A light-triggered protein secretion system

The use of light to control basic cellular functions has transformed experimental biology. Some of the first approaches relied on photolabile small molecule analogues of second messengers, second messenger chelators, or neurotransmitters to control cellular physiology and signaling pathways with ultraviolet (UV) light (Kaplan and Somlyo, 1989; Callaway and Katz, 1993; Ellis-Davies and Kaplan, 1994; Ellis-Davies, 2007). These “caged” compounds have been invaluable for dissecting numerous molecular pathways governing cellular physiology with unprecedented spatial and temporal control.

More recently, exogenously expressed photoreceptors from plants have been used to control cellular biochemistry by conditionally gating protein–protein interactions with light (Tucker, 2012). This approach has emerged as a new and powerful way to control cellular processes on fast timescales with fine spatial precision without the need for small molecules. Some of the first studies describing engineered optical control of cellular functions used the plant photoreceptor phytochromeB (PhyB; Shimizu-Sato et al., 2002; Tyszkiewicz and Muir, 2008; Levskaya et al., 2009). PhyB binds to members of the phytochrome-interacting family (PIF) of basic helix-loop-helix transcription factors when photoexcited with red (660 nm) light (Ni et al., 1998; Leivar et al., 2008). Remarkably, PhyB/PIF interactions can be reversed by near-infrared (730 nm) excitation, allowing fast and local toggling of PIF binding (Ni et al., 1999; Levskaya et al., 2009). However, PhyB-based systems require addition of an exogenous phycocyanobilin chromophore that is not normally present in yeast, flies, worms, or mammals, making it more difficult to implement than more recently developed systems that are entirely genetically encoded. These systems rely on blue light photoreceptor cryptochrome2 (Cry2), which binds to cryptochrome-interacting basic-helix-loop-helix 1 (CIB1) in response to blue light, and the light, oxygen, voltage (LOV) domain photoreceptors, which undergo a large conformational change when photoexcited (Harper et al., 2003, 2004; Liu et al., 2008; Strickland et al., 2008, 2012; Wu et al., 2009; Yazawa et al., 2009; Kennedy et al., 2010; Lungu et al., 2012).

These tools have proven to be extremely powerful for controlling a wide range of cellular processes, including cell migration/morphology, cell cycle progression, transcription, and DNA recombination. However, some fundamental cellular processes, such as protein secretion, have been difficult to engineer for optical control using currently available tools. Traditional approaches for conditionally controlling protein secretion, including temperature-sensitive trafficking mutants and more recent chemical–genetic strategies have been indispensable for dissecting the mechanisms of mammalian secretory trafficking (Lodish and Weiss, 1979; Presley et al., 1997; Rivera et al., 2000; Boncompain et al., 2012). Although extremely powerful, these approaches require extended incubation at nonpermissive temperatures or addition of small molecules, which can be costly, can have potential off-target effects, lack spatial precision, and can complicate large-scale screening efforts. Thus, we sought to circumvent these issues by engineering a system for optical control of protein secretion to complement and extend the current secretory trafficking toolkit.

Here we implement UVR8, a recently described plant photoreceptor protein, as an optogenetic actuator module for protein secretion (Rizzini et al., 2011; Christie et al., 2012; Wu et al., 2012). UVR8 has many unique properties that lend itself to this application, including constitutive formation of photolabile homodimers, slow reversal kinetics, and a UV-B absorption profile, which enables multicolor imaging of all widely used fluorescent proteins without activating the photoreceptor. We show that UVR8 can be used to conditionally sequester secretory cargo in the ER and that light triggers robust forward trafficking to the plasma membrane. We use this novel tool in neurons, where we demonstrate spatially constrained forward trafficking of secretory cargo at dendritic branch points.

