Zebrafish

Zebrafish were maintained and zebrafish experiments were performed according to standard protocols [21] and in conformity with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited facility. Fish were housed in a large recirculating aquaculture facility with 1.8 L and 6 L tanks. Water quality was routinely measured and proper parameters taken to maintain water quality stability. Fry were fed rotifers and adults were fed Gemma Micro 300 (Skretting). The following transgenic fish lines were used in this study: EK (wild-type), Tg(kdrl:gfp)la116 [22], Tg(kdrl:egfp-2a-rpl10a3xHA)y530 (this paper), Tg(uas:egfp-2a-rpl10a2xHA)y531 (this paper), Tg(xa210:gal4)y241 [23], Tg(fli1a:gal4ff)ubs4 [24], and Tg(huc:gal4) [25].

Sucrose gradient

Sucrose density gradients were prepared for sedimentation analysis of polysome profiles. 12 mL 5–50% sucrose gradients were prepared in 110 mM KOAc, 2M MgOAc, 10 mM HEPES pH 7.6 with BioComp Gradient Master (BioComp), and allowed to rest overnight at 4 °C. Dechorionated and deyolked embryos were dounce homogenized in 1.5 µl polysome fractionation buffer per embryo (10mM HEPES pH7.4, 110mM KOAc, 2mM MgOAc, 100mM KCl, 10mM MgCl2, 0.1% Nonidet P-40, 2mM DTT, 40U/mL RNasin, 200ug/mL cycloheximide, and protease inhibitors; adapted from [26]). Homogenates were centrifuged at 1000xg for 10 min at 4 °C. Protein concentration was quantified by Bradford assay (Sigma Aldrich) and equivalent amounts were loaded onto gradients. Samples were centrifuged in a SW41 Beckman rotor at 40,000 rpm at 4 °C for 2 h. 16 × 1 mL fractions were collected with an ISCO piercing apparatus connected to a BioLogic chromatography system at 0.5 mL/min, pushing 55% sucrose. Data were collected using LP DataView software (BioRad). For EDTA treatment, samples were treated with 200 mM final concentration EDTA prior to loading onto gradient.

TRAP protocol

Translating Ribosome Affinity Purification (TRAP) was performed as described previously [12] with modifications. For each larval TRAP sample, approximately 1200-1500 24 hpf zebrafish embryos were dechorionated, deyolked, and dounce homogenized in homogenization buffer consisting of 50 mM Tris pH 7.4, 100 mM KCl, 12 mM MgCl2, 1% NP-40, 1 mM DTT (Sigma, Cat. #646563), 1 × Protease inhibitors (Sigma, Cat. #P8340), 200 units/mL RNAsin (Promega, Cat. #N2115), 100ug/mL Cycloheximide (Sigma Cat. #7698), 1mg/mL Heparin (Sigma, Cat. #H3393-10KU). Lysates were cleared for 10 min at 10,000×g, 4 °C. 1 µl of anti-HA antibody (Abcam ab9110 Rabbit polyclonal, RRID: AB_307019) was added per 400 µg protein in an 800 µl lysate and the samples were orbitally rotated at 4 °C for 5 h. 60 µl of Dynabeads® protein G slurry (Novex/Life Technologies) was added per 1 µl of anti-HA antibody with homogenization. Before using, Dynabeads were washed with 800 µl homogenization buffer, rotating for 30 min to equilibrate the beads. The wash buffer was then removed from the Dynabeads and the lysate + Ab solution was added to the equilibrated Dynabeads. This Dynabead + lysate + Ab mixture was incubated overnight at 4 °C on an orbital rotator. Dynabeads were collected using a magnetic stand and washed 3 × 5 min on an orbital shaker with high salt homogenization buffer (50 mM Tris pH 7.4, 300 mM KCl, 12 mM MgCl2, 1% NP-40, 1 mM DTT, 1 × Protease inhibitors, 200 units/mL RNAsin, 100 µg/mL cyclohexamide, 1mg/mL heparin). RNA was collected from the Dynabeads and DNase treated using the Zymo ZR-Duet Kit (Zymo Research D7003). Note that careful attention to the protocol is required to keep lysates cold and free of RNases as RNA integrity must be carefully monitored. Large numbers of embryos are needed to provide the starting material for the TRAP protocol because (i) the endothelial or other cell types expressing the RiboTag represent a small fraction of the total cells in the animal, (ii) only a fraction of the ribosomes in the targeted cell type contain the RiboTag, with most ribosomes still being “untagged”; (iii) TRAP-purified RNA samples contain a large amount of co-purifying ribosomal RNAs in addition to the mRNAs, and (iv) we prepared largely unamplified libraries for sequencing.

