In-cell infrared difference spectroscopy of LOV photoreceptors reveals structural responses to light altered in living cells

Proteins are usually studied in well-defined buffer conditions, which differ substantially from those within a host cell. In some cases, the intracellular environment has an impact on themechanism, whichmight bemissed by in vitro experiments. IR difference spectroscopy previously has been applied to study the light-induced response of photoreceptors and photoenzymes in vitro. Here, we established the in-cell IR difference (ICIRD) spectroscopy in the transmission and attenuated total reflection configuration to investigate the light-induced response of soluble proteins in living bacterial cells. ICIRD spectroscopy on the light, oxygen, or voltage (LOV) domains of the blue light receptors aureochrome and phototropin revealed a suppression of the response of specific secondary structure elements, indicating that the intracellular environment affects LOV photoreceptor mechanisms in general. Moreover, in-cell fluorescence spectroscopy disclosed that the intracellular environment slows down the recovery of the light-induced flavin adduct. Segmentresolved ICIRD spectroscopy on basic-region leucine zipper (bZIP)-LOV of aureochrome 1a from the diatom Phaeodactylum tricornutum indicated a signal progression from the LOV sensor to the bZIP effector independent of unfolding of the connecting A9a-helix, an observation that stood in contrast to in vitro results. This deviation was recapitulated in vitro by emulating the intracellular environment through the addition of the crowding agent BSA, but not by sucrose polymers.We conclude that ICIRD spectroscopy is a noninvasive, label-free approach for assessing conformational changes in receptors in living cells at ambient conditions. As demonstrated, these near-native responses may deviate from the mechanisms established under in vitro conditions.

