ABSTRACT
Adult zebrafish fins develop and robustly regenerate an elaborately branched bony ray skeleton. During caudal fin regeneration, basal epidermal-expressed Sonic hedgehog (Shh) locally promotes ray branching by partitioning pools of adjacent progenitor osteoblasts (pObs). We investigated if and how Shh signaling similarly functions during developmental ray branching. As during regeneration, shha is uniquely expressed by basal epidermal cells (bEps) overlying pOb pools at the distal aspect of outgrowing juvenile fins. Lateral splitting of each shha-expressing epidermal domain followed by the pOb pools precedes overt ray branching. We use ptch2:Kaede fish and Kaede photoconversion to identify short stretches of shha+ bEps and neighboring pObs as the active zone of Hh/Smoothened (Smo) signaling. Basal epidermal distal collective cell migration continuously replenishes each shha+ domain with individual cells transiently expressing and responding to Shh. In contrast, pObs have constant Hh/Smo activity. Hh/Smo inhibition using the small molecule BMS-833923 (BMS) prevents branching in all fins, paired and unpaired, with minimal effects on fin outgrowth or skeletal differentiation. Staggered addition of BMS indicates Hh/Smo signaling acts throughout the branching process. shha+ bEps and pObs are tightly juxtaposed at the site of Hh/Smo signaling, as with regenerating fins. We use live time-lapse imaging and cell tracking to find Hh/Smo signaling restrains the distal migration of bEps by apparent ‘tethering’ to pObs. We conclude short-range Shh/Smo signaling enables ray branching by re-positioning pObs during both fin development and regeneration. We propose instructive basal epidermal collective migration and Shh/Smo-promoted heterotypic cell adhesion between bEps and pObs directs fin skeleton branching morphogenesis.
INTRODUCTION
Teleost fish fins have elaborately patterned skeletons comprised of bony rays, or lepidotrichia, that shape and structurally support fin appendages. The diversity and beauty of fish fins has long captivated aquarists while serving as a compelling model to consider morphological evolution. Danio rerio zebrafish are widely studied ray-finned teleost fish with branched rays in all paired and unpaired fins. For example, zebrafish caudal fins typically have 18 rays of which the central 16 branch into “daughter” rays (Figure 1 Supplement 1A). Juvenile fish form primary branches around 30 days post fertilization (dpf) followed by secondary and tertiary branches. Individual rays comprise two opposed hemi-rays that form cylindrical skeletal units segmented by joints and enveloped by a multilayered epidermis (Figure 1 Supplement 1B). Teleost fins and tetrapod vertebrate limbs evolved from a common ancestral appendage (Dahn et al., 2006; Freitas et al., 2006) with rays possibly sharing deep homology with digits (Nakamura et al., 2016). The relative simplicity of the fin’s skeletal structure makes zebrafish a valuable model for understanding mechanisms of appendicular skeletal patterning.
Zebrafish fully regenerate adult fins including restoring a branched ray skeleton within two weeks of injury. Empowered by versatile genetic and other tools, zebrafish have become a leading model for appendage regeneration research. Collective studies implicate many of the same signaling pathways,including Wnt, Bmp, Fgf, and Hh, involved in tetrapod limb development (Sehring et al., 2016; Stewart et al., 2014a; Wehner et al., 2014). Therefore, zebrafish fin development and regeneration provide accessible contexts to understand how cell signaling patterns the appendicular skeleton and how the pathways are reactivated for repair. Further, zebrafish fins and their rays enable studies of fundamental developmental questions, including how branched networks form – a common property of many organs including vasculature, lungs, kidneys, mammary glands, and pancreas.
Sonic hedgehog (Shh) signaling through its Smoothened (Smo) effector is one pathway dually involved in fin regeneration and appendage development. Shh is associated closely with tetrapod limb skeletal patterning (Zuniga, 2015) and Shh/Smo pathway perturbations cause syndactyly and polydactyly (Anderson et al., 2012; Malik, 2012). Shh is the Zone of Polarizing Activity (ZPA) secreted morphogen that pre-patterns the limb field into distinct skeletal units, including digits (Chiang et al., 2001; Riddle et al., 1993; Saunders & Gasseling, 1968). Zebrafish shha expression studies suggest a ZPA patterns pectoral and other paired fins but not unpaired fins including the caudal fin (Hadzhiev et al., 2007; Laforest et al., 1998). Nevertheless, shha is expressed in distal epidermal domains overlying each forming ray during caudal fin development and regeneration (Armstrong et al., 2017; Hadzhiev et al., 2007; Laforest et al., 1998; Y. Lee et al., 2009; Zhang et al., 2012). At least during regeneration, each shha-expressing domain splits prior to ray branching. We leveraged the highly specific Smo inhibitor BMS-833923 (BMS) to show Shh/Smo specifically promotes ray branching during zebrafish fin regeneration (Armstrong et al., 2017). However, a potential Shh/Smo role during developmental ray branching and underlying mechanisms are unresolved.
During caudal fin regeneration, distal-moving basal epidermal cells (bEps) adjacent to bony rays upregulate shha at the distal “progenitor zone” (Armstrong et al., 2017). Shh-expressing bEps activate Hedgehog/Smoothened (Hh/Smo) signaling in themselves and immediately adjacent progenitor osteoblasts (pObs) as marked by patched2 (ptch2), which encodes a Hh receptor and universal negative feedback regulator (Alexandre et al., 1996; Goodrich et al., 1996; Lorberbaum et al., 2016; Marigo et al., 1996). This short-range Hh/Smo signaling is required to split pOb pools and therefore ray branching without impacting proliferation or differentiation (Armstrong et al., 2017). While Hh/Smo signaling is active continuously, shha-expressing basal epidermal domains split laterally prior to pOb partitioning. We proposed Shh/Smo signaling might enhance physical associations between moving epidermal cells and pObs to enable the progressive partition of pOb pools (Armstrong et al., 2017).
