Tmem88 confines ectodermal Wnt2bb signaling in pharyngeal arch artery progenitors for balancing cell cycle progression and cell fate decision

Pharyngeal arch artery (PAA) progenitors undergo proliferative expansion and angioblast differentiation to build vessels connecting the heart with the dorsal aortae. However, it remains unclear whether and how these two processes are orchestrated. Here we demonstrate that Tmem88 is required to fine-tune PAA progenitor proliferation and differentiation. Loss of zebrafish tmem88a/b leads to an excessive expansion and a failure of differentiation of PAA progenitors. Moreover, tmem88a/b deficiency enhances cyclin D1 expression in PAA progenitors via aberrant Wnt signal activation. Mechanistically, cyclin D1-CDK4/6 promotes progenitor proliferation through accelerating the G1/S transition while suppressing angioblast differentiation by phosphorylating Nkx2.5/Smad3. Ectodermal Wnt2bb signaling is confined by Tmem88 in PAA progenitors to ensure a balance between proliferation and differentiation. Therefore, the proliferation and angioblast differentiation of PAA progenitors manifest an inverse relationship and are delicately regulated by cell cycle machinery downstream of the Tmem88-Wnt pathway. Using zebrafish as a model, Zhang et al. show that Tmem88a/b expression is required to balance proliferation and differentiation of pharyngeal arch artery progenitors into angioblasts by confining ectodermal Wnt2bb signaling.

Pharyngeal arch artery (PAA) progenitors undergo proliferative expansion and angioblast differentiation to build vessels connecting the heart with the dorsal aortae. However, it remains unclear whether and how these two processes are orchestrated. Here we demonstrate that Tmem88 is required to fine-tune PAA progenitor proliferation and differentiation. Loss of zebrafish tmem88a/b leads to an excessive expansion and a failure of differentiation of PAA progenitors. Moreover, tmem88a/b deficiency enhances cyclin D1 expression in PAA progenitors via aberrant Wnt signal activation. Mechanistically, cyclin D1-CDK4/6 promotes progenitor proliferation through accelerating the G1/S transition while suppressing angioblast differentiation by phosphorylating Nkx2.5/Smad3. Ectodermal Wnt2bb signaling is confined by Tmem88 in PAA progenitors to ensure a balance between proliferation and differentiation. Therefore, the proliferation and angioblast differentiation of PAA progenitors manifest an inverse relationship and are delicately regulated by cell cycle machinery downstream of the Tmem88-Wnt pathway.
Pharyngeal arch arteries (PAAs) are essential embryonic blood vessels that connect the heart with the dorsal aortae and undergo extensive remodeling, leading to the development of critical segments of the aortic arch and its branches at birth. It has been suggested that impaired PAA development may cause life-threatening congenital cardiovascular malformations [1][2][3] . During vertebrate embryogenesis, nkx2.5 + PAA progenitors are specified within anterior lateral plate mesoderm (ALPM) and then condense into several pharyngeal clusters 4,5 . Subsequently, these progenitors proliferate and differentiate into angioblasts, which undergo angiogenic sprouting and ultimately establish the functional vascular structures 5 .
Embryogenesis requires exquisite regulation of progenitor cell proliferation and differentiation to ensure the proper generation of a massively diverse range of cells. Indeed, a delicate balance between cell cycle dynamics and cell fate determinants has been found to be tightly controlled by multiple bidirectional interactions between the cell cycle machinery and factors driving differentiation [6][7][8][9] . Specifically, in the presence of developmental signaling, stem/progenitor cells preferentially make cell fate decisions in G1. This is because the G1 phase potentially establishes a favorable epigenetic and nuclear architectural environment, allowing for activation of transcriptional programs that direct cell fate 10 . During PAA development, the proliferation Article https://doi.org/10.1038/s44161-023-00215-z β-catenin accumulates in both the cytoplasm and nucleus to regulate target gene transcription 17,18 . Transmembrane protein 88 (TMEM88) is a two-transmembrane-type protein localized on the plasma membrane, Golgi, endocytic vesicles and multivesicular bodies (MVBs) 19,20 . Based on its C-terminal PDZ-binding motif, TMEM88 has been proposed to inhibit Wnt signaling by associating with Dvl proteins and/or sequestrating the Wnt signalosome to MVBs, which eventually fuses with the lysosome 19,20 . As an inhibitory factor of Wnt/β-catenin signaling, TMEM88 regulates embryonic axis formation in Xenopus laevis and cardiomyocyte differentiation of human embryonic stem cells 19,21 . However, whether and how the functional interaction between TMEM88 and and angioblast differentiation of PAA progenitors are determined by multiple cues in a spatiotemporal manner. Critically, the alteration of cell number and lineage commitment causes severe vascular defects [11][12][13][14] . However, there remains limited information about whether and how these two processes are orchestrated.
It is well established that Wnt/β-catenin signaling plays a crucial role in coordinating cell proliferation and differentiation in the intestinal stem cells and neural stem/progenitor cells 15,16 . During canonical Wnt signaling, the Wnt ligand binds to its co-receptors-Frizzled and LRP5/6-resulting in the activation of Dishevelled (Dvl) proteins. After Dvl activation, the degradation of β-catenin is inhibited. Consequently,    Article https://doi.org/10.1038/s44161-023-00215-z Wnt signaling pathway controls the fine balance between cell cycle and cell differentiation during PAA development has yet to be explored. Zebrafish have two tmem88 orthologs-termed tmem88a and tmem88b-both of which encode conserved proteins featured with the C-terminal PDZ-binding motif 22 . tmem88a transcripts are enriched in the developing blood and vascular cells 23,24 . Inactivation of tmem88a by antisense morpholino (MO) led to a loss of erythrocytes and myeloid cells 23 . The expression of tmem88a was also observed in the ALPM during embryo somitogenesis 21,25 . MO-mediated knockdown studies have shown that tmem88a was required to specify cardiac fate by restricting Wnt signaling 25 . However, in genetic mutants of tmem88a, the formation of blood cells remained unaffected, and no significant morphological heart defects were detected 22 . Thus, the functions of Tmem88 in embryonic development requires further investigation.
In this work, we generated tmem88a −/− and tmem88b −/− single mutants and tmem88a −/− ;tmem88b −/− (tmem88a/b −/− ) double mutant using the CRISPR-Cas9 system. The single mutants showed no apparent defects in heart development and blood cell formation, but a certain proportion of tmem88a/b −/− double mutants displayed impaired cardiomyocyte maturation and reduced blood cells. More notably, we found that simultaneous depletion of tmem88a and tmem88b led to overproliferation and impaired angioblast differentiation of PAA progenitors. We further reveal a key role for Tmem88 in governing the delicate balance between PAA progenitor proliferative expansion and angioblast differentiation by confining ectodermal Wnt2bb signaling.