Photoreceptors are of central importance for many organisms, from bacteria, fungi, and plants to animals, for the determination of intensity and frequency of light in their environment and for the adequate response to beneficial or harmful conditions. Moreover, attention has been attracted in recent years by the successful engineering of photoreceptors toward an application as optogenetic tools for the dissection of enzymatic and neuronal pathways in living organisms. For both the natural function and the application, it is of importance to elucidate the mechanism of these proteins. Accordingly, the light-induced response of photoreceptors has been studied extensively with a plethora of biophysical techniques in vitro (1). However, the proteins are usually investigated in a defined and simplified buffer solution, strongly differing from the intracellular conditions. The presence of small molecules, protein-protein interactions, the effect of macromolecular crowding, the different redox potential, and the different oxygen level might affect the light-induced response of a photoreceptor in cells. Indeed, it has been shown for plant cryptochrome 2 from Arabidopsis in intact insect cells that cellular conditions promote alternative reaction pathways (2). This effect has been attributed to the binding of nucleotides such as ATP to the plant cryptochrome (3)(4)(5). Accordingly, results from in vitro experiments should be scrutinized under intracellular conditions to provide for biological relevance.
To complement in vitro measurements, much effort has been put into establishing a broad field of techniques to study protein mechanisms in an intracellular environment (6). For soluble photoreceptors, such approaches include UV-visible spectroscopy on plant seedlings (7,8) as well as fluorescence and EPR spectroscopy on insect cells (9,10), which have been adapted to study the light-induced activation in situ. A limitation of these pioneering studies has been the focus on the cofactor and its photoreaction, whereas the structural response of the protein has not been addressed. For example, in vitro experiments on light, oxygen, or voltage (LOV) proteins revealed a signal propagation from the cofactor to the effector via structural changes in flanking helices (11,12). The question arises of whether this mechanism might be altered by the intracellular environment.
LOV proteins are blue light photoreceptors that can be found in various organisms, such as plants, algae, fungi, or bacteria (13). The first member identified was phototropin in plants, which is composed of two LOV sensory domains, LOV1 and LOV2, activating a C-terminal serine-threonine kinase upon illumination (Fig. 1A) (14). The LOV domains contain a noncovalently bound FMN, which reacts upon illumination with a nearby located cysteine to form an adduct (15). Aureochromes found in algae belong to the LOV protein family as well but have an unusual, inverted topology (Fig. 1B). The LOV domain of aureochromes is connected to a basic-region leucine zipper (bZIP) at the N terminus instead of at the C terminus. Aureochromes act as a blue light-dependent transcription factor regulating high-light acclimation, photomorphogenesis, and cell division in algae (16)(17)(18). The LOV domain of aureochromes has been applied successfully in optogenetics as a light-dependent dimerizer for receptor tyrosine kinase activation (19,20).
In the LOV domain of aureochrome, formation of the adduct leads to unfolding of a flanking Ja-helix (21,22) similar to phototropin-LOV (23). As a consequence of an allosteric regulation, the flanking A9a-helix then unfolds and exposes a hydrophobic b-sheet surface required for dimerization (22,24) or for rearrangement of the dark dimer structure (Fig. 1, C and D) (25). The structural change induces a signal progression from the A9a-helix to the bZIP domain (26), which then partially unfolds in the absence of DNA (27). The presence of random DNA leads to a bZIP stabilization and suppression of the unfolding. In contrast, in the presence of the ACGT core sequence, additional a-helical elements and loop elements are formed by light (27). Illumination of full-length aureochrome leads to an increase in affinity to the target DNA (16,22,28). The adduct in aureochrome-LOV spontaneously reverts to the dark state within minutes (16), depending on the oligomerization state (21). The recovery time of LOV domains can be strongly modulated by point mutations and accelerated by imidazole via a base-catalyzed mechanism (29)(30)(31)(32).
FTIR difference spectroscopy provides an established method to analyze structural changes of proteins in solution upon an external perturbation, such as light, electrical potential, temperature, or solute concentration (33)(34)(35). Common configurations for applying FTIR spectroscopy are the transmission mode, in which the IR light passes the sample sandwiched between two IR-transmissive windows, and the attenuated total reflection (ATR) configuration, in which total reflection within an internal reflection element (IRE) is used to generate an evanescent wave at the interface to the sample. These approaches are noninvasive, label-free, under ambient conditions, without any size limitation for the macromolecules investigated, and the acquisition time is fast. Therefore, it is highly promising to transfer FTIR difference spectroscopy from an in vitro to an in-cell application. Single cells have already been studied by using synchrotron radiation to investigate the effect of arsenite on human cells (36) and the response of rhodopsin to light in single animal rod cells (37). The latter cells represent an ideal sample, because of the high, native concentration of 2-3 mM of the insoluble, transmembrane photoreceptor. The application of ATR difference spectroscopy on rod cells showed changes in protein signals, which were assigned to the intracellular transducin movement toward the evanescent field (38).
To our knowledge, IR difference spectroscopy on soluble photoreceptors in living or intact cells has not yet been reported. These comprise a large variety of receptors, such as LOV proteins, BLUF (blue light using flavin) proteins, cryptochromes, phytochromes, or PYPs (photoactive yellow proteins) (1,39). In this study, we selected the LOV domain of aureochrome1a from the diatom Phaeodactylum tricornutum (PtAureo1a) (Fig. 1B) and the LOV1 domain of phototropin from the green alga Chlamydomonas reinhardtii (CrPhot) (Fig. 1A) for establishing incell investigations because they have already been characterized by FTIR difference spectroscopy in vitro (21,24,27,40). We present here in-cell IR difference (ICIRD) spectroscopy in the transmission mode and the ATR configuration as an application for the investigation of the light-induced structural response of soluble photoreceptors in living cells. We observed with ICIRD spectroscopy a specific suppression of changes in secondary Aureochromes show an unusual, inverted arrangement of the sensor and effector domains. C, the crystal structures of the PtAureo1a-LOV domain in the dark state (PDB entry 5A8B) and the light state (PDB entry 5DKL) are shown in red and blue, respectively. Upon illumination of LOV, the flanking Jahelix unfolds, leading to a repositioning and unfolding of the A9a-helix. D, the overlay of the LOV monomer with the light-state dimer in gray (PDB entry 5DKL) illustrates how this repositioning and unfolding of the A9a-helix provides access to the b-sheet surface for rotation or dimerization of the second LOV domain unit. structure elements caused by the intracellular environment. Investigation on the bZIP-LOV module leads us to postulate an alternative signal progression mechanism from the sensor domain to the effector. These findings emphasize the impact of the intracellular environment on LOV photoreceptors and need to be considered in future studies of physiological function as well as optogenetic application.

