Spatial and temporal control of expression with light-gated LOV-LexA

Abstract The ability to drive expression of exogenous genes in different tissues and cell types, under the control of specific enhancers, has been crucial for discovery in biology. While many enhancers drive expression broadly, several genetic tools were developed to obtain access to isolated cell types. Studies of spatially organized neuropiles in the central nervous system of fruit flies have raised the need for a system that targets subsets of cells within a single neuronal type, a feat currently dependent on stochastic flip-out methods. To access the same cells within a given expression pattern consistently across fruit flies, we developed the light-gated expression system LOV-LexA. We combined the bacterial LexA transcription factor with the plant-derived light, oxygen, or voltage photosensitive domain and a fluorescent protein. Exposure to blue light uncages a nuclear localizing signal in the C-terminal of the light, oxygen, or voltage domain and leads to the translocation of LOV-LexA to the nucleus, with the subsequent initiation of transcription. LOV-LexA enables spatial and temporal control of expression of transgenes under LexAop sequences in larval fat body and pupal and adult neurons with blue light. The LOV-LexA tool is ready to use with GAL4 and Split-GAL4 drivers in its current form and constitutes another layer of intersectional genetics that provides light-controlled genetic access to specific cells across flies.


Introduction
Patterned expression of genes is essential for differentiation of distinct cell types during development. Enhancers defining expression patterns have long been used in binary expression systems to study development and function of specific cell types. Binary expression systems couple enhancer-led expression of an exogenous transcription factor to expression of a transgene that sits downstream of promoter sequences exclusively bound by the exogenous transcription factor (del Valle Rodriguez et al. 2011). The GAL4-UAS system uses the yeast transcription factor GAL4 under the control of an enhancer that binds upstream activating sequences (UAS), which in turn drive expression of transgenes sitting downstream UAS (Brand and Perrimon 1993). Random insertions of P-elements carrying GAL4 into the genome were used to trap enhancers, with the expression of GAL4 dependent on neighboring regions in the genome Spradling and Rubin 1982;Bellen et al. 1989;Grossniklaus et al. 1989;Wilson et al. 1989;Perrimon et al. 1991;Lukacsovich, 2001;Bellen 2012, 2014). More recently, stretches of noncoding genomic DNA carved out of known gene enhancers, or from regions predicted to contain enhancers, have been extensively used to generate large collections of driver lines (Pfeiffer et al. 2008(Pfeiffer et al. , 2010Jenett et al. 2012;Kvon et al. 2014;Yáñez-Cuna et al. 2014;Tirian and Dickson 2017). Other binary expression systems were added to the fruit fly genetic toolbox. The LexA-LexAop (Lai and Lee 2006) and the QF-QUAS (Potter et al. 2010;Riabinina et al. 2015) systems rely on exogenous transcription factors and DNA binding sequences, and can be combined with GAL4-UAS, allowing for independent access to multiple cell types in the same organism (e.g. Feng et al. 2014Feng et al. , 2020Sen et al. 2017;Ribeiro et al. 2018). The spatial resolution, or cell type specificity of binary expression systems, is determined by the enhancer driving expression of the exogenous transcription factor. Given that it is still not possible to design enhancers specific for many cell types (Serebreni and Stark 2021), it is necessary to screen to obtain enhancers specific for the cell type of interest.
Several methods, under the umbrella of intersectional genetics, were developed to further restrict transgene expression in binary systems. The modular nature of the GAL4 activation and DNA-binding domains enables separation of GAL4 into two parts, with each split-GAL4 half placed under the control of a different enhancer (Luan et al. 2006). The final transgene expression occurs only in cells that express both split-GAL4 halves, which dimerize through added leucine zipper domains to form a fully functional transcription factor. Existing collections of split-GAL4 lines targeting single neuronal types were established by screening for enhancer pairs that together provide exclusive access to specific cell types of interest (e.g. Wu et al. 2016;Dionne et al. 2018;Namiki et al. 2018;Dolan et al. 2019;Schretter et al. 2020;Wang et al. 2020;Sterne et al. 2021). Other powerful methods of restricting expression to single or fewer cells include the recombinase-based systems for stochastic labeling (Lis et al. 1983;Golic and Lindquist 1989;Xu and Rubin 1993;Lee and Luo 1999;Hadjieconomou et al. 2011;Nern et al. 2015;Isaacman-Beck et al. 2020), temperature sensitive mutations of Gal80, the repressor of GAL4 (Nogi et al. 1977;Lee and Luo 2001;McGuire et al. 2003), and use of transcription factors modified to drive transcription in the presence of an ingestible drug (McGuire et al. 2004). In addition to providing temporal and spatial control, however, these methods either involve increase in temperature that unleashes a stress response in all cells of the organism (Lindquist 1986), or addition of drugs with potential off-target effects, both of which may affect experimental outcomes.
With several recent advances in optics and laser technology, light is now easily modulated at the level of its spectrum, intensity, and even beam shape (Chen et al. 2018). The high spatial and temporal precision of light pulse delivery to living organisms has the potential to take the spatial and temporal resolution of transgene expression to new levels. Several photosensitive proteins have been introduced into exogenous expression systems to amass the advantages of light as a precise trigger (Di Ventura and Kuhlman 2016;de Mena et al. 2018;di Pietro et al. 2021). Phytochromes (Phy) are sensitive to red and far-red light, and bind the phytochrome-interacting factor (PIF) in presence of light (Yamamoto and Deng 1999). The Photo-GAL4 tool capitalizes on the PhyB and PIF light-dependent interaction to reconstitute a complete GAL4 upon exposure to light (de Mena and Rincon-Limas 2020). To function, Photo-GAL4 requires addition of phycocyanobilin (PCB), a chromophore that is absent in animal cells (Yamamoto and Deng 1999), limiting its applicability (de Mena and Rincon-Limas 2020). The cryptochrome split-LexA system similarly uses cryptochrome 2 and its binding partner, the cryptochrome interacting protein, to gate reformation of split-LexA with blue light (Szuts and Bienz 2000;Chan et al. 2015). In ShineGal4, the pMagnet and nMagnet photoswitches derived from the blue light photoreceptor VVD endogenous to Neurospora crassa replace the leucine zippers in split-GAL4 halves and heterodimerize upon exposure to light (Kawano et al. 2015;di Pietro et al. 2021). ShineGal4 functions in several epithelia across developmental stages, with its current form limited to a few drivers.
To circumvent these limitations and expand the photosensitive toolbox in Drosophila, we developed a light-gated expression system based on the light, oxygen, or voltage (LOV) domain originally found in oat phototropin 1 (Avena sativa) (Christie et al. 1998(Christie et al. , 1999Crosson and Moffat 2002) and LexA (Horii et al. 1981;Walker 1985;Rhee et al. 2000;Masuyama et al. 2012), under the control of UAS sequences. LOV-LexA gates expression of transgenes with blue light in vivo, in several cell types in larval, pupal, and adult fruit flies. LOV-LexA can be directly crossed to split-GAL4 and GAL4 drivers, adding thus another layer of spatiotemporal control to transgene expression in Drosophila that is combinable with existent binary expression systems and transferable to other model organisms.