A similar TRAP protocol was utilized for the adult whole fish and organ samples as described above with a few additional modifications. SUPERase-In (100 U/µl) was added to the homogenization buffer as an additional RNase inhibitor. Organs were either dissected fresh or flash frozen for later use. Whole fish were flash frozen and ground into fish powder before homogenization. Four organs/whole fish were used per replicate and the amount of each tissue was weighed so that a ratio of only 0.75–3% tissue to homogenization buffer (weight/volume) was utilized. Tissue was homogenized using a Cole-Parmer LabGen 850 homogenizer at 13,000 rpm for 45 s. 5 µl of anti-HA antibody was used per sample. Antibody and Dynabeads incubation were the same as the larval samples, but RNA was collected from the Dynabeads using RNeasy Micro Plus Kit (Qiagen #74034) with RNA-only β-mercaptoethanol supplementation in the RLT buffer.

Quantitative RT-PCR

RNA was reverse transcribed using a high-capacity cDNA reverse transcription kit (Applied Biosystems™). cDNA was combined with the primers listed below and with LightCycler® 480 SYBR Green I (Roche). Reactions were run in a LightCycler® 480 Multiwell Plate 384 machine (Roche). Relative fold changes were calculated using ddCT calculations and by calculating standard deviations. The following primer pairs were used:

Gene

Forward Primer

Reverse Primer

cdh5

CAACAGACGCTGATGATTCC

GTCTTTGGCTTGAACAGCAA

kdrl

GAGTTCCAGCACCCTTTATCA

ATCGTCCTTCTTCACCCTTTC

desmin a

CGGTGGTTATCAGGACACTATC

TCCAGAGCCATCTTCACATTC

neurogenin 1

CGCATTGGATGCTTTGAGAAG

CGAAAGTGCCCAGATGTAGT

snap25

GGCTACTGTCATGCTCCTTATT

TGATTGTAAGTGCTCGTCGTATTA

actin b1

CGAGCAGGAGATGGGAACC

CAACGGAAACGCTCATTGC

RNAseq

RNA concentration and quality were measured on a Qubit (Thermo Fisher) and on an Agilent 2100 BioAnalyzer + RNA Pico Chip (Agilent), respectively. For our 24 hpf TRAP samples, a large proportion of the RNA samples we collected represented ribosomal RNAs co-purified with our polysome mRNAs, so 500 ng RNA was polyA-selected or rRNA-depleted using an Ribo-Zero rRNA Removal Kit (Illumina) before library construction. Sequencing libraries were constructed from the purified mRNA using TruSeq Stranded mRNA Library Prep Kits (Illumina). Libraries were sequenced using an Illumina HiSeq 2500 to generate approximately 50 million 2 × 75 bp or 2 × 100 bp pair end reads. Raw data were de-multiplexed and analyzed. RNAseq data were deposited into NCBI’s Gene Expression Omnibus (GEO) repository and assigned accession number GSE292080.

