Rbm8a deficiency causes hematopoietic defects by modulating Wnt/PCP signaling

Abstract

Defects in blood development frequently occur among syndromic congenital anomalies. Thrombocytopenia-Absent Radius (TAR) syndrome is a rare congenital condition with reduced platelets (hypomegakaryocytic thrombocytopenia) and forelimb anomalies, concurrent with more variable heart and kidney defects. TAR syndrome associates with hypomorphic gene function for RBM8A/Y14 that encodes a component of the exon junction complex involved in mRNA splicing, transport, and nonsense-mediated decay. How perturbing a general mRNA-processing factor causes the selective TAR Syndrome phenotypes remains unknown. Here, we connect zebrafish rbm8a perturbation to early hematopoietic defects via attenuated non-canonical Wnt/Planar Cell Polarity (PCP) signaling that controls developmental cell re-arrangements. In hypomorphic rbm8a zebrafish, we observe a significant reduction of cd41-positive thrombocytes. rbm8a-mutant zebrafish embryos accumulate mRNAs with individual retained introns, a hallmark of defective nonsense-mediated decay; affected mRNAs include transcripts for non-canonical Wnt/PCP pathway components. We establish that rbm8a-mutant embryos show convergent extension defects and that reduced rbm8a function interacts with perturbations in non-canonical Wnt/PCP pathway genes wnt5b, wnt11f2, fzd7a, and vangl2. Using live-imaging, we found reduced rbm8a function impairs the architecture of the lateral plate mesoderm (LPM) that forms hematopoietic, cardiovascular, kidney, and forelimb skeleton progenitors as affected in TAR Syndrome. Both mutants for rbm8a and for the PCP gene vangl2 feature impaired expression of early hematopoietic/endothelial genes including runx1 and the megakaryocyte regulator gfi1aa. Together, our data propose aberrant LPM patterning and hematopoietic defects as consequence of attenuated non-canonical Wnt/PCP signaling upon reduced rbm8a function. These results also link TAR Syndrome to a potential LPM origin and a developmental mechanism.

INTRODUCTION

Mutations in genes that lead to pleiotropic phenotypes, as observed in complex syndromic birth anomalies, remain challenging to mechanistically connect to a tissue- or cell type-specific developmental defect. Thrombocytopenia with absent radius (TAR) Syndrome (OMIM #274000) is a rare (0.42 per 100’000 live births), autosomal-recessive congenital disorder that at birth manifests as blood platelet deficiency (hypomegakaryocytic thrombocytopenia) and bilateral absence of the radius bone ^1,2. While thrombocytopenia is the most consistent phenotype, the limb defects vary considerably among patients and can involve both arms and legs. Concomitant anomalies in TAR syndrome patients can include heart malformations, such as Tetralogy of Fallot or atrial septal defects, floating or unilaterally absent kidneys, as well as mild craniofacial alterations ^2–5. Curiously, the predominant coagulation defect of newborn TAR patients commonly recedes with age, and patients show normal platelet counts and blood clotting responses a few years after birth ^1,2. The complex combination of organs and cell types affected in TAR patients has so far precluded any clear assignment of a developmental cause underlying the syndrome.

Genetically, RBM8A has been identified as causative gene in TAR syndrome patients. Sequence analysis of TAR patients has revealed that compound inheritance of a 1q21.1 deletion that includes RBM8A is required with a noncoding single-nucleotide variant in the remaining copy of the RBM8A gene ⁶. Corroborating the causal link between TAR Syndrome and RBM8A. The major RBM8A allele combinations in TAR patients suggest that one gene copy is a complete null by deletion and the other is hypomorphic, while individual patients with bi-allelic RBM8A variants in the absence of 1q21.1 deletions have been reported ^1,2,5–7. RBM8A encodes RBM8A/Y14 that together with eIF4A-III, MLN51, and MAGOH constitutes the exon–junction complex (EJC) ⁸. The EJC is involved in essential post-transcriptional mRNA control by associating with exon-exon junctions after splicing to support nuclear export and sub-cellular localization of specific transcripts, translation enhancement, and nonsense-mediated RNA decay (NMD) ⁹. NMD as major mRNA quality control step during translation leads to degradation of mRNAs with retained introns as routinely caused by incomplete splicing, in particular of long introns ^10–13. Which mechanism(s), mRNA target(s), and developmental process(es) are affected in TAR Syndrome upon reduced RBM8A function remain unknown.

Pioneering work on recessive zebrafish rbm8a and magoh mutants has documented the impact of EJC perturbation in which maternal mRNA and protein contribution to the embryo rescues development until lack of zygotic supply causes aberrant phenotypes ¹⁴. Zygotic rbm8a loss-of-function resulted in no discernible morphological phenotypes until late segmentation stages (approximately 19 hpf), when cell death in the brain and general muscle paralysis became detectable ¹⁴. Notably, zebrafish rbm8a mutants showed disrupted NMD and increased stability of mRNAs with 3’ UTR introns by 24 hpf ¹⁴. In Drosophila, tissue-specific disruption of NMD using rbm8a and magoh mutants has been linked to accumulation of individual mRNAs with retained large introns that result in perturbation of select developmental processes including RAS/MAPK signaling ^10,11. Together, these data document both quantitative as well as qualitative impact on individual mRNAs upon EJC perturbation in different models.

A possible lead to connect the seemingly pleiotropic TAR Syndrome phenotypes comes from the developmental origin of the affected organs. The lateral plate mesoderm (LPM) forms the progenitor cells for blood, heart, vasculature, kidney, craniofacial muscles, and limb connective tissue in the developing vertebrate embryo ^15–20. Hypomorphic perturbation of RBM8A in TAR Syndrome could therefore impact genes involved in a shared developmental process at the base of blood, limb, heart, and kidney development. The diversity of LPM cell fates could explain complex, yet developmentally connected co-morbidities of syndromic congenital anomalies as LPM diseases that result from early patterning defects²¹. By affecting blood, forelimbs, heart, and kidneys, the spectrum of TAR syndrome phenotypes seems to predominantly affect tissues of LPM origin. Yet, what developmental mechanisms altered RBM8A/Y14 dosage in TAR Syndrome triggers to possibly cause a LPM defect, and in particular thrombocytopenia, remains unknown.

Hematopoietic cell lineages emerge in close association with endothelial progenitors within the LPM from bilateral progenitors expressing the transcription factors Tal1/Scl, Lmo2, and Etv2, starting with a first wave of Gata1-expressing primitive erythrocytes ^19,22–25. Subsequently emerging intermediate hematopoietic progenitors form primitive myeloid cells including transient megakaryocytes that in mammals begin to shed anucleate thrombocytes (also called platelets) for coagulation ^26–29. The Runx1-expressing hematopoietic stem cell (HSC) precursors emerge from the ventral wall of the dorsal aorta and undergo various maturation steps before seeding the final hematopoietic niches ^30–33. Definitive megakaryocytes emerge from HSC-derived common myeloid progenitors that also form erythrocytes plus mast cells and myeloblasts ^25,34–36. The Gfi1 transcription factors have been implicated in controlling hematopoietic lineage progression in human and mouse: Gfi1b predominantly drives megakaryocyte differentiation by suppressing erythroid fate, with missense mutations in human GFI1B causing thrombocytopenia ^37–39. In zebrafish, the orthologs gfi1aa and gfi1b contribute to primitive erythropoiesis and are expressed in erythroid, intermediate erythroid-myeloid progenitors (EMPs), and emerging hematopoietic stem cells ^40–42. At which step embryonic perturbation of Rbm8a affects thrombocyte formation and how it connects to the other LPM-associated defects awaits clarification.

Among the highly dynamic LPM that patterns while the embryo undergoes dramatic changes in length and cell number, hematopoietic progenitors emerge under the influence of general patterning and morphogenetic signals. Most-prominently, BMP and canonical Wnt/beta-catenin signaling have been linked to various steps in LPM patterning and hematopoietic lineage differentiation ^{20,25,43–47}. In contrast, non-canonical Wnt/Planar cell polarity (PCP) signaling is a critical pathway coordinating the polarized orientation of fields of cells and relative orientation of cells among their neighbors during early embryo morphogenesis ^48–50. Wnt/PCP signaling coordinates the convergent extension of the embryo during somite stages that also influence the lateral-to-medial migration of the LPM ^51–54. Triggered by ligands including Wnt5 and Wnt11, select Frizzled receptors together with Celsr/Flamingo co-receptors and Vangl1/2 relay ligand binding through cytoplasmic components including Prickle1 and Dishevelled to control cytoskeletal dynamics without apparent transcriptional response ⁴⁹. Beyond gross trunk defects resulting from global perturbation of the pathway, more selective defects in non-canonical Wnt/PCP signaling have been linked to structural anomalies affecting the neural tube, kidneys, and heart ^{51,52,54– 61}. Non-canonical Wnt signaling has been implicated in supporting hematopoietic stem cell emergence from the dorsal aorta by activating Notch ligands in somites ^62,63, and in the maintenance of hematopoietic stem cells in their bone marrow niche ⁶⁴. How deregulation of non-canonical Wnt/PCP signaling could contribute to earliest blood progenitor formation, migration, and congenital hematopoietic disease remains unknown.

Here, harnessing the dosage range provided by maternal contribution, different mutant rbm8a alleles, and antisense morpholino knockdown in zebrafish, we investigate if Rbm8a perturbation akin to TAR Syndrome causes the thrombocytopenia phenotype by impacting hematopoietic development from LPM. We provide evidence that reduced rbm8a function causes intron retention and misexpression of mRNAs encoding components of the non-canonical Wnt/PCP pathway. Subsequently, rbm8a attenuation results in several, cumulatively significant defects in early LPM and endothelial/hematopoietic lineage patterning that we also document in embryos with classic PCP defects. Our data connect impaired PCP signaling with hematopoietic phenotypes as possible developmental origin of the phenotypes observed in TAR Syndrome.

RESULTS

Null and hypomorphic rbm8a perturbation in zebrafish

Pioneering previous work has established that zebrafish rbm8a is maternally contributed as mRNA and protein, persisting for at least 24 hpf ¹⁴. This waning maternal contribution has been used to assess hypomorphic and loss-of-function phenotypes from 19 hours post fertilization (hpf) on, yet rbm8a-mutant embryos deteriorate rapidly past 24–28 hpf and are severely deformed by 72 hpf ¹⁴. We therefore sought genetic means to decrease Rbm8a protein levels to establish hypomorphic Rbm8a function beyond these timepoints.

To generate mutant alleles, we targeted the zebrafish rbm8a locus by CRISPR-Cas9 mutagenesis to induce non-homologous end joining (NHEJ)-mediated lesions in the coding sequence (Fig. 1A). We raised surviving low-dose crispants and isolated two different mutant rbm8a alleles: g.5152_5156del and rbm8a g.5152_5154del, subsequently abbreviated as rbm8a^Δ5 and rbm8a^Δ3, respectively (Fig. 1A). rbm8a^Δ5 is a frameshift allele with no downstream alternative start codons that likely represents a null allele (Fig. 1A); no Rbm8a protein is detected by Western Blot at 24–26 hpf after the considerable maternal mRNA and protein deposition subsides (Fig. 1B, Supplementary Fig. 1A–D). Notably, our independently isolated allele is molecularly identical to the allele previously reported by Gangras and colleagues and generated with the same recommended sgRNA sequence, indicating preferential repair of this lesion with a 5 bp deletion ¹⁴. In contrast, our additional allele rbm8a^Δ3 deleted three base pairs, replacing phenylalanine (F) and proline (P) codons with a single serine (S) codon, substituting two non-polar side chains with a polar OH group (Fig. 1A). rbm8a^Δ3 substitutes amino acids right before the first alpha-helix in the Rbm8a N-terminus that is in proximity with the beta-sheet in Magoh ^65–67, proposing rbm8a^Δ3 as potential hypomorphic allele.

Figure 1: Genetic perturbations in the zebrafish rbm8a gene.

