p21, ccng1, foxo3b, and fbxw7 contribute to p53-dependent cell cycle arrest

Summary

p53 is a transcription factor and important tumor suppressor gene, yet its mechanism of tumor suppression remains unclear. While PUMA/BBC3, NOXA/PMAIP1, and p21/CDKN1A regulate apoptosis and cell-cycle arrest, zebrafish lacking puma, noxa, and p21 do not develop cancer, suggesting additional p53 targets contribute to tumor suppression. We show that p53 can still induce cell-cycle arrest in the absence of p21, either following DNA damage or mdm2 loss, implicating other transcriptional target in p53-dependent cell-cycle arrest. We conducted a cross-species analysis to identify 137 conserved p53-upregulated genes. Our analysis also stresses the importance of ortholog to paralog analysis across species, since in many cases the paralog but not ortholog in differing species is p53 dependent. Using a CRISPR-Cas9 G0 “crispant” screen in mdm2, puma, noxa, and p21 quadruple knockout zebrafish, we identified ccng1, fbxw7, and foxo3b that are involved in p53-dependent cell-cycle arrest.

Graphical abstract

Highlights

• •Unlike p53^−/− zebrafish, puma^−/−, noxa^−/−, p21^−/− zebrafish do not develop spontaneous tumor
• •p53 induces cell-cycle arrest even in the absence of p21
• •137 conserved p53-upregulated genes identified across zebrafish and mouse orthologs/paralogs
• •CRISPR G0 screen reveals ccng1, fbxw7, and foxo3b mediate p53-dependent cell-cycle arrest

Molecular biology; Cell biology; Omics

Introduction

The p53 pathway plays an essential role in tumor suppression. Evidence from The Cancer Genome Atlas (TCGA) confirms that TP53 is the most frequently mutated gene in cancers, with an average mutational occurrence of approximately 50% across cancers.¹ In some cancers, such as ovarian or triple negative breast cancers, p53 mutations have been found in over 95% of cancers, while in other cancers, where the TP53 gene remains wild-type, the p53 pathway is still repressed through alternative mechanisms, such as overexpression of MDM2 or MDM4, two negative regulators of p53.^2,3 The importance of p53 in tumor suppression is further underscored by Li-Fraumeni syndrome (LFS), an autosomal dominant cancer predisposition syndrome. LFS patients carry germline mutations in the TP53 gene, resulting in the loss of its functional activity. This syndrome is associated with a high penetrance of multiple cancer types, providing strong evidence for the critical role of p53 in preventing tumor development.^4,5,6,7,8,9 Furthermore, the tumor-suppressive activity of p53 is conserved across various vertebrate species. Studies in rats, mice, and zebrafish have demonstrated that animals with homozygous or heterozygous null p53 genotypes are prone to early onset of cancer.^{10,11,12,13,14,15} This conservation of p53 function among different species highlights its fundamental role in tumor suppression.

In “normal” cells, the levels of p53 protein are maintained at low levels primarily through the action of its negative regulator, MDM2. MDM2 functions as an E3 ubiquitin ligase, facilitating p53 degradation by tagging it with ubiquitin molecules.^16,17,18 However, p53 protein levels increase following a variety of cellular stresses, such as DNA damage and oncogene stress. This occurs because these stresses inhibit the MDM2-dependent degradation of p53.¹⁹ Upon p53 protein stabilization, p53 transcribes a variety of genes and biological effector processes to promote either elimination or arrest of injured cells to prevent progression to a cancerous state.^3,20 Among these biological processes, apoptosis (mediated by pro-apoptotic BCL-2 family members PUMA/BBC3 and NOXA/PMAIP1) and cell-cycle arrest (mediated by the CDK inhibitor p21/CDKN1A) are the most extensively studied and considered critical barriers against cancer development.^21,22 However, a study by Valente et al. utilized mice deficient for Puma, Noxa, and p21 (puma^−/−; noxa^−/−; p21^−/−) and found that these animals did not show an increased predisposition to cancer.²³ Further studies from the Gu lab, through the generation of p53 Lysine to Arginine mutants, demonstrated that loss of p53 induced cell-cycle arrest and apoptosis results in only a minor tumor predisposition.²⁴ These findings indicate that p53 is capable of suppressing tumors even in the absence of cell-cycle arrest and apoptosis; and highlighted the existence of other downstream targets and/or biological effector pathways that may be critical for p53-mediated tumor suppression.

Here, we investigated the tumor suppressive function of p53 in zebrafish by generating puma^−/−; noxa^−/−; and p21^−/− (referred to as pnp^−/−) zebrafish. Consistent to previous findings in mice, these zebrafish remained tumor-free; even those carrying the oncogenic mutation BRAF^V600E showed no tumor development. However, when subjected to DNA damage or loss of mdm2, these animals exhibited p53-dependent cell-cycle arrest even in the absence of p21, suggesting the involvement of additional transcripts in this process. To define these additional transcripts involved in p53-mediated cell-cycle arrest and other noncanonical pathways, we employed a combination of cross-species comparative transcriptomics and identified 137 potential candidate genes that are conserved between mice and zebrafish and are upregulated by p53 activation. Subsequently, through a CRISPR/Cas9 “crispant” screen that individually disrupt the top 24 cell proliferation-related candidate genes identified from the comparative analysis, we identified that, in addition to p21, ccng1, fbxw7, and foxo3b are important components in p53-dependent cell-cycle arrest.

Results

p53 efficiently suppresses tumor predisposition in the absence of puma, noxa, and p21

To further explore the conservation of the p53 tumor suppressive network across species, we generated triple knockout zebrafish for puma/bbc3, noxa/pmaip1,¹⁴ and p21/cdkn1a (Figure S1). These puma^−/−; noxa^−/−; p21^−/− triple knockout zebrafish, referred to as pnp^−/−, were then monitored for tumor predisposition (Figure 1A). Consistent with the findings in mice, none of the pnp^−/− zebrafish (N = 43) developed tumors within 450 days after birth. In contrast, all the p53^−/− zebrafish (N = 96) have succumbed to tumors, as depicted in Figure 1.¹⁴ In 2002, the Cancer Genome Project identified BRAF as the most frequently mutated gene in melanoma, with over 80% of BRAF mutations involving the hyperactivating variant BRAF^V600E.25 Previous studies showed that expressing human BRAF^V600E in zebrafish melanocytes led to ectopic melanocytic spots resembling nevi.²⁶ When combined with a p53 mutation, these nevi progressed to melanoma, demonstrating the requirement of p53 tumor suppressor pathways inactivation the the BRAF^V600E initiated melanoma.^27,28 Here, we bred BRAF^V600E into pnp−/− zebrafish and monitored them for tumor predisposition (Figure 1B). Surprisingly, despite carrying the oncogenic BRAF^V600E, triple knockouts of puma, noxa, and p21 did not develop melanoma (N = 42). In contrast, nearly all the BRAF^V600E; p53^−/− zebrafish (42 out of 44) succumbed to melanoma. This observation suggests that additional p53 transcriptional targets and/or biological processes regulate tumor suppression in the pnp^−/− background.

Figure 1.

puma^−/−; noxa^−/−; p21^−/− zebrafish are not predisposed to spontaneous tumors

(A) Kaplan-Meier tumor-free survival of p53^−/− (blue curve, N = 96, T50 = 261 days) zebrafish compared with puma^−/−; noxa^−/−; p21^−/− (called pnp^−/−, N = 43, green) and wildtype allele (N = 96, orange).

(B) Kaplan-Meier tumor-free survival of BRAF^V600E; p53^−/− (blue curve, N = 44, T50 = 271 days) zebrafish compared with BRAF^V600E; puma^−/−; noxa^−/−; p21^−/− (called BRAF^V600E; pnp^−/−, N = 42, green), BRAF^V600E; p21^−/− (N = 52, green), and wildtype allele (N = 44, orange). Long-rank statistic test was done. ∗∗∗∗, p-value between p53^−/− and pnp^−/− < 0.0001 and p-value between p53^−/− and p53^+/+ < 0.0001. ∗∗∗∗, p-value between BRAF^V600E; p53^−/− and BRAF^V600E; pnp^−/− < 0.0001, p-value between p-value between BRAF^V600E; p53^−/− and BRAF^V600E; p21^−/− < 0.0001, and BRAF^V600E; p53^−/− and BRAF^V600E; p53^+/+ < 0.0001.