Here, we explore the mechanisms underlying developmental fin ray branching in juvenile zebrafish. We show basal epidermal dynamics as well as Shh/Smo activity and function are largely the same as during regeneration. We use transgenic reporter lines for shha and its target gene ptch2 to refine developmental expression profiles. Kaede photoconversion of TgBAC(ptch2:Kaede) fish reveals continuous Shh/Smo signaling in distal ray Shh+ basal epidermal domains and neighboring pObs. We inhibit Shh/Smo signaling using the small molecule BMS-833923 to show the pathway is largely dedicated to ray branching in all fins, including the paired pectoral fins. Shh+ bEps and pObs are closely apposed at the site of Shh/Smo signaling where a basement membrane is incompletely assembled. bEps constantly move distally, trafficking through while contributing to shha-expressing domains that split laterally prior to ray branching. We use live time-lapse imaging to demonstrate Shh/Smo signaling restrains basal epidermal distal migration, possibly by promoting transient adhesion of shha+ bEps to distal ray pObs. We conclude the collective migration of bEps, constantly distal with progressive lateral domain splitting, and their atypical use of local Shh/Smo signaling re-positions pObs for skeletal branching during both fin development and regeneration. This reflects a unique branching morphogenesis process whereby movements of a neighboring cell type – the bEps – guides the tissue-forming cells – the pObs – into split pools.
RESULTS
Shha expression is progressively restricted to distal ray basal epidermal domains that split preceding ray branching
Developing fins express sonic hedgehog a (shha) in single basal epidermal domains adjacent to the tip of each hemi-ray up to 20 days post fertilization (dpf) (Hadzhiev et al., 2007; Laforest et al., 1998). During caudal fin regeneration, similar shha-positive basal epidermal domains split into distinct domains around 4 days post amputation (dpa) immediately preceding ray branching (Armstrong et al., 2017; Laforest et al., 1998; Y. Lee et al., 2009; Quint et al., 2002; Zhang et al., 2012). We expected a comparable pattern during fin development if branching during fin regeneration recapitulates developmental processes. We used the reporter line Tg(-2.4shha:gfpABC)sb15 (shha:GFP; (Ertzer et al., 2007; Zhang et al., 2012) to monitor shha expression in live zebrafish from its emergence through primary ray branching in juveniles around 30 dpf. As is well-established, shha:GFP expression in the caudal region was restricted to the developing floor plate and notochord through 5 dpf (Krauss et al., 1993). From 5-9 dpf, shha:GFP expression expanded into the caudal fin fold primordium through a ventral gap of melanophores (Figure 1A). By 9-10 dpf, shha:GFP+ cells were enriched along emerging immature rays (Figure 1B). At 10-12 dpf, shha:GFP expression became specifically associated with the distal aspect of maturing central rays while remaining along the lengths of immature peripheral rays (Figure 1C). shha:GFP was restricted to ray tips by 14 dpa, when all rays contained maturing, segmented bone (Figure 1D).
The shha:GFP domains began splitting around 30 dpf, immediately prior to branching of the corresponding ray (Figure 1I, I’, white brackets). We used whole mount immunostaining for GFP and the basal epidermal marker Tp63 (H. Lee & Kimelman, 2002) and 3D confocal reconstructions to confirm shha-driven GFP was expressed exclusively in basal epidermal cells (bEps) (Figure 1K, L; expanded data in Figure 1 Supplement 2). Where Tp63-marked cells were multi-layered (magenta dashed line, Figure 1K), only the innermost cells expressed GFP. Similar shha expression patterns, including shha+ bEp domain splitting, support a common Shh/Smo- dependent mechanism for both developmental and regenerative ray branching.
Ptch2 expression indicates Hh/Smo activity in basal epidermis and progenitor osteoblasts
We next assessed expression of ptch2, a Hh/Smo negative feedback regulator and activity marker (Alexandre et al., 1996; Goodrich et al., 1996; Lorberbaum et al., 2016; Marigo et al., 1996). Hh/Smo signaling induces ptch2 in pObs and neighboring bEps during caudal fin regeneration (Armstrong et al., 2017; Quint et al., 2002). ptch2 is also expressed in distal regions of late larval caudal fins, although cell-level expression is unresolved (Laforest et al., 1998). We imaged the TgBAC(ptch2:kaede)a4596 reporter line (ptch2:Kaede; (Huang et al., 2012) at the same developmental time points we examined shha:GFP. ptch2:Kaede expression was confined to the notochord and floor plate until ∼9 dpf (Figure 1E). By 9-10 dpf (Figure 1F), ptch2:Kaede was associated with nascent rays in the ventrally expanding fin primordia. At 10-12 dpf (Figure 1G), ptch2:Kaede was expressed the entire length of each ray. This pattern persisted through 14 dpf (Figure 1H) with notably higher ptch2:Kaede expression in joints and distal tip of each ray, matching its pattern in regenerating fins (Armstrong et al., 2017). As with shha:GFP, ptch2:Kaede+ domains split as ray branching initiated in 30-33 dpf juvenile fish (Figure 1J, J’).
To define the cell types expressing ptch2 during caudal fin development, we combined ptch2:Kaede with Tg(runx2:mCherry) (runx2:mCherry; Shannon Fisher Lab, unpublished) and shha:GFP to mark runx2+ pObs and shha+ bEps, respectively (Figure 1M-P and expanded data in Figure 1 Supplement 3). Confocal imaging of live 19 dpf double transgenic larval fish showed ptch2:Kaede co-localized with distal fin runx2:mCherry-expressing pObs (Figure 1M, N). Only adjacent and distal-extending presumptive bEps additionally expressed ptch2:Kaede. We used photoconversion to discern Kaede from GFP to reveal non-pOb ptch2:Kaede co-localized with shha:GFP-expressing bEps (Figure 1O, P). Therefore, ptch2 defines autocrine (in bEps) and short-range (in pObs) Shh/Smo signaling during caudal fin development.
Active Shha/Smo signaling is restricted to outgrowing distal ray regions
We photoconverted distal ray ends of 25 dpf ptch2:Kaede caudal fins and re-imaged 24 hours later to distinguish actively produced Kaede from perduring reporter fluorescence (Figure 2A-C). New, unconverted Kaede was produced exclusively in short, discrete domains at the ray tips. Therefore, active Shh/Smo signaling appears narrowly focused in close proximity with shha-expression bEps. We also observed a stretch of photoconverted Kaede+ cells distal to the ray tips in tissue newly formed over the 24-hour post-conversion period. We observed the same pattern during fin regeneration, suggestive of distal migration of previously Shh/Smo-responsive bEps that ceased ptch2 expression when moving beyond the Shh/Smo active zone (Armstrong et al., 2017).
