LiverZap: a chemoptogenetic tool for global and locally restricted hepatocyte ablation to study cellular behaviours in liver regeneration

Functional recovery after liver injury relies on the capacity to re-establish organ mass and tissue architecture. Liver regeneration studies have focused mostly on the source of cell replenishment and the molecular pathways of mass recovery (Forbes and Newsome, 2016; Michalopoulos, 2007; So et al., 2020). However, the mechanisms controlling the restoration of the hepatic tissue architecture and underlying morphogenetic behaviours of the constituent cell types are still poorly understood.

Liver function depends on the intricate organisation of all hepatic cell types (Arias et al., 2020). Hepatocytes carrying out the main metabolic functions of the liver interact with both the vascular and biliary networks (Fig. 1A). The biliary network is composed of biliary epithelial cells (BECs), and is responsible for transporting hepatocyte-produced bile to the gallbladder and intestine. Upon mild or acute liver injury, such as partial hepatectomy, the remaining hepatocytes drive tissue mass restoration by hypertrophy and proliferation (Michalopoulos, 2007; Miyaoka et al., 2012). However, if hepatocyte proliferation is impaired, for instance in chronic liver injury, BECs can contribute new hepatocytes in both mammals and zebrafish. In this process, BECs de-differentiate into liver progenitor-like cells (LPCs), often referred to as oval cells in mammals, which can give rise to hepatocytes and BECs (Choi et al., 2014; He et al., 2014; Manco et al., 2019; Raven et al., 2017; So et al., 2020). The appearance of LPCs is associated with an expansion and remodelling of the biliary network (Choi et al., 2014; Kamimoto et al., 2016, 2020; Kok et al., 2015); however, the significance of this remodelling and how it contributes to liver regeneration remains unclear. Elucidating the dynamic cellular processes of liver damage and repair requires both extended live imaging and high spatial resolution. Despite significant progress with intravital microscopy (Vats et al., 2021), extended in vivo imaging is still challenging in rodents (Cheng et al., 2021). In contrast, zebrafish are ideally suited for live imaging of cell and tissue dynamics, given their transparency and fast organ formation. The cellular composition, liver functions, and key cellular pathways controlling liver development and regeneration are largely conserved between zebrafish and mammals (Oderberg and Goessling, 2023; So et al., 2020; Wang et al., 2017). Yet, the majority of available zebrafish hepatocyte ablation models, such as the nitroreductase-metronidazole (NTR-MTZ) system, require prolonged treatments with pharmacological agents (e.g. 24-36 h of metronidazole; Curado et al., 2008). Moreover, MTZ toxicity can sensitise the larvae (Mathias et al., 2014) and render them incompatible with long-term live imaging. Furthermore, regeneration is initiated concomitantly with cell ablation in these models, which hinders discrimination of the two processes. Finally, investigating the interface between regenerating and healthy tissue, relevant to human disease in which tissue damage is heterogeneous across the liver (Martini et al., 2023), is not possible using global hepatocyte ablation systems.

Optogenetic approaches (Varady and Distel, 2020) provide precise spatial and temporal control of tissue manipulation. Genetically encoded photosensitisers, molecules that release reactive oxygen species (ROS) upon illumination, can impair cell functions via photo bleaching-induced damage or cell death (Bulina et al., 2006). Yet, in vivo, efficient ablation and activation in deeper tissues varies between different photosensitisers, representing the biggest challenges for optogenetic ablation models (Liu et al., 2021). The recently developed binary FAP-TAP system comprises a genetically encoded fluorogen-activating protein (FAP) that binds the Malachite Green derivate dye MG-2I to form a targeted and activated photosensitiser (TAP). Near-infrared (NIR) illumination then initiates ROS production, which has been shown to induce cell death in vitro and in vivo in zebrafish cardiomyocytes as well as causing mitochondrial damage upon organelle-specific targeting in neurons (He et al., 2016; Missinato et al., 2021; Xie et al., 2020). The superior penetration properties of NIR light are ideal for effectively targeting deep tissues such as the liver, including reduced off-target excitation of endogenous chromophores (Deliolanis et al., 2008).

Adapting FAP-TAP to the zebrafish liver, we developed LiverZap, a powerful chemoptogenetic tool for targetable hepatocyte ablation. Based on its optogenetic nature and a very short activation time, it is suitable for extended live imaging, and separating injury and repair processes. Spatially restricted LiverZap activation enables the study of local liver injury, relevant for elucidating human liver disease.

