Engineering strategy and vector library for the rapid generation of modular light-controlled protein-protein interactions

Optogenetics enables the spatio-temporally precise control of cell and animal behaviour. Many optogenetic tools are driven by light-controlled protein-protein-interactions (PPIs) that are repurposed from natural light-sensitive domains (LSDs). Applying light-controlled PPI to new target proteins is challenging because it is difficult to predict whether one the many available LSDs will yield robust light regulation. As a consequence, fusion protein libraries need to be prepared and tested, but methods and platforms to facilitate this process are currently not available. Here, we developed a genetic engineering strategy and vector library for the rapid generation of light-controlled PPIs. The strategy permits fusing a target protein to LSDs efficiently and in two orientations. The public and expandable library contains 29 vectors with blue, green or red light-responsive LSDs many of which have been previously applied ex vivo and in vivo. We demonstrate the versatility of the approach and the necessity for sampling LSDs by generating light-activated caspase-9 (casp9) enzymes. Collectively, this work provides a new resource for optical regulation of a broad range of target proteins in cell and developmental biology.


INTRODUCTION
Optogenetics has revolutionized research in neuroscience, cell biology and developmental biology by allowing the 'remote control' of cell and animal behaviour with extraordinary precision (1)(2)(3)(4)(5). This precision is achieved by utilizing light as a stimulus that offers unique advantages over pharmacological and genetic manipulation strategies. For instance, light permits unparalleled control in time (e.g., to modulate animal behaviour acutely or to target selected developmental or disease stages; Figure 1A) and in space (e.g., to target selected compartments in a cell or selected cells in a tissue; Figure 1B). Also, light can be readily applied and withdrawn given a sufficiently transparent matrix. Finally, light-activated molecular tools can be paired with genetic targeting to allow an even higher level of precision for specific cell types, tissues or developmental stages (6)(7)(8)(9)(10).
Optogenetics first flourished in the hands of neuroscientists that utilized animal and microbial opsins to dissect neural circuits through the bidirectional control of neuronal bioelectrical activity (8,11). More recently and in cell types other than neurons, light control of gene regulation and cellular signalling, together with associated cell behaviours, has emerged (12,13). The optogenetic tools that can regulate cell bioelectricity are fundamentally different from those applied to control biochemical and enzymatic processes. In the former case, ion conducting opsins, such as channelrhodopsin or halorhodopsin, turn neurons on or off by changing their membrane potential through an intrinsic light-gated ion channel or pump activity (7,8,14). In the latter case, a wide range of cellular processes have been rendered light-inducible by using LSDs that do not harbour catalytic activity but regulate intraor intermolecular binding events ( Figure 1C).
The plethora of cellular processes governed by PPIs currently far exceeds the number of available optogenetic tools. This is in part because generating functional fusion proteins of LSDs and target proteins is a non-trivial task. For instance, multiple LSD genes need to be obtained and validated to find a suited domain. Furthermore, the location of the fusion site as well as the length of linkers can be critical parameters that determine fusion protein function (35,36). As a consequence of combinatorial complexity, many genetic constructs need to be generated and tested, and currently no methods or libraries are available to facilitate this process.
Here, we developed a genetic engineering strategy and a vector library for the rapid and modular generation of light-controlled PPIs. The engineering strategy can produce LSD-target protein fusions in several domain orientations and with linkers in a single cloning step (a universal restriction enzyme digest followed by ligation) using inexpensive and readily available materials. The publicly available vector library contains a collection of prominent LSDs that are responsive to blue, green or red light and have been applied in the past ex vivo and in vivo. The design of the strategy and library allows for easy expansion either with further LSDs, targeting sequences or markers. Using this resource, we generated light-activated casp9 enzymes.

