Introduction

Although often studied as separate entities, the brain and spinal cord function in synergy to govern numerous bodily functions. The brain contains multiple functionally distinct neuronal populations known as spinal-projecting neurons (SPNs) that project through various regions of the spinal cord, terminating in specific segments along the rostrocaudal axis (Liang et al., 2016; Nyberg-Hansen, 1965; Wolters et al., 1982). These descending pathways orchestrate a wide array of motor, autonomic, and sensory functions (Kuypers, 2011, 1964; Winter et al., 2025). While some SPNs can carry motor commands ranging from basic locomotion to sophisticated fine movements involving muscles in the head or throughout the body (Kim et al., 2017; Leiras et al., 2022; Wang et al., 2017), other SPNs modulate the autonomic system regulating vital processes such as thermoregulation, breathing, cardiovascular function, and metabolism (Swanson and Sawchenko, 1980; Verstegen et al., 2017). Additionally, numerous SPNs modulate the sensory detection of tactile, thermal, proprioceptive, and painful stimuli (François et al., 2017; Liu et al., 2018).

The location and funicular trajectories of SPN nuclei have been well-documented via tract-tracing across many vertebrate taxa, including mammals, birds, reptiles, amphibians, lungfishes, elasmobranchs, and cyclostomi (Adli et al., 1999; Cabot et al., 1982; Cruce and Newman, 1981; Crutcher et al., 1978; Hermann et al., 2003; Lacroix-Ouellette and Dubuc, 2023; Ronan, 1989; Ronan and Northcutt, 1985; Sánchez-Camacho et al., 2001; Smeets and Timerick, 1981a; Ten Donkelaar, 1976; ten Donkelaar et al., 1980a, 1981a; Ten Donkelaar, 1982a). Comparative anatomical studies consistently demonstrate that the anatomical distribution of descending systems in early-derived vertebrates rivals that of mammals (Cruce and Newman, 1984; Ryczko et al., 2016). However, a significant gap remains in our understanding of the anatomy of these descending centers and their contribution to integrated motor and physiological responses.

The larval zebrafish (Danio rerio) has proven to be an ideal model for bridging this gap, offering a clear window into evolutionary conservation (Carbo-Tano et al., 2023) due to its unique experimental accessibility to its brain organization (Burgess and Burton, 2023; Pansera et al., 2025). Yet, in this genetic model organism the identity and characteristics of descending SPNs are not fully known. While horseradish peroxidase (HRP) tracing in various teleost species has shown a remarkable similarity to mammalian distributions (Bosch and Roberts, 1994, 2001; Hlavacek et al., 1984; New et al., 1998; Oka et al., 1986; Prasada Rao et al., 1987a), early anatomical studies in larval zebrafish predominantly labeled a very small (< 200) subset of neurons characterized by large cell bodies and large-caliber axons running in the medial longitudinal fasciculus (Kimmel, 1982; Kimmel et al., 1982). The distinctive morphology and stereotyped positions of these cells allowed for a systematic classification across individuals with names for distinct neurons identified according to its dorsoventral and anteroposterior coordinates (Kimmel, 1982; Kimmel et al., 1982; Lee et al., 1993; Lee and Eaton, 1991a). These early identifications of SPNs based on retrograde labeling suggested that the zebrafish descending system was possibly simpler in comparison to other vertebrates (Cruce and Newman, 1984). Retrograde labeling typically performed by injections in segment 14-15 highlighted primarily large reticulospinal neurons in the caudal midbrain and pontine region, with only sparse labeling in the medulla leading to a focus on these neurons for subsequent functional investigations (Collins et al., 2025; Gahtan and O’Malley, 2003; Huang et al., 2013; Lau et al., 2025; Orger et al., 2008).

Subsequent studies utilized photoconvertible proteins such as Kaede that change their emission spectrum from green to red upon UV light exposure (Ando et al., 2002). By illuminating axons in the spinal cord with UV light, the photoconverted Kaede diffuses retrogradely to cell bodies, allowing for ratiometric differentiation between spinal-projecting (red) versus non-spinal-projecting (green) neurons. By expressing Kaede in genetically-defined neuronal populations, such optical backfills of neurons projecting to the spinal cord have revealed within the population of neurons expressing vsx2 (former chx10) transcription factor, that the medulla of larval zebrafish contains many more reticulospinal neurons than previously thought (Kimura et al., 2013, 2006; Pujala and Koyama, 2019). Yet, these studies were limited to discrete, genetically identifiable neuronal subsets, leaving homological relationships across vertebrates unaddressed.

To label virtually all supraspinal neurons, we relied here on the photo-activable GFP (PAGFP) expressed pan neuronally. To map all spinal-projecting neurons in larval zebrafish, PAGFP can be activated using two photon illumination in a confined section of the spinal cord (Bianco et al., 2012; Ruta et al., 2010). This approach reveals far more neurons than conventional retrograde tracing strategies have previously identified, offering an unparalleled opportunity to revisit distinct spinal-projecting nuclei distributed across the hindbrain, midbrain, and diencephalic structures. By combining information on cell location, morphology, and projection patterns, we propose tentative anatomical homologies between these nuclei in fish and mammals. These proposed homologies will undoubtedly be refined by future studies integrating molecular, genetic, and functional data.

Methods

Animal care and transgenic lines

Animal handling and procedures were validated by the Paris Brain Institute (ICM) and the French National Ethics Committee (Comité National de Réflexion Éthique sur l’Expérimentation Animale; APAFIS no. 2018071217081175 and #51743-2024102515029150) in agreement with European Union legislation. To avoid pigmentation, all experiments were performed on Danio rerio larvae of AB background with the mitfa^−/− mutation. Adult zebrafish were reared at a maximum density of eight animals per liter in a 14/10 h light–dark cycle environment at 28.5 °C. Larval zebrafish were typically raised in Petri dishes filled with system water under the same conditions in terms of temperature and lighting as for adults. Transgenic larvae expressing PAGFP under the alpha-tubulin promoter used for the photoconversion experiments were raised in complete darkness to minimize global photoactivation at the early stages. Experiments were performed at 20 °C on animals aged between 5 and 6 days post fertilization (dpf). We used the double transgenic line referred to as Tg(Cau.Tuba1:PAGFP; elavl3:FusionRed) for simplicity throughout, by crossing Tg(Cau.Tuba1:c3paGFP)^a7437 with Tg(elavl3:stGtACR1-FusionRed) and screening for double-positive embryos. For anatomical interrogations, we used the double transgenic line Tg(oxt:GAL4-VP16:UAS:mCherry) referred to as Tg(oxt:GAL4;UAS:mCherry) by crossing Tg(oxt:GAL4-VP16)^zf3181Tg(Wee et al., 2019) with Tg(UAS:Eco.NfsB-mCherry)^rw0144Tg (Agetsuma et al., 2010) and Tg(GnRH2:EGFP) (Xia et al., 2014).

