Zebrafish breeding and care were performed as previously described (Westerfield, 1993). Embryos were kept in the dark at 28.5°C and raised in 1x E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄) with 0.2 mM phenylthiourea to prevent pigmentation. The staging was performed as previously described (Kimmel et al., 1995). The following transgenic lines were used in this study: th ^m1512Tg and th ^m1513Tg (this study); Tg(QUASr:GFP)c403 (Subedi et al., 2014); Tg(QUAS:nlsEGFP, myl7:TagRFP)m1523 (Altbürger and Driever, unpublished).

To generate a QF2-driver line for th, a previously reported knock-in technique was adapted (Li et al., 2015). Specifically, the plasmid th-P2A-Gal4 donor (Addgene #65563) was digested with BamHI and AgeI to excise the coding sequence for P2A-Gal4. The coding sequence for E2A-QF2 was amplified from the plasmid pBait-E2A-QF2 (as reported below) with the following primers: QF2 fwd2—TTCTCACAGATGCCCTGAATGTGTTGGCTGGATCCGGACAGTGTACTAATTATGCTCTCTTGAAATTGG; QF2 rev2—AAGCATAGAGGAATGATTAAGCAGAATTAATCACTGTTCGTATGTATTAATGTCGGAG. Using the NEBuilder^® HiFi DNA Assembly Cloning Kit, all the fragments were assembled. The plasmid pBait-E2A-QF2 was generated by (1) digestion of p3E-polyA (Kwan et al., 2007) with EcoRI and BamHI; (2) amplification of the eGFPbait + E2A sequence from the plasmid eGFPbait-E2A-KalTA4 donor (Auer et al., 2014) using the following primers: eGFPbait + E2A fwd—TAAGCTCGGGCCCTGCAGCTCTAGAGCTCGATAGTGGTACCATGGTGAGCAAG; eGFPbait + E2A rev—GCGCTTGGGTGGCATGGGACCTGGGTTGCTCTC; and (3) amplification of the QF2 sequence from the plasmid pME-QF2 (Addgene #83307) and primers QF2 fwd—AGCAACCCAGGTCCCATGCCACCCAAGCGCAAAAC; QF2 rev—CAAACTCATCAATGTATCTTATCATGTCTGTCACTGTTCGTATGTATTAATGTCGGA. All three fragments were assembled using the NEBuilder^® HiFi DNA Assembly Cloning Kit. To generate the th sgRNA, the plasmid pT7-th-sgRNA (Li et al., 2015) was linearized with HindIII, and the sgRNA was transcribed using the MEGAscript^TM T7 Kit, followed by ammonium acetate precipitation. For Cas9 mRNA transcription, the plasmid pCS2+hSpCas9 (Addgene #51815) (Ansai and Kinoshita, 2014) was linearized using NotI, and mRNA was transcribed using the mMESSAGE mMACHINE^TM SP6 Transcription Kit. To generate the knock-in of QF2 into the th locus, Cas9 mRNA, th sgRNA, and the th-E2A-QF2 plasmid were injected into single-cell stage embryos of an outcross of Tg(QUASr:GFP)c403 (Subedi et al., 2014) fish to ABTL. GFP-positive embryos were raised to adulthood and outcrossed to ABTL fish. GFP-positive F1 embryos from three independent G0 founders were raised to adulthood and used for the generation of stable F2 generations, of which th ^{m1512Tg(2A−QF2)} was used in this study. All three alleles show Mendelian segregation. The knock-in approach relies on non-homologous end-joining (NHEJ) and is expected to integrate the eGFPbait + E2A plasmid into the th locus (Li et al., 2015). The th ^m1512Tg was not confirmed by sequencing of the locus, given that the functional QF2 protein indicates that the knock-in occurred in frame to generate the Th-E2A-QF2 open reading frame.