In-cell recovery kinetics of LOV after illumination
To characterize our sample for in-cell spectroscopy, we first applied fluorescence spectroscopy to the isolated LOV domain of the blue light receptor PtAureo1a. The emission spectra of free and bound FMN can be readily differentiated by the hypsochromic shift of the fluorescence emission maximum from 538 to 497 nm with a shoulder at 528 nm by binding of the FMN to the LOV domain ( Fig. 2A) (41). The LOV domain was moderately expressed in Escherichia coli BL21 using low-level induction. Indeed, the fluorescence spectrum of E. coli BL21 expressing the LOV domain showed the characteristic emission of FMN bound to LOV with a negligible contribution from the signal of free flavin in the cells ( Fig. 2A). Additionally, we used the fluorescence signal to determine the copy number of the protein in the cells. We correlated the emission intensity at 495 nm of LOV in vitro with the absolute number of proteins in the probed volume by determining its absorbance at 447 nm and using the extinction coefficient of free FMN at 450 nm (42), the path length, and the sample volume (Fig. S1). To determine the cell number, the turbidity was analyzed and correlated with the counted cell number. We finally obtained a protein copy number per cell of ;300,000 (see "Experimental procedures"). The corresponding protein concentration of LOV in the cells is about 330 mM and 120 mM assuming a cell volume of 1.5 fl (starved) or 4 fl (in medium), respectively (43).
Illumination of LOV leads to formation of an adduct of the FMN with a cysteine, which reduces the absorbance at 447 nm as well as the fluorescence intensity at 497 nm (15,41). Therefore, the kinetics of the back reaction can be analyzed by studying the recovery of either absorbance or fluorescence intensity. Considering the strong turbidity of cells, fluorescence spectroscopy was used here to investigate the decay kinetics of the adduct in the cells. A cell suspension of E. coli BL21 expressing LOV was illuminated with blue light for 100 ms, and fluorescence recovery at 495 nm was monitored (Fig. 2B). Fitting with a monoexponential function, we obtained a time constant of 2800 6 160 s in cells and 2030 6 330 s in vitro at pH 8, 300 mM NaCl, and 25 mM LOV. Therefore, the presence of the intracellular environment stabilizes the adduct. It has been shown that the recovery is strongly concentration-dependent because of a monomer-dimer equilibrium. UV-visible studies on LOV recovery at 30 mM in vitro showed time constants of 320 s (14%) and 2290 s (86%), which were assigned to the monomer and dimer of LOV by comparison with variants that are constitutively monomeric (LOVDJa) (21) or dimeric (bZIP-LOV) (27), respectively. Even at 2.4 mM, a concentration much higher than in cells, the recovery was faster, with a time constant of 1660 6 400 s obtained by FTIR spectroscopy (21). The time constant of 2800 s in cells, therefore, cannot be explained by a concentration-dependent dimerization effect. Other contributions by the lower salt concentration or the lower pH value in cells were excluded, because a variation of these conditions in vitro led to an acceleration or a negligible effect, respectively (Figs. S2 and S3). Additionally, we tested whether macromolecular crowding via excluded volume affects the adduct decay. The addition of 100 mg·ml 21 Ficoll70 as a nonionic crowding agent resulted in a time constant of 2040 6 120 s (Fig. 2B), significantly faster than in cells. The alternative application of high concentrations of BSA as crowding agent failed, because at the resulting high viscosity, the signal from adduct decay is distorted by sample diffusion. In conclusion, the high concentration and typical cellular conditions, such as excluded volume, pH value, or salt concentration, do not explain the observation of a slower adduct decay, indicating the presence of other factors in the intracellular environment that affect the recovery kinetics of LOV domains.

In-cell IR difference spectroscopy on LOV
To study the light-induced response of LOV in cells, we then established ICIRD spectroscopy in transmission configuration. E. coli BL21 expressing LOV were sandwiched between two BaF 2 windows, and IR difference spectra were generated by recording the intensities directly before and after illumination with blue light (33). Very small signals were obtained in the difference spectra because of the low concentration of proteins in cells, making it necessary to extensively average 18,432 scans for a good signal/noise ratio (Fig. 3A). We estimate a concentration of 70 mM to generate this signal intensity by comparison with in vitro experiments with a low concentration of 365 mM and a scaling factor of 0.19, assuming equal path lengths and neglecting the extracellular volume. This estimated concentration fits well with the intracellular 120-330 mM calculated from fluorescence experiments. When comparing light-induced protein responses in cells and in vitro, the difference spectra almost perfectly match, which demonstrates the functionality of LOV in cells (Fig. 3A). Several characteristic signals in the photoreaction of LOV were observed, such as at 1728 (1) cm 21 /1713 (2) cm 21 , 1643 (2) cm 21 , and 1630 (1) cm 21 , which are attributed to the conversion of FMN (40,44), unfolding of the Ja-helix (21,45), and b-sheet dimerization (21), respectively, according to previous assignments (Table 1). However, additional signals in the amide I region (1695-1615 cm 21 ) and amide II region (1570-1520 cm 21 ) are present in cells, which are characteristic for changes in secondary structure (34).
The double difference spectrum calculated from in vitro minus in-cell differences shows the most prominent signal at 1657 (2) cm 21 (Fig. 3B). This signal has been assigned in vitro to a light-induced unfolding of the A9a-helix by comparing the difference spectra of LOV and LOVDA9a and calculating a double difference (24) and was confirmed by X-ray crystallography (22). Accordingly, the in-cell difference spectrum in this spectral region closely resembles that of LOVDA9a, for which a response of A9a can be excluded. Here, the observation of this signal in the double difference spectrum is attributed to a suppression of the unfolding in cells by the impact of the intracellular environment. In the difference spectra, the origin of this signal is more difficult to see (Fig. 3A), because the negative signal of the A9a unfolding in vitro is compensated by a strong positive contribution at 1666 cm 21 . In cells, the positive contribution is clearly visible, because of the lacking A9a unfolding. This observation is at odds with the mechanism proposed from in vitro studies postulating that A9a unfolding is required for b-sheet dimerization (24), because the signal attributed to the b-sheet dimerization at 1630 cm 21 is still found in the in-cell difference spectrum. The positive and negative signals in the amide II region of the double difference spectrum (Fig. 3B) provide support for the suggestion of altered protein responses in the secondary structure of LOV in cells compared with in vitro.

