Introduction

Thyroid hormones (THs) are essential for human central nervous system (CNS) development1. In humans, an inadequate supply of maternally derived THs (MTHs) during prenatal stages causes several neurological impairments, affecting a newborn’s psychomotor and cognitive development. The most severe condition results from mutations of the monocarboxylate transporter 8 (MCT8) gene, leading to a rare X-linked neurodevelopmental disorder, the Allan-Herndon-Dudley Syndrome (AHDS)2,3. AHDS patients present different severities of the condition that have been correlated with the particular mutations they carry4,5,6,7,8. One of the most prevalent pathological characteristics of AHDS is the low myelination levels found by brain magnetic resonance imaging (MRI) of these patients8,9,10.

The zebrafish slc16a2 (mct8) gene is also expressed during embryogenesis11,12. Notably, at physiological zebrafish temperature, Mct8 can only transport 3,5,3’-triiodothyronine (T3), showing that during zebrafish embryogenesis, the only form of MTH transported is T311. New evidence in human brain organoids derived from AHDS patient’s cells also shows impaired maternal T3 (MT3) transport by MCT8, highlighting that the lack of T3 is responsible for the pathophysiological outcome behind AHDS13. The emergence of zebrafish as a model species for AHDS brought into light another pathological aspect not previously found or looked for in humans or murine models. In these zebrafish models12,14, pronounced changes were found in the development of the hindbrain vasculature, which gave rise to the blood-hindbrain barrier (BHB). However, no TH signaling genes, including mct8, were expressed in zebrafish endothelial cells during development12, arguing that these cells are not direct targets of T3. In AHDS patient-derived induced pluripotent stem cells (iPSCs), the endothelial blood-brain barrier (BBB) in vitro differentiation suggests that it is not dependent on MCT8 expression and T3 function15, further indicating that endothelial BBB development is independent of T3 signaling. Although this study has shown that endothelial differentiation is not dependent on T3, it does not answer whether T3 is necessary for adequate brain vasculature development. Nonetheless, new evidence has emerged that shows impaired cortex vascularization in an 11-year-old patient with AHDS and also in a murine model of AHDS (mct8/dio2 knockout mice)16.

Previous evidence has pointed out that THs’ effect on angiogenesis is mediated by the αvβ3 integrin that serves as a membrane receptor for T417. TH has been found to promote angiogenesis18 via the PDGF-Akt pathway in adult mice hearts. In squamous cell carcinoma, TH promotes tumor angiogenesis19. In mice, in vivo and in vitro evidence shows that TH activating enzyme, Deiodinase type 2, is essential for muscle vascularization20. In all these cases, a common aspect of TH-induced angiogenesis involves downstream up-regulation of vascular endothelial growth factor-A (VEGFA) expression and secretion18,19,20. Post-natal induced hypothyroid mice present decreased brain vascularization of the CNS. Both in vivo and in vitro evidence shows that TH action is mediated by TH receptors and leads to increased VEGFA expression21. Together, these studies suggest that TH is essential for vascularizing several organs via VEGFA regulation.

Here, we provide evidence that MT3 is essential for the complete vascularization of the zebrafish hindbrain during embryogenesis. This action of MT3 is mediated via mct8 in a discrete pax6a+ neuroprogenitor cell (NPC) population, from 30 to 48 h post-fertilization (hpf), which are the source of vascular endothelial growth factor aa (vegfaa) that instruct the ingression of the sprouting central arteries (CtAs) into the hindbrain. Moreover, our findings suggest that the timing and combination of the expression of the thyroid hormone receptor aa (thraa) and thyroid hormone receptor ab (thrab) in pax6a + NPCs are essential for the chronological ingression of the sprouting CtAs into the hindbrain. Our findings show an unexpected role for MT3 in CNS angiogenesis in the zebrafish AHDS model. This work sheds light on the genetic and cellular mechanisms driven by MT3 in developing the brain vasculature.

Materials and methods

Zebrafish husbandry and spawning

Adult wildtype (AB strain), mutant and transgenic zebrafish were maintained in standard conditions in the Center of Marine Sciences (CCMAR) fish facility at the University of Algarve (Portugal) as described12. The zebrafish lines used were: wild-type AB strains, Tg(fli1:EGFP)22, Tg(kdrl:CaaX-mCherry)23, Tg(pax8:DsRed)24, Tg(gSA2AzGFF306A) (illuminating Cpne4+ cells)25,26 and mct8 knockout (KO, mct8(-/-))27. All experiments and fish husbandry were carried out by fully animal experimentation certified scientists according to the EU Directive 2010/63/EU and followed the Portuguese legislation for the use of laboratory animals (DL n°113/2013, 7 August). All experimental animals are younger than 5 days post-fertilization (dpf), and no ethical approval is required accordingly to Portuguese (DL n°113/ 2013, 7 August) and EU (2010/63/EU) law.

Morpholino injection

Upon spawning, embryos were immediately collected and microinjected at the 1-2-cell stage with 1 nL of morpholino solution containing either 0.8 pmol control morpholino (CTRMO) or mct8 morpholino (MCT8MO) (Gene Tools, USA) as described12. Embryos were grown on plastic Petri dishes containing E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl, 0.33 mM MgSO4) and reared until sampling time at 28.5 °C in an incubator (Sanyo, Germany) under 12 h:12 h light: dark cycles. Staging was done after Kimmel et al.28 observing developmental landmarks in CTRMO embryos.

T3 treatment

Upon spawning, embryos were immediately collected and distributed on plastic Petri dishes containing E3 medium and maintained at 28.5 °C in an incubator (Sanyo, Germany). After 10 h, the experimental group embryos were transferred to E3 medium containing 20 μM 3,3’,5-Triiodo-L-Thyronine (T3, Sigma-Aldrich, USA), while the control embryos were maintained in E3 medium. Embryos were reared at 28.5 °C until sampling under 12 h:12 h light: dark cycles. The staging was done after Kimmel et al. observing developmental landmarks in control embryos28.

Rescue experiments

One-cell stage embryos were injected with either 0.8 pmol CTRMO, 0.8 pmol MCT8MO or 0.8 pmol MCT8MO + 100 pg vegfaa-165 mRNA as described in ref. 12. The vegfaa-165 gene was isolated from the pCR®4-TOPO plasmid using EcoRI (Thermo Fisher Scientific, USA) restriction enzyme according to the manufacturer’s instruction and cloned into the pCS2+ plasmid. 5 μg of pCS2+ vegfaa-165 was linearized with NotI (Thermo Fisher Scientific, USA) restriction enzyme and purified by phenol-chloroform. 1 μg of the linearized pCS2+ vegfaa-165 was used as the template for in vitro transcription using the mMESSAGE mMACHINE® Kit (Ambion, UK) following the manufacturer’s instructions. mRNA was quantified and stored at −80 °C until use.

Analysis of gene expression

Five independent biological replicates (pools of 10 embryos) were sampled at 28, 30, 32, 36, 40, 44, 48, 54 and 72 hpf for each experimental condition (MCT8MO and CTRMO). Embryos were manually dechorionated, snap-frozen in liquid nitrogen and stored at −80 °C until use.

Total RNA extraction and cDNA synthesis were performed as described27.

The quantification method used with the quantitative real-time PCR (qPCR) method was the absolute quantification method, which determines the number of mRNA copies in the sample from a standard curve as described in ref. 27. Primers were designed using Primer 3 Plus29 using RNAseq data30. Supplementary Table 1 provides primer sequences, amplicon size and RefSeq for each gene included in the analysis. qPCR assays were run in a 384-well plate (Axygen, Life Sciences, USA) with 6 μL of reaction mixture per well (1 × SensiFAST™ SYBR, No-ROX Kit, Bioline, USA), 150 nM forward primer, 150 nM reverse primer and 1 μL cDNA (1/5)). PCR cycling condition consisted of 95 °C for 3 min, followed by 44 cycles of 95 °C for 10 s and 60 °C for 15 s. A standard melting curve analysis was included to confirm the production of a single amplicon and the absence of primer dimers that consisted of a gradient temperature increment of 0.5 °C for 5 s from 60 °C to 95 °C. A negative control was included to detect genomic DNA contamination. Total RNA input was used as a normalizer since no commonly used reference gene (18S and gapdh) presented invariable expression during the embryonic stages analyzed31.

Fluorescent immunohistochemistry (IHC)

Tg(fli1:EGFP)22 experimental embryos were fixed at selected stages in ice-cold 4% paraformaldehyde (PFA)/1 × Phosphate-buffered saline (PBS) overnight at 4 °C. IHC was carried out as previously described in ref. 27. Samples were incubated overnight at 4 °C with 1:1000 rabbit anti-GFP (ab290, Abcam, UK) primary antibody and detected with 1:400 goat-anti-rabbit 488H + L (111-545-047, Jackson Laboratory, USA).