To generate a chemoptogenetic hepatocyte ablation tool for extended live imaging, we developed LiverZap by adapting the FAP-TAP system (He et al., 2016). FAP expression, encoded by dL5**, was placed under the control of hepatocyte promoter fatty acid-binding protein 10a (fabp10a). In combination with MG-2I dye, FAP forms an active photosensitiser (FAP-TAP). Importantly, MG-2I was washed off prior to illumination with NIR light at 660 nm, which then generates ROS (Fig. 1A). tg(fabp10a:dsRed) transgenic larvae were crossed to stable tg(fabp10a:dL5**–mCer3), abbreviated tg(LiverZap), to allow visualisation of hepatocytes and hepatocyte-specific ROS production following 12 min NIR light exposure. Cell death was evaluated live 8 h post-illumination (hpi) by the membrane-permeable nucleic acid-binding dye Acridine Orange. Green Acridine Orange fluorescence and apoptotic bodies were detected in DsRed-expressing hepatocytes of tg(LiverZap) larvae incubated with MG-2I, compared with control larvae displaying no signal (Fig. 1B,C). Terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) further determined apoptosis as the main cell death mechanism, consistent with a previous report (He et al., 2016) (Fig. S1D,E), although the contribution of other cell death pathways cannot be excluded. To determine whether ROS production impacts adjacent cells, Acridine Orange incorporation was assessed in the tg(tp1:H2B-mCherry) reporter marking BECs by expressing H2B-mCherry under the Notch -responsive element Tp1 (Ninov et al., 2012). H2B-mCherry-positive cells were devoid of Acridine Orange signal (Fig. 1D,E), validating that LiverZap triggers hepatocyte-specific cell death, without a bystander effect in adjacent cells.

Having established the hepatocyte-ablation capacity and specificity of LiverZap, we examined ablation efficacy and the regeneration process using tissue mass recovery as read-out. Liver injury was triggered in 5 days post-fertilisation (dpf) tg(LiverZap) larvae and liver volume monitored over time by tg(fabp10a:eGFP) hepatocyte expression (Fig. 1F-Q). At 12 hpi, hepatocyte ablation was ongoing according to a 20-40% liver reduction compared with non-ablated livers (Fig. 1F,G,Q). Between 24 and 30 hpi, we observed two categories of hepatocyte ablation (Fig. 1H-J, Fig. S1A-C): (1) mild ablation, encompassing about 30% smaller livers devoid of apparent organ morphology changes or tissue integrity defects, and (2) severe ablation, marked by ≥70% liver size reduction accompanied by a loss of organ morphology, and severely disrupted tissue organisation, including detaching hepatocytes (Fig. 1J). At 50 hpi, a 14% size gain in livers with severe hepatocyte ablation demonstrated ongoing regeneration. Mildly ablated livers exhibited no significant change (Fig. 1K-M,Q). At 72 hpi, size recovery was also detectable in mildly ablated livers, whereas severely injured livers showed a drastic volume increase, doubling their size in 24 h (Fig. 1N-Q). By 7 days post-illumination (dpi), severely ablated livers had recovered to the control liver volume (Fig. 1Q). Quantification of total hepatic cell numbers mirrored and corroborated these liver regeneration dynamics (Fig. S1F).

A prerequisite for elucidating the cell behaviours mediating injury and repair is to determine the precise time point when regeneration begins. Employing light-sheet microscopy to visualise the regeneration process after LiverZap ablation, we imaged tg(LiverZap);tg(fabp10a:eGFP) larvae from 24 until 55 hpi (Fig. S1I, Movie 1). GFP-positive hepatocytes died from the first hours until maximum ablation was reached between 24 and 30 hpi with only a few round and detached hepatocytes remaining (Fig. S1I). From this point onwards, regeneration was evident by the restoration of liver mass and organ morphology. This demonstrates that LiverZap, owing to its optogenetic design, is compatible with extended live imaging, allowing the investigation of dynamic tissue morphogenesis after injury.

Because the FAP-TAP system kills cells by ROS production, we tested whether altering oxygen availability from 40% to 20% oxygen (normoxia) could be used to control hepatocyte ablation. Quantification at 24 hpi, when mild and severe categories are distinguishable also by stereomicroscope (Fig. S1A-C), revealed that hepatocyte ablation in 40% oxygen-containing medium was highly efficient, with ∼90% severely ablated livers, in contrast to ∼60% severely and ∼40% mildly ablated livers under normoxic conditions (Fig. 1R, Fig. S1G). In contrast, shorter illumination is unsuitable for modulating LiverZap ablation, since solely the number of control-like larvae increased, instead of those displaying mild injury (Fig. S1H).