For adult whole fish and organs, RNA samples were submitted to Novogene, Inc. for ultralow input mRNA sequencing. Samples underwent quality control, library construction, and sequencing. RNA libraries for RNAseq were prepared by Novogene using a SMART-seq V4 Ultra Low Input RNA kit and a non-directional library. Messenger RNA was enriched using a poly-T oligo-attached magnetic beads. Samples were sequenced on an Illumina NovaSeq 6000 to generate 150 bp pair end reads ranging from 24 to 94 million raw reads per sample. RNAseq data were deposited into GEO and assigned accession number GSE276280.

Assessing mRNA abundance from RNAseq data

For our larval TRAP and FACS samples, raw RNAseq reads were mapped to the zebrafish reference genome (GRCz11) using the STAR alignment software package (v. 2.7.11b). Only those RNAseq reads mapping to annotated protein-coding Ensembl gene models (release 99, https://ftp.ensembl.org) were used for gene expression profiling. Quantification of the mapped RNAseq data was done using the RSEM software suite (v. 1.3.3), obtaining expected and normalized RNAseq read counts (in transcripts per million, TPM). DESeq2 (v. 1.38.3) was used to compare mRNA abundance between samples using the RSEM-calculated expected counts for each gene as input data. For comparing mRNA abundance between TRAP-enriched or FACS-sorted samples versus their corresponding non-enriched or non-sorted input, a generalized linear model (GLM) based on the negative binomial distribution was used, as summarized by the formula [GeneExpression ~ condition], where the condition is either ‘enriched’ (for TRAP-enriched or FACS-sorted mRNAs) and ‘not enriched’ (for TRAP experimental mRNA input or mRNA from dissociated FACS samples, respectively). The Wald test, as implemented by DESeq2, was used to test for statistical significance. For identifying mRNA abundance differences between the two TRAP-enriched samples, i.e., endothelial cell TRAP-enriched mRNAs versus whole animal TRAP-enriched mRNAs, a GLM as summarized by the formula [ GeneExpression ~ experiment + condition + experiment:condition] was used, where ‘experiment’ indicates how mRNAs were isolated (by TRAP-enrichement or input mRNAs), and ‘condition’ indicates the different samples being compared (endothelial cell or whole animal mRNAs). The interaction term ‘experiment:condition’ is, in essence, the ratio of ratios [(TRAP-enriched endothelial cell polysome/input endothelial cell polysome)/(TRAP-enriched whole animal polysome/input whole animal polysome)]. As implemented in DESeq2, the likelihood ratio test was used along with a reduced model that excludes the ‘experiment:condition’ interaction term to test for differences in the TRAP enrichment of mRNAs between endothelial cell and whole animal samples.

For the adult TRAP samples, sequenced FASTQ files were trimmed with cutadapt and aligned with hisat2 using lcdb-wf pipeline (https://github.com/lcdb/lcdb-wf, v1.9rc). We used zebrafish genome assembly version GRCz11 and Ensembl gene annotation version 99. Subsequent downstream analyses were performed with the lcdb-wf pipeline utilizing R and packages DESeq2, clusterProfiler and GeneTonic (for fuzzy clustering). To determine which genes were endothelial enriched within a specific tissue, one-sided Wald t-tests were used to select for genes that were upregulated in the TRAP pulldown (IP) for that tissue compared to its total mRNA lysate (input), later referred to as the IP > Input comparison. To determine which endothelial enriched genes were shared among all tissues and the whole fish control, a multiple testing procedure for multi-dimensional pairwise comparisons was applied [27]. To control the false discovery rate (FDR) across genes at 10%, we extracted raw p-values for each gene and each tissue IP > Input comparison, with missing values being replaced by 1. A Holm correction was applied across comparisons for each gene separately. The FDR was calculated across genes using gene-wise minimal Holm-corrected p-values, and the fraction R of genes with FDR < 0.1 was determined. For each IP > Input comparison we selected genes whose within-gene Holm-corrected p-values were less than 0.1*R (and whose across-genes FDR was less than 0.1). Intersection of these genes across all IP > Input comparisons yielded a set of 350 genes. Out of this list, select genes were picked to be displayed in a heatmap in which their TRAP pulldown log2fold expression for each tissue was ≥ 1.2 compared to their own tissue lysate with a p-adjusted value < 0.05.