Animals that have lost of puma, noxa, and p21 are resistant to p53-dependent apoptosis

Before assessing other biological processes that may be involved in p53 tumor suppression, we wanted to assess the deficiency of p53-dependent apoptosis and cell-cycle arrest in our pnp^−/− zebrafish. Previously, we had demonstrated a robust apoptotic response in the neural tube of 1-day post fertilization (dpf) zebrafish embryos 6 hours post irradiation (hpi) with 30 Gy. This apoptosis response was absent in p53^−/− or puma^−/− embryos.^13,14 In this study, we performed cleavage caspase 3 staining on both ionizing radiation (IR) irradiated and un-irradiated wildtype, puma^−/−, noxa^−/−, pnp^−/−, and p53^−/− zebrafish embryos at 1, 3, and 6 hpi (Figures 2A and 2B). Comparing these results to wild-type zebrafish embryos at 30 hours post fertilization (hpf), we observed that p53^−/−, puma^−/−, noxa^−/−, and pnp^−/− embryos displayed resistance to p53-mediated apoptosis in response to irradiation. This indicated that puma and/or noxa are key mediators in p53-controlled apoptotic processes, and the pnp^−/− zebrafish exhibit resistance to p53-dependent apoptosis.

Figure 2.

Loss of puma, noxa, and p21 provide resistance to p53-mediated induction of apoptosis and partially resistance to p53-mediated cell-cycle arrest

(A) Experimental workflow showing how samples were harvested. 29-, 27- and 24-h post fertilization (hpf) wildtype, puma^−/−; noxa^−/−, pnp^−/− and p53^−/− zebrafish embryos were treated with 30 Gy IR-irradiation and fixed at 1-, 3-, 6-, 9- and 12-h post IR-treatment (hpi, 1hpi, 3hpi and 6hpi panels).

(B) Representative images of anti-active Caspase-3 staining on 30-hpf zebrafish embryos for each group. Arrows in WT points out active apoptotic area in head region at 3 and 6 hpi. Scale bar: 500μM.

(C) Representative images of phospho-histone H3 (pH3)-stained 30-hpf (1 and 3 hpi) or 36-hpf (12 hpi) zebrafish embryos for each group. Experimental design showing in Figures 2A and S2A. Scale bar: 500μM.

(D) Quantification of pH3 positive cells in treated and untreated WT, pnp^−/− and p53^−/− embryos for each group. Each dot represents an individual. The average number of pH3+ cells (Mean) were indicated in each group. Bars represent mean ± SEM. ∗, p < 0.05. ∗∗∗, p < 0.001.∗∗∗∗, p < 0.0001.

p53 dependent cell-cycle arrest in the absence of p21

To investigate p53-dependent cell-cycle arrest, we performed whole-embryo staining for Phospho-Histone H3 (pH3), a marker for M phase cells, on embryos with and without IR-irradiation at different time points (Figure 2A). We aimed to determine whether pnp^−/− embryos were resistant to p53-mediated cell-cycle arrest. At 30 hpf, embryos were fixed at 1- and 3-h post 30 Gy irradiation, followed by pH3 staining. Confocal imaging was then used to examine the number of pH3-positive cells in each embryo (Figures 2C and 2D). A decrease in pH3 staining at 1 and 3 h time points indicates cell-cycle arrest at the G2/M transition. We observed a substantial reduction in pH3-positive cells in wildtype (WT) control embryos after IR irradiation (53 and 61 pH3-positive cells at 1 and 3 hpi, respectively, compared to 334 in untreated embryos). However, the reduction in pH3-positive cells was less pronounced in both pnp^−/− (86 or 82 at 1 or 3 hpi, respectively) and p53^−/− zebrafish (104 or 102 at 1 or 3 hpi, respectively) (Figure 2D). These findings suggest two important conclusions: First, p53 is not the sole mediator of IR-induced cell-cycle arrest at the G2/M checkpoint, as evident by the presence of ∼100 pH3-expressing cells in p53^−/− embryos following IR treatment. Second, amongst p53-dependent arrest mechanisms, puma, noxa, and p21 are only partially functional in p53-dependent arrest, as indicated by the modest reduction in pH3-positive cells (∼80 versus ∼100) in the pnp^−/− embryos. This suggests that while p21 is an essential mediator for p53-mediated cell-cycle arrest, it is not the exclusive factor responsible for this process.

To access cell-cycle arrest at other phases of the cell cycle, we continued tracking the number of pH3-expressing cells at later timepoint (6, 9, and 12 hpi). Embryos at 24 hpf were exposed to 30 Gy IR, and after 6, 9, and 12 h, the number of pH3-positive cells was examined (Figures 2C, 2D, and S2). Changes in pH3 staining at the 9- and 12-h time points will reflect predominantly G1/S arrest. At 12 hpi, WT embryos exhibited significantly reduced pH3 staining compared to the untreated (266 vs. 361). In contrast, both p53^−/− and pnp^−/− embryos showed pH3 staining levels similar to the untreated group (422 and 387 versus 368 and 386, respectively). These findings suggest that the p53-p21 axis plays a role in G1 arrest; however, the absence of puma, noxa, and p21 does not completely abolish p53-dependent cell-cycle arrest, suggesting the involvement of additional p53-dependent mechanisms.

To further investigate the role of p21 in p53-regulated cell-cycle arrest, we utilized an mdm2 knockout strategy to activate p53 independently of stress signaling pathways. MDM2 is an E3 ubiquitin ligase that targets p53 for degradation, and its loss results in p53 protein accumulation and activation of p53 target genes.^16,17,18 The concurrent elimination of p53 rescues the embryonic lethality induced by mdm2 deletion in both mice and zebrafish (Figures 3A and 3B),^{14,29,30,31,32} establishing the lethality as p53-dependent. Previously, we demonstrated that mdm2-null zebrafish embryos exhibit early morphological defects and extensive apoptosis, which can be partially rescued by the loss of puma.¹⁴ However, embryonic lethality persists, suggesting the potential involvement of p53-dependent cell-cycle arrest. To address this, we generated quadruple knockout embryos for mdm2, puma, noxa, and p21 (referred as to mpnp^−/−). While these embryos showed improved morphological appearance compared to mdm2^−/− embryos, they still experienced developmental delays (Figure 3B), suggesting that the absence of puma, noxa, and p21 only partially rescues the embryonic lethality in mdm2-null embryos.

Figure 3.

Loss of p21 partially rescues p53-dependent mdm2-null induced cell-cycle arrest

(A) The conceptional diagram of mdm2-null induced embryonic lethality. Loss of mdm2 elevates p53 protein levels to induce downstream targets and effector functions to render the lethality.

(B) Representative gross images of 24-hpf mdm2^+/+, mdm2^−/−; puma^−/−; noxa^−/−; p21^−/− (mpnp^−/−) and mdm2^−/−; p53^−/− embryos. Scale bar: 500μM.

(C) pH3-stained mdm2^+/+, mdm2^−/−; mpnp^−/− embryos at 12-, 16-, 20- and 24-hpf. Scale bar: 200μM.

(D) Quantification of pH3 positive cells at 12 hpf (Top panel) and at 24 hpf (bottom panel). Each dot represents an individual. Bars represent mean ± SEM. ∗∗∗, p < 0.001.∗∗∗∗, p < 0.0001. Not statistical significance between mdm2^−/− and mpnp^−/− at 24 hpf.

We further stained mitotic cells in wild-type, mdm2^−/− and mpnp^−/− zebrafish embryos using a pH3 primary antibody at different time points (12, 16, 20, and 24 hpf) to assess the effect of p21 loss on cell-cycle arrest in the absence of mdm2 (Figures 3C and 3D). We observed a significant reduction in the number of cells undergoing cell-cycle arrest in the absence of puma, noxa, and p21 at 12 hpf (Figure 3D, 124 pH3-positive cells in mpnp^−/− vs. 54 in mdm2^−/−). However, compared to WT controls, a considerable number of cells were still arrested (124 in mpnp^−/− vs. 289 in +/+). This indicates that while p21 is important in the initial p53-dependent cell-cycle arrest, there are other p53 target genes capable of inducing cell-cycle arrest in the absence of p21. Furthermore, we quantified the number of pH3-positive cells in each embryo at 24 hpf and found no significant difference between mdm2^−/− and mpnp^−/− embryos (Figure 3D). Taken together, these results indicate that although p21 is involved in the initial stage of p53-dependent cell-cycle arrest, other p53 target genes are also capable of triggering cell-cycle arrest in the absence of p21.

Transcriptional analysis on a pnp^−/− background significantly reduced the number of p53 downstream genes

To define p53-dependent cell cycle regulatory transcripts, we first performed RNA-seq analysis of wild-type 30-hpf embryos 3 h after 30 Gy IR-irradiation and without IR (Figure S3). We identified 449 differentially expressed genes (DEGs) with a fold change (FC) cutoff of ≥2 or ≤ −2, p-value <0.05 and baseMean ≥100 (Figure S3B). Considering that many of these DEGs may be secondary or tertiary to the primary p53 targets and/or apoptosis/cell-cycle arrest, we again performed RNA-seq analysis of 30-hpf pnp^−/− embryos with and without IR at 3 hpi. With the same cutoff, we identified 162 DEGs (Figure S3B). This suggested that ∼72% of DEGs in the wild-type datasets were secondary to puma, noxa, and p21 and/or apoptosis/partial cell-cycle arrest; and encouraged future analysis in the pnp^−/− background. We did identify 36 DEGs present in the pnp^−/− datasets but not in the WT dataset (Figure S3C). However, we observed that most of these DEGs were just below the threshold (FC ≥ 2 or ≤ −2) in the WT dataset. Consequently, in subsequent analyses, we employed a reduced FC cutoff (≥1.5) to delineate primary p53 targets.