Across species, the extent of hepatic injury determines which regeneration programme restores liver mass (Forbes and Newsome, 2016; So et al., 2020). Like mammals, mild or acute injury in larval and adult zebrafish livers is resolved by hepatocyte proliferation, whereas BEC-derived LPCs will give rise to new hepatocytes after extensive hepatocyte death or when hepatocytes are unable to proliferate (Choi et al., 2014; He et al., 2014; Huang et al., 2014; Oderberg and Goessling, 2023). Given that LiverZap produces both mild and severe hepatocyte ablation, we examined which regenerative response is elicited by employing histone inheritance for short-term lineage tracing (Blanpain and Simons, 2013), similar to previous studies (Choi et al., 2014; He et al., 2014). We generated tg(LiverZap);tg(fabp10a:eGFP);tg(tp1:H2B-mcherry) fish to visualise hepatocytes by eGFP expression, and Notch-active BECs expressing tp1:H2B-mCherry (referred to as mCherry; Fig. 2A-A‴). Because mCherry is fused to the histone protein H2B, progeny derived from a mCherry^high cell inherit half of the fluorescent protein. Hence, if a Notch-active BEC/LPC gives rise to a Notch-inactive hepatocyte, the latter will inherit half the mCherry and be identified as mCherry^low cell, whereas Notch-active cells will maintain mCherry^high expression. As regeneration progressed at 48 hpi, control and mildly ablated livers showed mutually exclusive GFP and mCherry expression, marking hepatocytes and BECs, respectively (Fig. 2A-B‴). Notably, in severely ablated livers, there were two populations of mCherry expressing cells: mCherry^high and mCherry^low cells. The latter were all also fabp10a:eGFP-positive, suggesting their LPC origin and hepatocyte identity (Fig. 2C-C‴). This indicates that regeneration of mildly ablated livers occurs by hepatocyte proliferation, whereas in severely ablated livers, BECs de-differentiate into LPCs that give rise to new hepatocytes.

To determine whether BECs/LPCs proliferate before they differentiate into new hepatocytes or after to replenish the BEC population, cell type-specific proliferation was examined by 5-ethynyl-2′-deoxyuridine (EdU) incorporation in tg(LiverZap);tg(fabp10a:eGFP);tg(tp1:H2B-mCherry) over time. In severely ablated livers, surviving hepatocytes proliferated first, at 12 hpi (Fig. 2D), followed by BEC/LPC proliferation, which peaked at 30 hpi (Fig. 2E), and lastly both LPC-derived and remaining hepatocytes proliferated at 48 hpi (Fig. 2D,F). Proliferation of all cell populations returned to control levels by 72 hpi (Fig. 2D-F). Quantification of mitotic cells by staining for phospho-histone-3 (PH3) during regeneration in control and severely ablated samples showed similar dynamics (Fig. S2B-D). These findings show that BEC/LPC proliferation precedes differentiation of new hepatocytes and that the LPC response does not deplete the BEC pool at any time (Fig. S2A). In mildly ablated livers, hepatocyte proliferation increased after ablation at 12 hpi and 30 hpi and decreased to control levels by 72 hpi (Fig. 2D). Interestingly, at 12 hpi and 30 hpi, BECs also showed mildly increased proliferation (Fig. 2E), suggesting that BECs respond although they do not progress into LPCs. Together, these data demonstrate that LiverZap can elicit different ablation severities and distinct repair responses, recapitulating the predominant regeneration modes in zebrafish and mammals.

The zebrafish NTR-MTZ liver regeneration models elicit organ-wide tissue damage, but locally restricted hepatocyte ablation cannot be recapitulated (Curado et al., 2008), which is a limitation considering that clinically relevant hepatic injury is often regionally restricted (Martini et al., 2023). We therefore tested whether restricted NIR illumination could activate LiverZap locally instead of globally. A three-dimensional region of interest (ROI) was illuminated at 660 nm in livers of tg(fabp10a:DsRed);tg(LiverZap) and tg(tp1:H2B-mCherry);tg(LiverZap) larvae using a confocal microscope with a tuneable white light laser (Fig. 3A). At 17 hpi, Acridine Orange-marked hepatocytes and loss of tissue integrity were observed exclusively in the illuminated ROI, representing about 10-20% of the liver volume (Fig. 3B-D), whereas no Acridine Orange-positive BECs were detected (Fig. S3L-M‴). Likewise, no nonspecific hepatocyte death was detected in ROI-illuminated control livers (Fig. 3B), showing that LiverZap can mediate spatially restricted hepatocyte ablation.