Whole-mount immunofluorescence staining of larvae at 96 or 120 hpf stages was performed as previously described (Ronneberger et al., 2012). The following primary antibodies were used in this study at the indicated concentrations: chicken anti-GFP polyclonal (Thermo Fisher Scientific, #A-10262; 1:400); mouse anti-GFP monoclonal Living Colors JL-8 (ClonTech/Takara Bio USA, Inc., #632381; 1:400); rabbit polyclonal anti-Th directed against the polypeptide encoded by bp 274-1,097 of the th ENSEMBL transcript XM683987 (Ryu et al., 2007; Kastenhuber et al., 2010) (1:550), which does not crossreact with neurons exclusively expressing Th2 protein (Figures 4, S2, and S3 in Filippi et al., 2010); and rabbit anti-Gad1b/2 polyclonal (Abcam, ab11070; 1:500). The following secondary antibodies were used at a concentration of 1:1,000: goat anti-chicken IgY Alexa 488 (Thermo Fisher Scientific, #A-11039); goat anti-mouse IgG Alexa 488 (Thermo Fisher Scientific, #A-11001); and goat anti-rabbit IgG Alexa 555 (Thermo Fisher Scientific, #A-21430). Nuclei were stained using TOTO-3 as described before (Ronneberger et al., 2012). The immunofluorescence stainings were recorded using either an upright or inverted Zeiss LSM 880 system with the following lens: LD-LCI Plan Apochromat 25x/0.8 DIC. The histograms were adjusted linearly with ZEN. Lateral xz-views of dorsal z-stacks were created using the Reslice function in FIJI.

The heads of 30 dpf fish were fixed in 4% PFA overnight at 4°C. The fixation was stopped with several PBST (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 0.1% Tween-20) washes, and the brains were dissected in PBST. After dehydration through subsequent washes in increasing concentrations of MeOH (25, 50, and 75% MeOH in PBST), the brains were stored in 100% MeOH at −20°C. For immunofluorescence staining, the brains were rehydrated and washed three times with PBST. Subsequently, the brains were incubated in Solution 1.1 {10% THEED [aminoalcohol N,N,N′,N′-Tetrakis(2-hydroxyethyl)ethylenediamine; Sigma-Aldrich, 87600-100ML], 5% Triton X-100, and 5% urea in dH₂O (Pende et al., 2020)} overnight at 37°C. After several washes with PBST, the brains were blocked with 5% goat serum and 1% BSA in PBSTD (PBST with 1% DMSO) for 5 h at room temperature. The primary antibodies (chicken anti-GFP 1:400 and rabbit anti-Th 1:550, as detailed above) were applied in blocking solution (as detailed above) and incubated overnight at 4°C. After several washes with PBSTD, secondary antibodies (goat anti-chicken Alexa 488 1:1,000 and goat anti-rabbit Alexa 555 1:1,000) were applied in 1% blocking reagent (Roche, 11096176001) in PBSTD and incubated overnight at 4°C. Secondary antibodies were washed out with four washes of PBSTD and four washes of PBST. The stained brains were stored in 80% glycerol in PBST at 4°C. The brains were mounted in 1% agarose in 80% glycerol in glass-bottomed Petri dishes (MatTek, P35G-1.5-20-C) and recorded using an inverted Zeiss LSM 880 system with the following lens: LD-LCI Plan Apochromat 25x/0.8 DIC. The histograms were adjusted linearly with ZEN Black.

Whole mount FISH was performed on the dissected brains of larvae fixed at 120 hpf with 4% PFA overnight at 4°C. The staining procedure and the combination of fluorescent in situ hybridization with immunofluorescence staining were performed as previously described (Ronneberger et al., 2012), except that the permeabilization was achieved through 1 h incubation in 4% Triton X-100 in PBST instead of Proteinase K treatment. The following probes were used: ddc (Filippi et al., 2012); en1a (Ekker et al., 1992); en2a (Ekker et al., 1992); nr4a2a (Filippi et al., 2007); nr4a2b (Filippi et al., 2007); pax2a (Krauss et al., 1991a); pitx3 (Dutta et al., 2005); slc6a3 (Holzschuh et al., 2001); slc18a2 (Schredelseker et al., 2020). The whole mount fluorescent in situ hybridization coupled with immunofluorescence staining was recorded using either an upright or inverted Zeiss LSM 880 system with the following lenses: LD-LCI Plan Apochromat 25x/0.8 DIC or LD-LCI Plan Apochromat 40x/1,2 autocorr. The histograms were adjusted linearly with ZEN Black.

