Targeted and specific induction of cell death in an individual or groups of cells hold the potential for new insights into the response of tissues or organisms to different forms of death. Here, we report the development of optogenetically controlled cell death effectors (optoCDEs), a novel class of optogenetic tools that enables light-mediated induction of three types of programmed cell death (PCD)—apoptosis, pyroptosis, and necroptosis—using Arabidopsis thaliana photosensitive protein Cryptochrome-2. OptoCDEs enable a rapid and highly specific induction of PCD in human, mouse, and zebrafish cells and are suitable for a wide range of applications, such as sub-lethal cell death induction or precise elimination of single cells or cell populations in vitro and in vivo. As the proof-of-concept, we utilize optoCDEs to assess the differences in neighboring cell responses to apoptotic or necrotic PCD, revealing a new role for shingosine-1-phosphate signaling in regulating the efferocytosis of the apoptotic cell by epithelia.
Programmed cell death (PCD) controls the elimination of unnecessary, damaged, malignant, or infected cells during development and tissue homeostasis (Green, 2011). Apoptosis—the best-studied form of PCD—leads to the disintegration of dying cells into apoptotic bodies that maintain membrane integrity and are removed by phagocytes in an immunologically silent manner (Galluzzi et al., 2018). Mechanistically, apoptosis is initiated either by the extrinsic/death-receptor pathway or the intrinsic/mitochondrial pathway that controls the activation of apoptotic initiator and executioner caspases (caspase-8/-9, caspase-3/-7), a family of cysteine-aspartic proteases that cleave substrate proteins to promote cell death (Taylor et al., 2008). In contrast to apoptosis, programmed necrotic cell death, such as pyroptosis and necroptosis, results in cell lysis, which drives inflammation and immune responses (Galluzzi et al., 2018). Tumor necrosis factor receptor 1 (TNFR1) can initiate necroptosis if the activity of caspase-8 is inhibited by forming the necrosome, a signaling complex that contains the receptor-interacting kinase (RIPK)1 and RIPK3 (Vandenabeele et al., 2010). Within this complex, RIPK1 phosphorylates RIPK3, which in turn phosphorylates and activates the pseudokinase MLKL that translocates to the plasma membrane to form oligomeric ion channels that drive necroptosis. Pyroptosis is initiated by inflammasome complexes (Broz and Dixit, 2016), which assemble upon detection of pathogens or endogenous danger signals by cytosolic pattern recognition receptors, resulting in the activation of inflammatory caspases (e.g., caspase-1, -11, -4, and -5). These caspases process gasdermin D (GSDMD), thereby releasing the pore-forming GSDMDNT that permeabilizes the plasma membrane to trigger pyroptosis.
While the pathways and checkpoints that control PCD are well understood, comparably little is known about how distinct forms of PCD differ in their outcome for the dying cell, its neighbors, and the organism as a whole (Green, 2019). The effects of cell death are highly context-specific and can either induce or dampen immune responses, or induce cell proliferation and wound repair. It is generally assumed that necrotic cell death causes inflammation, while apoptosis is anti-inflammatory or even tolerogenic (Galluzzi et al., 2018). Yet studies that systematically compare and contrast different forms of PCD and the respective response of neighboring cells remain rare, particularly in an in vivo setting. Experimentally, such analyses are hindered by the low specificity of cell death-inducing stimuli, which affect both the targeted cell and its neighbors, the crosstalk between different PCD pathways, and the difficulty to induce death in a spatially and temporally controlled manner. Laser-induced ablation of single or multiple cells (Okano et al., 2020; Tirlapur et al., 2001; Uchugonova et al., 2008) has been used to overcome these obstacles, as it allows to study the signaling response of direct neighbors and dead cell extrusion. However, it is unclear what type of death is caused by laser ablation since dying cells often display features of both apoptosis (caspase activation) and necrosis (membrane permeabilization; Tirlapur et al., 2001). Another approach is to force-oligomerize cell death executors using fusions with a chemically-dimerizable domain (DmrB/FKBP; Oberst et al., 2010). Limits however are that reversion needs excess concentrations of washout ligands and the inability to selectively target and kill single cells in vitro or in vivo.
Here, we present a novel set of “clean cell death” tools for three major types of PCD—pyroptosis, necroptosis, and apoptosis—which are based on optogenetics (illumination with light), and hence named optogenetically controlled cell death effectors (optoCDEs). Light-mediated cell death induction offers several significant advantages: faster and easier signal delivery; ability to precisely control the intensity and duration of the cell death stimulus by varying illumination dose and duration; and the ability to restrict cell death induction to selected cells or tissues of interest. Different approaches exist to use optogenetics to activate proteins. Recently, light-sensitive protein domain (LOV2)-based tools to induce the separation of the catalytic subunits of caspase-3 and -7, and thus activation were reported (Mills et al., 2012; Smart et al., 2017). Our toolset on the other hand utilizes Cryptochrome 2 E490G (Cry2olig), an Arabidopsis thaliana–derived photosensitive protein, that undergoes rapid homooligomerization upon exposure to blue (450–488 nm) light (Taslimi et al., 2014). By fusing Cry2olig to the effector domains of human, mouse, or zebrafish inflammatory caspases, we designed light-activatable caspases (optoCaspases), which induce rapid and efficient pyroptosis within minutes post illumination. OptoCaspases are functional in various model systems, including multiple human and mouse cell lines, organotypic 3D cell cultures, and live zebrafish larvae, and can be utilized to precisely control, at a spatiotemporal level, the degree of caspase activation to drive sub-lethal pyroptosis activation or single-cell ablation. We extend this approach to induce apoptosis and necroptosis, by generating light-activated apoptotic effectors optoCaspase-8 and -9, and necroptotic effectors optoRIPK3 and optoMLKL. Using the optoCDEs, we then demonstrate how optogenetic induction of cell death can be used to study bystander cell responses and the extrusion of dying cells from epithelia, revealing fundamental differences in the fate of cells dying by apoptosis and programmed necrosis.
Design and characterization of optogenetically controlled human inflammatory caspases
In humans, pyroptosis is induced by inflammatory caspases-1, -4, or -5 that cleave the pore-forming cell death executor GSDMD (Fig. 1 A; Broz et al., 2020). To generate light-activatable human caspases (hereafter termed opto-hCaspases), we fused the mCherry-tagged A. thaliana photosensitive protein Cry2olig N-terminally to full-length or CARD-deficient human caspase-1, -4, or -5 (incl. splice variants of caspase-5; Fig. 1, B and C) and transiently expressed the constructs in GSDMD-transgenic (GSDMDtg) HEK293T cells (lack endogenous GSDMD) to assess their ability to induce pyroptosis upon blue light illumination. For all experiments where photoactivation was done by microscopy, we used 488 nm light at below 26.3 mW/cm2, as this allowed rapid clustering of Cry2olig in HEK293T cells (Fig. S1, A and B) without detectable phototoxicity. By contrast, opto-hCaspase-expressing (mCherry+) cells responded to blue light illumination by membrane blebbing, followed by cell swelling and nuclear rounding (Figs. 1, D and E, and S1 C; and Video 1), all being hallmarks of pyroptosis. Cells also rapidly lost membrane integrity (acquisition of the membrane-impermeable DNA dye CellTox Green), acquired Annexin-V staining (measures phosphatidylserine [PS] exposure), and progressively lost cytoplasmic mCherry signal (Figs. 1 E and S1 D; and Video 1). While all opto-hCaspase constructs induced death of more than 50% of mCherry+ cells within 30 min, full-length opto-hCaspase-1 and -5 were slightly less efficient. To minimize potential endogenous regulatory interactions via the CARD, we decided thus to use CARD-deficient opto-hCaspase-1 (aa 92–404), opto-hCaspase-4 (aa 92–377), and partially CARD-deficient opto-hCaspase-5 (aa 90–435) for all further experiments (of note, fully CARD-deficient opto-hCaspase-5 showed increased background cytotoxicity). The selected constructs induced death with comparable speed and reached similar levels of pyroptosis at later timepoints, with the majority of opto-hCaspase-expressing cells (i.e. mCherry-positive) dying within 10–15 min after illumination (Fig. 1, E and F).