Cassette design
Cassettes were introduced in pcDNA3.1-(Invitrogen/Life Technologies) to generate the vectors named pOVC1-3 (optogenetic vector core 1-3, Figure S4). A XmaI restriction site was removed from the backbone using site-directed mutagenesis (oligonucleotides 1 and 2, Table S2). Inverse polymerase chain reactions (PCR) (oligonucleotides 3 and 4, 5 and 6, and 7 and 8) were applied to remove the vector multiple cloning site and create ABC (pOVC1), ACB (pOVC2) and BAC (pOVC3) cassettes. In the inverse PCR procedure, PCR products were digested with DpnI, digested with EcoRI, XmaI or AgeI (NEB), respectively, ligated for 3 h at room temperature (RT) or overnight at 4°C using T4 ligase (Promega), and propagated in E.coli XL10 Gold cells (Agilent). All cassettes contain Kozak sequences, start codons and stop codons (for backbone ABC, the stop codon was introduced using sitedirected mutagenesis in a separate reaction (oligonucleotides 9 and 10)). For linker insertion, backbone pOVC1 was digested using EcoRI and BamHI. Linker fragments were generated by inverse PCR (oligonucleotides 57 and 58) or by annealing and phosphorylating single stranded oligonucleotides (59 to 64). All vector sequences (Table S3) were verified by Sanger sequencing (Micromon, Monash University) and deposited at Addgene.org.

LSD amplification and vector library
LSDs were amplified using PCR and oligonucleotides with AgeI and/or XmaI restriction site overhangs (oligonucleotides 11 to 34 and 45 to 52). Templates were previously described vectors from our laboratory or obtained from Addgene.org (Table   S1). In addition, gene fragments of AtCRY2-PHR, ScPH-1, AsLOV2-EcSsra, EcSSPB micro, AsLOV2-pep and HsPDZ1b were synthesized by a commercial supplier (Integrated DNA Technologies; Table S4). Restriction sites for AgeI and BamHI were removed from ScPH1-S and AtPHYB-S, respectively, as well as XmaI restriction sites from HsFKBP and AtCRY2-PHR using site directed mutagenesis (oligonucleotides 35 to 42). Site-directed mutagenesis was used to create EcSSPB nano (oligonucleotides 65 and 66). PCR products were digested with DpnI and with AgeI, XmaI or AgeI and XmaI depending on oligonucleotide overhangs. Backbone pOVC1 was digested with AgeI or XmaI for insertion into site A or C, respectively, and phosphatase treated.
Backbone and inserts were ligated either for 3 h at RT or overnight at 4°C using T4 ligase (Promega). All vector sequences (Table S5) were verified by Sanger sequencing (LGC Genomics) and deposited at Addgene.org. Note that for future subcloning of the generated genes, universal oligonucleotides can be designed that contain recognition sites for the enzymes AflII, ApaI, AscI, FseI, PacI, PspOMI or SbfI as these are not found in any of the genes.

Opto-casp9 constructs
The catalytic domain of casp9 (residues 135-416 of UniProt entry P55211) was synthesized (Integrated DNA Technologies; Table S4), amplified by PCR (oligonucleotides 43 and 44) and digested with XmaI. Vectors were digested with XmaI or AgeI, respectively, treated with phosphatase and gel purified. Backbone vectors and casp9 insert were ligated either for 3 h at RT or overnight at 4°C using T4 ligase.

Cell culture and transfection
HEK293 cells (Thermo Fisher Scientific; further authenticated by assessing cell morphology and growth rate) were cultured in mycoplasma-free Dulbecco's modified eagle medium (DMEM, Thermo Fisher Scientific) in a humidified incubator with 5% CO2 atmosphere at 37°C. Medium was supplemented with 10% FBS, 100 U/ml penicillin and 0.1 mg/ml streptomycin (Thermo Fisher Scientific). On the day after seeding, cells were transfected in DMEM supplemented with 5% FBS using polyethylenimine (Polysciences). Media was changed after 4 to 6 h and cells were stimulated with light starting 24 h after transfection for the durations specified below and at the intensities specified in the main text.   Figure 2B, sample numbers are 14, except for mock (26), HsFKBP (15) and DMSO (13). In Figure 2C, sample numbers are 16, except for mock (25), HsFKBP (19) and DMSO (12). In Figure 2D, sample numbers are 14, except for HsFKBP (26 and 12, dark and light). In Figure 2E, sample numbers are 16, except for mock (28) and HsFKBP (28 and 12, dark and light) and DMSO (12).