The larvae at 120 hpf or 15 dpf were anesthetized with 612 μM tricaine and mounted in 1% low melting agarose (Biozym, 850080) in 1x E3 medium in a glass-bottomed dish (Cellvis—D60-30-1.5-N), kept in E3 medium with 612 μM tricaine, and imaged at an upright Zeiss LSM880 system with a W-Plan Apochromat 20x/1.0 DIC objective and a 2-photon laser (Coherent Vision II) at the wavelength of 960 nm and 8.19 μs pixel dwell time. Z-stacks of two tiles with 10% overlap were recorded with a z-step of 1.23 μm. The histograms were adjusted with either ZEN Black or Adobe Photoshop. Sagittal views of dorsal z-stacks were created using the Reslice function in FIJI. The video was created using FIJI and Shotcut (www.shotcut.org).

Whole mount HCR RNA-FISH stainings were performed as described in the manufacturer's protocol (Molecular Instruments; https://files.molecularinstruments.com/MI-Protocol-HCRv3-Zebrafish-Rev7.pdf). The ssDNA oligos used in the corresponding oligo mixes for the mRNAs dbh, ddc, slc6a3, slc18a2, and th are listed in Supplementary Table S1. The following hairpin pairs (h1 and h2) were used: B4 Alexa Fluor^® 546; B2 Alexa Fluor^® 647; and B3 Alexa Fluor^® 647 (all purchased at Molecular Instruments). After the staining procedure, the larvae were stored in 80% glycerol in PBST at 4°C. The whole-mount HCR stainings were recorded using either an upright or inverted Zeiss LSM 880 system with the following lens: LD-LCI Plan Apochromat 25x/0.8 DIC. The histograms were adjusted linearly with ZEN Black.

All figures were assembled using Adobe Photoshop versions from CS4 to 24.1.1. Individual image planes, stack or sub-stack Z-projections, or orthogonal projections were exported as TIFF 8-bit images from ZEN or FIJI. Visualization of somata and projections from confocal image stacks was difficult due to the very different pixel intensity values, caused, first, by cytoplasmic GFP being predominantly localized to somata but not projections, and second, by fluorescence efficiency loss in deep ventral brain regions in the dorsal view recorded stacks. To better visualize both somata and projections across deep image stacks, we adjusted for individual image panels the whole image using non-linear gamma adjustments. The following steps were applied in Photoshop: whole individual images were selected and Photoshop Image Adjustment Levels menu applied, first, making minor linear adjustments to fill the 8-bit histogram without cutting information, and second, making non-linear adjustment of the mid-tone slider to make image content in the darker pixel range better visible. Therefore, signals in the image panels represent the anatomical location of fluorescence but do not linearly represent fluorescence intensities.

To drive the expression of genes of interest in catecholaminergic neurons in zebrafish, we utilized CRISPR/Cas9-mediated genome editing to insert the Q-system transcription factor QF2 (Potter et al., 2010; Riabinina et al., 2015) into the th locus. The Q-system in zebrafish has been reported to be less prone to transcriptional silencing in comparison with the widely used Gal4/UAS system (Subedi et al., 2014; Ghosh and Halpern, 2016; Burgess et al., 2020). We targeted the tyrosine hydroxylase (th) gene using a previously described knock-in strategy (Li et al., 2015) to generate an open reading frame fusion of Th with E2A-QF2. Injection of Cas9 mRNA, a sgRNA targeting the last intron of th, and a donor plasmid in single-cell stage embryos resulted in a successful knock-in of QF2 (Figure 1). The donor plasmid contains parts of the last intron of th such as the sgRNA target sequence, exon 13 of th fused with an E2A-QF2 encoding sequence, and 668 bp of genomic sequences downstream of the th gene (Figure 1A). After injection and rearing of G0 and F1 generation, we identified three independent th^Tg(2A−QF2)F2 alleles: th ^m1512Tg; th ^m1513Tg; th ^m1514Tg. All three independent alleles show similar expression of GFP when crossed with Tg(QUASr:GFP)c403 (shown for th ^m1513Tg in Supplementary Figure S1). We chose th ^m1512Tg for further analysis of our knock-in transgenic line and will use th ^{m1512Tg(2A−QF2)} as a synonym for this allele throughout this study.