(A) Allelic series of rbm8a in zebrafish with wild type allele, rbm8a^Δ5 allele and rbm8a^Δ3 allele induced by Cas9-mediated mutagenesis. (B) Western blot analysis of Rbm8a protein in mutants and morphants at different developmental stages, showing gradual decrease of maternal protein in mutants and translation block in morphants. (C) Representative images of zebrafish embryos carrying different rbm8a allele combinations during early development. Compared to wild type and different rbm8a^Δ5 and rbm8a^Δ3 combinations, homozygous rbm8a^Δ5/Δ5 carriers display severe microcephaly and corkscrew tail phenotype. rbm8a morphant (1:100 morpholino dilution) phenotypically recapitulates the homozygous rbm8a^Δ5/Δ5 embryos. Scale bars in (C): 250 μm, applies to all panels in C.

Embryos heterozygous for either allele (rbm8a^Δ3/+ and rbm8a ^Δ5/+) showed no obvious malformations at observed timepoints up to 5 days post-fertilization and adults were viable and fertile. As previously reported and compared to wild type siblings, homozygous rbm8a^Δ5 (rbm8a^Δ5/Δ5) zebrafish developed seemingly normal for the first 19 hpf, with subtle axis shortening during segmentation, before rapidly developing severe body axis hypoplasia (Fig. 1C). Around 26 hpf, the embryos started to display hypoplastic heads and tails with twisted axis, yet featured slowed heartbeat and rudimentary circulation before progressively deteriorating without survival past 60–72 hpf (Fig. 1C). In contrast, homozygous rbm8a^Δ3 (rbm8a^Δ3/Δ3) embryos developed and survived without obvious overt defects into fertile adults (Fig. 1C). By crossing, we derived trans-heterozygous rbm8a^Δ5/Δ3 embryos that harbor one allele as likely null while the other provides hypomorphic activity; rbm8a^Δ5/Δ3 embryos resembled wild type siblings during early development and continued to develop seemingly normal until adulthood. Incrosses of adult rbm8a^Δ5/Δ3 resulted in the expected phenotypes and allele combinations, as observed in heterozygous incrosses of the individual alleles.

To further control embryonic Rbm8a protein levels, we obtained the translation-blocking morpholino MO-rbm8aATG. Injection of 1 nL of 0.01 nmol/μL MO-rbm8aATG phenocopied rbm8a^Δ5/Δ5 embryos by morphology (Fig. 1C). Morpholino injection also resulted in rapid translation block and reduction of Rbm8a protein already at tailbud stage and absence by 17 hpf, as detected by Western blot (Fig. 1B, Supplementary Fig. 1A–D). Co-injection of MO-rbm8aATG with capped mRNA encoding human RBM8A that has no sequence overlap with the morpholino rescued the knockdown embryos to at least 5 dpf, while EGFP mRNA as control had no rescue capacity (Supplementary Fig. 1E–K, Supplementary Data 1). In sum, these data validate MO-rbm8aATG as complementary tool to our rbm8a allelic series⁶⁸. Altogether, these reagents establish a zygotic-mutant allelic series of zebrafish rbm8a and a potent translation-blocking knockdown reagent to study the impact of rbm8a perturbations on early development (Supplementary Fig. 1L).

To gain insights into specific phenotypes in our mutants, we used mRNA in situ hybridization (ISH) for diverse markers related to TAR Syndrome-associated structural defects (Supplementary Fig. 2A). We did not observe any overt phenotypes in rbm8a-mutant embryos at 17 hpf (around 14–15 somite stage), 48 hpf, and 72 hpf (Supplementary Fig. 2B–D): we observed no striking differences in the expression patterns of cardiovascular and hematopoietic (pu.1, fli1a, vcana, gata1), mesothelial (hand2, wt1a), kidney (pax2a, wt1a) and broader LPM (bmp4) markers in rbm8a^Δ5/Δ5 or rbm8a^Δ3/Δ5 embryos compared to wild type at 14–15 somite stage, despite the increasing deterioration in overall morphology observed in rbm8a^{Δ5/Δ5 14}. We noted in individual rbm8a^Δ5/Δ5 embryos at 17 hpf that expression of the LPM genes hand2 and gata1 showed short, subtle interruptions in the continuous bilateral expression domains, which were however of variable penetrance among embryos (Supplementary Fig. 2B; see also below). Based on this first, albeit limited, gene expression analysis, we conclude that rbm8a-mutant zebrafish do not feature overt phenotypes in several TAR Syndrome-associated tissues or cell types at early stages of development, including the forelimbs (pectoral fins), kidney, or circulatory system. These observations are consistent with residual maternal contribution of Rbm8a-encoding mRNA and protein in our zygotic mutant conditions (Fig. 1)¹⁴.

Hypomorphic rbm8a expression results in thrombocyte reduction

The functional equivalent of cytoplasmic platelets shed by megakaryocytes in humans, nucleated thrombocytes in zebrafish are labeled by the transgenic reporter cd41:EGFP starting from 40–46 hpf^69,70. Notably, cd41:EGFP transgenics harbor EGFP-high thrombocytes and EGFP-low prospective hematopoietic stem cell precursors (HSPCs) that can be distinguished by sorting and fluorescence imaging^71,72. As rbm8a ^Δ5/Δ5 mutants and MO-rbm8aATG morphants deteriorate beyond 28–30 hpf (Fig. 1C) ¹⁴, we sought to test thrombocyte numbers in genetic combinations with the hypomorphic rbm8a^Δ3 allele. We additionally established an injection titration curve for MO-rbm8aATG to reduce Rbm8a protein levels and to maintain viability: the suboptimal dose ranged between 1:125 and 1:200 stock dilution and resulted in slight to no detectable developmental defects in the morphants with survival beyond 7 dpf, compared to a dose of 1:100 that fully phenocopied rbm8a^Δ5/Δ5 mutants (Fig. 1C).

We next sought to quantify thrombocytes in wild type controls and upon mutant and morpholino-induced rbm8a perturbation in cd41:EGFP-transgenic larvae at 3 dpf and 6 dpf, when thrombocytes are clearly detectable. Previous work had applied live-imaging of circulating cd41:EGFP cells on a chosen field of view and counted cells during a pre-determined time window in the resulting video capture^73,74. To increase throughput and to capture all cd41:EGFP-positive thrombocytes in individual larvae, we devised an alternative imaging-based workflow. To transiently arrest the heartbeat and stop circulation, we treated the larvae with 2,3-Butanedione monoxime (BDM) and quantified EGFP-positive cells from fluorescent images of whole larvae using custom Fiji scripts (Fig. 2A). This approach provided a semi high-throughput quantification to distinguish the circulating, high EGFP-expressing thrombocytes from the immobile, sparser low EGFP-expressing HSPCs in the caudal hematopoietic territory (CHT) (Fig. 2A–I). Our quantification documented a significant decrease in the number of high EGFP-positive cells at both 3 and 6 dpf in the morphants with a clear response to the dose of morpholino compared to wild type animals (Fig. 2B–E,J,K, Supplementary Data 2). Notably, viable mutant allele combinations for rbm8a showed no significant change to cd41:EGFP-expressing cell numbers at 3 dpf (Fig. 2F,H,J, Supplementary Data 2), in line with unperturbed maternal deposition of rbm8a mRNA and translated protein (Fig. 1C) ¹⁴. In contrast, at 6 dpf, rbm8a^Δ3-homozygous larvae showed a significant reduction of cd41:EGFP cell counts, while rbm8a^Δ3/Δ5 larvae showed an even further reduction (Fig. 2G,I,K, Supplementary Data 2). This reduction was significant in larvae derived from incrosses of rbm8a^Δ3/Δ5 parents in which the oocytes only harbor the hypomorphic rbm8a allele as functional copy, further underlining the influence of maternally contributed wild type rbm8a transcript. These data are in line with a potential hypomorphic quality of the rbm8a^Δ3 allele. Of note, counting thrombocytes at 6 dpf using video-based methodology as per previous work ⁷³ resulted in comparable observations and trends: wild type and heterozygous rbm8a^Δ5/+ larvae showed comparable numbers to the semi-automated count (Supplementary Fig. 2E, Supplementary Data 2). Homozygous rbm8a^Δ3/Δ3 and trans-heterozygous rbm8a^Δ3/Δ5 larvae showed slightly lower overall cd41:EGFP-expressing cell numbers using the video-based count, which further resulted in mild, but not significant, reduced thrombocyte numbers in the homozygous rbm8a^Δ3/Δ3 and trans-heterozygous rbm8a^Δ3/Δ5 larvae compared to wild type (Supplementary Fig. 2F, Supplementary Data 2). Together, our data indicate that reducing rbm8a function in zebrafish causes a reduction in cd41:EGFP-expressing thrombocytes.

Figure 2: rbm8a perturbation reduces thrombocyte numbers in zebrafish larvae.

(A) Workflow schematic of cd41:EGFP-positive thrombocyte progenitor quantification from circulation and from the caudal hematopoietic territory (CHT). Created with Biorender.com (Subscription: Individual, Agreement number: KV25WQCJKF). (B-I) Representative fluorescent dissecting scope and confocal images of zebrafish embryos transgenic for the thrombocyte marker cd41:EGFP, greyscale and color-inverted to reveal GFP-positive cells; anterior to the left, insert depicts brightfield image of larvae for reference. (J,K) cd41:EGFP-positive thrombocyte counts at 3 dpf (J) and 6 dpf (K) for each analyzed condition. Wild type larvae have significantly more cd41:EGFP-positive cells at 3 and 6 days compared to the high morpholino dose (1:125 and 1:150, only at 3 dpf), while lower dose (1:200) remained at wild type levels (J,K). At 6 dpf (K), viable trans-heterozygous allele combinations for rbm8a also show a significant thrombocyte reduction. Note that morpholino injections will block translation from maternal mRNA, while mutant combinations will retain maternal mRNA function from the wild type allele carried by the mother. Individual datapoints (total number of cd41:GFP-positive cells per embryo) shown with mean and standard deviation, significance calculated by Mann-Whitney test: 3 dpf (J) wild type vs. control MO p=0.7756 (not significant), wild type vs. MO-rbm8aATG 1:125 p<0.0001, wild type vs. MO-rbm8aATG 1:150 p<0.0001, wild type vs. MO-rbm8aATG 1:175 p=0.2589 (not significant), wild type vs. MO-rbm8aATG 1:200 p=0.5628 (not significant), rbm8a^+/+ vs. rbm8a^Δ5/+ p=0.605 (not significant), rbm8a^+/+ vs. rbm8a^{Δ3/ Δ3} p=0.7655 (not significant), rbm8a^+/+ vs. rbm8a^{Δ3/ Δ5} p=0.4289 (not significant); 6 dpf (K) wild type vs. control MO p=0.9567 (not significant), wild type vs. MO-rbm8aATG 1:125 p<0.0001, wild type vs. MO-rbm8aATG 1:150 p=0.5198 (not significant), wild type vs. MO-rbm8aATG 1:175 p=0.7147 (not significant), wild type vs. MO-rbm8aATG 1:200 p=0.288 (not significant), rbm8a^+/+ vs. rbm8a^Δ5/+ p=0.9462 (not significant), rbm8a^+/+ vs. rbm8a^{Δ3/ Δ3} p=0.0208, rbm8a^+/+ vs. rbm8a^{Δ3/ Δ5} p<0.0001 (see Supplementary Data 2 for details). Scale bars in (B): 200 μm, applies to panels B,D,F,H; (C): 150 μm, applies to panels C,E,G,I.

We further tested the functionality of thrombocytes with reduced rbm8a function. Following laser-mediated injury of the dorsal aorta, cd41:EGFP-positive thrombocytes aggregated at the new wound site to occlude the damaged artery. The resulting time to occlusion (TTO) and number of aggregating thrombocytes provide a measure for a functional coagulation response ^74,75. We did not observe any significant differences in TTO or number of aggregating thrombocytes between 6 dpf wild type and heterozygous rbm8a^Δ5 larvae (Supplementary Fig. 2G,I, Supplementary Data 2), homozygous rbm8a^Δ3/Δ3 or trans-heterozygous rbm8a^Δ3/Δ5 (Supplementary Fig. 2H,J, Supplementary Data 2).

