An active traveling wave of Eda/NF-κB signaling controls the timing and hexagonal pattern of skin appendages in zebrafish Free

Skin appendages are a key feature of vertebrate anatomy, and have a myriad of functions and forms, from hair in mammals to feathers in birds (Chuong et al., 2013; Aman et al., 2018; Painter et al., 2021). Formation of these structures is complex and includes periodically and dynamically delivered patterning sequences. The emergence of periodic patterns is a fundamental process in development and drives the formation of a variety of tissues, both internal and external. Periodic patterns can arise from a homogeneous field of cells through symmetry breaking mechanisms such as self-organization or changes in gene expression at the cellular level. Experimental access to this dynamic sequence is a challenge that often is limited by a lack of direct access to the process (Chuong et al., 2013; Painter et al., 2021). Many of the mechanisms that regulate periodic patterning occur early in development, when tissues such as the skin or gut are inaccessible. Mammals and birds, two common model systems to study skin development, develop in utero or in ovo, making longitudinal studies difficult (Kowalczyk-Quintas and Schneider, 2014). Here, we develop a system for exploration via live imaging of zebrafish scales as they develop.

Zebrafish scales are skin appendages that cover the fish in a highly organized, hexagonal array. Their development process is similar to other skin appendages and the pathways that regulate this are conserved across vertebrates. Unlike skin appendages in other vertebrate systems, zebrafish scales develop relatively late in ontogeny, ∼4 weeks postfertilization, and are transparent, providing a unique opportunity for longitudinal studies examining tissue growth and patterning (Sire and Akimenko, 2004; Aman et al., 2018).

Zebrafish scales consist of osteoblasts surrounding a mineralized bone matrix and mirror mammalian dermal bone development by forming via intramembranous ossification (Sire et al., 1997; Cox et al., 2018). Scales do not appear all at once over the fish, but sequentially in a spreading wave. The first scales begin forming in two locations, one in the trunk of the fish and one more anterior on the belly, then spread out along the anterior-posterior (AP) and dorsal-ventral (DV) axes until the entire fish is covered (Fig. 1A-D). Recent studies have begun to identify the molecular pathways involved, but exactly how this process is regulated remains unclear (Sire and Akimenko, 2004; Aman et al., 2018, 2021). Recently a potential role for somites in determining the positioning of scales in snakes has been revealed (Tzika et al., 2023).

Although bony scale development is unique to fish, the pathways that coordinate this process, namely Ectodysplasin A (Eda), Wnt and Fgf, regulate skin appendage formation across vertebrates (Kowalczyk-Quintas and Schneider, 2014; Ho et al., 2019). Eda is a member of the TNF family and binds to activate its receptor, Edar. This leads to NF-κB then translocating into the nucleus to activate target genes, including Wnt, Shh and Fgf genes (Sadier et al., 2014) (Fig. 1G). In zebrafish, eda is expressed in dermal fibroblasts and null mutations lead to a complete loss of scales. On the contrary, edar is expressed in epidermal cells above the scale papillae (Harris et al., 2008; Aman et al., 2018).

The process by which cells migrate and differentiate to form skin appendages is likewise conserved. In zebrafish, mesenchymal cells in the dermis condense to form placodes. At the same time, signals from the epidermis trigger growth and differentiation in these cells. The placode then develops into a layer of osteoblasts surrounding a mineralized matrix of bone (Chuong et al., 1996; Sire and Akimenko, 2004; Aman et al., 2021) (Fig. 1H,I).

Although much has been discovered regarding the formation of individual hair follicles or feathers, establishing the regulators of the temporal and spatial patterning has remained elusive. Signaling waves are emerging as a general mechanism by which developing tissues are organized. In systems that work in a periodic and sequential manner, waves contribute to directionality, sequentiality and spacing of the patterns (Widelitz et al., 2006; Ho et al., 2019).