We used the potent Smo inhibitor BMS-833923 (henceforth abbreviated BMS) (Armstrong et al., 2017; Lin & Matsui, 2012) to confirm ptch2:Kaede reports Shh/Smo activity during fin development. As expected, caudal fins of 25 dpf fish treated with 1.25 µM BMS and photoconverted 3 hours later produced no new Kaede 24 hours post-conversion (hpc; Figure 2 Supplement 1G-L, n=8/8). Curiously, we no longer observed distally displaced photoconverted Kaede+ bEps and the remaining photoconverted Kaede had weakened since conversion (Figure 2 Supplement 1H, I, K, L). We surmise the 24-hour Shh/Smo-inhibition caused Kaede+ bEps to shed prematurely from the fin and therefore only photoconverted Kaede+ pObs remained. Therefore, Shh/Smo signaling may retard distal bEp collective movements.
Sustained Shh/Smo signaling promotes ray branching during fin development
We next investigated if Shh/Smo is required for ray branching during fin development as during regeneration (Armstrong et al., 2017). Tg(sp7:EGFP)b1212 osteoblast reporter fish (sp7:EGFP; DeLaurier et al., 2010) treated with 0.63 µM BMS from 25 to 42 dpf failed to branch their caudal fin rays (Figure 3A-D, n=5 per BMS and DMSO control groups). In contrast, Shh/Smo-inhibition did not disrupt caudal fin outgrowth or skeletal maturation of the central 16 rays (Figure 3 Supplement 1). Curiously, the non-branching principal peripheral rays uniquely were shorter in BMS-treated fish (Figure 3A, C white arrows and Figure 3 Supplement 1). BMS-treatment of 29 dpf shha:GFP fish exposed to EdU for 12 hours did not change the fraction of EdU+ intra-ray cells, i.e. pObs and mesenchyme nestled between the epidermal Shh domains of each hemi-ray (Figure 3 Supplement 2, n=5 per group). We conclude Shh/Smo signaling is largely dedicated to ray branching with minimal proliferation or bone maturation effects during both fin development and, as shown previously, regeneration (Armstrong et al., 2017).
Shh/Smo signaling may act transiently to initiate ray branching or continuously during the branching process. To distinguish between these possibilities, we staggered the start of BMS treatment to “before”, “during”, or “after” branching (Figure 3E), identified by a priori screening 24-35 dpf sp7:EGFP clutchmate fish. Expectedly, BMS-exposure initiated prior to ray branching prevented said rays from branching (“before” group, Figure 3F, G, G’, n=4/4) and rays that had already fully branched remained so after BMS treatment (“after” group, J, K, K’, n=4/4). However, rays that recently initiated branching (“during” group, Figure 3H) re-fused upon BMS exposure, forming “gapped” ray segments (Figure 3I, I’, n=4/5). Therefore, sustained Shh signaling acts throughout ray branching morphogenesis rather than as a switch that initiates branching.
Shh/Smo signaling does not substantially contribute to initial fin ray patterning
shha and ptch2 expression during the initial stages of ray formation (Figure 1B, E) suggest Shh/Smo influences early fin skeletal patterning in addition to promoting later, juvenile-stage ray branching. To explore this possibility, we inhibited Shh/Smo signaling from as early as 2 dpf, when the larval fin fold entirely comprises soft tissue absent of any ray structures. As expected, ptch2:Kaede-marked Shh/Smo signaling was restricted to the notochord and floor plate (Figure 3 Supplement 3A-C). Photoconversion experiments confirmed BMS fully inhibited production of new ptch2:Kaede in 14 dpf larval caudal fins (Figure 3 Supplement 3D-E’, total n=33-44 per group), as with embryos, juvenile fins, and regenerating adult fins (Figure 2 Supplement 1G-L; (Armstrong et al., 2017). We treated sp7:EGFP;runx2:mCherry fish with 1.25 μM BMS from 2 until 14 dpf, when all 18 rays were clearly established. Their caudal fins developed the correct complement of 18 rays, each which largely maintained position, dorsal-ventral pattern, and identity (Figure 3 Supplement 3F-G’). As such, Shh/Smo may have only minor or no roles in initial caudal fin skeletal patterning.
Shh/Smo signaling promotes ray branching in all fins
Zebrafish have 3 unpaired (dorsal, anal, and caudal) and 4 paired (pectoral and pelvic) fins, all which have a branched dermoskeleton. We tested if Shh/Smo signaling promotes ray branching in all seven fins by treating shha:GFP;runx2:mCherry fish with BMS starting at 21 dpf, prior to asynchronous ray branching across fins. Both DMSO control and BMS-treated fish showed shha:GFP+ domains at the distal end of every ray of all fins at 42 dpf (Figure 4). However, BMS-treated fish showed no or, at best, severely delayed and sporadic branching in all fins (n=6 per group). Therefore, all fins employ a common Shh/Smo signaling-dependent mechanism for ray branching regardless of evolutionary or morphological divergence.
Shha+ bEps and pObs are intimately associated in developing caudal fins
We next aimed to identify how sustained, local Shh/Smo signaling affects bEp and/or pOb cell behaviors to promote ray branching morphogenesis. The close proximity of these two Shh-responsive cell types suggested their movements might be physically coupled in a Shh/Smo-dependent manner to promote branching. To assess potential physical contacts between bEps and pObs, we first stained longitudinal sections of 32 dpf juvenile fin rays from shha:GFP fish with GFP, the osteoblast marker Zns-5, and Laminin, a component of the epidermal-osteoblast separating basement membrane. As expected, Shha:GFP+ bEps were directly adjacent to pObs (Figure 5 Supplement 1). A thin Laminin-containing basement membrane separated pObs and the proximal-most Shha:GFP+ bEps that had recently arrived in the active zone and initiated shha expression. More distally, the double staining for Shha:GFP+ bEps and Zns-5+ pObs produced even partially overlapping signal (Figure 5 Supplement 1D and D’), suggesting the two cell types are intimately associated. Here, the laminin+ basement membrane was less dense and sometimes fragmented, likely reflecting its nascent production (asterisks, D’).
