solvent in LOV chemistry and signal propagation may be a key limiting factor to improving optogenetic devices and understanding the broader role of LOV/Period-Arnt-Singleminded domain proteins in signal transduction pathways.
In the present study we have employed ENV1 as a model protein for examining the role of a residue at the juncture between the LOV core and an N-terminal cap (position 101) on adduct decay kinetics and solvent dynamics. Position 101 has been implicated in tuning LOV domain kinetics in numerous systems, but its role in activating water has not been explored (18,31,36,46). Similarly, its effect on signaling mechanisms is not well known. Kinetic studies and molecular dynamic simulations of ENV1 indicate that position 101 is ideally positioned to tune adduct decay pathways by altering solvent-HN5 interactions. Further, solvent access to HN5 is coupled to local order within the N-terminal cap through Thr 101 , as well as conformational changes within a conserved Gln residue (Gln 204 ) implicated in signal transduction in LOV proteins. Combined, these studies have implications for adduct decay pathways and signal transduction in LOV proteins.
Thr 101 leads to stabilization and recruitment of water adjacent to HN5. Solvent recruitment directly correlates with a decrease in stability of the Cys-flavin C4a adduct, leading to acceleration in adduct decay kinetics. Thus, water likely func-

Ordered Water Tunes LOV Chemistry and Function
tions as the intrinsic base deprotonating HN5 in the rate-limiting step of adduct decay in LOV proteins.
In addition, Thr 101 may mediate signal transduction in some LOV proteins. Activation of LOV proteins is typically coupled to signal transduction through two primary mechanisms: 1) alteration of protein stability through order-disorder transitions following adduct formation and 2) conformational changes in N-or C-terminal cap elements, which hinge on flipping the side chain of an active site Gln (Gln 204 ) following adduct formation and N5 protonation (16).
Molecular dynamic studies of ENV1 indicate that the residue identity at position 101 directs light-induced changes in local order. Specifically the presence of the additional methyl functionality in T101I variants leads to ordering of position 101 in both the light and dark. In contrast, Thr 101 is more disordered in the dark but following adduct formation and water recruitment is more ordered ( Table 2). Thus, side chain identity at position 101 may dictate order-disorder transitions in some LOV proteins. Similarly, water recruitment to HN5 is coupled to rotation of Gln 204 , whereby adduct formation leads to Gln 204 rotation from the buried conformation present in all LOV structures (Fig. 5A), to a more exposed position interacting at the flavin O4 position (Fig. 5, C and D). These movements are reminiscent of results of other MD simulations, which indicated that the active site Gln rotates out of the flavin binding pocket following adduct formation (39). Notably, such conformations have not been observed in any crystal structures, so their role in signaling pathways remains unknown. We note that the role of position 101 in possibly both signal transduction and photocycle tuning indicates great care needs to be employed in studying rate altering variants, because many sites may have unexpected consequences in signal transduction (41,42,47).
Indeed recent studies indicate that such residues modulate signal transduction in response to other environmental sensing pathways such as oxidative stress (11). These studies identify that LOV proteins are implicated in adaptation to general stress responses and osmotic stress in diverse organisms (bacteria (48,49), plants (50), and fungi (11)). Currently a direct sensing mechanism by LOV proteins in these responses has remained elusive, and most focus on two-component signal transduction pathways, where LOV proteins act to attenuate kinase function in a light-dependent manner. Attenuating these light responses based on the environmental stress (e.g. salt/oxygen) is likely, but mechanisms are currently unknown. Further, how sLOV proteins that lack the signal transduction element (histidine kinase) adapt to environmental stress is poorly understood.
Here we show how residues in an N-terminal hinge region directly alter local order of a key signaling residue in response to local water. These effects directly correlate with in vivo responses to osmotic stress. As a result the N-terminal region of LOV proteins likely plays a key role in modulating light responses to other environmental or cellular cues.
Specifically, in ENV1 two residues in an N-terminal hinge region mediate adaptation to changing levels of light, nutrients, and environmental stress. The current study identifies that Thr 101 directs adaptation to osmotic stress in vivo. Recent research had previously identified a nearby residue Cys 96 in direct detection of oxidative stress (11). In this manner, LOV proteins are able to modify signal transduction based on direct light-independent detection of environmental or cellular cues. Similar mechanisms have likely arisen in different organisms. Indeed alignment of ENV1 homologues from different genera and phylogeny revealed evolutionary selection of Thr 101 in Trichoderma (Fig. 4). Hence, the unique presence of a Thr residue at position 101 in fungal and some bacterial LOV proteins may reflect evolutionary adaptation to allow integration of multiple stress response pathways.

Experimental Procedures
Cloning and Protein Purification-ENV1 constructs were designed based on sequence homology to N. crassa VVD. Initial constructs focused on full-length ENV1  and an N-terminally truncated construct homologous to VVD-36, ENV-64 (65-207). All constructs were cloned from cDNA obtained from T. reesei grown on cellulose using standard PCR approaches. Envoy constructs were cloned into pGST vector using NcoI and XhoI cut sites and were verified by DNA sequencing (Genewiz). Point mutation T101I was introduced into ENV1 (1-207) using the QuikChange protocol (Stratagene). All mutants were also verified by DNA sequencing (Genewiz).