Using in vivo 2-photon microscopy, we generated an overview of GFP expression in th ^{m1512Tg(2A−QF2)}, Tg(QUASr:GFP)c403 transgenic larvae at 5 dpf (Figures 1B, C), which revealed GFP in all previously described larval DA groups in the forebrain: in the olfactory bulb, subpallium, pretectum, optic recess region (ORR; previously called the preoptic area in teleosts), and in the groups numbered 1–7 in the ventral thalamus, posterior tuberculum, and hypothalamus. Furthermore, all NA groups in the hindbrain (medulla oblongata, area postrema, locus coeruleus), and NA sympathetic neurons in the peripheral nervous system, express GFP (for CA group nomenclature, Rink and Wullimann, 2002; Filippi et al., 2014). Furthermore, we observed prominent GFP expression in the NA cells of the carotid body (data not shown). In addition, we detected GFP+ cells in the midbrain and the rostral hindbrain (rHb) close to the midbrain–hindbrain boundary (MHB), which do not belong to previously identified Th+ populations. These GFP+ cells are not a population that would only transiently express GFP, as we also observed these cells at more advanced larval stages, such as 15 dpf (Supplementary Video S1) and 30 dpf (Supplementary Figure S4). Moreover, the number of rHb GFP+ cells close to the MHB increases during development into juvenile stages (Figures 1B, C; Supplementary Video S1). The rHb GFP+ cells adjacent to the MHB appear to send out axons to the contralateral side and project to the dorsal cerebellum (Figures 1B, C; Supplementary Video S1). For technical reasons, we did not investigate th^{m1512Tg(2A−QF2)} expression in the adult zebrafish brain; however, we note that analyses of adult stages will be important to better understand the significant changes in the catecholaminergic systems from larval to juvenile and adult stages. The QF2-driver line in combination with photoactivation of Cre^ERT (Sinha et al., 2010) and QUAS:floxed fluorescent transgenic responders, when available, may provide an opportunity to address the anatomical correlation and differences between larval and adult catecholaminergic groups.

To determine whether all GFP+ cells also express Th, we performed whole-mount immunofluorescence for GFP and Th on th ^{m1512Tg(2A−QF2)}; Tg(QUASr: GFP)c403 heterozygous embryos at 96 hpf (Figure 2). An analysis of the DA neurons in the forebrain showed a strong coincidence of GFP and Th expression. Very few cells expressed GFP but were not Th immunoreactive (Figures 2C–C”: a cell at the telencephalic midline ventricular wall marked by an asterisk), which may represent cells that only transiently express Th but have retained GFP because the binary QF2 expression system enhances both the level and time of GFP expression or may represent ectopic transgene expression. Because we did not consistently observe such GFP-positive Th-negative cells in different larvae, we favor the idea that these cells transiently express QF2 from the th locus. Transient TH-expressing populations have also been documented in the developing human brain (Puelles and Verney, 1998).

All DA neurons in the pretectum express both GFP and Th (Figures 2A–A”). The DA neurons in the olfactory bulb and the subpallium show both GFP and Th signals (Figures 2B–B”). This is also the case for the DA neurons in the optic recess region (Supplementary Figure S2). For the DA groups initially numbered 1–7 (Rink and Wullimann, 2002), we will refer to them as dopamine cluster (DAC) 1–7 to indicate that we are focusing on the DA cluster. We use the DAC1–DAC7 nomenclature because the exact correlation of the early larval DA clusters with the anatomically defined clusters in the adult brain is still not fully understood (Kaslin and Panula, 2001; Rink and Wullimann, 2002; Yamamoto and Vernier, 2011). The DA clusters in the prethalamus (DAC1), in the posterior tuberculum and hypothalamus (DAC2-6; Figures 2C–C”), and the posterior recess region (DAC7; Supplementary Figure S2) coexpress GFP and Th. We found that Th is expressed at much lower levels in SP, DAC1, and DAC7 DA neurons, while the amplification by QF2 in the transgene drives higher levels of GFP expression, generating bright GFP+ signals. We also analyzed the previously uncharacterized GFP+ cells in the midbrain for Th coexpression (Figures 2D, E). We find that these Mb cells also express Th, albeit at low levels compared with the catecholaminergic groups. The magnification of two exemplary GFP+ cells in the midbrain (Figures 2, D1–D3, E1–E3) demonstrates that the brightly stained GFP+ somata also have a clear cytoplasmic halo of Th immunoreactivity. To determine whether these cells are located in the midbrain, we first performed immunofluorescence in combination with nuclear staining using TOTO-3 (Supplementary Figure S3). These data show that the Th+ and GFP+ cells are indeed located in the midbrain and not in the diencephalon (Supplementary Figures S3A”', B”', C”'). These cells most likely reside in the tegmental part of the midbrain, as their location appears ventral to the optic tectum in sagittal optical sections (Supplementary Figures S3C–C”').