Taken together, using our available genetic tools, these observations indicate that while rbm8a-perturbed larvae show reductions in thrombocyte numbers that are dependent on functional Rbm8a levels, the function of the thrombocytes is not affected.

rbm8a perturbation impairs mRNAs encoding planar cell polarity components

To identify developmental mechanisms influencing thrombocyte formation and general hematopoiesis upon rbm8a perturbation, we sought to define the transcriptome of rbm8a-mutant zebrafish embryos at early developmental stages. Prior work has described the transcriptome of 21 hpf and 27 hpf rbm8a^Δ5/Δ5 embryos, when mutants show clear signs of deterioration including onset of widespread apoptosis and necrosis¹⁴. We therefore analyzed the transcriptome of tailbud (9 hpf) rbm8a^Δ5/Δ5 embryos as well as 24 hpf as comparison, capturing with the former the end of gastrulation when maternal Rbm8a protein and rbm8a mRNA levels start to wane (Fig. 3 A–C). The RNA-seq data can be browsed in our R/Shiny-based app RNA-seq Explorer with an interactive interface (http://imlspenticton.uzh.ch:3838/mosimann_p2452/; see Methods for details).

Figure 3: Intron retention in mRNAs encoding non-canonical Wnt/PCP components in zebrafish rbm8a mutants.

(A) Experimental design of bulk RNA-seq experiment. Created with Biorender.com (Subscription: Individual, Agreement number: SN25WQAKB9). (B,C) Volcano plots of tailbud (B) and 24 hpf (C) comparisons between wild type versus rbm8a-mutant embryos; significantly down- (magenta) and up-regulated (blue) genes with named examples (see Supplementary Data 3 and 4 for details). (D) Qualitative assessment of retained introns between wild type versus rbm8a-mutant embryos; note that retained introns are already apparent at tailbud stage (see Supplementary Data 6 for details). (E, F) Read coverage plot of mRNA sequencing reads of non-canonical Wnt/PCP components celsr2, wnt11f2 (former wnt11), and vangl2 that feature mRNAs with retained introns (E,F) and/or differential transcript usage (G) in rbm8a-mutant embryos. (H) Differential transcript usage of vangl2 transcripts at 24 hpf (exon 1 followed by long intron, asterisks).

Comparing rbm8a^Δ5/Δ5 versus wild type embryos, we identified 850 differentially expressed genes (720 of which were protein coding) with FDR <= 0.05 at tailbud stage (Fig. 3B, Supplementary Data 3), and 1988 genes (1818 of which were protein-coding) with FDR <= 0.05 at 24 hpf (Fig. 3C, Supplementary Data 4). We used Metascape ⁷⁶ to identify statistically enriched functional annotations of deregulated genes and their accumulative hypergeometric p-values (Supplementary Fig. 3A,B, Supplementary Data 5): similar to the reported transcriptomes of 21 and 27 hpf rbm8a^Δ5/Δ5 embryos ¹⁴, at 24 hpf we observed enrichment for cell death and responses to stress among the upregulated genes, as well as downregulation of cell cycle processes and chromatin organization (Supplementary Fig. 3A,B, Supplementary Data 5). These transcriptional changes are in line with the progressive deterioration of rbm8a^Δ5/Δ5 embryos as visible by the mutant and morphant phenotypes at that stage (Fig. 1) ¹⁴. In contrast, at tailbud stage, we observed broad deregulated gene categories associated with morphogenesis and general organ development (Supplementary Fig. 3A,B, Supplementary Data 5).

As part of the EJC, Rbm8a functions in NMD that removes mRNAs with retained introns or other splicing defects that occur routinely during native post-transcriptional processing ^12,13,77,78. In Drosophila, loss of rbm8a causes selective retention of mRNAs with mis-spliced large introns due to defective NMD ^10,11, and previous analysis of zebrafish rbm8a mutants documented stabilization of mRNAs with introns in their 3’ UTRs ¹⁴. Together, these data document both quantitative as well as qualitative impact on individual mRNAs upon EJC perturbation in different models. Given the depth of sequencing and early versus late timepoint in our dataset, we next asked if qualitative differences in individual transcripts could link Rbm8a deficiency with TAR Syndrome-associated processes.

We therefore scanned the tailbud stage and 24 hpf transcriptome of rbm8a-mutant embryos for differential exon and intron usage of individual mRNAs using DEXSeq (Supplementary Fig. 3C) ⁷⁹. After filtering (retaining only protein-coding transcripts and further excluding U12 splice introns, transcripts with read counts below 100, adjusted p value per gene <= 0.05 and differential exon/intron usage event <= 0.1), we found 697 differential intron usage events in 564 genes (Supplementary Data 6) and 713 differential exon usage events in 502 genes (Supplementary Data 7) in rbm8a^Δ5/Δ5 mutants compared to wild type at tailbud stage (Supplementary Fig. 3C). In contrast, and again in line with progressive deterioration, we found 2220 differential intron usage events in 1594 genes (Supplementary Data 6) and 5439 differential exon usage events in 2316 genes (Supplementary Data 7) in rbm8a^Δ5/Δ5 compared to wild type at 24 hpf (Supplementary Fig. 3C). This analysis reveals a considerable qualitative difference in numerous mRNAs following rbm8a perturbation in zebrafish, including mRNAs with retained introns. Curiously, most of the retained introns were larger than 1 kb at tailbud stage, compared to retained introns of all sizes at 24 hpf when the rbm8a^Δ5/Δ5 embryos were declining (Fig. 3D). We speculate that at tailbud stage, given Rbm8a involvement in the EJC, problems with longer introns start to accumulate while smaller introns seem to still being spliced properly; at 24 hpf, introns of all sizes are affected.

We again used Metascape ⁷⁶ to identify statistically enriched functional annotations of the genes with differential intron usage events at tailbud stage and 24 hpf (Supplementary Fig. 3C). We observed statistically enriched annotations for cell-cell adhesion and cell junction organization at tailbud stage and at 24 hpf stage amongst others (Supplementary Fig. 3E, Supplementary Data 8). When we looked manually through the genes associated with the enriched GO terms, we found celsr2, wnt11, and prickle1, as well as several additional mRNAs encoding components of the non-canonical Wnt/PCP signaling, including wnt5b and vangl2 (Fig. 3E–G, Supplementary Fig. 3F), among irregular transcripts with developmental contributions at both analyzed stages. While per se not enough to evoke a GO term involving non-canonical Wnt signaling, even though the GO terms cell-cell adhesion and cell junction organizations are cellular processes linked to PCP signaling, significant intron retention of these genes indicates possibly perturbed non-canonical Wnt pathway involved in PCP signaling. PCP signaling equips tissues with a polarity axis resulting in collective morphogenetic events, such as the orientation of subcellular structures and cell rearrangements ⁴⁹. These mRNAs showed variable, yet considerable sequencing reads in individual introns, indicating seemingly minor but significant accumulation of mis-spliced transcripts with retained introns (Fig. 3E–G, Supplementary Fig. 3F). Such intron retention is predicted to lead to unproductive protein translation or truncated, functionally perturbed polypeptides ^12,13, potentially lowering the effective dose of individual non-canonical Wnt/PCP components. We also noted that one transcript isoform of celsr2, encoding one of several Flamingo family receptors for non-canonical Wnt/PCP ligands ^49,80, was among the top-upregulated mRNAs at tailbud stage in rbm8a^Δ5/Δ5 embryos (logFC 10.4, FDR=0.013); at 24 hpf, celsr2 mRNAs showed selective intron retention of intron 15 (logFC 2.4, p=0.00015) (Fig. 3B,C, Supplementary Fig. 3F). Transcripts of the key PCP component vangl2 showed differential variant use between our sequenced timepoints (Fig. 3G,H).

Taken together, our analysis indicates that rbm8a-mutant embryos already show developmental anomalies at the bulk transcriptome level at the end of gastrulation, when maternal rbm8a mRNA and Rbm8a protein contribution is starting to fade. From our qualitative assessment of mRNAs at tailbud and 24 hpf stage, we hypothesized that rbm8a-perturbed zebrafish embryos feature a mild, yet functionally significant attenuation of non-canonical Wnt/PCP signaling involved in convergence and extension movements and other cell polarity coordination. Defective PCP signaling could impact the proper migration, and subsequently influence the cell fate determination of, the LPM stripes and the hematopoietic/endothelial progenitors.

Zebrafish Rbm8a deficiency causes aberrant LPM morphology

As the phenotypes in TAR Syndrome arise in LPM-derived organs and cell types, we next sought to determine if the early LPM develops normally upon rbm8a perturbation. To observe LPM formation and patterning after gastrulation, we performed light sheet-based live imaging using the transgenic reporter scl:EGFP that marks the emerging hematopoietic and endothelial progenitors in the medial LPM from early somitogenesis stages (Fig. 4A) ^19,81,82. In wild type, the bilateral scl-expressing LPM stripes gradually widen and thicken before starting an anterior-to-posterior midline migration around the 5 somite-stage (Fig. 4B,E,F). In contrast, rbm8a knockdown with MO-rbm8aATG resulted in reduced area and volume of the scl:EGFP-expressing LPM in the same developmental timespan (Fig. 4C,E,F). At the imaging endpoint at 10–11 somites, EGFP fluorescence from scl:EGFP showed no discernible difference in intensity between wild type and morphants, yet their overall morphology in mutants appeared wider apart and less converged to the midline (Fig. 4B,C,E,F). This reduced convergence is unlikely due to developmental delay, as we stage-matched the analyzed embryos by morphology and not absolute time post-fertilization.

Figure 4: rbm8a and vangl2 perturbations reduce the rate of posterior LPM growth.

(A) Schematic of the workflow to image the zebrafish using Light sheet microscope from 4 different angles, assemble in a 3D image (represented by the globe) and measure volume and area of the PLPM scl:GFP surface rendering (represented by the map). Created with Biorender.com (Subscription: Individual, Agreement number: IG25WQA6FD). (B-D) Surface rendering of GFP signal from light sheet-based timelapse imaging of scl:GFP transgenic zebrafish embryos. Wild type (A), rbm8a morphant (B), and vangl2 morphant (C) embryos at 5, 7, 9 and 11 somite stages, dorsal view, anterior to the top, posterior end of the LPM at bottom. (D,E) Area and volume of scl:GFP-expressing LPM territory measured with Imaris, comparing wild type (n=4), rbm8a morphant (n=5) and vangl2 morphant (n=4) measurements depicted in color groups corresponding to sample groups. Note how area in wild type increases by x^2 (D) and volume by x^3 (E), revealing reduced growth rates of the posterior scl:GFP territory upon rbm8a and vangl2 perturbations. Datapoints shown are average with standard deviation, significance calculated by 2-way Anova. Area (p<0.0001) and volume (p<0.0001) are significantly different in MO-rbm8aATG- and MO1-vangl2-injected embryos compared to stage-matched wild type embryos at the analyzed timepoints (see Supplementary Data 9 for details). Scale bar in (B): 150 μm, applies to all panels in B-D.

Reflective of the central role of Vangl2 in relaying PCP signaling ^83–85, zebrafish embryos mutant for vangl2 and morphants show severe convergence and extension defects, resulting in embryos with shortened, wider posterior trunks and compacted tails ^{60,84,86–91}. Consistent with these defects, vangl2 MO knockdown resulted in reduced volume and width of the scl:EGFP-expressing LPM stripes with no discernible impact on fluorescent reporter levels (Fig. 4D–F, Supplementary Data 9). Together, these observations indicate that the scl-expressing LPM in both rbm8a- and in vangl2-perturbed embryos develops an aberrant morphology.

rbm8a-mutant embryos show convergence and extension defects

Non-canonical Wnt/PCP signaling controls cell migration during convergence and extension movements that drive zebrafish embryonic axis elongation ^{83,85,89,91–93}. Hallmarks of defective convergence and extension include reduced length of the anteroposterior axis and increased width of the neural plate and somites ^{51,60,61,89,90,93–96}. To document and quantify if rbm8a perturbation causes convergence and extension defects in zebrafish, we measured morphometric parameters during early somite stages of wild type and rbm8a-mutant embryos: measured parameters included axis length, somite and neural plate width, three morphometric parameters affected by reduced convergence and extension in PCP mutants. To assign landmarks for these measurements, we combined in situ hybridization for dlx3b, tbxta, myoD, and hgg1 as markers of the neural plate border, chordamesoderm, somites, and prechordal plate, respectively ^97–101.