tmem88a/b transcripts are enriched in PAA progenitors
To uncover the function of tmem88 genes during PAA development, the expression of tmem88a and tmem88b in zebrafish embryos was first examined by whole-mount in situ hybridizations from 22 hours to 42 hours post-fertilization (hpf). Consistent with previous reports, tmem88a was expressed in the primordial heart, the developing vasculature and caudal hematopoietic tissue (Extended Data Fig. 1a) 23,24 . Interestingly, tmem88b showed similar spatiotemporal expression patterns (Extended Data Fig. 1b). In addition, we noticed that, in the pharyngeal region, tmem88a/b transcripts exhibited discontinuous distribution reminiscent of the PAA progenitor clusters (Extended Data Fig. 1a,b).
To further confirm the expression of tmem88a/b in PAA progenitors, fluorescence in situ hybridization combined with immunofluorescence staining was performed in Tg(nkx2.5:ZsYellow) embryos. The transcripts for tmem88a/b were specifically localized in the ventral root of ZsYellow + sprouts, where the PAA progenitors are situated ( Fig. 1a,b) 11 . Because the PAA progenitor clusters undergo angioblast differentiation in a cranial-to-caudal manner 5 , the absence of tmem88a/b expression in the anterior PAA sprouts at 38 hpf might indicate a cell fate commitment (Fig. 1a,b). We next investigated the reciprocal relationship between the expression of these two tmem88 genes, the PAA progenitor marker nkx2.5 and the PAA angioblast marker tie1 during PAA morphogenesis. The expression of tmem88a and tmem88b appeared nearly simultaneously with nkx2.5 expression in most of the PAA clusters, with cluster 6 having even earlier expression ( Fig. 1c). It has been shown that MO-mediated inactivation of nkx2.5 in zebrafish embryos suppressed angioblast differentiation, resulting in an accumulation of PAA progenitors 5 . As expected and when compared to control animals, nkx2.5 morphants exhibited increased expression of tmem88a/b in the pharynx (Fig. 1d). Our recent work has shown that the cloche gene npas4l is essential for the specification of PAA progenitors from the pharyngeal mesoderm 11 . Consistently, the expression of tmem88a/b was completely abolished in cloche −/− mutants (Fig. 1e). Collectively, these results demonstrate that both tmem88a and tmem88b are expressed in PAA progenitors and can be used as new progenitor marker genes. tmem88a and tmem88b synergistically support PAA formation To explore the role of tmem88a and tmem88b in PAA development, zebrafish mutants for these two genes were generated using the CRISPR-Cas9 system. The mutant allele of tmem88a or tmem88b carried a DNA deletion near the guide RNA (gRNA) targeting sequence and was predicted to encode a truncated protein lacking the transmembrane domains along with the C-terminal PDZ-binding motif (Extended Data Fig. 2a,b). Homozygous mutants of tmem88a and tmem88b were then successfully identified from the offspring of heterozygous parents (Extended Data Fig. 2c-e). In line with a previous report 22 , genetic inactivation of either tmem88a or tmem88b caused no morphological defects and had no apparent effects on the formation of the heart and blood cells (Extended Data Fig. 2f-h).
We next examined the regional blood flow in the pharynx in embryos carrying transgenic reporters flk:EGFP and gata1:DsRed. We found that approximately 80% of tmem88a −/− mutants and 10% of tmem88b −/− embryos exhibited obvious PAA vascular stenosis accompanied by interrupted blood flow (Fig. 2a,b), suggesting that both tmem88a and tmem88b are involved in PAA development, with tmem88a playing a primary role in this process. Notably, the expression of tmem88a at both the messenger RNA (mRNA) and protein levels was not increased in tmem88b −/− mutants and vice versa (Extended Data Fig. 3a-d), excluding a possible genetic compensatory interaction between these two genes. These observations led to the hypothesis that tmem88a and tmem88b synergistically regulated PAA development. To verify this, a tmem88a/b −/− double mutant was generated by intercrossing tmem88a −/− and tmem88b −/− fish (Extended Data Fig. 3d). A fluorescent tracer-rhodamine-dextran-was next injected into the common cardinal vein of wild-type and tmem88a/b −/− embryos to visualize blood flow. When compared to control animals, most tmem88a/b −/− mutants had a total blockage of blood flow in the pharynx at 72 hpf ( Fig. 2c), indicating dysplastic PAAs. Simultaneous depletion of tmem88a and tmem88b resulted in either non-patent PAAs or a complete lack of these embryonic vessels at 60 hpf ( Fig. 2d-g). This effect was much more serious than that observed in the tmem88a −/− mutants. The expression of endothelium marker genes tie1 and flk in PAAs at 60 hpf was notably decreased or completely abolished in approximately 90% of tmem88a/b −/− mutants (Fig. 2h,i). Collectively, these results indicate that tmem88a and tmem88b function redundantly and are indispensable for PAA morphogenesis. Interestingly, the PAA Article https://doi.org/10.1038/s44161-023-00215-z stenosis in tmem88a −/− or tmem88b −/− embryos was recovered at 96 hpf (Fig. 2j). On the contrary, most tmem88a/b −/− mutants still showed significantly less blood flow in the PAAs or lack of PAA structures and died by 6 days post-fertilization (Fig. 2k), ruling out the possibility that the PAA defects resulted from developmental delay.
In addition, approximately 75% of tmem88a/b −/− embryos displayed evident pericardial edema from 36 hpf (Extended Data Fig. 3e). Interestingly, a small group of tmem88a/b-deficient embryos showed more severe morphological defects, including the absence of notochord and shortened posterior trunk, and died before 48 hpf 39    Article https://doi.org/10.1038/s44161-023-00215-z (Extended Data Fig. 3e). Moreover, loss of tmem88a/b resulted in a significant reduction in cmlc2 expression in the cardiac chambers and led to an obvious decrease in erythrocytes (Extended Data Fig. 3f,g), demonstrating that tmem88a and tmem88b function together in heart and blood cell development.