Quantification of signal intensity in cell culture
Five images per well, in 24-well plates, were obtained from immuno-stained cells under an inverted fluorescence microscope (Leica DM IL LED), with a 5Â objective, with green and red filter cubes. The open-source software CellProfiler (version 4.1.3) (Carpenter et al. 2006;McQuin et al. 2018) was used to segment individual cells in each image, based on the test construct fluorescent tag signal with the Otsu method, and measure the amount of LexAop reporter in each segmented cell, as a proxy for transcription levels of LexAop reporters by test constructs, LexAtransactivator-eLOV-tag. We used the mean intensity of segmented cells in Cell Profiler, Object MeanIntensity, as a measure of mean pixel intensity per segmented cell, in the green and red channels. This measure is a normalized value by default in Cell Profiler and is plotted in Fig. 1, c and d and Supplementary Fig. 1, b and d. Scripts written in Python (version 3.8, http://www.py thon.org) were then used to read data values per Object, average all segmented cells across at least 2 independent experiments and plot the data.

Drosophila culture and genetics
Fly stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. All the strains of the fruit fly Drosophila melanogaster used in this study are listed in Supplementary Table 2. Fruit flies were maintained on standard cornmeal-agar medium supplemented with baker's yeast and incubated at 18 or 25 C with 60% humidity and 12 h light/ dark cycle. Males and females were tested indiscriminately throughout experiments. Larvae of the second and third instar were used for tests on fat body with Cg-GAL4; 2-4 days after puparium formation (APF), pupae were used for tests in neurons in the central brain and dorsal abdominal oenocytes; adults ranging from 1 to 6 days old were used for tests in adult neurons.

Light stimulation in fat body
To test UAS-LOV-LexA, UAS-eLOV-nls-tdTomato-mLexA:GAD, and UAS-eLOV-nls-tdTomato-mLexA:VP16 constructs in fat body cells, second-to-third instar larvae from crosses with Cg-GAL4 (Asha et al. 2003) were removed from the food, washed in water, and placed in a well with 40 ml of 15% sucrose in water solution; 1 larva per well in a 96-well plate wrapped in aluminum foil to shield larvae from light. Pulses of blue light were delivered to individual wells, with the 96-well plate mounted on an inverted microscope. Light from a blue LED (LEICA DM IL LED) was delivered at 11.7 mW, at 1 Hz for 30 s (Supplementary Table 3). Half the plate was not exposed to light and larvae in such wells served as controls. The fat bodies were dissected 6-12 h after light delivery, fixed, and immunostained with anti-GFP chicken antibody dilution 1:1,000 (Rockland, catalog # 600-901-215S; RRID: AB_1537403) and anti-RFP rabbit antibody dilution 1:2,000 (Rockland, catalog # 600-401-379, RRID: AB_11182807). Secondary antibodies were goat anti-chicken Alexa Fluor 488 1:1,000 (Thermo Fischer Scientific, catalog # A-11039; RRID: AB_2534096) and goat anti-rabbit Alexa Fluor 568 1:1,000 (Thermo Fischer Scientific, catalog # A-11011; RRID: AB_143157). Stained fat bodies were mounted with Vectashield Antifade Mounting Medium (BIOZOL, Ref H-1000) and imaged under a Leica TCS SP8 confocal microscope.

Light stimulation in neurons
To stimulate neurons in pupal stages, pupae were recovered from vials by adding water to the vial wall to dissolve the glue-binding pupal cases to the pupation site. Pupae aged between 2 and 3 days APF were then dried on a kimwipe tissue and glued on a double-side sticky tape spread on a cover slip (Fig. 3c) that was attached to a microscope slide with plasticine. Adult flies were glued to a custom-made aluminum or plastic plate with a hole, with a diameter ranging from 300 to 400 mm, large enough to expose part of the head of the adult fly and shield the rest of the fly from light. Melted Eicosane 99% (Aldrich 219274-5G) was added to the thorax and part of the head to immobilize the adult fly and shield part of the head from light (Fig. 4f). Such custom holders were mounted on a microscope slide with plasticine, separating the flies from the slide.
Slides bearing pupae or adult flies were then mounted on an upright confocal microscope (Leica TCS SP8) for preprogrammed serial light delivery with the 458-nm laser at 10% power at 5.75 mW (Supplementary Table 3). Each light pulse was composed to 30-50 scans across a depth of 200-400 mm. Using the xyzt mode of Leica software together with position mapping, it was possible to deliver light serially to many pupae or adult flies prepped together. After light delivery, cover slips with pupae were vertically inserted into new food vials and incubated for 2 days for expression in neurons in the fly brain, before dissection. Adult flies were removed from holders after light delivery, by breaking the brittle Eicosane, placed in fresh food vials, and incubated at 25 C for 1-2 days before dissection.
Adult brains were dissected in cold PBS, fixed in 4% PFA for 20-40 min at room temperature, washed in PBS with 0.5% Triton X-100 2 times, incubated with DAPI (Invitrogen, D1306) at dilution 1:3,000 in PBS with 0.5% Triton X-100 for 10 min, washed again, mounted with Vectashield Antifade Mounting Medium (BIOZOL, Ref H-1000), and imaged on the same day under a Leica TCS SP8 confocal microscope.