261 p53 unregulated genes were defined in early response to irradiation in zebrafish

Our pH3 staining results demonstrated that p21 is not the only mediator in p53-regulated cell-cycle arrest at 1 and 3 hpi (Figures 2C and 2D). Therefore, to distinguish p53-dependent transcripts in pnp^−/− embryos that were not induced in p53^−/− embryos, we conducted RNA-seq analysis on pnp^−/− and p53^−/− embryos at 1 and 3 hpi. To avoid the identification of developmentally related DEGs, all embryos (∼30–35 embryos per sample) were collected at 30 hpf and subjected to irradiation at either 29-hpf (1 hpi) or 27-hpf (3 hpi). We set the threshold for DEGs at fold change of ≥1.5 (noting this lower cutoff compared to the previous analysis) and a q-value of <0.05 for each comparison. When comparing pnp^−/− treated embryos with untreated embryos, we identified 76 upregulated DEGs at 1 hpi and 324 DEGs at 3 hpi (Figure 4C; Tables S1 and S2). Volcano plots comparing pnp^−/− treated versus untreated at 1 hpi and 3 hpi revealed that well-established p53 targets such as bbc3/puma, cdnk1a/p21, and mdm2 are significantly upregulated at both time points (Figures 4A and 4B). To determine if the identified DEGs are p53-dependent, we conducted additional RNA-seq analyses on 30-hpf p53^−/− embryos, both with and without IR treated, at 1 and 3 hpi. Fold change in pnp^−/− embryos (treated versus untreated) at 1 and 3 hpi were compared to those in p53^−/− embryos at the corresponding time points. Genes with a fold change difference of ≥1.5 were classified as p53-upregulated DEGs in response to IR-irradiation. With this approach, we identified 60 genes as p53-induced DEGs in pnp^−/− treated versus untreated embryos at 1 hpi and 242 at 3 hpi (Figure 4C; Tables S1 and S2). By merging these, we identified 264 genes as p53-induced following IR-irradiation in zebrafish (Figure 4D). Among the identified genes, well-known p53 targets such as gadd45aa, mdm2, and ccng1 were induced and demonstrated a progressive increase in expression levels in pnp^−/− datasets in response to IR treatment, but not in p53^−/− datasets (Figure 4E). Of these, 204 genes were exclusively identified at the 3-hpi timepoint, suggesting that they either represent late-induced p53-dependent genes or potentially secondary targets that are downstream of the primary p53-transcribed genes.

Figure 4.

Defining IR induced zebrafish early responsive p53-upregulated genes

(A and B) Volcano plots showing 30-hpf zebrafish pnp^−/− embryos with the treated versus untreated at 1 (A) and 3 hpi (B). The cutoff was set as fold change ≥2 or ≤ −2 and p value <0.05. Upregulated DEGs were color-labeled with magenta and the downregulated were labeled with blue. The gene symbol of some TOP DEGs was indicated on the plot. The -log₁₀(p-value) of phlda3, foxo3b and mdm2 treated versus untreated at 3 hpi is above 300. Their log₂(Fold change) values were pointed out (top right square).

(C) Schematic of the method used to create Venn diagrams for p53-upregulated DEGs in pnp^−/− at 1 and 3 hpi (left panel). Venn graphs for the DEGs (right panel). The cut-off is fold change ≥1.5 and q < 0.05.

(D) Venn graph showing 264 p53-induced genes in pnp^−/− between 1 and 3 hpi.

(E) Representative plots showing well-established p53 targets, including gadd45aa, mdm2 and ccng1, in pnp^−/− but not p53^−/− datasets in response to IR treatment.

2,442 IR-irradiation induced p53 upregulated early response genes in mice

The absence of p53 in both mice and zebrafish led to the spontaneous development of tumors with 100% penetrance,^10,13,14,33 demonstrating that the transcriptional networks regulated by p53, which prevent tumor formation, are conserved across these species. To narrow down the list of potential p53 target gene candidates, we conducted a cross-species comparative transcriptomic analysis using datasets from both zebrafish and mice. We performed RNA-seq analysis on E9.5 days postcoitum (dpc) p53^+/+ and p53^−/− mouse embryos, a developmental stage roughly equivalent to that of zebrafish embryos at 30 hpf based on organ development, both with and without exposure to 30 Gy IR-irradiation (Figures S4A and S4B). Consistence with our zebrafish findings, the volcano plots showed that Bbc3, Cdkn1a, and Mdm2 were all induced within 1 h and showed increased expression at the 3-hpi timepoint (Figures S4C and S4D). In comparison to the untreated, we identified 55 upregulated DEGs at 1 hpi and 1685 at 3 hpi in mice. Among these, 49 DEGs at 1 hpi and 1602 DEGs at 3 hpi were determined to be p53-dependent (Figure 5A; Tables S3 and S4), with a combination of p53-induced genes count at 1617 (Figure 5B). Conceptually, like in zebrafish, in mice, there were 34 DEGs present at both the 1-hpi and 3-hpi timepoints, 15 DEGs unique to the 1hpi timepoint, and 1568 DEGs exclusive to the 3hpi timepoint (Figure 5B). Many of the DEGs identified exclusively at the 3-hpi timepoint likely represent secondary transcripts resulting from p53 transcripts or biological outcomes. Unfortunately, in contrast to our work with zebrafish, we did not have access to Puma^−/−; Noxa^−/−; p21^−/− mice, which would have allowed us to filter out these secondary targets following p53 activation.

Figure 5.

Defining conserved p53-upregulated genes in zebrafish and mouse

(A) Venn graphs representing p53-upregulated DEGs in mouse p53^+/+ at 1 and 3 hpi. The cut-off is fold change ≥1.5 and q < 0.05.

(B) Venn graph showing 1,617 p53-induced genes in mouse p53^+/+ between 1 and 3 hpi.

(C) The diagram showing the analysis on mouse orthologs of p53-upregulated DEGs in pnp^−/− zebrafish embryos at 1 and 3 hpi. For 264 p53-upregulated DEGs defined in zebrafish at 1 or 3 hpi, 247 of them are with mouse orthologs. Among them, 226 genes have one ortholog, and 21 of them are with multiple orthologs. 12 did not define orthologs in mouse. Five of them are non-coding genes. And 247 zebrafish p53-upregulated DEGs are corresponding to 323 mouse orthologs and 1,804 mouse paralogs. Among them, 232 genes are upregulated in mouse WT but not in p53^−/− treated versus untreated. Finally, defining 137 zebrafish p53-induced DEGs are also conserved upregulated by p53 in mouse.

137 conserved p53-upregulated genes were identified in mouse and zebrafish

To define p53-upregulated DEGs conserved between the two species, we referred to the Alliance of Genome Resources database (https://www.alliancegenome.org/), and Ensembl Genome Browser dataset (https://useast.ensembl.org).³⁴ Our analysis revealed that out of the 264 p53-upregulated DEGs identified in zebrafish, 247 have one or more orthologs in mice, corresponding to 323 mouse genes (Figure 5C; Tables S5 and S6). It is intriguing that in some instances, one or more paralogs, rather than the mouse ortholog(s), of certain zebrafish DEGs are upregulated by p53. For example, in zebrafish, sesn1, sesn3, and sesn4 (si:zfos-80g12.1) ranked highly among the upregulated DEGs, yet in mice, only Sesn2 showed significant induction (Figure S5). To analyze the conservation of gene regulation more thoroughly by p53, we identified 1,804 paralogs corresponding to the 323 mouse orthologs (Figure 5C; Table S6) and assessed whether these ortholog/paralogs are p53-induced genes in mice. Of these, 232 were identified as p53-upregulated in mice (Table S7), which correspond to 137 DEGs in zebrafish (Figure 5C; Table S8). This identifies the 137 conserved genes that are upregulated by p53 in both zebrafish and mice, potentially contributing to the tumor-suppressing activity of p53. From our analysis of the dataset, several unique observations were made: (1) 27 conserved zebrafish DEGs only correspond to p53-upregulated mouse orthologs, including genes like ccng1, phlda3, tp53inp1, and mdm2. (2) 32 conserved zebrafish DEGs have both orthologs and paralogs induced by p53 in mice. For example, ptp4a3a is a p53-induced gene in zebrafish, and it corresponds to three p53-upregulated genes in mice: Ptp4a3 (the orthologous gene), as well as Ptp4a1 and Ptpdc1 (the paralogous genes). The 59 gene can be identified through conservation analysis alone, without the need to consider paralogs analysis. (3) 78 conserved zebrafish DEGs are associated with one or multiple p53-upregulated paralogs in mice, but not with their ortholog. For instance, while the zebrafish p53-dependent gene isg20 was identified, its paralog Aen (sharing 48% sequence identity with zebrafish isg20) was p53-induced in mice, rather than the ortholog Isg20. Fischer et al. identified AEN as a target gene upregulated by p53, as evidenced in 11 out of 16 human genome-wide datasets.³⁵ Investigating the expression of the zebrafish p53-induced genes corresponding to the p53-induced paralogs in mice would provide valuable insight into the conservation of the regulatory networks of p53 across species. Without considering the mouse paralogs, we would have failed to identify these 78 DEGs are conserved, thereby potentially missing significant genes mediated by p53. We are also curious about the expression of the zebrafish orthologs for these p53 induced mouse paralogs. All of zebrafish orthologs were non-DEGs upon p53 regulation, except for one gene, Txnip. The zebrafish ortholog Txnipa was just below the FC cutoff, registering an FC of 1.48 and a highly significant q-value of 1.37E-14.