Mutation of the catalytic cysteines in opto-hCaspase-1, -4, or -5 completely abrogated cell death induction upon illumination (Fig. 1 G), similarly to treatment with the pan-caspase inhibitor Z-VAD (Fig. 1 H), validating the specificity of the tools. Mutating D387 of Cry2olig, which strongly decreases its light-mediated clustering (Kennedy et al., 2010), also led to the reduction in pyroptosis levels (Fig. 1 G). Light-induced activation of opto-hCaspase-1, -4, or -5 in wild-type (WT) HEK293T cells, which cannot undergo pyroptosis due to the lack of endogenous GSDMD expression, did not induce membrane permeabilization and CellTox influx (Fig. S1, E and F). Instead, we observed a typical apoptotic morphology that correlated with the activation of a genetically encoded caspase-3/7 reporter VC3AI (Zhang et al., 2013; Fig. S1 G). This is in line with reports showing that caspase-1 activation in Gsdmd-deficient mouse macrophages engages rapid apoptosis via Bid cleavage (de Vasconcelos et al., 2020; Heilig et al., 2020; Tsuchiya et al., 2019). While mouse caspase-11 is not able to induce this type of apoptosis (Heilig et al., 2020), we found that opto-hCaspase-4 and -5 caused GSDMD-independent apoptosis, possibly due to their wider substrate profile (Bibo-Verdugo et al., 2020). Opto-hCaspase activation efficiently induced pyroptosis in commonly used cell lines, such as MCF7, HeLa, Caco-2, HT-29, and HaCaT (Figs. 1 I and S1 H). In summary, we demonstrate that fusion with the photoresponsive protein Cry2 can drive inflammatory caspase activation and that optogenetically activatable inflammatory caspases induce GSDMD-dependent pyroptosis across different cell types.
Inflammatory optoCaspases induce efficient pyroptosis and downstream substrate processing in macrophage-like cell lines
We next validated opto-hCaspases in PMA-differentiated U937 cells, as macrophages remain the most commonly used in vitro model to study pyroptosis. Blue light stimulation of U937 stably expressing optoCaspases induced rapid membrane permeabilization, Annexin-V acquisition, and pyroptotic morphology in the majority of mCherry-positive cells, while Cry2-expressing controls remained unaffected (Fig. 2, A and B). To compare the efficacy of optogenetically-induced pyroptosis to pyroptosis caused by regular inflammasome triggers, we assessed cell lysis, cytokine secretion, and substrate processing in cells stimulated by “classical” canonical and non-canonical inflammasome activation versus bulk cell blue light illumination in a LED light plate apparatus (Gerhardt et al., 2016). Nigericin (NLRP3 activator) induced around 20% of LDH release and IL-1β secretion in WT (non-transduced), mCherry-Cry2olig-, and opto-hCaspase-expressing macrophages (Fig. 2 C). By contrast, blue light illumination efficiently killed ∼50% of opto-hCaspase-1-expressing cells within 1 h but did not induce LDH or IL-1β release from WT or Cry2olig-expressing cells. Cell lysis correlated with the percentage of mCherry-positive cells in the population (Fig. S2, A and B), indicating that most opto-hCaspase-1-expressing cells had died. Immunoblot analysis confirmed opto-hCaspase-1 processing after blue light illumination, but not after Nigericin treatment (Fig. 2 D), validating that the fusion protein was functioning orthogonally to the endogenous inflammasome pathway. In line with the LDH and IL-1β data, opto-hCaspase-1 efficiently processed pro-IL-1β and GSDMD (Fig. 2 D). Consistent with the rapid opto-hCaspase-1 activation, we found that also ∼50% of opto-hCaspase-4 or -5-expressing cells died within 1–3 h of blue light illumination (Fig. 2 E), whereas activation of the non-canonical inflammasome by LPS transfection required 8 h to induce even moderate levels of cell death. Light-induced activation of opto-hCaspase-4/5 also resulted in caspase auto-processing and GSDMD cleavage (Fig. 2 F).
As mice are commonly used to study inflammasomes both in vitro and in vivo, we also generated optogenetically activatable mouse inflammatory caspases, e.g., opto-mCaspase-1 and opto-mCaspase-11 (Fig. S2 C). Like human optoCaspases, opto-mCaspase-1 and opto-mCaspase-11 induced high levels of pyroptosis in hGSDMDtg HEK293T cells (Fig. S2, D and E), and pyroptosis induction was completely abrogated by mutating the catalytic cysteines. Opto-mCaspase-1 and opto-mCaspase-11 activation induced pyroptotic morphology and membrane permeabilization (Annexin V and CellTox co-staining) in mouse macrophage-like RAW264.5 cells and primary bone marrow–derived macrophages (BMDMs; Fig. S2, F and G). Thus, optoCaspases efficiently function in human and mouse myeloid cell models and yield a downstream response comparable with or stronger than the regular inflammasome activators.
Sub-lethal activation of optoCaspases by limited illumination
A major advantage of optogenetics is the ability to precisely control the level of protein activation or localization by changing spatial and temporal illumination parameters (i.e., illumination intensity or duration, number of light pulses, or the size of the illuminated region). We thus next assessed how these parameters affected optogenetic pyroptosis induction. Gradual increase in 488 nm light intensity resulted in a concomitant increase in the speed by which cells underwent pyroptosis, as well as the total number of pyroptotic cells at 25 min after the start of illumination (Fig. 3 A). A similar dose-dependent response was observed in cells transiently illuminated with increasing numbers of light pulses at a single timepoint (Fig. 3 B), and in opto-hCaspase-1tg U937 macrophages in which increasing blue light intensity or illumination time increased the levels of pyroptosis induction (Fig. 3, C and D).
Interestingly, we noticed that upon illumination with a single pulse of moderate-intensity light, a fraction of cells displayed signs of early pyroptosis (e.g., membrane blebbing) and moderate DRAQ7 influx (DRAQ7low), but nevertheless survived and reverted to normal morphology within 30–40 min post illumination (Fig. 3 E and Video 2). By contrast, other cells showed either no membrane permeabilization (viable nonresponding cells, DRAQ7−) or complete permeabilization (fully pyroptotic cells, DRAQ7high), the latter being followed by the acquisition of pyroptotic morphology (Fig. S2 H). Since the DRAQ7low cells remained viable (at least up to 6 h after illumination), they were able to undergo normal cell division (with both daughter cells retaining slight DRAQ7 staining; Fig. S2 I). Limited DRAQ7 uptake suggests that the plasma membrane of these cells was temporarily permeabilized by GSDMD pores, but that they afterward repaired the damage as observed for necroptosis and pyroptosis previously (Gong et al., 2017; Rühl et al., 2018). Illumination with a second round of light that was 10 times more intense induced death in the majority of these recovered cells (Fig. 3, E and F), confirming that they retain the ability to transition into the lytic stage and that their initial survival was not due to intrinsic defects in pyroptosis execution.
To quantitatively assess the ability of these cells to survive the transient opto-hCaspase-1 activation, we determined the percentage of DRAQ7-negative (live), DRAQ7low (“sub-lethal”), and DRAQ7high (pyroptotic) cells. Around 40% of opto-hCaspase-1 cells responded to the first pulse by gaining nuclear DRAQ7 staining, and among them around 50% underwent pyroptosis while 50% survived this low-level illumination (Fig. 3 G). A second round of longer illumination induced permeabilization in 80% of the cells, and among them 75% cells underwent pyroptosis and 25% survived. The second laser pulse affected DRAQ7low cells more strongly than the DRAQ7− cells since more DRAQ7low cells underwent pyroptosis after illumination. Consistent with the amount of caspase activation determining whether a cell survived or died, we observed that the fraction of “sub-lethal/DRAQ7low” cells was more prominent under low stimulation conditions (1–5 light pulses), yet some were present even after high-intensity stimulation (30–50 pulses; Fig. 3 H). In line with this observation, opto-Caspase-1 expression levels also correlated with the probability to induce sub-lethal pyroptosis (Fig. 3, I and J).