Establishing the ATR configuration for ICIRD spectroscopy on LOV
Although LOV was expressed in E. coli, we cannot verify in the transmission experiments whether the cells are viable, whether the protein is indeed located in the cells, and whether the sample is fully hydrated. Therefore, we established ATR-ICIRD spectroscopy. We placed a concentrated cell suspension of E. coli BL21 expressing LOV on an internal reflection element and tightly sealed the small compartment to prevent any evaporation (Fig. 4A). The comparison of difference spectra of LOV in cells and in vitro obtained with the ATR approach shows that the spectra are nearly congruent but with the same deviating, positive contribution at 1657 cm 21 as observed in the transmission mode (Fig. 4B). The ATR approach suffers from low light intensities because of losses by coupling to and absorption in the internal reflection element, leading to a strong increase in noise. Additionally, the slightly reduced effective pathlength of about 6 mm in the ATR approach results in an about 20% lower absorbance (Figs. 3A and 4B). Therefore, the obtained difference spectrum of LOV in cells had to be averaged from 73,728 scans, which is 4 times more scans than required in the transmission mode.
The ATR approach allowed us to recover the sample from the compartment after the IR spectroscopic experiments and to record fluorescence spectra of the cells as well as of the supernatant of the centrifuged cell suspension (Fig. 4C). As a result, the characteristic fluorescence emission of LOV was detected only in the cell suspension, whereas the emission of the supernatant did not show any fluorescence of LOV, providing evidence for the intracellular location of LOV. Furthermore, the cells were plated out after the ATR-ICIRD experiments. We found that 30% of the cells remained reproductive, indicating that our experiments were performed mostly on living cells. It should be noted that the total number of metabolically active cells is higher than 30%, because the stress conditions during the protein expression and the experiment cause cells to switch to the viable but nonculturable state (46,47).
The deviating signal in the ATR difference spectrum of LOV in cells at 1657 cm 21 recorded at full hydration supports the above results regarding the suppression of A9a unfolding and additionally confirms the localization of LOV in cells and the presence of living cells.

Emulating cellular conditions by adding crowding agents
The impaired unfolding of the A9a-helix in the cells points to a stabilization of the helix fold in the light state by the intracellular environment. The total protein concentration in cells can be .300 mg·ml 21 (48), which leads to macromolecular crowding effects (49). Therefore, we aimed to emulate the intracellular environment with the common crowding agent Ficoll70 in vitro. Ficoll70 does not show any absorbance in the amide I and amide II regions, which would lead to an increase in noise in the FTIR difference spectroscopic experiments (Fig. S4). The difference spectrum of LOV in vitro with ;250 mg·ml 21 Ficoll70 indeed showed some suppression of the A9a-helix unfolding at around 1660 cm 21 (Fig. 5A). However, the crowding agent failed to emulate the intracellular environment, because all signals from secondary structure changes were Table 1 Band assignment of ICIRD spectra and in vitro FTIR difference spectra of PtAureo1a-LOV in the spectral region of carbonyl stretches including amide I Band positions are given in cm 21 .

In cell
In vitro  strongly suppressed nonspecifically. Ficoll70 is an uncharged, synthetic sucrose polymer interacting mainly via hydrogen bonding (50). To emulate other interactions, including charged interactions, we chose the protein BSA as crowding agent. The absorbance of BSA in the amide I and amide II regions (Fig. S4) leads to a lower accessible path length in the experiments and thereby significantly reduces the signal/noise ratio. In the presence of ;300 mg·ml 21 BSA, the difference spectrum of LOV in vitro shows a specific suppression of A9a-helix unfolding at around 1660 cm 21 (Fig. 5B). In addition, signals of other changes in secondary structure remain unaffected. As a result, the difference spectrum of LOV with BSA in vitro is nearly congruent to the spectrum in cells. Whereas Ficoll70 failed to emulate the intracellular environment (Fig. 5A), BSA was able to mimic in-cell conditions in FTIR difference experiments on LOV (Fig. 5B).

In-cell spectroscopy on LOV with bZIP effector
The discrepancy in mechanism between the light-induced protein response of LOV in vitro and in cells motivated us to investigate whether the signal progression to the effector domain bZIP remains operational in cells. Therefore, we studied bZIP-LOV with ATR-ICIRD spectroscopy and compared the difference spectrum with that of LOV in cells (Fig. 6A). The difference spectrum of bZIP-LOV shows the dimerization signal at 1628 (1) cm 21 and a strong a-helical unfolding at 1642 (2) cm 21 . For segment-resolved difference spectroscopy, we separated the contributions caused by the presence of the bZIP domain from the other signals by calculating a double difference of the spectrum of bZIP-LOV minus that of LOV. Thereby, we isolated the light-induced structural response of the effector bZIP (Fig. 6B). Upon illumination, bZIP responds with pronounced a-helix unfolding at 1651 (2) cm 21 and additional turn structural changes at 1675 (1) cm 21 . These marker bands have been assigned to be the result of a signal progression from the LOV sensory domain to the N-terminal bZIP effector domain (27). Therefore, this result evidences a functional signal progression of bZIP-LOV in cells. Some differences are found in cells as a reduced amplitude of the turn structural changes and the concomitant loss of amide II signal at 1569 (1) cm 21 . A further impact on the response of bZIP-LOV has been shown by the presence of DNA in vitro (27). Comparison of the segment-resolved spectrum of bZIP in cells with that in vitro shows that bZIP response in cells resembles a situation in vitro in the absence of DNA, although genomic DNA is present in the bacterial cells (Fig. 6B). Accordingly, the additional signal at 1642 (1) cm 21 in the presence of random STAT (signal transducers and activators of transcription)-box DNA was not observed in the double difference spectrum in cells (Fig. S5). Such direct comparison, however, should be taken with caution because additional signals were detected in cells at 1549 (2) cm 21 and 1520 (1) cm 21 in the amide II region as well as at 1450 (1) cm 21 and 1431 (2) cm 21 , which do not fully match those recorded in vitro (Fig. 6B). Summarizing, the signal progression from the LOV domain to bZIP is functional in cells, even though unfolding of the A9a-helix is suppressed in the intracellular environment.