We further explored the relative positioning of bEps and pObs at the onset of ray branching by 3D confocal reconstructions of live imaged fins of 28 dpf shha:GFP;runx2:mCherry fish (Figure 5A-C). shha:GFP+ bEps and runx2:mCherry+ pObs were tightly juxtaposed in both hemi-rays of a single lepidotrichia (Figure 5 Movie 1). Focusing on one hemi-ray, we observed extensive apparent heterotypic surface contacts, including areas where shha:GFP+ bEps enshrouded a ridge of pObs (Figure 5 Movie 2). Single sagittal optical slices and reconstructed slice equivalents examined multi-dimensionally (Figure 5D-F) showed intertwined bEps and pObs unresolvable by conventional confocal microscopy.
We considered if Shh/Smo signaling promotes the close juxtaposition of bEps and pObs. However, BMS treatment of shha:GFP;runx2:mCherry fish from 24-34 dpf did not alter the intimate association between Shh+ bEps and Runx2+ pObs in static images of live fins even though the same drug exposure prevented ray branching to 42 dpf (Figure 5 Supplement 2). As expected, Runx2+ pOb pools failed to split upon BMS exposure. Interestingly, the shha:GFP domains of BMS-treated fish variably remained as one cluster per hemi-ray (Figure 5 Supplement 2E, J; 4/10 split, 6/10 un-split for dorsal ray 3). In contrast, our fin regeneration study indicated Shha-domain splitting is always Shh/Smo independent (Armstrong et al., 2017). We speculate Shh/Smo signaling has an ancillary role in epidermal domain branching that is less apparent in larger rays, including of adult fins. Regardless, Shh/Smo signaling does not support ray branching by promoting close proximity between bEps and pObs per se.
Shh/Smo signaling restrains basal epidermal collective movements while adjacent to pObs
Shh/Smo inhibition appeared to increase the rate of bEp shedding due to accelerated distal collective movements (Figure 2 Supplement 1I, L). Therefore, we considered if Shh/Smo promotes transient adhesion (direct or indirect) between bEps and pObs that impedes bEp movements while they neighbor relatively stable pObs. Such regulated heterotypic adhesion, which may not be evident by static imaging, coupled with force-generating bEp collective movements during shha+ domain splitting, could re-position pOb pools over time. We time-lapse imaged caudal fins of ptch2:Kaede fish at late larval stages (22-24 dpf) to assess heterotypic cellular dynamics in outgrowing rays. ptch2:Kaede+ bEps moved distally over ptch2:Kaede+ pObs, which remained stationary over the 30 minute imaging period, and in more distal fin tissue (Figure 6 Movie 1, Figure 6 Supplement 1A-C). We used semi-automated tracking of individual ptch2:Kaede+ bEps to determine ptch2:Kaede+ bEps of BMS treated fish moved significantly faster (3-6 cells per fish and n=8 fish per group) (Figure 6 Supplement 1D). Therefore, Shh/Smo signaling restrains the distal movement of Shh/Smo-responsive bEps.
We imaged caudal fins of shha:GFP;runx2:mCherry larval fish with and without Shh/Smo inhibition to monitor movement dynamics of shha:GFP-expressing bEps relative to Runx2+ pObs (representative fish in Figure 6A-D’ and Figure 6 Movie 2; all fish shown in Figure 6 Supplement 2). We resolved individual cells at higher detail in distal ray regions by capturing full confocal z-stacks every 2 minutes over 30 minutes. Assisted by semi-automated cell tracking, we noted bEps moved faster when beyond the field of pObs in DMSO-treated control fins. We observed slow moving bEps in contact with pObs before detaching and rapidly moving distally. In contrast, BMS exposure caused rapid distal bEp movement irrespective of proximal-to-distal position or proximity to pObs.
We quantified positional dynamics of individual bEps and plotted their average normalized speed compared to starting position relative to the end of ray-forming Runx2+ pObs (DMSO: n=5 fish, 26-38 cells per fish, total of 159 cells; BMS: n=4 fish, 26-41 cells per fish, total of 135 cells). shha:GFP-expressing bEps located distal to pObs moved faster than pOb-associated bEps in control animals, producing a clear upward velocity shift at the pOb border (Figure 6E). In contrast, BMS treatment caused evenly distributed bEp velocities before and after the pOb-containing region. Taken together, we propose local Shh/Smo signaling enhances heterotypic cell adhesion that transiently restrains the continuous distal movement of bEps when they pass over pObs. For ray branching morphogenesis, Shh/Smo-enhanced cell adhesion between pObs and successive waves of bEps could enable the pOb pool to gradually follow laterally splitting shha-expressing bEp domains. Eventually, the divided pOb pools would then form separate daughter rays connected at a branch point.
DISCUSSION
Basal epidermal movements and Shh/Smo signaling direct skeletal branching morphogenesis during zebrafish fin ray development and regeneration
Zebrafish fin ray branching provides an accessible context to define mechanisms of appendage patterning and skeletal morphogenesis. Our current and earlier study (Armstrong et al., 2017) extends previous research to demonstrate the same Shh-dependent branching morphogenesis mechanism branches developing and regenerating rays. In both contexts, a gradually splitting domain of Shh-expressing basal epidermal cells (bEps) at the distal aspect of each outgrowing fin ray partitions the immediately adjacent pre-osteoblast (pOb) pool. Highly localized, continuous Shh/Smo activity allows a given ray’s distal pOb population to gradually follow the separating Shha+ basal epidermal domains. Eventually fully split, divided pOb pools continue to promote outgrowth of now two rays connected at a branch point. The shared mechanism of pOb positioning for ray branching underscores that fin regeneration re-activates developmental mechanisms. Strikingly, Smo-dependent Shh signaling appears largely dedicated to ray branching in all fins with pathway inhibition producing minimal or no effects on initial fin patterning, outgrowth, or skeletal differentiation during fin development or regeneration.