In summary, caspase activation can lead to either sub-lethal or lethal GSDMD pore formation, as hypothesized before (Broz et al., 2020), and the optogenetic control of caspase activity yields a unique approach to study the features and mechanisms that allow cells to avoid or revert from cell death induction, such as the expression of pro-survival factors or plasma membrane repair (Gong et al., 2017; Rühl et al., 2018).
Optogenetic pyroptosis induction enables precise single-cell ablation in 2D and 3D cell culture
Laser-induced single cell ablation has been previously used to study the response of direct neighbors to dying epithelial cells (Kuipers et al., 2014). To test whether optoCaspase activation enabled precise single-cell pyroptosis induction and cell ablation in populations of closely attached epithelial cells, we transiently illuminated 10–15 µm2 regions of individual opto-hCaspase-1tg HaCaT cells within confluent monolayers in 15-min intervals (Fig. 4 A). Illuminated cells underwent pyroptosis within minutes (Fig. 4, A and B and Video 3), while neighboring cells remained viable and maintained membrane integrity (Fig. S3 A) despite similar opto-hCaspase-1 expression levels. The neighboring cells responded to the pyroptotic events by rapidly migrating toward the pyroptotic cell, extruding it from the monolayer and resealing the gap (Fig. 4, A–C).
We next tested optogenetic pyroptosis induction in opto-hCaspase-1tg Caco-2 cells cultured to form polarized acini-like structures (spheres). We selectively illuminated ROIs within the sphere wall (Fig. 4 D) and monitored cell death and neighboring cell response by time-lapse microscopy. As in 2D cell cultures, illumination of single cells within the sphere wall induced their pyroptosis, highlighted by membrane blebbing and DRAQ7 influx, and the extrusion of illuminated cells into the sphere lumen (apical side; Fig. 4 E and Video 4), followed by gap closure by neighboring cells. A defined luminal space was not required for extrusion to the apical side, since pyroptotic cells were still extruded toward the spheroid center in non-lumenized spheroids, causing the formation of a lumen around them (Fig. S3 B). When larger subpopulations of the cells were killed simultaneously (Fig. 4 F), pyroptotic cell extrusion and resealing still occurred, and involved a directed migration of the viable neighboring cells toward the lesion, which physically pushed out the dead cells into the lumen. Lumen size was a limiting factor as once the lumen was completely filled with dead cells, extrusion and gap closure could not proceed properly (Fig. 4 F), resulting in a breach in the basal membrane and the spilling of dead cells from the sphere.
In summary, these data demonstrate that optogenetic caspase activation can be used for a precise single-cell ablation in 2D and 3D cell culture models and highlight the immense potential of the new toolset for the study of bystander cell responses and cell extrusion in these models.
Optogenetic induction of apoptosis
To expand our toolset to optogenetic apoptosis induction (Fig. 5 A), we fused Cry2olig N-terminally to either full-length or ΔCARD/ΔDED hCaspase-8/-9 (Fig. 5 B). Illumination of opto-hCaspase-8- or opto-hCaspase-9-expressing HEK293T cells induced apoptotic features, such as cell shrinking, detachment, and membrane blebbing in the majority of mCherry-positive cells within 30 min (Figs. 5, B and C, and S3, C and D; and Video 5). Since the DEDs of opto-hCaspase-8 and to a lesser extent the CARD of opto-hCaspase-9 delayed apoptosis induction (Fig. 5 C), we used ΔCARD/ΔDED opto-hCaspase-8/-9 for further characterization.
Validation of opto-hCaspase-8/-9 showed rapid activation of a genetically encoded caspase-3/-7 activity reporter confirming efficient activation of executioner caspases (Fig. 5, D–F). Catalytic dead or dimerization-deficient mutants failed to induce cell death (Fig. 5 G), and the pan-caspase inhibitor Z-VAD abrogated opto-hCaspase-8/9-induced apoptosis (Fig. 5 H). Unexpectedly, we observed a strong variation in opto-hCaspase-8- and opto-hCaspase-9-induced apoptosis levels between different tested cell lines (Fig. 5 I), potentially due to variable levels of endogenous caspase inhibitors (cIAPs/XIAP) or incorrect folding of the construct. Consistent with the former, treatment with the SMAC mimetic AZD5582 facilitated apoptosis induction in at least some but not all cell lines (Fig. S3 E). In summary, optogenetically activatable caspase-8/-9 can induce apoptosis in a variety of cell lines and might provide a new approach to study endogenous regulatory mechanisms that control apoptosis execution.
Optogenetic activation of necroptosis by RIPK3 and MLKL
To complete the toolbox of optogenetic cell death inducers, we next constructed optogenetic inducers of necroptosis (Fig. 6 A) by C- or N-terminal fusing mCherry-tagged Cry2olig to WT or RIP homotypic interaction motif (RHIM)-deficient (QIG449–451AAA mutation) hRIPK3 or the N-terminal bundle and brace-protein kinase domain (NBB-PKD) of hMLKL (Fig. 6 B). Illumination of opto-hRIPK3 or opto-hMLKL-expressing HEK293T cells resulted in cell rounding and PS exposure (Annexin-V staining), indicating that cell death had been induced (Fig. 6, C and D). Among the four hRIPK3 constructs, the N-terminal Cry2 fusion to opto-hRIPK3ΔRHIM (in the following termed opto-hRIPK3) proved the most efficient in inducing cell death (Fig. 6 C), while the other constructs proved less efficient or led to dark-state protein aggregation and cell death in a significant fraction of mCherry-positive cells (Fig. S3, F–H).
Time-lapse microscopy of HEK293T cells expressing either opto-hRIPK3 (co-expressed with hMLKL) or opto-hMLKL revealed rapid induction of necrotic cell death upon blue light illumination, which was characterized by cell rounding, Annexin-V staining, and at later stages by CellTox acquisition (Fig. 6, D and E and Video 6). Also, opto-hRIPK3-induced necroptosis proceeded with slower kinetics than opto-hMLKL (Fig. 6 F). Opto-hRIPK3 did not induce any phenotypical changes in the absence of hMLKL co-expression (Fig. S3 I), confirming that morphological changes and cell death resulted from RIPK3-dependent MLKL phosphorylation. The mutation of hRIPK3 D142 or hMLKL L162/L165 significantly decreased the proportion of necroptotic cells upon illumination (Fig. 6 G), as did RIPK3 or MLKL inhibitors GSK’872 and necrosulfonamide (NSA; Fig. 6 H). Both constructs were highly active in a range of cell types commonly used for necroptosis studies (Fig. 6 I).
Interestingly, necroptosis proceeded through two distinct phases: (1) A “sub-lytic” phase of circa 1 h, during which cells displayed membrane blebs, acquired Annexin-V staining, and rounded up, but maintained membrane integrity, and (2) a lytic phase characterized by the sudden and complete rupture and disappearance of the cell membrane and rapid CellTox acquisition (Fig. 6, D, E, J, and K and Video 6). This differed strikingly from the pyroptotic cell, which acquired CellTox during blebbing and ballooning but never displayed a complete plasma membrane rupture and the disappearance of the plasma membrane. Thus during pyroptosis, membrane permeabilization is an early event, while it is a late event during necroptosis, and it is preceded by a sub-lytic phase during which MLKL channels do not yet reach the threshold to induce full plasma membrane rupture (Samson et al., 2020).