Live imaging in oenocytes and neurons
The pupal case covering the most anterior abdomen in the case of oenocytes (w; 109(2)-GAL4, UAS-CD8:GFP/þ; UAS-LOV-LexA/þ), or the head in case of neurons (w, LexAop-CsChrimson:Venus;þ; UAS-LOV-LexA/fru-GAL4), of pupae lined up on a double side sticky tape on a slide (see above, Fig. 4a), was removed under low light conditions, or as low as possible since pupal cuticle is transparent. The slide was mounted on a Leica TCS SP8 confocal microscope, and positions for serial imaging were marked. The pupae were then kept in the dark for 30 min before live imaging was initiated.
Oenocytes were first scanned in the red channel alone, followed by exposure to blue light (see Supplementary Table 3). Afterward, oenocytes were imaged in the red channel every 5 min for at least 1 h and 40 min. Oenocytes were imaged one last time in the red and green channels, obtain CD8:GFP signal to delineate the oenocyte cell body, in addition to LOV-LexA. To image fruþ neurons in pharate adult pupae (4 days APF), a scan in the green and red channels preceded the exposure to light (see Supplementary Table 3), after which pupal heads were scanned in the red and green channels every hour. Most pupae eclosed after 12-14 h under the confocal microscope.

Quantification of signal intensity in flies
Mounted fat body and brain tissue were imaged under a Leica TCS SP8 confocal microscope with a 20.0Â objective, using the lasers 405, 488, and 568 nm to image DAPI, Venus, and tdTomato respectively. The same laser power, gain, and line averaging were used within each experiment to compare fluorescence levels across light and dark conditions. Z-stacks thus obtained were cropped in XY and Z with ImageJ/Fiji (Schindelin et al. 2012), to isolate cell bodies expressing the test construct. Z-projections of these crops were loaded with the Scikit-image image processing package into Python (version 3.8, http://www.python.org), to obtain pixel intensity values in a 2-dimensional matrix for each color channel in RGB. The ratio of the mean pixel intensity in green and red channels, or green and blue channels, was used to compare relative fluorescence levels of reporter gene to test constructs.

