Methods to conditionally control protein function in cells can be valuable for researchers studying complex biological systems. Over the last several years, we and others have developed technologies to regulate proteins post-translationally using small molecule-ligands.¹ However, some experiments require tight spatial or temporal control of protein function, which makes the use of ligands sub-optimal as they can diffuse freely within and between cells and can be difficult to remove. The use of light as a perturbant has several advantages over small-molecule ligands as it can be delivered instantly and can be applied in a spatially restricted manner to cells or organisms.

Several methods employing light-responsive protein domains have recently been described, but significant trial-and-error can be required to make these systems useful for the particular proteins being studied. One strategy employs the Light-Oxygen-Voltage (LOV) domains that are found in plant photoreceptor proteins and respond to blue light via a flavin cofactor.² The LOV2 domain of phototropin 1 from Avena sativa (AsLOV2) possesses a C-terminal alpha helix that is tightly bound to the LOV core domain in the absence of light. Exposure to blue light induces formation of a flavin-cysteine adduct, resulting in unfolding of the helix. The AsLOV2 domain has been used to regulate the activity of proteins by sterically inhibiting interactions with an effector protein or by conformationally restricting a specific protein state.^3-5 These methods are suitable for the engineered proteins reported, but not generally applicable to any protein-of-interest. Engineered photosensory domains have also been used to establish light-regulated protein-protein interactions.^6-14 Light-induced translocation to the cell or nuclear membrane has been reported to regulate location-specific protein activity and light-induced gene expression, respectively. These light-dependent translocation strategies, however, generally require two or more genetic manipulations. We set out to develop a method for posttranslational control of protein levels in mammalian cells that (1) is generally useful for any protein, (2) makes use of a single regulatory domain, and (3) is regulated by non-toxic blue light, so that this technique does not require a small-molecule ligand that may have unintended effects.

We envisioned that a small peptide degron could be fused to the C-terminal alpha helix of the AsLOV2 domain to engineer a conditional Blue-Light Inducible Degradation (B-LID) domain. Renicke et al recently showed that such approach can be applied in Saccharomyces cerevisiae.¹⁵ We have previously shown that a small four-amino acid peptide degron, RRRG, fused to the C-terminus of a protein of interest results in fast proteasome-mediated degradation in mammalian cells.¹⁶ We reasoned that the C-terminal alpha helix would interact with the LOV core domain in the dark state, thus sequestering the degron away from cellular quality control proteins and rendering it cryptic. Exposure to blue light should cause the C-terminal helix to dissociate from the LOV core domain, thus revealing the degron and inducing degradation of the B-LID domain and its fusion protein through the processive activity of the proteasome (Figure 1a).

We began by using full length (residues 404-546) as well as truncated versions of the AsLOV2 domain of phototropin 1 with the RRRG degron fused to the C-terminus of the photo-responsive J-alpha helix [Supporting Information (SI) Figure S1]. These constructs were fused to the C-terminus of yellow fluorescent protein (YFP) to facilitate easy analysis of B-LID stability. NIH3T3 cells stably transduced with these constructs were analyzed by analytical flow cytometry. We used a blue LED light source to illuminate the cells. The four constructs tested (LOV5-8, SI Figure S1) displayed low fluorescence in both the dark and illuminated states, likely because the RRRG degron was too strong or insufficiently sequestered by the core LOV2 domain, leading us to investigate the use of the weaker RRRGN degron. Cells expressing YFP fused to LOV2 domains encoding the RRRGN degron displayed fluorescence when not illuminated, and a slight decrease in fluorescence was observed for cells that were exposed to blue light (465 nm) for two hours (LOV1-4, SI Figure S1).

In order to improve the dynamic range of the B-LID domain, we investigated the effects of mutations that are known to modulate the properties of the LOV domain. As the interaction of the C-terminal helix to the LOV2 core domain is in equilibrium with helix undocking, we aim at either stabilizing the conformation in the dark or in the lit state. We first introduced into all eight constructs a mutation (I532A) that results in enhanced helix binding to the core LOV2 domain with the goal of decreasing degron exposure, and thus decreasing degradation, in the unilluminated state.¹⁷

Additionally, another mutation (V416A) that lowers the rate of flavin deprotonation was explored to increase the time of degron exposure in the lit state.¹⁸ We tested these mutations alone or in combination by transducing cells with each construct and exposing them to blue light. Cells stably expressing these constructs displayed varying levels of light-regulated YFP expression (LOV9-32, SI Figure S1).