During skin development, a tissue-wide priming wave could shift a homogenous system of cells into a state to allow for symmetry breaking and facilitate the spatial and temporal organization of the pattern. This priming wave could be either an active or passive wave. Active waves depend on diffusion and biochemical feedback to travel across a tissue, whereas passive waves are kinematic and are controlled by pre-determined spatial delays (for example, delays imposed by nonuniform oscillators) (Winfree, 1972; Deneke and Di Talia, 2018). Although early studies have suggested that the wave which patterns feather placodes might be passive, more recent work has shown that the morphogenetic signals traveling ahead of placode formation are in fact active waves (Davidson, 1983b; Patel et al., 1999; Ho et al., 2019). As these early experiments were carried out in chicken skin explants without the benefit of live molecular markers, it is likely that the signals regulating feather formation were not interrupted. Due to the dynamic nature of these waves, the timing of experiments to interrupt wave propagation is crucial. Longitudinal studies with molecular markers are also necessary for distinguishing passive and active waves (Davidson, 1983b; Painter et al., 2021).

During feather development, EDA is expressed as a wave in the dorsal skin and drives feather patterning (Houghton et al., 2005; Ho et al., 2019). EDA upregulates FGF20 expression, which promotes dermal cell condensation. These condensates in turn activate more FGF20 signaling through mechanical feedback. Dermal cell aggregation also occurs in a wave-like pattern. Although EDA and FGF20 promote the formation of cell condensates in the dermis, it remains unclear whether the formation of condensates precedes the wave of EDA expression or vice versa and what the dependencies among these signals are (Ho et al., 2019). Moreover, it has also been suggested that the formation of feather placodes is initiated by mechanical signals rather than biochemical ones (Shyer et al., 2017). Due to the lack of live imaging techniques and longitudinal studies, it has been difficult to dissect which processes precede and drive one another and the precise relationship between these processes remains unclear. The ability to study scale formation longitudinally in living animals using transgenic reporters and pharmacological perturbations provides a unique opportunity to dissect these dependencies in the formation of zebrafish scales.

The transcription factor NF-κB acts downstream of Eda signaling and is important for skin appendage development (Drew et al., 2007; Sadier et al., 2014; Aman et al., 2018). However, the dynamics of NF-κB activity in appendage development have yet to be quantitatively described. To that end, we used an osteoblast-marker to label scales (osx:H2A-mcherry) (Cox et al., 2018) and a previously established transcriptional reporter transgenic line Tg(6xHsa.NFKB:EGFP) (Kanther et al., 2011), hereafter NF-κB:eGFP, to visualize NF-κB activity and its dynamics during scale development. This line features six NF-κB binding sites fused to a minimal promoter to drive GFP expression when NF-κB is in the nucleus. By live imaging skin patterning every 24 h, we found that NF-κB is transcriptionally active in the epidermis before osteoblast differentiation. Although NF-κB acts downstream of multiple signaling pathways and labels multiple tissues, including the fish lateral line cells and immune cells in the skin, NF-κB activity can be specifically detected in a primordium that precedes the specification of scale osteoblasts by about a day. As scales grow and mature, NF-κB activity becomes restricted to the epidermis overlying the posterior edge of the scale (Fig. 1B,E,F). We estimate the wavefront to have a size comparable with that of a placode, which is roughly 20-30 cells in diameter (∼100 μm). Thus, about a few hundred cells come together to form a dermal condensate, which then gives rise to osteoblasts (Fig. 1B,H,I). Here, we chose to focus on scale development immediately anterior to the caudal fin, as this area was most amenable to our imaging methods. Scale formation proceeds in a very organized AP pattern in this region.

Using this line as a reporter, we investigated whether scale development and Eda/NF-κB activity traveled as an active or passive wave. In the case of an active wave, communication between cells is required to spread the signal throughout a tissue, whereas in a passive wave, each scale is specified autonomously in a predetermined temporal pattern. These two models can be differentiated by their behavior when a barrier is introduced to the system (Deneke and Di Talia, 2018; De Simone et al., 2021). As active waves require a (diffusible) signal for cell-cell communication and some form of positive feedback to propagate across a tissue, the signal will be unable to pass through the barrier and scale development will be halted. A passive wave, on the other hand, does not require cell-cell communication as it is controlled by pre-determined delays and thus the wave can cross the barrier and scale development will continue unaltered (Fig. 2A).