Tmem88 coordinates cell proliferation and differentiation
We next sought to determine whether attenuation of tmem88a/b would affect the emergence of PAA progenitors. Surprisingly, we observed a marked increase of nkx2.5 transcripts in tmem88a/b-deficient embryos (Fig. 3a). Consistently, there appeared to be more nkx2.5 + cells in the PAA clusters of tmem88a/b −/− mutants when compared to that of control embryos (Fig. 3b). Additionally, we found no obvious changes in nkx2.5 expression in the ALPM of tmem88a/b −/− embryos (Extended Data Fig. 3h), suggesting a normal segregation of the pharyngeal mesodermal progenitor population from the cardiac precursors. Thus, tmem88a and tmem88b are not required for PAA progenitor specification but necessary to restrict their expansion. We hypothesized that the loss of tmem88 genes might change the proliferative property of PAA progenitors. Indeed, the results of BrdU incorporation assays showed a substantial increase in the number of proliferative PAA progenitors in tmem88a/b −/− mutants (Fig. 3c,d). We then further analyzed the effects of tmem88a/b on cell cycle progression in Tg(EF1α:mKO2-zCdt1 (1/190)) and Tg(EF1α:mAG-zGem(1/100)) transgenic embryos, where cells in G1 and S/G2/M phases were labeled with mKO2 and mAG fluorescence, respectively 26 . We found a significant decrease in the number of PAA progenitors with mKO2 fluorescence in tmem88a/b −/− mutants (Fig. 3e,f). In contrast, the number of mAG + cells in PAA clusters was clearly increased upon tmem88a/b depletion (Fig. 3g,h). These results suggest a shortened G1 cell cycle phase that enables rapid progenitor cell proliferation.
Considering the severe malformations of PAA structures in tmem88a/b −/− mutants, we next determined whether these tmem88 genes also regulated PAA progenitor differentiation. We detected significantly decreased expression of the earlier angioblast lineage marker genes etv2 and scl in tmem88a/b-depleted embryos at 38 hpf, indicating an impaired angioblast transition for PAA progenitors (Fig. 3i). Consistent with this, almost no tie1 + PAA angioblasts were found in tmem88a/b −/− mutants at a later developmental stage (Fig. 3i). Together, these data suggest that Tmem88 coordinately regulates the cell cycle progression and angioblast differentiation of PAA progenitors.
Based on the above observations, we hypothesized that suppression of the overactivation of Wnt/β-catenin signaling in tmem88a/ b −/− mutants would rescue the PAA defects. To verify this, tmem88a/ b −/− embryos were treated with the selective Wnt/β-catenin signal inhibitor CCT036477 (ref. 28 ). BrdU incorporation assays showed that CCT036477 treatment eliminated the tmem88a/b-depletioninduced overproliferation and excessive expansion of PAA progenitors ( Fig. 5a-c). Interestingly, blocking Wnt/β-catenin signaling in tmem88a/ b −/− embryos using either CCT036477 treatment or heat-shock-induced overexpression of a secreted Wnt inhibitor-DKK1-significantly alleviated PAA progenitor differentiation defects ( Fig. 5d-f). In addition, both the impaired PAA morphogenesis and the interrupted blood flow in tmem88a/b −/− embryos were partially recovered by CCT036477 treatment (Fig. 5g,h). These findings indicate that there is an inverse relationship between proliferation and angioblast differentiation of PAA progenitors, which is developmentally balanced by Tmem88-Wnt signaling. Indeed, pharmacological treatment of wild-type embryos with BIO-a selective chemical activator of Wnt/β-catenin signaling 29 accelerated the expansion of PAA progenitor cells at the expense of angioblast differentiation in a dose-dependent manner (Fig. 5i,j). Similar angioblast formation defects were also observed in the heatshocked Tg(hsp70l:wnt8a-EGFP) embryos (Extended Data Fig. 4a,b).
Taken together, the above results suggest that Tmem88 balances PAA progenitor proliferation and angioblast differentiation by negatively regulating Wnt/β-catenin signaling.