Furthermore, Gene Ontology (GO) analysis on the 137 conserved p53-regulated DEGs identified several significant biological processes and KEGG pathways, with regulation of cell-cycle ranking highly among them (Figure S6A). To enhance our time-course analysis, we collected samples from pnp^−/− zebrafish embryos that were subjected to 30 Gy treatment and compared them to untreated embryos at 2 hpi (Figure S6B). Hierarchal clustering accurately illustrated the progression of gene expression changes at 1, 2 and 3 hpi (Figure S6B). For a more detailed investigation of the dynamics of gene expression, we organized them into eight clusters (Figure S7). Clusters 1, 3, 5, and 6 show a continual increase in gene expression, but with different dynamics (Figure S7A). Clusters 7 and 8 show a delayed increase, with Cluster 7 peaking at 3 hpi and Cluster 8 at 2 hpi (Figure S7A). Cluster 2 has a unique profile, with an early increase followed by a decrease by 3 hpi (Figure S7B). Cluster 4 exhibits an increase up to 2 hpi, followed by a plateau at 3hpi (Figure S7A). These subclusters demonstrate variations in transcriptional dynamics that are likely reflective of differences in p53 promoter interaction, transcriptional co-regulation, or changes in RNA stability.

We next highlighted 24 of the 132 DEGs associated with the regulation of the cell cycle or proliferation, identified either through GO term analysis (Figure S6A) or manual curation based on relevant literature. Many of these 24 p53-induced genes are also involved in a variety of downstream effector processes that extend beyond the regulation of the cell cycle (Figure S6C); e.g., the gene gadd45aa is involved in multiple cellular processes, including p53 signaling pathway, FoxO signaling pathway, regulation of cell cycle, positive regulation of apoptotic processes, regulation of DNA-templated transcription, cellular senescence, cell differentiation, MAPK signaling pathway, and DNA damage response. These 24 genes are denoted in Figures S6B and S7A, with their mRNA expression levels progressively increasing over time.

Defining the IR-induced p53-dependent transcriptional network in response to mdm2 loss in vivo

Studying cell-cycle arrest in the mpnp^−/− mutants provides multiple advantages: (1) the activation of p53 is in the absence of cellular stress signal and (2) mdm2-null induced cell-cycle arrest is in a p53-dependent manner. To determine which of the 137 conserved p53 target genes are implicated in cell-cycle arrest and/or embryonic lethality induced by mdm2-null mutations, we analyzed RNA-seq data comparing mpnp^−/− with its sibling controls (referred to as mpnp^+/+ that includes both mdm2^+/+; pnp^−/− and mdm2^+/−; pnp^−/− genotypes) at 18 hpf. 18 hpf represents the earliest stage at which mpnp^−/− embryos can be morphologically distinguished from mpnp^+/+ embryos. We identified 2,582 DEGs that were upregulated in the mpnp^−/− group compared to the mpnp^+/+ group, with a fold change of ≥2 and a q-value of <0.05 (Figure 6A; Table S9). The extensive number of DEGs includes direct targets of p53 but also contains numerous secondary and downstream genes induced as a result. This is due to the use of morphological differences to identify mdm2-null animals long after the initial molecular changes have occurred; for example, cell-cycle defects were observable as early as 12 hpf (Figure 3C). Among the 137 conserved p53-dependent IR-induced DEGs, 112 were also present in the DEGs of mpnp^−/− versus mpnp^+/+ at 18 hpf, demonstrating strong conservation of these genes between both p53 activation methods. The 24 genes of interest (GOIs) involved in the cell cycle or proliferation regulation are among the 112 overlapped DEGs.

Figure 6.

Comparing 137 conserved p53 dependent IR induced genes with DEGs in mpnp^−/− datasets

(A) Venn graph displaying the overlapping genes between 137 conserved UP DEGs in both zebrafish and mouse with IR-irradiation and 2,582 upregulated (UP) DEGs in mpnp^−/− versus sibling controls at 18 hpf. Experimental timeline showing the timepoint that distinguish mpnp^−/− embryos from sibling controls and how to harvest RNA samples at the time (Top panel).

(B) Heatmap showing transcriptional changes for 2,582 DEGs over time in early mpnp^−/− datasets (8, 10, 12, 14, and 16 hpf). Each timepoint was measured in duplicate, and the averages of the duplicates were used for each group. Z-scores, calculated from TPM values calculated across all samples, were used to standardize gene expression levels before clustering. The experimental workflow for sample collection is shown in the top panel. Note that mutant samples at each timepoint were diluted 4-fold by their sibling controls (¼ mpnp^−/− and ¾ sibling controls). Genes with similar expression patterns over time were grouped into eight clusters, with their trends and cluster sizes shown in the right panel.

(C) Line graphs showing the kinetics of 8 out of 24 GOIs in early mpnp datasets, with the dashed line indicating a fold change of 2. Expected counts calculated with RSEM were used for creating line graphs. Line graphs for the remaining 16 GOIs are shown in Figure S8B.

To define direct and likely early-induced transcripts, we performed RNA-seq analysis on pools (N = ∼50 embryos) of un-genotyped progeny from mdm2^+/−; pnp^−/−intercrosses at various developmental stages (8, 10, 12, 14, and 16 hpf) as the mutant group (“Mutant” in Figure 6B), with same-stage progeny from pnp^−/− intercrosses serving as the control group (“Control” in Figure 6B). While this approach generated a dilution of RNA transcripts, since only ¼ of the embryos were mpnp^−/−, it nonetheless allowed us to glimpse at the dynamics of p53 target gene induction in mpnp^−/− animals. Without accounting for dilution effects, we created a heatmap using Transcripts Per Million (TPM) values to represent these DEGs upregulated in mpnp^−/− at 18 hpf (Figure 6B). We identified eight distinct clusters, with clusters 5 and 7 being the most predominant. The 24 GOIs are distributed across all eight clusters, ranging from one gene in Cluster 8 to five genes in Cluster 6. (Table S10). For the 24 GOIs, we produced individual plots (Figures 6C and S8) to visualize and analyze their expression patterns. We utilized an equation to compensate for the dilution and to monitor the changes more accurately in gene expression (Figure S8A). The equation is founded on two premises: first, that one-fourth of the embryos in the mutant samples are mpnp^−/−; second, that the gene expression levels in embryos with the genotypes mdm2^+/−; pnp^−/− (referred to as m^+/−pnp^−/− in Figure 6B) and mdm2^+/+; pnp^−/− (denoted as m^+/+pnp^−/−) are equivalent to those in pnp^−/−. While the expression trends for many of these genes are distinct, a general observation is that all of them show a continuous increase in expression from the 8 hpf to the 16 hpf timepoints (Figures 6C and S8B). This increase occurs well in advance of any observable morphological differences. Most importantly, the induction of these 24 genes occurs early, with fold changes reaching or exceeding 2 prior to 12 hpf, suggesting their potential role in driving the cell-cycle arrest observed in mpnp^−/− embryos.

fbxw7, foxo3b, and ccng1 partially rescue p53-dependent cell-cycle arrest in the absence of p21

We have previously demonstrated the use of CRISPR/Cas9 F0 “crispant” analysis to rapidly define genes responsible for ciliopathies in zebrafish.³⁶ To evaluate whether this technique could be applied to mitigate the embryonic lethality associated with mdm2-null mutations, we initially assessed if targeting p53 in crispants could rescue the mdm2-null phenotypes. The co-injection of Cas9 protein along with four p53-specific guides (forming p53 crispants) into mdm2-null embryos successfully rescued the lethal phenotype (Figure S9A). About 25% of the injected mdm2 null embryos appeared morphologically normal at 24 hpf, resembling zygotic wild-type or mdm2^−/−; p53^−/− embryos. The rest of the injected embryos showed significant rescue, albeit with some minor deformities that were distinguishable under microscope examination (Figure S9A). Subsequently, we stained for mitotic cells using the pH3 antibody in both p53 crispants and un-injected mpnp^−/− embryos, as well as their sibling controls at 21 hpf (Figure S9B). We observed that F0 crispants injected with p53 did not undergo cell-cycle arrest, mirroring the behavior of un-injected wildtype embryos (18 pH3-expressing cells in un-injected mpnp^−/− embryos compared to 263 in p53-injected mpnp^−/− and 266 in p53-injected mpnp^+/+ sibling controls). To quantify the extent of rescue from cell-cycle arrest, we introduced the concept of the rescue ratio (RR), which calculates the percentage of the average number of pH3-expressing cells in mutant embryos compared to their sibling controls (Figure S10). An RR value that approaches 100 indicates a more complete rescue of cell-cycle arrest, signifying that the intervention effectively restored cell-cycle progression to levels similar to those in control embryos. The calculated RR for p53-injected group was 99.21, suggesting a nearly complete rescue of cell-cycle arrest in mpnp^−/− embryos. These findings collectively suggest that CRISPR/Cas9 crispants, injected with four specific guides, can effectively be used to rescue both the cell-cycle arrest and lethal phenotype associated with mdm2 loss.