Design of the Tg(elavl3:stGtACR1-FusionRed) stable line

To generate the Tg(elavl3:stGtACR1-FusionRed) transgenic line, the stGtACR1-FusionRed coding sequence was subcloned from pAAV-CKIIa-stGtACR1-FusionRed (Addgene, plasmid #105679) into a Tol2 destination vector under the control of the pan-neuronal elavl3 promoter. Transgenic zebrafish were generated using the Tol2 transposon-based method (White et al., 2008a); the construct was co-injected with Tol2 transposase mRNA into one-cell stage zebrafish embryos in a casper background (Urasaki et al., 2008; White et al., 2008b). Injected fish (F0) were raised to adulthood and outcrossed to casper fish to identify germline-transmitting founders. Positive F1 offspring were identified by screening for FusionRed fluorescence, and stable transgenic lines were established from confirmed carriers.

Optical backfills

Transgenic Tg(Cau.Tuba1:PAGFP; elavl3:FusionRed) zebrafish larvae expressing cytosolic PAGFP and membrane tagged FusionRed proteins pan-neurally under the alpha-tubulin and elavl3 promoter respectively were raised in darkness. At 5 dpf, larval zebrafish were anesthetized in 0.02% MS-222 and mounted in glass-bottom dishes (MatTek, Ashland, MA, USA) dorsal side up in 1.5% low melting point agarose. The dish was filled with external bath physiological solution (134 mM NaCl, 2.9 mM KCl, 2.1 mM CaCl₂-H₂O, 1.2 mM MgCl₂, 10 mM glucose and 10 mM HEPES, pH adjusted to 7.4 and osmolarity to 290 mOsm). The PAGFP protein in the rostral spinal cord (segments 4–9) was photoactivated in a two-photon microscope (2p-vivo, Intelligent Imaging Innovations, Inc., Denver, CO, USA, https://www.intelligent-imaging.com/) including a Mai Tai Deepsee laser (Spectra-Physics) tuned to 800 nm and a Zeiss Axio-Examiner Z1 with a Zeiss W Plan-APO 20X water-immersion objective (NA = 1). Line scans were controlled using Slidebook software (3i Intelligent Imaging Innovations, Inc., Denver, CO, USA). Before the photoactivation session, using minimum power the region of the spinal cord was recognized, and a region of interest was drawn in such a way that only half of the spinal cord was imaged. The photoactivation was done by scanning at 2 Hz with 4 µs dwell time using a laser power of 10 mW. The full depth of the spinal cord was scanned at a 1 µm interval and each plane was imaged for approximately 2 seconds. After the photoactivation session, larvae were unmounted and kept in the dark for 12 - 24h.

Retrograde labeling experiments using biocytin

For the retrograde labeling experiments, 5-dpf Tg(GnRH2:EGFP) larvae were anesthetized with 0.02% MS-222 (catalog no. A5040, Sigma-Aldrich) and laterally embedded in 1.5% low melting point agarose (Sigma, A4018) prepared with external bath solution (as described above). A small portion of the agarose was removed on top of the tail around myomere 9, and all the liquid was then extracted from the dish. A few crystals of biocytin Alexa Fluor 594 (catalog no. A12922, Thermo Fisher Scientific) were diluted in 5 µl distilled water and then allowed to recrystallize on the tip of insect pins (stainless steel Minutiens, 0.2-mm diameter, Austerlitz). Under a stereo dissecting microscope, a unilateral single lesion in the spinal cord between myomere 8 and myomere 10 was done using the dye-soaked insect pins. After the procedure, the dish was filled with external solution, the larvae were unmounted and allowed to recover for 2 h in an oxygenated external solution. Larvae were allowed to recover overnight in an external bath solution. The next day, the larvae were anesthetized again with 0.02% MS-222 and mounted dorsally in 1.5% low-melting-point agarose. The retrogradely labeled neurons were imaged in a confocal microscope as described above.

Whole-brain imaging

At 6 dpf, larval zebrafish were anesthetized in 0.02% MS-222 (catalog no. A5040, Sigma-Aldrich) and mounted in glass-bottom dishes dorsal side up in 1.5% low melting point agarose. The brain and rostral spinal cord were imaged in an upright confocal microscope (Olympus FV1200 confocal system, Shinjuku City, Tokyo, Japan) with a 20X water dipping objective NA = 0.5 (UMPLFLN 20XW) for capturing the general brain features and a 40X water dipping objective NA = 0.80 (LUMPLFLN 40XW) for further detailed imaging. Using the 20X objective, the whole brain of larval zebrafish was captured in two consecutive tiles with dimension 500 x 510 µm (10% overlap). Voxel size was set to ∼ 0.9 x 0.9 x 2 µm. Detailed imaging with the 40X objective was done with a voxel size of ∼ 0.2 x ∼0.2 x 2 µm. Due to the somewhat diffuse PAGFP signal when photoactivated, natural tissue autofluorescence when imaging with 488nm light, and because all imaging was performed from the dorsal aspect, the image quality is degraded in the ventral regions of the brain.

Reference brain creation and 3D registration

To generate the reference brain space, the consecutive tiles of each brain were stitched using Fiji (i.e., ImageJ) software (https://fiji.sc/) applying the stitching plugin (Preibisch et al., 2009). Our template brain was a 6-dpf Tg(elavl3:stGtACR1-FusionRed);mitfa^-/- larvae. Registration of image volumes was performed using ANTs (http://stnava.github.io/ANTs/) through the SlicerANTs extension (Revision: 4e622b5) of 3DSlicer software (Fedorov et al., 2012), www.slicer.org, version 5.8.1). The soma targeting of the opsin GtACR1 was not effective in the line Tg(elavl3:stGtACR1-FusionRed), which enabled us to have a membrane signal providing a robust scaffold of the brain architecture. The FusionRed signal of 6 larvae was registered using affine and SyN registration steps. We created the mean-stack reference brain by calculating the mean across all registered fish and normalizing to the maximum-intensity. Registration steps followed and ANTs parameters used are described in (Bhandiwad et al., 2024). All registrations were manually assessed for global and local alignment accuracy. In order to align our results with the well-established effort of the mapZbrain (Kunst et al., 2019) and Zbrain (Randlett et al., 2015) atlases, we registered the Tg(elavl3:lynTagRFP) signal from the mapZbrain Atlas and Anti-Zrf2 signal from the Zbrain atlas was registered to our Tg(elavl3:FusionRed) space.

Cell detection

Single-cell detection and segmentation were performed manually due to the complex nature of the backfill signal. Automatic segmentation methods failed to accurately identify individual neurons because the backfilled signal included not only somata but also dendrites and axons, resulting in a poor signal-to-noise ratio. Individual neurons were identified and manually segmented based on soma morphology and labeling intensity using napari software (Sofroniew et al., 2026) (https://github.com/napari/napari). Each neuron was annotated using the point layer tool in napari, with points placed at the center of each soma. Neurons were traced across consecutive optical sections to ensure accurate identification and avoid double counting the same neuron. The spatial xyz coordinates of each identified neuron were recorded for subsequent anatomical registration and analysis.