CDK4/6 connects progenitor proliferation and differentiation
Two families of G1 cyclins are known to operate in mitotic cells: D-type (D1, D2 and D3), which activates the cyclin-dependent kinases CDK4 and CDK6, and E-type (E1 and E2), which activates CDK2 (ref. 30 ). Wnt/βcatenin signaling is a well-known promoter of cell cycle progression, which is achieved by upregulation of cyclin D1 expression 31 . To this end, we found a marked enhancement of cyclin D1 expression in the PAA progenitor clusters of tmem88a/b −/− embryos (Fig. 6a). Quantitative reverse-transcription polymerase chain reaction (PCR) analysis of ZsYellow + cells sorted from the pharynx region of Tg(nkx2.5:ZsYellow) embryos further indicated that, upon depletion of tmem88a/b, the expression of cyclin D1 was increased in PAA progenitors, but this change disappeared after the progenitors differentiated into PAA angioblasts (Extended Data Fig. 5a-c). In contrast, loss of tmem88a/b had no effect on the expression of cdk4 during PAA development (Extended Data Fig. 5a-c). Thus, the stage-specific upregulation of cyclin D1 expression in tmem88a/b −/− mutants was consistent with the shortened G1 phase in PAA progenitors.
To determine whether the increased cyclin D1-CDK4/6 activity was responsible for the failure of PAA angioblast differentiation, we TopFlash luciferase reporter and the indicated plasmids were treated with or without Wnt3a conditional medium (CM) and then harvested for luciferase assays (a) and western blots (b). The relative luciferase activity was the mean with ± s.d. from three independent biological repeats. Student's t-test was used. NS, non-significant. c,d, The expression of etv2 and scl was examined in tmem88a/b −/− mutants injected with 200 pg of tmem88a or tmem88a-ΔC mRNAs (c). Bar graphs showing the average number of etv2 + or scl + angioblast clusters that formed in each group (d). n ≥ 15 per group. Student's t-test was used. Data are mean ± s.d. e-g, Tg(nkx2.5:ZsYellow) embryos were co-immunostained with anti-ZsYellow and anti-β-catenin antibodies at 28 hpf (e) and 38 hpf (f). Nuclei were counterstained with DAPI. The images shown in the insets were amplified from the boxed areas in the corresponding images. These experiments were repeated independently three times. Scale bars, 50 µm. Average fluorescent intensity of β-catenin over the ZsYellow + cell region in each group was calculated and shown in g. n = 5 per group. Student's t-test was used. Data are mean ± s.d. h-j, Tg(nkx2.5:ZsYellow;TOP:dGFP) embryos stained with anti-ZsYellow antibody were hybridized with gfp probe at 28 hpf (h) and 38 hpf (i). These experiments were repeated independently three times. Scale bar, 50 µm. Average fluorescent intensity of gfp over the ZsYellow + cell region in each group was calculated and shown in j. n = 5 per group. Student's t-test was used. Data are mean ± s.d. WT, wild-type.
Article https://doi.org/10.1038/s44161-023-00215-z next applied PD0332991, a specific CDK4/6 inhibitor 32 , and CYC202, a selective CDK2 inhibitor 33 , to tmem88a/b −/− embryos. As expected, both PD0332991 and CYC202 treatments eliminated the excessive expansion of PAA progenitors (Fig. 6b). However, the impaired PAA angioblast differentiation was recovered after only PD0332991-and not CYC202treatment (Fig. 6b). In addition, treatment of tmem88a/b −/− embryos with the G2/M phase inhibitor RO-3306 or the potent S phase inhibitor PHA-767491 achieved similar effects as CYC202 treatment (Extended Data Fig. 5d). Even more important, tmem88a/b −/− embryos treated with PD0332991 exhibited greatly improved vascular morphology and blood flow (Fig. 6c, d). These results suggest that the angioblast differentiation of PAA progenitors is closely tied to proliferative expansion Relative luciferase activity Because Nkx2.5 has been implicated in the regulation of PAA development in mouse and human 5,34 , we speculated that CDK4/6 kinases might affect the function of zebrafish Nkx2.5 (zNkx2.5). To test this hypothesis, CDK4 and zNkx2.5 were co-overexpressed in mammalian cells. Unexpectedly, both the protein stabilization and subcellular distribution of zNkx2.5 were not affected in the presence of CDK4 (Extended Data Fig. 6a,b). We then considered whether the transcriptional activity of zNkx2.5 was affected by CDK4. To this end, a reporter   BrdU ZsYellow  Article https://doi.org/10.1038/s44161-023-00215-z assay was performed in HEK293T cells using the widely employed Nkx2.5-responsive reporter ANF-luciferase construct (ANF-Luc) 35 . We found that the Nkx2.5-induced activation of ANF-Luc was markedly repressed by overexpression of wild-type CDK4 but not its kinase-dead K38M mutant 36 (Extended Data Fig. 6c,d). The presence of cyclin D1 further enhanced the inhibitory effects of CDK4 on ANF-Luc expression (Extended Data Fig. 6e). These results indicate that cyclin D1-CDK4/6 negatively regulates the transcriptional capacity of Nkx2.5 in a kinase activity-dependent manner.
Next, co-immunoprecipitation experiments demonstrated a physical association between CDK4 and zNkx2.5 in HEK293T cells (Fig. 6e). In vitro binding assay using purified proteins further confirmed a direct interaction between these two proteins (Fig. 6f). Given that CDK4 is a proline-directed serine/threonine kinase 37 , we sought to determine whether CDK4 could phosphorylate zNkx2.5. Multiple sequence alignments showed that threonine 11 (T11), followed by a proline, in zebrafish Nkx2.5 might be a conserved CDK4 phosphorylation site (Fig. 6g). Indeed, ectopic expression of wild-type CDK4-and not its kinase-dead mutant-resulted in a significant increase in the phosphorylation level of zNkx2.5 (Fig. 6h). Meanwhile, tmem88a/ b −/− embryos exhibited a marked increase in the phosphorylation of endogenous Nkx2.5, which could be suppressed by CCT036477 treatment (Fig. 6i). Notably, the CDK4-induced hyperphosphorylation of zNkx2.5 was heavily reduced when alanine was substituted for T11 (Fig. 6h). Furthermore, in vitro phosphorylation assays showed that, when recombinant cyclin D1-CDK4 was added, notable phosphorylation was detected in purified zNkx2.5 protein but not in its T11A mutant (Fig. 6j). Based on these observations, we conclude that CDK4 phosphorylates zNkx2.5 directly at a conserved threonine residue.
We further examined the activity of ANF-Luc after overexpression of either wild-type zNkx2.5 or its T11A mutant in HEK293T cells and wild-type embryos. Impressively, and when compared to wildtype zNkx2.5, the phosphorylation-resistant mutant more efficiently enhanced ANF-Luc expression (Extended Data Fig. 6f,g). These results indicate an inhibitory function for CDK4-mediated phosphorylation on threonine 11 for the transcriptional activity of zNkx2.5. We next examined the rescue effects of wild-type zNkx2.5 and its T11A mutant on PAA angioblast differentiation in tmem88a/b −/− embryos. To turn on zNkx2.5 expression in a temporally restricted fashion, we used an antisense photo-cleavable morpholino (AS-Flag-photo-MO), which targets to the start codon and subsequent Flag sequence of relevant mRNA and blocks translation, presumably by interfering with the progression of ribosomes on the mRNA until UV exposure 38 . We found that overexpression of zNkx2.5-T11A in tmem88a/b −/− mutants from 20 hpf more effectively reduced nkx2.5 + PAA progenitors, increased etv2 + angioblasts and recovered the defects in PAA formation than its wild-type counterpart (Fig. 6k,l). These data suggest that, after Tmem88 deficiency, cyclin D1-CDK4-mediated hyperphosphorylation of zNkx2.5 restrained its transcriptional activity.