Impact of the intracellular environment on the structural changes of phototropin-LOV1
The altered light-induced protein response of PtAureo1a-LOV led us to investigate other LOV domains in cells, because we wondered whether the intracellular environment has an impact on LOV proteins in general. To analyze a different type of LOV in cells, we studied the LOV1 domain of phototropin from C. reinhardtii via ATR-ICIRD spectroscopy. The difference spectrum of LOV1 in cells revealed a specific suppression of signals at 1675 (2) cm 21 , 1655 (1) cm 21 , 1552 (2) cm 21 , and 1540 (1) cm 21 compared with in vitro recordings, whereas other signals were not affected by the intracellular environment (Fig. 7). The characteristic frequencies of amide I signals and the corresponding signals in the amide II region indicate that the losses at 1675 (2) cm 21 and 1655 (1) cm 21 are caused by a suppression in turn and helix signals, respectively (34). Similar to the PtAureo1a-LOV, the intracellular environment appears to stabilize specific secondary structure elements in LOV1. These results support a general impact of the intracellular environment on the light-induced structural response of LOV proteins.

Discussion
Broadening the range of available methods for in-cell spectroscopy The investigation of soluble proteins in cells has been extensively advanced with different spectroscopic techniques. Many experiments have targeted the (co)localization of proteins by, for example, FRET, the redox state of the cofactor, such as heme in cytochrome c, or the conversion of ligands, such as quinones, by UV-visible and EPR spectroscopy. Here, we will focus in the following on techniques that provide information on the structure of proteins inside the cell. Fluorescence spectroscopy has been used for sensing the state of folding and aggregation of proteins in mammalian cells (51). Structural information on intraprotein distances can be obtained in vivo by analyzing spin-labeled proteins with in-cell double electronelectron resonance (DEER) (52,53). By establishing in-cell NMR spectroscopy, not only protein conformation but also three-dimensional structures can be determined (54,55). Strategies for selective isotopic labeling for NMR range from reintroducing labeled proteins into cells such as oocytes (56) to inducing expression after exchange of medium (54,55). Recent development of dynamic nuclear polarization solid-state NMR enables the investigation of proteins in intact mammalian cells at endogenous protein concentrations injected by electroporation (57). Structures of proteins in eukaryotic cells with .90% cell viability over 24 h by using a specialized bioreactor system were resolved with three-dimensional NOESY (58).
In this study, we expanded the range of in-cell methods with IR difference spectroscopy to analyze the light-induced response in conformation of soluble proteins in living cells, such as the 18-kDa LOV domain and the 29-kDa bZIP-LOV module. Previous IR spectroscopic approaches have targeted the membrane protein rhodopsin by using synchrotron radiation (37) or the catalytic center of hydrogenases by collecting cellular absorption spectra in a transparent frequency window (59,60). Complementary to in-cell fluorescence or UV-visible spectroscopy in which the chromophore is observed, ICIRD spectroscopy now additionally enables the investigation of the light-induced changes in secondary structure. ICIRD spectroscopy in ATR configuration suffers from increased noise compared with the transmission configuration but ensures a fully hydrated sample, living cells, and access to an investigation of the sample after ICIRD experiments. We could demonstrate  that 300,000 protein copies/cell, corresponding to an intracellular protein concentration between 120 and 330 mM, are sufficient to record difference spectra of proteins in cells (43). Accordingly, an increase in sensitivity by a factor of about 10 would enable future investigations of some proteins at endogenous expression level.
IR spectroscopy offers some fundamental advantages for an in-cell application. First, ICIRD spectroscopy can be performed label-free, without any purification step and without the requirement of a D 2 O buffer, in contrast to NMR and DEER experiments. Second, ICIRD spectroscopy has no protein size limitation in contrast to NMR, allowing the investigation of large full-length receptors in a native environment. Third, this noninvasive method provides detailed information about lightinduced changes of the chromophore as well as the secondary structure.