Building on the success of the pilot CRISPR/Cas9 crispant experiments, we proceeded to examine if injection of a four-guide cocktail against any one of the 24 GOIs could rectify the morphological abnormalities observed in mpnp^−/− embryos. The crispants generated for each of the 24 GOIs failed in rescuing the embryonic lethality observed in mpnp^−/− embryos (Figure S11). We next tested whether any of the crispants for the 24 GOIs could influence the number of pH3-positive cells in mpnp^−/− embryos. Amongst them only the crispants targeting fbxw7, foxo3b, and ccng1 showed an increase in the number of pH3-positive cells in mpnp^−/− embryos. Specifically, the number of pH3-positive cells increased from 18 in the control group to 53, 40, and 41 in the corresponding crispant groups (Figures 7A and 7B). Representative images of embryos injected for each gene are present in Figure S12. The RR values for these genes were calculated and ranked to show their respective impacts on cell-cycle arrest (Figure S10). fbxw7, foxo3b, and ccng1 have RR values of 19.21, 16.18, and 15.77 respectively. Notably, si:dkey-204L11.1 (also known as tnfrsf27, corresponding to the human EDA2R gene and referred to as eda2r in the figures), fosab, tob1a, and slc4a2a showed fewer pH3+ cells and a reduced body area compared to uninjected controls, as shown in Figures 7B and S10–S12. This finding suggests that these genes may play an important and independent role in development. Interestingly, in the embryos injected with mmrn2b- and tp53inp1-guides, we observed more developed embryos (Figure S11), characterized by the beginning of eye formation and the presence of more somite. This suggests that these genes may play a role in the observed lethality independent of the cell-cycle defect.

Figure 7.

fbxw7, foxo3b and ccng1 G0 crispants mitigate p53-mediated cell-cycle arrest

(A) Representative images showing pH3-stained un-injected (control, un-inj) and injected mpnp^−/− embryos at 21 hpf. Scale bar: 250μM.

(B) Quantification of pH3 positive cells in injected mpnp^−/− embryos for 24 GOIs. un-inj (negative control) and the p53 guides-injected (p53, positive control).

(C) Representative images representing pH3-stained, IR-irradiation treated or untreated, four-guide injected pnp^−/− embryos at 30 hpf. Scale bar: 500μM.

(D) Quantification of pH3 positive cells in injected pnp^−/− embryos for fbxw7, foxo3b and ccng1. p21-inj (negative control) and p53-inj (positive control). Each dot represents an individual. Bars represent mean ± SEM. ∗, p < 0.05. ∗∗, p < 0.01.∗∗∗, p < 0.001.∗∗∗∗, p < 0.0001.

To broaden our research, we explored the influence of fbxw7, foxo3b, and ccng1 on IR-induced, p53-dependent cell-cycle regulation. Our goal was to determine whether the deletion of fbxw7, foxo3b, and ccng1 in pnp^−/− embryos could confer resistance to IR-induced, p53-mediated cell-cycle arrest. Crispant-injected embryos were fixed at 30 hpf, 3 h after exposure to 30 Gy IR irradiation. This time point was selected based on previously findings that p21 is not the sole gene involved in p53-mediated cell-cycle arrest. Following fixation, the embryos were stained with a pH3 antibody to assess cell-cycle regulation (Figures 7C and 7D). Embryos injected with p21 in the pnp^−/− background served as an injection-negative control to establish the baseline level of proliferative cells. Meanwhile, pnp^−/− embryos injected with p53 were used as positive controls, illustrating the extent of p53-mediated cell-cycle arrest. With this setup, we revealed that foxo3b and ccng1 were able to partially rescue the IR-induced, p53-mediated cell-cycle arrest. On the other hand, fbxw7 exhibited a more profound effect by completely counteracting the IR-induced, p53-mediated cell-cycle arrest.

Discussion

Multiple transcriptomic studies have been performed to define p53 target datasets. Here, we compare and contrast several of these studies that have indicated p53 core gene sets. In human cancer cell lines, the Fischer group identified a p53 core of 116 genes by analyzing common genes induced across 16 published p53 transcriptional datasets. These datasets were derived from human cancer cell lines (HCT116, MCF7, U2OS, Cal51, and HFK) exposed to various p53-inducing agents, including 5-FU, etoposide, doxorubicin, IR, and Nutlin-3 (a mdm2 inhibitor) which, when combined, these datasets suggest a total of 3,509 potential p53 target genes.³⁵ Additionally, the Espinosa group defined a p53 core of 103 genes through transcriptional comparison between p53 wild-type and p53-null human cell lines (HCT116, MCF7, and SJSA) following exposure to Nutlin-3.³⁷ This analysis integrated multiple omics approaches, including GRO-seq to define actively transcribed genes, CHIP-seq to map p53 binding sites in the genome, and polysome-bound RNA-seq to define actively translated mRNAs. In a separate study, the Espinosa group also investigated p53 transcriptional targets using “normal” (non-tumor) human IPSCs and two differentiated cell populations.³⁸ In this study, they defined a 49-gene core that is consistently induced in IPSCs and the two differentiated cell populations, out of 677 induced p53-induced genes that were upregulated in at least one dataset. This study draws attention to the necessity of examining multiple cell types to fully elucidate the complete repertoire of p53 tumor-suppressive functions. The Lozano group utilized conditional, tissue-specific mouse knockouts of Mdm2 to define transcriptional profiles of p53 target gene across five different tissues.³⁹ Interestingly, they identified only seven upregulated genes common across these tissues, highlighting the transcriptional variability of p53 target genes in different cell types. By refining their analysis to include genes upregulated in at least two of five tissues, a core set of 143 genes was identified.

Our dataset is unique in that (1) we utilized puma^−/−; noxa^−/−; p21^−/− zebrafish, which excluded a significant portion (∼72%) of secondary and tertiary targets typically activated in response to p53, thereby refining the focus to primary targets. (2) We studied DEGs at 1- and 3-h post-IR, enabling the identification of early p53-responsive genes while minimizing secondary transcript interference. (3) We analyzed both mouse orthologs and paralogs of zebrafish p53-upregulated DEGs, preventing the omission of important target genes (78 of the 137). (4) Our analysis used whole embryos from both mouse or zebrafish embryos, which captures the cellular complexity and normal physiology of the organism. This cross-species approach led to the identification of a 137-gene core set.

When we compare direct orthologs (Venn diagram Figure S13), our core set of 137 genes shares 18 genes with the Espinosa 103, 24 with the Fischer 116, 16 with Espinosa core 49, and 15 with the Lozano core 143. When comparing the five described cores, only seven genes are common across all five core sets. Note Lozano’s seven core genes across five different tissues are partially encompassed within these seven genes. 14 genes are shared by four of the five cores (nine of which are included in our core), 16 genes are shared by three of the five cores (six of which are in our core), and 53 genes are shared by two of the five cores (six of which are in our core). Notably, the majority of these 53 genes (36 of 53) are shared between the Espinosa 105 and Fischer 116 cores, primarily due to their overlap in cancer cell line datasets. The differences among these datasets could reflect several factors, including in vivo studies verse cell culture studies; species differences (mouse, human, or zebrafish); cell-of-origin variations; and timing after induction, as studies span from 1 h to 24 h post treatment. Additionally, methodological differences in bioinformatics analyses contribute to variations. For example, some studies filter for genes with CHIP-seq-based p53 binding sites near promoters, with thresholds such as ±2.5 kb (Fischer 116 and Espinosa 103) or ±10 kb (Lozano); Specialized sequencing methods like GRO-seq, which focus on actively transcribed gene, are employed in studies such as Espinosa 103. Different fold change cutoffs are also applied; for instance, our group analyzes only p53-upregulated genes with a fold change ≥1.5. p-value cutoffs vary as well, including q < 0.05 (used in our study, Fischer, Espinosa 103, and Lozano) and q < 0.1 (used in Espinosa 49).