CDK4 phosphorylates Smad3 to suppress PAA differentiation
It was suggested that genetic depletion of nkx2.5 in zebrafish did not lead to obvious PAA defects, probably due to the activation of a compensatory network 39,40 . Given that the PAA defects were steadily observed in tmem88a/b −/− mutants, it is likely that Tmem88-Wnt signaling regulates PAA angioblast differentiation through other factors aside from Nkx2.5. One possible candidate is Smad3, as this transcription factor is a major physiological substrate of the G1 cyclin-dependent kinases and functions as a cellular effector of TGF-β signaling to induce PAA angioblast differentiation 13,41 .
The three CDK4 phosphorylation sites (Thr 8, Thr 178 and Ser 212) that have been identified in human Smad3 are highly conserved in its zebrafish homologs, termed Smad3a and Smad3b 41 (Extended Data Fig. 7a). Indeed, we found that CDK4 bound to and phosphorylated zebrafish Smad3 proteins (Extended Data Fig. 7b-d). Moreover, CDK4induced phosphorylation of these Smad3 proteins was suppressed by treatment with the CDK4/6 inhibitor PD0332991 (Extended Data Fig. 7d). In agreement with the observed upregulated cyclin D1 expression in tmem88a/b −/− embryos, the CDK4-mediated phosphorylation of Smad3 was significantly increased in tmem88a/b-depleted PAA progenitors (Fig. 7a). Moreover, such Smad3 hyperphosphorylation was eliminated in tmem88a/b −/− embryos treated with CCT036477 (Fig. 7a). These results indicate that Tmem88-Wnt signaling controls the G1 CDK-mediated Smad3 phosphorylation in PAA progenitors. . Proliferating cells were visualized by BrdU immunofluorescence (a). Scale bar, 50 µm. This experiment was repeated independently three times. The percentage of BrdU + PAA progenitors for each group is shown in b. n ≥ 5 per group. Student's t-test was used. Data are mean ± s.d. c,d, Expression changes of nkx2.5 (c), etv2 and scl (d) in wild-type and tmem88a/b −/− embryos treated with or without 10 µM CCT036477 (CTT). e,f, Tg(hsp70l:dkk1b-EGFP) embryos were heat-shocked at 20 hpf for 30 minutes and then harvested at 38 hpf for in situ hybridization (e). The average numbers of PAA angioblast clusters were quantified in f. n ≥ 18 per group. Student's t-test was used. Data are mean ± s.d. g,h, Tg( flk:EGFP;gata1:DsRed) embryos were treated with or without 10 µM CCT036477 and then harvested for live confocal images. Representative images of different classes of PAA morphology are shown in g. Scale bar, 50 µm. This experiment was repeated independently three times. The ratios of different classes in each group are presented in h. n > 20 per group. i,j, The expression of nkx2.5 and etv2 in wild-type embryos treated with increasing concentrations of BIO (i). The average numbers of etv2 + PAA angioblast clusters were quantified in j. n ≥ 20 per group. Student's t-test was used. Data are mean ± s.d. WT, wild-type. . c,d, Representative images of different classes of PAA morphology in embryos treated with or without 10 µM PD0332991 (c). Scale bar, 50 µm. This experiment was repeated independently three times. Statistical data for each group are presented in d. n > 18 per group. e, Immunoprecipitation assays to detect the interaction of CDK4 and zNkx2.5 in HEK293T cells. f, Direct binding of purified CDK4 to zNkx2.5 proteins. g, Conserved CDK substrate motifs in Nkx2.5 proteins from different species. Red stars indicate critical residues in the CDK substrate motifs. h, CDK substrates were immunoprecipitated from HEK293T cells using a phospho-threonine-proline antibody and blotted with anti-Flag antibody. i, Embryos were harvested at 32 hpf and lysed for immunoprecipitation with a phospho-threonine-proline antibody. The phosphorylation level of zNkx2.5 was detected by western blotting with anti-Nkx2.5 antibody. j, In vitro phosphorylation of zNkx2.5 protein by cyclin D1-CDK4. Flag-zNkx2.5 protein and its T11A mutant were purified from HEK293T cells by anti-Flag M2 agarose beads and then incubated with recombinant HA-cyclin D1 and Myc-CDK4 proteins that were purified from bacterial cells. k,l, tmem88a/b −/− embryos were injected with 200 pg of indicated mRNA and 1 ng of AS-Flag-photo-MO at the one-cell stage and then subjected to UV exposure at 20 hpf. The resulting embryos were harvested for in situ hybridization (k) and for live confocal images at 60 hpf (l). Both the representative images of different classes of PAA morphology and statistical data are shown in l. Scale bar, 50 µm. This experiment was repeated independently three times. n ≥ 20 per group. WT, wild-type; IP, immunoprecipitation; WB, western blot; WCL, whole cell lysate.
Article https://doi.org/10.1038/s44161-023-00215-z Because Etv2 and Scl are essential transcription factors for the initiation of the angioblast program 42,43 , we next investigated the effects of zebrafish Smad3 proteins on etv2 and scl transcription in HEK293T cells. We constructed two reporter plasmids, etv2-Luc and scl-Luc, which harbored critical cis-regulatory elements that drove transgene expression in developing vasculatures 44,45 . Luciferase assays showed        Article https://doi.org/10.1038/s44161-023-00215-z that overexpression of Smad3a promoted the activity of etv2-Luc-and not scl-Luc-in a dose-dependent manner (Extended Data Fig. 8a). We further found that overexpression of CDK4, but not its kinase-dead mutant, suppressed Smad3a-induced etv2-Luc reporter expression. Treatment with the CDK4/6 inhibitor PD0332991 relieved this suppression (Extended Data Fig. 8b). Furthermore, simultaneous mutation of the three CDK phosphorylation sites in Smad3a abolished CDK4mediated inhibition on etv2-Luc reporter activation (Extended Data Fig. 8c). Together, these results suggest that CDK4 phosphorylates Smad3 to repress etv2 transcription.
To identify the potential cis-regulatory elements responsible for Smad3, a series of truncations of the etv2 promoter were generated. The results of reporter assays suggested that the region between −2,516 bp and −2,270 bp within the etv2 promoter contains Smad3-responsive elements (Extended Data Fig. 8d). Smad3 directly binds to the DNA sequence CAGAC, Smad-binding element (SBE), within the promoters of its target genes 46 . Sequence analysis showed two potential SBEs in the −2,516-bp to −2,270-bp fragment (Extended Data Fig. 8e). Electrophoretic mobility shift assay (EMSA) experiments revealed that purified Smad3 specifically bound to synthesized probes that contained the SBEs but not to the mutants with one nucleotide replacement in the binding motif (Extended Data Fig. 8f). It is worth noting that mutation of either of the two SBEs resulted in a clear decrease of etv2-Luc reporter expression in HEK293T cells and wild-type embryos. Consistently, mutations in both of the two SBEs nearly abolished the capacity to respond to Smad3 (Extended Data Fig. 8g,h). These results suggest that Smad3 directly binds to SBEs in the promoter to accelerate etv2 expression.
To test whether CDK4-mediated Smad3 hyperphosphorylation is responsible for the failure of angioblast differentiation, Flag-Smad3a or its CDK phosphorylation-resistant mutant Flag-Smad3a-3A were temporally overexpressed in tmem88a/b −/− embryos by using AS-Flag-photo-MO. Compared with wild-type Smad3a, its CDK phosphorylation-resistant mutant had better rescue effects on the excessive expansion of progenitors, impaired angioblast differentiation and defective PAA formation (Fig. 7b,c). In addition, an interesting question that remains to be solved is how the CDK4 activity in PAA progenitors is inhibited after the proliferative burst. It is well known that TGF-β signaling inhibits cell proliferation by activating the transcription of genes, including p15 and p21, that encode CDK4 and CDK2 inhibitory proteins, respectively 41 . Specifically, we found that p15 expression was substantially upregulated while the expression of p21 remained unchanged during PAA progenitor-to-angioblast transition ( Fig. 7d and Extended Data Fig. 9). Furthermore, treatment with selective TGF-β signaling inhibitor LY-364947 suppressed the developmental upregulation of p15 expression and, thereby, increased the CDK4-mediated phosphorylation of Smad3 (Fig. 7e,f). These results imply an important role of TGF-β signaling in slowing PAA precursor proliferation by inducing p15 expression. Consistent with previous findings 13 , wild-type embryos treated with LY-364947 displayed excess PAA precursors and impaired angioblast differentiation (Fig. 7g). It is important to note that blocking TGF-β signaling using LY-364947 abolished the rescue effects of CCT036477 on the overexpansion of PAA precursor and the failure of angioblast differentiation in tmem88a/b −/− embryos (Fig. 7h). Thus, these data suggest that TGF-β pathway may act as the developmental signal to arrest cell cycle of the proliferating progenitors and induce angioblast differentiation during PAA formation.