Because UVR8 has not been characterized for use as an optical tool in live mammalian cells, we evaluated the efficiency of UVR8 dimerization and its excitation properties in HEK293T cells using a plasma membrane recruitment assay that has been described in previous studies (Fig. 1 A; Inoue et al., 2005; Levskaya et al., 2009; Yazawa et al., 2009; Kennedy et al., 2010). We expressed UVR8 fused to EGFP containing a C-terminal CaaX prenylation motif (memGFP), which localizes to the plasma membrane, along with UVR8 fused to mCherry (UVR8-mCh). UVR8-mCh is normally cytosolic, but when coexpressed with UVR8-memGFP, a large fraction of the protein is localized to the plasma membrane, even after the cells have been exposed to room light, or laser illumination at 405, 488, or 561 nm used to visualize CFP, GFP, and mCh (Fig. 1 B).

To test if excitation energy at shorter wavelengths could dissociate the UVR8 dimerization pair in a mammalian system, we exposed cells expressing UVR8-memGFP and UVR8-mCh to a brief (3 s) pulse of UV-B (0.3 mW/cm², 312-nm narrow band) illumination. Upon UV-B illumination, UVR8-mCh immediately began dissociating from the plasma membrane and accumulating in the cytosol (Fig. 1, B–D; Video 1). To determine the minimal amount of UV-B required for dissociation, we measured the baseline cytoplasmic level of UVR8-mCh and then exposed cells to UV-B for durations ranging from 0.5 to 10 s (Fig. 1, E and F). After the first UV-B pulse, we incubated cells in darkness for 10 s to allow UVR8-mCh to dissociate, and then delivered a second, saturating dose of light to assay how much UVR8-mCh still remained associated with the plasma membrane. We found that 7 s of illumination (0.3 mW/cm², 312-nm narrow band) was sufficient to translocate 90% of plasma membrane–bound UVR8-mCh to the cytosol, with 50% dissociation occurring after 2–3 s (Fig. 1 F).

To test if UVR8, once monomerized, would reassociate over time, we dissociated UVR8-mCh from the plasma membrane with a 7-s UV-B exposure and then incubated the cells in the dark for varying times from 0.5 to 8 h. The cells were treated with 25 µM cycloheximide during the dark incubation period to prevent new protein synthesis and apparent recovery of plasma membrane–bound UVR8-mCh due to newly synthesized UVR8-memGFP and UVR8-mCh. We found that UVR8-mCh did not appreciably redimerize with UVR8-memGFP over the time scale of the experiment (8 h; Fig. S1).

A major concern is that UV treatment can cause serious damage to cells that can ultimately lead to cell death. Thus, it is important to assay toxicity at the levels of light required to dissociate UVR8 dimers. Although no immediate effects on cell morphology were observed in the minutes or hours after UV-B treatment, we determined long-term cell viability following UV-B treatments identical to those used in our plasma membrane dissociation studies (Fig. S2 A). We tested for UV-B toxicity in Cos7 cells, HEK293 cells, and primary cultures of hippocampal neurons treated with UV-B for varying durations and assayed for cell viability using two different standard assays: propidium iodide and annexin V staining (Fig. S2). For cells treated with 3 or 5 s of UV-B (illumination that triggers 55 and 70% dissociation, respectively), cell viability was identical to dark controls. We did observe limited toxicity in all cell types when we treated cells for longer durations. For example, COS7 cells exposed to UV-B for 7 and 10 s (illumination that triggers ∼90 and 99% dissociation, respectively) caused a 12 ± 9% and 16 ± 7% loss in cell viability, respectively (Fig. S2 A). However, neurons and HEK293 cells were more resistant to these UV-B doses (Fig. S2, C and D). It is of note that UVR8 has also been used in yeast two-hybrid assays, indicating that levels of UV-B required for UVR8 dissociation were also minimally toxic to yeast (Rizzini et al., 2011).

As a potential alternative to UV-B excitation, we also tested whether UVR8 could be dissociated with two-photon (2-p) illumination using a femtosecond pulsed laser source (Chameleon Ultra II; Coherent Inc.) tuned to the shortest wavelength possible (690 nm). This would offer many advantages over single photon excitation including better tissue penetration, less potential for toxicity, and a more precise focal excitation volume. However, 690-nm excitation failed to appreciably monomerize UVR8 at a range of laser intensities or pixel dwell times (Fig. S3). It may be possible that 2-p excitation at shorter wavelengths (∼600 nm) will trigger dissociation of UVR8 dimers, but this will require the use of more sophisticated optical hardware such as optical parametric oscillators to generate shorter wavelength pulsed excitation energy.