Optogenetic activation of zebrafish caspases allows spatial and temporal controlled induction of pyroptosis and apoptosis in vivo
Zebrafish (Danio rerio) embryos are a powerful model to visualize biological processes like cell differentiation and death in real-time and in vivo. Zebrafish feature homologs of mammalian apoptotic and inflammatory caspases, such as zf(zebrafish)Caspa which induces pyroptosis in zebrafish skin cells downstream of zfASC-dependent inflammasomes (Forn-Cuní et al., 2019) and zfCaspb that forms a non-canonical inflammasome together with zfNLRP3 (Li et al., 2020; Fig. S4 A). We designed optogenetically activatable opto-zfCaspa, opto-zfCaspb, and opto-zfCaspase-8 by replacing their PYD or DED by Cry2olig and evaluated these constructs in GSDMDtg or WT HEK293T cells (Figs. 7 A and S4 B). Activation of opto-zfCaspa/b in HEK293T cells induced rapid pyroptosis and CellTox influx, while opto-zfCaspase-8 activation caused a typical apoptotic morphology (Fig. S4, B–D), confirming both functionality and evolutionary conservation of substrate specificity of these proteins.
We next generated an inducible system for the expression of optogenetic constructs under the control of a heat shock responsive element (HSE; Bajoghli et al., 2004), and created stable zebrafish lines by tol2-based transgenesis. A cardiac myosin light chain 2 (cmlc2) promoter driving tagRFP expression in the heart was used to identify transgenic larvae (Fig. 7 B). Heat-shock-induced efficient opto-caspase expression, yielding a typical mosaic expression pattern in different tissues, without any detectable spontaneous cell death induction (Fig. 7 C). While skin cells (keratinocytes and basal cells) form inflammasomes and die in response, muscle cells do not express zfASC, inflammatory caspases, or gasdermins. The mosaic expression of opto-zfCaspases in the different tissues (depicted in Fig. 7 D) allowed us to compare the effect of different optoCDEs in inflammasome-responsive and unresponsive tissues. Blue light illumination of opto-zfCaspa-expressing keratinocytes (Fig. 7, D and E and Video 7) rapidly induced pyroptosis, as judged by the appearance of a characteristic pyroptotic morphology highlighted by shrinking and eventually membrane ballooning (yellow arrow). By contrast, opto-zfCaspase-8 activation in keratinocytes and muscle cells induced apoptosis, which was highlighted by the formation of apoptotic bodies (Fig. 7 F white arrows, and Video 8). While effects of opto-caspase activation in skin cells were visible within ∼40 min, it only became apparent in muscle cells after several hours (Fig. 7, E and F). Unexpectedly, opto-zfCaspb also induced apoptosis in keratinocytes (Fig. S4 E) despite the ability of Caspb to cleave hGSDMD in HEK293 cells. Both opto-zfCaspa as well as opto-zfCaspase-8 induced apoptotic cell death in muscle cells (Fig. 7, E and F; visible as a contraction without ballooning of the cell), likely due to the absence of gasdermins in these cells (Kuri et al., 2017). In line with the observations in the mammalian cells, induction of pyroptosis in zebrafish skin triggered the rapid extrusion of the pyroptotic cells from the monolayer, followed by acquisition (Fig. 7 G and Video 9). Finally, we also used a 2-photon laser to stimulate ROIs (Fig. 7 H, light blue squares), demonstrating that we can restrict the activation of cell death to single cells. In summary, optoCDEs allow both temporal and spatial activation of cell death within living tissue down to a single-cell level.
Distinct cell death modes induce differential fates of epithelial cells
To highlight the usefulness and versatility of the novel tools for the study of complex biological questions, we compared the fates of cells dying by lytic (pyroptosis or necroptosis) versus non-lytic (apoptotic) cell death within confluent epithelial monolayers. Previous studies reported that apoptosis is followed by the extrusion of dead cells from monolayers, but death was usually induced non-specifically by laser ablation that causes a mixed apoptotic/necrotic phenotype (Okano et al., 2020; Tirlapur et al., 2001; Uchugonova et al., 2008), and the fate of necroptotic/pyroptotic cells were never studied on a single cell level before.
To avoid confounding effects of low-level cell death activation in neighboring cells due to light diffusion, we used confluent mosaic monolayers of opto-hCaspase-1-, opto-hCaspase-8-, or opto-hMLKL-expressing Caco-2 cells co-cultured with 20–50-fold excess of WT cells (Fig. 8 A). Illumination of optoCDE-expressing cells initiated the respective cell death programs, which was followed by a rapid rearrangement of neighboring cells and the re-establishment of epithelial integrity (Fig. 8 B and Video 10). As observed before (Fig. 3), both pyroptotic and necroptotic cells were rapidly extruded within minutes after the appearance of the first morphological signs of cell death, Annexin-V staining and, in case of pyroptosis, CellTox influx (Figs. 8, B–E; and S4, F and G), with some Annexin-V-positive membrane remnants being left behind and taken up by the neighboring cells.
By contrast, the majority of apoptotic Caco-2 cells were retained within the monolayer, rapidly fragmented into apoptotic bodies and either partially or fully engulfed by their neighbors (Fig. 8, B–E). Apoptotic bodies persisted within neighbors for up to 12 h before being degraded (Fig. S4, H–J). Unexpectedly, the fragmentation of apoptotic Caco-2 cells was only observed in the presence of neighboring cells, as when isolated Caco-2 cells were induced to die by apoptosis they did not form apoptotic bodies, despite displaying other apoptotic features, i.e., cell shrinking and Annexin-V acquisition (Fig. S5, A–C). Apoptotic cell fragmentation and efferocytosis preceded PS exposure and were not blocked by incubation with Annexin-V (Fig. S5, A–D), suggesting that, unlike in the case of efferocytosis by professional phagocytes, PS sensing was not a major driver of apoptotic cell engulfment by epithelial cells. Importantly, WT cells that spontaneously underwent apoptosis were also retained, fragmented, and efferocytosed (Fig. S4 E), confirming that engulfment was not due to laser illumination or the lack of natural apoptotic signals.
Since a hallmark of programmed necrosis is the loss of membrane integrity, we investigated if extrusion of pyroptotic and necroptotic Caco-2 cells correlated with membrane permeabilization. While membrane permeabilization and CellTox influx always preceded extrusion of pyroptotic cells, the majority of necroptotic cells were fully extruded even before acquiring CellTox signal, and marginally faster than pyroptotic cells (20 vs. 26 min on average; Figs. 8, F–H, and S4 G), implying that, cell lysis cannot be the only factor determining cell extrusion. Apoptotic cells were also efferocytosed in zebrafish larvae, while optogenetic induction of pyroptosis resulted in cell extrusion (Figs. 7 G and 8, I–K). In summary, our data demonstrate that unlike unspecific induction of mixed apoptotic/necrotic cell death modes by laser ablation, the specific cell death induction using optoCDEs allows us to identify previously unknown differences in the ways the cells respond to apoptotic and necrotic cells.
Extrusion and uptake of dying cells require active cytoskeleton remodeling
Both efferocytosis and extrusion of dead cells involve extensive cytoskeletons and the formation of characteristic structures like phagocytic cups or contractile actomyosin rings (Green et al., 2016; Kuipers et al., 2014; Le et al., 2021; Rosenblatt et al., 2001) to visualize actin dynamics in the neighbors of apoptotic or pyroptotic/necroptotic cells; we thus used Lifeact-GFP Caco-2 cells. Resting neighbors showed an actin-rich cortex and dynamic randomly oriented lamellipodial protrusions at the basal side (Fig. S5 F). By contrast, neighbors involved in the extrusion of pyroptotic or necroptotic cells formed polarized lamellipodia at the basal cell surface (0 µm, asterisk) as well as contractile purse strings, visible as thick Lifeact-positive structures around the dying cell at the apico-lateral side (arrowheads; Fig. 9, A, B, and F), consistent with cytoskeletal structures previously implicated in the extrusion of laser-ablated dead cells (Duszyc et al., 2021; Le et al., 2021).