Data analysis and visualization

Analysis was done either in R (v.4.5.3) using the RStudio IDE (v.2025.9.0.387) (Posit team, 2025) or Python v.3.7. Figures were assembled in Inkscape (v.1.4.3, https://inkscape.org/).

Results

Brainwide distribution of spinal descending neurons

To investigate the distribution of neurons sending descending axons to the spinal cord in larval zebrafish, we employed optical backfilling in 5-day post fertilisation (dpf) double transgenic larvae Tg(Cau.Tuba1:PAGFP; elavl3:FusionRed). These transgenic larvae pan-neuronally express cytosolic photoactivatable GFP (PAGFP) and membrane-targeted FusionRed (Figure 1a). We performed unilateral photoactivation at the rostral spinal cord using a two-photon scanning system at 800 nm on a region extending caudally from the obex and spanning approximately 10 spinal segments (Figure 1a, Methods). At 6-dpf, larvae were mounted in agarose, and their brains were imaged using confocal microscopy (Figure 1a). To enable direct comparison of results across individual animals and to integrate our findings with existing publicly available zebrafish brain atlases, we generated an averaged brain template from six whole-brain images based on the FusionRed signal (Figure 1b). We then bridged our brain template with the well-established zebrafish atlases registering our brain template with the Zbrain Anti-Zrf2 signal (Randlett et al., 2015) and the mapZbrain Tg(elavl3:lyn-TagRFP) signal (Kunst et al., 2019). Following the registration to the template brain, PAGFP-positive cell bodies were manually segmented from the 3D image stacks. The resulting whole-brain distribution map revealed a substantial population of spinal-projecting neurons in larval zebrafish (Figure 1c, Figure 2, Supp. video 1), displaying a clear ipsilateral prominence with an average of 654.5 ± 53 (mean ± SD) spinal-projecting neurons per larva (estimated 1,309 ± 106 neurons for the complete spinal-projecting population (mean ± SD)). To regionalize the locations of the segmented neurons, we utilized the anatomical masks of the Zbrain atlas (Randlett et al., 2015), which provides better anatomical annotation in the hindbrain region where the rhombomere boundaries are chosen (see below). Consistent with the distribution of spinal-projecting neurons in other vertebrates, we found that the hindbrain contains the largest population of spinal descending neurons (Figure 1d), with the medulla oblongata harboring most of these cells (Figure 1c). Some anatomical regions displayed a clear unilateral distribution pattern, while other neuronal populations were primarily bilateral in nature (Figure 2c).

Figure 1.

Figure 2.

The identification and characterization of spinal-projecting neurons (SPNs) in zebrafish relied on multiple complementary criteria. The primary consideration was the spatial organization of labeled neurons, including both their absolute coordinates in the brain atlas and their relative positioning with respect to known nuclei and fiber tracts (medial longitudinal fasciculus, MLF; lateral longitudinal fasciculus, LLF; posterior cerebellar tract, PC; and vestibulo-spinal tract, TVS). Cellular morphology provided additional distinguishing features, particularly the characteristics of labeled cell bodies, including soma size, and density. When detectable, the trajectory patterns of axons emerging from labeled cell bodies served as another crucial identification criterion. We documented the distribution patterns, noting both symmetrical (reflecting ipsilateral and contralateral projections) and asymmetrical (reflecting either ipsilateral or contralateral projections) labeling. The relative intensity of labeling within cell groups provided supplementary discriminating information, as consistent differences in signal strength emerged between distinct neuronal populations. Unlike conventional retrograde labeling methods employing HRP or conjugated dextrans, optical backfilling reliably labels somatic compartments but does not yield sufficient signal to resolve dendritic or axonal morphology. We attribute this limitation to two factors: the simultaneous labeling of large numbers of neurons, which likely increases background fluorescence, and the restricted diffusion of photoactivated PAGFP into dendritic compartments. Therefore, all our subsequent analyses focus primarily on somatic information. Our analysis extended beyond the anatomical characteristics of nuclei to encompass broader organizational principles, as we anchored our observations with expression patterns from reference transgenic lines and hybridization chain reaction (HCR) staining data available in the Zbrain and mapZbrain atlases (https://zebrafishexplorer.zib.de/; https://mapzebrain.org/). In the following sections, we provide a detailed description of identifiable cellular groups of SPNs organized in rostrocaudal order.

Telencephalic projections to spinal cord

Using classical tracing techniques, mammals appeared as the main taxon of vertebrate species with obvious telencephalic projections to the spinal cord (Cruce et al., 1999) along with salamanders and sharks where few telencephalic spinal projection neurons (Kokoros and Northcutt, 1977). In some of our specimens (n = 4 out of 8), few telencephalic neurons in the olfactory bulb sent axons to the spinal cord (Figure 3a). By exploring the recently published ZExplorer atlas (Du et al., 2025), which contains thousands of single neuron tracings in the larval zebrafish brain, we searched for telencephalic neurons projecting to the spinal cord. We confirm that a few spinal-projecting neurons in the olfactory bulbs send both ipsilateral and contralateral descending axons, while in the subpallium spinal projection neurons send only ipsilateral descending axons (Figure 3a). The neurons we found in the olfactory bulbs matched those present in the ZExplorer atlas (Figure 3a). In contrast, we could not detect subpallial neurons in our dataset, likely because these neurons are very sparse or because the photoactivation level was insufficient to distinguish their somas from the background signal.

Figure 3.

Diencephalic projections

The preoptic area (POA) contains only ipsilateral projecting neurons (Figure 3b, Figure 2c). These neurons correspond to the rostral oxytocinergic cluster as shown by the expression pattern of the Tg(oxt:GAL4;UAS:mCherry) transgenic line (Figure 3b). In agreement with the results of our optical backfills, the oxt+ neurons send exclusively ipsilateral descending axons to the spinal cord (Figure 3b). Caudally, the posterior tuberculum (PT) has also mostly ipsilateral projection neurons (Figure 2c) segregated into two clusters, one anterior and the other posterior to the caudal oxytocin cluster (Figure 3b). The location of these two spinal-projecting clusters corresponds to the diencephalic dopaminergic (DA) groups referred to as DC2 and DC4 in zebrafish (Tay et al., 2011), and is in line with the results found in lampreys and other vertebrates (Ryczko et al., 2016). In the hypothalamus, spinal-projecting neurons were localized in the rostral (rHyp) and intermediate subdivisions (iHyp). While spinal-projecting neurons in the rHyp are exclusively ipsilateral, those in the iHyp show a slight contralateral bias (Figure 3c, Figure 2c). The thalamus (Th) hosts mainly ipsilateral spinal-projecting neurons located in the caudal pole (Figure 3d, Figure 2c). We found many spinal-projecting neurons in the accessory optic system (AOS). The AOS sends bilateral projections to the spinal cord with a strong contralateral prominence (Figure 3d, Figure 2c). The preganglionic complex (PGC) contains the soma of many spinal-projecting neurons with ipsi- and contralateral projections to the spinal cord, with the highest density near the midline adjacent to the caudal pretectum (Figure 3e, Figure 2c). At this level, the pretectum (PrT) hosts the soma of many neurons projecting to the spinal cord that are densely packed along the midline with two easily recognizable nuclei, a large one with strong ipsilateral labeling and a smaller one located more medially with contralateral projections (Figure 3e, Figure 2c).