Our study of the readily observable developing caudal fin highlights how collective migration of Shh+ bEps positions pObs to generate branched rays. Our Kaede photoconversion and time-lapse imaging show bEps continuously move distally in growing fins, activating shha expression upon reaching the distal zone that includes pObs. Individual bEps pass through the shha-expressing domain, down-regulate shha when moving beyond the pObs, and then are shed from the end of the fin. In turn, proximal bEps enter the distal zone and activate shha to replenish the shha-expressing basal epidermal domain adjacent to pObs. Shha produces a constant Smo-dependent response in neighboring pObs and an autocrine, transient response in shha-producing bEps as represented by upregulated ptch2 in both cell types. This continuous, localized Hh/Smo signaling regulates bEp collective movement dynamics and promotes ray branching by enabling concomitant separation of pOb pools with lateral splitting shha basal epidermal domains.
Shh/Smo signaling is involved in branching morphogenesis of other organs, including the lung (Bellusci et al., 1997; Fernandes-Silva et al., 2017; Pepicelli et al., 1998) and submandibular salivary gland (Jaskoll et al., 2004). In the lung, Shh/Smo also mediates interactions between mesenchymal and epithelial populations although likely by promoting local proliferation and/or differentiation (Kim et al., 2015). Collective cell migration also is broadly implicated in branching morphogenesis, including for renal tubes, mammary glands and blood vessels (Ewald et al., 2008; Riccio et al., 2016; Spurlin & Nelson, 2017). Unlike those contexts, we propose a neighboring cell type – basal epidermis – that does not directly contribute to the final tissue provides the instructive collective movements. This unusual arrangement may reflect regenerating fin pObs having a mesenchymal state during patterning before returning to their differentiated epithelial state (Stewart et al., 2014).
Shh/Smo signaling may position pre-osteoblasts by promoting their adhesion to moving basal epidermal cells
Continuous Shh/Smo signaling through fin development and retained splitting of Shh-expressing basal epidermal domains when the pathway is inhibited indicate Shh/Smo has a permissive role in ray branching. We favor a model whereby Shh/Smo’s function is to promote transient cell adhesion between bEps and pObs. The transient nature may result from the moving bEps rapidly terminating their shha expression and ptch2-defined Shh/Smo activity. A slight lateral component to bEp movements away from the midline of each forming ray would then successively tug contacting pObs to follow. Over the course of several days, pObs eventually are pulled into two pools. The pOb pools become sufficiently and irreversibly separated to now generate branched daughter rays.
Our 3-D reconstructions showing Shh-expressing bEps and Runx2-expressing pObs likely share extensive and intimate physical contacts are consistent with this heterotypic cell adhesion model. The time-lapse imaging of developing caudal fins provides functional support by showing Shh/Smo signaling impedes bEp distal movements. Notably, Shh-expressing bEps accelerate when they pass beyond Shh-responding pObs. Chemical inhibition of Shh/Smo signaling significantly increases overall Ptch2-positive bEp migration rates and eliminates the characteristic velocity decrease when Shh-expressing bEps pass adjacent to pObs. While inhibiting Shh/Smo signaling accelerates individual bEp cell movements, a steady-state Shh-expressing basal epidermal domain persists and at least partially splits. However, the pObs cannot follow without Shh/Smo signaling to promote adhesion with bEps and therefore remain as a single pOb pool that forms an unbranched ray.
We favor Shh/Smo-promoted adhesion-based skeletal positioning over alternative hypotheses for additional reasons. First, Shh/Smo-dependent cell proliferation is not observed during development or regeneration (Armstrong et al., 2017), arguing against a model whereby Shh promotes localized proliferation at the margins of a given pOb pool progressively dividing it. Notably, previous conclusions that Shh/Smo signaling is a pro-proliferative factor in regenerating fins (Y. Lee et al., 2009; Quint et al., 2002) may reflect off-target cyclopamine effects (Armstrong et al., 2017). In contrast, fins develop and regenerate to normal size when using BMS-833923 to block Shh/Smo signaling with the intriguing exception of the two peripheral rays and their surrounding tissue. Second, all Shh/Smo-responsive cells remain outwardly specified upon Shh/Smo inhibition, including osteoblasts that still differentiate to produce ray skeletal units complete with joints. Any additional Shha or other Hedgehog ligand roles, if existent, would seem Smo-independent, as with Indian Hedgehog A (Ihha) and bone maturation during fin regeneration (Armstrong et al., 2017). Third, we use chemical genetics to map the Shh/Smo time-of-function and show ray branching requires persistent Shh/Smo signaling from initial hints of Shh-expressing basal epidermal domain splitting until daughter rays are fully separated. Therefore, Shh/Smo signaling promotes a continuous rather than switch-like mechanism acting throughout the morphogenesis process.
Short range Shh promoting cell adhesion may be a common Shh/Smo signaling mode
Our proposed short range Shh/Smo signaling mode promoting heterotypic cell adhesion differs from Hh’s more typical role as a gradient-forming morphogen. Providing precedence, Hh acts on neighboring cells in several well-established contexts. For example, Hh famously mediates interactions between directly adjacent cells during Drosophila embryo segment boundary formation (Ingham, 1993). Short-range Shh/Smo signaling also occurs in vertebrates, including mammalian hair follicle development (Millar, 2002; Sato et al., 1999; Woo et al., 2012), avian limb patterning (Sanders et al., 2013), and zebrafish retina development (Shkumatava et al., 2004). Perhaps most germane, shha+ epidermal cells organize directly underlying Hh-responsive dermal cells during zebrafish scale morphogenesis (Aman et al., 2018).
Shh/Smo has also been tied to cell adhesion in other settings. For example, Hh’s archetypal role in Drosophila wing disc compartment boundary establishment (Ayers et al., 2010) may be through increased “cell bonding” (Rudolf et al., 2015). Shh alters neural crest cell adhesion and migration during avian neural tube morphogenesis (Fournier-Thibault et al., 2009; Jarov et al., 2003; Testaz et al., 2001). Further, misregulated Shh/Smo signaling is linked to invasive cell migration associated with liver, breast, ovarian, and skin cancers (Chen et al., 2013; Chen et al., 2014; Hanna & Shevde, 2016; Zeng et al., 2017).