Design of an expression system gated by light
The LOV2 domain of A. sativa phototropin 1, AsLOV2, is photosensitive (Harper et al. 2003;Lungu et al. 2012;Zayner et al. 2012;Diensthuber et al. 2014). Exposure to blue light causes the JA helix to unfold, thereby freeing its C-terminus (Harper et al. 2003). This property arises from interactions with flavin, the blue lightabsorbing chromophore present in animal cells, and can be used The constructs for GAD were LexA:GAD-tdTomato and mLexA:GAD-tdTomato, for p65 were LexA:p65-tdTomato and mLexA:p65-tdTomato, and for VP16 were LexA:VP16-tdTomato and mLexA:VP16-tdTomato. Mutagenizing NLS-like sequence reduces the transcriptional activity of mLexA-transactivator chimeras compared to LexA-transactivator chimeras. At least 200 cells with medium levels of expression of mCherry or tdTomato, from at least 2 transfections of S2Rþ cells are represented for each condition. d) S2Rþ cells were transfected with a reporter, LexAop-myr:GFP, together with the driver MET-GAL4 and the test construct, to examine light-gated transcription for test constructs. e) Expression of LexAop reporter myr:GFP in relation to expression of mLexA-transactivator chimeras combined with eLOV. Placement of eLOV-nls N-terminal followed by the fluorescent protein tdTomato and the mLexA-transactivator chimera yielded the best signal for cells exposed to pulses of blue light, while maintaining a low reporter signal in cells kept in the dark. As in (b), cotransfection of UAS-mCherry with UAS-CD8:GFP (second bar in plot) or LexAop-myr:GFP (first bar in plot) served as a positive and negative control, respectively. The test constructs were eLOV-nls-tdTomato-mLexA:GAD (GAD), eLOV-nls-tdTomato-mLexA:p65 (p65), and eLOV-nls-tdTomato-mLexA:VP16 (VP16). At least 200 cells with medium levels of expression of mCherry or tdTomato, from 2 to 5 transfections of S2Rþ cells are represented for each condition. c, e) *** represents P-values <0.001; n.s. represents P-values >0.05, obtained with Student's t-test. f) Representative examples of S2Rþ cells expressing test constructs indicated above the images under the control of MET-GAL4. The eLOV-nls-tdTomato-mLexA:GAD forms clusters in the cytoplasm, whereas both eLOV-nls-tdTomato-mLexA:p65 and eLOV-nls-tdTomato-mLexA:VP16 are evenly distributed in the cytoplasm, and sometimes nucleoplasm, like mCherry. g) Schematic representation showing how eLOV-nls-tdTomato-mLexA:p65 (or LOV-LexA) works.
to expose a small peptide of up to 10 amino-acid residues long, added to or integrated into the JA C-terminus (Huala et al. 1997;Christie et al. 1998Christie et al. , 1999Salomon et al. 2000;Harper et al. 2003). This photosensitive system has been used to cage several peptides in genetic tools, including the nuclear-localizing signal (NLS) to shuttle proteins to the nucleus, the tobacco etch virus protease (TEVp) cleavage site for an integrator of neuronal activity and reporters of protein-protein interactions (Wang et al. 2010;Strickland et al. 2012;Motta-Mena et al. 2014;Niopek et al. 2014Niopek et al. , 2016Guntas et al. 2015;Yumerefendi et al. 2015Yumerefendi et al. , 2016Jayaraman et al. 2016;Reade et al. 2017;Smart et al. 2017;Salinas et al. 2018;van Haren et al. 2018;Zhao et al. 2018;Cavanaugh et al. 2020). Recent work employed directed evolution on the native AsLOV2 to develop the evolved LOV (eLOV) that presents improved stability in the dark state due to 3 single-nucleotide mutations (Kim, Wang, et al. 2017;Wang et al. 2017;Kim et al. 2019). We added the short NLS from SV40 (Pfeiffer et al. 2010), to make eLOV-nls and regulate availability of the NLS to the cell milieu with blue light (Niopek et al. 2014).
To build a transcription factor gated by light, we selected the binary expression system LexA/LexAop (Szuts and Bienz 2000;Loewer et al. 2004;Lai and Lee 2006), that is complementary to the widespread GAL4-UAS system and has been successfully incorporated in diverse model organisms (Lai and Lee 2006;Emelyanov and Parinov 2008;Nonet 2020). LexA is a repressor of transcription endogenous to Escherichia coli (Horii et al. 1981), where it regulates the SOS response (Walker 1985). Addition of an activation domain to the C-terminal of LexA renders such LexA-transactivator chimeras capable of activating transcription of transgenes sitting downstream of the LexA operator (LexAop) (Rhee et al. 2000;Lai and Lee 2006). In Drosophila, the most common LexA-transactivator chimeras contain the activation domains GAL4 activation domain (GAD, LexA:GAD), p65 (LexA:p65), or VP16 (LexA:VP16) (Rhee et al. 2000;Szuts and Bienz 2000;Lai and Lee 2006;Emelyanov and Parinov 2008;Yagi et al. 2010). Despite its bacterial origin, LexA carries an NLS-like sequence that allows it to shuttle to the nucleus when expressed in eukaryotic cells (Rhee et al. 2000;Pfeiffer et al. 2010;Masuyama et al. 2012). To make the translocation of LexA to the nucleus solely dependent on eLOV-nls, we mutagenized the NLS-like sequence in the LexA codon optimized for D. melanogaster (Pfeiffer et al. 2010;Rhee et al. 2000) and created a mLexA (Fig. 1a, see Materials and Methods). We examined the propensity to translocate to the nucleus of mLexA-transactivator chimeras by transfecting such constructs into the S2Rþ Drosophila cell line (Echalier 1997), together with the metallotheionein-GAL4 (MET-GAL4) that drives ubiquitous expression upon addition of CuSO 4 (Velichkova et al. 2010), and the reporter myr:GFP under control of LexAop sequences (Pfeiffer et al. 2010) (Fig. 1b). All three chimeras of unmodified LexAtransactivator drove expression of the LexAop-myr:GFP to levels similar to UAS-CD8:GFP (Fig. 1c), confirming their ability to shuttle to the nucleus (Rhee et al. 2000;Pfeiffer et al. 2010;Masuyama et al. 2012). In contrast, mLexA-transactivator chimeras led to reduced expression of the reporter transgene (Fig. 1c), confirming that the NLS-like sequence in LexA plays a major role in shuttling LexA to the nucleus.
We combined eLOV-nls with the 3 mLexA-transactivator chimeras, and a fluorescent protein (Shaner et al. 2004) placed each combination under the control of UAS and tested their performance in S2Rþ cells cotransfected with MET-GAL4 and LexAop-myr:GFP for constructs tagged with a red fluorescent protein or LexAop-myr:tdTomato for constructs tagged with GFP ( Supplementary Fig. 1, a and c). Several mLexA constructs carrying eLOV-nls at the C-terminal led to the expression of myr:GFP in the dark ( Fig. 1e and Supplementary Fig. 1, b and d), indicating that NLS is frequently uncaged with eLOV-nls at the C-terminal end, even in the absence of light. On the other hand, many of the mLexA constructs with eLOV-nls N-terminal were unable to drive the expression of myr:GFP upon presentation of blue light ( Supplementary Fig. 1, b and d). The combinations made with LexA:GAD chimera formed clusters in the cytoplasm irrespective of the fluorescent protein used as a tag (Supplementary Fig.  1, e and h), while most other combinations were homogenously distributed in the cytoplasm and occasionally in the nucleoplasm ( Fig. 1f and Supplementary Fig. 1f-j). Of note, cells with high levels of expression of many of the mLexA constructs tested, presented the expression of myr:GFP irrespective of the light regime delivered (data not shown), indicating that eLOV is unstable if expressed at high levels, as previously observed (Kim, Wang, et al. 2017, 2019. Cells expressing eLOV-nls-tdTomato-mLexA:p65 and eLOV-nls-tdTomato-mLexA:VP16 at moderate levels presented no to very little LexAop-myr:GFP reporter expression in the dark and displayed an increase in the expression of LexAop-myr:GFP upon exposure to blue light (Fig. 1e). These two constructs thus gathered the characteristics necessary for a light-gated expression system and were used to create transgenic flies. Despite its shortcomings, the eLOV-nls-tdTomato-mLexA:GAD was also injected since mLexA:GAD is suppressible by Gal80, potentially providing another level of regulation of a light-gated expression system.