Compared to stage-matched wild type siblings, rbm8a^Δ5/Δ5 mutants as well as rbm8a morphants showed significantly reduced axis lengths concomitant with a decrease in axis angle at the analyzed 6–11 somite stages (Fig. 5A, Supplementary Data 10). The axis length and angle discrepancy was already detectable at the first developmental timepoint we measured. While rbm8a^Δ5/Δ5 mutant and rbm8a morphant embryos continued to extend in length, they fell short of reaching wild type lengths at the end of our stage series (Fig. 5B,C). In contrast, rbm8a^Δ3/Δ5 and rbm8a^Δ3/Δ3 mutant embryos did not show significant changes in axis length or axis angle at the 6–7 somite timepoint, but continued to being slightly shorter compared to wild type towards the end of our measurements (Fig. 5B,C). Further, compared to wild type embryos, rbm8a^Δ5/Δ5 mutant embryos featured wider neural plates (Fig. 5D) throughout our timecourse, while the rbm8a morphants, as well as rbm8a^Δ3/Δ5 and rbm8a^Δ3/Δ3 mutant embryos did not, despite a slight increase at the 8–9 somite stage (Fig. 5D). Lastly, rbm8a^Δ5/Δ5 mutant embryos had significantly wider somites throughout the timecourse compared to wild type embryos (Fig. 5E). The somite width of the rbm8a morphants was significantly reduced as well compared to wild type (except at the 8–9 somite stage), however their somites were not as wide as the somites of the rbm8a^Δ5/Δ5 mutants (Fig. 5E). In rbm8a^Δ3/Δ5 and rbm8a^Δ3/Δ3 mutants, the somite width was unchanged compared to wild type embryos at all timepoints (Fig. 5E).

Figure 5: rbm8a perturbation results in convergence and extension defects.

(A) mRNA in situ hybridization for myoD, dlx3b, tbxta, and hgg1 as landmarks to measure morphometric parameters associated with convergence and extension as influenced by non-canonical Wnt/PCP signaling. Zebrafish embryos are shown as lateral views (odd columns, anterior to the top, ventral to the left) and dorsal views (even columns, anterior to the top). Black arrowheads depict anterior and posterior extent of the body axis in lateral views, pink and grey arrowheads depict otic placodes and somites at measured position in dorsal views. Axis length and axis angle (marked by black arrowheads) were measured from the odd panels. Neural plate (between the otic placodes at beginning of notochord, pink arrowheads) and somite width (last 3 somites, grey arrowheads) were measured from the even panels. (B-E) Morphometric measurements for axis length (B), axis angle (C), neural plate width (D), and somite width (E) from individual embryo (wild type, mutant, morphant) images at different developmental stages. Datapoints shown are average with standard deviation, significance calculated by Mann Whitney test: axis length (B) 6–7 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.4394 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.9347 (not significant), 8–9 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.0048, wild type vs. rbm8a^Δ3/Δ3 p=0.1643 (not significant), 10–11 somite stage wild type vs. MO-rbm8aATG p=0.0001, wild type vs. rbm8a^Δ5/Δ5 p=0.0003, wild type vs. rbm8a^Δ3/Δ5 p=0.0945 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.1219 (not significant); axis angle (C) 6–7 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.3325 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.4449 (not significant), 8–9 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.7492 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.2633 (not significant), 10–11 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.439 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.342 (not significant), neural plate width (D) 6–7 somite stage wild type vs. MO-rbm8aATG p=9978 (not significant), wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.4437 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.3001 (not significant), 8–9 somite stage wild type vs. MO-rbm8aATG p=0.0086, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ3 p=0.0008, 10–11 somite stage wild type vs. MO-rbm8aATG p=0.8324 (not significant), wild type vs. rbm8a^Δ5/Δ5 p=0.0198, wild type vs. rbm8a^Δ3/Δ5 p=0.5477 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.5448 (not significant); somite width (E) 6–7 somite stage wild type vs. MO-rbm8aATG p<0.0001, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.2296 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.7951 (not significant), 8–9 somite stage wild type vs. MO-rbm8aATG p=0.1853 (not significant), wild type vs. rbm8a^Δ5/Δ5 p<0.0001, wild type vs. rbm8a^Δ3/Δ5 p=0.2642 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.6899 (not significant), 10–11 somite stage wild type vs. MO-rbm8aATG p=0.0028, wild type vs. rbm8a^Δ5/Δ5 p=0.0003, wild type vs. rbm8a^Δ3/Δ5 p=0.5769 (not significant), wild type vs. rbm8a^Δ3/Δ3 p=0.4317 (not significant) (see Supplementary Data 10 for details). Scale bar in (A): 250 μm, applies to all panels in A.

Together, while milder than phenotypes observed in mutants for core regulators of the process such as kny or tri/vangl2 ^89,96, our results indicate that somite-stage rbm8a-deficient zebrafish embryos show phenotypes consistent with convergence and extension defects.

Non-canonical Wnt/PCP signaling is sensitive to Rbm8a levels

We next tested whether non-canonical Wnt/PCP signaling is sensitive to attenuated Rbm8a levels. We used previously established morpholino oligonucleotides to reduce the levels of the intron-retained PCP ligands, wnt5b and wnt11f2, as well as transmembrane PCP core components fzd7a and vangl2: each morpholino has a well-established injection dose that recapitulates the recessive mutant phenotypes of each respective gene as previously validated per current guidelines ^{52,68,102,103}.

To determine whether reduction of non-canonical PCP core components sensitizes zebrafish embryos to reduced Rbm8a levels, we co-injected MO-rbm8aATG (Fig. 6A,B) together with each PCP core component-targeting morpholino at sub-threshold concentrations into wild type embryos (Fig. 6C–F). We used morphology of the body axis as a read-out to score the level of interaction between rbm8a and PCP core components at 48 hpf. Combined knockdown of individual PCP core component genes and rbm8a resulted in phenotypes that approached the full-dose knockdown of each Wnt/PCP gene (Fig 6C–F). For fzd7a, vangl2, and wnt11f2, over 90% of all co-injected embryos became phenotypic, indicating a strong genetic interaction with rbm8a resulting in a predominant severe phenotype class and lower number of moderate phenotypic embryos (Fig. 6G, Supplementary Data 11). wnt5b attenuation showed the mildest, yet still significant, response to rbm8a co-attenuation with 40% embryos becoming phenotypic (Fig. 6E,G). The combined attenuation resulted in embryos with shortened body axes and concomitant circulation defects, as common to perturbations of PCP signaling that affects embryo elongation by convergence and extension ^{49,51,89–91,93}.

Figure 6: rbm8a attenuation sensitizes zebrafish to non-canonical Wnt/PCP defects.

(A-F) 2 dpf lateral views of representative zebrafish larvae analyzed for each genetic perturbation. Anterior to the left. (A) Uninjected wild type as phenotype reference. (B) Suboptimal (1:500) MO-rbm8aATG injection that does not result in a visible phenotype (as compared to higher doses) and provides a baseline for rbm8a attenuation. (C-F) Representative phenotypes following sequential injection of suboptimal MO-rbm8aATG and suboptimal doses of individual, validated morpholinos against the non-canonical Wnt/PCP components fzd7 (C), vangl2 (D), wnt5b (E), and wnt11f2 (F). (G) Observed phenotypes in percent of embryos. (H-K) Reduced function of vangl2 and wnt5b results in reduced cd41:GFP-expressing thrombocytes at 6 dpf. (H-J) Lateral, inverted greyscale views of 6 dpf cd41:GFP-transgenic embryos confocal-imaged for GFP fluorescence, inserts depict brightfield view of imaged embryos and overall phenotype reference. Note how suboptimal dosing retains longer viability and less morphological impact (compare to D,E). (K) Thrombocyte quantification at 6 dpf reveals mild, yet dose-dependent reduction in cell number upon vangl2 and wnt5b perturbation. Individual datapoints (total number of cd41:GFP- positive cells per embryo) shown with mean and standard deviation, significance calculated by Mann-Whitney test: 6 dpf (K) wild type vs. control MO p=0.4981 (not significant), wild type vs. MO2-wnt5b 1:10 p=0.0407, wild type vs. MO2-wnt5b 1:15 p=0.2364 (not significant), wild type vs. MO1-vangl2 1:5 p=0.0533 (not significant), wild type vs. MO1-vangl2 1:10 p=0.3828 (not significant) (see Supplementary Data for 11 for details). Scale bar in (A): 100 μm, applies to panels A-F; (H): 150 μm, applies to panels H-J.

To deduce if reducing PCP signaling has a downstream impact on thrombocyte formation, we revisited our cd41:EGFP-based thrombocyte counts (Fig. 2). We injected cohorts of cd41:EGFP embryos with two sub-optimal doses of morpholinos against vangl2 and wnt5b, respectively; injection of each resulted in viable larvae with shortened body axis while establishing seemingly normal blood circulation in the observed timeframe (Fig. 6H–K). Compared to 6 dpf wild type and control morpholino-injected larvae, both vangl2- and wnt5b-attenuated zebrafish harbored reduced numbers of cd41:EGFP-expressing thrombocytes with a clear dose response (Fig. 6K, Supplementary Data 2), and a significant decrease in MO2-wnt5b (1:10)-injected larvae.

Taken together, our injection-based genetic interaction series established that reduced levels of rbm8a sensitize zebrafish embryos to perturbations in components of the non-canonical Wnt/PCP pathway during early development, with consequences to later stages. These observations connect with our morphometric data (Fig. 4,5), indicating that rbm8a-mutant zebrafish show features of defective convergence and extension that also affect the hematopoietic progenitors among the LPM.

Perturbing rbm8a or vangl2 impacts hematopoietic gene expression

Following bilateral stripe emergence, scl:EGFP-positive cells in the converging drl:mCherry-positive LPM form hematopoietic, and endothelial fates in the trunk (Fig. 7A) ^20,82. We therefore sought to evaluate possible changes to hematopoietic and endothelial progenitor formation and patterning upon rbm8a and vangl2 perturbation, respectively. If attenuation of rbm8a results in PCP signaling defects that subsequently cause LPM migration and hematopoietic fate disruptions, we would also anticipate that interrupted PCP signaling in vangl2 mutants reveals comparable phenotypes. We performed mRNA in situ hybridization (ISH) on rbm8a^Δ5/Δ5, rbm8a^Δ3/Δ3, MO-rbm8aATG-injected, vangl2^vu67/vu67, and wild type control embryos with a panel of marker genes expressed in in hematopoietic and endothelial progenitors (sox7, kdrl, gata1, gfi1aa, gfi1b, and runx1) (Fig. 7B–H). In situ hybridization for the pan-trunk muscle marker myoD provided a reference across conditions and for overall embryo morphology (Fig. 7B). We stage-matched embryos to wild type and control MO-injected controls by somite number and overall morphology to ensure valid comparisons and quantifications.

Figure 7: Select changes in hematopoietic marker expression in rbm8a- and vangl2-mutant zebrafish embryos.