The ability to conditionally trigger protein trafficking through the cellular secretory pathway with light would be an extremely powerful experimental tool, but currently there are no available methods to accomplish this. Previous approaches have relied on either temperature-sensitive mutants of secreted proteins (Lodish and Weiss, 1979; Presley et al., 1997) or chemical–genetic strategies that use small molecules to release proteins sequestered in the ER (Rivera et al., 2000; Jaiswal et al., 2009; Al-Bassam et al., 2012; Boncompain et al., 2012). We based our strategy on a previously developed chemical–genetic strategy that uses an FK506-binding protein mutant (FKBP_F36M) that forms homodimers that can be dissociated with the rapamycin derivative AP21998 (Rollins et al., 2000). In this system, tandem copies of FKBP_F36M are fused to a secreted protein, which promotes the formation of high-order oligomers that are efficiently trapped in the ER. Disrupting these oligomers with AP21998 allows forward trafficking from the ER to the Golgi apparatus and ultimately to the plasma membrane (Rivera et al., 2000). Given its slow reversibility, fast dissociation kinetics, and UV excitation profile, we reasoned that UVR8 would be the ideal photoreceptor module to engineer a similar system that uses light instead of a small molecule to trigger synchronous secretory trafficking.

To test if UVR8 could be implemented as a photoswitch for secretory trafficking, we fused one, two, or three copies of UVR8 to the C-terminal intracellular domain of vesicular stomatitis virus glycoprotein (VSVG) tagged with YFP (Fig. 2 A; Fig. S4). VSVG was chosen because its trafficking properties are well documented (Nishimura and Balch, 1997; Presley et al., 1997; Cole et al., 1998; Toomre et al., 2000; Schmoranzer and Simon, 2003). As expected, we observed large clusters of YFP signal in the ER of cells expressing VSVG-YFP fused to multiple (2× or 3×) copies of UVR8, which were very similar in size and distribution to VSVG-YFP-4×FKBP_F36M (Fig. 2 B; Fig. S4 A). A brief (7 s) exposure to UV-B triggered massive redistribution of the YFP signal, which quickly diffused throughout the ER (Fig. 2 B; Videos 2 and 3). We compared VSVG-YFP-2×UVR8 with VSVG-YFP-4×FKBP_F36M and found that whereas a brief pulse (7 s) of UV-B completely dissolved VSVG-YFP-2×UVR8 clusters within 5 min, 1 µM AP21998 treatment incompletely disrupted VSVG-YFP-4×FKBP_F36M clusters 5 min after application (Fig. 2 B). Imaging for longer time periods demonstrated that VSVG-YFP-2×UVR8 efficiently trafficked to the Golgi apparatus within 30 min, similar to previous reports using either temperature-sensitive VSVG mutants or chemical–genetic strategies to release VSVG from the ER (Fig. 2 C; Hirschberg et al., 1998; Toomre et al., 2000; Schmoranzer and Simon, 2003; Jaiswal et al., 2009). After 120 min, cells displayed robust surface expression of VSVG-YFP-2×UVR8 and nearly complete “emptying” of the Golgi complex (Fig. 2, D and E). We estimated that 95 ± 2% of ER-retained VSVG-YFP-2×UVR8 is trafficked to the plasma membrane 2.5 h after UV-B treatment. Disappearance of signal from the Golgi complex was accompanied by formation of mobile post-Golgi carriers (Video 4). The size of these carriers was determined by a Gaussian fit of pixel intensities along a line drawn through the widest aspect of the vesicle. The full width at half maximal (FWHM) sizes ranged from ∼200 nm (near the diffraction limit of light) to ∼600 nm (Fig. 2 F), similar to a previous study (Hirschberg et al., 1998). Insertion of VSVG-YFP-2×UVR8 into the plasma membrane was confirmed by staining with an antibody that recognizes an extracellular VSVG epitope at various times after UV-B exposure and revealed a 5–10-fold increase in surface VSVG 150 min after UV-B treatment (Fig. 2, G and H). As an independent verification of trafficking to the plasma membrane, we performed total internal reflection fluorescence (TIRF) microscopy to directly visualize post-Golgi carriers fusing with the plasma membrane (Fig. 2 I; Videos 5 and 6). Cos7 cells were imaged in TIR mode 35–40 min after UV-B treatment to allow time for cargo to accumulate in post-Golgi carriers destined for the plasma membrane. Fig. 2 I and Videos 5 and 6 show mobile post-Golgi carriers fusing with the plasma membrane in UV-B–treated cells. We never observed fusion events in cells not treated with UV-B. Fusion events in UV-B–treated cells were consistent with previous studies using TIRF microscopy to observe VSVG plasma membrane insertion (Schmoranzer et al., 2000; Toomre et al., 2000).