Addressing whether restricted hepatocyte ablation is sufficient to trigger regeneration, ROI illumination was performed in tg(LiverZap);tg(fabp10a:eGFP);tg(tp1:H2B-mCherry) embryos (Fig. 3E-J′). Similar to whole-organ ablation, the source of new hepatocytes can be deduced from the inheritance of BEC-derived mCherry expression. Monitoring the ablation efficacy at 24 hpi by live imaging showed that efficient hepatocyte ablation was associated with BEC aggregation in the ROI (Fig. 3F,I), consistent with global ablation. By 72 hpi, hepatocytes had repopulated the ablated ROI (Fig. 3J), indicating that regeneration had been triggered. Remarkably, most hepatocytes in the ablated region, identified by fabp10a:eGFP and a characteristic round nuclear shape (Russell et al., 2019), exhibited mCherry^low expression (Fig. 3J′), suggesting their BEC origin, and the activation of the LPC programme. Imaging distant regions in the same ROI-ablated specimen revealed no apparent morphological changes, corroborating that the LPC programme was exclusively activated in the illuminated ROI (Fig. S3A-D′). This finding is very surprising, because current understanding in the field links the generation of new hepatocytes from BECs/LPCs with global hepatocyte death.

ROI-restricted LiverZap activation elicited the LPC response locally in a comparatively small volume, which appeared to recover tissue mass within 72 h. This raised the question of whether solely BECs within the illuminated ROI gave rise to new hepatocytes. To trace BECs exclusively in the ROI, we turned to a photoconversion strategy (Arrenberg et al., 2009). To express photoconvertible Dendra specifically in BECs, we identified the gene-trap line gSAlzGFFM1349A, which expresses the activator Gal4FF (Asakawa and Kawakami, 2009) in BECs (E.M.A., K. Kawakami and E.A.O., unpublished) and combined it with the effector line UAS:Dendra-KRAS. Photoconverted Dendra was specific to BECs and still detectable after 72 h (Fig. S3E-J). In tg(LiverZap);tg(tp1:H2B-mcherry);gSALzGFFM1349A;tg(UAS:Dendra-KRAS) larvae, photoconversion was immediately followed by LiverZap activation in the same ROI (Fig. 3K). Photoconversion and NIR light illumination of control larvae without MG-2I caused no injury (Fig. 3L-N″). Tissue morphology alteration concomitant with hepatocyte death and the condensation of BEC nuclei was apparent 24 hpi after ROI LiverZap activation in tg(LiverZap);tg(tp1:H2B-mcherry);gSALzGFFM1349A;tg(UAS:Dendra-KRAS) larvae (Fig. 3O,P). At 48 hpi, BEC/LPC-derived hepatocytes appeared, distinguishable by their mCherry^low expression and hepatocyte-typical round nuclear shape (Fig. 3Q-Q″). Some of the LPC-derived hepatocytes had inherited converted red Dendra, highlighting new hepatocytes derived from BECs/LPCs in the ablated ROI (Fig. 3Q-Q″). Strikingly, a subset of BEC/LPC-derived-hepatocytes (mCherry^low) showed non-converted green Dendra (Fig. 3Q,Q′), indicating that, in addition to BECs/LPCs in the ablated region, BECs within an average distance of 78 µm in adjacent uninjured tissue (Fig. S3K) respond to the injury, de-differentiate into LPCs and contribute to replenishing new hepatocytes after local injury.

Although we are starting to understand the dynamic processes of BEC network formation (Dimri et al., 2017; Lorent et al., 2010), the cellular behaviours and morphological changes elicited by liver injury and driving regeneration are largely unknown. Given the crucial role of BECs in regeneration, understanding the morpho-dynamic processes related to the biliary network, such as the de-differentiation of BECs into LPCs, is of particular interest.

First, to quantify the change of biliary network structure during injury and regeneration, distances to the nearest neighbour were measured in livers using BEC nuclei following LiverZap activation in 5 dpf tg(LiverZap);tg(tp1:H2B-mCherry) larvae (Fig. 4A-I). In control livers, BEC nuclei (mCherry^high) were detected at all analysed time points distributed throughout the organ, with 40% having the closest neighbour at a 0-8 µm distance and 50% at 8-16 µm. In contrast, at 12 hpi, midway through ablation, BEC nuclei commenced clustering with 60% displaying a 0-8 µm distance. This distribution was maintained in mildly ablated livers at 30 hpi, the maximum ablation time point. Severely ablated samples, however, showed more extensive clustering of BEC nuclei as 85% were 0-8 µm apart. This doubling of the small distances compared with control demonstrates increased clustering of BECs during injury. At the network scale, biliary Anxa4 expression at 24 hpi corroborated these morphological changes (Fig. S4). As regeneration was initiated, biliary network organisation was progressively restored. By 72 hpi, mildly ablated livers showed a BEC inter-nuclear distance distribution that was comparable to control larvae. Severely ablated livers at 72 hpi showed increased distances between BECs; 40% showed a surprisingly larger distance of greater than 24 µm between neighbours, which was rare in control livers, suggesting that network restoration is still ongoing (Fig. 4I). These data show that gradual condensation of BECs accompanies progressive hepatocyte death and provides a quantitative baseline for future functional studies of BEC network remodelling during injury and regeneration.