The LOV24 construct displayed the best combination of stability in the dark state and instability when illuminated with blue light. The LOV24 domain lacks the last three amino acids of the original AsLOV2 domain, it contains the stronger RRRG degron, and it contains the I532A mutation which results in stronger helix binding to the AsLOV2 core domain. We tested additional mutations that are known to increase the helix binding affinity (e.g., the G528A/N538E tandem mutant)¹⁷ but these did not improve the dynamic range of the LOV24 construct (SI Figure S2).

When illuminated with blue light for two hours NIH3T3 cells stably expressing the YFP-LOV24 fusion protein showed a 5-10 fold decrease in fluorescence as measured by flow cytometry and immunoblotting, whereas YFP alone was not affected by illumination (Figure 1b and SI Figure S3). Addition of the proteasome inhibitor MG132 prevented the light-dependent decrease in fluorescence (Figure 1b), indicating the involvement of the proteasome. The maximum amount of YFP-LOV24 degradation was observed approximately 90 minutes after starting illumination, and no further reduction in YFP levels was observed as illumination was continued (Figure 1c and 1d). Since conformational dynamics of the LOV2 domain is on the millisecond timescale¹⁹ it is likely that the capacity of the proteasome to degrade the large pool of substrate that is generated upon illumination is the rate limiting step. To test the ability of the LOV24 domain to confer light-dependent stability to other proteins we fused the LOV24 domain to a β-actin-mCherry fusion protein as well as to the mCherry fluorescent protein alone. Cells stably expressing these constructs were illuminated with blue light for two hours resulting in low fluorescence levels for both β-actin-mCherry-LOV24 (SI Figure S4) and the mCherry-LOV24 fusion as evidenced by fluorescence microscopy, flow cytometry and immunoblot (Figure 2a-c).

An important feature of light-regulated conditional protein alleles such as the B-LID domain is the ability to perturb these proteins in a temporary or spatially selective manner in living organisms. This can be particularly useful in transparent model organisms such as zebrafish or C. elegans. Illumination is instantaneous, which allows for rapid regulation of light-sensitive proteins. The use of small molecule ligands to regulate proteins may be troublesome in these organisms, as ligands may elicit unanticipated or undesired effects (e.g., toxicity) and may not be evenly distributed within the organism. We demonstrated the feasibility of using the B-LID domain in vivo by injecting mRNA encoding mCherry-LOV24 into zebrafish embryos. The embryos were then cultured with or without blue light illumination. Fluorescence was detectable 6 hours post-injection for embryos grown without illumination, and mCherry expression became more apparent after 24 hours (Figure 2d). Embryos continually illuminated with blue light displayed low mCherry fluorescence at all time points, and light-induced degradation of mCherry-LOV24 was confirmed by immunoblotting (Figure 2e). We next raised embryos in the dark for six hours, at which point illumination was started to promote degradation of the mCherry-LOV24 fusion protein (SI Figure S5). A significant decrease in fluorescence was observed compared to the embryos that were not illuminated. This confirms the potential of the B-LID domain and light to reversibly regulate a protein of interest in transparent organisms.

The spatial regulation of specific genes in zebrafish embryos have shown to be a promising approach to investigate the spatiotemporal differences in gene function.²⁰ Here, uncaging of a small molecule by a UV pulse generated the active molecules instantly allowing spatial activation. The disadvantage of the B-LID system may be the long-term illumination that is needed to ensure complete spatial degradation of the fusion protein. In some cases however, reversible gene regulation is desired which is difficult to achieve in life organisms with the uncaging strategy thereby making the B-LID strategy more appropriate.

This new method allows protein levels to be rapidly and reversibly controlled by light on a posttranslational level. The B-LID is simple to use through fusion to the 3’-end of any gene under study. Additionally, only one genetic manipulation is required, which makes this strategy attractive for use in organisms that are not easily amenable to high-efficiency gene targeting. We have demonstrated the utility of the B-LID domain in cultured cells and in zebrafish embryos, but we envision its use in other organisms.