Neurons are the largest and most morphologically complex cells in the body, with membrane surface areas ∼10,000 times that of typical epithelial cells (Horton and Ehlers, 2003b). Their immense size, polarity, and architectural complexity place extraordinary demands on the neuronal secretory pathway. However, it has been difficult to study the neuronal secretory pathway with traditional tools. For example, primary neuronal cultures respond poorly to the extended 40°C incubation required for ER accumulation of the temperature-sensitive ts045 variant of VSVG. Other methods hold promise, but require addition of expensive chemical ligands (Rivera et al., 2000; Al-Bassam et al., 2012) or specially formulated media lacking biotin, which is included in commercial neuronal media supplements (Boncompain et al., 2012). Our system overcomes these issues by circumventing the requirement for added chemicals or elevated temperatures.

Expression of VSVG-YFP-UVR8 in neurons resulted in the expected punctate expression pattern in the soma, but we also observed clusters throughout neuronal dendritic arbors due to the continuity of ER throughout the somatodendritic compartment (Fig. 3 A). Exposure of neurons to UV-B illumination dissolved these clusters and resulted in a uniform localization that extended throughout the soma and dendrites 5 min after UV-B treatment (Fig. 3 B). After 30 min, we observed that most VSVG-YFP-2×UVR8 signal had redistributed to perinuclear Golgi membranes in the soma (Fig. 3, A and B). We confirmed that VSVG-YFP-2×UVR8 ultimately accumulated at the neuronal plasma membrane by surface labeling neurons with an antibody against an extracellular VSVG epitope at various times after UV-B treatment (Fig. 3, B and C). In neurons we observed an ∼30-fold increase in VSVG surface label 150 min after UV-B.

After UV-B treatment, we often observed accumulation of VSVG cargo at dendritic branch points (Fig. 3 D; Video 7). These branch point clusters have been previously described as dendritic “Golgi outposts,” and are thought to represent platforms for local trafficking of dendritic secretory cargo (Horton and Ehlers, 2003a; Horton et al., 2005). However, given the somatodendritic continuity of ER, it is impossible to know whether accumulated cargo at branch point Golgi membranes originates from branch point ER (locally), or whether it is trafficked to branch points from distal sites, such as somatic ER. We thought this an important distinction because synaptic activity drives local protein synthesis in dendrites, which leads to incorporation of important neuronal proteins in the dendritic plasma membrane. Indeed, local protein synthesis in dendrites is thought to be critical for important forms of plasticity, but if newly synthesized secreted proteins quickly diffused away from sites of synthesis (in the ER membrane) before they could be captured for forward trafficking, any activity-induced spatial enrichment of newly synthesized proteins would be lost. A recent study demonstrated that the morphological complexity of the ER near dendritic branch points can act as a diffusion trap, presumably to restrict diffusion of ER proteins and to limit forward trafficking to nearby Golgi membranes (Cui-Wang et al., 2012).