Double fluorescent immunohistochemistry (IHC)

A double IHC procedure was performed in Tg(gSA2AzGFF306A)25 and double transgenic (Tg(gSA2AzGFF306A);Tg(kdrl:CaaX-mCherry))23 experimental zebrafish embryos at 32, 36 and 48 hpf. Antibody detection and development of the signal were carried out sequentially. IHC was carried out as described previously. Primary antibodies used were rabbit anti-GFP (Abcam, UK, 1:1000). Second primary antibodies used were 1:100 mouse anti-mCherry-Tag (STJ34373, St John’s Laboratory, UK), 1:100 mouse anti-Zrf1 (ZDB-ATB-081002-46, ZIRC, USA), 1:500 mouse anti-HuC/D (16A11, Invitrogen, USA). Secondary antibodies were goat-anti-rabbit 488H + L (Jackson Laboratory, USA, 1:400) and goat anti-mouse IgG-CF594 (SAB4600321, Sigma-Aldrich, USA, 1:400). Samples were stored in 1 × PBS containing 0.1% Dabco (CarlRoth, Germany).

Colorimetric whole-mount in situ hybridization (WISH)

Riboprobe synthesis, hybridization, and imaging of colorimetric WISH were performed as described12. The vegfaa and vegfab plasmids were kindly provided by Brant Weinstein, the pdgfrb plasmid was kindly provided by Ching-Ling Lien, the gad67a, gad67b, glyt2a and glyt2b plasmids were kindly provided by Wolfgang Driever, the cadherin5 plasmid was kindly provided by Wiebke Herzog, the dsRed-1 plasmid was kindly provided by Raquel Andrade, the thraa and thrab plasmids were kindly provided by Sachiko Takayama. The zebrafish pax8 (XM 001339857.2) riboprobe was generated as described in Campinho et al. 12. The zebrafish pax6a riboprobe was generated after PCR cloning against the pax6a transcript30 (Supplementary Table 2). Images were acquired using an Olympus SX7X stereoscope with an Optika digital color camera.

Double fluorescent whole-mount in situ hybridization (WISH)

Riboprobes were generated as described in the previous section and labeled with either digoxigenin (Dig) (gad67a, gad67b, glyt2a and glyt2b) or fluorescein (Fluo) (dsRed-1). WISH was carried out as described in ref. 27. The signal from the first probe was generated after using the FastRed/Naphthol AS-MX substrate (Sigma-Aldrich, USA) and the signal from the second probe was generated with FITC-Tyramide (Perkin-Elmer, USA). Samples were stored until imaging in 1 × PBS containing 0.1% Dabco (CarlRoth, Germany).

Fluorescent whole-mount in situ hybridization (WISH) with immunohistochemistry (IHC)

Riboprobes were generated as described in the previous section and labelled with digoxigenin (Dig) (pdgfrb, pax6a, pax8 and vegfaa). A fluorescent WISH with IHC was performed using Tg(fli1:EGFP)22 or Tg(gSA2AzGFF306A)25 zebrafish embryos (32, 36, 48 hpf).

The procedure was carried out as described27, and fluorescent color development was carried out using FITC-Tyramide (Perkin-Elmer). The primary antibody used was rabbit anti-GFP (Abcam, UK, 1:500) and secondary antibody fluorescent labeling was carried out using goat anti-rabbit CF594 (Sigma-Aldrich, USA, 1:400). Samples were stored until imaging in 1 × PBS containing 0.1% Dabco (CarlRoth, Germany).

Double fluorescent whole-mount in situ hybridization (WISH) with immunohistochemistry (IHC)

Riboprobes were generated as described in the previous section and labelled with digoxigenin (Dig) (mct8, thraa and thrab) or fluorescein (Fluo) (pax6a). A double hybridization procedure combining one Dig and one Fluo probe was performed using Tg(fli1:EGFP)22 zebrafish embryos (30, 32, 36, 42, 48 hpf) as described27. Afterwards, IHC against GFP was carried out with a rabbit anti-GFP (Abcam, UK, 1:500) anti-serum. The secondary antibody fluorescent labelling was done using goat anti-rabbit CF633 (Sigma-Aldrich, USA, 1:400). Samples were stored until imaging in 1 × PBS containing 0.1% Dabco (CarlRoth, Germany).

Evans blue staining

For live observation of hindbrain blood vessel, 48 hpf wild-type (n = 15) and mct8(-/-) (n = 13) embryos were incubated for 10 min at room temperature in E3 medium with 0.1% Evans blue accordingly to Eliceiri et al. 32. After incubation, the embryos were washed five times for 5 min in an E3 medium without Evans blue. After staining, the embryos were mounted in 3% methylcellulose/E3 medium and imaged in a Zeiss Z2 microscope coupled to Hamamatsu Orca V2 digital camera.

Generation of pax6a loss-of-function mutant

CRISPRScan33 was used to design a guide RNA (gRNA) against exon 7 of the zebrafish pax6a locus (GGTTGAGGTTGTGCCCGAGG). gRNAs were in vitro transcribed as described33 and purified by sodium acetate and ethanol precipitation. pax6a gRNA was diluted to 300 ng/μL in a 600 ng/μL Cas9 protein (Weissman Institute, Israel) and injected in 1-cell stage zebrafish Tg(kdrl:CaaX-mCherry)23 embryos.

At 24 hpf, eight embryos per injection clutch were collected for genotyping by PCR. Genomic DNA extraction was carried out after overnight digestion at 50 °C in genomic extraction buffer (10 mM Tris pH8.2, 10 mM EDTA, 20 mM NaCl, 0.5% sodium dodecyl sulfate (SDS), 200 ng/mL Proteinase K), followed by centrifugation and washing with 70% ethanol, air dried, and resuspended in 20 μL of Tris-EDTA buffer (TE) pH 8. PCR was carried out with primers (0.2 μM) flanking the gRNAs binding sites (For – GATAGTGCACATTGTAGCAA; Rev – CAGCCCAGCCAGACCTCATCC) using the DreamTaq polymerase kit (Thermo Fisher Scientific, USA). Thermocycling was carried out as follows: 5 min 95 °C, 35 cycles of 30 s 95 °C, 30 s 58 °C, 30 s 72 °C, followed by a final 5 min step at 72 °C. PCR products were fractioned in a 3.5% agarose/1 × Tris-acetate-EDTA buffer (TAE) gel. gDNA of a control wild-type embryo was used as a negative control. Bands were isolated from the gel and extracted with a gel extraction band kit (OMEGA Biotek, USA), followed by Sanger sequencing using the Big-dye termination method.

The isolated band sequence was confirmed after BLAST analysis and alignment to the zebrafish pax6a locus. That ensured that injected clutches had embryos carrying the desired genetic lesions on the pax6a locus. Injected embryos were reared until adulthood. After isolation, adult-injected PCR genotyped fish after fin-clipping to identify carriers of genetic lesions on the pax6a locus. After sequencing, only carriers of mutations were allowed to cross with wild-type siblings to give rise to non-mosaic F1 carrier lines. Adult Tg(kdrl:CaaX-mCherry)23 F1 carriers were genotyped by PCR after fin-clipping and sequenced. In-crosses were carried out to generate F2 homozygous mutants for the pax6a locus.

Fluorescent live images of the hindbrain vasculature from Tg(kdrl:CaaX-mCherry)23 wild-type and pax6a(-/-) were taken at 4 dpf in a Zeiss Z2 microscope coupled to Hamamatsu Orca V2 digital camera.

Live imaging

Zebrafish Tg(fli1:EGFP)22 were mated, and 1-cell stage eggs were injected with either 0.8 pmol CTRMO or MCT8MO (Gene Tools, USA) as described12. At 28 hpf, zebrafish embryos were screened for GFP fluorescent protein driven by the fli1 promoter, a marker for endothelial cells. Zebrafish Tg(kdrl:CaaX-mCherry)23 and wild-type AB were mated and 1-cell stage eggs were injected with pax6a gRNA+Cas9 (300 ng/μL + 600 ng/μL) (Weissman Institute, Israel) as described previously. At 28 hpf, zebrafish embryos were screened for the membrane-bound Cherry fluorescent protein driven by the kdrl promoter, a marker for endothelial cells. Afterwards, embryos were screened for positive pax6a mutation by selecting embryos with a small eye34.

Imaging was carried out by light-sheet microscopy, Lightsheet Z1 (ZEISS, Germany). Briefly, embryos were anesthetized with 0.08% tricaine pH 7.4 buffered, mounted alive in 0.3% (w/v) low-melting agarose (LMA) (CarlRoth, Germany) in E3 medium containing tricaine (0.08%) into FEP tubes closed with a 1% LMA. Three animals per group, CTRMO and MCT8MO or not injected control and pax6a crispants, were imaged in the same tube. Two independent experiments were carried out. Time lapse images were taken from 32 until 50 hpf with a z-step of 0.740 μm to acquire the complete hindbrain and were imaged every 20 min for 20 h. The hindbrain was imaged with a 10 × lens, 1 × zoom with dual illumination. For image analysis, dual illumination images from the Z1 were merged using Dual side Fusion (Zen Black, Zeiss). Afterwards, images were imported into Fiji35 and analyzed.

Image acquisition and analysis

Fluorescent imaging of the hindbrain was carried out using a Zeiss Z2 microscope coupled to a Zeiss digital camera or a Lighsheet Z1 microscope (Zeiss, Germany). Samples for Z2 imaging were mounted in 0.3% agarose (Sigma-Aldrich, USA) and imaged using a 20 × lens and a z-step of 0.850 μm. Afterwards, images were deconvoluted in SCI Huygens software 4.4 (Scientific Volume Imaging, The Netherlands, http://svi.nl). Maximum projections were generated in Fiji35. Samples for Z1 imaging were mounted in 1% Low melting agarose (CarlRoth, Germany) and imaged using a 10 × lens, dual illumination and a z-step of 1.5 μm (optimal distance option) to acquire the complete hindbrain using 1 × zoom. After image acquisition, dual illumination images were merged using Dual side Fusion (Zen Black, Zeiss, Germany). Images were then imported into Fiji35 and analyzed.