Previous studies have found that cutting explanted chick embryo skin immediately ahead of newly formed placodes did not impede the formation of new placodes, suggesting that in that system the wave might be passive (Davidson, 1983a,b). However, the timing of these experiments is key, and these studies predated the existence of live molecular markers for feather formation, so it is possible that the cut might have been introduced after the initiating wave had already traveled across the cut region. Consistent with this scenario, subsequent experiments introducing incisions further from the initiating line of feather placodes blocked the formation of new placodes, indicating that the wave might in fact be an active one (Patel et al., 1999).

To determine whether NF-κB waves during scale formation in zebrafish are active or passive, we made a shallow incision with a razor blade along the DV axis of juvenile fish to interrupt cell-cell communication. Cuts were made posterior to Nfkb:eGFP activity and scale formation at a stage in which scale induction proceeds anterior to posterior in the trunk of the fish (Fig. 2B). Fish were imaged once a day until the cut fully healed.

In each fish, cuts caused a significant delay in the formation of new scales. Before cut healing, no scales formed posterior to cuts that bisected the entire AP axis. Although the injury did impact the growth of scales in the cut site, scales anterior to the cut continued to develop and grow normally. Only after the cuts were fully healed did new scales begin to form posterior to the cut site. We also performed cuts that did not completely bisect the AP axis, allowing signaling to propagate above the cut. Unlike the full cuts, NF-κB activity and scale development proceeded around the cut and continued in the expected pattern (Figs 2 and 3). Importantly, cuts made parallel to the AP axis did not impact scale development outside of the cut site (Fig. S1), indicating that the delay in scale formation is not triggered by a global injury response.

These results suggest that new scales only form in response to a signal from neighboring cells, rather than being instructed by a pre-existing pattern. If the cuts had no effect on scale development outside of the injury site, this would suggest that a cell-autonomous mechanism dictated the timing of scale development. The fact that the experimental cuts impeded the traveling of NF-κB activity and subsequent scale development suggests that scales develop via an active wave mechanism. Furthermore, as scales only formed following the emergence of NF-κB activity, this suggests that NF-κB may itself act as the primary factor instructing scale initiation.

To further examine the role of NF-κB in this system, we used the pharmacological inhibitor Bay11-7085 to manipulate NF-κB activity during scale development (Drew et al., 2007; Mishra et al., 2020). Fish expressing Nfkb:eGFP and osx:H2A-mcherry were imaged and treated with Bay11-7085 over the course of several days. We found that Bay11-7085 caused a reduction in Nfkb:eGFP signal and a significant reduction in the rate of scale formation and growth (Fig. 4A-C). A transient treatment with Bay11-7085 revealed that scale development stopped during treatment but recovered following drug washout and the return of NF-κB activity. Moreover, this temporary inhibition did not significantly affect the patterning of scale development, as new scales formed sequentially in the pre-established pattern set before the inhibition of NF-κB (Fig. 4D-F).

In an active model in which NF-κB contributes to the mechanism triggering the wave, the speed of the wave depends on the timescale of self-sustained NF-κB activation. In a passive wave model, however, NF-κB does not participate in wave propagation, as the wave is predetermined by an upstream factor, and thus the speed of the traveling wave would be independent of NF-κB activation kinetics. We found a significant correlation between NF-κB activity and the rate of scale formation (used here as a proxy for wave speed), arguing that NF-κB activity contributes to the propagation of the wave. This correlation and the observation that NF-κB activity precedes new scale formation suggest that NF-κB activity plays a central role in the timing of scale development (Fig. 4G).