To expand these core gene set comparisons to the individual gene level, BAX, BBC3, CCNG1, CDKN1A, MDM2, PHLDA3, and TP53INP1 are consistently induced across our 137 core genes, as well as in Espinosa 103, Espinosa 49, Fisher 116, and Lozano 143. While FAS, GDF15, LIF, SESN2, TNFRSF10B, and ZMAT3 are commonly induced genes in Espinosa 103, Espinosa 49, Fisher 116, and Lozano 143, they are not induced in our core set. Zebrafish lack orthologs of GDF15 and LIF, as determined through Ensembl dataset analyses. Zebrafish possess orthologs of SESN2, FAS, TNFRSF10B and ZMAT3, suggesting the induction of these genes by p53 is an evolutionary adaptation. However, while zebrafish sesn2 was not induced in our analysis, sesn1, sesn3, and sesn4 (si:zfos-80g12.1) were among the top p53-induced DEGs, as identified through our paralog analysis. Additionally, BTG2, GADD45A, PLK3, RPS27L, and SESN1 are consistently induced across our 137 core, as well as in Espinosa 103, Espinosa 49, and Fisher 116, but not in Lozano 143. DDB2, RRM2B, TNFRSF10B, TRIAP1, POLH, ASCC3, CMBL, DRAM1, EPS8L2, and FDXR are induced in Espinosa 103, Espinosa 49, and Fisher 116, but are absent from Lozano 143 and in our core (or our mouse dataset). This suggests that these genes may be human-specific or associated with cell culture conditions. Notably, while POLH is not induced in mice or zebrafish, its paralog Polk is induced in mice (Lozano 143 and our mouse dataset), and its paralog rev1 is induced in zebrafish. MAP3K20, MMP2, and RNF169 are present exclusively in Lozano and our core, suggesting they may be specific to in vivo studies, non-human species, or cell types not represented in the other cores. While these different datasets define distinct cores with limited overlap, our analysis highlights the importance of comparing orthologs and paralogs across species, to fully capture the breadth of p53-regulated genes.

Other p53 transcriptional targets have been reported, including SLC7A11, SAT1, and GLS2, which are key p53-regulated genes associated with ferroptosis.⁴⁰ However, none of these three genes are induced in any of the cores analyzed. Notably, SLC7A11 is absent in both our zebrafish and mouse datasets, though its paralogs slc7a1a and Slc7a1 are induced in zebrafish and mouse, respectively. TIGAR plays a dual role in p53 regulation by modulating metabolism and reducing oxidative stress, thereby promoting cell survival under stress conditions.^41,42 Despite its critical role, TIGAR is induced only in Fischer 116 and Espinosa 49.

Furthermore, other p53 transcriptional targets have been reported to regulate the cell-cycle, such as 14-3-3σ, PCNA, PP2A, LATS2, and BubR1.³ The genes encoding 14-3-3σ, PCNA, PP2A are not induced in any of the five core datasets. The gene encoding LATS2 is not induced by IR at 1 or 3 hpi in zebrafish but is induced at 6 hpi, and it is upregulated by p53 in mice at 3 hpi. This suggests that LATS2 may be a late-responsive gene or, more likely, a secondary downstream transcript of the puma, noxa, and p21 effector pathway. Interestingly, its paralogs sgk2b and mast3a are induced in zebrafish, while Lats2, Mastl, and Mast4 are induced in mice. In contrast, Bub1b, which encodes BubR1, is inhibited (rather than induced) by p53 in both zebrafish and mice, suggesting it is likely a secondary, indirect p53 target.

Mice deficient in Puma, Noxa, and P21 do not develop spontaneous tumors, suggesting that p53-mediated apoptosis and cell-cycle arrest are dispensable for tumor prevention. More importantly, this indicates that other p53 downstream biological processes may play a critical role in tumor suppression.²³ Further supporting this, p53 3KR mutant mice, which lack the ability to induce apoptosis and cell-cycle arrest, exhibit only mild cancer predisposition.²⁴ In addition, TAD1/2 mutant mice, which lack transcriptional activity, show tumor predisposition similar to p53 null mice, indicating that p53’s tumor suppression function relies on its transcriptional activity.⁴³ Notably, TAD1-only mutant mice, which lack apoptosis and cell-cycle arrest, do not demonstrate tumor predisposition, suggesting these processes are not essential for tumor suppression. Consistent with mouse studies, our findings reveal that the loss of puma, noxa, and p21 does not predispose zebrafish to tumors. However, in contrast to the above mouse studies, we observed cell-cycle arrest even in the absence of p21, suggesting alternative p53-dependent mechanisms are at play. This discrepancy could reflect differences between mice and zebrafish; however, an important caveat in the mouse studies is that the ablation of p53-mediated cell-cycle arrest was assessed only in MEFs,^23,24,43 rather than in the whole organism. Therefore, it remains unclear whether p53-dependent cell-cycle arrest is truly absent across multiple cell types or tissues in various p53 mouse mutants. Interestingly, p53 TAD1 mutant MEFs are competent for p53-dependent senescence under oncogenic stress, which aligns with our findings. This observation does not rule out the importance of other effector pathways, including but not limited to senescence, ferroptosis, or autophagy/mTOR regulation. It also suggests that further investigation is needed to fully understand the mechanisms of p53 cell-cycle arrest. In fact, some of the other effector pathways could be impacting cell-cycle arrest either directly or indirectly, ultimately contributing to tumor suppression. Our data specifically highlights that fbxw7, foxo3b, and ccng1 play significant role in p53-dependent cell-cycle arrest, further emphasizing the complexity of these pathways.

The only relevant metric to determine which genes and effector pathways are involved in tumor suppression is to look at how loss of a gene affects tumor formation in vivo.²⁶ In this regard, the idea is to ablate the p53 target gene and monitor the cohort for tumor onset. If the ablation renders tumor formation equivalent to p53 loss, then the underlying process is crucial for p53 tumor suppression. If not, it is viewed as being unnecessary or there are redundancies. However, these experiments can require 2-3 years to investigate a single gene and require significant costly space to maintain the animals during this period. Zebrafish alleviate the majority of the burden associated with the latter but still require multiple years. The rescue of mdm2 null phenotype can be used as a surrogate for tumor survival analysis. In our opinion, this is an ideal model suited for studying p53 tumor suppression in that there are no actual oncogenic signals or stress stimuli involved, but the mis-regulation of p53 to induce transcription and effector pathways. The observed lethality in this model is due to systemic tumor-suppressive effects, thereby emulating how p53 prevents tumor initiation and progression. The mouse Mdm2 knockout model has been used to demonstrate that the loss of Bax can slightly rescue Mdm2-null mice.⁴⁴ Crossing the p53 3KR mutant mouse with Mdm2 null partially rescues the lethal phenotype, indicating that apoptosis and cell-cycle arrest are associated with the lethality, while other p53 effector pathways remain activated.⁴⁵ Consistent with these findings, our previous work demonstrated that loss of puma and noxa provides a mild rescue of mdm2-associated lethality.¹⁴ In this study, we further demonstrate that the combined loss of puma, noxa, and p21 can offers slightly greater rescue but remains early lethal. Furthermore, we leveraged the mpnp^−/− zebrafish model, which still undergoes p53-dependent cell-cycle arrest, to demonstrate that the knockout of fbxw7, foxo3b, or ccng1 can partially rescue cell-cycle defects. Note none of them rescued the morphological abnormalities, suggesting that alleviating mpnp^−/− induced lethality will likely require the inactivation of multiple p53 downstream targets and effector pathways. The Gadd45 family comprises three genes GADD45α, GADD45β, and GADD45γ, of which GADD45α is a well-known p53 target gene known to arrest cells at both G1/S and G2/M phases.^10,46,47 However, the loss of gadd45aa did not lead to an increase in pH3+ cell numbers in mpnp^−/− embryos (RR = 5.97), suggesting that gadd45aa may not serve as a key p53 downstream mediator of cell-cycle arrest in mpnp^−/− embryos. Our data, however, indicates that unlike in mice, gadd45aa, gadd45bb and gadd45ga are p53-upregulated genes in zebrafish, raising the possibility of functional redundancies within the gadd45 family.

With the mpnp^−/− model, we quantified pH3 staining results following the knockout of 24 p53-upregulated GOIs. Among these, the knockout of fbxw7 (RR = 19.21), foxo3b (RR = 15.77), or ccng1 (RR = 16.18) resulted in an increased number of pH3 positive cells, indicating a reduction in p53-dependent cell-cycle arrest. FBXW7 is a haploinsufficient tumor suppressor gene that plays a well-documented role in cell-cycle regulation through Cullin-dependent E3 ubiquitin ligase complex, mediating the proteasome-mediated degradation of key cell-cycle regulators such as cyclin E, C-MYC, and Notch. It also impacts cell-cycle progression indirectly via translational control through mTOR, which affects cyclin E levels.^48,49,50,51 Consistent with our analysis, FBXW7 has been demonstrated to be a direct transcriptional target of p53.⁵² Together, this suggests that p53 can induce FBXW7, leading to the degradation of cell-cycle regulators and subsequent cell-cycle arrest. Conversely, the inactivation of FBXW7 results in reduced cell-cycle arrest. However, FBXW7 also binds to and degrades p53 in a feedback loop similar to the one established for MDM2.^53,54,55 This raises a question: given this feedback loop, why does the loss of FBXW7 reduce cell-cycle arrest instead of enhancing it via increased p53 activation.