The dynamic nature of BEC aggregation and re-distribution raised the question of whether this is the result of active remodelling or whether the loss of the scaffold normally provided by hepatocytes could lead to a passive collapse of the network. Live imaging of tg(tp1:eGFP) in LiverZap-ablated livers during the injury phase from 5 to 20 hpi (Fig. 4J), revealed highly dynamic BEC behaviours, including cell rearrangements and filopodia-like extensions, whereas control BECs appeared mostly static (Fig. 4K-R; Movie 2), suggesting that BECs actively rearrange in response to the loss of adjacent tissue mass and do not solely collapse passively.

Here, we introduce LiverZap, a new chemoptogenetic hepatocyte ablation tool for studying liver injury in zebrafish. LiverZap ablation triggers severity-dependent regeneration programmes corresponding to those in zebrafish and mouse models (Forbes and Newsome, 2016; Michalopoulos, 2007). In contrast to existing tools, LiverZap does not require long incubation with pharmacological agents and has a short activation time, enabling live imaging up to 30 h, crucial for uncovering key morphogenetic events in liver repair. Showcasing the experimental potential of LiverZap, we demonstrate that targeted ablation of 10-20% of the liver volume triggers LPC-mediated regeneration, which commonly occurs only upon extensive global hepatocyte ablation, and, strikingly, involves the conversion of BECs outside the ablated region into LPCs to restore the lost tissue.

LiverZap can elicit both mild and severe hepatocyte ablation, which reflect the two main modes of zebrafish and mammalian liver regeneration depending on the type and extent of injury: the regenerative potential of the liver following mild injury relies on the proliferation of hepatocytes, whereas following extensive injury or impaired hepatocyte proliferation regeneration is largely dependent on de-differentiation of BECs (Forbes and Newsome, 2016; Michalopoulos, 2007; So et al., 2020; Wang et al., 2017). Short-term lineage tracing of BEC progeny and proliferation studies support our current understanding of cellular contributions to liver regeneration according to injury extent. Importantly, the high spatiotemporal control of the LiverZap system allowed interrogation of distinct, cell type-specific proliferation phases: hepatocytes proliferate first, then BECs/LPCs and, finally, LPC-derived hepatocytes. The LiverZap-induced LPC response is thus reminiscent of a ductular reaction with LPC response in murine models of severe hepatocyte loss, senescence or chronic liver disease induced, for instance, by thioacetamide (TAA), choline-deficient, ethionine-supplemented (CDE) or carbon tetrachloride (CCl₄) diets (So et al., 2020). Both the cellular sources and the proliferation dynamics resemble those of other injury models, including the widely used NTR-MTZ system in zebrafish (Choi et al., 2014; He et al., 2014). LiverZap findings will therefore be comparable with and expand currently available tools with its suitability for live imaging.

A key advantage of LiverZap is that incubation with the MG-2I dye to form the FAP-TAP photosensitiser occurs prior to the short 12-min ROS-inducing NIR illumination. This contrasts with the long drug treatments concomitant with cell ablation in the existing NTR-MTZ models (Choi et al., 2014; He et al., 2014; Huang et al., 2014), which are prone to obscuring regeneration onset. Thus, the LiverZap ablation mechanism allows (1) the separation of continued injury and starting regeneration phases, and (2) circumvention of possible nonspecific MTZ effects on larval survival (Mathias et al., 2014), which can be detrimental to extended live imaging. This is clearly shown by live imaging for more than 30 h, which is essential for determining dynamic BEC behaviours and cell–cell interactions underlying de- and reconstruction of the liver architecture following injury. We anticipate that cell ablation in the adult liver will be feasible, like previous work in the heart (He et al., 2016). Employing optically more transparent fish, such as the Casper strain (White et al., 2008), together with the superior tissue penetration of NIR light would enable the deeper positions of the liver in the body cavity to be reached.