(A) Schematic of basic hematopoietic lineage and marker relationships. Created with Biorender.com (Subscription: Individual, Agreement number: IZ25WQERBS). AGM: aorta-gonadal-mesonephros area; ICM: intermediate cell mass; PBI: posterior blood island; CHT: caudal hematopoietic territory; VDA: ventral dorsal aorta. (B-F) Expression patterns for individual marker genes in wild type, morpholino-injected, rbm8a- and vangl2-mutant zebrafish embryos at stage-matched 18 somites. Schematic of zebrafish embryo in the last panel in B-H shows where marker expression is expected and analyzed. (B) myoD as marker for somitic muscles shows comparable expression across conditions, with vangl2 mutants depicting the short, deteriorating tail typical for these mutants (asterisk). (C) The early endothelial marker sox7 is reduced in the trunk upon rbm8a perturbation (arrowheads) and seemingly normal in vangl2 mutants (asterisk). (D) The endothelial marker kdrl is decreased similarly as sox7. (E) The erythroid marker gata1 is reduced in all rbm8a-perturbed embryo conditions (arrowheads) and only retained in posterior-most LPM cells of vangl2 mutants (asterisk). (F) The hematopoietic progenitor marker gfi1aa is reduced in all rbm8a-perturbed embryo conditions (arrowheads) and barely detectable in vangl2-mutant embryos (asterisk). (G) In contrast, gfi1b is unchanged in all conditions. (H) The hematopoietic progenitor marker runx1 in rbm8a morphants and mutants is reduced or absent in the trunk and posterior blood island (PBI) (embryo midline, future dorsal aorta, arrowheads); vangl2 mutants also show reduced trunk expression (arrowhead) yet retain PBI runx1 expression (asterisk). (I) mRNA in situ hybridization-based quantification n of gata1, gfi1aa, and gfi1b expression levels; individual data points (normalized signal intensity/embryo) shown with mean and standard deviation, significance calculated by Mann-Whitney test: gata1 ISH signal intensity in wild type vs. MO-rbm8aATG p<0.0001, gfi1aa ISH signal intensity in wild type vs. MO-rbm8aATG p=0,031, gfi1b ISH signal intensity in wild type vs. MO-rbm8aATG p=0.7727 (not significant). (J-M) mRNA in situ hybridization-based quantification of runx1 in select areas (aorta-gonadal-mesonephros area (AGM): embryo midline, future dorsal aorta), posterior blood island (PBI), neurons). (J) Representative scale of runx1 signal ranked from low to high in rbm8a wild type siblings. The numbers on the bottom represent the average scores for AGM, PBI and neurons, respectively. (K) Signal in AGM. (L) Signal in PBI. (M) Signal in dorsal neurons. Individual data points (average of 4 scores per embryo) shown with mean and standard deviation, significance calculated by Mann-Whitney test: runx1a signal intensity in AGM (K) in wild type vs. rbm8a^Δ5/+ p=0.0048, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, rbm8a^Δ5/+ vs. rbm8a^Δ5/Δ5 p=0.0009; PBI (L) in wild type vs. rbm8a^Δ5/+ p=0.0077, wild type vs. rbm8a^Δ5/Δ5 p<0.0001, rbm8a^Δ5/+ vs. rbm8a^Δ5/Δ5 p=0.0007; neurons (M) wild type vs. rbm8a^Δ5/+ p=0.1604 (not significant), wild type vs. rbm8a^Δ5/Δ5 p=0.005, rbm8a^Δ5/+ vs. rbm8a^Δ5/Δ5 p=0.0481. Violin plot in background depicts population of the not averaged scores. Numbers (n)=average/total (see Supplementary Data 12 for details). Scale bar in (B): 100 μm, applies to all panels in B-H.

Compared to wild type and control MO-injected embryos, we observed decreased and patchy expression of the early pan-endothelial cell markers sox7 (Fig. 7C) and kdrl (Fig. 7D) in the trunk of rbm8a^Δ5/Δ5 genotyped embryos (n=2/2 sox7, total n=14/48; n=2/8 kdrl, total n=22/93), as well as in MO-rbm8aATG-injected embryos (n=12/23 sox7, n=10/55 kdrl) ^19,104,105. In comparison, the anterior endothelium and endocardial progenitor fields showed variable reduction of sox7 and kdrl expression in MO-rbm8aATG-injected embryos (Fig. 7C,D). While still expressing sox7 and kdrl at overall comparable levels to wild type siblings, vangl2^vu76/vu67 embryos (n=6/12 sox7, n=2/10 kdrl) showed no discernible midline fusion of the bilateral kdrl-expressing progenitors in the trunk at 18 somite stage (ss), in line with perturbed convergence and extension and midline migration (Fig. 7C,D).

The erythroid progenitor marker gata1 at 18 ss reveals the emerging primitive red blood cells in the intermediate cell mass of the trunk (Fig. 7A,E) ^19,106. In both stage-matched, genotyped rbm8a^Δ5/Δ5 (n=6/11, total n=47/68) and MO-rbm8aATG-injected embryos (n=32/57), gata1 expression was significantly reduced or patchy compared to wild type or control MO-injected embryos by visual inspection (Fig. 7E) and in situ hybridization signal quantification (wild type n=25, MO-rbm8aATG-injected n=24) (Fig. 7I, Supplementary Data 12). Similarly, gata1 expression in vangl2^vu76/vu67 embryos remained in the posterior-most LPM, yet was greatly reduced in the trunk (n=7/17) (Fig. 7E).

The transcription factors GFI1 and GFI1B are key determinants of myeloid fate potential towards erythroid/megakaryocyte and lymphoid fates in mammals, and mutations in human GFI1B cause thrombocyte deficiencies ³⁷. In zebrafish, the orthologs gfi1aa and gfi1b influence primitive erythroid precursor formation and are expressed in intermediate and definitive precursors (Fig. 7A) ^40,41,107. Compared to wild type and control MO-injected embryos, both genotyped rbm8a^Δ5/Δ5 (n=4/6, total n=27/41) and MO-rbm8aATG-injected embryos (n=12/20) displayed significantly reduced gfi1aa expression, which we further verified by signal intensity comparison (wild type n=23, MO-rbm8aATG-injected n=21). (Fig. 7F,I). gfi1aa expression was also markedly reduced in vangl2^vu76/vu67 embryos compared to wild type siblings (n=16/27) (Fig. 7F). In contrast, gfi1b expression showed no significant quantitative differences (Fig. 7I), yet we observed individual genotyped rbm8a^Δ5/Δ5 embryos with patchy gfi1b expression (n=3/6, total 14/45) (Fig. 7G).

Lastly, runx1 expression that broadly marks prospective hematopoietic progenitors including future stem cell precursors in the trunk at 18 ss (Fig. 7A) ¹⁰⁸ was reduced in genotyped rbm8a^Δ5/Δ5 (n=16/19) and MO-rbm8aATG-injected embryos (n=26/27) (Fig. 7H). We further verified our observations by signal intensity comparison of runx1-positive cells in the aorta-gonadal-mesonephros area (AGM) and posterior blood island (PBI), which showed a significant decrease in runx1-positive cells in those areas in both rbm8a^Δ5/+ and rbm8a^Δ5/Δ5 mutant embryos compared to wild type siblings (Fig. 7J–L). Notably, the spinal cord neurons and other domains expressing runx1 seemed unaffected in number (Fig. 7F), a hallmark of mutants with selective hematopoietic defects ¹⁰⁹, but showed slightly higher expression levels in our quantitative signal intensity comparison (Fig. 7M, Supplementary Data 12). Although vangl2^vu67/vu67 embryos maintained runx1 expression in the eye, ear, and dorsal spinal cord neurons, they displayed selectively reduced or absent runx1 expression in the presumptive hematopoietic stem cell progenitors at the midline as well as in the PBI (n=6/8) (Fig. 7H). This impaired runx1 expression in both rbm8a^Δ5/Δ5 and vangl2^vu76/vu67 embryos is unlikely to be caused by developmental delay or the severely perturbed midline migration in vangl2^vu76/vu67 embryos, as runx1 expression is stronger in earlier bilateral LPM progenitors before confining to hematopoietic precursors in the trunk (Fig. 7A,H) ^108,109.

While only including a limited sampling of perturbations and timepoints, together these data document select changes in hematopoietic progenitor gene expression upon rbm8a and vangl2 defects. Our data proposes a model in which attenuated Rbm8a levels and concomitant EJC reduction result in retention of mis-spliced mRNAs encoding non-canonical Wnt/PCP genes (among others), lowering the effective protein concentration of several PCP pathway components; this results in sub-optimal convergence and extension and cell migration including LPM progenitors, with subsequent impact on hematopoietic progenitor numbers and/or gene expression. In line with this model, zygotic rbm8a-mutant and -knockdown zebrafish feature select and significant blood and endothelial lineage defects already prior to the onset of heartbeat and circulation.

DISCUSSION

The striking phenotype combination of TAR syndrome – affecting selective blood and limb skeleton features as well as heart, kidney, and craniofacial structures – has remained challenging to consolidate with a unifying disease mechanism. The identification of hypomorphic defects in the ubiquitous EJC factor Rbm8a in TAR syndrome patients has further complicated possible explanations of how the selective phenotypes arise. Investigating the developmental impact following different means of reducing Rbm8a protein levels in zebrafish, we propose that TAR syndrome involves early embryonic perturbations in LPM formation and patterning via attenuated PCP signaling. Our work links sensitizing reduction of pathways essential for embryo morphogenesis to retained introns in individual mRNAs upon Rbm8a reduction, proposing a developmental mechanism to consolidate the seemingly divergent phenotypes of TAR syndrome as LPM-affecting birth defect.

Disruption of a molecular mechanism that is shared in the development of diverse cell fates by means of clonal cell relationship, evolutionary co-option, or signaling interactions between progenitors can lead to complex syndromic phenotypes. Mutations in pleiotropic factors such as ribosomal or DNA repair components have been linked to hematological disorders with additional structural defects that include the limbs as observed in Diamond-Blackfan Anemia and Fanconi’s Anemia, respectively ^110–115. Curiously, distinct rare congenital cases with combined amegakaryocytic thrombocytopenia and radial-ulnar anomalies have been reported (CTRUS or RUSAT, OMIM # 605432) ¹¹⁶. Mutations in HOXA11 have been linked to thrombocytopenia with concurrent proximal fusion of radius and ulna in affected patients ¹¹⁷, and Hox gene mutations are hypothesized to underlie similar phenotypic manifestations in rare cases ¹¹³. Individual patients with initial platelet deficiency that improves with age and deficient pronation-supination of the forearm carry mutations in MECOM1 encoding the transcription factor EVI-1¹¹⁸. Both Hoxa11 and Evi-1 have been associated with hematological, forelimb, and other LPM-linked disorders ^119–128. Independent of underlying molecular mechanism, establishing a developmental connection between affected cell types in syndromic congenital disease provides a potent framework to expand diagnosis, phenotype assessment, and long-term care of affected patients. The multilineage potential of early LPM provides a developmental concept to connect seemingly disparate syndromic phenotypes as outlined above and to potentially regard them as LPM diseases ²¹.

rbm8a perturbation in zebrafish has increasingly pleiotropic defects as described here and in previous work ¹⁴, in line with Rbm8a protein function in the ubiquitously deployed EJC that regulates diverse mRNA biology ^9,12,78,129. Elegant genetic experiments have documented how the 3’ UTR intron-containing foxo3b mRNA that encodes a pro-apoptotic transcription factor is stabilized in rbm8a-mutant embryos, while double rbm8a;foxo3b mutants showed a significant rescue of the motor neuron defects ¹⁴. Not surprisingly, recessive null combinations of human RBM8A mutations have never been reported and are likely not viable; all genetically investigated TAR patients reported to date harbor at least one hypomorphic allele of RBM8A, suggesting reduced EJC levels or activity result in the combination of syndromic phenotypes ^1,5,6,130. The combination of waning maternal contribution, morpholino-based dose attenuation, and our hypomorphic rbm8a^Δ3 allele in zebrafish provides in vivo means to approximate the assumed hypomorphic reduction of RBM8A levels in human TAR patients. In line with morpholino guidelines ⁶⁸, MO-rbm8aATG injection morphologically phenocopy rbm8a^Δ5/Δ5, reduces Rbm8a protein levels, is rescued with human RBM8A mRNA while expression lasts, and can be titrated for genetic interactions (Fig. 1,2,6, Supplementary Fig. 1,2).