Mechanism of signal propagation in aureochrome revisited
Fluorescence recovery of the illuminated PtAureo1a-LOV domain showed that the adduct decay was slowed down by a factor of 1.4 in the intracellular environment compared with in vitro experiments. The effect could not be emulated by changing conditions, such as pH, excluded volume, or salt concentration. These results implicate the presence of small molecules or nonspecific protein-protein interactions in the intracellular environment that slow the adduct decay down. Emulation of the protein-protein interaction in fluorescence experiments by BSA was not successful because the high viscosity of the concentrated protein solution leads to an additional diffusion-dependent component in the recovery kinetics. The dimerization of LOV has been demonstrated to slow down the recovery considerably (21), but clearly not to the extent of the 2800 s observed in cells. Moreover, differences in recovery times were similarly found between the dimeric bZIP-LOV domain in cells and in vitro (Fig. S6). These findings speak against an effect by dimeric interactions in the cells and against any involvement of the A9a-helix in mediating a change in the recovery kinetics.
A strong acceleration of the recovery kinetics by the cellular environment was observed for the LOV-protein YtvA in bacterial cells by fluorescence spectroscopy and shown to be caused by dehydration (61). Similarly, the presence of imidazole in vitro leads to a strong acceleration of the adduct decay in LOV domains (29). However, stabilizing compounds have not yet been identified either in vitro or in cells. Internal steric and electronic effects have been shown to be relevant in adduct stabilization by up to 2 orders of magnitude in time, which opens the possibility of tuning LOV recovery by point mutations (30)(31)(32). Therefore, it is conceivable that compounds in the cells, which still have to be identified, change the structural environment of the FMN in the light state.
Studying the LOV domain of PtAureo1a in E. coli with ICIRD spectroscopy revealed a suppression of the light-induced A9ahelix unfolding in cells (Fig. 8). These observations are contrary to the proposed mechanism in vitro, in which A9a unfolding is required for b-sheet dimerization and formation of the light dimer structure (22,24). Other characteristic responses, such as the Ja-helix unfolding and the dimerization signal, remain unperturbed by the intracellular environment. These findings exclude the presence of an unfolded A9a-helix already in the dark, because it would lead to a concomitant loss of the dimerization signal, as observed for LOVDA9a (24). Therefore, we suggest that the dynamic equilibrium of folded and unfolded flanking helices demonstrated for phototropin-LOV-Ja (62) is shifted in the cells to promote a predominantly folded state of A9a in the dark and in the light. It should be noted that this conclusion is drawn from experiments on the isolated LOV domain without effector. Studies on the full-length protein by hydrogen-deuterium exchange MS support a destabilization of the A9a fold by light in vitro (22), whereas FTIR spectroscopic experiments have been inconclusive. Furthermore, the effect of binding of specific DNA sequences to the aureochrome on the A9a unfolding needs to be studied in the future by in-cell spectroscopy.
The segment-resolved double difference spectrum of bZIP-LOV in cells showed a loss of a-helical structure in the bZIP domain upon illumination, which demonstrates the functionality of the signal progression in cells (27). These results suggest that the signal progression to bZIP is independent of the A9ahelix unfolding. Therefore, light-induced A9a-helix unfolding does not reflect the native response of LOV in an intracellular environment. Instead, we propose that in the light state the A9a-helix is reoriented to enable a formation of the light-state dimer (Fig. 8). In IR difference spectroscopy, reorientation of secondary structure elements is not directly detectable. Such reorganization without unfolding for LOV dimerization was observed previously for the effector 4a-helix in the bacterial blue light transcription factor EL222 (63). The major advantage of ICIRD spectroscopy is that it covers the physiologically relevant conditions for a protein to be functional. For example, Heintz and Schlichting (22) showed in vitro that the light state of PtAureo1a has a much higher affinity to the target DNA in the presence of MgCl 2 . Moreover, nonspecific DNA binding to the bZIP domain of the cAMP-responsive element-binding protein is dramatically decreased (64). These findings might explain why, according to the double difference spectrum, bZIP is not bound to genomic DNA of E. coli. Accordingly, with ICIRD spectroscopy, the protein of interest is studied in a close to native environment, which ensures its functionality and prevents artifacts from in vitro conditions. Importantly, the intracellular environment not only affects the response of the PtAureo1a-LOV domain, but also has an impact on CrPhot-LOV1. For both photoreceptors, ICIRD experiments showed a specific suppression of secondary structures, indicating an impact of the intracellular environment on LOV protein mechanism in general. The deviations of FTIR difference spectra between in-cell and in vitro experiments demonstrate the relevance of ICIRD spectroscopy and encourage more investigations on other photoreceptors in cells.