Identification and characterization of Shh/Smo-upregulated adhesion molecules would strengthen a heterotypic cell adhesion model for ray branching. One intriguing possible mechanism is that Shh/Smo-upregulated Patched directly binds to non-secreted Shh retained on bEp surfaces to increase high-affinity contacts between pObs and bEps. This mechanism could apply elsewhere given Patched is an evolutionary-conserved Shh/Smo-target gene (Alexandre et al., 1996; Goodrich et al., 1996; Lorberbaum et al., 2016; Marigo et al., 1996) while placing Patched in the curious position as both a Shh/Smo effector and negative feedback regulator. Alternatively, Shh/Smo activity could increase expression of more traditional cell adhesion factors or promote cell features (e.g. shape, polarity, or interconnectivity) that indirectly favor heterotypic adhesion. Regardless, our adhesion model assigns Shh/Smo a permissive role with shha+ basal epidermal domain splitting instructing branching. Therefore, how shha is activated when bEps enter distal fin outgrowth zones containing pObs and how shha+ bEp domains laterally split are key unresolved questions.
Fin ray branching as an ancestral mechanism of Shh-mediated appendage patterning and skeletal morphogenesis
How vertebrates pattern skeletal appendages (fins, limbs) is a textbook question of evolutionary and developmental biology. Fin rays comprise dermal bone opposed to endochondral bone found in tetrapod limbs with rays considered lost in tetrapod lineages. However, teleost fin dermal skeleton and tetrapod digits may share deep evolutionary homology (Nakamura et al., 2016). If so, our demonstration Shh/Smo signaling supports developmental ray branching morphogenesis is intriguing given Shh’s long-appreciated but mechanistically distinct role in vertebrate digit patterning. Shh is the secreted morphogen produced by the zone of polarizing activity (ZPA) at the posterior edge of developing limb buds that directs anterior-to-posterior patterning of skeletal elements including digits (Cohn et al., 2002; Riddle et al., 1993; Tickle, 2017). Polarized shha expression in zebrafish pectoral fin buds indicates paired fins may follow ZPA-like skeletal patterning (Akimenko & Ekker, 1995; Krauss et al., 1993; Neumann et al., 1999). In contrast, caudal fin primordia lack polarized shha (Hadzhiev et al., 2007; Laforest et al., 1998), our results). Consistently, we found disrupting Shh/Smo signaling even prior to formation of the caudal fin field does not alter the initial complement of 18 rays. Moreover, we demonstrate Shh/Smo signaling is required for ray branching in all fins, whether paired and unpaired. The unpaired medial fins (dorsal, caudal, anal) evolved prior to paired fin appendages (Dahn et al., 2006; Desvignes et al., 2018; Freitas et al., 2006; Larouche et al., 2017). Therefore, Shh-dependent ray branching may reflect an ancestral skeletal morphogenesis mechanism that predates emergence of ZPA-based appendage patterning. Interestingly, our proposed ray branching morphogenesis mechanism also hearkens a classic view that limb skeletal patterns progressively form by the unfolding of a series of three events: de novo cartilage condensations, branching, and segmentation (Oster et al., 1988; Shubin & Alberch, 1986).
Shh/Smo signaling promotes skeletal morphogenesis in many contexts. In zebrafish, Shh/Smo patterns craniofacial dermal bones, as illustrated by the opercle (Huycke et al., 2012), and both developing and regenerating scales (Aman et al., 2018). Shh/Smo signaling also impacts mesenchymal cell movements to pattern bird feathers, another albeit non-ossified skin appendage (Li et al., 2018). Shh further supports patterning of the axial skeleton (Chiang et al., 1996; Choi et al., 2012; Dworkin et al., 2016; Hu et al., 2015; Hu & Helms, 1999; Jeong et al., 2004; Swartz et al., 2012) as well as teeth (Ahn et al., 2010; Dassule et al., 2000; Seppala et al., 2017). Our discovery Shh/Smo signaling enables neighboring cells to position pObs during fin ray branching suggests similar mechanisms act in other skeletal patterning contexts. If so, manipulating Shh/Smo pathway to position therapeutically delivered or endogenous progenitor cells could enhance skeletal regenerative medicine.
MATERIALS AND METHODS
Zebrafish
Danio rerio zebrafish were maintained in 28-29°C circulating fish water within the University of Oregon Aquatic Animal Care Services (UO AqACS) fish facility. The following lines were used: wildtype AB, Tg(sp7:EGFP)b1212 (DeLaurier et al., 2010), TgBAC(ptch2:Kaede)a4596 (Huang et al., 2012), Tg(-2.4shha:gfp:ABC)sb15 [previously known as Tg(-2.2shh:gfp:ABC)] (Ertzer et al., 2007; Shkumatava et al., 2004), Tg(runx2:mCherry) (Shannon Fisher Lab, unpublished). The University of Oregon Institutional Animal Care and Use Committee (IACUC) approved zebrafish experiments.
Microscopy
Larval and juvenile fish were anesthetized with 74 µg/ml tricaine (MS-222, Syndel) in fish facility system water. Fish or dissected fins were transferred immediately to a 35 mm glass bottom FluoroDish plate (World Precision Instruments). Two or three drops of 1% low-melt agarose, stored at 38°C and cooled before application, were placed on the caudal fin. Fins were quickly flattened to the FluoroDish with a single-hair paintbrush before the agarose hardened. The following microscopes were used: Nikon Eclipse Ti-E widefield and Nikon Eclipse Ti2-E with Yokogawa CSU-W1 spinning disk attachments, and Zeiss LSM 880 laser scanning confocal microscope. Confocal image stacks were processed using Imaris software to generate single optical slice digital sections, surface renderings, and 3D reconstructions. Adobe Photoshop was used to adjust levels with identical image acquisition and processing settings for a given experiment. Live fish promptly were euthanized or returned to tanks after imaging.
Kaede photoconversion and imaging
ptch2:Kaede fish were anesthetized and placed on FluoroDish plates as described above. Fins were viewed with a Nikon Eclipse Ti-E widefield microscope or Nikon Ti2-E/ Yokogawa CSU-W1 spinning disk confocal microscope. Kaede-expressing regions of interest (ROIs) were photoconverted using a metal halide light source and DAPI excitation filter or with 405 nm laser illumination from 10 seconds to 2 minutes, depending on ROI size and fish age. Before and after images were acquired to ensure complete photoconversion of Kaede from green (518 nm) to red (580 nm) emission. Fish were returned to system water and then similarly re-imaged after defined periods.