Under pathological conditions, liver damage is often locally restricted; for example, non-alcoholic fatty liver diseases originate pericentrally in human, whereas autoimmune hepatitis starts periportally (Martini et al., 2023). Regenerative processes consequently take place at the junction of healthy and diseased tissue. The precise and restricted ROI activation of LiverZap allows these processes to be investigated live and at high resolution. Extraordinarily, ROI hepatocyte ablation of 10-20% liver volume activated the LPC programme in a regionally restricted fashion, suggesting that LPC activation does not exclusively happen after extensive, global hepatocyte death as previously shown (Choi et al., 2014; He et al., 2014; Manco et al., 2019; Raven et al., 2017; So et al., 2020). Instead, it suggests that the ratio of lost hepatocytes to BECs in a defined area may be sufficient to elicit this repair programme. This could also explain recent observations of small foci of LPC-derived hepatocytes in NTR-MTZ-mediated injury in adult zebrafish (Oderberg and Goessling, 2023). Therefore, we propose that LPC activation may be a consequence of BECs sensing their environment and reacting to vast hepatocyte cell death in their immediate surroundings. Elucidating the cellular and morphological changes as well as the cell–cell signalling occurring at the interface of injured and healthy tissue will be crucial to compare with behaviours in hepatic disease such as cirrhotic nodules, where restricted proliferation is insufficient to fully restore liver function (Forbes and Newsome, 2016).

Furthermore, we show that restricted hepatocyte ablation activates the LPC programme in BECs up to 100 µm beyond the ablated region, which contribute to new hepatocytes. This strongly suggests that BECs sense and respond to signals from their environment, potentially through monitoring molecules released by dying hepatocytes (Brenner et al., 2013; Eguchi et al., 2014; Jung et al., 2010). Likewise, CCl₄, known to kill hepatocytes around the central vein in mammalian livers, can, upon extended treatment, trigger an LPC response in distant periportal BECs (Manco et al., 2019), suggesting that activation of liver regeneration processes at a distance may be a conserved mechanism between zebrafish and mice. This could be explained by different mechanisms. Systemic signals, although attractive (Johnson et al., 2018), seem unlikely given that only BECs adjacent to the injury site and not throughout the liver exhibit an LPC response. Given that diffusion often occurs only over a short range in dense tissues (Müller et al., 2013), signalling could also occur by long filopodia from the injury site, as observed frequently during development (Caviglia and Ober, 2018). The signal could be dispersed by trigger waves, like ERK in the murine skin and zebrafish scales (De Simone et al., 2021; Hiratsuka et al., 2015). Although Egfr-ERK signalling inhibits hepatocyte differentiation from LPCs (So et al., 2021), it may still mediate the dedifferentiation of BECs into LPCs. A final alternative could be oscillatory signalling, for instance of the Notch pathway, which represents one of the few known factors mediating the BEC-to-LPC conversion (He et al., 2014) and displays oscillatory behaviour in other systems (Bosman and Sonnen, 2022). Our work thereby contributes to the exciting view that tissue regeneration is more complex than a local response, and may in fact require the integration of neighbouring, long-distance or systemic contributions (Sun et al., 2023).

Nuclear clustering and biliary network remodelling occur after extensive LiverZap and NTR-MTZ-triggered hepatocyte death (Choi et al., 2014), which is also a hallmark of mammalian ductular reaction (Kamimoto et al., 2016; Kaneko et al., 2015). Although nuclear aggregation is more evident in severe samples, we show that it also occurs in mild LiverZap-ablated livers, suggesting it may be the first response of BECs to hepatocyte death, possibly as a result of local changes in the mechanical support. Similarly, ductular reaction in murine models can occur without the generation of new hepatocytes (Kamimoto et al., 2020), suggesting that specific thresholds control the activation of the LPC programme. Therefore, extensive BEC network remodelling, marked by enhanced BEC condensation and protrusive activity, dominating in severely ablated livers, suggests that active cell rearrangement plays a key role in the remodelling of the biliary network upon liver injury.

In summary, LiverZap is a chemoptogenetic hepatocyte ablation model based on 12-min NIR-light activation that enables (1) extended live imaging for capturing tissue remodelling during ablation and repair, (2) separation of injury and regeneration processes, and (3) ROI hepatocyte ablation. LiverZap elicits common hepatocyte- and LPC-mediated regenerative responses. Its experimental properties extend the existing regeneration toolbox and will enable analysis of dynamic cellular behaviours key for understanding how liver architecture is deconstructed upon injury and rebuilt during repair.

Adult zebrafish and embryos (Danio rerio) were kept and raised according to standard laboratory conditions (Westerfield, 2000). To prevent pigmentation, embryos were grown in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂ and 0.33 mM MgSO₄) with 0.2 mM N–phenylthiourea (PTU; Sigma-Aldrich, P7629) from 24 hpf onwards and under standard laboratory conditions. All experiments were performed according to ethical guidelines approved by the Danish Animal Experiments Inspectorate (Dyreforsøgstilsynet; approval #2018-15-0201-01431).