Efficient genetic engineering strategy
A major challenge in the optical control of PPIs is to achieve functional coupling of LSD oligomerization state changes to activity of target proteins. For most target proteins, it is initially unclear if a suited LSD can be identified and in what orientation LSDs are best attached because steric compatibility and effects on protein folding are difficult to predict. In the majority of previous studies, LSD-target protein fusions were constructed by inserting several LSD genes into vectors that contain the target protein ( Figure S1A, top). This approach requires selecting candidate LSDs, obtaining the corresponding genes from collaborators or commercial sources, validating LSD sequences, delineating domain boundaries and preparing amplicons that adapt each LSD to the vector ( Figure S1A, bottom). Furthermore, generation of both N-and Cterminal fusion proteins may require additional modification of the vector and/or amplicons. We propose an inverted strategy in which the target protein is inserted into a series of vectors that already contain LSDs ( Figure 1D; see below for a comprehensive LSD vector library). The advantages of this strategy are that only a single amplicon of a familiar and available target gene is required and that multiple LSD-target protein fusions can be generated in a simple standardized reaction that is easily parallelized. As a consequence, multiple time-consuming steps that require analysis of sequences and reagents specific to each LSD are not required and the workflow is greatly simplified ( Figure S1B).
To achieve this strategy, we designed a modular cloning cassette termed ABC that harbours three insertion sites (A, B and C; Figure 1E). Importantly, sites A and C contain recognition sequences for restriction enzymes that produce compatible cohesive overhangs (in both cases a CCGG overhang after AgeI or XmaI digestion at site A and C, respectively; Figure 1E). Consequently, a target protein amplicon flanked by either of these restriction sites in any combination can be inserted into site A as well as C and thus N-and C-terminally of a LSD (start and stop codons are already contained in the cassette). Site B contains recognition sequences for restriction enzymes of different families (EcoRI and BamHI) for incorporation of additional domains (e.g., fluorescent proteins) or epitopes. In order to provide additional flexibility, we engineered ABC vectors to include four different flexible or stiff linkers ( Figure 1F). We also prepared compatible ACB and BAC cassettes that permit insertion of flanking targeting sequences or fluorescent proteins in terminal B sites.

LSD vector library
Employing above genetic engineering strategy, we generated 29 vectors that contain one of 11 LSDs or one of five LSD binding partners inserted into site A and C ( Figure   1G) thaliana (21,22)). The library also includes binding partners for the heterodimerizing LOV domains, CRY and PHY, which are the minimal proteins EcSspB of E.coli with different affinities, HsPDZ1b of H. sapiens, AtCIB and AtPIF6 of A. thaliana (20,22,(39)(40)(41); Figure 1G) (sequence information and protein database identifiers can be found in Table S1). Collectively, these vectors provide coverage of methods to induce homodimerization, homooligomerization, heterodimerization with binding partners, or monomerization in response to different wavelengths of light. Many of these domains have been previously utilized ex vivo and in vivo but the library also contains less frequently applied domains (e.g., CrPH-LOV or RsLP-LOV). Vectors are available with all proteins inserted into the site A and separately the site C (i.e. N-terminal and C-terminal of the target protein insertion site), except in cases where N-terminal attachment is incompatible with protein function (AsPT1-LOV2 and AtPHYB). In the future, the library is expected to grow as its modular design allows direct expansion with additional LSDs (23,42).