To directly test whether ER-to-Golgi secretory trafficking could be spatially restricted in dendrites, we fused VSVG-2×UVR8 to the photoswitchable protein mEOS2 (Zhang et al., 2012), which can be photoconverted from green to red with 405-nm excitation (Fig. S5; Fig. 4 A). This allows us to locally “tag” ER-retained secretory cargo in different subcellular domains (e.g., dendrites, branch points, or soma) with focal 405-nm excitation before ER release with UV-B (Fig. 4 B). We first confirmed that photoswitching with focal 405-nm illumination did not itself trigger UVR8 dissociation (Fig. 4 A). After 405-nm photoconversion, VSVG-mEOS2-2×UVR8 clusters remained intact, indicating that UVR8 was not sensitive to focal photoswitching illumination. Conversely, we also confirmed that global UV-B illumination used to dissociate UVR8 does not also photoconvert mEOS from green to red, as this would complicate tracking locally photoswitched secretory cargo. After UV-B treatment, we observed rapid dissociation of VSVG-mEOS2-2×UVR8 clusters, but no increase in overall red signal, confirming that UV-B does not trigger mEOS2 photoswitching; however, we did observe limited photoactivation of the green form of mEOS2 with UV-B (Fig. 4 A; Fig. S5).

Dendritic branch points have been implicated as trafficking “hot spots” for newly synthesized proteins based on the prevalence of ribosomes, secretory organelles, and high ER surface area/morphological complexity at branch points compared with surrounding dendritic regions (Horton et al., 2005; Cui-Wang et al., 2012). However, no study has directly addressed whether ER cargo originating near branch points is trafficked to nearby Golgi with any preference over cargo originating at distant sites. To address this issue, we locally photoconverted small regions (∼1 µm in diameter) of ER-retained VSVG-mEOS2-2×UVR8 at dendritic branch points, or at dendritic sites >5 µm from branch points, and analyzed subsequent trafficking behavior after ER release (Fig. 4, B–E; Video 8). A significant fraction of cargo released from branch points redistributed to nearby Golgi membranes over the course of 35 min (Fig. 4 C; Video 8). Accumulation in Golgi outposts occurred with kinetics nearly identical to the redistribution of cargo to the somatic Golgi, indicating no major differences in the rates of secretion for dendritic versus somatic proteins. This experiment demonstrates that forward trafficking from ER to Golgi membranes can occur for a significant fraction of cargo before it diffuses away from branch points in the ER membrane. We also tracked VSVG-mEOS2-2×UVR8 released from dendritic sites >5 µm from branch points. Little or no local redistribution/accumulation of VSVG was observed (Fig. 4, D and E). Thus, dendritic branch points selectively retain a significant fraction of cargo released from nearby ER, further supporting a model for restricted local secretory trafficking at dendritic branch points. In these experiments we used VSVG, a nonneuronal cargo molecule, to investigate spatial determinants of neuronal protein secretion. Future experiments will test whether specific neuronal receptors are more or less efficiently trapped and trafficked at branch points as a mechanism for spatially controlling the composition of cell surface receptors at nearby dendritic branches and synapses.

Here we developed a novel optogenetic tool for light-triggered control of protein secretion in mammalian cells. We selected the plant photoreceptor UVR8 for this application because it has distinct properties that favor it over currently available optogenetic tools: It can be used in conjunction with all popular fluorescent and photoswitchable proteins, operates on a fundamentally different principle than existing tools (protein dissociation rather than association), and once photoactivated is essentially irreversible for hours.

As a first demonstration of the utility of this approach, we tracked secretory cargo originating in dendritic branches to determine whether dendritic ER proteins undergo local secretory trafficking. Our data suggest that cargo released from the ER near branch points can be locally trafficked to dendritic Golgi membranes at branch points. Further experiments using fluorescent reporters for cargo delivery to the cell surface (e.g., superecliptic phluorin) will determine whether subsequent post-Golgi trafficking is also locally restricted. In these experiments we used VSVG, a nonneuronal cargo molecule, to investigate the spatial determinants of neuronal protein secretion. Future experiments will test whether specific neuronal receptors, cell adhesion molecules, or ion channels are more or less efficiently trapped and trafficked at branch points as a mechanism for spatially controlling the molecular composition and fundamental properties of nearby dendrites and whether synaptic activity influences this process.