We investigated whether midbrain th mRNA expression had been previously detected in high-quality HCR mRNA in situ expression data and identified an entry in the “mapzebrain” gene expression atlas (Shainer et al., 2023), showing th HCR data in 6 dpf larvae. The data (https://api.mapzebrain.org/media/Lines/th/average_data/T_AVG_th.zip) also revealed th-expressing cells in a similar location of the midbrain. To determine whether the GFP+ cells in the midbrain still express Th at juvenile stages, we performed anti-Th and anti-GFP immunofluorescence on dissected brains of 30 dpf zebrafish (Supplementary Figure S4). As a control, we recorded GFP and Th immunofluorescence in the telencephalon and pretectum (Supplementary Figures S4A, B). We found a strong coincidence of GFP and Th expression in both brain regions, with very few cells that are only Th+ but not GFP+ in the subpallium (Supplementary Figure S4A). These neurons could potentially be recently added new dopaminergic neurons, which may start to express GFP only after a short delay due to the time required for the QF2 expression and its re-entry into the nucleus to initiate GFP expression. The analysis of GFP+ cells in the midbrain revealed that these cells express low levels of Th at 30 dpf, similar to 5 dpf larval stages, (Supplementary Figure S4C and magnified area in C1–C3).

Catecholaminergic and serotonergic neurons share common enzymes and transporters for transmitter synthesis and processing (Flames and Hobert, 2011). To characterize the newly identified Mb and rHb Th-expressing cells, we analyzed the expression of the dopamine transporter slc6a3/dat, which is important for the reuptake of dopamine released into the synaptic cleft (Figures 3A–A”). slc6a3 is often expressed in DA neurons, and even though some mismatches between the expression of Th and slc6a3 have been reported (Holzschuh et al., 2001; Yamamoto et al., 2011), their combined expression may be a good indicator of a functional DA neuron. Whole brain FISH revealed no coexpression of GFP and slc6a3 in the midbrain of th ^{m1512Tg(2A−QF2)}; Tg(QUASr:GFP)c403 brains at 5 dpf (Figure 3A”, cell marked with Mb). The enzyme Dopa decarboxylase, Ddc, and the monoamine transporter, Slc18a2/Vmat2, are required for the synthesis of dopamine and serotonin and their uptake into vesicles, respectively (Flames and Hobert, 2011). Therefore, we analyzed their expression using whole brain FISH, in th ^{m1512Tg(2A−QF2)}; Tg(QUASr:GFP)c403 brains at 5 dpf. GFP-expressing cells in the midbrain and close to the MHB do not express the monoaminergic marker slc18a2, in contrast to other known monoaminergic neuron groups, e.g., in the pretectum or the LC (Figures 3B–B”). Similarly, we found no coexpression of ddc with GFP in the Mb and rHb clusters (Figures 3C–C”).