Tmem88 confines ectodermal Wnt2bb signaling
We next sought to explore which Wnt ligand triggers the cascade in PAA progenitors. The expression of a total of eight wnt genes, includ- ing wnt1, wnt2, wnt2ba, wnt2bb, wnt3, wnt3a, wnt8a and wnt8b, was examined by in situ hybridization. Interestingly, wnt2bb was the sole wnt gene expressed in the pharynx (Fig. 8a). Its transcripts were symmetrically localized in the pharynx from 20 hpf to 36 hpf (Fig. 8a,b).
In particular, transverse sections showed that wnt2bb was expressed in the outermost tissues known as the pharyngeal ectoderm (Fig. 8b). Co-localization analysis further confirmed the pharyngeal ectoderm expression of wnt2bb, as its transcripts were located in a cell layer adjacent to the PAA progenitors (as indicated by the expression of nkx2.5, tmem88a and tmem88b, respectively) and the chondrogenic neural crest cells (as indicated by sox10 expression) (Fig. 8c-f).
It has been shown that wnt2bb, which is also expressed in the lateral plate mesoderm adjacent to the foregut endoderm, is required for liver specification 47 . Thus, we hypothesized that the pharyngeal ectodermderived Wnt2bb may be responsible for activating Wnt/β-catenin pathway in PAA progenitors, and reducing the expression of wnt2bb should rebalance the proliferation and differentiation of PAA progenitors in tmem88a/b −/− mutants. Hence, a tmem88a/b −/− ;wnt2bb −/− triple mutant was generated by crossing tmem88a/b −/− fish with wnt2bb −/− fish 47 . Intriguingly, deletion of one allele of the wnt2bb gene did not cause an apparent decrease in cytoplasmic and nuclear β-catenin levels, nor did it result in overt developmental defects in PAA formation (Extended Data Fig. 10a,b). However, haploinsufficiency of wnt2bb in tmem88a/b −/− embryos eliminated the unusual accumulation of nkx2.5 + PAA progenitors, restored the emergence of etv2 + angioblasts and rescued the phenotypic defects of PAAs ( Fig. 8g-i). Likewise, and by using a previously validated antisense MO 47 , knockdown of wnt2bb in tmem88a/b −/− mutants similarly reduced the excessive proliferation of PAA progenitors, which regained the ability to differentiate into angioblasts (Extended Data Fig. 10c-f). Surprisingly, the expression of nkx2.5 was reduced, and the etv2 + angioblasts and their derivative PAAs were absent in tmem88a/b −/− ;wnt2bb −/− triple mutants (Fig. 8g-i). Consistent with this, inactivation of wnt2bb significantly attenuated Wnt/β-catenin signaling in PAA progenitors and severely impaired the formation of PAAs (Extended Data Fig. 10a,b), suggesting that Wnt/βcatenin signaling triggered by ectodermal Wnt2bb is required for PAA development, presumably by supporting precursor cell proliferation. However, it must be confined by Tmem88 in PAA progenitors to maintain the delicate balance between cell proliferation and differentiation (Fig. 8j).