Apoptosis induction, on the other hand, led to a rapid polarized apical movement of the neighbors on top of the apoptotic cell, starting immediately after the appearance of the first signs of apoptosis. This was accompanied by the simultaneous engagement of basal lamellipodia (asterisk), which dynamically extended below the apoptotic cell, sealing the basal gap. Neighboring cells polymerized actin from the apico-lateral to the basal side, leading to the formation of structures resembling phagocytic cups all around the apoptotic cell (arrows), followed by the engulfment of apoptotic fragments (Fig. 9, C–E). Occasionally, we observed the formation contractile rings around apoptotic cells, which typically resulted in full or partial corpse extrusion and was associated with a lack of full apical closure or failed corpse engulfment (Figs. 9 F and S5, G and H).
To mechanistically assess the contribution of lamellipodia-based cell motility versus actomyosin contractility in extrusion or efferocytosis, we used the well-characterized Myosin II inhibitors, blebbistatin and ML-7; the ROCK inhibitors, Y-27632 and thiazovivin; the Rac1 inhibitor, NSC-23766; and the CDC42 inhibitor ML-141 (Fig. 9 G). Consistent with previous work on cell extrusion, both myosin II and ROCK inhibition significantly delayed necrotic cell extrusion and increased the fraction of retained necrotic corpses 1 h post-death induction (Figs. 9 H and S5 I). Blocking lamellipodia formation by Rac1 and CDC42 inhibitors only weakly affected necrotic cell extrusion. Blebbistatin and ML-7 treatment also affected apoptotic cell efferocytosis by reducing the number of apoptotic cell fragments and moderately increasing the fraction of fully or partially extruded apoptotic cells (Fig. 9, I and J). Together, these data suggest that in Caco-2 cells, actomyosin contractility plays a dominant role in both necrotic cell extrusion and apoptotic cell fragmentation.
Sphingosine-1-phosphate signaling is needed for efferocytosis of apoptotic cells in monolayers
The release of sphingosines-1-phosphate from apoptotic cells and its sensing through S1P receptor 2 (S1PR2) by neighboring cells has been proposed to be a major driver of apoptotic cell extrusion (Atieh et al., 2021; Gu et al., 2011; Santacreu et al., 2021; Fig. 10 A). We thus tested whether inhibition of this pathway also blocks pyroptotic or necroptotic cell extrusion or apoptotic cell efferocytosis. Although treatment with JTE-013 (S1PR2 inhibitor) or SKI-II (Sphingosine kinase 2 inhibitor) delayed the extrusion of necroptotic cells (Fig. 10, B and C) from epithelia, it had no significant effect on the extrusion of the pyroptotic cells, even though pyroptotic cells released the highest levels of S1P (Fig. 10 D). By contrast, both JTE-013 and SKI-II treatment strongly reduced the levels of apoptotic cell efferocytosis (highlighted by the reduced numbers of apoptotic cell bodies being formed). This loss in cell fragmentation correlated with the reduced uptake of apoptotic cells by neighbors and an increase in the number of extruded cells (Fig. 10, E–G and Video 11), suggesting that in the absence apoptotic cell engulfment, other apoptotic cell-derived signals may trigger cell extrusion. The phenotypes for necroptotic cell extrusion and apoptotic cell efferocytosis were recapitulated upon knocking down S1PR2 expression in the neighboring cells (Fig. 10, H and I).
As the finding that apoptotic cells can be efferocytosed in an S1P-dependent manner contradicts previous work that reported the extrusion of apoptotic MDCK and MCF10A cells (Gagliardi et al., 2018; Gu et al., 2011; Rosenblatt et al., 2001), we tested our optoCDE approach in these cell lines. Unlike in zebrafish keratinocytes or Caco-2 cells, we observed that both MCF10A monolayers extruded apoptotic, pyroptotic, and necroptotic cells and that S1P signaling was required for this process (Fig. 10, J and K). Similar data were obtained for MDCK cells (data not shown). These findings are in line with previous works and imply that the fate of a dying cell is highly cell-line specific. Nevertheless, S1P signaling appears to play an important role both for efferoctosis and extrusion of apoptotic cells.
Here, we report the development, optimization, and characterization of a new toolset of optogenetically controlled cell death effectors (optoCDEs) for the induction of three major PCD types—apoptosis, necroptosis, and pyroptosis—in human and mouse cells and zebrafish larvae. OptoCDEs allow the specific induction of these forms of cell death orthogonally to and with faster kinetics and higher efficiency than the endogenous cell death pathways, offering several advantages over other currently used methods of “clean” cell death induction such as chemical activators or laser ablation. Compared with forced dimerization and the activation of caspases (Gargett and Brown, 2014; Oberst et al., 2010; Straathof et al., 2005) and RIPK3/MLKL (Hu et al., 2021; Wang et al., 2014; Wu et al., 2014) by modified FKBP domains, optoCDE excel not only in their rapid kinetics of activation and the possibility to exactly titrate the amount of stimulus, but above all in their use for single-cell manipulation and imaging in 2D, 3D, or in vivo settings. Furthermore, as Cry2 is known to become inactive within minutes of ceasing illumination, they also offer the promise of fast inactivation of signal. As we show, focused laser illumination yields superior spatiotemporal resolution that allows to rapidly and selectively kill individual cells without directly affecting the neighbors or the organism, even if these also express optoCDEs. Although other laser-based methods (such as laser ablation, laser-induced DNA-damage, or photosensitizers, such as KillerRed) can induce cell death rapidly in selected cells and in vivo (Jewhurst et al., 2014; Riani et al., 2018; Teh et al., 2010; Williams et al., 2013; Xu and Chisholm, 2016), they often fail to elicit specific cell death modes, rather relying on less specific types of cellular damage (Wang et al., 2021). Laser ablation for example induces mixed apoptotic and necrotic phenotypes (Okano et al., 2020; Uchugonova et al., 2008), and thus cannot be used to study cellular or tissue responses to specific PCD types. By contrast, the optoCDE approach allows highly specific activation of selected PCD programs directly at the effector protein level, increasing the specificity of cell death programs and reducing the impact of endogenous regulation and inter-pathway crosstalk. In contrast to optogenetic strategies for apoptosis induction via clustering of death receptors (Kim et al., 2020), recruitment of Bax (Hughes et al., 2015) to mitochondria or allosteric non-proteolytic activation of effector caspases via interdomain linker extension (Mills et al., 2012; Smart et al., 2017) allows our system to utilize a similar molecular approach and illumination parameters for different death effectors, enabling the comparison of signaling events induced by different types of cell death both in vitro and in vivo. Finally, the Cry2-based activation of cell death could easily be expanded to other death pathways that involve oligomerization of signaling components, such as lysosomal cell death or autophagy-driven cell death, and recently identified light-cleavable proteins allow to extend this principle to cleavage-activated cell death executors, such as Bid and Gasdermin D (He et al., 2021; Lu et al., 2020 Preprint).
Although “classical” methods for PCD induction might take hours and require co-stimulation with multiple ligands, direct optogenetic activation of cell death on the effector level takes just a few minutes, allowing for mechanistic studies of cells undergoing cell death with extremely high temporal resolution and independently of confounding effects associated with regularly used activators (e.g., TNFa + SMAC or LPS + nigericin). Combined with advanced live imaging techniques, this approach could offer unprecedented insights into the early cellular and metabolic events occurring during PCD; and careful titration of illumination offers a new way to study cellular mechanisms used to avoid or revert from cell death (anastasis; Tang et al., 2012; Tang and Tang, 2018), or the effects of sub-lytic/sub-lethal activation of cell death effectors like caspases and gasdermins. Such sub-lethal cell death induction recently gained attention as low-level caspase activation results in “minority mitochondrial outer membrane permeabilization” (MiniMOMP) that can drive DNA damage and genome instability (Ichim et al., 2015) or is required for differentiation (Oberst et al., 2016). Finally, the most valuable application and the largest advantage over current methods lies in the use of optoCDEs in specific single-cell death induction in 2D epithelia, 3D organoids, and live animals, and in the study of how these multicellular systems respond to distinct forms of PCD.