The PGC and the PrT are two closely located regions with boundaries that are difficult to define. The region comprising the caudal PGC, PrT, and rostral tegmentum (Teg) is also known as the interstitial tegmentum or central tegmental field (Ten Donkelaar, 2020; Warner, 1947), an ill-defined area with nuclei that project to large regions of the central nervous system (Walter and Shaikh, 2014). There is still confusion about whether these areas belong to the forebrain or the midbrain (Watson et al., 2019a). In most vertebrate species investigated, conspicuous spinal-projecting neurons have been described in this area (Auclair et al., 1993; Castiglioni et al., 1978; Kuypers and Maisky, 1975; Nudo and Masterton, 1988a; ten Donkelaar et al., 1980b; Ten Donkelaar, 1982b). In zebrafish, the spinal-projecting neurons in this region have been classically named the nucleus of the medial longitudinal fasciculus (nMLF) (Kimmel, 1982; Kimmel et al., 1982) and have been regarded as located in the midbrain (Berg et al., 2023). Nevertheless, recent reports have relocated the nMLF, using molecular approaches (Wullimann et al., 2023), to prosomere 1 of the diencephalon, aligning its location with the corresponding locus in mammals (Watson et al., 2019a). Since the anatomical masks used in this study still rely on the uncorrected region boundaries, we opted to assign the spinal-projecting neurons corresponding to the rostral medial tegmentum to the caudal PrT.

The pretectal interstitial nuclei are composed of the centrally projecting Edinger-Westphal nucleus (EW), the nucleus of Darkschewitsch (ND), and the interstitial nucleus of Cajal (INC) laterally (Ten Donkelaar, 2020) (Figure 4a). All three nuclei send projections to the spinal cord, unilaterally for the ND and INC, and with a bilateral tendency for the EW (Liang et al., 2011a; Nudo and Masterton, 1988b). In mice, the EW is a facultatively peptidergic nucleus that expresses numerous peptides, including Cocaine- and amphetamine-regulated transcript (CART) and Urocortin (Ucn) (Bluett et al., 2025; Priest et al., 2023). Anatomically, the centrally projecting EW has extensive projections throughout the brain and spinal cord (Topilko et al., 2022). In an attempt to find zebrafish transgenic lines useful for targeting the nMLF, we identified the Tg(GnRH2:EGFP) (Xia et al., 2014) transgenic line as a possible candidate (Figure 4b). At 6 dpf, the Tg(GnRH2:EGFP) transgenic line labels only a single cluster of neurons in the interstitial tegmental region with widespread projections throughout the brain and spinal cord (Xia et al., 2014; Yamamoto et al., 1995a). However, labeling neurons in the nMLF with spinal dye injections revealed that the GnRH2⁺ cluster lies medial to the nMLF (Figure 4c1) (Yamamoto et al., 1998). We previously proposed that the nMLF of zebrafish larvae might correspond to the INC (Carbo-Tano et al., 2023). Given that the Tg(GnRH2:GFP)⁺ cluster is located medially to the INC and has an extensive projection pattern, we speculated that the Tg(GnRH2:GFP)+ cluster corresponds to the EW nucleus. We tested this hypothesis by examining the expression pattern of two peptides associated with the EW nucleus in mammals, using the whole-brain distribution of mRNA for cocaine- and amphetamine-regulated transcript 2 (cart2) and urotensin 1 (uts1) from the mapZbrain atlas (homologs to CART and UCN in mammals). We found that the signals for cart2 and uts1 match the GnRH2⁺ cluster, supporting our hypothesis that the GnRH2⁺ cluster corresponds to the peptidergic EW nucleus (Figure 4c2). Using the EW as an anchor point, we now propose that the interstitial pretectal nuclei are composed of the EW medially, the ND adjacent laterally, and the INC lateral to the MLF (Figure 4d). Since the ND is included as part of the ventral paraqueductal gray (PAG) (Ingram and Ranson, 1935; Zhao et al., 2024), the spinal-projecting neurons of the ND in the most dorsal planes could be part of the ventral PAG, which is known to have ipsilateral projections to the spinal cord (Liang et al., 2011a).

Figure 4.

Mesencephalic projections

In the mesencephalon, retrogradely labeled neurons were observed in the optic tectum (OT) with an ipsilateral prominence (Figure 5a, Figure 2c). The contralaterally projecting neurons are few and sparse and not spatially organized into a cluster. The highest density of neurons is located ipsilateral to the backfill in the ventrocaudal periventricular layer near the tegmentum (Teg). Retrogradely labeled neurons were also found in the torus semicircularis (TS) on both sides of the brain, but with a slight ipsilateral bias (Figure 5b).

Figure 5.

The Teg contains spinal-projecting neurons on both sides of the brain. At the midline, we found ipsilateral labeling of a group of neurons that could correspond to the medial accessory oculomotor nucleus (MA3) of mammals (Liang et al., 2011b). Lateral to the INC, we consistently found a group of well-labeled neurons sending ipsilateral axons to the spinal cord (Figure 5c-d). Based on their location and projection pattern, these neurons could belong to the midbrain reticular nucleus (mRt) or nucleus of the lateral lemniscus (nLL) as described in teleost (New et al., 1998; Yáñez et al., 2024) (Figure 5c-d). Sending exclusively contralateral projections, we found the red nucleus (RN) in the mediolateral Teg (Figure 5e). Consistent labeling of the RN was difficult to achieve, with RN neurons showing variable labeling intensity, possibly because the RN in 6 dpf larval zebrafish does not project far caudally beyond the obex region. We found that the RN consists of 5 neurons arranged in a C shape in the sagittal plane (Figure 5e). Unfortunately, due to the consistently low level of photoactivation, we could not trace the axons of the RN neurons to confirm their projection characteristics. To further confirm our identification of this group of neurons as the RN, we investigated whether the cerebellum (Cb) sends contralateral efferents to the area we identified as the RN (Finger, 1978). To verify Cb to RN connections, we used the mapZbrain single cell atlas to search for projections to the region we identified as the RN. Confirming our hypothesis, we found neurons with somas in the Cb sending axons that terminate or pass through the RN (Figure 5f).