As NF-κB activity propagates as an active traveling wave to induce scale formation, we suggest that NF-κB promotes downstream signals promoting both the activation and inhibition of the Eda/NF-κB pathway to create a transient signal that can move across the skin (Gelens et al., 2014; Di Talia and Vergassola, 2022). As such, active NF-κB in a cell must be able to promote the pathway in a neighboring cell, likely through the production and release of extracellular ligand(s). As the developmental ligand Eda is upstream of NF-κB (Fig. 1G), it is natural to consider the possibility that it is involved in the positive feedback driving the observed NF-κB activity wave. Consistent with this, a wave of eda expression has been observed during feather formation. However, how this wave is propagated is unclear (Aman et al., 2018; Ho et al., 2019). In zebrafish, NF-κB is active in the epidermis, where the Eda receptor is expressed, whereas eda is solely expressed in the dermis (Harris et al., 2008; Aman et al., 2018); it is therefore unlikely that NF-κB can directly regulate Eda transcription. Thus, a simple feedback loop of Eda activating NF-κB to produce more Eda is improbable. This suggests that the wave is propagated through an intermediary molecule that is expressed in the epidermis and active in the dermis. A strong candidate to participate in the wave is Wnt/β-catenin signaling, which is required for osteoblast differentiation, is active in the epidermis and dermis during scale development, and is capable of activating Eda in other systems (O'Brown et al., 2015; Aman et al., 2018; Jacob et al., 2021). Wnt signaling is required for scale development and is active in the placode before osteoblast differentiation, as shown by the expression of its targets Lef1 and Axin in the epidermis and dermis of the developing placode before scale formation (Aman et al., 2018).

To investigate the potential role of Wnt/β-catenin signaling in this process, we used the pharmacological inhibitor IWR-1-endo, which has been shown previously to affect scale development in juvenile zebrafish (Aman et al., 2018; Iwasaki et al., 2018; Jacob et al., 2021). Similar to blocking NF-κB activity, inhibition of Wnt signaling halted scale formation and the formation of new NF-κB placodes (Fig. 4H). Upon release of inhibition, scale and placode development proceeded as normal (Fig. 4I,J), as observed following NF-κB transient inhibition. The similar effects of inhibition of Wnt and Eda/NF-κB signaling on scale formation suggests that they might cooperate in the propagation of a signaling wave that triggers a wave of scale patterning.

In addition to Wnt and Eda, Fgf signaling is also required for scale development (Harris et al., 2008; Daane et al., 2016; Aman et al., 2018). A null mutation in fgfr1a results in abnormally large scales, but in combination with an fgf20a mutation, scales are smaller and fewer, suggesting that multiple Fgfs and Fgfrs regulate scale development (Daane et al., 2016). In other systems Fgf20 has been shown to act downstream of Eda and to promote mesenchymal aggregation and placode formation (Kowalczyk-Quintas and Schneider, 2014; Sadier et al., 2014; Ho et al., 2019). In zebrafish, Fgf20 may act as a chemoattractant to drive dermal cell condensation, which occurs in a wave-like pattern (Aman et al., 2021).

Using a transcriptional reporter of fgf20a expression (Cox et al., 2018), we found that fgf20a expression closely mirrors NF-κB activity. Fgf20a is expressed early in the scale placode before osterix (also known as sp7) expression and then becomes restricted to the exterior edge of the scale as the tissue grows. Following inhibition of NF-κB, formation of new fgf20a positive placodes stopped, indicating that NF-κB is required to activate fgf20a expression (Fig. 5). We also found that inhibiting Fgfr and NF-κB activity blocked the formation of dermal cell condensates (Fig. S2). These results indicate that NF-κB most likely activates Fgf20a, which then promotes dermal cell migration and condensation to drive scale development.