To determine which genes were uniquely enriched in the endothelium of specific organs, we pre-screened for genes within each tissue whose endothelial TRAP pulldown (IP) was upregulated compared to its total tissue mRNA lysate (input) and, separately, whose endothelial TRAP pulldown was upregulated compared to the whole fish endothelial TRAP pulldown, using a similar multiple testing procedure for multi-dimensional pairwise comparisons as described above, and taking an intersection of the two comparisons. This yielded a list of 3304 genes combined across all tissues (excluding whole fish). Based on this dataset, we determined genes dominantly expressed in each tissue's TRAP pulldown (IP) sample. The criteria for selection were as follows: (1) a gene was selected in a given tissue during the pre-screening, (2) it passed the omnibus FDR < 0.1 testing overall gene change across IPs (using minimum Holm-corrected p-values of all pairwise IP vs. IP comparisons), (3) statistically verified greater IP expression than in all other tissue IPs (with cross-IP Holm-corrected p-values < 0.1 × R, where R is the fraction of genes passing the omnibus test), (4) the minimum across the tissue IP replicates is above any other IP replicates in all other tissues. Gene lists for each tissue were then further filtered as follows: (1) endothelial gene expression was at least 1.5-fold greater than that of the organ lysate, (2) endothelial gene expression was at least twofold greater than the endothelial gene expression of the next highest tissue, and (3) the difference in endothelial gene expression between the tissue of interest and the next highest expressing tissue was at least twofold greater than their tissue lysate differences. A heatmap was generated using the top 7–8 genes that fit these criteria for each organ. Known information on selected genes was obtained using ZFIN and UniProt databases [28, 29].

Functional annotation of gene lists determined from RNAseq analyses

Gene Ontology (GO) term enrichment analyses were performed using the PANTHER classification system (v18.0, https://pantherdb.org/). Statistical overrepresentation of biological processes GO terms was done using gene lists that were determined from the RNAseq analyses performed on larval samples. The Fisher’s exact test with Bonferroni correction was used for statistical testing.

Analysis of gene features

For calculating codon usage, the coding sequences (CDSs) for all annotated transcripts were obtained from Ensembl (release 99, https://ftp.ensembl.org) in FASTA format. Custom Perl scripts were used to retrieve specific transcript sequences and to count codons from these sequences. Only Ensembl CDSs having a length consistent with a genetic code made up of three-nucleotide codons were considered. Additionally, only CDSs with the start codon ATG and stop codons TAA, TAG, and TGA were considered. The selection criteria for a representative annotated transcript for each annotated gene were as follows: (1) The transcript with the longest CDS length was selected as the representative transcript sequence for its corresponding annotated gene, (2) when attempting to select for the longest transcript, if there was a tie in CDS lengths of two or more transcripts for a given annotated gene, the transcript with the longest mRNA length was then selected. Upon selection of the representative CDSs for each annotated genes, the codon usage of ‘more highly translated’ genes and ‘less highly translated’ genes was determined from the Ensembl CDS. Partitioning of genes into one of these two categories was done using the results from the RNAseq analysis between the whole animal TRAP-enriched mRNA versus the corresponding input mRNA. Codon usage was determined by counting the number of times a codon was seen among all CDSs of genes that had higher mRNA abundances in the whole-animal TRAP-enriched samples—the ‘more highly translated’ genes. Likewise, the same codon counting method was used for genes found to have lower mRNA abundances among whole animal TRAP-enriched samples, with these genes being considered ‘less highly translated.’

Western blot

Protein samples were collected and boiled in Laemmli buffer + 2-mercaptoethanol and run on 12% polyacrylamide gels, transferred onto PVDF membranes, and then blocked in 5% BSA. Blots were probed with either anti-RPL11 (Abcam, Cat. # ab79352, RRID: AB_2042832) or anti-HA (Sigma, H9658, RRID: AB_260092) antibodies. Blots were exposed on film (Amersham Hyperfilm ECL, GE28906836) using Amersham ECL Western Blotting Reagent (GE, RPN2106).