Strengths and limitations of macromolecular crowding agents
The altered protein response of LOV photoreceptors in cells motivated us to emulate the intracellular environment in vitro. Ficoll70 is a common crowding agent, but it failed to reproduce the specific changes in signals of the A9a-helix. Instead, Ficoll70 induced a strong suppression of all changes in secondary structure elements, which might be attributed to the entropic preference for a compact structure of LOV by hard core repulsions in the absence of electrostatic interactions between LOV and the synthetic sucrose polymer (50). In contrast, the protein BSA at 300 mg·ml 21 (4.5 M) emulated the intracellular environment well with respect to the light-induced response of LOV. Therefore, BSA or lysozyme as crowding agents seem to be favorable to include enthalpic contributions to the interaction (50). The specific effect on the A9a-helix is explained by the stabilizing effect of macromolecular crowding on the protein fold, particularly for helices (65).
One might argue that at the typical millimolar concentration of FTIR difference spectroscopic experiments, the condition for a crowding environment is reached already. In the native environment, however, the receptor is present at much lower concentrations, which are obtained by diluting the protein with crowding agent. The routine application of these crowding agents in FTIR difference spectroscopic experiments is limited, because their absorbance in the amide I and water region at around 1650 cm 21 strongly reduces the signal/noise ratio. More importantly, the imitation of the intracellular environment with crowding agent does not include the mixture of other cellular components, such as metabolites or ions. Therefore, ICIRD spectroscopy provides a more realistic response in a heterogeneous environment and should be applied in addition to in vitro experiments.

Conclusions
Key steps in the signaling mechanism of LOV proteins comprise a repositioning or unfolding of flanking helices. The establishment of ICIRD spectroscopy in the transmission mode and in the ATR configuration allowed us to study the light-induced response of LOV proteins in living E. coli cells, including the 18-kDa PtAureo1a-LOV, 29-kDa PtAureo1a-bZIP-LOV, and 15-kDa CrPhot-LOV1. For both LOV and LOV1, we observed a specific suppression of secondary structure changes in living cells, demonstrating a general influence of the intracellular environment on the mechanism of LOV photoreceptors. The altered light-induced response was successfully emulated in vitro with the crowding agent BSA, but not Ficoll70, demonstrating the strong impact of macromolecular crowding by proteins in the intracellular environment.
Despite these deviations in response, characteristic structural changes of functional PtAureo1a-LOV are preserved in cells, such as the Ja-helix unfolding and the dimer rearrangement at the b-sheet. Moreover, studying the sensory domain together with the effector, PtAureo1a-LOV-bZIP, confirmed the a-helical unfolding in the bZIP domain previously assigned in in vitro experiments. Therefore, we propose that the signal in PtAureo1a propagates from LOV to bZIP via a repositioning of the A9a-helix instead of an unfolding to uncover the dimerization site for the light-state dimer. The A9a-helix unfolding has been assigned a key role in the mechanism of phototropin from plants and algae as well (66,67), underlining the necessity for in-cell studies.
The wide range of other soluble photoreceptors encourages further investigations in cells of the light-induced response with ICIRD spectroscopy. Current acquisition of ICIRD spectroscopy is on the time scale of minutes but in the future might be performed with much higher time resolution applying the rapid-scan technique. Implementation of modern quantum cascade lasers (68) might dramatically increase probe light intensity to lower the noise level in the ATR approach and would, as broad-band frequency combs, even reach a time resolution in the submicrosecond range (69). Further development of ICIRD spectroscopy might allow us to lower the number of copies per cell from the current 300,000 (C 300 mM) to reach endogenous levels and to investigate photoreceptors in the native organism, for example diatoms. In addition, ICIRD spectroscopy might be extended to studying the mechanism of proteins other than photoreceptors in cells by applying state-ofthe-art optogenetic approaches.

Experimental procedures
Expression and purification LOV (amino acids 238-378) and bZIP-LOV (amino acids 145-378) of aureochrome1a from P. tricornutum were expressed with an N-terminal His 6 tag in E. coli BL21 (DE3) and E. coli Arctic Express (DE3), respectively, using a pET28a (1) vector (21,27). LOV1 (amino acids 16-133) of phototropin from C. reinhardtii was expressed with an N-terminal His 15 tag in E. coli BL21 pLysE (DE3) with a modified pMAL-c2x vector (70). The cells were cultivated at 37°C and 120 rpm in DYT medium up to an OD 600 of 0.5 and subsequently cooled to 18°C. At an OD 600 of 0.8, expression was induced with 10 mM isopropyl-b-D-thiogalactopyranoside. LOV and LOV1 were expressed for 20 h in the dark, whereas expression of bZIP-LOV went on for over 40 h. For in vitro experiments, LOV, bZIP-LOV, and LOV1 were expressed and purified as described previously (21).

Preparation for in-cell spectroscopy
The cells were centrifuged at 3500 3 g and 4°C. For fluorescence spectroscopy, cells were washed two times with a 50 mM phosphate buffer, pH 8, 300 mM NaCl and diluted to an OD 600 of 1.5. For IR spectroscopy, the pellet was washed two times with a saline solution of 150 mM NaCl and 5 mM KCl. After centrifugation, most of the supernatant was discarded, and the pellet was gently homogenized with the remaining buffer, resulting in a concentrated cell suspension with an OD 600 of ;200-300.