Colocalization Colormap plugin36 was used to determine the colocalization of Dig and Fluo cRNA probes.

Statistical analysis

All statistical analyses were done in GraphPad Prism version 8.4.0 software for Mac (San Diego, USA, www.graphpad.com). Values are represented as means ± SD. The levels of statistical significance were expressed as p-values, *p < 0.05; **p < 0.01; ***p < 0.001; ns: non-significant. Due to the role played by the genes analyzed in embryonic development, the present work did not intend to determine their temporal expression patterns, only the effect of mct8 knockdown on their expression at specific time points. To determine gene expression differences between CTRMO and MCT8MO embryos, statistical significance was determined by unpaired Students t-test: two-sample, assuming equal variances. One-way analysis of variance (ANOVA) or Fisher’s exact test was used to determine statistically significant differences (p < 0.05) between control and experimental embryos.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Zebrafish hindbrain vasculature is altered in mct8 morphants and loss of function mutants

A previous study showed that the mct8 morphant zebrafish embryo at 48 hpf presents an impaired hindbrain vasculature development, displaying only 3 of the 7 central arteries (CtAs) (Supplementary Fig 1a)12. These results were further confirmed in the recently developed mct8 knockout mutant (mct8(-/-))27 using Evans blue staining (Supplementary Fig 1b). This evidence shows that MT3 is involved, via Mct8 transport, in CtAs and hindbrain vasculature development.

MT3 action during zebrafish hindbrain vasculature development

Previous transcriptomic analysis between control morpholino (CTRMO) and mct8 morpholino (MCT8MO) zebrafish embryos at 25 hpf, showed that before hindbrain vasculature development, angiogenic genes are differentially expressed (Supplementary Fig 2a, p < 0.01; FDR < 0.05)30. We extended gene expression analysis to a time window from 28 to 72 hpf (Supplementary Fig 2b) that encompasses the development of the hindbrain vasculature37. Despite different expression dispersion, whole-embryo quantitative real-time PCR (qPCR) gene expression analysis also did not find significantly different expression between CTRMO and MCT8MO embryos from 28 to 72hpf (Supplementary Fig 2b–d).

Although CtAs development is affected in MCT8MO embryos, this does not occur in intersegmental trunk vessels, indicating that potential MT3 action on CtAs is specific to the brain (Supplementary Fig 1a). Moreover, neither mct8 nor T3 receptors thyroid hormone receptor aa (thraa), thyroid hormone receptor ab (thrab), and thyroid hormone receptor b (thrb) are expressed in endothelial cells before 48 hpf12, suggesting an indirect effect on CtAs. Nonetheless, given the vascular hindbrain phenotype, we decided to analyze the in situ expression of the vascular endothelial growth factor a (vegfa) ligands given their essential role in zebrafish CtA development38. Whole-mount in situ hybridization (WISH) analysis revealed a decrease in vegfaa hindbrain expression at 36 hpf in MCT8MO embryos (Supplementary Fig 3a) and in a region where CtAs are expected to sprout (green arrowheads in Supplementary Fig 3a). WISH analysis for vegfab revealed a more widespread expression in the brain than vegfaa (Supplementary Fig 3). The expression pattern of vegfab between CTRMO and MCT8MO zebrafish embryos was mostly maintained. Still, the signal was less intense in MCT8MO than in CTRMO zebrafish embryos (Supplementary Fig 3b). The results suggest that MT3 is required for appropriate vegfaa and, to some extent, vegfab expression in the hindbrain.

Given that vegfaa is the principal angiogenic factor involved in CtA development38, we analyzed the spatial relationship between hindbrain vegfaa expression and the position of CtA ingression using the Tg(fli1:EGFP) transgenic zebrafish line (Fig. 1a). In CTRMO embryos expression of vegfaa at 32, 36 and 48 hpf can be observed juxtaposed to specific CtAs, showing the importance of vegfaa for CtA sprouting and migration. In MCT8MO embryos, vegfaa hindbrain expression was not lost, but significantly reduced (Fig. 1a–d). At 32 hpf, the knockdown of mct8 significantly reduced the expression of vegfaa in rhombomeres 2, 3, and 5 (Fig. 1b); at 36 hpf, it reduced considerably in rhombomeres 2, 3, and 4 (Fig. 1c); at 48 hpf, it was significantly reduced in rhombomeres 1 to 5 (Fig. 1d). These results argue that MT3 regulates vegfaa hindbrain expression during CtA development.

Fig. 1: MT3 regulates vegfaa expression during hindbrain vascular development.
figure 1

a Lateral maximum projections of the hindbrain of Tg(fli1:EGFP) after WISH against vegfaa (magenta) and immunostained against GFP (endothelial marker, white) in CTRMO and MCT8MO zebrafish embryos at 32, 36 and 48 hpf. Yellow arrowhead indicates agglomeration of vegfaa expression around a CtA or the putative location where a CtA is supposed to ingress into the hindbrain. The red box indicates the hindbrain area for every embryonic stage analyzed adapted from (Kimmel et al., 1995). Graphical representation of the number of times, defined as frequency (%), that vegfaa expression can be observed around a specific CtA or where a CtA is supposed to develop at b) 32 hpf, c) 36 hpf, d) 48 hpf between CTRMO and MCT8MO embryos; Fisher’s exact test.n = 22 (32hpf and 48hpf CTRMO), 15 (32hpf MCT8MO), 23 (36hpf CTRMO), 21 (36hpf MCT8MO), 18 (48hpf MCT8MO). e At 48 hpf, Tg(fli1:EGFP) zebrafish embryos co-injected with vegfaa-165 mRNA and MCT8MO (n = 24) are able to partially rescue, in a rhombomere specific manner, some but not all CtAs. CTRMO (n = 10), MCT8MO (n = 14). f Statistical analysis of the number of CtAs present in each experimental group; One-way ANOVA followed by Bonferroni’s multiple comparison analysis. Data are presented as box-and-whisker plot, where the black thick horizontal line represents the median. The first and third quartiles are marked by the lower and upper edges of the boxes, respectively. Error bars represent standard deviations (smallest and higher value). g Frequency (%) of each CtA to develop in each experimental group; Fisher’s exact test. **p < 0.01; ***p < 0.001; ****p < 0.0001. For detailed statistics, see Supplementary Data 2. The red arrowhead indicates the mid-cerebral vein (MCeV), and the green arrowhead indicates the primordial hindbrain channels (PHBC). Numbers in yellow 1 – 8 indicate the CtA in its respective rhombomere. Scale bar: 50 μm. Zebrafish drawings used from Kimmel et al (1995)(Ref. 28) under CC licence number 6041951178663.

To functionally determine if MT3 acts upstream of vegfaa during hindbrain vasculature development, we co-injected zebrafish vegfaa-165 mRNA with MCT8MO (Fig. 1e). Embryos injected with CTRMO have 5 to 7 CtAs present, as expected at this developmental stage37 (Fig. 1f). The number of CtAs is significantly reduced in the MCT8MO embryos, with an average of 2 CtAs present at 48 hpf (Fig. 1f). Analyzing the frequency of each CtA developed between CTRMO and MCT8MO embryos shows that all 7 CtAs were significantly affected (Fig. 1g). Exogenous supply of vegfaa-165 mRNA significantly increases the number of CtAs to an average of 4 CtAs compared to MCT8MO alone (Fig. 1f), being able to rescue CtAs 1, 2, 3 and 7 (Fig. 1g). The embryos co-injected with MCT8MO and vegfaa-165 mRNA partially rescued the phenotype (Fig. 1e, f) even though these were still significantly different. CtAs 5 and 6 were partially recovered by the exogenous supply of vegfaa-165 mRNA, but CtA 4 cannot be recovered (Fig. 1g). This evidence argues that for these CtAs vegfaa is likely not the only signal under the regulation of MT3 required for ingression into the hindbrain. Noteworthy, the ingression of CtAs into the hindbrain in co-injected MCT8MO+vegfaa-165 mRNA was not identical to CTRMO embryos (Fig. 1e), and these present a shallow ingression in the hindbrain. These can be due to impaired CtA development after ingression or a consequence of altered hindbrain development27,30, or both.