Next, we used the pan-Fgfr inhibitor BGJ398 (Aman et al., 2018; De Simone et al., 2021) to analyze the contributions of Fgf signaling to both dermal condensation and wave propagation. Treatment with this inhibitor completely blocked scale development as anticipated but, similar to observations in feather formation (Ho et al., 2019), did not block the Eda/NF-κB activity wave from traveling. The Eda/NF-κB activity wave not only continued to travel along the AP axis but exhibited an ectopic pattern. Instead of distinct placodes of NF-κB transcription, we observed a continuous front of reporter expression (Fig. S3). This suggests that Fgf signaling is not required for the propagation of the Eda/NF-κB wave and may play a role in activating the signals responsible for turning off Eda and NF-κB (Fig. 6A-D).

As the Eda/NF-κB activity wave continues to travel in the absence of Fgf signaling, but no obvious placodes could be detected, we used a transient treatment with BGJ398 to test the effects on patterning of decoupling wave propagation and placode formation. Following drug washout, the extended propagation of the Eda/NF-κB activity wave led to significant changes in scale patterning. Scales formed post-washout arose more synchronously, resulting in a similar number of scales compared with controls (Fig. 6C,D). In addition, the spatial pattern of these new scales was significantly altered as well (Fig. 6E-H). Rather than a strict hexagonal arrangement, there was greater variation in the number of neighbors each scale had and the spacing between scales. This result suggests that the sequential timing of scale development is required to properly pattern scales. Thus, coupling the traveling Eda/NF-κB activity wave to dermal condensation and scale formation is necessary to facilitate accurate patterning and growth.

Zebrafish scales are an ideal system to study skin patterning as their external development and the use of transgenic reporters and live imaging are conducive to dynamical analyses. Developing scales form relatively late in ontogeny. Thus, they can be easily imaged to conduct long-term experiments and manipulating the processes that drive scale development is unlikely to affect the formation of other essential structures, which might in principle confound the analysis of these perturbations. As the pathways and developmental mechanisms that regulate skin appendage development are conserved across vertebrates, this system can improve our quantitative understanding of how these pathways lead to patterning.

Here, we have shown that skin appendage development in zebrafish is coordinated by the propagation of a biochemical wave. We have found that an active wave of NF-κB activity regulates the spatial and temporal patterning of scale development. Interfering with cell-cell communication using a physical barrier blocked the propagation of Eda/NF-κB activity and scale development. Although this is likely due to a reaction-diffusion process between nearby cells generating a traveling wavefront (Deneke and Di Talia, 2018), this experiment does not rule out the possibility of long-range diffusion or other mechanisms for cell-cell communication (e.g. mechanical communication).

Both NF-κB and Wnt activity are required for scale development and propagation of the wave. Fgf signaling, although required for scale development, is not involved in how this wave travels. Although it is possible to drive scale development through ectopic Fgf signaling in the absence of Eda, this results in aberrant scale patterning and ectopic osteoblast formation (Aman et al., 2018). As Fgf signaling is required for the formation of the dermal placodes that precede scale formation, these results demonstrate that the signaling waves and the cellular processes that lead to placode formation (for example, migration) must be tightly coordinated to properly pattern scale development (Fig. 7).

Our results suggest a mechanism by which the wave of Eda travels across the skin to drive this process. As NF-κB is active in the epidermis and Eda is expressed in the dermis, it is unlikely that NF-κB directly activates transcription of Eda (Aman et al., 2018). Thus, NF-κB likely induces Wnt, which can then diffuse to the dermis to trigger expression of Eda. Eda can in turn diffuse to the epidermis to activate NF-κB and complete the positive feedback loop. At the same time, NF-κB also induces Fgf20a, which can signal to drive dermal cell aggregation and possibly promote the inhibitors responsible for turning off Eda/NF-κB signaling late in scale development. Although these inhibitors have yet to be identified, Bmp signaling has been shown to inhibit Eda and could be acting to block scale formation (Kowalczyk-Quintas and Schneider, 2014; Sadier et al., 2014; Ho et al., 2019).