In addition to studying the secretory system in morphologically complex cells such as neurons, we anticipate that this approach will be broadly adopted in multiple fields. For example, we see great potential for its use in screening for proteins or drugs influencing secretory trafficking, or for conditional and local release of morphogens or growth factors from light-targeted cell populations in developing embryos. UVR8-based tools will expand and complement existing systems by offering a new spectral channel for protein control and unique properties that will enable optical manipulation of a new set of cellular processes.

UVR8 cDNA was obtained from the TAIR consortium for Arabidopsis. UVR8-memGFP and UVR8-mCh were generated by PCR amplifying UVR8 and cloning into GFP-CaaX (Kennedy et al., 2010) or mCherry vectors. All VSVG-YFP-UVR8 clones were generated by PCR amplifying UVR8 and using standard cloning techniques to generate the 1×, 2×, and 3× versions. mEOS2 was obtained from Addgene. VSVG-mEOS2-2×UVR8 was generated by exchanging YFP for mEOS2 using the VSVG-YFP-2×UVR8 construct. All constructs used CMV promoter to drive expression. TGN38 was a kind gift from K. Howell (University of Colorado School of Medicine, Aurora, CO).

HEK293T and COS7 cells were maintained in DMEM with 10% FBS. Cells were transfected using Lipofectamine 2000 according to the manufacturer’s instructions (Invitrogen). Cells were imaged 12–16 h after transfection. Longer transfection times and high expression levels led to higher basal levels (∼30% of total) of VSVG-YFP-2×/3×-UVR8 at the plasma membrane; thus, we confined our experiments to 12–16 h after transfection.

Primary hippocampal neurons were prepared from neonatal Sprague-Dawley rats. Hippocampi were dissected from the brains of postnatal day 0–2 rats and dissociated by papain digestion. Neurons were plated at 60,000–70,000 cells/ml in MEM, 10% FBS (Hyclone) containing penicillin/streptomycin on poly-d-lysine–coated glass coverslips. After 1 d the media was replaced with Neurobasal-A supplemented with B27 (Invitrogen). The neurons were then fed with Neurobasal-A, B27, and mitotic inhibitors (uridine + fluoro-deoxyuridine [Ur+FdUr]) by replacing half the media on day 4 or day 5 and then twice weekly. For Lipofectamine 2000 transfections, neurons were transfected on DIV 5–6 with 1 µg of plasmid/1 µl of Lipofectamine according to the manufacturer’s protocol and imaged on DIV 6–7. All animal procedures were performed in accordance with the University of Colorado School of Medicine guidelines.

Live-cell imaging was performed at 32°C on a microscope (model IX71; Olympus) equipped with a spinning disc scan head (Yokogawa Corporation of America) with a 60×/NA 1.4 objective. Excitation illumination was delivered from an AOTF controlled laser launch (Andor Technology) and images were collected on a 1024 × 1024 pixel EM-CCD camera (iXon; Andor Technology). Data acquisition and analysis were performed with MetaMorph (Molecular Devices) and ImageJ (National Institutes of Health). Some images were smoothed (averaging over 3 × 3 pixels) using ImageJ for display, but never before quantification. COS7 or HEK293T cells were plated on 18-mm coverslips and imaged in a buffer containing the following (mM): 10 Hepes, 10 d-glucose, 140 NaCl, 3 KCl, 1 MgCl₂, and 1 CaCl₂. To trigger forward trafficking of ER-retained VSVG, cells were treated with diffuse illumination from a UV-B (312 nm) source (model EB280C; Spectroline) positioned 7 cm above the sample for 7–10 s. The intensity of light at the sample was measured at 0.3 mW/cm²_. For light titration experiments, the UV-B excitation was shuttered with a Uniblitz shutter controlled with a Master-8 stimulator (A.M.P.I.). For local photoactivation of mEOS2, 405-nm laser illumination was steered using galvometric laser-scanning mirrors (FRAPPA; Andor Technology). Two-photon excitation was performed on a microscope (model LSM510; Carl Zeiss) equipped with a laser (Chameleon Ultra II; Coherent Inc.) tuned to 690 nm and 50–65% of maximum power with 250-µsec pixel dwell times. Higher excitation energy or slower pixel dwell times caused immediate membrane blebbing and cell death likely due to sample heating.