We further investigated if the impaired development of the hindbrain vasculature could give rise to changes in the neurovascular unit that is formed. To determine the involvement of MT3 in establishing a functional blood-hindbrain-barrier (BHB) in zebrafish we used brain pericyte recruitment for hindbrain vasculature as a proxy. We quantified the number of pericytes following WISH for platelet-derived growth factor receptor-b (pdgfrb) after T3 supplementation and in mct8 knockdown (Fig. 2a, b). Treatment with T3, compared with control embryos, did not affect the recruitment of pericytes to the hindbrain vasculature (Fig. 2a). However, compared to CTRMO, in mct8 morphants, there was a delay in the pericyte presence on the primordial hindbrain channels (PHCB) and basilar artery (BA) (Fig. 2b). In the PHBC of MCT8MO embryos, the pericyte number was significantly reduced at 48 hpf (Fig. 2f). In the BA, the pericyte number increased significantly at 36 hpf and then decreased significantly at 48 hpf (Fig. 2g, h). Interestingly, at 36 hpf, the BA in the MCT8MO was not yet completely formed, and the pericytes formed agglomerates in the posterior region of the BA (Fig. 2b). Nevertheless, the BA was completely formed at 48 hpf in MCT8MO embryos but presented fewer pericytes (Fig. 2b, g). This observation strongly suggests that impaired pericyte migration in MCT8MO embryos is not due to direct effects on these cells but due to delayed BA development. Pericytes in CtAs were only observed at 48 hpf in CTRMO embryos, while in MCT8MO embryos, no pericytes were observed (Fig. 2b, h). Since there is no increase in pericyte number after adding T3 to the medium, we suggest that MT3 does not act directly on pericytes, but instead, the delayed/impaired vascular development seems to affect the number of pericytes being recruited to these vascular structures.

Fig. 2: MT3 seems to affect indirectly pericyte number due to impaired development of the hindbrain vascular structures.
figure 2

Lateral view of Tg(fli1:EGFP) zebrafish embryos after WISH for pdgfrb (pericyte marker, magenta) and immunostained against GFP (endothelial marker, white) at 32, 36 and 48 hpf. Zebrafish embryos were submitted to two experimental conditions: a Increase of T3 (20 mM) availability in the medium (Control vs. T3 treatment) and b Knockdown of the Mct8 transporter by a morpholino-based system (CTRMO vs. MCT8MO). The red arrowhead indicates the mid-cerebral vein (MCeV), the green arrowhead indicates the primordial hindbrain channels (PHBC), the blue arrowhead indicates the basilar artery (BA), and the yellow arrowhead indicates pericytes. Yellow numbers 1-7 indicate the central arteries (CtAs) in their respective rhombomere. Scale bars: 100 µm. Quantification of the pericyte numbers in Control and T3 treatment condition in the c PHBC, d BA, and e CtAs at 32 hpf (n = 13, 11 (C, T3)), 36 hpf (n = 17, 14 (C, T3)) and 48 hpf (n = 20, 18 (C, T3)). Quantification of the pericyte numbers in CTRMO and MCT8MO condition in the f) PHBC, g) BA, and h) CtAs at 32 hpf (n = 14, 12 (CTRMO, MCT8MO)), 36 hpf (n = 15, 12 (CTRMO, MCT8MO)) and 48 hpf (n = 16, 11 (CTRMO, MCT8MO)). Data are presented as box-and-whisker plot, where the black thick horizontal line represents the median. The first and third quartiles are marked by the lower and upper edges of the boxes, respectively. Error bars represent standard deviations (smallest and highest value). Statistical significance determined by t-test p < 0.05. For detailed statistics, see Supplementary Data 2.

pax8 and cpne4 hindbrain neurons are under MT3 regulation

The previous results have shown that the angiogenic factor vegfaa is under MT3 regulation and thus responsible for CtA development. However, the question is, which neural cells depend on MT3 signaling and are thus responsible for vegfaa expression? Previous zebrafish spinal cord vascularization studies have shown that neurons coordinate intersegmental vessel development by expressing neuron-derived vegfaa and soluble fms-like tyrosine kinase 1 (sflt1)39. One hindbrain population previously shown to be under MTH influence are pax8 inhibitory interneurons12. We explored the spatial relationship between pax8 expression and CtA development. In CTRMO embryos, pax8 expression at 32 and 36 hpf was located in the posterior region of the hindbrain and the midbrain-hindbrain boundary (MHB) (Fig. 3a). At 48 hpf, pax8 was expressed in the hindbrain juxtaposed to CtAs 4 to 7 (Fig. 3a, b). In MCT8MO embryos, some pax8 expression was detected at 32 and 36 hpf in the MHB, but in the hindbrain, only at 48 hpf pax8 expression was present but in a more reduced expression field (Fig. 3a, c). Most of the remaining pax8-expressing cells were located more ventrally in MCT8MO embryos than in CTRMO embryos, suggesting a spatial redistribution of these cells and likely a different identity and function (Fig. 3b, c). This observation suggests a relation between the lack of or reduced pax8 expression in the posterior region of the hindbrain in MCT8MO zebrafish embryos and the lack of sprouting CtAs from rhombomeres 4 to 7. To understand if pax8-neurons are involved in CtA development, we used a homozygous transgenic Tg(pax8:DsRed) zebrafish line that presents pax8 loss of function24. WISH analysis for the vascular endothelial cadherin (VE-cadherin or cadherin-5) between heterozygous pax8+/- (control) and hypomorph pax8-/- embryos, which is specifically expressed in endothelial cells, shows that pax8-/- hypomorph mutants develop a normal hindbrain vasculature (Fig. 3d). Comparative analysis of the length of the PHBC (Fig. 3f) and CtAs (Fig. 3e) between control pax8+/- and hypomorph pax8-/- mutant embryos showed no significant differences between them, except for CtA 7, usually the last to develop37, which can lead to its length variance at 48 hpf. These results demonstrate that pax8-neurons are not involved in CtA ingression.

Fig. 3: pax8 expression is downregulated in MCT8MO zebrafish embryos but is not involved in hindbrain vascular development.
figure 3

a Lateral view of maximum projections of the hindbrain vasculature structures at 32, 36 and 48 hpf in Tg(fli1:EGFP) in CTRMO and MCT8MO zebrafish embryos after WISH for pax8 (magenta) and immunostaining for GFP (endothelial marker, white). In CTRMO zebrafish embryos, pax8 expression appears in the posterior hindbrain region, juxtaposed to CtAs 4 to 7. In MCT8MO zebrafish embryos, pax8 expression was clearly reduced and only appears at 48 hpf (n = 9 – 18). Magnification of the selected area at 48 hpf for b CTRMO and c MCT8MO embryos are shown. d Lateral view of the hindbrain of Tg(pax8:dsRed)+/- (control group (pax8+/-)) and Tg(pax8:dsRed)-/- (hypomorph group (pax8-/-)) zebrafish embryos after WISH for cadherin-5 at 48 hpf. All 7 CtAs were present in control pax8+/- and hypomorph pax8-/- zebrafish embryos. Image J software was used to measure the length of e each CtA and f the PHBC between control pax8+/- (n = 18) and hypomorph pax8-/- (n = 14) zebrafish embryos. No changes were observed; Unpaired t-test (Mann-Whitney test). Data are presented as box-and-whisker plot, where the black thick horizontal line represents the median. The first and third quartiles are marked by the lower and upper edges of the boxes, respectively. Error bars represent standard deviations (smallest and highest value). For detailed statistics, see Supplementary Data 2. The red arrowhead indicates the mid-cerebral vein (MCeV), and the green arrowhead indicates the primordial hindbrain channels (PHBC). Yellow numbers 1 – 7 indicate the developed CtA in its respective rhombomere. Scale bar 50 μm for figures (a) and (d) and 20 μm for (b) and (c).

Since pax8 gene function is lost in hypomorph pax8-/- mutant embryos and they developed a normal BHB, this raises the question of whether other glycinergic (glyt2a and glyt2b positive) and GABAergic (gad67a and gad67b positive) interneurons (from now on called “inhibitory”) in the hindbrain can be behind CtA ingression. In control, pax8+/- embryos, colocalization between DsRed expressing cells and inhibitory cells can be observed, while in hypomorph pax8-/- mutant zebrafish embryos, most of this colocalization was lost (Supplementary Fig 4a). That indicates that the remaining DsRed cells in hypomorph pax8-/- mutants are likely not neurons. We used the confined displacement algorithm (CDA) for colocalization significance analysis to compute the Pearson correlation coefficient between control pax8+/- and hypomorph pax8-/- mutant zebrafish embryos. The distribution of correlation coefficients in the control pax8+/- group follows a weak positive correlation [0.0107– 0.1280], while the hypomorph pax8-/- mutant group follows a weak negative correlation [0.0432–0.1503] (Supplementary Fig 4b). This result shows that the inhibitory cell fate in hypomorph pax8-/- mutant embryos has diverged from that of the control pax8+/- zebrafish embryos. These results indicate that, although the inhibitory cell fate may have changed, it does not seem to affect CtA development, indicating that these cells are not responsible for CtA ingression into the hindbrain.