Discussion
TMEM88 is highly induced during cardiovascular progenitor development and is functionally required for the cardiac differentiation of human embryonic stem cells 21 . In this study, we successfully generated two single zebrafish mutants-tmem88a −/− and tmem88b −/− -and the double mutant tmem88a/b −/− . Consistent with previous findings 22 , genetic inactivation of tmem88a or tmem88b caused no apparent cardiac defects. Conspicuously, a certain proportion of tmem88a/ b −/− double mutants displayed impaired cardiomyocyte maturation. Thus, TMEM88 might play an evolutionarily conserved role in heart development across species.
Progression of both proliferation and differentiation of PAA cells is indispensable for proper vascular development [11][12][13][14] . It remains unknown whether proliferation and differentiation are regulated in a coordinated manner during PAA morphogenesis. We found that both tmem88a and tmem88b were dynamically expressed in PAA progenitors, and simultaneous depletion of these two genes led to severe defects in PAA formation. Interestingly, tmem88a/b-deficient PAA progenitors showed hyperproliferation and impaired angioblast differentiation. Furthermore, treatment with the CDK4/6 inhibitor PD0332991 not only suppressed the excessive expansion of PAA progenitors in tmem88a/ b −/− embryos but also improved angioblast differentiation. These findings suggest an inverse linkage between PAA progenitor proliferation and differentiation, which is generally observed when stem/progenitor cells respond to differentiation cues 9,10,48 .
Until recently, only a few developmental signaling pathways have been connected to the cell cycle progression to cell fate determination 9,49 . In this work, we demonstrate that Tmem88 confines Article https://doi.org/10.1038/s44161-023-00215-z Wnt signaling to ensure a delicate balance between PAA progenitor proliferative expansion and angioblast differentiation. Several lines of evidence support this conclusion. (1) Wnt signaling is aberrantly activated in tmem88a/b-deficient PAA progenitors. (2) Overexpression of wild-type Tmem88a-but not its ΔC mutant-relieves the loss of PAA angioblasts in tmem88a/b −/− mutants. (3) Pharmacological blockade of Wnt/β-catenin signaling or haploinsufficiency of wnt2bb in tmem88a/ b −/− embryos rebalances the proliferation and differentiation of PAA progenitors in tmem88a/b −/− mutants. (4) Wild-type embryos treated with the Wnt signal activator BIO exhibit an excessive expansion of PAA progenitor cells at the expense of angioblast differentiation. Given that the development of many organs hinges on the important function of Wnt/β-catenin signaling in regulating cell proliferation and cell fate determination 17,18 , we hope that these findings will motivate additional work to explore whether Wnt/β-catenin signaling has a general role in coordinately regulating cell proliferation and differentiation during embryonic development.
In general, the activation of Wnt signaling does not appear to be cell cycle regulated 10,50 . An important scientific question remains regarding how the Wnt signaling pathway guides the cell cycle machinery to regulate developmental genes. Recent advances have characterized multiple bidirectional interactions between the G1 cell cycle machinery and factors driving cell fate decisions 6,9,10,51 . In line with the above findings, we found that the expression of cyclin D1 is strongly activated in tmem88a/b-deficient PAA progenitors. Cyclin D1-CDK4/6 accelerates G1/S transition to promote PAA progenitor proliferation. However, such G1 phase cyclin-CDKs phosphorylate Nkx2.5/Smad3 and suppress their transcription activity, thereby inhibiting angioblast commitment. Nkx2.5 is essential for PAA formation in mouse embryos, and multiple mutations of Nkx2.5 have been identified in patients with great vessel malformations 5,34 . Although Nkx2.5 may be dispensable for zebrafish PAA development, likely owing to genetic compensation 39 , our finding that cyclin D1-CDK4/6 controls Nkx2.5 transcription activity by phosphorylation provides critical insights into PAA development.
TGF-β/Smad3 signaling serves as a critical regulator of angioblast lineage commitment 13 . However, how Smad3 functions to promote PAA angioblast differentiation remains unknown. Here we show that TGF-β signaling may slow PAA precursor proliferation by inducing p15 expression. Moreover, p15-mediated CDK4 inhibition releases the repression of Smad3 transcriptional activity and then actives etv2 expression to drive angioblast differentiation. These findings fill a gap in our understanding regarding the role of TGF-β/Smad3 signaling in PAA development.
Furthermore, upon tmem88a/b depletion, the precursor cells in PAA clusters 3-6 show a clear increase in Wnt signal activation, cyclin D1 expression and cell proliferation. However, the expression of evt2 and scl was not obviously decreased in PAA cluster 6, implying that the angioblast differentiation of progenitors in PAA cluster 6 is less responsive to the tmem88a/b-deficient-induced hyperactivation of Wnt signaling. Nonetheless, the angioblast differentiation in PAA cluster 6 is definitely regulated by Wnt signaling because excessive activation of Wnt/β-catenin pathway in wild-type embryos by treatment with higher concentration of BIO impairs the angioblast differentiation in all the PAA clusters.
During embryonic development, one key question remains regarding how cells differentially respond to either morphogens or signaling molecules to obtain their regional identities for specific organ fates 52,53 . For instance, it has been shown that epithelial cell adhesion molecule (EpCAM) is the key molecule that confers competence to specific endodermal cells to respond to the liver-inductive Wnt2bb signal 52 . We found that wnt2bb is also expressed in the pharyngeal ectoderm overlying the PAA progenitors. However, in contrast to EpCAM, which enhances Wnt pathway in liver primordial cells, Tmem88 constitutes a key regulatory layer that is responsible for confining the response of PAA progenitors to ectodermal Wnt2bb signaling, thereby fine-tuning the balance between cell proliferation and angioblast differentiation.