Rhombencephalic projections

From this point caudally, the number of labeled neurons increases considerably (Figure 1d). For the brainstem description, we followed the nomenclature and subdivisions proposed by Watson et al. 2019 (Watson et al., 2019a). Although originally developed for mammals, we found this nomenclature and the brainstem boundaries based on developmental gene expression to be well-suited for comparative studies across vertebrates (Bhandiwad et al., 2022; Carbo-Tano et al., 2023). The brainstem starts with the prepontine (PP) region, composed of the isthmus and rhombomere 1 (r1), with the trochlear nucleus (4N) lying at the rostral border of the isthmus and the locus coeruleus (LC) at the caudal pole of r1 (Watson et al., 2019a). In the PP region, we found spinal-projecting neurons on both sides with nearly equal laterality and a slight ipsilateral bias (Figure 6a, Figure 2c). At the ventral planes, we found large neurons with ipsilateral projecting axons that are part of the pontine oralis group of mammals (Liang et al., 2011a). At the medial level, the LC is strongly labeled bilaterally, lying at the caudal border of r1 (Figure 6b). Rostral to the LC, we found a group of spinal-projecting neurons with bilateral projections to the spinal cord and an ipsilateral prominence that resembles the parabrachial group of mammals (PB) (Liang et al., 2011a) (Figure 6b). In mice, at the level of the trochlear nucleus in the isthmic region, the cuneiform nucleus (CuF) shows ipsilateral projections to the spinal cord (Liang et al., 2011a). At the same isthmic level, the midbrain reticular formation (mRt) shows bilateral projections to the spinal cord (Liang et al., 2011a). In larval zebrafish, we found a similar topographical organization for the spinal-projecting neurons located rostrolaterally to the PB neurons, where both ipsilateral and contralateral projecting groups can be identified (Figure 6b). The ipsilateral projecting group could correspond to the cuneiform nucleus (CuF) or the caudal pole of the mRt, while the contralateral projecting group likely corresponds to mRt (Figure 6b). Rostral to the LC, labeled somata were present with an ipsilateral prominence, likely belonging to the pedunculotegmental nucleus (PTg) or/and the laterodorsal tegmental nucleus (LDT) (Figure 6b) in the location where we previously functionally identified the mesencephalic locomotor region (MLR, (Carbo-Tano et al., 2023)). Along the midline, we found faint PAGFP+ neurons located at the caudal pole of the superior raphe nucleus (SR) (Figure 6b, d). These raphe spinal-projecting neurons had an ipsilateral prominence but were difficult to label consistently. By performing bilateral optical backfills, the labeling intensity was enhanced, revealing the sublocation of the raphe spinal-projecting neurons in the caudal pole (Figure 6d).

Figure 6.

The pontine (P) region is defined rostrally by r2, containing the rostral trigeminal motor nucleus (5Nr), and caudally by rhombomere 4 (r4), which includes the Mauthner cell (M cell). In mammals, the bulk of spinal-projecting neurons in the pontine region is located in the pontis oralis nucleus of the reticular formation (PnO) (Liang et al., 2015, 2011a). This region is not well subdivided in mammals and contains spinal-projecting neurons with relatively large cell size (Liang et al., 2015). We found the same anatomical pattern in larval zebrafish, with spinal-projecting neurons with large somas located in the ventral planes, likely belonging to the PnO (Figure 7c). Dorsocaudally to the trigeminal motor nucleus, we found intensely labeled projecting neurons (Figure 6b, Figure 7c) likely belonging to the sensory trigeminal nuclei (s5) (Liang et al., 2015, 2011a). Caudal to the nuclei s5 at the ventrolateral alar plate in r4, the intermediate octavomotor nucleus (ION) shows numerous strongly labeled projecting neurons ipsilateral to the photoactivation site (Figure 7c) (Bussières et al., 1999).

Figure 7.

The retropontine (RP) region consists of rhombomere 5 (r5) and rhombomere 6 (r6) containing the abducens (6N) and facial nucleus (7N), respectively. At this level, we found large and well-labeled neurons that belong to the pontine caudalis nucleus (PnC) (Figure 7b, 7c). Apart from the reticulospinal neurons, our backfill method in larval zebrafish also revealed the vestibulospinal system. In addition to the vestibulospinal neurons of the intermediate octavomotor nucleus (ION), we found that the posterior octavomotor nucleus (PON) contained both ipsilateral and contralateral spinal-projecting neurons (Figure 7c-d). The ipsilateral vestibulospinal neurons of the PON are located caudally and smaller than the contralateral ones (Figure 7d) (Bussières et al., 1999).

The medulla is rostrally delimited by rhombomere 7 (r7) and caudally by the rostral spinal cord (Watson et al., 2019b). Medially, it houses the gigantocellular reticular formation, a structure comprising multiple nuclei: the gigantocellular reticular nucleus (Gi), the alpha part of the gigantocellular reticular nucleus (GiA), the ventral part of the gigantocellular reticular nucleus (GiV), and the lateral paragigantocellular reticular nucleus (LPGi) (Liang et al., 2016, 2011b). Within the Gi, ipsilateral descending neurons are concentrated in the ventral region, while contralateral spinal-projecting neurons are prominent at its rostral pole (Liang et al., 2011b). Ventral to the Gi, the GiA, GiV, and LPGi all contain substantial numbers of spinal-projecting neurons with a strong ipsilateral tendency (Liang et al., 2016, 2011b). In mammals, the GiA/V and LPGi are ventrally located nuclei running alongside an extra raphe serotonergic column (VanderHorst and Ulfhake, 2006).

Our results show that in zebrafish, spinal-projecting neurons of the medulla display a slight ipsilateral predominance (Figure 8a, Figure 1d, Figure 2c). Ventrally, the medially located inferior raphe nucleus (IR) is flanked by two columns of spinal-projecting neurons (Figure 8b, c), which most likely correspond to the GiA/V or LPGi, given their colocalization with serotonergic neurons as labeled in the Tg(pet1:GFP) transgenic line (Figure 8ci). Dorsal to these, a mass of densely packed spinal-projecting neurons with a sharp rostral boundary at r7 projects predominantly ipsilaterally to the spinal cord, and likely corresponds to the mammalian Gi (Figure 8b, c). Laterally, two well-defined bilaterally projecting columns likely correspond to the intermediate reticular nucleus (lRt), which separates the Gi from the more laterally situated parvocellular reticular nucleus (PCRt) (Figure 8b, c). At the caudal medulla, spinal-projecting neurons are more densely packed, making it challenging to distinguish putative homologous nuclei. Nevertheless, we propose that the ventral medullary reticular nucleus (MdV) lies caudal to the Gi, while the dorsal medullary reticular nucleus (MdD) is positioned lateral to the caudal pole of the lRt (Figure 8b, c). Finally, the spinal trigeminal tract (Sp5) appears to be delimited by well-labeled bilaterally projecting neurons located at the medullary border (Figure 8c).