Fluorescence activated cell sorting

For Fluorescence Activated Cell Sorting (FACS), 24hpf AngioTag embryos were anaesthetized with Tricaine and dechorionated and deyolked. Embryos were then washed in PBS and incubated in 0.25% trypsin at room temperature with trituration until dissociated into a single cell suspension. Following embryo dissociation, trypsin was quenched with Leibovitz L-15 phenol-free media with 10% FBS (Gibco). Cells were passed through a 70 µM strainer and collected by centrifugation at 600xg for 1 min. Cells were resuspended in phenol-free L-15 + 10% FBS, and GFP + cells were collected by FACS sorting on a BD FACSAria™ III machine using BD FACSDiva™ software (Becton Dickinson Biosciences). Sorted cells were collected at 500×g for 5 min and RNA was isolated and DNase treated using the ZR-Duet kit (Zymo Research D7003).

Whole-mount RNA in situ hybridization

DIG-labeled antisense riboprobes for the below genes were generated using DIG Labeling Kit (Roche). In situ hybridization was performed as described [3]. BM purple (Sigma) was used for DIG-labeled probes. Riboprobes were generated for the following genes:

ENSDARG00000069998 (F′ GGTGCAGATAACTGGGAAGGTGATAG, R′ TCAGTGTGAAGACGTACACC),

ENSDARG00000076721 (F′ CAGATGAAGTAAAGTCAGTATCTGTGATG, R′ GTAGTCTGGTTGGTGAATGAATAAGC),

ENSDARG00000098293 (F′ GACCATGTGCTGAGAAATGTGAAGAG, R′ TGTTAGCTCCATTTCCGCAG),

ENSDARG00000098129 (F′ GCTGTCTGTGGAGCGCTAAGTGTTTGTCT, R′ TAATACGACTCACTATAGGGAGAATGTCACATCCGACCAATCAGAAT),

ENSDARG00000008414/exoc312a (F′ GAGGCTGAAGGTAGATTTGGACAGATCGAC, R′ TGTGGATCCCACTATTTTTACAGTG),

ENSDARG00000056643/slc22a7b.1 (F′ ACAACTTTATCGCCGCCATC, R′ AGCCCTCCAGTCATTCACAA),

ENSDARG00000099980/bpifcl (F′ CAGAAGCAGATGAAGTTCATTAGTTCATTA, R′ CTCCATGTTAGTGACTGCTTGTTGG)

Hybridization chain reaction (HCR) in situ

HCR probes were designed and purchased from Molecular Instruments or designed using the in-situ-probe-generator_v.0.3.2 [30] and purchased from IDT. The molecular instruments protocol for fixed whole-mount zebrafish embryos and larvae was followed with a few modifications. Whole fish were fixed in 4% paraformaldehyde (PFA) for 2 h at room temperature, with a cavity opened for better PFA penetration, then organs dissected and rinsed in 1 × phosphate-buffered saline. The methanol steps were not performed and instead organs directly underwent proteinase K treatment after the 1 × phosphate-buffered saline rinses which ranged from 30-50µg/ml for 10–15 min depending on the organ. 4 pmol of each probe was utilized for hybridization. Organs were incubated with 1μg/ml of DAPI in 5 × SSCT for one hour before a final 5 × SSCT only wash step.