The following strains were used: tg(EPV.Tp1-Mmu.Hbb:hist2h2l-mCherry)^s939 abbreviated to tg(tp1:H2B-mCherry) (Ninov et al., 2012), tg(EPV.tp1-Mmu.Hbb:eGFP)^um14 abbreviated to tg(tp1:eGFP) (Parsons et al., 2009), tg(-2.8fabp10a:eGFP)^as3 abbreviated to tg(fabp10a:eGFP) (Her et al., 2003), tg(UAS:Dendra-kras)^s1998t abbreviated to tg(UAS:Dendra) (Arrenberg et al., 2009), tg(fabp10a:dsRed)^gz15 abbreviated to tg(fabp10a:dsRed) (Dong et al., 2007) and gSAlZGFFM1349A:Gal4FF (Kawakami et al., 2010).

The fluorogen-activating protein dL5** (FAP) fused to monomeric cyan fluorescent protein mCer3 was excised from the pCS2+ MBIC5-mCer3 construct (He et al., 2016) and placed under the control of the fabp10a promoter (Her et al., 2003) by insertion into the pHD157 -2.8fabp10a:Cre;cryaa:Venus construct (Ni et al., 2012). Stable transgenic carriers were generated by injecting 25-50 pg DNA with I-SceI meganuclease into one-cell-stage embryos as previously described (Soroldoni et al., 2009). Multiple transgenic founders were tested for hepatocyte ablation efficiency, of which two transgenic lines of similar efficiency were established. Tg(fabp10a:dL5**-mCer3)^cph10 and ^cph11, abbreviated tg(LiverZap), are currently maintained in F6 generations. All experiments were carried out with heterozygous tg(fabp10a:dL5**-mCer3)^cph10 larvae.

Specifically, larvae were fixed in 4% paraformaldehyde (Sigma Aldrich, 6148) overnight at 4°C. Embryos were then washed three times, 5 min each wash, in PBST 0.1% (0.1% Triton X-100 in PBS). The yolk was manually removed followed by permeabilisation with PBST 2% (2% Triton X-100 in PBS) for 1 h at room temperature. Blocking and whole-mount antibody staining was performed in PBST 1% (1% Triton X-100 in PBS) with 10% serum (goat or donkey) and 1% dimethyl sulphate (DMSO; Sigma-Aldrich, D8418). Antibody incubation was performed at 4°C for two nights with gentle shaking. Antibodies used were: mouse anti-Anxa4/2F11 (1:1000; gift from Julian Lewis, CRUK-LRI, London, UK), goat anti-Hnf4a (1:100; Santa Cruz Biotechnology, sc-6556, lot B1605), rabbit anti-pH3 (1:500; Millipore, 631257, lot 2825969), chicken anti-GFP (1:500; Abcam, ab13970), rat anti-mCherry (1:1000; Invitrogen, m11217, lots TL276838 and UJ287711). Fluorophore-conjugated secondary antibodies (Jackson ImmunoResearch; Alexa Fluor 488 donkey anti-chicken, 703-545-155; Alexa Fluor 405 donkey anti-rabbit, 711-475-152; Cy3 donkey anti-rat, 715-166-150; Alexa Fluor 647 donkey anti-mouse, 715-605-151) were added at 1:500 to PBST 1% with 10% serum and 1% DMSO and incubated for one to two nights at 4°C. To visualise nuclei, DAPI (Sigma-Aldrich, D9542) was added to the secondary antibody solution. Whole-mount immunostained embryos were mounted either in 1:2 benzyl alcohol:benzyl benzoate (Sigma-Aldrich B6630 and 305197) clearing mixture after methanol dehydration or in Vectashield (Vector Laboratories, H-1900) prior to imaging. The embryos were imaged either with a Leica SP8 or Leica Stellaris confocal microscope. Long-term live imaging was performed using an LS1 Live light-sheet microscopy system (Viventis Microscopy Sàrl). Images were processed using Imaris (Bitplane) image analysis software, Fiji (Schindelin et al., 2012) and Adobe Suite CS 2019.

Embryos were incubated in E3 supplemented with 0.2 mM PTU and 500 nM MG-2I (synthesised as described by He et al., 2016) protected from light, overnight or for a minimum of 3 h at 28°C, 20% or 40% O₂. The next day, larvae were transferred into a 6 cm diameter Petri dish with fresh 20% or 40% oxygen-enriched E3/PTU. The free-swimming larvae were placed under an NIR LED light at a 2 cm distance, inside a custom-built box (as described by He et al., 2016) equipped with an exhaustion fan to dissipate heat. Larvae were illuminated for 12 min, and subsequently returned to the incubator at 28°C with ambient oxygen levels. Control larvae without MG-2I underwent the same treatment. Ablation severity phenotypes were scored at 24 hpi under a fluorescent stereomicroscope (Leica, M205 FCA) based on liver size, as described in the Results section.