Figure 8.

Discussion

This study provides a comprehensive description of the descending supraspinal system of larval zebrafish. Despite the proven track record of larval zebrafish as a powerful vertebrate model for systems neuroscience, descriptions of brain nuclei that project to the spinal cord have been surprisingly incomplete. Our approach revealed a complex descending system with an organization that rivals the one described in mammals (Cruce and Newman, 1984; Liang et al., 2011b; Nudo and Masterton, 1989; Tohyama et al., 1979). Using optical backfilling combined with two-photon photoactivation of PAGFP in transgenic larval zebrafish, we built the first comprehensive whole-brain map of spinal-projecting neurons (SPNs), identifying approximately ∼1,300 SPNs per larva distributed across all major brain regions. The hindbrain, particularly the medulla oblongata, contains the vast majority of SPNs, with the caudal reticular formation (Gi, GiA/V, LPGi) constituting the largest population. Strikingly, we detected telencephalic spinal projections originating from the olfactory bulb and subpallium, confirming that this pathway represents a conserved vertebrate feature rather than a mammalian innovation. In the diencephalon, distinct neuromodulatory populations were identified as SPNs, including oxytocinergic neurons in the preoptic area, dopaminergic neurons in the posterior tuberculum, and neurons of the accessory optic system. A key highlight of our work is the anatomical clarification of the interstitial pretectal nuclei: we propose that the GnRH2⁺ cluster corresponds to the Edinger-Westphal nucleus (EW), supported by co-expression of cart2 and uts1. In the mesencephalon, the red nucleus was identified as a compact cluster of just 5 neurons projecting contralaterally, and its connectivity with the cerebellum was confirmed using single-cell atlas data. Throughout the brainstem, our study identifies putative zebrafish homologs of key mammalian reticulospinal, vestibulospinal, and raphe-spinal populations, anchored by transgenic line expression data and brain atlas registration. Taken together, our findings offer a detailed description of the descending system in larval zebrafish that we hope will serve as a useful reference for both comparative vertebrate neuroanatomy and future functional work. Our findings indicate that the perceived ‘simplicity’ of the larval zebrafish brain reflects the constraints of classical labeling methods rather than an intrinsic biological property.

Achievements and limitations of the study

The incomplete description of the zebrafish descending system reflects the limitations of earlier labeling methods. Retrograde tracing relies on tracer uptake through the axon, and older techniques preferentially labeled large, myelinated axons while missing smaller neurons with unmyelinated fibers (Beresford, 1965; Papez, 1926; Torvik and Brodal, 1957). While tracers like horseradish peroxidase (HRP) and small fluorescent dyes can allow visualization of finer reticulospinal projections (Cabot et al., 1982; Sánchez-Camacho et al., 2001; Ten Donkelaar, 1976), the original zebrafish description (Kimmel, 1982; Kimmel et al., 1982), performed alongside similar HRP backfills in other vertebrates, still yielded a low neuron count that made comparative analysis difficult (Cruce and Newman, 1984). This likely reflects a combination of factors: the early developmental stage of the larvae, the effectiveness of the mechanical damage caused by pressure-injecting dye into the spinal cord, and the possibility that small larval axons close too quickly for effective dye incorporation. In contrast, optical backfilling with two-photon imaging avoids many of these issues by allowing controlled axon labeling. The method does, however, have limitations. In particular, somatic signal intensity depends on the ratio of photoactivated PAGFP ascending from the axon to the volume of the cell body. Neurons with large somata projecting to spinal cord remain faint, which is why the Mauthner cells, despite having the largest axons, were consistently the faintest, a pattern also seen in large neurons of the ICN, ND, and ventral SPNs at the PnO and PnC. Similarly, long-range neurons in the rostral diencephalon and telencephalon project thin axons to the spinal cord, resulting in faint somatic labeling and likely, underrepresentation in our data.

Traditionally, nuclei of supraspinal neurons projecting to spinal cord have been identified by first delineating brain regions using cytoarchitectonic criteria, and then confirming spinal projections through retrograde tracing (Cruce et al., 1999). Here, we took the opposite approach. Based on observations that the well-established principle that the vertebrate descending system is highly conserved across vertebrate species (Cruce et al., 1999; Cruce and Newman, 1984; Yamamoto et al., 2017); (Adli et al., 1999; Cabot et al., 1982; Cruce and Newman, 1981; Crutcher et al., 1978; Hermann et al., 2003; Sánchez-Camacho et al., 2001; Smeets and Timerick, 1981a; Ten Donkelaar, 1976; ten Donkelaar et al., 1980a, 1981a; Ten Donkelaar, 1982a), we used the location and projection patterns of retrogradely labeled neurons as an anchor for defining brain areas homology. This was a practical necessity, as the cytoarchitectonic boundaries in the brain of larval zebrafish are not as well defined as in other vertebrates, making anatomy alone an insufficient basis for homology comparisons.

The brain of 6-dpf larvae is actively developing, which could complicate comparisons with adult vertebrates. However, evidence from fish and other vertebrates suggests that the broad organization of the supraspinal descending system is established early in development (Auclair et al., 1993; Bosch and Roberts, 1994; Gross and Oppenheim, 1985), allowing us to make confident comparisons even at this developmental stage. Undoubtedly, the compact brain of a 6-dpf larva presents challenges, as many distinct nuclei are closely packed, making it difficult to delineate clear boundaries between adjacent groups. Whenever anatomical criteria were insufficient, we relied on molecular markers from the brain atlases to aid our comparative assessments. Altogether the homologies we proposed here should be considered as working hypotheses to be further tested using molecular developmental approaches to confirm homologies. We expect future studies integrating refined molecular, genetic, and functional data to challenge, expand, and refine the identifications proposed here (Pompeiano, 1973).

Telencephalic descending projections

Comparative studies across vertebrates have consistently described a broadly conserved organization of descending nuclei and spinal pathways (Cruce et al., 1999; Cruce and Newman, 1984). The most rostrally located area consistently found to have descending projections is the diencephalic preoptic area (Cruce and Newman, 1981; New et al., 1998; Prasada Rao et al., 1987b; Smeets and Timerick, 1981b; Ten Donkelaar, 1982a, 1976; ten Donkelaar et al., 1981b, 1980a). Historically, no descending cells had been observed in the telencephalon of non-mammalian vertebrates (Ronan and Northcutt, 1985; ten Donkelaar, 1988; Ten Donkelaar, 1982a), with the exception of studies in nurse sharks, tiger salamander, iguana, and lamprey that found telencephalic projections reaching the rostral spinal cord using anterograde labeling (Ebbesson and Schroeder, 1971; Kokoros and Northcutt, 1977; Ocaña et al., 2015; Ten Donkelaar, 1982a). Our optical backfilling here revealed a few neurons with descending projections from the telencephalon in larval zebrafish. Later examination of the ZExplorer larval zebrafish brain and spinal cord atlas (Du et al., 2025) corroborated these sporadic observations. The morphology of the neurons in the atlas shows that these few telencephalic cells only project to the most rostral levels of the spinal cord. This explains why they are so difficult to capture retrogradely. Our findings challenge the established assumption that the telencephalic projections to the spinal cord are a derived feature of mammals (Cruce and Newman, 1981; New et al., 1998; Prasada Rao et al., 1987b; Smeets and Timerick, 1981b; Ten Donkelaar, 1982a, 1976; ten Donkelaar et al., 1981b, 1980a), and go in line with recent results from lampreys showing that the telencephalic (cortical) descending pathways appeared very early in vertebrate evolution (Ocaña et al., 2015).