BMS-833923 treatments
BMS-833923 (“BMS”, Cayman Chemicals) was dissolved in DMSO to a concentration of 6.3-12.5 mM. This stock was diluted to a final concentration of 0.63-1.25 µM in system fish water for both larval and juvenile zebrafish treatments. Equal volumes of DMSO were used for control group treatments.
To test Hh/Smo requirements for caudal fin ray branching, 25 dpf sp7:EGFP fish (n=6 per group) with unbranched rays were treated initially for 24 hours in BMS or DMSO-alone water and then returned to standard housing. Fish were exposed to BMS for 4 hours every other day until the experimental end point at 42 dpf. To assess Hh/Smo roles in all fin appendages, shha:GFP;runx2:mCherry fish were treated with 1.25 µM BMS or DMSO (n= 6 per group) from 21 to 42 dpf. Fins were dissected and imaged as described above.
For staggered-start juvenile fish treatments, 25 dpf sp7:EGFP fish were anesthetized, fluorescently screened, and sorted into groups of those having caudal fins with “unbranched” rays or fins in which branching had initiated but was incomplete (“during”). Fish from the two groups were then treated with BMS as described above. Untreated clutchmate sp7:EGFP fish were returned to standard housing and screened every other day until all fish had developed branched rays. Drug treatment of the “branched” group of fish was started at 35 dpf. All treatments ended at 42 dpf, when fins were mounted and imaged as described. For BMS-treated fish (unbranched, during branching, and branched), n=4 or 5 fish per group with n=2 or 3 for DMSO-treated control groups.
For early larval development studies, sp7:EGFP;runx2:mCherry and ptch2:Kaede fish were bathed in 1.25 µM BMS starting at 2 dpf alongside DMSO-treated controls (n=33-44 fish per group, per clutch). The same drug exposure regiment described for juvenile fish was used. From 2-4 dpf, larvae were treated in 40 mL embryo media in petri dishes. From 5-14 dpf, fish were drug-exposed in beakers containing 125 mL embryo media. Drug efficacy on larval fish was assessed by photoconverting distal fin ROIs of ptch2:Kaede fins (photoconversion methods described above) at 13 dpf and re-imaging those regions at 14 dpf (n=3-5 per group). All fish were screened for skeletal patterning phenotypes by widefield microscopy. Across clutches, 35/44 (79.5%) BMS-treated larvae developed normally (9/44 or 20.5% were runted) compared to 26/33 (78.8%) DMSO-treated larvae (7/33 or 21.2% were runted). The ∼20% incidence of developmentally delayed larvae was likely caused by extended periods in 250 mL beakers instead of larger nursery tanks. Regardless, nearly all larvae irrespective of size in both groups developed the normal complement of 18 caudal fin rays.
Ray morphometrics
Ray lengths were assessed for sp7:EGFP clutch mates treated from 25-42 dpf with BMS-833923 or DMSO (experiment described above, n=6 per group). Using Fiji-ImageJ software, the Principal Peripheral Ray (unbranching lateral ray) and Dorsal Ray 3 (longest branching ray) were measured from 42 dpf endpoint caudal fin images from the proximal base of the fin to the distal fin end. Raw and normalized data were graphed with GraphPad Prism V8 and significance assessed with a Student’s unpaired t-test.
Whole mount immunostaining
shha:GFP caudal fins were harvested at 22-23 dpf and immediately fixed in 4% PFA/PBS overnight at 4°C or for 4 hours at room temperature. Fins were washed extensively in PBS + 0.1% Tween-20 and blocked in 1x PBS, 1% Triton X-100, 5% Normal Goat Serum, and 10% DMSO buffer overnight at 4°C. Fins were incubated with primary antibodies in blocking buffer overnight at 4°C. Primary antibodies were anti-GFP (1:1000; AVES, GFP-1020), anti-Tp63 (1:100; Thermo Fisher, PA5-36069) and anti-Runx2 (1:100; Santa Cruz Biotechnology, sc-101145). Fins were washed in a high-salt 500 mM NaCl buffer for 30 min followed by extensive washes in PBS + 0.1% Tween-20. Secondary antibody incubations using Alexa Fluor conjugates (Thermo Fisher) were performed overnight protected from light at 4°C at a concentration of 1:1000 in blocking buffer. Fins were then washed extensively in PBS + 0.1% Tween-20, nuclei stained with Hoechst (Thermo Fisher), and mounted with SlowFade Diamond Antifade (Thermo Fisher).
Paraffin section immunostaining
Dissected 32 dpf shha:GFP caudal fins were fixed in 4% PFA/PBS overnight at 4°C. After extensive PBS washing, fins were decalcified for 4 days in 0.5M EDTA, pH 8.0 with daily solution changes. Fins then were dehydrated in an ethanol series and tissue cleared with xylenes prior to longitudinal embedding in paraffin wax. 7 µm sections were cut on a Leica RM255 microtome. Antigen retrieval was performed on rehydrated sections using 1 mM EDTA + 0.1% Tween-20 for 5 minutes in a pressure cooker. Following PBS washes, sections were blocked in 1x PBS, 10% nonfat dry milk, 2% normal goat serum, and 4% fetal bovine serum for a minimum of 1 hour. Sections were incubated overnight at 4°C with primary antibodies in blocking solution. Primary antibodies were: anti-GFP (1:3000; AVES, GFP-1020), anti-Tp63 (1:100; Thermo Fisher, PA5-36039), anti-Laminin (1:40; Sigma, L9393), and anti-Zns5 (1:5, ZIRC). Sections were washed in PBS containing 500 mM NaCl + 0.1% Tween-20. Alexa Fluor conjugated secondary antibodies (Thermo Fisher) were diluted 1:1000 in blocking buffer and incubated for 1 hour at room temperature protected from light. Sections were washed, nuclei stained with Hoechst, and mounted with SlowFade Gold Antifade (Thermo Fisher). Images were acquired on a Zeiss LSM 880 laser scanning confocal microscope and images processed with Fiji-ImageJ, Imaris, and Adobe Photoshop.