Given that pax8-neurons are not involved in CtAs ingression, we argued that another neuron population might be responsible for instructing BHB development. We identified the Tg(gSA2AzGFF306A) line, which presents a discrete population of hindbrain GFP(+) cells located dorsally juxtaposed and perfectly aligned along the PHBC. Genomic data from these fish indicated that the enhancer trap construct containing GFP locates in the cpne4 locus (https://shigen.nig.ac.jp/zebrafish/strain/strainTopJa.jsp)25. In mice, Cpne4 is an embryonically expressed gene labeling specifically the intrinsic digit-innervating excitatory glutamatergic motor neurons40,41,42,43. In mice, blood vessel development in the spinal cord is promoted by motor neurons by a combination of attractive and repelling signaling cues44,45. So, we argued if these Cpne4 cells could constitute the neuronal cell population responsible for CtAs ingression. We analyzed the hindbrain of double transgenic zebrafish embryos (Tg(gSA2AzGFF306A;Tg(kdrl:CaaX-mCherry)) where Cpne4/GFP(+) cells were located dorsally to the PHBCs and juxtaposed to the developing CtAs in CTRMO embryos (Fig. 4a). It was visible that the number of Cpne4/GFP(+) cells increases over development in CTRMO embryos (Fig. 4a–d). In MCT8MO embryos at 36 hpf, the number of Cpne4/GFP(+) cells was significantly reduced in rhombomeres 5 and 7, compared to CTRMO embryos (Fig. 4c). That was also evident at 48 hpf, where the number of Cpne4/GFP(+) cells was significantly reduced in rhombomeres 2, 4, 5, 6 and 7 (Fig. 4d). Given the close location of these cells with the CtAs and the fact that these cells were significantly reduced in MCT8MO embryos, the correlation between the presence of Cpne4/GFP(+) cells and the presence of CtAs in CTRMO and MCT8MO at 48 hpf was analyzed. The data showed a positive correlation between Cpne4/GFP(+) cells and the development of CtAs 4, 6 and 7 in the hindbrain (Fig. 4e). We analyzed vegfaa expression in Tg(gSA2AzGFF306A) control embryos. However, no colocalization was found between Cpne4/GFP(+) cells and vegfaa (Fig. 4f), arguing that these cells do not instruct CtA ingression into the hindbrain.

Fig. 4: MT3 regulates Cpne4+ cells during hindbrain vasculature development.
figure 4

a Maximum projection of the hindbrain vasculature structures in double transgenic zebrafish embryos (Tg(kdrl:CaaX-mCherry);Tg(gSA2AzGFF306A)) immunostained against mCherry (green) and GFP (Cpne4+ cells, magenta) are shown. In CTRMO zebrafish embryos, Cpne4/GFP(+) cells were in close contact with the central arteries (CtAs), while in MCT8MO zebrafish embryos some Cpne4/GFP(+) cells were lost. The white arrowhead indicates the mid-cerebral vein (MCeV), the yellow arrowhead indicates the primordial hindbrain channels (PHBC), and the blue arrowhead indicates the basilar artery (BA). White numbers 1 – 7 indicate the developed CtA in its respective rhombomere. The number of Cpne4/GFP(+) cells between CTRMO and MCT8MO zebrafish embryos was analyzed at b) 32 hpf (n = 9, 8 (CTRMO, MCT8MO)), c) 36 hpf (n = 5, 6 (CTRMO, MCT8MO)) and d) 48 hpf (n = 7, 10 (CTRMO, MCT8MO)); Unpaired t-test; Error bars represent standard deviation. e Graphical view of the correlation between the presence of Cpne4/GFP(+) cells and the presence of each CtA in CTRMO and MCT8MO at 48 hpf shows that a correlation between these two conditions exists for CtAs 4, 6 and 7. Fisher’s exact test. n = 7 (CTRMO), 11 (MCT8MO). For detailed statistics, see Supplementary Data 2. f Dorsal view of the hindbrain of Tg(gSA2AzGFF306A) zebrafish embryos at 32, 36 and 48 hpf after WISH for vegfaa (white) and immunostained against GFP (Cpne4 cells, green) are shown. Heatmap colocalization analysis was used using normalized mean deviation product (nMDP) values and a combined image between GPF and vegfaa expression is shown. Color bar chart indicating no colocalization (-1, blue color) to colocalization (1, red color). Scale bar: 50 μm.

Together, this evidence argues that neurons were likely not the neuronal cells instructing the ingression of CtAs into the hindbrain.

pax6a NPCs are under MT3 regulation and express vegfaa

Previously, we found that a discreet hindbrain population of ventral pax6a NPCs depend on MT330, arguing that these cells could be responsible for CtA development. We analyzed the expression pattern of pax6a during hindbrain vascular development in CTRMO and MCT8MO zebrafish embryos (Fig. 5a). In CTRMO zebrafish embryos, pax6a+ cells were found along the hindbrain, and the ventral pax6a+ cells were near the developing CtAs (Fig. 5a). In MCT8MO embryos, pax6a+ cells were less abundant, and at 36 and 48 hpf the ventral pax6a+ cells were lost (Fig. 5a).

Fig. 5: Ventral pax6a+ cells express vegfaa in the zebrafish hindbrain.
figure 5

aLateral view of Tg(fli1:EGFP) CTRMO and MCT8MO zebrafish embryos after WISH for pax6a (magenta) and immunostained against GFP (endothelial marker, cyan) at 32, 36 and 48 hpf are shown. The ventral population of pax6a expressing cells were lost in MCT8MO zebrafish embryos, compared to CTRMO embryos. bLateral view of Tg(fli1:EGFP) fluorescent maximum projection images of double WISH for pax6a (magenta) and vegfaa (yellow) and immunostained against GFP (cyan) in CTRMO and MCT8MO zebrafish embryos at 30, 32, 36, 42 and 48 hpf are represented. The hindbrain of CTRMO and MCT8MO zebrafish embryos were analyzed for colocalization of pax6a and vegfaa co-expressing cells (white dotted circles) during BHB development at different time points. Colocalization was determined by using the colormap colocalization plugin of Fiji software in the region of every CtA. The white arrowhead represents the mid-cerebral vein (MCeV), and the red arrowhead represents the primordial hindbrain channels (PHBC). Numbers 1 – 7 indicate the CtA in its respective rhombomere. Scale bar: 50 μm. During the different time points of hindbrain vasculature development (c-CtA1, d-CtA2, e-CtA3, f-CtA4, g-CtA5, h-CtA6, i-CtA7), the presence and absence of CtAs and pax6a/vegfaa co-expressing cells were analyzed and the frequency determined. CtAs 2, 4, 5 and 7 correlate with CtA development and vegfaa/pax6a co-expressing cells. Fisher’s exact test. n = 8 (30 hpf CTRMO, 32 hpf MCT8MO), 9 (30 hpf MCT8MO, 36 hpf MCT8MO), 10 (all other stages and conditions). For detailed statistics, see Supplementary Data 2.

Given these findings, we analyzed if these pax6a+ hindbrain cells express vegfaa and are thus responsible for CtA sprouting into the zebrafish hindbrain. Colocalization analysis, using the image J plugin Colormap36 showed that some hindbrain pax6a+ cells express vegfaa (Fig. 5b). This was evident in the time course analyzed, 30, 32, 36, 42 and 48 hpf. Analyzing the surrounding area of every CtA in every stage shows that colocalizing pax6a/vegfaa cells occurs near or dorsal to the developing CtAs (Supplementary Fig 5, individual images of the CtAs). Notable was the observation that the developing CtAs seem to migrate towards these pax6a/vegfaa colocalizing cells (individual images, Supplementary Fig 5c, e, g and i), indicating that these cells might be responsible for the chemoattraction of the developing CtAs. We also analyzed the relationship between the developing CtAs and the co-expressing pax6a/vegfaa cells between CTRMO and MCT8MO zebrafish embryos (Fig. 5c–i). The statistical analysis revealed no relationship between CtA 1 development and pax6a/vegfaa co-expressing cells in the analyzed time points (Fig. 5c). In rhombomere 2, pax6a/vegfaa co-expressing cells were present from 30 hpf onwards in CTRMO embryos, while these cells were absent in MCT8MO; at 48 hpf only 30% of the embryos developed CtA 2. The statistical analysis argues that the existence of pax6a/vegfaa co-expressing cells might favor the development of CtA 2 (Fig. 5d). CtA 3 developed in most CTRMO and MCT8MO zebrafish embryos. However, this CtA developed later in MCT8MO embryos, at 36 hpf. This explains the statistical significance observed at 32 hpf (Fig. 5e). This strengthens the fact that pax6a/vegfaa co-expressing cells in this rhombomere lead to the successful development of CtA 3 in MCT8MO embryos. CtA 4 is one of the first CtAs to develop37, but in MCT8MO zebrafish embryos, only at 42 hpf 10% of the embryos analyzed had developed this CtA. In MCT8MO zebrafish embryos in which CtA 4 developed, pax6a/vegfaa co-expressing cells were always present. This evidence shows a dependency of CtA 4 development on pax6a/vegfaa co-expressing cells (Fisher’s exact test p < 0.05, Fig. 5f). In rhombomere 5, pax6a/vegfaa co-expressing cells were present in CTRMO from 30 hpf and in MCT8MO embryos but in lesser quantity. Between 32 and 42 hpf, the presence of CtAs and pax6a/vegfaa co-expressing cells between CTRMO and MCT8MO embryos differed significantly, showing that pax6a/vegfaa co-expressing cells are needed for CtA development (Fig. 5g). In rhombomere 6, pax6a/vegfaa co-expressing cells were present in CTRMO and MCT8MO zebrafish embryos from 30 hpf. There is a dependency between CtA 6 development and pax6a/vegfaa co-expressing cells at 36 and 42 hpf (Fisher’s exact test, p < 0.05). In rhombomere 7 of CTRMO zebrafish embryos, pax6a/vegfaa co-expressing cells were present early during hindbrain vascular development, while in MCT8MO embryos, pax6a/vegfaa co-expressing cells were only present from 42 hpf. At 42 hpf, the statistical analysis revealed a significant positive relationship between pax6a/vegfaa co-expressing cells and CtA 7 development (Fig. 5i). In summary, pax6a/vegfaa co-expressing cells are likely required to develop CtAs 2, 4, 5 and 7.