In recent years, signaling waves have emerged as a common mechanism for the regulation of developmental processes (Gelens et al., 2014; De Simone et al., 2021; Bailles et al., 2022). Waves can facilitate the rapid propagation of biochemical signals and help organize cellular responses across large tissues. In cellular systems, this process involves ligand release and often de novo production mediated by transcription. The timescales of such processes dictate the speed of the wave. In systems in which wave propagation involves transcription, the typical speed of traveling waves is about 10 μm/h, that is the wave travels from one cell to the other in ∼1 h (Deneke and Di Talia, 2018). Consistent with this idea, we found that the wave of Eda/NF-κB travels about 200 μm in a day (Fig. 6B; Fig. S3), suggesting that the wave is likely driven by the transcription and release of ligands at the wavefront.

Previous work has shown that the coupling of signaling waves and chemotactic cellular processes can produce regular patterns in vertebrate skin. During feather patterning, a wave of EDA lowers the threshold of cell density required for follicle development by transiently inducing FGF20, which is believed to act as a chemoattractant driving the aggregation of cells into condensates. Subsequently, the condensates upregulate FGF20 and BMP, which acts to inhibit FGF20 and limit condensate growth. As feathers are formed row by row, the formation and signaling set up by the previous row dictate the offset pattern of feather follicles in the adjacent row (Ho et al., 2019).

Similarly, in zebrafish scales, coupling a signaling wave to the formation of condensates is required for regularly patterned scales. Allowing the Eda wave to propagate while inhibiting cell aggregation leads to ectopic NF-κB activity and irregularities in scale patterning. This also suggests that Fgfs and/or cell aggregation are required for the production of the inhibitors that limit placode size and Eda signaling.

Cell shape has also been shown to be important for setting up the conditions to pattern a self-organized system. Specifically, cell shape anisotropy correlates with a higher level of cell motility, which is required to create an ordered pattern in feather development (Curantz et al., 2022). Although we did not investigate cell shape dynamics in this study, it is possible that they also influence pattern formation in zebrafish skin.

In other systems, cell density and mesenchymal aggregation are important for regulating wave dynamics and primordia formation (Bailleul et al., 2019; Ho et al., 2019). In this study, we did not observe significant differences in wave speed along the AP axis when dermal condensation was blocked (Fig. 6B). Interestingly, we did observe that the wave appears to travel faster along the AP axis than along the DV axis. Although the reasons for this difference in wave spreading are unknown, it is possible the lag is due to differences in cell migration or molecular diffusion in the AP versus DV directions. For this study, we chose to focus on scale development immediately anterior to the caudal fin, as this area was the most suitable for our imaging methods. It is possible that the wave dynamics are less AP directed in the anterior of the fish due to differences in body shape.

Our results indicate both the Eda signaling wave and cell migration are required for scale primordia formation and osteoblast differentiation. However, it is difficult to disentangle the role of Fgfs in osteoblast differentiation and cell migration. Although we found that a lack of cell condensation did not affect wave speed, it did impact the inhibition of Eda/NF-κB. As we did not completely restrict cell migration in this system, it is possible that a complete loss of cell movement could affect wave dynamics. An overproduction of Fgf signaling likely interferes with cell migration dynamics and causes ectopic osteoblasts to create disorganized bony tissue rather than regularly ordered scales (Aman et al., 2018). Conversely, loss of thyroid hormone limits dermal cell migration and causes a delay in scale development (Aman et al., 2021). It has yet to be investigated how this delay affects Eda/NF-κB wave dynamics.

Coupling wave dynamics and chemotactic signals also appears to influence emerging tissue structures, such as tooth emergence. Previous work has found that Edar expression dynamics and chemotaxis regulate placode formation and retraction in tooth development (Sadier et al., 2019). In scale development, placodes form sequentially and thus their pattern is likely dictated by the placodes formed immediately prior. It is also likely that this patterning influences the final size and shape of mature scales, but this has yet to be fully investigated.