In vivo EdU incorporation assays
29 dpf shha:GFP juvenile fish were treated with DMSO or 1.25 μM BMS for 4 hours in groups of n=5. Anesthetized fish were injected intraperitoneally with 5 μl of 1 mg/mL EdU (Thermo Fisher) in sterile PBS, monitored for recovery for 10 minutes in fresh facility water, and then returned to treatment tanks. 12 hours post-injection, caudal fins were amputated and fixed for 4 hours at room temperature in 4% PFA/PBS. Fins were washed thoroughly with PBS and blocked overnight at 4°C in PBS/1% Triton X-100/5% Normal Donkey Serum/10% DMSO. EdU signal was detected with Click-iT Plus Alexa Fluor 647 Picolyl Azide (ThermoFisher) at 2.5 μl/ mL according to the manufacturer’s protocol. Following EdU detection, whole-mount GFP immunostaining and Hoechst nuclear staining was performed as described below. Whole mount confocal images were acquired using a Zeiss 880 LSM and 3D reconstructions prepared using Imaris. EdU+ and total intra-ray nuclei, i.e. from cells located in between the epidermal Shh domains of each hemi-ray, were identified and scored for Rays 2 and 3 using the Imaris “Spots” function and the following parameters: ROI around length of Shha:GFP+ domain, Quality Threshold 0.642, cell diameter 3 microns. Quantification of EdU+ cells is expressed as the number of EdU+ intra-ray cells over total number of Hoechst-stained nuclei.
Cell migration imaging and analysis
Fish were anesthetized sequentially in freshly prepared 74 µg/ml tricaine solution for 3 minutes and monitored for slowed opercular movements. Anesthetized fish were transferred to a 35 mm FluroDish plate and mounted in 3% low melt agarose as described earlier. Set agarose was carefully removed from the most distal region of the caudal fin to allow for free movement of the epidermis while the trunk remained adhered to the FluoroDish. 74 µg/ml of tricaine solution was added to maintain anesthesia and cover the fin. After imaging, fish were returned to system water to confirm recovery and then promptly euthanized. Fish that did not recover were excluded from downstream cell migration analyses. We occasionally observed extremely rapid epidermal movements in which entire shha:GFP+ domains would be shed from rays in <15 min. We suspect this phenomenon results from elevated stress, anesthesia intolerance, and/or damage from plate surface contact or agarose application. We excluded these animals from analyses.
For bulk cell migration assays, 22-24 dpf ptch2:Kaede fish were treated with 0.63 µM BMS or DMSO (n=8 per group). 24 hours later, fish were mounted and imaged with a Nikon Eclipse Ti-E widefield microscope every 1 minute for 30 minutes. Imaris was used to automatically track cells for 3-6 single ptch2:Kaede+ basal epidermal cells on dorsal rays 2-5 for each fish. All tracks were quality checked to confirm individual cell tracking. Individual average cell speeds (n=38 cells per group) and then averages for each animal were determined. Statistical significance tests comparing all cells tracked (n=38 cells per group) and mean cell speed per fish (n=8 fish per group) used Student’s unpaired t-tests.
For position-dependent cell migration assays, 21-24 dpf shha:GFP;runx2:mCherry fish were treated with DMSO or 1.25 µM BMS for 24 hours prior to imaging. Fish were imaged every 2 minutes for 30 minutes with full z-stacks using a Nikon Ti2-E with a Yokogawa CSU-W1 SoRa spinning disk confocal unit. A single hemi-ray of Ray 2 or Ray 3 was analyzed for each time-lapse video. If both rays were captured, the ray with more pObs in frame was analyzed to avoid oversampling individuals. GFP+ cells were automatically tracked using Imaris software “Spots” algorithms with the following parameters: estimated cell diameter 5 microns, maximum distance between frames 6 microns, maximum gap between frames 3 time points. Each cell track was quality checked using 3D reconstructions and edited if Imaris assigned multiple cells to one track or fragmented the track of a given cell. 26-41 cells were tracked across 9 fish (n=5 for DMSO and n=4 for BMS groups, respectively) for a total of n=159 for DMSO-treated and n=135 for BMS-treated fish. Data was normalized for each fish by dividing the track speed of a single cell by the average of all cells tracked for that fish. Positional data was determined by setting the X-position of the most distal Runx2+ pOb as “0” and assigning a relative initial X-position for each shha:GFP+ cell. Cells with a negative starting position were therefore pOb-associated while those with a positive starting position had already migrated beyond the pOb pool when video acquisition began. Imaris was used to determine each cell’s total X-displacement and track speed. Graphs were generated using GraphPad Prism V8. Fourth-order best-fit polynomial curves were added to position/speed graphs to help visualize data trends.
FOOTNOTES
Competing interests
None.
Author contributions
J.A.B, A.E.R. and K.S. designed experiments with input from S.S.; J.A.B and A.E.R. performed experiments; J.A.B, A.E.R. and K.S. prepared and wrote the manuscript.
Funding
The National Institutes of Health (NIH) provided research funding (1R01GM127761 (K. S. and S. S.). J.A.B. had an Experiencing Science Practices through Research to Inspire Teaching (ESPRIT) award supported by the National Science Foundation’s Robert Noyce Teacher Scholarship Program, a Mary G. Alden Scholarship, and an Institute of Molecular Biology Summer Scholarship. A.E.R. received support from the University of Oregon Genetics Training Program (5T32GM007413).
Data and material availability
Requests for materials should be addressed to K. S.
ACKNOWLEDGEMENTS
We thank the University of Oregon AqACS Facility for zebrafish care; the University of Oregon zebrafish community for support; A. Delaurier and C. Kimmel for Tg(sp7:EGFP)b1212 fish; S. Megason for providing the TgBAC(ptch2:Kaede)a4596 fish developed in A. Schier’s lab and the Tg(-2.4shha:gfp:ABC)sb15 fish developed in U. Strahle’s lab; G. Crump for providing the runx2:mCherry fish generated by S. Fisher’s group; and C. Kimmel, A. Saera-Vila, and the Stankunas lab for input.