The whole brain FISH technique is suitable for detecting medium to high levels of expression but often fails to detect low-level gene expression. As the Mb and rHb GFP+ cells express low levels of Th protein, and Th protein in some cells is undetectable, we made use of the recently established HCR RNA-FISH 3.0 technique (Choi et al., 2018), which enables the detection of low-level transcripts. We designed HCR probes for th, dbh, ddc, slc6a3, and slc18a2 (Supplementary Table S1) and analyzed th ^{m1512Tg(2A−QF2)}; Tg(QUASr:GFP)c403 larvae at 5 dpf (Figure 4). Our HCR RNA-FISH analysis confirmed that GFP+ cells in the midbrain (Figures 4A, A', A”' and magnified cell in A1-A3) and rostral hindbrain (Figures 4C–C” and magnified cells in C1-C3) express low levels of th mRNA. To exclude that sequences of the QF2 transgene, including the plasmid backbone, might cause ectopic expression of th in these cells, we performed th HCR RNA-FISH on wild-type 96 hpf zebrafish (Supplementary Figure S5). As expected, we observed high expression of th in the A11-type neurons in the posterior tuberculum and hypothalamus (Supplementary Figure S5A). In addition, we found cells expressing low levels of th in the midbrain (Supplementary Figures S5B, B') and in the rostral hindbrain close to the MHB (Supplementary Figures S5C, C'). These findings confirm that Mb and rHb th+ are indeed endogenously expressing th in wild-type zebrafish. However, these GFP+ and th+ cells do not express the DA marker slc6a3, as we detected no transcripts in these cells (Figures 4A”, A”'). NA neurons express, in contrast to DA neurons, the enzyme dopamine beta-hydroxylase, Dbh, which catalyzes the conversion of dopamine to noradrenaline. Therefore, dbh expression is used to identify NA neurons in zebrafish (Guo et al., 1999). Expression analysis of dbh through HCR RNA-FISH revealed that the Mb and rHb GFP+ cells do not express dbh (Figures 4B–B”'). Thus, we do not consider these cells to be NA neurons. In addition, we performed HCR RNA-FISH for ddc and slc18a2, but did not observe any coexpression of these markers in Mb or rHb GFP+ cells at 5 dpf (Figures 4D–E”'). Catecholaminergic neurons synthesize, in addition to dopamine, other neurotransmitters like glutamate or GABA (Filippi et al., 2014). We found the neurons in the midbrain to be GABAergic, as immunofluorescence at 96 hpf revealed that they express the enzyme glutamate decarboxylase, Gad1b, or Gad2 (Supplementary Figure S6). In conclusion, the newly identified Th-expressing cells in the midbrain and rostral hindbrain lack the expression of key enzymes of dopamine and noradrenaline synthesis but may be GABAergic.

DA neurons in the mammalian midbrain are characterized by the expression of several transcription factors that define their regional identity and differentiation (Filippi et al., 2007; Arenas et al., 2015; Blaess and Ang, 2015). To determine whether the zebrafish Mb Th-expressing cells might share the expression of homologs of these factors, we analyzed the expression of en1a/2a, pitx3, and nr4a2a/b (Figure 5). We also analyzed the expression of pax2a and en1a/2a as anatomical markers of the MHB to clarify the exact location of the rHb GFP+ cells. The pax2a expression domain appears rostrally adjacent and in part intermingled with the rHb GFP+ cells at 5 dpf, indicating that the rHb GFP+ cells indeed are located in the rostral hindbrain (Figures 5A, B). Consistent with this finding, en1a/2a expression is rostrally adjacent to rHb Th+ cells (Figure 5C). Neither the Mb nor the rHb GFP+ cells express pitx3 (Figures 5D, E). In contrast, the Mb GFP+ cells express nr4a2a/b or are at least located in a nr4a2a/b-positive region of the midbrain (Figures 5F–F” and magnified cells in F1-F3). To clarify whether the Mb GFP+ cells are bona fide neurons, we imaged them at higher magnification (Figure 5B and magnified cells in Figure 5B'). These two GFP+ cells do indeed send out axons and are therefore likely to be neurons and not immature progenitor cells (Figure 5B'). Unfortunately, because the Mb GFP+ cells are very rare and express low levels of GFP, a mapping of potential projection targets using mosaic analysis (Tay et al., 2011) is not feasible. In conclusion, we show that th ^{m1512Tg(2A−QF2)}; Tg(QUASr:GFP)c403 larvae and juvenile fish develop previously unidentified GFP-expressing GABAergic neurons in the midbrain and GFP-expressing neurons in the hindbrain, both of which express Th/th, but no other catecholaminergic or mammalian midbrain DA marker genes, except nr4a2a/b.