To highlight this application of optoCDEs, we studied the differential fates of individual epithelial cells upon apoptosis, necroptosis, or pyroptosis induction. Dying cells release a number of “find-me” and “eat-me” signals that allow professional phagocytes to migrate toward and engulf dead cells (efferocytosis; Ravichandran, 2011). Efferocytosis by non-professional phagocytes, including epithelial cells, was reported as well (Davies et al., 2018), but is less well understood as most epithelia, such as the gut epithelium, are thought to react to cell death only by extruding dying cells (Ohsawa et al., 2018). For example, multiple in vitro studies show extrusion of dying cells in response to different proapoptotic triggers, such as etoposide treatment (Andrade and Rosenblatt, 2011; Teo et al., 2020), UV illumination (Andrade and Rosenblatt, 2011; Gu et al., 2011; Rosenblatt et al., 2001), or starvation (Gagliardi et al., 2021). While we found that apoptotic MCF10A and MDCKO cells are extruded as reported before (Andrade and Rosenblatt, 2011; Atieh et al., 2021; Duszyc et al., 2021; Eisenhoffer et al., 2012; Gu et al., 2011; Le et al., 2021; Rosenblatt et al., 2001; Teo et al., 2020), our results show that apoptosis induction in Caco-2 cells and zebrafish keratinocytes results in efferocytosis by their neighbors. This implies that the decision to extrude or engulf apoptotic cells is highly cell line–specific. By contrast, we found that necroptotic and pyroptotic cells are generally extruded from monolayers, in line with previous study. showing extrusion of epithelial cells after inflammasome activation (Rauch et al., 2017; Sellin et al., 2014) and that this extrusion involves the simultaneous lamellipodia-based neighbor motility and contractile purse-string actin structures, as reported previously for laser-ablated cells (Duszyc et al., 2021; Kuipers et al., 2014; Le et al., 2021). Additional studies will be required to determine which pathway controls efferocytosis or extrusion of epithelial cells, but S1P release and signaling appear to be key signaling pathways determining both cell fates.
In summary, optoCDEs provide a versatile and specific approach to investigate PCD pathways at a spatiotemporal level, not only in individual cells but also in their neighbors in a multicellular setting, and their application might allow the identification and study new biological processes during PCD and a better understanding of associated phenomena such as membrane repair, sub-lethal cell death, or cell extrusion and migration.
Materials and methods
The following constructs were obtained from Addgene: Cry2olig-mCherry (plasmid #60032), mouse RIPK3-GFP (plasmid #41382), and Lifeact-miRFP703 (plasmid #79993) and used as the cloning templates. The vector containing human caspase-5 was obtained from Sino Biological (cat. #HG11152-M). The genes encoding human caspase-1, -4, -5, -8 or -9, RIPK3 and MLKL, mouse caspase-1 and caspase-11 were amplified from human or mouse cDNA. Amplified fragments were fused C- or N-terminally to Cry2olig (as indicated in the text and figures) using overlap-extension PCR and subcloned into NheI/BstBI cloning sites of pLJM1 (plasmid 19319; Addgene) for constitutive expression or EcoRI site of pLVX (cat. #632164; TaKaRa) for doxycycline-inducible expression using InFusion HD cloning kit (TaKaRa). An additional GGGS linker was introduced between the Cry2olig and caspase/kinase domains or Cry2olig and mCherry to reduce the potential sterical interference between the domains. The exact construct design is indicated in the corresponding figures. For in vivo studies in zebrafish, the catalytic domains of zf caspa, caspb, and caspase-8 were fused N-terminally to Cry2olig and subcloned into the pTH2 vector backbone containing a bidirectional heat-shock element (HSE) as a promoter (Bajoghli et al., 2004), allowing heat-shock-induced expression of the cassette. The plasmids also contain the cmlc2:tagRFP as a transgenic marker and the insertion cassette is flanked by Tol2 sites for transgenesis (Dick et al., 2016; Fig. 7 A). All inserts were verified by sequencing to ensure the absence of unwanted mutations.
Cell lines and tissue culture
HaCaT cell line was a generous gift from Gian-Paolo Dotto (University of Lausanne, Lausanne, France), Caco-2 cell line was a gift of Shaynoor Dramsi (Institut Pasteur, Paris, France), and MCF7 cell line was a gift from Nouria Hernandez (University of Lausanne). HT-29 cell line was purchased from Sigma-Aldrich (cat. #91072201). The human GSDMD-transgenic cell line was described previously (Rühl et al., 2018). HEK293T, HeLa, Caco-2, HT-29, MCF7, 3T3, and RAW264.7 cells were maintained in DMEM (Gibco) supplemented with the 10% FCS (Bioconcept), 200 U/ml penicillin, and 200 μg/ml streptomycin (Bioconcept). HaCaT, U937, and THP-1 cells were maintained in RPMI medium supplemented with the 10% FCS and penicillin-streptomycin mix. Cells were maintained at 37°C, 5% CO2.
Generation of stable cell lines
All stable cell lines were generated using an optimized lentiviral transduction protocol. For the production of lentiviral particles, 1 × 106 HEK293T cells were transiently transfected using 1.9 μg of lentiviral expression vector (pLJM1 or pLVX), 1.9 μg of third-generation packaging vector PsPax2, 0.2 μg of VSVg, and 5 μl of JetPRIME transfection reagent. The supernatants were collected 24–48 h after transfection, filtered, and either used immediately or preserved at −80°C for long-term storage. For transduction, approximately 1 × 106 cells of each cell line were spin-infected for 1 h at 1.9 × 103g (3,000 rpm) in the presence of 10 μg/ml polybrene (Merck) to facilitate the infection. The virus-containing medium was replaced 24 h later, and the cells were left to recover for an additional 24–48 h in fresh medium. The stably transduced population of cells was selected using appropriate antibiotics (5 µg/ml puromycin or 50–100 μg/ml hygromycin Gold, both from Invivogen) for at least 5 d. To maintain stable expression, cell lines were subjected to the regular rounds of antibiotic treatment, and expression was monitored using fluorescence microscopy. Lifeact-GFP-expressing Caco-2 cells were additionally sorted based on GFP fluorescence to achieve a more uniform transgene expression level.
Generation and lentiviral transduction of primary murine bone marrow–derived macrophages
The primary bone marrow–derived macrophages were isolated from 6- to 8-wk-old wild-type C57BL/6 mice and differentiated in DMEM supplemented with 20% conditioned 3T3 cell supernatant (as a source of macrophage colony-stimulating factor M-CSF), 10% FCS, 10 mM Hepes (BioConcept), penicillin/streptomycin (Gibco), and non-essential aminoacids (Gibco). To express the opto-hCaspase-1/11, the macrophages were transduced with the lentiviral particles after 48 and 72 h of culture, as described previously (Dick et al., 2016), and kept in the dark afterward to avoid spontaneous construct activation. Photoactivation and imaging were performed on days 6–8.
3D cell culture
Single-cell Caco-2 suspension was mixed with 1.5–2 volumes of ice-cold Phenol Red Free Matrigel (Cornig) and a 40 μl drop of the resulting mixture was placed in the middle of each well of a preheated four-well μ-Slide (Ibidi), followed by immediate transfer at 37°C to induce Matrigel polymerization. After 10 min, 250–300 μl of complete medium was added to each well and the cells were maintained for 7–10 d to induce sphere formation. The medium was replaced every 2–3 d. To induce the transgene expression, the medium was supplemented with 1 μg/ml doxycycline approximately for 16 h before the experiment. For visualization of pyroptotic cells, the imaging medium was supplemented with DRAQ7 for at least 1 h before imaging to allow for dye diffusion in Matrigel.