Characterization of eLOV-nls-tag-mLexA chimera constructs in vivo
Drosophila larvae have transparent cuticle that allows for internal tissues to be exposed to unabated light. The bilateral, multilobed fat body running along the larva, is visible underneath the body wall musculature and is targeted by the collagenase enhancer (Cg-)GAL4 (Asha et al. 2003). Distribution of eLOV-nls-tdTomato-mLexA:GAD, eLOV-nls-tdTomato-mLexA:p65, or eLOV-nls-tdTomato-mLexA:VP16 in larval fat body followed the trend observed in S2Rþ cells (Fig. 2, a and d and Supplementary Fig. 2, a-f), with eLOV-nls-tdTomato-mLexA:GAD forming clusters ( Supplementary Fig. 2, c and d) and eLOV-nls-tdTomato-mLexA:p65 or eLOV-nls-tdTomato-mLexA:VP16 distributing evenly in the cytoplasm, and occasionally in the nucleoplasm (Fig. 2, c, d, f, and g and Supplementary Fig. 2, f and g). To test the ability to induce expression of a reporter under control of LexAop sequences, LexAop-CsChrimson:Venus (Klapoetke et al. 2014) (hereafter referred to as Venus), second and third instar larvae reared at 18 C were placed in 96-well plates in a 15% sucrose solution, to repress their tendency to wander, and exposed to several pulses of low intensity blue light (Fig. 2, a Table 3). Larvae were then incubated at 25 C for 7-11 h, after which the fat body was dissected, fixed, and stained. Despite considerable expression of eLOV-nls-tdTomato-mLexA:GAD and eLOVnls-tdTomato-mLexA:VP16 in fat body cells, exposure to blue light failed to elicit expression of the reporter (Supplementary Fig. 2, a-h). In contrast, exposure of larvae expressing eLOV-nls-tdTomato-mLexA:p65 to as little as three pulses of blue light led to increase in reporter expression under control of LexAop sequences ( Fig. 2a-i). Surprisingly, a lower number of pulses of blue light combined with longer incubation at 25 C resulted in maximum increase in reporter expression (Fig. 2i). This suggests that despite low intensity, exposure to too many pulses of blue light leads to less efficiency of light-gated expression. Given that expression of eLOV-nls-tdTomato-mLexA:p65 in S2Rþ and fat body cells kept in the dark presented no or very low expression of the reporter gene (a) Fig. 2. LOV-LexA is gated by light in vivo. a) Drosophila larvae expressing LOV-LexA in the fat body were exposed to blue light and examined for expression of the LexAop reporter, as well as the construct selected for LOV-LexA. b) Schematics showing the timeline of the experiment, light regime. Second or young third instar larvae were selected from vials kept at 18 C and transferred to 15% sucrose solution. The first light pulse was delivered immediately after this transfer, with a blue LED on an inverted microscope (Supplementary Table 3). Larvae were placed in the dark at 25 C between light pulses (see text for details), until dissection. Fat bodies expressing LOV-LexA with Cg-GAL4 for second to third instar larvae kept in the dark (c-e) or exposed to 3 30-s pulses of blue LED light at 1 Hz (f-h). Exposure to blue light appears to alter LOV-LexA cellular distribution (d, g) and leads to the expression of LexAop-CsChrimson:Venus in fat body cells as detected with anti-GFP antibody (e, h). i) Ratio of fluorescence, measured as pixel intensity in confocal-acquired images, of anti-GFP signal/anti-RFP signal for stained fat bodies from larvae with the genotype w, LexAop-CsChrimson:Venus; Cg-GAL4/ þ; UAS-LOV-LexA/þ that were kept in the dark (N ¼ 7, representative example in (c)-(e), exposed to 6 light pulses and dissected after 7 h (N ¼ 10), or 11 h (N ¼ 11), or exposed to 3 light pulses and dissected 11 h later (N ¼ 16, representative example in (f)-(h). Varying number of light pulses and the incubation period at 25 C before dissection led us to conclude that LOV-LexA gates expression with light in fat body and that LOV-LexA light-gated expression is highest with 3 light pulses and an 11-h incubation period at 25 C. *** represents P-values <0.001, n.s. represents P-values >0.05, 2-tailed Mann-Whitney tests. Exposure to blue light leads to an increase in the amount of Venus relative to LOV-LexA levels. j) Schematics showing the timeline of the experiment, light exposure, and functional imaging. k) Drosophila pupae expressing LOV-LexA and CD8:GFP in oenocytes were mounted on double-side sticky tape, and an opening in the pupal case that exposes oenocytes was created. Pupae expressed LOV-LexA and CD8:GFP in oenocytes with the following genotype: w; 109(2)-GAL4, UAS-CD8:GFP/þ; UAS-LOV-LexA/þ. l) Representative images of pupal oenocytes showing LOV-LexA before (before) and immediately following exposure to blue light (after), 60 min after exposure to blue light (recovery), and 120 min after light exposure (final scan). The final scan included the green channel to capture CD8:GFP, coexpressed with LOV-LexA, and used to delineate the cell body. The light used to capture GFP is blue and elicited another translocation of LOV-LexA to the nucleus, thereby demonstrating that the oenocytes were healthy after imaging. m) Mean nuclear tdTomato fluorescence over time, imaged live every 5 min. Shades represent standard error of the mean (SEM). LOV-LexA translocates to the nucleus upon exposure to blue light within minutes in oenocytes and slowly leaks out of the nucleus after exposure to blue light. and that exposure to blue light led to increase in reporter expression, the construct eLOV-nls-tdTomato-mLexA:p65 was selected for further studies and named LOV-LexA (Fig. 1g).