Light-activated caspase-9
We employed the engineering strategy and vector library to develop a light-induced variant of casp9, an initiator caspase in apoptosis induction. The function of casp9 is mediated by homomeric assembly through the N-terminal caspase recruitment domain (CARD) (43), and casp9 has been rendered inducible by substitution of CARD with orthogonal homodimerization domains (44,45). This work demonstrated that dimerization by an N-terminal domain is sufficient for casp9 activation and resulted in a chemically-induced casp9 (iCasp9) that is employed as a cellular safety and suicide switch (46). To generate casp9 activated by blue light (Opto-casp9), we inserted a casp9 amplicon N-terminally and C-terminally of four LOV domains and AtCRY2-PHR ( Figure 2A). We focused on these blue light-sensitive domains because they represent commonly applied optogenetic tools and because their flavin co-factors are ubiquitously available in cells of virtually all organisms. As a control, we employed casp9 fused to an engineered chemical dimerization domain derived from human FK506 binding protein (HsFKBP) analogous to iCasp9. We first tested if these proteins exhibit constitutive activity (i.e. dark activity) by metabolically assessing the viability of human embryonic kidney 293 (HEK293) cells using the fluorescent viability dye resazurin ( Figure 2B,C). As constitutive activity was not observed, we next tested if these proteins can be used to induce cell death. To analyse cell death while controlling for transfection efficiency, we co-transfected cells with Opto-casp9 and a genetic viability reporter (Renilla luciferase under the control of a constitutive promoter). We chose a luciferase over a fluorescent protein as the reporter gene because of the high signal-to-noise ratio in luminescence detection and to avoid undesired excitation of the reporter by stimulation light. Twenty-four h after transfection, cells were stimulated for 7 h with blue light (l » 470 nm, intensity (I) = 200 µW/cm 2 ) in a tissue culture incubator equipped with light emitting diodes, and luminescence signals were measured subsequently. We found strongly reduced viability for cells that were transfected with casp9 fused to VfAU1-LOV or AtCRY2-PHR domains but not the other domains ( Figure 2D,E). To confirm the specificity of the observed effect using VfAU1-LOV-casp9 as an example, we demonstrated that apoptosis increases with increasing light dose (the half maximal effective light dose was 5.5 µW/cm 2 ; Figure S2). We further verified that light stimulation resulted in apoptosis using flow cytometry analysis with propidium iodide (PI) and Annexin markers ( Figure 3A). For VfAU1-LOV-casp9 and AtCRY2-PHR-casp9 but not for mock transfected cells we observed robust induction of apoptosis ( Figure 3B,C). This result demonstrates that by linking a casp9 amplicon to multiple LSDs functional Opto-casp9 enzymes could be quickly designed.

Specificity in light-induced PPIs
The modularity of the genetic engineering strategy provides the possibility to perform additional experiments, such as negative controls and immunodetection, that complement the efficient fusion protein generation demonstrated above. In optogenetics, negative controls typically consist of the application of light to naïve cells or to cells that were transfected with inactivated optogenetic tools (e.g., through lossof-function mutations). The latter control is required to obtain baseline signals and to ensure that overexpression of LSDs or target proteins does not alter cellular sensitivity.
The most commonly applied loss-of-function mutations for inactivation either target photochemically-active LSD residues or residues involved in light-induced conformational changes. However, targeting LSD photochemistry can be incomplete with persistent LSD activation through alternative reaction mechanisms or generation of chemical photoreaction side products (47,48). In addition, because of the diversity in the structures and activation mechanisms of LSDs, generalizable loss-of-function mutations do not exist, and thus negative controls cannot be studied under identical conditions. To address these limitations, we developed a universal inactivation strategy for light-controlled PPIs, which is based on testing the function of constructs in which the LSD and target protein have been uncoupled (e.g., uncoupling of VfAU1-LOV and casp9 should result in a loss in light activation). We realized this strategy by taking advantage of the availability of site B in all generated vectors. Into this site, we inserted a self-cleaving peptide sequence of porcine teschovirus-1 2A (P2A) that will effectively dissociate the two domains resulting in a loss of light sensitivity ( Figure 4A).
As expected for a P2A-modified Opto-casp9, we observed that light-induction of cell death by was abolished completely with self-cleavage effectively producing the same experiment outcome as removal of the catalytic activity of casp9 ( Figure 4B, Figure   S3). Immunoblotting against epitope tags that flanked the P2A sequence verified cleavage as we only detected the single LSD and casp9 domains but not the full protein ( Figure 4C). These results demonstrate a new control strategy that preserves target protein and LSD expression and LSD photochemistry taking advantage of linker and epitope incorporation into site B.

CONCLUSIONS
Optogenetics is one of few techniques that permits the regulation of cell behaviours with high precision in space and time. We developed a resource for the generation of light-induced PPIs and demonstrated its applicability by engineering Opto-casp9 enzymes. This resource will contribute to the broader use of optogenetics in cell and developmental biology and pave the way to novel optogenetics studies. For instance, experiments on the scale of entire families of LSDs or target proteins require efficient and modular genetic engineering approaches that are now within reach. Opto-casp9 enzymes may provide a test bed for optogenetic hardware development and testing, a process that entails optimization of light parameters (e.g. wavelength, intensity, duration) and culture conditions, because cell death can be assessed with different assays. Finally, the engineering strategy and empty cassettes may also be of use in areas other than optogenetics, such as for the rapid and modular design of fluorescent sensors and protein probes.