Our observations that different PCP perturbations result in select hematopoietic defects (Fig. 3,6) provide a first mechanistic direction to consolidate the TAR defects. Our broad mRNA ISH panel for several developmental markers documented that rbm8a-mutant embryos do not feature broad LPM anomalies (Supplementary Fig. 2), and do not show the full complement of reported TAR phenotypes early in development ^2,5,6. Notably, the typical limb and joint anomalies associated with the syndrome ^1,2 cannot be fully modeled in zebrafish due to the evolutionary differences in extremity formation, limiting the use of zebrafish to model TAR Syndrome beyond basic developmental mechanisms. Nonetheless, our cell counting of cd41:EGFP-positive thrombocytes using two different methods including our new whole-larvae imaging (Fig. 2, Supplementary Fig. 2) provides evidence that rbm8a perturbation in zebrafish reduces overall thrombocyte numbers, while their ability to move, adhere and aggregate at sites of endothelial injury is not significantly perturbed (Supplementary Fig. 2). We further do not see evidence for disrupted endothelium mechanisms in our rbm8a mutants, as our ISH using fli1a shows no difference in endothelium/blood vessels (Supplementary Fig. 2).

Together, our data indicates that perturbed rbm8a affects the regulation of thrombocyte numbers and not their functionality, as appears to be the case for TAR syndrome patients ^2,5. While TAR cases are rare and severe thrombocytopenia precludes standard platelet aggregometry, limited previous work has also found no overt functional defects in thrombocytes of TAR patients ¹³¹.

Convergence and extension in an embryo influences nearly all tissues and overall body architecture ^60,89,132. Our morphometric analysis of 6–11 somite stage rbm8a-deficient embryos showed that rbm8a^Δ5/Δ5 mutant and rbm8a morphant embryos displayed phenotypes consistent with convergent and extension phenotypes (Fig. 5), even though the phenotypic characteristics were milder. The somites in classic convergent and extension mutants, such as tri/vangl2, amongst others, are generally wider; for example, in 7 somite stage tri/vangl2 mutants, the somites are 1.27x wider than those in stage-matched wild type embryos ^90,133–136. At the same timepoint, we observed that somites in rbm8a^Δ5/Δ5 mutants are 1.15x wider (Supplementary Data 10). Nonetheless, this paraxial mesoderm phenotype seems to be further amplified into the adjacent LPM (Fig. 4), which undergoes dramatic cellular and structural changes during convergence and extension movements, resulting in condensed bilateral stripes at the periphery of the embryo. The subsequent midline migration of hematopoietic, endothelial, and kidney progenitors is sensitive to perturbations of guidance cues, yet the exact mechanism(s) that trigger and drive midline migration remain incompletely resolved. PCP signaling defects prevent condensation of cells across all germ layers along the anterior-posterior axis, keeping especially the LPM as lateral-most cells widely dispersed ^{55,60,89,91,137,138}.

How could attenuation of Wnt/PCP signaling influence hematopoietic progenitors? One possible explanation could be that as hematopoietic progenitors migrate as medial-most LPM stripe and are exposed to different signals and cellular interactions during their migration ²⁰, attenuated PCP could interrupt hematopoietic progenitor arrangement and migration speed. This model is in line with our observations of altered scl:EGFP-expressing LPM in rbm8a and vangl2 perturbations (Fig. 4). However, in both rbm8a and vangl2 mutants hematopoietic progenitors do still form, albeit with altered expression of individual regulatory genes (Fig. 7). During their medial migration, LPM cells are in constant contact with endoderm and overlaying somites that have been implicated in influencing hematopoietic progenitor patterning and fate potential ^{62,63,139,140}. Non-canonical Wnt signaling has been implicated in controlling somite-expressed Notch ligands that are required for hematopoietic stem cell progenitor emergence ⁶². While we do see a reduction of runx1-positive progenitors (Fig. 7), they are perturbed already before runx1 expression refines to HSPCs in the dorsal aorta after 24–26 hpf, suggesting an earlier influence on hematopoietic progenitors following attenuated Wnt/PCP in rbm8a-deficient embryos. Nonetheless, the involvement of non-canonical Wnt/PCP signaling at different steps of hematopoiesis might contribute to the selective thrombocyte anomalies observed in TAR patients, warranting further experimental follow-up.

We therefore hypothesize that, following Rbm8a perturbations, attenuated PCP through retention of mis-spliced mRNAs encoding non-canonical Wnt/PCP components affects migration or relative position of hematopoietic progenitors within the embryo as they migrate through signaling gradients. Notably, mice with defects in the EJC component Srsf3 also display thrombocytopenia ¹⁴¹. The progressive recovery of the thrombocytopenia phenotype in TAR patients is compatible with an early defect in embryonic or fetal hematopoiesis when transient or intermediate progenitor waves are forming; the high plasticity and clonal expansion even of potentially reduced hematopoietic stem cells could subsequently replenish the diminished megakaryocyte levels during definitive hematopoiesis ^{1,2,25,142–145}. The reduced gfi1aa expression (Fig. 7) and later reduced numbers of cd41:EGFP-expressing thrombocytes in our quantifications following rbm8a and PCP perturbations show signs of compensation over time (Fig. 2,6), providing an experimental starting point to pursue this model.

Connecting morphological alterations by attenuating non-canonical Wnt/PCP signaling as major pathway involved in embryo morphogenesis provides a possible mechanistic link to the impact on LPM-derived structures in TAR Syndrome and possibly other congenital conditions. While we here focus on consequences of perturbed rbm8a on hematopoiesis, non-canonical Wnt/PCP disruptions have the potential to affect a variety of tissues ⁴⁹. PCP signaling does not elicit an obvious transcriptional response, confining effects of its perturbation to morphological defects; however, we hypothesize that incorrect or delayed migration or altered arrangement of LPM progenitors could manifest as defects in individual downstream fates by altering the spatio-temporal dynamics of exposure to developmental signals and cell-cell interactions. Such a model would be consistent with recent observations linking emerging cell fates to overall morphogenesis that is a prerequisite for the correct formation of patterning gradients ¹⁴⁶. In addition to zebrafish not forming a radius bone, future work is warranted to decode the basic impact on heart, forelimbs (pectoral fin), kidney, and other LPM-derived structures upon rbm8a perturbation. PCP defects have been linked to anomalies in each of these structures in mice ^{49,52,54,64,103,147–150}, further suggesting defective PCP signaling in LPM patterning as corroborating developmental link for the TAR syndrome phenotypes.

MATERIALS AND METHODS

All research described herein complies with all relevant ethical regulations as reviewed and approved by the University of Colorado Anschutz Medical Campus, the University of Zurich, the MDC Berlin, and Baylor College of Medicine.

Zebrafish husbandry

Zebrafish (Danio rerio) husbandry and experiments were performed according to IACUC regulations at the University of Colorado Anschutz Medical Campus (protocol number 00979) and the University of Michigan (protocol number PRO00010679), FELASA guidelines ¹⁵¹, the guidelines of the Max-Delbrück Center for Molecular Medicine in the Helmholtz association, and the local authority for animal protection (Landesamt für Gesundheit und Soziales, Berlin, Germany). The ‘Principles of Laboratory Animal Care’ (NIH publication no. 86–23, revised 1985) as well as the current version of German Law on the Protection of Animals and EU directive 2010/63/EU on the protection of animals were followed. All zebrafish were raised, kept, and handled according to established protocols if not noted otherwise ¹⁵².

Transgenic zebrafish lines

The following established transgenic lines were used in this study: Tg(scl:EGFP)^{zf255 82}, Tg(cd41:EGFP) ⁶⁹, and Tg(RGB) ¹⁵³ (combined transgene insertions from Tg(−6.35drl:EGFP)^{cz3331 102}, Tg(lmo2:loxP-dsRED2-loxP-EGFP)^{rj2 154}, and Tg(myl7-loxP-AmCyan-loxP_ZsVenus)^{fb2 155}.

sgRNA template generation, RNP complex assembly and injection

Oligos were obtained from Sigma/Millipore and Life Technologies as standard primers except where otherwise noted. A short guide RNA (sgRNA) template was generated by oligo-based PCR amplification (Bassett et al., 2013) using the sgRNA forward primer containing a T7 promoter sequence 5’-GAAATTAATACGACTCACTATAGGGAGGCGAAGACTTTCCTAGTTTTAGAGCTAGAAATAGC-3’ and the invariant reverse primer 5’-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3’ (PAGE-purified). A primer extension reaction was performed using Phusion DNA polymerase (M0530L, NEB) followed by QIAquick purification (28106, Qiagen) with elution of the PCR product in DEPC-treated water. In vitro transcription (IVT) of the sgRNA-based PCR template was performed using the MAXIscript T7 Kit (AM132, Ambion) with an overnight reaction run at 37°C, followed by ammonium acetate precipitation or clean up via RNA isolation kits as per the manufacturer’s instructions and as described previously ^156,157. The only change to the manufacturer’s protocol of the MAXIscript T7 Kit was to add 100mM NTPs instead of the recommended 10 mM as this greatly increases sgRNA yield. Before use, the sgRNA was quality controlled on a denaturing 2.5% MOPS gel. The CRISPR-Cas9 ribonucleoprotein complex (RNPs) was assembled using appropriate salt concentrations to keep the RNPs solubilized (Burger et al., 2016). Salt-solubilized RNPs were then injected at sub-optimal concentrations to increase viability and founder identification (Burger et al., 2016).

Mutant zebrafish lines

rbm8a mutant alleles were generated as per our previous work ¹⁵⁸ using the target site 5’-GGGAGGCGAAGACTTTCCTA-5’ as proposed by CHOPCHOP ¹⁵⁹. Details about sgRNA template generation and RNP complex assembly and injection are described in the Supplemental Data. The two alleles rbm8a^Δ5 (also described in Gangras et al., 2020) and rbm8a^Δ3 were selected. Both alleles were genotyped by sequencing PCR products using primer rbm8a fwd 5’-AAACAGCAGACGGCGAGG-3’ and primer rbm8a rev 5’-GCTGAATCACTACAACGCG-3’. Additionally, the rbm8a wild type allele was genotyped by restriction digest of the PCR product with Xmn1 (R0194S, NEB) and the rbm8a^Δ3 allele by Hinf1 (R0155S, NEB) while the rbm8a^Δ5 allele stayed intact upon Xmn1 or Hinf1 digest.

Further, the established vangl2^vu67 mutant was used⁸⁸ (a kind gift from Dr. Liliana Solnica-Krezel). Embryos were genotyped by sequencing PCR products using primer vangl2 fwd 5’-ATTCCCTGGAGCCCTGCGGGAC-3’ and primer vangl2 rev 5’-AGCGCGTCCACCAGCGACACAGC-3’ or restriction digest of the PCR products with Alu1 (R0137S, NEB). The vangl2 wild type allele stayed intact while the vangl2^vu67 mutant allele was identified by a digested PCR product.

Morpholino injections and rescue experiments with capped mRNA

Capped mRNA for mutant rescue was generated using the mMESSAGE mMACHINE^™ SP6 Transcription Kit (AM1340, Ambion) as per the manufacturer’s instructions. As templates, pCS2+:EGFP¹⁶⁰ and pCS2FA:hs_RBM8A were used. To generate the original pCS2FA:hs_RBM8A, the human RBM8A ORF was amplified from HEK293T cDNA by PCR with primer fwd 5’-TTAGGATCCATGGCGGACGTGCTAGATC-3’ containing a BamHI site (underlined) and primer rev 5’-TTGCTAGCTCAGCGACGTCTCCGGTCT-3’ containing a NheI site (underlined). The PCR product was cloned with BamHI (R0136S, NEB), NheI (R3131S, NEB) restriction digest into Tol2kit vector 396 (pCS2FA:Tol2)¹⁶¹ cut with BamHI and XbaI (R0145S, NEB).