All constructs were expressed in Escherichia coli JM109 cells. The cells were grown at 37°C until reaching an optical density at 600 nm of 0.6. The temperature was then decreased to 18°C for 40 min to assure thermal equilibrium. At 18°C 0.3 mM isopropyl ␤-D-thiogalactopyranoside (obtained from RPI) was added to initiate protein expression. After 22 h the bacterial pellets were harvested and stored in100 mM NaCl, 50 mM Hepes (pH 8), and 10% glycerol. All constructs were purified with glutathione affinity resin (Qiagen). After binding, the columns were treated with 2 mg of TEV protease/ml of resin at 22°C for 2 h to cut the GST tag from the proteins. Cleaved proteins were eluted in buffer containing 100 mM NaCl, 50 mM Hepes (pH 8), 10% glycerol. An additional round of nickel-nitrilotriacetic acid chromatography was conducted to remove His 6 -TEV prior to final purification with a Superdex S200 size exclusion column equilibrated with 100 mM NaCl, 50 mM Hepes (pH 8), and 10% glycerol.
Spectroscopy and Kinetics-UV-visible absorbance spectroscopy for all constructs and mutants of ENV were conducted on an Agilent 8453 spectrophotometer. Photophysical properties of all constructs were verified to show LOV type chemistry exhibiting a single characteristic peak ϳ380 -400 nm for C4a adduct and two broad absorption peaks for the ground state. The ground state spectra consisted of a peak at 450 nm and two vibrational bands at 425 and 475 nm. Light-dark recovery rates were studied with imidazole to expedite the conversion rates. Imidazole acts as a base to catalyze adduct scission by abstraction of N5. Stock solutions containing 1.0 M imidazole, 100 mM NaCl, 50 mM Hepes (pH 8.0), and 10% glycerol were prepared. Light-dark recovery rates were measured at various concentrations of imidazole ranging from 0 -500 mM imidazole. Each data set at a particular concentration of imidazole was repeated in triplicate. All samples were exposed to a broad spectrum white floodlight source (150 W), while being incubated on ice to populate the light state. For Eyring analysis light-dark recovery rates were studied at temperatures between 288 and 306 K. Kinetics of thermal and base catalyzed reversion were obtained from the absorbance at 450 and 478 nm as a function of time. All values for 450 and 478 nm were corrected for deviations in the baseline by subtracting the absorbance at 600 nm. Full spectra were collected at varying times, such that a minimum of 10 data points per half-life was obtained. The data were fit using mono-and biexponential equations as required to extract kinetic parameters. All time constants were reported as 1/ k adduct_scission that are averaged between the values obtained at 450 and 478 nm. We note that biexponential kinetics were only observed in data sets where reduction competed with adduct formation. Experimental methods for dealing with the biexponential nature of those data sets are discussed in the main text. Accessibility constants (a) are defined as rate ϭ a(k adduct_base ), and the values of a were extracted as defined previously (10,17).
Molecular Dynamics simulations-Molecular simulations were run on the University of California, Santa Cruz, computing cluster, using the charmm27 force field (52) as implemented in the GROMACS 4.6.5 program suite (53)(54)(55)(56)(57), with the addition of flavin-cysteinyl adduct parameters (39). The starting coordinates were taken from the crystal structure of ENV-64 wild type, and a mutation at Thr 101 was generated using Chimera (58).
Starting configurations with one water molecule inside the flavin pocket were manually constructed: an equilibration procedure of energy minimization, then 100 ps of simulation in the fixed N,V&T (N indicates number, V indicates volume, and T indicates temperature) ensemble with the protein atoms positions restrained, then 100 ps with fixed N,P&T (P indicates pressure) and protein position restraints, followed by 750 ps of NPT ensemble with no position restraints was used; the last step was a 1-ns production run. 100 individual production runs were stitched together using the gromacs utility trjcat (56). Using the GROMACS radial distribution function tool (56), a plot of the density of water molecules around flavin N5 was created. From the radial distribution function, g(r), we calculated the energy, w(r), to pull a water from infinite distance to a given r from N5, using the equation ϪRT ln g(r) ϭ w(r) (59). The difference between the lowest and highest part of these curves was taken as the energy barrier to bound water exchange with the bulk solvent. The correlation between water-HN5 distance and HN5-Gln 204 distance was examined by measuring the two distances in all frames of the trajectory, using the gromacs utility g_dist and a simple Python script to extract the distances and couple the distances from the same frame together. These bivariate data sets were then binned into hexagonal bins and heat maps constructed, with a number of frames having values within the bin defining the color of said bin, using the R (R Core team, 2015) package hexbin (60). Using a Python script, frames from the trajectories were binned into one of two bins according to water-HN5 distance, distances of less than 360 picometers were considered to be hydrogen bonds. All frames in the bin were then stitched together into a new trajectory, then the gromacs utility g_chi was called, and the order parameter for residue 101 was extracted.
Strains-The parental strain used for the osmotic stress assay was T. reesei QM6a (ATCC13631). The mutant strain QM6a⌬env1, lacking the whole open reading frame of ENV1 (61), was complemented with the mutated allele.
Stress response-Analysis of response to oxidative stress was performed as described previously (11). The osmotic strain assay for the parental and mutant strains was performed on plates with malt extract medium (3% w/v) or with Mandels-Andreotti minimal medium with carboxymethylcellulose sodium salt (Sigma; 1%, w/v) or glucose (Roth; 1%, w/v) as a carbon source supplemented either with NaCl (Roth; 1 M), sorbitol (Sigma; 1 M), or KCl (Sigma; 1 M) to impose osmotic stress. Strains were grown in constant darkness, constant low light (1,500 lux; Philips Master TL5 HO) and in constant high light (8,000 lux) at 28°C for 5 days. Because this prescreening indicated a relevance of Thr 101 for tolerance of osmotic stress in high light, we repeated the experiment for this condition with 14 days of incubation. The control strains used for all combinations of the assay were QM6a and QM6a⌬env1, and three biological replicates were used in every experiment. Three different recombinant strains harboring the T101I mutations were considered in each experiment and used at least in duplicate. Statistical analysis of hyphal extension after 14 days was performed using PSPP version 0.9.0 (one-way analysis of variance, post hoc tests Tukey, Bonferroni, and Scheffe).