To generate mosaic co-cultures for cell extrusion studies, the optoCDE-expressing Caco-2 cells were trypsinized, mixed 1:20 with the wild-type Lifeact-GFP-expressing “neighbor” cells, and plated on the tissue culture–treated eight-well µ-chambers (ibidi) at a concentration of 2 × 105 cells/well to form a confluent monolayer. OptoCDE expression was induced using 1 µg/ml doxycyclin treatment overnight and imaging/photostimulation were performed the following day. To accurately determine the dying cell and neighboring cell borders, the cell membranes were pre-stained with 5 ng/ml CellMask DeepRed (Invitrogen) for 60 min.
Live cell imaging
All in vitro imaging experiments were performed using LSM800 point-scanning confocal microscope (Zeiss) equipped with 20× air immersion objective (for sphere imaging) and 63 × oil immersion Plan Apo objective (for all other experiments), as well as 405, 488, 563, and 630 nm lasers. Temperature, CO2, and humidity were controlled throughout live imaging using an automated temperature control system and a gas mixer. For time-lapse imaging, Zeiss Definite Focus.2 system was used to ensure image stabilization. Images were acquired using Zen 2 software (Zeiss) at 16-bit depth, and the acquisition settings were kept constant for the same type of experiments to aid statistical analysis.
Photoactivation of optoCDE constructs
The photoactivation experiments were performed using 488 nm laser. The laser intensity was determined using FieldMaxII laser power meter (Coherent), and the light power density was calculated as follows (Dessauges et al., 2021 Preprint). Light power density (W/cm2) = Laser peak power [W]/effective focal spot area [cm2]−1. The illumination settings were kept constant for the experiments of the same type, unless indicated otherwise.
For transient stimulation (Figs. 3 and 4), cells were illuminated at the selected timepoint with the indicated number of pulses using Zen “experimental regions” and “bleaching” functions. For sustained activation, the stimulation was performed repeatedly every 15 s (unless otherwise indicated in figure legends) or every 90–180 s (cell extrusion experiments). Initial testing and validation of the optoCDE constructs was performed in HEK293T cells. Briefly, 4 × 104 cells were seeded in each well of collagen-coated eight-well µ-chamber (ibidi) and transfected with 150–300 ng of indicated construct using XtremeGene 9 transfection reagent (Roche) according to the manufacturer’s recommendation, and photostimulation/imaging was performed 24 h later. Prior to imaging, the cells were briefly washed and the cell culture medium was replaced with the pre-warmed optiMEM. To visualize PS exposure and membrane permeabilization, the imaging medium was supplemented with CellTox Green (50,000 × dilution; Promega), 1 µM DRAQ7, and/or 1 µg/ml PacificBlue-conjugated Annexin V (BioLegend). For cell death induction in transgenic cell lines, the cells were plated at a concentration of 1.5–2.0 × 105 cells/well ∼24 h before the experiment, and optoCDE expression was induced by overnight treatment with 1 µg/ml doxycycline. To avoid cell death due to spontaneous construct activation, cells were protected from light following transfection or expression induction, and all manipulations were performed under dim light.
Activation and live imaging of optogenetic caspase variants in zebrafish
Selected larvae were heat-shocked at 2.5 dpf by incubation at 39°C for 30 min and kept under light-shielding conditions. For imaging, larvae were anesthetized by adding 40 µg/ml ethyl m-aminobenzoate methanesulfonate (tricaine; Sigma-Aldrich) into the medium. They were then mounted in 1% low-melting-point agarose (PEQ LAB Biotechnologie) on glass-bottom dishes (MaTek). A Zeiss Biosystems 780 inverted confocal microscope was used for live imaging at RT. For skin and muscles cells, we used a 40× water objective (LD C-Apochromat 40×/1.1 W Corr M27 or C-Apochromat 40×/1.2 W Corr M27; Zeiss Bio-systems). For whole larvae and overnight imaging, we acquired tile scans using a 20×air objective (Plan-Apochromat; Zeiss Biosystems) and stitching. The 488-laser was used at a power between 3% and 5% to induce optogenetic activation and excitation of GFP. The 2-photon laser was used at 950 nm wavelength and 20% laser intensity. For visualization of pyroptotic cells, the anesthetizing medium and mounting agarose solution was supplemented with DRAQ7 (1:100).
Phototoxicity and Cry2olig-induced cytotoxicity assessment
To ensure the absence of phototoxicity and Cry2olig-induced toxicity, the cells were transiently transfected with Cry2olig-mCherry plasmid and whole-imaging field was repeatedly scanned with 488 nm laser of varying intensity (0.5–4.5% laser power, 4.8–40.9 mW/cm2) every 15 s for at least 1 h. The cell viability was assessed at the end timepoint based on cell morphology and the presence of Annexin V and CellTox staining. Based on these data, the illumination intensity range between 4.8 and 25 mW/cm2 was defined as non-toxic for the cells and used for all following experiments. To assess Cry2olig-induced toxicity, the percentage of mCherry-positive cells displaying altered morphology, Annexin V staining, or CellTox influx was compared with the non-transfected (mCherry-negative) cells in the same field of view.
Generation and maintenance of Zebrafish strains
Experimental animals were cared for in accordance with EMBL guidelines and regulations and according to standard procedures (Westerfield, 2000). The following transgenic lines were created by co-injecting embryos at the one-cell stage with Opto-caspase plasmids with transposase mRNA (100 ng/μl). Transgenic larvae were selected based on heart specific expression of tagRFP under the cardiac myosin light chain (cmlc2) promoter. The following lines were generated: tg(mCherry-Cry2olig-caspa), tg(mCherry-Cry2olig-caspb), and tg(mCherry-Cry2olig-zfCaspase-8). Additionally, we used the asc:asc-gfp (Forn-Cuní et al., 2019) for skin cell labeling. For visualizing cell boundaries the Opto-lines were crossed with tg(Krt4:AKT-PH-GFP; Gault et al., 2014). To follow actin dynamics in basal retaining cells, we used the tg(krt4:Gal4) line (Wada et al., 2013) crossed to tg(6xUAS:mneonGreen-UtrCH) generated in the lab of D. Gilmour by Jonas Hartmann. To inhibit pigmentation in larvae for imaging, we treated embryos with 0.2 mM 1-phenyl-2-thiourea (PTU; Sigma-Aldrich) in E3 medium.
The following inhibitors were used in the study: 20 μM Z-VAD-FMK (Invivogen), 1 μM GSK’872 (Selleck Chemicals), 5 μM Necrosulfonamide (Calbiochem), 50 µM blebbistatin (Sigma-Aldrich), 30 µM Y-27632 (StemCell), 15 µM thiazovivin (StemCell), 50 µM ML-141 (Sigma-Aldrich), 30 µM ML-7 (Tocris), 200 µM NSC-23766 (Tocris), 20 µM JTE-013 (Sigma-Aldrich), and 30 µM SKI-II (Tocris). All inhibitors were dissolved in DMSO and diluted to 10× concentration in OptiMEM, then added directly to the cells to achieve 1× concentration at least 1 h before imaging.
Canonical and non-canonical inflammasome activation in U937 cells
For canonical inflammasome activation, PMA-differentiated U937 cells grown in 24-well plates were primed with LPS for 2 h, followed by the treatment with the 15 μM Nigericin for 3 h to induce NLRP3 inflammasome activation. For non-canonical inflammasome activation, the cells were primed with human IFNγ (10 ng/ml) overnight and LPS transfection was performed as described previously (Heilig et al., 2020). Briefly, 15 μg of smooth LPS from Escherichia coli O111:B4 (Invivogen) was diluted in 200 μl of OptiMEM and incubated with Lipofectamine 2000 (5 μl per well) for 20 min at room temperature. Transfection complexes were then added directly on top of the cells in 200 μl of OptiMEM medium. To facilitate the LPS uptake, the cells were centrifuged at 500 g for 15 min, followed by incubation at 37°C for 8 h.