pax6a/vegfaa co-expressing cells are essential for CtA migration

We investigated whether the presence of pax6a/vegfaa co-expressing cells influences the direction of the growing CtA in the hindbrain. For that, the angle of the developing CtA was measured using the PHBC as origin (0 degrees). The angle was measured until the directional change of the CtA towards the BA or its ipsilateral CtA neighbor. In CTRMO zebrafish embryos, CtA 1 developed later in development (42 hpf; Fig. 6a), and this CtA was not attracted by the pax6a/vegfaa co-expressing cells, mirroring the previous results (Fig. 5c). The growth angle between CTRMO and MCT8MO zebrafish embryos differed significantly because only 10% of the MCT8MO zebrafish embryos developed CtA 1 at 48 hpf (Fig. 6a). In CTRMO embryos, CtAs 2 to 5 grew towards the pax6a/vegfaa co-expressing cells (Fig. 6b-e). However, mostly at 42 and/or 48 hpf, these CtAs changed the migration direction (indicated by the grey box in Fig. 6b-e), thus not being attracted anymore by the pax6a/vegfaa co-expressing cells (Fig. 6b–e, grey boxes). In rhombomeres 6 and 7 pax6a/vegfaa co-expressing cells were absent in CTRMO at 48 hpf (Fig. 6f, g). The analysis showed that CtAs 2, 4, 5, 6 and 7 migration angles differed significantly between CTRMO and MCT8MO embryos (Fig. 6b, d–g). The only exception was for CtA 3 (Fig. 6c). In MCT8MO embryos, CtA 3 developed after 32 hpf. Although this CtA developed later than the control group, it grew towards the direction of the pax6a/vegfaa co-expressing cells until 48 hpf (Fig. 6c, h).

Fig. 6: pax6a/vegfaa co-expressing cells guide central arteries migration.
figure 6

The angle of the developing CtA 1 (a), CtA 2 (b), CtA 3 (c), CtA 4 (d), CtA 5 (e), CtA 6 (f), and CtA 7 (g) were measured using the PHBC as the basis (0 degrees). The angle was measured before the turnover of the CtA towards the BA or directional change of the CtAs towards its ipsilateral neighbors. CtAs and pax6a/vegfaa co-expressing cells that were absent are indicated as 0°. Grey boxes show the CtAs that changed the directional migration during hindbrain vasculature development. a–g Statistical significance was determined using 2-way ANOVA. n = 8 (30 hpf CTRMO, 32 hpf MCT8MO), 9 (30 hpf MCT8MO, 36 hpf MCT8MO), 10 (all other stages and conditions). Data are presented as box-and-whisker plot, where the black thick horizontal line represents the median. Each dot represents a biological replicate; error bars represent standard deviations (smallest and highest value). For detailed statistics, see Supplementary Data 2. h Graphical representation of the CtA directionality and the position of the pax6a/vegfaa co-expressing cells during BHB development. The mean values were used to construct these graphs.

Hindbrain pax6a NPCs are regulated by MT3 in a cell-autonomous way

The previous results indicated that MT3 signaling is important for developing a specific population of ventral pax6a + NPCs likely involved in CtA ingression into the hindbrain. T3 cellular signaling genes are expressed in NPC of the developing cerebral cortex of mice46, and mct8, thraa, and thrab are already expressed in the zebrafish neuroepithelium at 12 hpf12. From 30 hpf onwards, several MT3-sensitive pax6a + NPCs are present in the hindbrain of CTRMO zebrafish embryos from rhombomeres 3 to 7 but are lost in mct8 morphants pax6a + MT3 sensitive (pax6a/mct8 + ) (Fig. 7). Likewise, ventral hindbrain pax6a+ cell loss is also observed in mct8(-/-) loss of function homozygous embryos (Supplementary Fig 1c).

Fig. 7: pax6a-expressing cells colocalize with T3 transporter mct8.
figure 7

a Lateral view of maximum projection images of fluorescent double WISH against pax6a (magenta) and mct8 (yellow) and immunostaining against GFP (endothelial marker, cyan) in Tg(fli1:EGFP) CTRMO and MCT8MO zebrafish embryos at 30, 32, 36, 42 and 48 hpf. The hindbrain of CTRMO and MCT8MO zebrafish embryos were analyzed for colocalization of pax6a with mct8-expressing cells (white dotted circles) during BHB development at different time points. Colocalization was determined by using the colormap colocalization plugin of Fiji software in the region of every CtA. The white arrowhead represents the mid-cerebral vein (MCeV) and the red arrowhead represents the primordial hindbrain channels (PHBC). Numbers 1 – 7 indicate the CtA in its respective rhombomere. Scale bar: 50 μm. During the different time points of BHB development, the presence and absence of CtA 1 (b), CtA 2 (c), CtA 3 (d), CtA 4 (e), CtA 5 (f), CtA 6 (g), CtA 7 (h) and pax6a/mct8 co-expressing cells were analyzed and the correlation was determined. Statistical significance was determined using Fisher’s exact test. n = 3 (36 hpf MCT8MO), 4 (30 hpf CTRMO/MCT8MO, 32 hpf CTRMO, 36 hpf CTRMO), 5 (32 hpf MCT8MO, 42 hpf CTRMO/MCT8MO, 48 hpf CTRMO/MCT8MO)). For detailed statistics, see Supplementary Data 2.

CtA 4, 5, and 6 ingression coincides or is preceded by the development of pax6 + MT3 sensitive cell (pax6a/mct8 + ; Fig. 7) with the ability to respond to MT3 (pax6a/thraa/thrab + ; Fig. 8, Supplementary Fig 6 and 7). This evidence shows direct MT3 signaling is involved in CtA 4, 5, and 6 ingressions.

Fig. 8: pax6a-expressing cells colocalize with THalpha receptors.
figure 8

During the different time points of hindbrain vasculature development, the presence and absence of CtA 1 (a), CtA 2 (b), CtA 3 (c), CtA 4 (d), CtA 5 (e), CtA 6 (f), CtA 7 (g) and pax6a/thraa co-expressing cells were analyzed and co-localization determined. Fisher’s exact test. n = 5, except for 32 hpf CTRMO and 42 hpf MCT8MO were n = 4. During the different time points of BHB development, the presence and absence of CtA 1 (h), CtA 2 (i), CtA 3 (j), CtA 4 (k), CtA 5 (l), CtA 6 (m), CtA 7 (n) and pax6a/thrab co-expressing cells were analyzed and and co-localization determined. Fisher’s exact test. n = 3 (30 hpf CTRMO, 36 hpf MCT8MO), 4 (32 hpf CTRMO), 5 (all other stages and conditions).

CtA 1 and 2 develop independently of MT3 signaling, given that no mct8/pax6+ cells coincide with CtA ingression (Fig. 7). In rhombomeres 1 and 2, co-expressing pax6a/thrab+ cells were found before CtA ingression (Fig. 8h, I, Supplementary Fig 7). These observations shows that MT3 responsivity is not involved in CtA 1 and 2 ingressions (Fig. 7). Still, it cannot be excluded that thrab aporeceptor function is not involved in these CtA ingressions (Fig. 8h, i). In the case of CtA 7, pax6a/mct8/thraa+ cells (Figs. 7 and 8, Supplementary Fig 6) are present in rhombomere 7 before ingression, but the vessel ingresses in CTRMO embryos at a time where MT3-sensitive pax6a+ cells are not present and independently of mct8 knockdown (Fig. 7). This argues that CtA7 develops independently of MT3 signaling and thyroid receptor aporeceptor function.

CtA 3 presents the most complex situation. The pax6a/thraa/thrab+ cells (Fig. 8g, n) are present before and during CtA 3 ingression, irrespective of pax6a/mct8+ cells’ presence (Fig. 7h). This evidence does not exclude that thraa and/or thrab or both functions are not involved in CtA 3 ingression. This hypothesis will have to be answered in future studies.

However, for CtAs 4, 5, and 6, there is a temporal pattern of thraa and thrab expression in pax6a/mct8+ cells (Figs. 7e, f and g and 8e, f, g, k, l, and m). The presence of pax6a/mct8/thraa/thrab+ cells coincides with vessel ingression, indicating that MT3 signaling via thraa and thrab is likely necessary for pax6a+ cells attraction of CtAs via vegfaa signaling.

Loss of pax6a function leads to impaired hindbrain CtA development

To confirm the involvement of hindbrain pax6a NPCs in hindbrain vasculature development, we knocked out pax6a. A guide RNA (gRNA) was designed against the zebrafish pax6a exon 7 locus, thus leading to a frameshift before the homeobox and the proline/serine/threonine rich (PST) domains that prevents DNA binding47,48 (Supplementary Fig 8). Zebrafish embryos were injected with the pax6a gRNA and Cas9 protein (Weissman Institute, Israel) in 1-cell stage zebrafish embryos. The efficiency of the injection was verified after phenotyping for small eye size, characteristic of impaired pax6a function34, and confirmation by genotyping by PCR (Supplementary Fig 8) of the F0 pax6a mutant zebrafish embryos (hereafter called crispants) at 24 hpf.