The size of post-Golgi carriers was determined by fitting the pixel intensities of a line drawn across the widest aspect of the vesicle with a Gaussian fit using Igor Pro software (Wavemetrics). The full width at half maximal value for post-Golgi carriers was extracted from the fits.

TIRF microscopy was performed on a microscope (model IX81; Olympus) equipped with a 10-mW argon laser fiber-coupled to a manual TIRFM illuminator. The temperature of the cells was controlled at 32°C. Images were recorded at 2 Hz on a CCD camera (Orca ER; Hamamatsu Photonics) using Slidebook software.

Surface VSVG was labeled with an antibody that recognizes an extracellular epitope (clone 8G5F11; Kerafast). After UV-B exposure, cells were incubated for various times before fixation for 10 min with 4% paraformaldehyde in PBS. Cells were washed in PBS and incubated with anti-VSVG overnight at a dilution of 1:1,000 in PBS with 5% BSA. Cells were washed and incubated with anti–mouse secondary antibody conjugated to Alexa Fluor 647.

Primary hippocampal neurons, HEK293 cells, or COS7 cells were exposed to varying durations of UV-B illumination (312 nm, 0.3 mW/cm²) from a distance of 7 cm. For primary cultures of hippocampal neurons, 100 mM glutamate was added to the media as a positive control for neuronal cell death. 10% methanol was used as a positive control for COS7 and HEK293 cells. A nontreated negative control was included. After UV treatment and a subsequent 14-h incubation period in the dark, COS7 cells were resuspended in PBS and incubated in the dark with 2 µg/ml propidium iodide (Sigma-Aldrich) or annexin V Alexa Fluor 488 conjugate (5%; Invitrogen). Cells were then transferred to a microscope slide and coverslipped. Neurons were not resuspended before labeling. The number of live cells versus dead cells was determined using a cell counter on a fluorescence microscope with excitation at 488 nm and emission at 516 nm for annexin V, and excitation at 516 nm and emission at 610 nm for propidium iodide.

HEK293 cells expressing UVR8-mCh and UVR8-memGFP were exposed to 7 s of UV-B light (312 nm, 0.3 mW/cm²) from a distance of 7 cm and then incubated for 2, 4, 6, and 8 h in the dark. Samples were treated with 25 µM cycloheximide to prevent new protein synthesis. After dark incubation, samples were fixed for 10 min with 4% paraformaldehyde, washed, and mounted using Prolong Gold (Invitrogen). Confocal microscopy was performed to determine UVR8-mCh membrane localization.

Fig. S1 demonstrates a lack of reversibility of UVR8 dissociation. Fig. S2 shows UVB toxicity. Fig. S3 shows that UVR8 cannot be monomerized with conventional two-photon excitation at 690 nm. Fig. S4 compares ER retention of VSVG fused to 1, 2, or 3 copies of UVR8. Fig. S5 shows that UVR8 can be used with the photoswitchable protein mEOS2. Video 1 shows the dissociation kinetics of UVR8 in response to UVB excitation. Video 2 demonstrates light-triggered dissociation of ER-retained VSVG-YFP-2×UVR8 clusters. Video 3 shows light-triggered VSVG-YFP-2×UVR8 trafficking from ER to Golgi membranes. Video 4 shows the formation and mobility of VSVG-YFP-2×UVR8 post-Golgi carriers. Video 5 shows TIRFM imaging of post-Golgi carriers. Video 6 shows TIRFM imaging of post-Golgi carriers fusing with the plasma membrane. Video 7 shows trafficking of VSVG-YFP-2×UVR8 at dendritic branch points. Video 8 shows local photoconversion and release of VSVG-mEOS2-2×UVR8 at branch point and non-branch point dendritic regions.

We would like to thank Chandra Tucker for critical discussions, Radu Moldovan and Greg Glazner of the Advanced Light Microscopy Core Facility at the University of Colorado School of Medicine, and Kathryn Howell and John Caldwell for providing constructs.

This work was supported by grants from the Brain and Behavior Research Foundation (M.J. Kennedy) and the Boettcher Foundation (M.J. Kennedy).