and b and Supplementary
The use of the AsLOV2 domain to cage a NLS signal has been previously demonstrated to effectively move coupled proteins into the nucleus in a light-dependent manner (Niopek et al. 2014;Yumerefendi et al. 2015). To determine the kinetics of LOV-LexA nuclear translocation, we expressed LOV-LexA in oenocytes, which are large cells sitting underneath the cuticle with roles in secretion and metabolism (Makki et al. 2014). Adult oenocytes arise in pupae and reach their final locations through several bouts of migration during metamorphosis. We imaged stationary oenocytes in pupae aged between 2 and 3 days APF, through the transparent cuticle, after removal of the overlying pupal case (Fig. 2, j and k). After preparation of the samples, LOV-LexA was present in the cytoplasm as well as in the nucleus in abdominal oenocytes (Fig. 2l "before"). Exposure to blue light (485 nm, 2.53 mW, 30 slices, Supplementary Table 3) leads to a rapid accumulation of LOV-LexA in the nucleus, that decreases over time ( Fig. 2l "after," "recovery," Fig. 2m). To determine the location of the cytoplasm, cells were imaged to detect CD8:GFP as well as tdTomato in LOV-LexA 100 min after exposure to blue light ( Fig. 2l "final scan," not depicted in the graph in Fig. 2m).
LOV-LexA accumulated again in the nucleus in all oenocytes imaged (n ¼ 14), indicating that the reduction of LOV-LexA levels in the nucleus over time is not due to the general degradation of the live preparation. LOV-LexA thus exhibits fast translocation to the nucleus, which peaks 5 min after exposure to blue light, and a slower movement out of the nucleus, reaching minimum levels after 20 min in the dark.  Fig. 3. LOV-LexA gates expression with light in neurons. a, b) Schematic representation outlining the experiment. Pupae reared at 18 C aged 2-3 days APF were removed from vials, mounted on double side sticky tape on a cover slip and kept in the dark (a), or pasted onto a slide and exposed to blue light (b). Mounted pupae kept in the dark or exposed to light were shifted to 25 C until dissection. c) Schematic representation showing pupae lined on double side sticky tape for light delivery. Adult brains showing expression of LOV-LexA (red in d and f) and LexAop-CsChrimson:Venus (Venus in d and f and dedicated image in e and g), as detected by native fluorescence of tdTomato and Venus, from pupae kept in the dark (d, e) or exposed to light (f, g) at 3-4 days APF, as shown in (b). h) Ratio of Venus signal intensity over DAPI signal intensity for fruþ neuronal cell bodies located in the anterior brain. Pupae exposed to pulses of blue light (N¼12) express the LexAop reporter Venus at higher levels compared to pupae kept in the dark (N ¼ 13), demonstrating that exposure to light leads to higher LOV-LexA transcriptional activity. Adult brains showing expression of LOV-LexA (red in i and k) and LexAop-CsChrimson:Venus (Venus, green in i and k and dedicated image in j and l), as detected by the native fluorescence of tdTomato and Venus, from w, LexAop-CsChrimson:Venus;þ/LC10a-SS1.AD; UAS-LOV-LexA/LC10a-SS1.DBD pupae kept in the dark (i, j) or exposed to light at 3-4 days APF (k, l), as shown in (b). m) Ratio of Venus signal intensity over DAPI signal intensity for LC10a neuronal cell bodies. Pupae exposed to pulses of blue light (N ¼ 12) express the LexAop reporter Venus at higher levels compared to pupae kept in the dark (N ¼ 6). n) Ratio of Venus over LOV-LexA native fluorescence from adult brains w, LexAop-CsChrimson:Venus;þ; UAS-LOV-LexA/fru-GAL4 exposed to 0, 2, 4, or 8 pulses of blue light as 2-3 days APF pupae (N ¼ 3, 4, 11, and 5, respectively). *** represents P-values <0.001, * represents P-values <0.05, n.s. represents P-values >0.05, 2-tailed Mann-Whitney tests.

LOV-LexA behavior in diverse neuronal types
(a) LC10s-SS2 and LC10a-SS1 drivers (Ribeiro et al. 2018) led to moderate expression of Venus in the dark if flies were raised at 25 C ( Supplementary Fig. 3, a-e), but not if flies were raised at 18 C in the dark (Supplementary Fig. 3, f-j). This indicates that the dark state of LOV-LexA is unstable in flies reared at 25 C. The leakiness of LOV-LexA at 25 C could arise from an elevated accessibility to the NLS at higher temperatures or increased LOV-LexA expression as previously observed in S2Rþ cells ( Supplementary  Fig. 1) and in other eLOV-based tools (Kim, Wang, et al. 2017;Kim et al. 2019). The stability of LOV-LexA in the dark was further tested with the panneuronal driver GMR57C10-GAL4 (Jenett et al. 2012). In many neuronal types, rearing flies at 18 C prevented the accumulation of the Venus reporter in flies expressing LOV-LexA panneuronally ( Supplementary Fig. 3, k-m). Several neuronal types, including neurons in the optic lobe, mushroom body, antennal lobe, and suboesophageal region, were an exception to this rule and presented high levels of Venus expression. Differences in expression strength across neuronal types represented in the GMR57C10-GAL4 expression pattern partially account for the observed variability in Venus expression in the dark.
On the other hand, differential expression pattern of genes involved in nucleocytoplasmic transport in different neuronal types could potentially underlie these discrepancies. Alpha importins function as adaptors that bind NLS peptides, bringing proteins with NLS in contact with b importins, which in turn mediate transport into the nucleus. The A importin A Karyopherin 4 (AKap4, CG10478) is highly expressed in Kenyon cells and other neuronal types ( Supplementary Fig. 3n) (Venken et al. 2011;Larkin et al. 2021). Expression of the A importin karyopherin A1 (Kap-A1, CG8548) and the b importins cadmus (cdm, CG7212) and Chromosome segregation 1 (Cse1, CG13281) are limited to a small number of neuronal types in the central brain ( Supplementary  Fig. 3, o-q). The presence of AKap4, or other importins, in certain neuronal types could potentially explain the selected leakiness of LOV-LexA dark state. To test this, we coexpressed Kap-A1 (Jang et al. 2015;Larkin et al. 2021) with LOV-LexA in LC10a-SS1 neurons in flies reared at 18 C in the dark. Coexpression of LOV-LexA with Kap-A1 in LC10a-SS1 neurons led to the expression of Venus reporter gene ( Supplementary Fig. 3, r-v), suggesting that increase in nucleocytoplasmic transport may facilitate the translocation of LOV-LexA to the nucleus, in the dark. We tested several GAL4 and split-GAL4 drivers in flies raised at 18 C and compared the expression of LOV-LexA and the reporter Venus. Like in other cell types, above certain levels of expression of LOV-LexA, the amount of Venus detected in neurons correlated with that of LOV-LexA ( Supplementary Fig. 3z). Together these observations suggest that the LOV-LexA tool has a stable dark state in drivers of weak to moderate expression strength, which constitute the majority of GAL4 and split-GAL4 lines available for genetic access to single neuronal types.