Diencephalic descending projections

The location and laterality of descending neurons in the hypothalamic preoptic area, rostral and intermediate hypothalamus, thalamus, and pretectum are consistent with descriptions in other early-diverging vertebrates, including tetrapods (Cruce and Newman, 1981; New et al., 1998; Prasada Rao et al., 1987b; Smeets and Timerick, 1981b; Ten Donkelaar, 1982a, 1976; ten Donkelaar et al., 1981b, 1980a). Our optical backfill results, combined by the expression patterns of various transgenic lines, corroborate that oxytocinergic and dopaminergic spinal inputs represent strongly conserved descending pathways across vertebrate species (Bohic et al., 2026; Oti et al., 2021; Ryczko et al., 2016; Sharples et al., 2014; Teclemariam-Mesbah et al., 1997). We paid particular attention to the delineation of the pretectal interstitial nuclei. Building on the premise that gnrh2⁺ neurons correspond to the EW nucleus (Kauffman, 2004), we were able to more precisely define the nuclei situated lateral to EW. Across many vertebrate species, EW neurons express a wide range of peptides (Priest et al., 2023) regulating many behavioural aspects (Kauffman, 2004; Kozicz et al., 2011; Priest et al., 2023). Anatomically, the mammalian EW nucleus has wide range projections through the brain profusely projections to the spinal cord (Topilko et al., 2022; Yamamoto et al., 1995b). Besides a role in anxiety and depression-like behavior, mammalian EW-GnRH2 neurons are involved in regulating food intake (Kauffman, 2004; Kozicz et al., 2011). Similarly, in fish, modulation of GnRH2 signalling alters feeding behavior (Hofmann, 2006; Marvel et al., 2021, 2019; Nishiguchi et al., 2012), demonstrating a conserved functional role.

In mammals, the INC and ND are two interstitial nuclei that share considerable anatomical and functional similarities (Fukushima, 1991, 1987; Zhao et al., 2024). Both nuclei belong to the accessory oculomotor complex, classically implicated in the control of eye and neck movements (Fukushima, 1991, 1987; Zhao et al., 2024). A key anatomical criterion for distinguishing INC from ND is their relative position with respect to the MLF axons. We used this landmark to delineate these two nuclei as distinct clusters on either side of the MLF. In zebrafish, the INC and ND have historically been studied as a single entity under the designation of the nMLF (Kimmel et al., 1982). However, a recent study in larval zebrafish demonstrated that the nMLF can be subdivided into lateral and medial clusters based on differential expression of VGLUT2 and VGLUT1, respectively (Berg et al., 2023). This molecular segregation is further supported by physiological evidence, as the two clusters serve distinct functions in locomotor control. Such functional specialisation aligns well with our anatomical delineation of the INC and ND based on anatomical evidence. Similar to what we show here in larval zebrafish, the INC-ND sends descending projections to the caudal medulla in cats and mice (Skinner et al., 1984; Zhao et al., 2024). The role of the INC-ND in controlling locomotion (Berg et al., 2023; Kashin et al., 1974; Severi et al., 2014; Thiele et al., 2014) is not exclusive to zebrafish as electrical stimulations of this area shows locomotor promoting effects in cats and rats (Berezovskii, 1992; Berezovskiĭ, 1991; K.V.Bayev et al., 1986). Moreover, studies in rats showed that INC-ND is necessary for the execution of locomotion upon electrical stimulation of the hypothalamic locomotor region (Sinnamon and Benaur, 1997), further supporting the evolutionary conservation of the pretectal interstitial nuclei role in locomotion.

The location of the putative PAG in larval zebrafish was proposed solely based on relaxin 3a (rln3a) gene expression (Donizetti et al., 2009, 2008; Spikol et al., 2024), without consideration of anatomical landmarks or comparative neuroanatomical criteria (Donizetti et al., 2009, 2008; Spikol et al., 2024). A recent report indicates that the neurons located in the above region tentatively labeled as PAG send ipsilateral descending spinal projections (Spikol et al., 2024). We challenge this designation of PAG in larval zebrafish. In mammals, RLN3 expression within the PAG is sparse and largely confined to its dorsal subdivision (Smith et al., 2010), a region devoid of spinal-projecting neurons (Liang et al., 2011b). By contrast, RLN3 is far more prominently expressed in the mRT, located laterally to the RN, with clear descending projections to the spinal cord (Liang et al., 2011b; Smith et al., 2010). The population described by Spikol and collaborators (Spikol et al., 2024) aligns anatomically with the characteristics of the mRT rather than the PAG, and we therefore propose that their designation represents a misidentification of these two structures. Instead, we propose that the PAG in larval zebrafish is located medially and dorsal to the interstitial nuclei INC, ND, and EW, a position that better respects its defining anatomical relationship to the cerebral aqueduct and is consistent with the location described across vertebrates (Olson et al., 2017). It is worth noting that as development progresses, the expansion of the cerebral aqueduct displaces the PAG away from the midline, creating the impression of a more lateral position. However, its relative location with respect to surrounding nuclei remains unchanged throughout (Goodson and Bass, 2002; Kittelberger and Bass, 2013)

Mesencephalic pathways

The tectospinal neurons were not not identified in teleosts (Kimmel et al., 1982; Lee and Eaton, 1991b; Oka et al., 1986; Prasada Rao et al., 1987b), but found in elasmobranchs (Smeets and Timerick, 1981a) and lungfishes (Ronan and Northcutt, 1985). The descending tectal afferent neurons reach only the most rostral spinal levels (Altman and Carpenter, 1961; Foster and Hall, 1975; Luiten, 1981; Rubinson, 2008). In contrast to previous reports (New et al., 1998; Oka et al., 1986; Prasada Rao et al., 1987b), our approach revealed, that actinopterygian fishes possess tectospinal projections matching the locations and laterality described for other vertebrate species. The red nucleus (RN) is defined by its relative position in the tegmentum mesencephali, at the level of the oculomotor nerve, rostral to the decussation of the brachium conjunctivum. Classically, the criteria used to identify a structure as RN are, apart from its relative position, its contralateral spinal projection (ten Donkelaar, 1988). Here, we identified the RN in larval zebrafish as a small and compact cluster of neurons located in the tegmentum, lateral to the oculomotor nucleus, with clear contralateral spinal projections. In teleosts, the contralaterally projecting RN is defined as the “nucleus ruber of Goldstein” (NRg) (Goldstein, 1905; Prasada Rao et al., 1987b; Yamamoto et al., 2017).