Optogenetic cell death induction in cell populations
The U937 cell were seeded onto black tissue culture–treated 24-well plates (4titude) at a concentration of 4 × 105 cells/well. The differentiation toward the macrophage-like phenotype was induced using PMA treatment (5 ng/ml) for 24 h, after which the cells were washed once and left to recover for additional 24 h prior to induction of optoCDE expression. The HaCaT cells were seeded at a concentration of 4 × 105 cells/well ∼24 h before an experiment. Expression of the constructs was induced in both cell lines by treatment with the 1 μg/ml doxycycline overnight. To induce cell death, each of the wells was illuminated with blue-light LEDs (450 nm) using a custom-built Light Plate Apparatus device (Gerhardt et al., 2016). To manipulate light intensity and illumination duration, custom illumination programs were created using Iris software (http://taborlab.github.io/Iris/).
U937 were seeded at a density of 4 × 105 cells/well and stimulated as indicated in the figure legends. After stimulation, the cells were lysed using pre-heated NuPage LDS sample buffer (Thermo Fisher Scientific) supplemented with 66 mM tris-Cl (pH 7.4), 2% SDS, and 10 mM dithiothreitol (DTT). Cell supernatants were also collected, precipitated using methanol and chloroform, and combined with cell lysates. Proteins were separated using 12% polyacrylamide gels and transferred onto a PVDF membrane using Trans-Blot Turbo (Bio-Rad). The following primary antibodies were used: rabbit anti-cleaved IL-1β (83186, CST; 1:1,000), mouse anti-IL-1β (12242, CST; 1:1,000), rabbit anti-GSDMD (ab210070; 1:1,000; Abcam), rabbit anti-cleaved N-terminal GSDMD (ab215203; 1:1,000; Abcam), mouse anti-caspase-1 (clone Bally-1 AG-20B-0048-C100; 1:1,000; AdipoGen), mouse anti-mCherry (ab125096; 1:2,000; Abcam), and HRP-conjugated mouse anti-tubulin (ab40742; 1:5,000; Abcam). Secondary antibodies conjugated to HRP (1:10,000; Southern Biotech) were used for the chemiluminescent detection.
Evaluation of cell lysis and IL-1β release
Cell lysis was assessed by measuring lactate dehydrogenase (LDH) activity in cell supernatants using the LDH cytotoxicity detection kit (TaKaRa). To obtain the positive control, the cells were fully lysed using 1% Tryton. To normalize for spontaneous cell lysis, the percentage of cell death was calculated as follows: (LDHsample − LDHnegative control)/(LDHpositive control − LDHnegative control) × 100. The level of IL-1β in cell culture supernatants was measured using ELISA (R&D Systems) according to the manufacturer’s protocol.
Image analysis of in vitro data
All image analyses were performed using FiJi (https://imagej.net/Fiji) and Zen 2 (Zeiss) software. To aid visualization, the brightness and contrast of the representative images of the same panels were adjusted and set to similar values, and illumination correction for the DIC images was performed using the Bandpass filter option (FiJi). For Figs. 1, 2, 4, and 6, the quantification of CellTox-positive and Annexin-positive cells at selected timepoints was performed manually due to the detachment and loss of late pyroptotic and necroptotic cells from the imaging plane. Apoptotic cells were defined based on morphology (cell shrinking and blebbing) and Annexin V staining (Figs. 1, 4, and S4). Quantification of DRAQ7-positive nuclei per field (Fig. 3, B and C) was performed automatically using a particle analysis tool and custom-written FiJi Macro, and the total number of DRAQ7+ nuclei per frame was divided by the total number of the mCherry-positive cells. The quantification of single-cell DRAQ7 intensities (Figs. 3 and 7) was performed by manually segmenting cell area and measuring the intensity density of selected regions over time, and intensity was normalized to the region intensity at t = 0. Pyroptotic and necroptotic cell area was quantified by the manual segmentation and tracking of dying cell borders based on CellMask and mCherry signals and additionally validated using DIC channel. The extrusion time for Figs. 8, 9, and 10 was defined based on complete closure of the area previously occupied by the mCherry-positive cell, and for all experiments the t = 0 was defined as the beginning of 488 laser stimulation. Pyroptotic and necroptotic cells were quantified as “extruded,” when the cell body and nucleus were fully removed from the plane of imaging and the remaining gap was completely closed by the neighbors (determined by membrane staining) and “retained” if the nucleus remained in the imaging plane at the end of the experiment. Apoptotic cells were quantified as “fully extruded,” if the majority (over 80%) of apoptotic bodies were extruded by the end of the experiment, “retained” if less than 20% of the cell body was extruded, and “partially extruded,” if at least 20%, but not more than 80% of cell body remained within the monolayer.
Quantification of apoptotic and pyroptotic events in vivo
Quantification of in vivo optoCaspase induced apoptotic and pyroptotic events in zebrafish keratinocytes was performed manually in several independent experiments for each of the three lines (tg(mCherry-Cry2olig-caspa), tg(mCherry-Cry2olig-caspb), and tg(mCherry-Cry2olig-zf_caspase-8), and cells were classified as pyroptotic or apoptotic based on the appearance of dead cells/cell debris. For the quantification of epithelial closure of either retention or extrusion of cells, we counted the time starting at the frame before cell death–related cell shape changes that occurred to the timepoint at which surrounding cells have closed the gap completely at the apical site.
All quantitative data analysis was performed using Microsoft Excel and GraphPad Prism 8. Statistical significances are referred as * (P < 0.05), ** (P < 0.01), *** (P < 0.001), and **** (P < 0.0001). Data distribution was assumed to be normal but was not formally tested. For comparison of the two groups, a two-tailed Student t test was used. For comparison of three or more groups P values were determined using the one- way analysis of variance (ANOVA) for multiple comparisons, and correction for the multiple comparisons was performed using Dunnett’s method. Comparison of extrusion probabilities of apoptotic, pyroptotic, and necroptotic cells was performed using a Mantel-Cox test.
Online supplemental material
Fig. S1 contains additional information on optoCaspase-1/-4 and -5 activation in human cells. Fig. S2 shows the testing of human and mouse optoCDEs in macrophages. Fig. S3 shows additional information on the activation of optoCaspase-1 in 2D and 3D cell culture systems, and additional information on the generation and testing of light-activatable apoptosis and necroptosis systems. Fig. S4 describes the generation and testing of zebrafish optoCDEs and contains additional information on the elimination of pyroptotic, apoptotic, or necroptotic cells. Fig. S5 contains additional information on the elimination of pyroptotic, apoptotic, or necroptotic cells. Video 1 shows the induction of pyroptosis by light-induced activation inflammatory caspases in HEK293T. Video 2 shows sub-lethal induction of pyroptosis on a single-cell level. Video 3 shows optogenetic pyroptosis induction in single cells. Video 4 shows optogenetic pyroptosis induction in 3D organotypic cell cultures. Video 5 shows optogenetic induction of apoptosis in HEK293T cells. Video 6 shows optogenetic induction of necroptosis in HEK293T cells. Video 7 shows optogenetic induction of pyroptosis in zebrafish larvae. Video 8 shows optogenetic induction of apoptosis in zebrafish larvae. Video 9 shows pyroptotic cell permeabilization and extrusion in zebrafish. Video 10 shows necrotic cell extrusion and apoptotic cell efferocytosis in human epithelial cells. Video 11 shows that the sphingosine-1-pathway regulates apoptotic cell efferocytosis by neighbors.
We thank the UNIL Cellular imaging facility, UNIL Flow Cytometry Facility and EMBL Advanced Light Microscopy Facility for technical assistance and to D. Gilmour and J. Hartmann for the tg(6xUAS:mneonGreen-UtrCH) zebrafish line.
This work was supported by grants from the ERC (ERC2017-CoG-770988-InflamCellDeath), the Swiss National Science Foundation (175576 and 198005), the the OPO Stiftung and Novartis to P. Broz.
The authors declare no competing financial interests.
Author contributions: K. Shkarina and P. Broz conceptualized the study. K. Shkarina, J.C. Santos, and S. Ramos performed in vitro experiments; M. Leptin and E. Hasel de Carvalho designed and performed in vivo experiments. All authors contributed to data analysis and manuscript writing.