The forkhead box O3 (FOXO3) transcription factor is a crucial transcription regulator of multiple cellular processes, including cell-cycle arrest, cell death, DNA repair, autophagy, and aging.^56,57,58,59 Pointing the possibility that FOXO3 directly affects the cell cycle or exerts its influence indirectly through another effector pathway. In terms of cell-cycle regulation, FOXO3 induces specific cell-cycle-related molecules such as CDKN1A/p21, CDKN1B/p27, and Gadd45, which are involved in DNA repair and activation of cell-cycle checkpoints in response to DNA damage.⁶⁰ In addition, Cyclin G1, encoded by CCNG1, was one of the earliest p53 targets.^61,62,63 Previous studies have demonstrated that cyclin G1 plays an important role in cell-cycle machinery, influencing cyclin-dependent kinases (CDKs) that control progression through different phases of the cell cycle.⁶⁴

While none of these genes alone provides a complete rescue from cell-cycle arrest, the identification of three independent genes—fbxw7, foxo3b, and ccng1—suggests that they may have overlapped or compensatory roles alongside p21 in p53-dependent cell-cycle arrest. Further investigations into the effects of fbxw7, foxo3b, and ccng1 on IR-induced, p53-dependent cell-cycle arrest reveal that fbxw7 can fully counteract the p53-mediated cell-cycle arrest. These results emphasize the significant role of fbxw7 within the mechanisms employed by p53 to arrest the cell cycle, often in conjunction with p21. These findings also suggest that alleviating the morphological lethality observed in mdm2 nulls may require the involvement of extra p53-mediated downstream effector pathways. An intriguing question arises: in the absence of these three genes in a pnp^−/− background, would tumor predisposition match that of p53-null models? While this experiment, involving six gene knockouts, appears technically challenging, the genetic advantages of zebrafish make it a feasible avenue for future investigation.

Comparison of IR-induced DEGs with mpnp DEGs revealed 25 early IR-responsive genes absent in our mpnp^−/− data. For example, myl7 and myh7 are conserved p53-upregulated genes after IR-irradiation but they are undetected in our time-course mpnp^−/− data. Myl7 and myh7 encode myosin regulatory heavy chain proteins, primarily expressed in the heart, however mpnp^−/− data were harvest from 8 to 18 hpf, a timepoint when the heart is not formed. Suggesting they maybe tissue specific p53 regulated genes. In contrast, Sec14L8 is induced in the mpnp^−/− dataset as early as 8 hpf (Fold change = 6.31 and p-value = 3.32E-05) but is absent in the IR-induced dataset. This discrepancy may be attributed to the gene being involved in very early Wnt signaling and axis elongation processes. These findings underscore the importance of studying tissue specific differences that exist in the p53 network.

While mice have become the predominant model for studying p53 tumor suppression, zebrafish provide several strengths that could help understand p53 tumor suppression. One of the primary advantages of using zebrafish in the study of p53 tumor suppression is their small size and minimal housing requirements. This allows for the execution of large-scale, multiple gene knockout cohort tumor screens to delineate key tumor suppressive pathways efficiently. Furthermore, the ease in generation of gene knockouts or use of multiple guide crispant can provide rapid functional in vivo screening. Moreover, we identified 137 conserved p53-regulated zebrafish genes that correspond to 232 conserved genes in mice, highlighting an expansion of paralog induction in the mouse model. Future studies on these genes may be more feasible and efficient using the zebrafish system.

Limitations of the study

First, zebrafish offer genetic advantages, but species differences may limit p53 pathway conservation. To enhance cross-species analysis, we identified paralogs. Additionally, whole-organism p53 response was analyzed to avoid missing potential p53 targets, yet some IR-induced p53 genes were absent in mpnp^−/− embryos, suggesting potential tissue-specific or developmental stage differences. Furthermore, our focus on early p53 responses (1–3 hpi post-IR, 8–16 hpf in MPNP datasets) may have excluded later-acting p53 targets. Finally, while fbxw7, foxo3b, and ccng1 mediate p53-dependent cell-cycle arrest, their interplay with other pathways remains unclear. We identified these three genes through screening but did not generate stable knockout models to further validate and investigate their roles in p53 network. A six-gene knockout (fbxw7, foxo3b, and ccng1 in puma^−/−, noxa^−/−, p21^−/− embryos) could further assess tumor predisposition but remains technically challenging.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, John M Parant (jparant@uab.edu).

Materials availability

All materials are available upon request from the lead contact, John M Parant (jparant@uab.edu)

Data and code availability

• •Data: The RNA sequencing data that support the findings of this study have been deposited in the NCBI Gene Expression Omnibus (GEO) and are publicly available as of the data of publication. Accession numbers are listed in the key resources table.
• •Code: The original code generated for this study has been deposited at GitHub, named as “Homology Data Fetcher”, and publicly available as of the data of publication. An accession link is provided in the key resources table.
• •Additional information: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

The authors would like to acknowledge the members of the Parant lab for technical help and critical reading of the manuscript. J.M.P. is supported by NIH R01CA216108, NIH U54OD030167, and UAB School of Medicine AMC21 pilot funds. We use the following core facilities: UAB Zebrafish Research Facility, UAB Heflin Center for Genomic Science, UAB Core Grant for Vision Research (P30 EY003039).

Author contributions

J.M.P. and B.K.Y. designed and oversaw the study. H.R.T. generated zebrafish mutants determined Mendelian inheritance, and validated alleles. J.W. and R.G.T. monitored tumor cohorts. J.W. and R.G.T. performed cleavage-caspase 3 staining and imaging. J.W. and Z.L. performed pH3 staining, imaging, and quantification. J.W., Z.L., K.F., and D.K. performed bulk-RNA sequencing and analysis. J.W. and Y.M. performed the CRISPR G0 screen and phenotypic analysis. J.W., with consultation from J.M.P., made all figures and performed statistical analysis. J.M.P. and J.W. wrote the manuscript with revision by all authors.

Declaration of interests

The authors declare no potential conflicts of interest.

Declaration of generative AI and AI-assisted technologies in the writing process

No generative AI was employed in the preparation of the manuscript. J.W. employed ChatGPT only for language refinement.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
active caspase 3	BD	559565 (RRID: AB_397274)
Phospho-Histone H3 (Ser10)	Cell Signaling	9701 (RRID: AB_331535)
Goat Anti-Rabbit lgG (H+L) Antibody, Alexa Fluor 488 Conjugated	Invitrogen	A-11008 (RRID: AB_143165)
Critical commercial assays
QIAGEN RNeasy Plus Mini Kit	QIAGEN	74134
tracrRNA	IDT	1072532
Alt-R S.p. Cas9 Nuclease V3	IDT	1081058
pT3TS-nCas9n	Addgene	46757
Invitrogen mMESSAGE mMACHINE™ SP6 Transcription Kit	Fisher Scientific	AM1340
MEGAclearTM Transcription Clean Up Kit	Fisher Scientific	AM1908
LC Green Plus Melting Dye	Biofire Defense	BCHM-ASY0005
Taq DNA Polymerase (1000 U) with 10 mM dNTP Mix (0.5 ml)	Genscript	E00101
Deposited data
RNA-seq data of IR-treated mouse embryos	This paper	GSE288659
RNA-seq data of IR-treated zebrafish embryos	This paper	GSE288660
RNA-seq data of zebrafish mdm2-/-;puma-/-;noxa-/-;p21-/- embryos at 8 to 16 hpf	This paper	GSE288666
RNA-seq data of zebrafish mdm2-/-;puma-/-;noxa-/-;p21-/- embryos at 18 hpf	This paper	GSE288667
Experimental models: Organisms/strains
D. rerio strain: p53-/-	Wang et al.14	PMID: 34193827
D. rerio strain: puma-/-	Wang et al.14	PMID: 34193827
D. rerio strain: noxa-/-	Wang et al.14	PMID: 34193827
D. rerio strain: p21-/-	This paper	PMID: 34193827
D. rerio strain:BRAFV600E	Craig J Ceol et al.28	PMID: 21430779
D. rerio strain: mdm2-/-	Wang et al.14	PMID: 34193827
D. rerio strain: puma-/-; noxa-/-;p21-/-	This paper
D. rerio strain: mdm2-/-; puma-/-; noxa-/-;p21-/-	This paper
D. rerio strain: puma-/-; noxa-/-	Wang et al.14	PMID: 34193827
Mouse: p53-/-	Jackson Labs	002101
Mouse: p53+/+	Jackson Labs	000664
Oligonucleotides
Primers for zebrafish p53, puma, noxa, mdm2	Wang et al.14	PMID: 34193827
p21 Forward Primer	This paper	CAGCTGCAGCGTGAGTACC
p21 Reverse Primer	This paper	GGAAGTCTCCGCCCTCTAGT
Software and algorithms
GraphPad Prism 10	Dotmatics	https://www.graphpad.com/updates/prism-10-1-1-release-notes
Nikon NIS Element	Nikon	https://www.microscope.healthcare.nikon.com/products/software/nis-elements
Fuji	Schindelin et al.65	https://imagej.net/software/fiji/downloads
Homology Data Fetcher	This paper	https://doi.org/10.5281/zenodo.15272571
FastQC (version 0.11.7, Java-1.8.0_74)	Andrews66	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
STAR (version 2.7.3a, GCC 6.4.0-2.28)	Dobin et al.67	https://github.com/alexdobin/STAR
HTSeq (version 0.12.3, foss-2018b, Python-3.6.6)	Anders et al.68	https://htseq.readthedocs.io/en/latest/
RSEM (version 1.3.3)	Li and Dewey69	https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-12-323
R (version 4.3.3)	R Foundation	https://www.r-project.org/
R studio (2024.09.1-394)	Posit	https://posit.co/download/rstudio-desktop/
DESeq2 (version 1.42.1)	Love et al.70	https://pubmed.ncbi.nlm.nih.gov/25516281/
GOplot (version 1.0.2)	Walter et al.71	GOplot: an R package for visually combining expression data with functional analysis
DAVID Bioinformatics Resources 6.8	Huang et al.72	https://davidbioinformatics.nih.gov/
SRplot	Tang et al.73	SRplot: A free online platform for data visualization and graphing.
Zhang lab gRNA design tool	Zhang Lab, MIT	http://crispr.mit.edu
IDT predesigned CRSPR-Cas9 guide RNA Tool	Integrated DNA Technologies	https://www.idtdna.com/site/order/designtool/index/CRISPR_PREDESIGN
Other
IDT designed IDT crRNAs	This paper	Table S10