Reticulospinal neurons: conserved core of motor control

The reticular formation is a phylogenetically conserved collection of neurons dispersed throughout the medulla, pons, and isthmus, embedded within a dense network of longitudinally and commissurally projecting fibers (Balcells, n.d.; Brodal, 1957; Pompeiano, 1973). While Meessen and Olszewski (Meessen and Olszewski, 1949) established the modern mammalian nomenclature of the reticular formation based on cytoarchitectonically distinct nuclei, this precise parcellation cannot be directly extended to non-mammalian vertebrate species, for which the reticular formation has not been characterized with the same anatomical resolution (Cruce and Nieuwenhuys, 1974; Nieuwenhuys, 1974, 1972). Reticulospinal neurons constitute the most ancestral descending system involved in motor control in all classes of vertebrates (ten Donkelaar, 1988). By using the location and projection patterns of retrogradely labeled neurons as criteria for defining tentative homologous brain areas, we were able to confidently assign designations to the reticular formation of larval zebrafish, relying on comparisons with mammalian nuclear identification.

To place our findings in a broader evolutionary context, we compiled and compared data on reticulospinal neurons across vertebrate species (Figure 9). For this comparison, we excluded both vestibulospinal neurons, which fall outside the reticular formation proper (Brodal, 1957; Pompeiano, 1973; Ten Donkelaar, 1982a), and raphespinal neurons, whose representation is inconsistent across specimens due to their unreliable labeling with classical tracing techniques. To further facilitate cross-species comparisons, we follow the parcellation proposed by Hoevell (van Hoevell, 1911), who divided the reticular formation into three regions based on the arrangement of its largest cells: the superior reticular nucleus (SRN), spanning from the isthmus to the rostral trigeminal nucleus in rhombomere 2; the middle reticular nucleus (MRN), extending from rhombomere 3 to rhombomere 6 at the level of the facial nucleus; and the inferior reticular nucleus (IRN), running from rhombomere 7 to the obex. We also follow the convention of excluding midbrain and diencephalic structures from the reticular formation (Brodal, 1957; Pompeiano, 1973). This comparative analysis reveals a striking degree of conservation in the topographical organization of reticulospinal neurons across vertebrates. In the IRN, these neurons are densely packed and medially concentrated, with their distribution terminating abruptly at rhombomere 6. Within the MRN and SRN, reticulospinal neurons fan out into two distinct extensions toward the isthmic region, giving the overall distribution a characteristic Y-shape in dorsal view (Figure 9). Our zebrafish results fit seamlessly within this conserved organizational plan, reinforcing the view that the fundamental architecture of the reticulospinal system was established early in vertebrate evolution (ten Donkelaar, 2009). Within this conserved structural plan, early-diverging vertebrate species tend to have fewer but larger reticulospinal neurons compared to mammals (Cruce and Nieuwenhuys, 1974; Heijdra and Nieuwenhuys, 1994; Newman and Cruce, 1982; Nieuwenhuys, 2011; Nieuwenhuys and Oey, 1983; Opdam et al., 1976), and the progressive increase in neuron number across lineages likely reflects the growing complexity of the body plan, the emergence of appendages, and the accompanying expansion of the motor repertoire.

Figure 9.

Data availability

Data will be shared when manuscript is under review.

Acknowledgements

We deeply thank Professor Yonathan Zohar for sharing the Tg(GnRH2:EGFP) transgenic line. We thank Camille Lejeune, Antoine Arneau and Nolwenn Jezequel, Noémie Frère and Sophie Nunes Figueiredo, from Phenoparc their fish care. We thank Dr. Olivier Renaud, Astou Tangara and Claire Lovo from the ICM.Quant core facility.

Additional information

Contributions

M.C.T: Conceptualization, methodology, formal investigation, formal analysis, plotting, writing original draft, review and editing. K.F: Reviewing and editing. T.W: Assisted performing initial backfills experiments. S.N, and M.A. Generated Tg(elavl3:stGtACR1-FusionRed) transgenic line. R.D: Review and editing. C.W: Conceptualization, review and editing, funding acquisition.

Funding

This work was supported by the Richard Mille Fund for the project “From genetic mutations to motor defects and recovery”, the Fondation Bettencourt-Schueller (FBS-don-0031), the European Research Council (ERC Consolidator ERC-CoG-101002870), the team grant Fondation pour la Recherche Médicale (FRM-EQU202003010612), the Agence Nationale pour la Recherche (ANR) LOCOCONNECT (ANR-22-CE37-0023 LOCOCONNECT), RocSMAP (ANR-23-CE16-0017-02 RocSMAP), ASCENTS (ANR-21-CE13-0008 ASCENTS), MOTOMYO (ANR-21-CE14-0042 MOTOMYO), CIRCOLO (ANR-24-CE16-7992 CIRCOLOCO). The Wyart team was also awarded European Union’s Horizon 2020 Research and Innovation program under the Marie Skłodowska-Curie Grant No. 813457 (https://zenith-etn.com) at the Paris Brain Institute (Institut du Cerveau).

Funding

Richard Mille Foundation (RMF) (DBS-2023)

• Kevin Fidelin

• Martin Carbo-Tano

• Claire Wyart

Agence Nationale de la Recherche (ANR) (ANR-24-CE16-7992 CIRCOLOCO)

• Claire Wyart

Agence Nationale de la Recherche (ANR) (ANR-23-CE16-0017-02 RocSMAP)

• Claire Wyart

• Kevin Fidelin

Agence Nationale de la Recherche (ANR) (ANR-22-CE37-0023 LOCOCONNECT)

• Kevin Fidelin

• Claire Wyart

Agence Nationale de la Recherche (ANR) (ANR-21-CE14-0042 MOTOMYO)

• Kevin Fidelin

• Claire Wyart

• Martin Carbo-Tano

Agence Nationale de la Recherche (ANR) (ANR-21-CE13-0008 ASCENTS)

• Claire Wyart

EC | European Research Council (ERC)

https://doi.org/10.3030/101002870

• Martin Carbo-Tano

• Claire Wyart

• Kevin Fidelin

Fondation pour la Recherche Médicale (FRM) (FRM- EQU202003010612)

• Claire Wyart

Fondation Bettencourt Schueller (Bettencourt Schueller Foundation) (FBS-don-0031)

• Claire Wyart

Additional files

Supplemental Video 1.