Generation of constructs and transgenic lines

The RiboTag and AngioTag constructs were generated using Gateway Technology [31]. Rpl10a was PCR amplified from 24 hpf zebrafish cDNA using forward primer GTG AGA GGG GAG ATA TCA CG and reverse primer CTA AGC GTA ATC TCC AAC ATC GTA TGG GTA GTA GAG GCG CTG TGG TTT TCC CAT G, and TOPO TA cloned into the PCR-II vector. Bridging PCR was then used to add EGFP and viral 2A sequence [32] to generate the “RiboTag” construct pME-egfp-2a-rpl10a-3xHA. Gateway LR reactions were used to combine pDEST-IsceI-Flk7kb, pME-egfp-2a-rpl10a-3xHA, and p3E-polyA. The kdrl:egfp-2a-rpl10a3xHA “AngioTag” construct was digested with IsceI enzyme and microinjected into the blastomere of one-cell stage zebrafish embryos. A stable Tg(kdrl:egfp-2a-rpl10a3xHA)y530 germline transgenic line was established by screening through multiple generations.

The UAS:RiboTag construct was generated using SLiCE technology [32]. The pT1ump-14xUAS-MCS-POUT [34] was digested with EcoRI and XhoI and then forward primer TCC CAT CGC GTC TCA GCC TCA CTT TGA GCT CCT CCA CAC GAA TTC GCC ACC ATG GTG TCA AAA G and reverse primer ACA TGT TCA GTT AAC GGT GGC TGA GAC TTA ATT ACT AGT CTC GAG TTA AGC GTA ATC TGG AAC ATC were used to slice clone the RiboTag cassette downstream from the 14 × UAS sequence, using pME-egfp-2a-rpl10a-3xHA as the template.

The Gene C (ENSDARG00000098293) and Gene B (ENSDARG00000076721) mutants were generated using CRISPR-Cas9 technology, as outlined in Gagnon et al. [35]. For ENSDARG00000098293, the guide primer sequence was TAATACGACTCACTATAGGAATTGGGCGACTTACTGCGTTTTAGAGCTAGAAATAGCAAG, for ENSDARG00000076721, the guide primer sequence was TAATACGACTCACTATAGGTTTGGACCTCATGAGAGTGTTTTAGAGCTAGAA. Genotype was determined using an ABI 3130 (Applied Biosystems) with screening primers TGTAAAACGACGGCCAGTATGGCTGTAGATGAATGAAGACT and GTGTCTTTCTCAGCACATGGTCAGAGG for ENSDARG00000098293 and screening primers TGTAAAACGACGGCCAGTCAGGTGTGTTTGGTGCTGAT and GTGTCTTCACGGGCATTAACTCACCAT for ENSDARG00000076721.

The Gene C mutant described in this manuscript has a 20bp deletion in exon 2.

37TTGGGCGACTTAA—GTGAAGGAGTTAAGCGAAGCTCAGACC76

37TTGGGCGACTTACTGCAGGAGTTTAATGATGTTGTGAAGGAGTTAAGCGAAGCTCAGAC96

The Gene B mutant described in this manuscript has a 5bp deletion in exon 2.

192TTATCAGATACTGTGGATGTTTGGACCTCATGAG—GATAGCTGAAATCTATAAGCA247

192TTATCAGATACTGTGGATGTTTGGACCTCATGAGAGTCGGATAGCTGAAATCTATAAGCA252

Microscopic image acquisition and processing

Larvae used for imaging were anesthetized using 168 mg/L Tricaine (1X Tricaine) and mounted in 0.8–1.5% low melting-point agarose dissolved in embryo buffer and mounted on a depression slide. Confocal fluorescence imaging for larvae and adult organs was performed with an LSM 880 (Zeiss) or a Nikon TI2 with CSUW1 spinning disk microscope. Confocal images were processed using either Zen software or Nikon Elements. Schematics and figures were made using Adobe Illustrator 2023 and BioRender software.

Caudal plexus measurements

The dorsal–ventral width of the caudal plexus was measured under the first ISV posterior to the anal pore, using Fiji [36]. Measurements to the left, underneath, and to the right of the ISV were taken, and then averaged and normalized to the overall dorsal–ventral height of the embryo at the first ISV posterior to the anal pore, which was measured in a similar manner to the caudal plexus. The values for each genotype were then averaged, normalized to the wild-type measurement, and the standard error of the mean was determined. A student’s t-test was run to determine significance.