LOV-LexA mediates light-gated expression in neurons
The pupal case and the adult cuticle are tanned and block light, leading to the decreased exposure of internal tissues to light. To uncage the NLS in LOV-LexA expressed in pupal and adult brain, we used a 1-photon laser with 458-nm wavelength in a confocal microscope (see Materials and Methods and Supplementary Table  3). The driver fru-GAL4, a GAL4 knock-in in the locus of the gene fruitless (fru) (Gailey and Hall 1989;Stockinger et al. 2005), targets approximately 100 neuronal types, collectively called fru neurons, many of which were shown to regulate courtship behavior (among others, Cachero et al. 2010;Yu et al. 2010;Lu et al. 2012;Thistle et al. 2012;Toda et al. 2012;Bath et al. 2014;Inagaki et al. 2014;Ribeiro et al. 2018;McKellar et al. 2019). Expression is initiated in pupal development with low expression strength at late pupal stages. Pupae reared at 18 C and expressing LOV-LexA in fru neurons were exposed to a series of 4 preprogrammed light pulses (Supplementary Table 3), after which they were placed at 25 C for 2 days (Fig. 3, a-c). Expression of Venus was significantly increased in most fru neurons in pupae that were exposed to blue light (Fig. 3, d-h). Importantly, pupae kept in the dark displayed little or no expression of Venus (Fig. 3, d and e). Similar outcomes were observed for the LC10a-SS1 driver. Like fru-GAL4, LC10a-SS1 drives expression in pupal stages at low levels (Ribeiro et al. 2018, and data not shown). Exposure of pupae to 4 pulses of 1-photon laser 458 nm light spaced over 30 min (Supplementary Table 3) elicited light-dependent expression of Venus in LC10a neurons (Fig. 3, i-m). Delivery of 4-8 pulses of blue light, but not 2, proved to be sufficient for appreciable increase in Venus expression (Fig. 3n). Increase in expression of the reporter Venus in fruþ and LC10-group neurons exposed to light, and its absence in the same neurons kept in the dark, demonstrates that LOV-LexA gates expression with blue light in neurons in the pupal brain.
Precise control of the time of initiation of transgene expression has numerous advantages, including allowing for embryonic and pupal development to occur undisturbed in the absence of ectopic expression and for regulation of the level of transgene expressed. We measured the time it takes for LOV-LexA to drive the transcription of LexAop-controlled Venus after exposure to blue light. The head in pupae expressing LOV-LexA with fru-GAL4 was uncovered by removing the encapsulating pupal case and exposed to pulses of blue light (Supplementary Table 3 and Fig. 4a). The pupal brain was then imaged every hour for 12 h to determine the timing at which Venus starts to be expressed. Venus expression doubled 12 h after blue light pulse delivery (Fig. 4, b-d). Detection of expression with native protein fluorescence in adult brains was reliably observed 24 h after exposure to blue light during late pupal stages (Fig. 3, d-h), indicating that LOV-LexA lightgated expression takes 12-24 h to accumulate enough LexAop Venus reporter to be visualized with native levels.
Drivers that initiate expression at adult stages, such as LC10s-SS2, were exposed to light at adult stages. Adult flies expressing LOV-LexA in LC10-group neurons were immobilized with lowmelting wax on a custom-made opaque plastic coverslip with 300-400-mm holes (Fig. 4, e and f). The area of cuticle above the cells of interest was placed under one of the holes (Fig. 4f). Somata for the LC10-group neurons are located in the dorso-posterior side of the head, in an area bordering the rim of the retina. Immobilized flies with the cuticle covering somata of LC10-group neurons on one side of the adult head exposed were delivered 4-6 pulses of 485 nm 1-photon laser light over the course of 1 h (Supplementary Table 3). Detection of native fluorescence revealed the accumulation of Venus in LC10-group neurons exclusively on the side exposed to light (Fig. 4, g-j). Similar light deliveries to adult flies expressing LOV-LexA in LC10a (Fig. 4, k-n), LC6 (Fig. 4, o-r), and LC9 neurons (Fig. 4, s-v) resulted in unilateral Venus expression. Importantly, most flies prepared in this fashion showed unilateral expression in LC neurons (Fig. 4, j, n, r, and v), indicating that LOV-LexA allows for consistent genetic access to the same cell types within an expression pattern.