Morpholinos were obtained from Gene Tools (Philomath, USA) and prepared as a 1 mM stock solution according to the manufacturer’s instructions. To target rbm8a, a translation blocking morpholino, MO-rbm8aATG with sequence 5’-GAAGATCCAGTACGTCCGCCATGA-3’ (sequence complementary to predicted start codon is underlined), was designed and diluted as indicated. To achieve full rbm8a knockdown a dilution of 1:100 was used, and is referred to as optimal dilution. To achieve suboptimal rbm8a knockdown, dilutions of 1:125, 1:150, 1:175, 1:200, and 1:500 were used. Of note, optimal and suboptimal dilutions differed between labs, and are indicated in the respective figures. All other morpholinos used in this study have been published previously and morpholino details and injection procedures are described below.

Rescue experiments were performed by injecting MO-rbm8aATG (1:100 dilution) and 100 ng/μL capped hs_RBM8A(altered) mRNA and EGFP mRNA as control. As the codons encoding the N-terminus of zebrafish and human RBM8A protein are identical in the respective mRNAs and therefore also human RBM8A mRNA is recognized by the MO-rbm8aATG, 5’ codons in the morpholino-targeted region were altered to synonymous codons where possible: wild type human RBM8A 5’-ATGGCGGACGTGCTAGATCTTCACGAGGCTGGG-3’ to 5’-atggcAgaTgtCTtGgaCTtGCAC-3’ (alterations in capital letters, resulting in synonymous codons, ATG start codon underlined in both sequences). The resulting RBM8A(altered) ORF was ordered from Twist Biosciences and swapped into pCS2FA:hs_RBM8A with flanking BamHI and NheI sites as above, resulting in pCS2FA:hs_RBM8A(altered). mRNA was generated and injected as outlined above.

The embryos were observed for phenotypes at 24, 48 and 72 hpf and representative images were taken with a 10x lens on a Zeiss LSM880. Observed phenotypes were plotted in percent using GraphPad Prism (10.0.3).

Additional morpholinos and injection procedures

MO1-vangl2 (ZFIN ID: ZDB-MRPHLNO-041217–5, 5’-GTACTGCGACTCGTTATCCATGTC-3’), MO2-wnt5b (ZFIN ID: ZDB-MRPHLNO-051207–1, 5’-TGTTTATTTCCTCACCATTCTTCCG-3’), MO2-wnt11f2 (formerly known as MO2-wnt11 or wnt11-MO, ZFIN ID: ZDB-MRPHLNO-050318–3, 5’-GTTCCTGTATTCTGTCATGTCGCTC-3’) and MO1-fzd7a (ZFIN ID: ZDB-MRPHLNO-050923–5, 5’-ATAAACCAACAAAAACCTCCTCGTC-3’) which we have described previously and validated^{52,68,162–165}, have been used at indicated dilution. Lastly, the Gene Tools pre-made standard control oligo morpholino (100 nmol; diluted to 1:200) was used.

Prior to injections, the 1mM morpholino stock solution was heated up to 65 °C, vortexed for 5 minutes and diluted to achieve the desired concentrations. The diluted morpholino solution was again heated to 65 °C and vortexed for 10 minutes before injecting. All morpholino solutions were stored at room temperature. Microinjections were performed using MPPI-3 microinjectors (Applied Scientific Instrumentation) together with a Narishige micromanipulator (MN-153). Morpholino oligonucleotides were injected into the yolk of one-cell stage zebrafish embryos. The injection volume was calibrated to the final volume of 1 nL in total. In case of double knockdown approaches to study genetic interactions, both morpholinos were injected sequentially. Embryos were kept in E3 embryo medium under standard laboratory conditions at 28.5 °C.

Zebrafish protein sample collection

Zebrafish embryos were collected at indicated developmental stages (tailbud stage, 10 somite stage, 16 somite stage, 24 hpf). Individual embryos were transferred into tubes kept on ice filled with 200 μL cell dissociation solution (CDS; 5 ml of Ringers’ solution (96724, Sigma), 1 tablet of Complete Mini EDTA-free (11836153001, Roche). The embryos in CDS were flash frozen in liquid nitrogen and samples were stored at −80 °C until they were genotyped and could be combined according to their genotype for further usage. For genotyping, a few cells from individual embryos were collected and genotyped as described above.

After combining 10 embryos with the same genotype for each developmental stage, the samples were centrifuged at 1500 rpm for 2 minutes at 4 °C. The tubes were turned by 180° and centrifuged for 2 additional minutes. The supernatant was removed, 200 μL CDS was added, and centrifuged again as above. Afterwards, the supernatant was again removed, and the samples centrifuged once more with additional removal of the remaining supernatant using fine gel loading tips. Then, 20 μL 2x SDS Laemmli buffer (1610747, BioRad) supplemented with 5% 2-betamercaptaethanol (M3148, Sigma) was added to each sample, vortexed rigorously, and cooked for 5 min at 95 °C.

Electrophoresis and immunoblot analysis

Samples were loaded on a precast Mini-PROTEAN TGX gel (4569034, Biorad) and electrophoresis was performed as per the manufacturer’s instructions. Afterwards, the gel was transferred onto a PVDF membrane (1704272, Biorad) using the trans-blot turbo transfer system (1704150, Biorad).

Zebrafish Rbm8a and Tubulin proteins were detected using standard immunodetection protocols, in short: blocking with 1XTTBS 5% BSA for 1 hour, overnight incubation with anti-Rbm8a (1:1000; 05–1511, Millipore), anti-Tubulin (1:1000; 05–829, Millipore) in 1XTTBS 5% BSA at 4 °C, followed by 3 × 5 min wash in TTBS and 30 min incubation with IRDye 800CW goat anti-Mouse (1:10000; 926–32210, Li-Cor) in 1XTTBS 5% BSA at room temperature followed by 3 × 5 min wash in 1XTTBS. Images of Western blots were obtained using an Odyssey CLx imager (Li-Cor) and quantified using ImageJ (https://www.yorku.ca/yisheng/Internal/Protocols/ImageJ.pdf). All full gel images are shown Supplementary Fig. 1.

cd41:EGFP-positive cell counting

For live imaging of 3 and 6 dpf cd41:EGFP zebrafish larvae, the heartbeat was stopped with 30 mM 2,3-butanedione monoxime (BDM; B0753, Sigma) in E3, in addition to anesthesia with Tricaine. Images of the zebrafish larvae used for thrombocyte counting were taken on a Leica M205 FA with Plan Apo 1.0x lens. 3 dpf larvae were imaged directly in E3+BDM+Tricaine on a glass plate, while the 6 dpf larvae were placed on a petri dish covered with a 2% agarose layer. Larvae were placed anterior to the left. Images were taken consistently at 2x zoom for groups of two-four zebrafish larvae at 3 dpf and 3x zoom for a single zebrafish larvae at 6 dpf. For analysis, the images were exported from LAS Leica, and using Fiji/ImageJ macro ¹⁶⁶, the GFP channels were separated. At 3 dpf, the cells were counted manually by adjusting brightness and contrast as needed, using the Fiji/ImageJ cell counter plugin (https://imagej.nih.gov/ij/plugins/cell-counter.html). At 6 dpf, a second Fiji/ImageJ macro was run that thresholds and automatically counted the cells. Representative images of 6 dpf larvae for figure panels were taken on a Zeiss LSM880 with 10x lens. In 6 dpf larvae, Fiji/ImageJ counts were manually recounted for verification. Individual cd41:EGFP values, means, and ± S.D. were plotted using GraphPad Prism (10.0.3).

Macros to run on imaging files in an individual folder:

##Macro 1##
           // ask user to select a folder
           dir = getDirectory(“Select a folder”);
           // prepare a folder to output the images
           output_dir = dir + File.separator + “ output” + File.separator;
           File.makeDirectory(output_dir);
           list = getFileList(dir);
           for (i=0; i<list.length; i++) {
           showProgress(i+1, list.length);
           open(dir+list[i]);
           }
           for (i=0; i<200; i++) {
               imageTitle=getTitle();
	   selectWindow(imageTitle);
           run(“Split Channels”);
           selectWindow(“C3-”+ imageTitle);
           close();
           selectWindow(“C1-”+ imageTitle);
           close();
           selectWindow(“C2-”+ imageTitle);
           run(“Subtract Background...”, “rolling=10”);
           saveAs(“tiff”, output_dir+imageTitle);
           close();
           }
##Macro 2##
           // ask user to select a folder
           dir = getDirectory(“Select A folder”);
           list = getFileList(dir);
           for (i=0; i<list.length; i++) {
           showProgress(i+1, list.length);
           open(dir+list[i]);
           }
           for (i=0; i<200; i++) {
                imageTitle=getTitle();
                selectWindow(imageTitle);
                run(“16-bit”);
                setAutoThreshold(“Default dark”);
                run(“Threshold...”);
                waitForUser(“done”);
                setOption(“BlackBackground”, false);
                run(“Convert to Mask”);
                run(“Watershed”);
                setTool(“polygon”);
                run(“Analyze Particles...”, “ show=Outlines display clear summarize add”);
                setTool(“hand”);
                   close();
                   close();
           }

Upon completion, toggle the top bar until all cd41:EGFP-bright cells are covered, press “ok” in dialog window. Resulting data was processed in Excel for next steps.

Circulating thrombocyte video quantitation and functional analysis

Circulating cd41:EGFP-positive thrombocytes were measured as previously described^73,74. 6 dpf larvae cd41:EGFP were anaesthetized in tricaine and submersed in a solution of 0.7% low-melt agarose. 4 to 6 larvae were aspirated into 1.5–1.8 mm diameter capillary tubes (Pyrex) using a manual pipette pump, each separated by an air bubble. The capillary was mounted on modeling clay, submerged in 1x E3 medium in a 10 cm Petri dish, and then placed under a stereomicroscope (Leica) with an Olympus DP22 2.8-megapixel digital camera. Magnification was set at 85x and a lateral view of the anterior tip of the larva to the end of yolk sac extension was visualized. Fluorescent videos were captured for a duration of one minute and then analyzed using VirtualDub (v1.10.4) and Fiji ¹⁶⁶. The first 1000 frames were quantified, averaged, and then background subtracted to eliminate non-circulating cells. Finally, zebrafish larvae were removed from the capillary, lysed, and genotyped.

Laser-mediated arterial endothelial injury was performed using a pulsed-dye laser system (Micropoint, Andor Technology), as previously described^74,75. 6 dpf zebrafish larvae carrying the cd41:EGFP transgene were anesthetized in tricaine, positioned laterally in 0.8% low-melt agarose on glass cover slips and visualized on an inverted microscope (Olympus 1X73) using a 20x objective. The endothelium of the dorsal aorta was ablated at the fifth somite posterior to the anal pore. The time to occlusion was recorded up to 120 seconds and the number of cd41:EGFP-positive thrombocytes adhering at the site of injury during that period was recorded. After laser injury, zebrafish larvae were recovered from the agarose, lysed, and genotyped. Individual values, means, and ± S.D. were plotted using GraphPad Prism (10.0.3).

Bulk RNA sequencing data analysis

RNA sequencing reads were initially quality checked with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and MultiQC¹⁶⁷ and deemed of high quality. We quantified transcript-level expression using Salmon (version 0.8.2) ¹⁶⁸ against the Ensembl GRCz10.87 catalog ¹⁶⁹ and performed gene-level differential expression using tximport (version 1.4.0) ¹⁷⁰ and edgeR (version 3.18.1) ¹⁷¹ using the quasi-likelihood F test ¹⁷² and a Benjamini-Hochberg multiple testing correction ¹⁷³; thresholds were set at FDR <= 0.05. Volcano plots were produced using the ggplot2 (version 2.2.1) package. RNA-seq reads were also mapped against the GRCz10 reference genome with STAR (version 2.5.1b_modified) ¹⁷⁴, using the Ensembl GRCz10.87 catalog. Coverage tracks (in bigWig format) were created with bedtools (version 2.17.0)¹⁷⁵ and bedGraphToBigWig (version 4)¹⁷⁶ and plotted with Gviz (version 1.20.0) ¹⁷⁷. Differential exon/intron usage analysis was performed using DEXSeq (version 1.22.0) ⁷⁹. Transcripts contributing less than 5% of the abundance for a given gene in all samples were removed from the gtf file ¹⁷⁸ and a gff file with disjoint exon bins were generated from the remaining transcripts using DEXSeq. Finally, introns (regions of a gene locus not belonging to any annotated transcript) were explicitly added as separate bins in the gff file before quantification and differential exon/intron usage analysis with DEXSeq. To explore the RNA-seq data interactively, we created the R/Shiny-based RNAseq Explorer app (link: http://imlspenticton.uzh.ch:3838/mosimann_p2452/). The app contains tabs for various visualizations and analyses: individual gene expression levels (by typing the gene name), a PCA plot for an overall view of the similarity of gene expression profiles, normalized base-level coverage tracks for individual genes, differential expression results, differential transcript usage results, and gene set enrichment analyses.