We performed live imaging of pax6a crispants to evaluate CtA development in Tg(kdrl:CaaX-mCherry)23. We observed that pax6a crispants developed a normal PHBC (yellow arrowheads in Fig. 9a lower panel), mid-cerebral vein (MCeV, red arrowheads in Fig. 9a lower panel), and lateral dorsal aorta (LDA, blue arrowhead in Fig. 9a, lower panel). Like MCT8MO (Tg(fli1:EGFP) background) zebrafish embryos (Fig. 9b), pax6a crispants (Fig. 9a) displayed fewer CtAs than non-injected control embryos. In pax6a crispants, the first CtAs sprouts were present at 36 hpf. At 40 hpf, a third CtA developed, and these were the only CtAs that developed until 48 hpf (Fig. 9a, second panel). CtAs in CTRMO zebrafish embryos developed similarly to those in non-injected Tg(kdrl:CaaX-mCherry) control embryos (Fig. 9a, b, first panels). MCT8MO zebrafish embryos developed CtAs after 36 hpf, presenting two CtAs at 40 hpf. At 44 hpf, a third CtA developed, and these were the only CtAs that developed until 48 hpf (Fig. 9b second panel). This shows that both pax6a crispants and MCT8MO zebrafish embryos presented CtA defects that arise in a similar chronological order (Fig. 9a, b). We further confirmed the results obtained with crispant embryos for pax6a loss-of-function in F2 pax6a homozygous loss-of-function zebrafish larvae at 4 dpf (Fig. 9c). Further analysis of pax6a expression in mct8 knockout embryos confirms the loss of some but not all hindbrain pax6a+ cells. Together, this data confirms that pax6a + NPCs loss is behind impaired hindbrain vasculature development in both pax6a+ and mct8 loss of function models and that MT3 is involved in some but not all pax6a + NPCs development and function.

Fig. 9: pax6a mutant zebrafish embryos show similar hindbrain vascular development defects as MCT8MO zebrafish embryos.
figure 9

a Live imaging of 32 hpf zebrafish embryo at the start of imaging between non-injected (control) and pax6a crispant zebrafish embryos are represented. The vascular system (kdrl) is shown in white (mCherry) in the reporter line Tg(kdrl:CaaX-mCherry). Dorsal view of maximum projection images of the mentioned time points is shown. b Live imaging of 32 hpf zebrafish embryo at the start of imaging between CTRMO and MCT8MO zebrafish embryos are represented. The vascular system is shown in white (GFP) in the reporter line Tg(fli1:EGFP). Dorsal view of maximum projection images of the mentioned time point is shown. n = 2. Scale bar: 50 μm. c Comparison between control (not injected) and mutant pax6a CRISPR zebrafish larvae at 4 dpf are presented. n = 5 (control), 9 (pax6a CRISPR). Mutant pax6a knockout zebrafish larvae present only 4 CtAs, while the control zebrafish have 7 CtAs. The red arrowhead represents the mid-cerebral vein (MCeV), the yellow arrowhead represents the primordial hindbrain channels (PHBC), and the blue arrowhead represents the lateral dorsal aorta (LDA). White * represents sprouting projections of the PHBC to the BA (due to the inclination of the hindbrain imaging, we can visualize these structures). Numbers 1 – 7 indicate the CtA in its respective rhombomere. d In mct8(-/-) embryos at 36 hpf it is observed a loss of pax6a hindbrain cells coincident with CtAs underdevelopment. Scale bar 50 μm.

Discussion

Several studies have shown a relationship between THs and brain vascularization during development. In adult dogs, it has been demonstrated that chronic hypothyroidism leads to increased BBB permeabilization49,50. Impaired brain vascularization was also observed in post-natal induced hypothyroid rats21,51. In zebrafish, impaired hindbrain vascularization development has been shown in MCT8MO embryos12 and larvae14. Importantly, new evidence from AHDS patients shows a reduced vascularization of the cortex16, further confirming that MCT8-dependent T3-signaling is essential for human brain vascularization.

Here we show that hindbrain vasculature development in the zebrafish requires MT3 signaling. Our evidence outlines a mechanism where MT3-dependent development of a specific hindbrain pax6a + NPCs population is involved in the vegfaa-dependent chemoattraction of endothelial cells ingression into the hindbrain that enables ingression of CtAs (Fig. 10). Our work highlights the integrative role of MT3 in embryonic development, which is necessary for CNS development12,27,30 but also for integrating the vasculature in the CNS to give rise to a fully functional organism.

Fig. 10: Proposed model for MT3 Mct8-mediated hindbrain development in zebrafish embryogenesis.
figure 10

a Illustration of the zebrafish hindbrain with proposed action of MT3 signaling through Mct8 on CtA development. The different pax6a neural progenitor cells (NPC) identified in the different rhombomeres with the TH machinery mct8, thraa and thrab are illustrated. In rhombomere 1 we suggest the involvement of another central nervous system (CNS) cell type that is MT3 dependent and the source of vegfaa required for CtA development. In rhombomere 3, pax6a NPCs are present, but MT3 signaling is not involved in CtA 3 development (dashed arrow line) but enhances its development. Vascularization of rhombomeres 4, 5 and 6 suggests that, besides vegfaa, another angiogenic factor is involved in the angiogenic sprouting of these CtAs. b Illustration of a lateral view of the zebrafish hindbrain showing the effect of Mct8 knockdown and MT3-impaired signaling. The most frequently developed CtAs in the hindbrain of 48 hpf zebrafish embryos are represented. Although significantly reduced in rhombomere 2, vegfaa signaling is still present in MCT8MO zebrafish embryos, showing that another CNS cell type, independent on MT3 signaling, is responsible for releasing vegfaa or able to compensate for the lack of pax6a + /MT3-responsive cells. The inability to rescue the vascularization of rhombomere 4 by vegfaa-165 mRNA suggests that another angiogenic factor is necessary for its development. C: cerebellum; chb: caudal hindbrain; MHB: Midbrain-hindbrain boundary; MT3: maternal T3; PHBC: Primordial Hindbrain Channels; r1 – r8: rhombomere 1 – 8; sc – spinal cord. The endothelial cell model was adapted from https://bioart.niaid.nih.gov/discover?q=endothelial.

Although we did not analyze CtA development after 48 hpf, this phenotype was not a consequence of the late development of the MCT8MO embryos, given that other developmental landmarks, such as heart development, co-occurred with CTRMO zebrafish embryos. Impaired brain vascularization was also observed in MCT8MO zebrafish larvae at 120 hpf14, confirming that the vascular phenotype persists throughout development. Moreover, our results with the newly generated mct8 knockout also support the persistent consequences found in the mct8 knockdown model.

MT3 regulates hindbrain vegfaa expression for CtA development

Vertebrate brain angiogenesis strongly depends on VEGF and their endothelial tyrosine kinase receptors (VEGFR)38,39,52,53,54. In zebrafish two vegfa paralogues exist, vegfaa and vegfab38. First, we analyzed the expression of vegfaa and vegfab by qPCR; however, no significant changes were observed during the BHB developmental time window (28 to 72 hpf) (Supplementary Fig 2c, d). However, WISH expression analysis showed that both genes were reduced in the brain in MCT8MO embryos (Supplementary Fig 3a, b). Nevertheless, in MCT8MO zebrafish embryos, vegfab was only mildly affected. At the same time, vegfaa expression was significantly reduced in the hindbrain, indicating that MT3 regulates both vegfa paralogues but was affecting more the expression of vegfaa. Similarly to our results, immunohistochemistry analysis of the ventricular zone of prenatal hypothyroid mice showed significant differences in cerebral microvasculature, COL IV, PECAM-1, and F-actin expression between control and PTU-treated mice51. These studies found no differences when bulk transcriptomic analyses were carried out, indicating that T3 action on brain vascularization occurs discretely cellular and context-dependently.

We also showed that vegfaa-165 mRNA partially rescued the MCT8MO hindbrain vasculature phenotype (Fig. 1e, f). This revealed that MT3 through Mct8 regulates CtA development through vegfaa expression. However, the incapacity of vegfaa-165 mRNA injection to ultimately rescue all CtAs, argues that MT3 might regulate another factor involved in CtA development. Rossi et al. 38 showed that although vegfaa is the necessary angiogenic factor in CtA development, vegfab knockout embryos also lose some CtAs. The expression of vegfab is also affected in mct8 knockdown embryos, although to a much lesser extent (Supplementary Fig 3b). Another possible factor responsible for the partial rescue of the hindbrain CtAs is the extracellular matrix (ECM) composition. In the 25 hpf MCT8MO zebrafish transcriptome30, ECM constitutes one of the most affected genetic pathways, arguing for overall differences in ECM composition. Transcriptomic analysis of primary cerebrocortical cells of mice55 and from microdissected brain tissues of the ventricular zone of hypothyroid rats51 also identified ECM genes regulated directly by TH. Moreover, the Vegfaa-165 isoform binds to the ECM56. Modifying the ECM components might affect the binding capacity of Vegfaa-165 to the ECM57, leading to a lower concentration gradient or effectivity of vegfaa signaling to initiate the angiogenic sprouting of the endothelial cells from the venous PHBC. These results further highlight that the hindbrain vasculature defects observed in MCT8MO zebrafish embryos result from an indirect regulation of MT3 on the endothelial cells. Still, it regulates MT3-responsive hindbrain cells that express vegfaa, and likely the extracellular environment that enables adequate Vegfa signaling.