In summary, we have shown the efficacy of using zebrafish as a model for vertebrate skin appendage development and demonstrated that a wave of Eda/NF-κB activity is required to form the precise hexagonal pattern in scale development. We decoupled wave propagation from the formation of the placodes by transiently inhibiting Fgf signaling. Consequently, the Eda/NF-κB wavefront travels across a large region of tissue but no placode forms. Following washout of the inhibitor, a large region of the tissue becomes competent to form placodes simultaneously. As a result, several scales form at the same time but they do so in a less precise and regular pattern. Thus, we propose that a biochemical wave is upstream in the patterning process and that its coupling with the mechanochemical process of placode formation ensures that sequential formation of scales proceeds in precisely controlled hexagonal array.

Zebrafish (Danio rerio) of Ekkwill and Ekkwill/AB strains were maintained at 26-28.5°C with a 14 h light:10 h dark cycle. Juvenile fish aged 3.5-5 weeks postfertilization were used for experiments. Both males and females were used for experiments and randomly assigned to control and experimental groups. As scale development does not start at an exact age, all fish were checked for the onset of placode or scale development before use in experiments. For each experiment, control and experimental groups were taken from the same clutch. All fish experiments were approved by the Institutional Animal Care and Use Committee at Duke University and followed all the relevant guidelines and regulations. Transgenic lines used in this study were Tg(osx:H2A-mCherry)_pd310 (Cox et al., 2018), Tg(fgf20a_EGFP)/HGN21A (Shibata et al., 2016), Tg(6xHsa.NFKB:EGFP) (Kanther et al., 2011).

Live juvenile fish were imaged using a Leica Sp8 confocal microscope and LAS X 2.01.14392 software with a 10× air objective at 1.25× zoom. Fish were anesthetized with 0.015% tricaine (Sigma-Aldrich, E10521-50G) in system water and transferred to a 1% agarose bed in a 40 mm petri dish. After imaging, fish were revived and placed back in their original tanks. For longitudinal time courses, fish were imaged once every 24 h for 3-10 days. All images focused on the trunk of the zebrafish, immediately anterior to the fin. As the area of scale development is larger than the field of view of our microscopy setup, multiple overlapping z-stacks (1-3 with variable number of plans) were used. Images were acquired at 1024×1024 resolution (0.909 μm pixel size) and a z-step of 0.909 μm. Fluorescent proteins were imaged using the following lasers: Nfkb:eGFP, 488 nm; osx:H2A-mCherry, 561 nm; fgf20a:GFP, 488 nm.

Fish expressing Nfkb:eGFP and osx:H2a-mcherry were anesthetized using 0.015% tricaine in system water. Once fish were fully anesthetized (checked by tail pinch test), fish were transferred to a cotton pad soaked with the same concentration of tricaine. A fluorescent dissecting scope was used to visualize the Nfkb:eGFP and thin, shallow incisions were made posterior to Nfkb placodes using a razor blade. Fish were imaged using our live imaging setup immediately after incisions were made. Following imaging, fish were revived and placed back on the system.

Fish were imaged on day 1 before treatment, then placed in aquarium water with the pharmacological compound or vehicle control diluted to working concentration for the duration of treatment. Fish were maintained off the aquarium system in the dark. For Bay11-7085 and IWR-1-endo, fish were treated continuously in aquarium water off system. Fish were fed and the treatment medium was refreshed once every 24 h. Bay11-7085 (ApexBio, B3033) inhibits IκBα phosphorylation, which prevents NF-κB from entering the nucleus to activate transcription. The efficacy of inhibition was determined by using the Nfkb:eGFP reporter line, which showed a loss of eGFP signal after full inhibition. Fish were treated with 0.5 μM Bay11-7085 or 0.0005% ethanol in 500 ml of aquarium water off system. IWR-1-endo (Cayman Chemical Company, 13659) broadly inhibits Wnt signaling through Axin stabilization. Fish were treated with 10 μM IWR-1-endo or 0.01% DMSO in 100 ml of aquarium water off system.