In-cell fluorescence spectroscopy
Fluorescence experiments were performed in a Jasco FP8300 spectrometer (Jasco, Gross-Umstadt, Germany) after transferring 1.5 ml of the sample into a fluorescence cuvette (Hellma 119.00F-QS; 10 mm 3 4 mm) equipped with a magnetic stirrer bar (8 mm 3 1 mm 2 ) rotating at 800 rpm. Excitation light was attenuated with a 5% hole filter and a long pass filter (31% transmission at 447 nm). The excitation shutter was opened only during measurements to minimize sample exposure. Emission spectra were recorded for 1 s every 10 s with excitation at 447 nm and using emission and excitation slit widths of 2.5 nm at 20°C. Recovery kinetics were obtained at 495 nm with excitation at 447 nm at 20°C after illumination with a blue LED (451 nm, 60 milliwatts·cm 22 ; Phillips Lumileds) for 100 ms. Excitation and emission slit widths of 2.5 and 20 nm, respectively, were chosen. In vitro experiments were performed with identical settings as the in-cell experiments. Emission spectra of 25 mM LOV were obtained in 50 mM phosphate buffer, pH 8, 300 mM NaCl, and spectra of 8 mM FMN were obtained in water. The recovery kinetics of 25 mM LOV was recorded with and without 100 mg/ml Ficoll70 in a 50 mM phosphate buffer, pH 8, 300 mM NaCl.

Determination of the protein copy number
To obtain the average number of proteins per cell, the fluorescence intensity I 495 at 495 nm with excitation at 447 nm of 1.5 ml of a cell suspension expressing LOV was determined at an OD 600 of 0.5 and 1.5, respectively. The fluorescence intensity (I 495 ) was then correlated to the concentration of LOV by recording a series of dilutions of purified protein from a stock solution of 1.5 ml with a defined absorbance at 448 nm (A 448 ) and using the extinction coefficient of free FMN (42) at 450 nm (e 450 ) = 12,200 M 21 ·cm 21 as well as the path length d in the UV-visible spectroscopic experiment (Fig. S1). This procedure yields A 448 of the proteins in the cells by recording I 495 and dividing by the correlation factor of 516,000 counts. It should be noted that scattering of the cells leads to a small deviation in this correlation, which was estimated to be 1.2 by recording the turbidity of a mixture of FMN and polystyrene particles (100 nm, Nanosphere Size Standards, Fisher Scientific, Schwerte, Germany) and its fluorescence after dilution. The protein den-sity (n protein ) was calculated by Equation The cell number per volume (n cell ) of the cell suspension was determined with an improved Neubauer hemacytometer slide, resulting in n cell = 4,92·10 11 cells·liter 21 at OD 600 = 0.5 and n cell = 1.48·10 12 cells·liter 21 at OD 600 = 1.5. The protein copy number per cell (N) was then calculated by Equation 2.
N ¼ n protein n cell (Eq. 2) In-cell IR difference spectroscopy Spectra were recorded on a Bruker IFS 66v or IFS 66/S spectrometer with a mercury cadmium telluride detector at a resolution of 4 cm 21 for LOV and bZIP-LOV, whereas the resolution was 2 cm 21 for LOV1. A long pass filter was mounted in front of the detector to block light above 2000 cm 21 for a better signal/noise ratio and protection of the detector. FT was performed with a zero filling factor of 4. For FTIR spectroscopic experiments, ;2 ml of the concentrated cell suspension was enclosed between two BaF 2 windows (Korth Kristalle, Altenholz, Germany) sealed with grease. The sample thickness was adjusted to an absorbance of 0.8-1.1 at 1650 cm 21 . 3072 scans were performed on LOV before and after 20 s, illuminating the sample with a blue LED (455 nm, 27 milliwatts·cm 22 ; Phillips Lumileds) at 10°C. The difference spectrum resulted from different biological replicates with a total number of 18,432 scans. For ATR-FTIR experiments, 15 ml of the concentrated cell suspension was placed on a diamond/ZnSe internal reflection element with nine active reflections and an effective pathlength of ;6 mm (DuraSamplIRII, Smiths, CT, USA). 1024 scans for bZIP-LOV, 1024 scans for LOV1, and 3072 scans for LOV were acquired before and after 10-s illumination with a blue LED (455 nm, 25 milliwatts·cm 2 ; Phillips Lumileds). All samples were kept at 10°C except for LOV1, which was investigated at 20°C. The spectra resulted from experiments on different biological replicates with a total number of 73,728 scans for LOV, 26,624 scans for bZIP-LOV, and 9216 scans for LOV1. In vitro difference spectra of 1.6 mM LOV1 and 356 mM LOV with and without ;250 mg/ml Ficoll70 or 300 mg/ml BSA as crowding agents were measured in a 50 mM phosphate buffer, pH 8, 300 mM NaCl with the same settings.

Data availability
All data presented and discussed are contained within the article or in the supporting information.
Acknowledgments-We thank Ina Ehring for technical assistance and Dominik Cholewa for help with the cell-counting procedure.