Light sheet sample preparation, microscopy, and image analysis

The Zeiss Z.1 microscope equipped with a Zeiss W Plan-Apochromat 20×/0.5 NA objective was used for all light sheet microscopy. Embryos within the chorion were embedded in 1% LMA in a 50 μL glass capillary. For the multi-angle imaging data sets, the four individual angles per embryo were manually registered before applying the Fiji Multiview Reconstruction and BigStitcher plugins ^166,179 for fusion and deconvolution of the images. Images and movies were further processed using Fiji/ImageJ and Imaris (9.7)¹⁸⁰.

Morpholino-injected embryos were raised at 28 °C and imaged in the Z.1 from 4 somite stage to 12 somite stage, alongside uninjected scl:EGFP controls. In Imaris, the baseline and background were subtracted and surface rendering was performed to automate volume and area determination of the fluorescent signal.

Quantifications of area and volume from wild type, MO-rbm8aATG and MO1-vangl2-injected embryos were averaged for each timepoint. Average values and ± S.D. were plotted using GraphPad Prism (10.0.3).

Whole mount in situ hybridization

Whole-mount in situ hybridization (WISH) was performed according to the Thisse lab protocol ¹⁸¹. Zebrafish embryos were fixed overnight in 4% PFA (47608, Sigma) in PBS (EC-A500, Life Sciences), dechorionated, rinsed in PBS and dehydrated into methanol. Samples were then rehydrated into PBT (PBS + Tween (P1379, Sigma), treated with Proteinase K (1:10000) (P8108S, NEB) for 2 min, washed in PBT and postfixed in 4% PFA/0.2% glutaraldehyde (G5882, Sigma) in PBS for 20 min at RT. Then the samples were incubated for at least two hours in hybridization solution with 50% formamide (in 0.75 M sodium chloride, 75 mM sodium citrate, 0.1% Tween-20 (P1379, Sigma), 50 mg/mL heparin (H3393, Sigma), and 200 mg/mL tRNA (AM7119, Invitrogen)) at 70°C, then hybridized overnight at 70°C with antisense probes diluted approximately 1 ng/ml in hybridization solution. Samples were washed gradually into 2X SSC buffer (0.3 M sodium chloride, 30 mM sodium citrate or AM9763, Invitrogen), and then gradually from SSC to PBT. Samples were blocked at room temperature for several hours in PBT with 2% goat or sheep serum (S2263, Sigma) and 2 mg/mL bovine serum albumin (BSA, 196941, Fisher), then incubated overnight at 4°C with anti-DIG antibody (11093274910, Roche) at 1:5000 in blocking buffer. Samples were rinsed extensively in PBT and staining buffer (100 mM Tris-HCl pH 9.5, 50 mM MgCl2, 100 mM NaCl, 0.1% Tween 20, 0.3mg/mL tetramizole hydrochloride (T1512, Sigma) in dH20), prior to staining with BM Purple AP staining solution (62321800, Roche). Staining was stopped by washing embryos in PBT.

Antisense riboprobes for in situ hybridization

The templates for all in situ hybridization (ISH) probes were generated by PCR amplification using primers and complementary DNA (cDNA) or linearized plasmids. First-strand cDNA was generated from pooled zebrafish RNA isolated from different developmental stages using SuperScript III first-strand synthesis kit (18080051, Invitrogen). Subsequently, DNA templates for ISH probes were generated using first-strand cDNA as a PCR template and primers as specified for each gene of interest; for in vitro transcription (IVT), either the T7 (5′-TAATACGACTCACTATAG-3′) or T3 (5′-AATTAACCCTCACTAAAG-3′) promoter sequence was added to the 5′ ends of the reverse primers. PCR reactions were performed under standard conditions as per manufacturer’s protocol using Phusion DNA polymerase.

The following primer or plasmids to generate templates for ISH probes were used: kdrl (using pBS-kdrl (a kind gift from Dr. Leonard I. Zon), linearized with EcoRI (R0101S, NEB), T7 polymerase for IVT)¹⁸²; sox7 (using primer fwd 5’-CGACCAAAACTCCCTTCCG-3’ and primer rev 5’-AAAATAATACGACTCACTATAGGGTTGTTGTAGTAGGCTGC-3’, T7 polymerase for IVT), gata1 (using primer fwd 5’-TGGGAAAGACAGTCCCAGG-3’ and primer rev 5’-AAAAATAATACGACTCACTATAGGGCCTTCACACTAGTGTGGG-3’, T7 polymerase for IVT); runx1 (using pBS-runx1a, linearized with HindIII (R0104S, NEB), T7 polymerase for IVT)¹⁸³ (a kind gift from Dr. Teresa V. Bowman); gfi1aa (using primer fwd 5’-TTATCATCAGCCCCGTTACC-3’ and primer rev 5’-AAAATAATACGACTCACTATAGGGAATGGACGGCTTTATGTTGC-3’, T7 polymerase for IVT)⁴²; gfi1b (using primer fwd 5’-ACCAACCTCAAACGAGAGC-3’ and primer rev 5’-AAAATAATACGACTCACTATAGGGATTGTCCATCAACTTCTGTC-3’, T7 polymerase for IVT) ⁴², myoD (linearized with BamHI, T7 polymerase for IVT) ⁹⁹, dlx3b (in pBS, linearized with SalI, T7 polymerase for IVT) ¹⁰⁰, tbxta (in pBS-KS, linearized with XhoI, T7 polymerase for IVT ) ⁹⁷, and hgg1 (linearized with EcoRI, T7 polymerase for IVT) ¹⁰¹.

Antisense ISH probes were transcribed from their template overnight at 37 °C using either MAXIscript T7 Kit (AM132, Ambion) or MAXIscript T3 Kit (AM1316, Ambion) and digoxigenin (DIG)-labeled NTPs (11277073910, Roche) as per the manufacturer’s instructions. The resulting RNA probe was treated with TURBO DNase (AM2238, Invitrogen), precipitated with lithium chloride (9480G, Invitrogen) and EtOH, dried, rehydrated in 10–20 μL DEPC-treated water and stored at −80 °C until further use.

DNA extraction from fixed WISH zebrafish embryos for genotyping

Residual PBT was removed from embryos and replaced with 50 μL 300 mM NaCl. The sample was incubated for 4 hours at 65 °C for reverse crosslinking. Afterwards, NaCl was removed and replenished with 50 μl ELB (50 ml: 500 μL 1M Tris pH 8.0, 2.5 ml 1M KCl, 150 μL Tween-20 (P1379, Sigma), 47 ml water) and 4 μL 10mg/ml Proteinase K (P8108S, NEB). The samples were incubated for 8 hours at 55 °C followed by 10 min at 98 °C to inactivate the Proteinase K. The samples were centrifuged and stored at −20 °C until genotyping procedure (see above).

Image analysis for morphometric measurements

WISH-stained embryos with myoD, dlx3b, tbxta, and hgg1 were transferred into a clear flat-bottom dish with methylcellulose and imaged using a Nikon SMZ18 stereomicroscope. After imaging, embryos were transferred back in PBT for storage at 4 °C or immediately genotyped as described above.

ImageJ was used to visualize and manipulate images of WISH-stained embryos and measurements were made manually using the line and angle tools ^166,184. Axis length and angle were collected by measuring from the anterior to posterior aspects of the dlx3b expression domain in lateral-view images. Neural plate width was collected by measuring from one side of the dlx3b expression domain to the other side at the level of the otic placodes and beginning of the notochord in dorsal-view images. Somite width was taken from dorsal-view images by measuring the left to the right side of the myoD expression domain at the same last three somite pairs per embryo.

Quantification of axis length, axis angle, neural plate width, and somite width across conditions were averaged for each timepoint. Average values and ± S.D. were plotted using GraphPad Prism (10.0.3).

Image analysis of hematopoietic markers and quantification of ISH signal

WISH-stained embryos with kdrl, sox7, gata1, runx1, gfi1aa, and gfi1b were transferred into a clear flat-bottom dish with glycerol and imaged using a Leica M205 dissecting scope with TL3000 Ergo base, swan neck lamps, Planapo 1.0x M-series objective, and FLEXACAM C1 camera.

Quantification of the ISH signal for gata1, gfi1aa, and gfi1b was done as previously described using ImageJ ¹⁸⁵. Briefly, after imaging the WISH-stained embryos with consistent parameters, the images were inverted to negative and converted to 8-bit grayscale. A region of interest (ROI) containing the ISH expression signal was drawn manually for each embryo using the freehand selection tool in ImageJ. To define the background, a second ROI with the same shape and area was created in an unstained region of the trunk, above the ROI with the signal. A value for each region was determined by measuring the average pixel intensity. The pixel intensity of the ISH signal was determined by subtracting the value of the background region from the value of the stained region. Individual data points, means, and ± S.D. were plotted using GraphPad Prism (10.0.3). Quantification of the ISH signal intensity in AGM, PBI, and neuron territories in runx1-stained embryos was done by giving the signal intensity a score from 1 to 5 by three independent researchers, totaling four counts per embryo. Counts were averaged and are presented in the graph as individual data points. Individual data points, means and ± S.D. were plotted using GraphPad Prism (10.0.3).

Experimental study design

All assays were treated with identical experimental conditions across species and performed at least twice or more times. All attempts at replication were successful.

No data were excluded in the zebrafish studies.

Data analyses of the cd41-transgenic reporter quantification was based on morpholino injections into zebrafish embryos and based on defined genotypes of zebrafish embryos from crosses. No other randomizations were applicable.

Data collection for the cd41:EGFP-positive cell counting was blinded to avoid researcher bias and counted by at least two independent researchers. Data collection for the circulating thrombocyte video quantitation and functional analysis was recorded by a researcher blinded to the genotype. Data collection of ISH analyses were blinded and morphometrics and ISH signal intensity were measured by at least two independent researchers. Zebrafish embryos were not selected by gender as sex determination happens later in development.

Statistical analysis

The authors declare that key measures of statistics and reproducibility are built into the work throughout. For the zebrafish assays, sufficient embryos were analyzed to achieve statistical significance based on previous experience in these studies. Experimental sample sizes were chosen by common standards in the field and in accordance with solid phenotype designation.

HIGHLIGHTS.

• Zebrafish mutants for the TAR Syndrome gene rbm8a show thrombocyte reduction

• Attenuated Rbm8a function results in retained introns in mRNAs encoding PCP components

• Early PCP defects result in lateral plate mesoderm anomalies and hematopoietic defects

• PCP anomalies selectively impact cell fate patterning

Data and code availability

All mRNA sequencing data has been deposited at ArrayExpress with accession E-MTAB-12503. The related GitHub repository (link: https://github.com/markrobinsonuzh/rbm8a-rnaseq-explorer) hosts code and R objects to support the R/Shiny app for exploring rbm8a mutant RNA-seq experiments interactively. All primary data from the cd41:EGFP-positive cell counts, SPIM imaging of the scl:GFP-positive territory to determine area and volume, ISH morphometrics and quantifications are included as tables in the Supplementary Data. The western blots are included in Supplementary Fig. 1.

The code for the R/Shiny-based RNAseq Explorer app and instructions for its use are available at https://github.com/markrobinsonuzh/rbm8a-rnaseq-explorer, with track data downloadable from https://doi.org/10.5281/zenodo.7680271.