Bitplane Imaris software was used to calculate the liver volume by manual segmentation of the liver using the ‘surface’ module. Nuclear counts were determined with the ‘spots’ module, including manual correction.

Using image analysis software Bitplane Imaris (version preceding 9.0), all BECs were marked using the ‘spots’ function. For each spot, the xyz coordinates were exported and employed to calculate the nearest neighbour distance using a custom MATLAB script. Since the release of Imaris V9.0, this quantification can also be performed automatically through the software's statistics tab.

Larvae were incubated in Acridine Orange (5 µg/ml; Sigma Aldrich, A6014) in E3/PTU for 30 min protected from light at 28°C and washed twice, 5 min each. Zebrafish larvae were anesthetised with Tricaine (160 mg/l; Sigma-Aldrich, A5040) and mounted in 1% low melting agarose (Lonza, 50080) supplemented with Tricaine. They were mounted on their left side to increase tissue accessibility and imaged using a Leica Stellaris confocal microscope.

Embryos were incubated for 1 h at 28°C with 400 µM EdU (Invitrogen, A5040) and 5% DMSO in E3/PTU. The fish were then fixed with 4% paraformaldehyde overnight at 4°C with gentle shaking. EdU incorporation was visualised using the Click-iT™ EdU Alexa Fluor™ 647 Imaging Kit (Invitrogen, C10340) prior to secondary antibody incubation.

Fish with the desired transgenic expression were incubated at 5 dpf with MG-2I as described for LiverZap activation. The next day, individual fish were transferred into fresh oxygen-rich E3/PTU (incubated overnight at 40% O₂). The fish were anesthetised with Tricaine and mounted on their left side in low melting agarose 1% in E3/PTU media supplemented with Tricaine. Once the agarose solidified, the chamber was filled with oxygen-rich E3/PTU with Tricaine to prevent drying and maintain the anaesthesia. For ROI ablation, fish were illuminated using a Leica Stellaris confocal microscope equipped with a white light laser (WLL) tuned at 660 nm. Fish were imaged using a 40× water objective. The ROI was selected by digital zoom into the tissue (3× zoom). LiverZap activation in the ROI was achieved using a laser speed of 10 Hz in combination with a low resolution of 256×256 pixels. The total z-stack of the ROI was kept at 90 µm, illuminated at 5 µm intervals, which in the majority of cases encompasses the entire thickness of the tip of the left liver lobe. The WLL power was set up at either 85% or 100% and laser intensity at 100%. Each ROI was illuminated for a total of 20 or 40 min, equivalent to five or ten cycles of imaging for the complete z-stack. We observed an increasing need for laser power or imaging length as the laser aged from daily wear.

After illumination, individual larvae were carefully recovered from the agarose, placed in E3/PTU and ambient oxygen levels and kept in an incubator until the desired time point.

Larvae were prepared and mounted in the same fashion as for ROI LiverZap activation. For photoconversion, we illuminated the fish for 5 min with 10% laser power using a 405 nm laser in the Leica Stellaris confocal microscope. We simultaneously acquired both non-converted and converted Dendra signal and confirmed that all Dendra-expressing cells were converted. To prevent nonspecific photoconversion, imaging of non-converted Dendra was performed using very low 488 nm laser power (0.1%). We used a Leica Stellaris confocal microscope and acquired the images using a 40× water objective.

For live imaging with the LS1 Live light-sheet microscope (Viventis), fish were mounted ventrally to improve tissue illumination and imaging acquisition. A 25× objective was used and acquisition was performed every 20 min.

Biological replicates represent different cohorts of larvae, bred on different days and/or different pairs and are indicated in each figure legend as N. Technical replicates are indicated with the letter n and represent individual embryos. Statistical analysis was performed using GraphPad Prism software and specific tests used are stated in figure legends.

We thank all members of the Ober group for fruitful discussions, T. Piotrowski for invaluable support, K. Kawakami for the gSAlZGFFM1349A:Gal4FF line, D. Y. R. Stainier for the tg(UAS:Dendra-KRAS) line and S. E. Fraser for suggestions on oxygen concentrations. We also thank A. Vanoosthuyse for expert technical support, J. Bulkescher, A. Sheshtra and J. Dreier (DanStem Imaging core facility) for training and continued advice with image acquisition, analysis and quantification, and the department of experimental medicine for expert fish care. We are grateful to Drs A. Boni and P. Strnad (Viventis) for light-sheet microscopy. The Novo Nordisk Foundation Center for Stem Cell Biology was supported by a Novo Nordisk Foundation grant number NNF17CC0027852.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202217.reviewer-comments.pdf