For BGJ398 (SelleckChem, S2183), fish were treated in 100 ml of aquarium water off system for 1 h per day. Between treatments, fish were fed once every 24 h and kept in 500 ml of water off system. BGJ398 is a pan-FGFR inhibitor. Fish were treated with 1μM BGJ398 or 0.001% DMSO in 100 ml of aquarium water off system.

For transient treatments, fish were treated post-imaging on day 1 until post-imaging on day 3 (around 48 h). After washout, fish were rinsed with fresh water and placed back on the system. For treatment followed by DAPI staining, fish were live imaged before treatment to observe Nfkb:eGFP signal. Fish were then treated for 48 h as established above with either BGJ398 or Bay11-7085. Fish were then fixed and stained with DAPI and imaged with the same settings.

Whole juvenile fish were fixed using 4% paraformaldehyde (PFA) for 48 h at 4°C, then rinsed with PBS and transferred to a 30% glucose solution overnight at 4°C. Tissues were frozen in Tissue Freezing Medium (General Data, TFM-5) and sectioned using a cryostat at 14 μm/section. Slides were rehydrated in PBS with 0.1% Tween and stained with DAPI. Sections were imaged using a 20× oil objective on a Leica Sp8 confocal microscope.

For wholemount DAPI staining, fish were fixed using 4% PFA for 48 h at 4°C, then rinsed with PBST for 3×5 min. Juveniles were incubated in DAPI for 20 min, then rinsed in PBS and imaged using the same settings as the live imaging methods above.

Images from live juveniles were stitched using a custom MATLAB (Mathworks) script adapted from De Simone et al. (2021). Each acquired z-stack contained a portion of the fish trunk, positioned diagonally with respect to the x-y plane. Overlapping z-stacks were stitched into a single z-stack using image cross-correlation and position data provided by the LAS X software. Images of histological sections were assembled using the Stitching plugin in ImageJ (Preibisch et al., 2009). For longitudinal experiments, time points were registered to each other either manually or using the imregister function in MATLAB, which aligns images on a single coordinate plane based on image intensity. To register images obtained on different days, images were placed over a black background so that images from each day were the same size in x and y. Registration settings were adjusted as needed for accurate alignment and checked manually. Nfkb or fgf20a placodes and scales were counted manually in ImageJ. As images obtained were not always the exact same size, the same size area was used between fish of the same experiment. Scale area and centroid location were obtained manually in ImageJ using the polygon function to outline each scale. Centroid locations were used to construct a Voronoi tessellation in MATLAB. The output of the tessellation was used to determine the number of scales adjacent to a single scale and the distance between neighboring scales.

NF-κB activity was calculated by taking the integral of the GFP intensity profile for Nfkb:eGFP. The intensity profile was obtained by measuring the intensity of the Nfkb:eGFP channel across a single row of developing placodes along the AP axis. When integrating, lines of the same length were used across fish. To determine the distance travelled by the Nfkb:eGFP reporter, the posterior edge of the wave or most posterior placode of the bottom row was used to mark the NF-κB wave.

All tests were carried out in MATLAB using standard functions for unpaired t-tests or Chi-square tests. Some animals were excluded from analysis based on aberrations in Nfkb:eGFP activity or scale development. Some treated fish did not respond to treatment (no visible decline in reporter activity or effect on scale development) or controls did not form scales at a normal rate (and had a similar lack of reporter activity). These were excluded from analysis due to likely underlying biological differences. Group sizes were based on our experience and individual group standard deviations.

We thank Ken Poss and John Rawls for sharing fish strains. We thank Jim Burris, Lawrence Frauen and Colin Dolan for zebrafish care, and Duke University Zebrafish Core Facility staff. We also thank the Duke Light Microscopy Core Facility for technical support. We thank Michel Bagnat, Bernard Mathey-Prevot, David McClay and all members of the Di Talia lab for comments on the manuscript. We thank Priyom Adhyapok, Boris Shraiman and Massimo Vergassola for discussions on the physical mechanisms of skin patterning. We thank Alessandro De Simone for help with computational image analysis.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201866.reviewer-comments.pdf