Ethics statement
Our zebrafish embryo studies were approved by the Animal Care and Use Committee at the Institute of Zoology, Chinese Academy of Sciences (permission number: IOZ-13048).

Generation of tmem88a and tmem88b mutants
The tmem88a and tmem88b mutants were generated according to the procedures described previously 38 . We designed gRNAs targeting the second exons of tmem88a and tmem88b genes, respectively, using the CRISPR Design website (http://crispor.tefor.net/). These gRNAs were synthesized by T7 RNA polymerase in vitro and purified using mirVanaTM miRNA Isolation Kit (Ambion). Cas9 mRNA and gRNAs were co-injected into the one-cell stage zebrafish embryos. These founder (F0) embryos were raised to adulthood and outcrossed with wild-type fish to generate F1 heterozygous lines. For screening of the F1 fish with mutant alleles, genomic DNA (gDNA) was isolated from the tail of individual fish. gDNA was extracted from the resulting embryos at 24 hpf. A 453-bp DNA fragment spanning the target site for tmem88a was amplified from the gDNA by PCR (forward primer, 5′-TCCACTCCA CATCCAAAAAAACATA-3′; reverse primer, 5′-CTCTCAGCTCCAGTGG ACACTCGTT-3′) and sequenced for genotyping. Meanwhile, the forward primer 5′-CAACAGATACCTGACATTGGCACTT-3′ and reverse primer 5′-CCACCACAAACCATACATCAGTGC-3′ were used to amplify the tmem88b gRNA target sequence.
Microinjections were performed as previously described 38 . For smad3a mRNA overexpression experiments, embryos were co-injected with pre-mixed Flag-smad3a mRNA and AS-Flag-photo-MO (200 pg of Flag-smad3a mRNA and 1 ng of AS-Flag-photo-MO per embryo) at the one-cell stage and then treated with 365-nm UV 10 minutes by Lightbox (Gene Tools) at 20 hpf and then harvested at 38 hpf for in situ hybridization experiments. These experimental processes were conducted under constant dark conditions. For observation of blood flow, rhodamine-dextran (D7139, Invitrogen) was injected into the common cardinal vein of either wild-type or tmem88a/b −/− embryos 10 minutes Article https://doi.org/10.1038/s44161-023-00215-z before 72 hpf. The embryos were then embedded with 1% low melting agarose and imaged using a Nikon A1R+ confocal microscope.
Confocal images were captured using a Nikon A1R+ confocal microscope equipped with NIS Elements AR 4.13.00 software. Immunofluorescence images were analyzed using ImageJ version 1.48.

BrdU incorporation
Embryos developed to 26 hpf were incubated with 10 mM BrdU solution and shielded from light on ice for 30 minutes. Then, these embryos were transferred to Holtfreter's solution at 28.5 °C and harvested at the desired developmental stages for subsequent immunostaining.

Dual luciferase assay
Dual luciferase assay was performed according to the procedures described previously 38 . HEK293T cells were transfected or zebrafish embryos were injected with the indicated plasmids together with the internal control Renilla luciferase reporter vector. HEK293T cells were stimulated with or without Wnt3a conditional medium for 12 hours before harvesting for later luciferase activity assays. Each experiment was performed at least in triplicate, and the average activity was calculated.
Bound proteins were then separated by SDS-PAGE and visualized using western blot.
For GST pulldown assays, GST-zNkx2.5 proteins were immobilized using Glutathione Sepharose 4B beads and incubated with purified Myc-CDK4 at 4 °C for 4 hours. Samples were washed with TNE buffer three times, resuspended in 1× loading buffer and subjected to SDS-PAGE.

In vitro kinase assay
For in vitro kinase assays, either Flag-zNkx2.5 protein or its T11A mutant were purified from HEK293T cell lysates by immunoprecipitation and incubated with Myc-CDK4 and HA-cyclin D1 that had been affinity purified from E. coli in the kinase buffer (25 mM Tris-HCl pH 7.5, 5 mM β-glycerophosphate, 0.1 mM Na 3 VO 4 , 10 mM MgCl 2 and 2 mM dithiothreitol) with or without 50 µM ATP at 30 °C for 30 minutes. Then, the samples were fractionated in SDS-PAGE and visualized using either Coomassie blue staining or immunoblotting.

EMSA
EMSA was performed as previously described 58 . In brief, binding reactions were conducted for 30 minutes at room temperature in 10 µl of binding buffer containing 25% glycerol, 50 mM KCl, 0.5 mM EDTA, 10 mM DTT and 5 mM Tris-HCl pH 8.0. Purified GST-Smad3a (10 ng) protein was incubated with 0.5 ng of carboxyfluorescein (FAM)-labeled probes with or without 20-fold excess unlabeled oligonucleotide. The samples were resolved on a 5% non-denaturing polyacrylamide gel (29:1 acrylamideto:bisacrylamide ratio). The gels were run in 0.5× binding buffer at 4 °C and then visualized using a Typhoon FLA9500 scanner.

Statistical analysis
Student's t-test was used to analyze all datasets (GraphPad Prism 6.01 or Microsoft Excel 2013 software). Experiments were all performed at least in triplicate. Results were considered statistically significant at P < 0.05.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. anti-Flag antibody. d, CDK4 phosphorylates Smad3a and Smad3b. HEK293T cells were transfected with the indicated plasmids, and then treated with or without 0.5 µM PD0332991 for 5 h prior to harvest for Western blot. CDK4 substrates were blotted with an antibody against phosphorylated human Smad3 (Thr178). Tubulin was used as the loading control. WT, wild-type; KD, kinase dead.