A further example of how indirect regulation of MT3 of the hindbrain vascularization occurs is the recruitment of the pericytes to the different hindbrain vascular structures. Pericytes are essential for blood vessel maturation and BBB stabilization as they express tight junction proteins to promote vascular permeability58, allowing for the establishment of a competent neurovascular unit. In MCT8MO zebrafish embryos, pericyte numbers were significantly reduced in the PHBC, BA, and CtAs at 48 hpf, and treatment with T3 could not accelerate pericyte migration. These observations are further evidence of an indirect role of MT3 in vascular structures development. MT3 did not directly regulate pericyte recruitment by inducing pdgfrb expression, but rather, it affected pericyte recruitment due to impaired hindbrain vasculature development used by these cells to achieve their final position.

Hindbrain pax6a + NPCs are responsible for the chemoattraction of the CtAs in a cell-autonomous way

Maternal THs deficiency leads to an underdeveloped CNS. Several reports have highlighted the importance of maternal THs for NPC expansion, neurogenesis, and the differentiation of neuronal and glial cells, among others1,46,59. Supporting the hypothesis that impaired MT3 signaling results in impaired neurodevelopment is the source of chemoattracting CtAs via vegfaa; we identify the vegfaa signaling source by examining MT3-dependent hindbrain CNS cells. We verified that pax8 (inhibitory neuron marker) mutant zebrafish embryos presented a complete BHB (Fig. 3d) and that copine 4 (cpne4) interdigit innervating neurons do not express vegfaa (Fig. 4f), thus confirming they are not responsible for CtA development. Nonetheless, we identified these neurons (pax8 and cpne4) as MT3-responsive, indicating that their progenitor cells likely depend on MT3 for differentiation.

In mammals, after perineural vascular plexus (PNVP) formation, the sprouting of the blood vessels into the neural tissue is promoted by NPCs that secret Vegf-A60,61,62. In fetal hypothyroid rats, Pax6 NPCs were reduced46, like in the hindbrain of MCT8MO zebrafish embryos30, leading us to hypothesize that these cells are the source of vegfaa and responsible for CtA development in the zebrafish hindbrain. Moreover, in mice, Pax6 NPCs express thyroid hormone receptor alpha-1 (THRα1)46, consistent with being under the regulation of MT3. We found a reduced expression of pax6a in the hindbrain of MCT8MO zebrafish embryos and also that some of these cells colocalize with vegfaa. Our findings show that in contrast to the spinal cord in mice44 and zebrafish39, where neurons contribute to vascularization, in the zebrafish hindbrain, this role lies with the NPCs. Our evidence corroborates previous studies showing63 that zebrafish brain and trunk tip cells express different types of matrix metalloproteinase, leading to different specifications of blood vessel characteristics and developmental programs regulating vascularization.

We identified that CtA migratory routes correlate with pax6a/vegfaa co-expressing cells in zebrafish embryos. In MCT8MO embryos, altered routes occurred due to reduced pax6a/vegfaa cells, except CtA 3, which maintained its BA connection (Fig. 6). CtA 3’s unique persistence in MCT8MO embryos correlated with residual vegfaa in rhombomere 3 (Fig. 1a–d) and its essential BA interconnection for hindbrain circulatory closure. CRISPR-generated pax6a mutants confirmed CtA 3 development is independent of pax6a NPCs (Fig. 9a).

CtA 1 (rhombomere 1) was rescued by vegfaa-165 mRNA (Fig. 1g) but showed no correlation with pax6a/vegfaa cells (Fig. 5c), suggesting another MT3-regulated vegfaa-secreting cell type guides its migration into the hindbrain. CtA 2 development ( ~ 50% in MCT8MO) relied on residual vegfaa signaling (Fig. 1b-d) and correlated with pax6a/vegfaa cells (Fig. 5d), indicating MT3 regulation.

Rhombomere 4 vascularization failed vegfaa rescue (Fig. 1g) but correlated with pax6a/vegfaa cells (Fig. 5f) and pax6a/mct8 co-expression (Fig. 7e), confirming Mct8-mediated MT3 control of CtA 4. Rhombomeres 5 and 6 showed partial vegfaa rescue (Fig. 1g) and correlation with pax6a/vegfaa cells (Fig. 5g, h) and pax6a/mct8 for CtA 5 (Fig. 7f), suggesting additional angiogenic factors. CtA 7 rescue via vegfaa (Fig. 1g) and pax6a/vegfaa correlation confirmed MT3 regulation via pax6a NPCs.

pax6a loss-of-function experiments validated mct8 knockdown/knockout results, demonstrating pax6a + NPCs direct hindbrain vessel ingression (Fig. 9). These findings reveal cell/tissue-specific roles for MT3 signalling.

Interestingly, mct8, thraa, and thrab were expressed in a spatiotemporal manner, indicating that these cells are responding autonomously to MT3. For CtAs 4, 5, and 6, vessel ingression occurred only when pax6a + NPC co-expressed the three T3-signaling genes. This evidence supports that a change from thyroid receptor apo- to receptor function is involved in coordinating the sprouting and ingression of the CtAs for each rhombomere in a timely coordinated manner. Nonetheless, this new evidence needs to be functionally validated in future studies. This shows that, as has been described37, while CtAs sprouting is variable, it is not random.

Our analysis shows that an MT3-dependent function on pax6a + NPCs is required for CtA 4, 5, and 6 ingressions. For the remaining CtA, the situation is more complex. Except for CtA 1 and 7, independent of MT3 action, there is a correlation between vessel ingression and pax6a + NPCs co-expressing thrab. Nonetheless, thrab is considered to have a dominant negative function64, thus making a definitive conclusion about the role of MT3 in these NPCs challenging to determine from our data. These need additional studies in the future. This study indicates that not all MT3-dependent NPCs can mediate brain vascularization after MT3-signaling. Notably, in our dataset, only mct8/thraa/thrab+ pax6a + NPCs seem to be involved in hindbrain vascularization, suggesting that a particular thyroid receptor-mediated MT3-action is crucial for giving pax6a + NPCs the functional capacity to attract blood vessels.

Relevance to AHDS

T3-signaling genes are not expressed in the developing zebrafish vasculature until 48 hpf12, but impaired MT3-signaling in the CNS affects hindbrain vasculature development and angiogenic-related developmental pathways. Similarly, human endothelial cells derived from AHDS patient iPSC do not present altered differentiation15, and recent evidence from AHDS patient-derived organoids shows that the neurological effects of impaired MCT8 function are the result of MT3 action on neural cells and not impaired BBB function13. However, and at least in humans, it cannot be discarded that T4 might have a role in AHDS. Our work demonstrates that MT3 regulates a particular population of pax6a + NPCs, defined by co-expression of mct8/thraa/thrab, that results in vegfaa expression and chemoattraction and ingression of endothelial cells from the PHBC, allowing for the sprouting of CtAs 4, 5 and 6 (Fig. 10). We did not analyze whether MT3 directly affects endothelial cell proliferation during this study. However, this seems unlikely considering previous results showing that no TH-signaling genes are expressed in these cells at the zebrafish developmental stages12. Nonetheless, our results indicate that there are indirect effects. Specifically, our findings show a decrease in vegfaa-mediated attraction in the endothelial sprouting and migration mechanism.

The present study, together with evidence from AHDS patients16, argues that the pathophysiological consequences of MT3-impaired signaling are not only related to decreased neural cell diversity13,27,30, but also to the consequences on CNS vascularization. Given the irreversible neurological damage found in AHDS patients that cannot be recovered after birth1,65 it is very likely that the consequences on brain vascularization also cannot be recovered in these patients. Moreover, given the widespread prevalence of thyroid deficiencies that constitute the most common cause of preventable impaired cognitive development66, it cannot be excluded that some of the impacts in affected individuals also comprise impaired brain vascular development.

The fact that the BHB is disrupted in zebrafish argues that brain permeabilization might be compromised, as has been observed in hypothyroid dogs49,50, leading to the access of several factors/hormones/wastes to the brain. Nonetheless, the human AHDS brain maintains a hypothyroid state while the peripheral tissues are hyperthyroid67, suggesting impaired BBB development maintains some barrier function. This seems to be the case in zebrafish as well. We never observed any brain edema in mct8 morphant zebrafish embryos12. Although pericyte recruitment is delayed into the BHB, it still occurs, further arguing that in zebrafish, the BHB permeability is maintained, at least at the end of embryogenesis.

Conclusion

The presence of MT3 allows the survival and proliferation of specific pax6a NPCs in the zebrafish hindbrain, which are necessary for the timely ingression of the CtAs through the expression of the angiogenic factor vegfaa. The presence of the different TH signaling genes (mct8, thraa and thrab) in pax6a NPCs throughout development and specific rhombomeres indicates that MT3 acts upstream of CtA ingression, contributing to a fully functional hindbrain vasculature development (Fig. 10).

Limitations of this study

Although we have uncovered the developmental mechanism by which MT3 controls the development of the hindbrain vasculature and identified the cells under the hormone’s control, we cannot determine the exact identity of the pax6a + NPC that responds to and mediates MT3’s action on CtA ingression. Moreover, we cannot determine the genetic and developmental causes that lead to the loss of the specific pax6a + NPC under the regulation (or lack) of MT3 signaling. This is the subject of ongoing research.