Discussion
We developed LOV-LexA, a light-gated expression system based on the photosensitive eLOV domain , and the modified transcription factor mLexA (Emelyanov and Parinov 2008;Pfeiffer et al. 2010). In the absence of light, LOV-LexA proteins reside in the cytoplasm of larval and adult cells. Delivery of blue light causes the LOV JA helix to uncage an NLS, which then mediates translocation of LOV-LexA to the nucleus. Once in the nucleus, LOV-LexA drives expression of transgenes under the control of LexAop sequences. The use of light as a trigger enables control of expression with high spatial and temporal resolution in live larvae and adult flies, making LOV-LexA an important addition to the Drosophila genetic toolbox that will expand the use of existent broad drivers as well as allow targeting subsets of cells within single tissues or cell types.
Several forms of LexA-transactivator chimeras are used in different animal models (Lai and Lee 2006;Emelyanov and Parinov 2008;Nonet 2020). Surprisingly, the ability to remain outside the nucleus in the dark and to elicit reporter expression upon light exposure varied widely among different combinations of mLexAtransactivator chimeras, eLOV-nls, and fluorescent tag. Replacing tdTomato with the FLAG tag in LOV-LexA, to make eLOV-nls-FLAG-mLexA:p65, leads to high levels of leakiness in the dark in S2Rþ cells (data not shown), suggesting that intraprotein interactions among the different components of LOV-LexA play an important role in stability of the JA helix in the dark (Kim, Wang, et al. 2017;Wang et al. 2017). Experiments in cell culture suggest that at high levels of expression, LOV-LexA proteins are more likely to translocate to the nucleus and drive expression of the LexAop reporter transgene. Rearing flies expressing LOV-LexA at 25 C similarly leads to unwanted expression of the LexAop reporter, imposing limits on the temperature used to raise fruit flies and the available driver lines. Further improvements of the eLOV domain have to be implemented to circumvent this limitation (Kim et al. 2019). Roughly 12-24 h separate the delivery of blue light and the accumulation of LexAop transgene expression in neurons, giving the fly time to recover from potential adverse effects of exposure to blue light, that include temporary blindness (Montell 2012). This temporal separation might preclude the use of transgenes encoding proteins with a short half-life. However, this time allows for other light and genetic manipulations to be performed on the same animal, without the need to perform all manipulations simultaneously on a tethered fly (Kim, Rouault, et al. 2017).
Replacing the transcription factor in LOV-LexA with QF2 (Riabinina and Potter 2016), testing NLS sequences of varied strengths, or using other LOV-based domains might improve the stability of LOV-LexA at higher temperatures and expression levels and change the time required for reporter expression. Addition of another protein domain that counterbalances nuclear import, such as a nuclear export signal (Niopek et al. 2014) or a membrane tethering domain (Kim, Wang, et al. 2017), might provide more stability to LOV-LexA. On the other hand, some neuronal types present LexAop reporter expression even if LOV-LexA is expressed at low levels. The uneven expression of importins across the fly brain, similar to what is observed in the mouse brain (Hosokawa et al. 2008), suggests that different neuronal types might express nucleocytoplasmic transport machinery to different extents. We predict that this variability is likely to influence how LOV-LexA functions across neuronal and cell types, making cells with high nucleocytoplasmic transport capabilities less suitable for light-gated expression with LOV-LexA.
Other light-gated expression systems have been developed in Drosophila, including the cryptochrome split-LexA, Photo-Gal4, and ShineGal4 (Chan et al. 2015; de Mena and Rincon-Limas 2020; di Pietro et al. 2021). We expressed the cryptochrome split-LexA with the same driver used to test LOV-LexA, LC10s-SS2 (Ribeiro et al. 2018), and found that cryptochrome split-LexA system is leaky in flies raised at 18 C and kept in the dark ( Supplementary  Fig. 3, x and y). Given that Photo-GAL4 relies on PhyB and requires addition of the chromophore PCB, normally absent in animal cells, it is currently limited to ex vivo studies (de Mena and Rincon-Limas 2020). The chromophore providing LOV with light sensitivity, flavin, exists in animal cells, making the LOV-LexA system solely dependent on the delivery of light. The limited number of enhancers driving ShineGal4, mostly targeting embryonic and pupal epithelia, prevents its widespread testing without re-cloning under other promoters. LOV-LexA is currently under the control of the UAS promoter and is one cross away from being tested with the myriad of GAL4 and split-GAL4 driver lines available.
There are thousands of enhancer-LexA or -GAL4 drivers targeting several cell types simultaneously (Jenett et al. 2012;Kockel et al. 2016;Robie et al. 2017;Tirian and Dickson 2017;Kockel et al. 2019). The LOV-LexA can be placed downstream of broadly expressed enhancers, to restrict transgene expression in the cell type of interest. Moreover, LOV-LexA downstream of an enhancer can be combined with GAL4 and QF binary expression systems, to genetically target two or more single neuronal types independently in the same animal, enabling several different experiments, including simultaneous monitoring of neuronal activity or determining dependency relationships among different neuronal types. Many neuronal types are composed of dozens of cells that are topographically organized to represent the visual field (Fischbach and Dittrich 1989;Otsuna and Ito 2006;Wu et al. 2016). Topographic organization of neuropiles processing sensory information is also observed in other animals, like the mouse superior colliculus, visual cortex, and for other sensory modalities (Engelmann et al. 2021;Huberman et al. 2008;Nassi and Callaway 2009;Petersen 2019). LOV-LexA is an ideal tool to test the role of topography, by providing consistent genetic access to the same subsets of somata within a single neuronal type, with little stochasticity. We demonstrate consistent targeting of several LC neurons unilaterally with LOV-LexA by targeting their somata. Applying this strategy to all visual projection neurons will elucidate how each contributes to guiding visual behavior.
Compared to D. melanogaster, many model organisms in which it is possible to create transgenics have smaller repertoires of enhancer driver lines that give access to different tissues and cell types. Implementing LOV-LexA in such model organisms will greatly amplify the number of specific cell types that can be genetically manipulated, expanding the landscape of possible experiments in emerging model organisms and the knowledge we can acquire from them.

Data availability
The DNA plasmid for LOV-LexA is deposited in DGRC (stock # 1583) and is available upon request. The D. melanogaster LOV-LexA flies were deposited in VDRC (stock # 311200) and are also available upon request. Data sets are available upon request.
Supplemental material is available at G3 online.
Imaging Facility, A. Fabritius for assistance with cloning, L. Groschner for assistance with mounting adult flies, and A.H. Ali, R.M. Vieira, and G. Ammer for critically reading the manuscript. The Bloomington Drosophila Stock Center, the Drosophila Genomics Resource Center, VDRC, and FlyBase were instrumental for this research project and countless others, and the authors wish that these institutions remain fully funded.

Funding
This work was supported by the Max Planck Gesellschaft (to AB, WE, and RK) and Rosa-Laura und Harmut Wekerle Foundation (to IMAR).

Conflicts of interest
None declared.
Literature cited