Experimental model and study participant details

Zebrafish lines and maintenance

All zebrafish (Danio rerio) work was performed in the Zebrafish Research Facility (ZRF) of the University of Alabama at Birmingham (UAB). Adult fish and embryos are maintained as described by Westerfield et al. (1995) by the ZRF Animal Resources Program which maintains full American Association for Accreditation of Laboratory Animal Care (AAALAC) accreditation and is assured with the Office of Laboratory Animal Welfare (OLAW). All zebrafish were of the AB stain. Adult zebrafish were housed in groups within 3.5-L tanks (approximately 6 fish per liter) in a circulating water system under a 14-hour light/10-hour dark cycle at 28°C. Embryos were kept in E3 medium containing 0.7 μM methylene blue, also under a 14-hour light/10-h dark cycle at 28°C. All animal studies have UAB Institutional Animal Care and Use Committee (IACUC) approval and were approved under protocol number IACUC-20705. All experiments were conducted on embryos at the specific developmental stages as described. All knock-out alleles were generated and maintained on the AB stain.

Zebrafish sex as a biological variable

Sex determination is not possible in embryonic zebrafish, so embryos were randomly selected regardless of sex.

Generation and validation of a new p21/cdkn1a knockout allele in zebrafish

The p21 knockout was generated as described previously.⁷⁴ gRNA target sites were identified using the Zhang lab gRNA design tool (http://crispr.mit.edu). The target site and PAM motif were shown in Figure S1. The procedures to prepare the gRNA and Cas9 mRNA for microinjection followed previously described protocols.¹⁴ Approximately 1-2 nL of sgRNA/Cas9 mRNA were microinjected into the yolk of one-cell-stage zebrafish embryos. For indel efficiency evaluation, genomic DNA was randomly selected to extract from approximately 24 5 dpf injected F0 embryos and evaluated with HRM curve (see below for details). The remaining injected embryos (F0s) were raised and later crossed with wild-type AB stain at ∼3 months post fertilization to generate F1 progeny. Several out of frame alleles were identified in the F1 generation. The Δ2 allele was sequenced by the UAB Heflin Center for Genomic Sciences Sanger Sequencing Core and maintained for further propagation. To “cleanup” genetic background, F1 fish were bred at least one more generation with the wildtype AB stain.

Mouse lines and maintenance

All mouse studies were conducted in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the IACUC at UAB (protocol number IACUC-20705). The p53 knockout (KO) allele was obtained from Jackson Labs (Strain #002101) and maintained on a C57BL6/J genetic background (Strain #000664), which were also purchased from Jackson Labs. Mice had ad libitum access to food and water under 12:12 light/dark cycle at 21-22°C. The breeding p53 heterozygous KO mice used in this study were housed one male and two female per cage, and fertilization was checked daily around 9:00 AM. Pregnant females were irradiated with 30 Gy at the E9.5 development stage of the embryos. Dissections were performed at the specific time points as described. A small tail biopsy of the embryos was harvested for genotyping, and the remaining embryos were homogenized in RLT buffer for RNA extraction.

Mouse sex as a biological variable

Sex cannot be determined or predicted in E9.5-stage mouse embryos, so selection was based solely on genotype.

Method details

Genotyping with high resolution melt analysis (HRMs)

Adult genomic DNA was isolated as previously described.³⁶ For genotyping stained zebrafish embryos, genomic DNA was extracted from stained embryos and treated as follows: stained whole embryos were incubated in 30 μL Extraction Lysis Buffer (ELB), which consists of 10 mM Tris (pH 8.3), 50 mM KCl, 0.3% Tween 20, 0.3% NP40, 1 mg/mL Proteinase K. The mixture was incubated at 55°C for 2 hours, followed by heat inactivation of Proteinase K at 95°C for 15 minutes. After this, the samples were stored at 4°C. PCR reagents and reaction conditions were set up as described previously.^36,75 Genotyping reagents, along with primers used for zebrafish p53, puma/bbc3, noxa/pmaip1, p21/cdkn1a and mdm2, and mouse p53, are listed in the key resources table.

Establishing tumor cohorts

Our tumor cohorts were derived by natural breeding of a single set of parents (one male and one female). p53^+/+ were generated by inbreeding a single set of wildtype AB stain parents. p53^-/- were generated by in breeding homozygous p53^-/- individuals. puma^-/-; noxa^-/-; p21^-/- (referred to as pnp^-/-) fish were generated by inbreeding homozygous pnp^-/- individuals. The cohort size (N) is indicated in the figure legends. At 4 months of age, all fish were separated into 4 tanks, with 24 fish per tank. Adult fish were screened weekly for the presence of tumors and/or missing or decreased fish. Fish visually identified as tumor-burdened were euthanized according to IACUC protocols. Kaplan-Meier survival analysis was performed using GraphPad Prism 9 software, with statistical details provided in the figure legends.

Ionizing radiation irradiation

Zebrafish embryos or pregnant mice carrying E9.5-stage embryos were placed at the closest position to the ionizing radiation source in an X-RAD 320 X-ray irradiator, exposing the embryos to 5 Gy/min of radiation for a total dose of 30 Gy.

Whole-mount immunohistochemistry

Embryos were fixed overnight at 4°C in 4% paraformaldehyde (PFA). For pH3 staining on embryos older than 24 hours postfertilization (hpf), dehydration and rehydration were performed using a graded series of methanal solution (25%, 50%, 75% and 100% methanal in PBST), followed by the reverse order for rehydration. Dehydrated embryos can be stored at 100% methanal at -20°C for several months. For pH3 staining, embryos were permeabilized in prechilled acetone for 7 minutes at -20°C. They were then washed three times for 5 minutes each in 0.1% PBST (PBS + 0.1% Tween 20) and blocked for 1 hour at room temperature in blocking solution (10% Sheep Serum, 0.1% of DMSO into 1X PBS). For active caspase 3 staining, embryos were permeabilized for 2 hours at -20°C in 100% methanal, then washed twice for 15 minutes in PDT (0.3% Triton-X, 1% DMSO into 0.1% PBST) and blocked for 1 hour at room temperature with blocking solution containing 10% heat-inactivated FBS and 2% BSA in 0.1% PBST. After blocking, embryos were incubated overnight at 4°C with primary antibodies. For pH3 staining, the anti-pH3 primary antibody (Cell Signaling, 9701) was used at 1:200 dilution. For active Caspase-3 staining, the anti-active Caspase-3 antibody (BD, 559565) was used at a 1:500 dilution. After that, for pH3 staining, embryos were washed four times for 15 minutes each in 0.1% PBST. For active Caspase 3 staining, embryos were washed four times for 15 minutes each in PDT. For active Caspase-3 staining, 1-hour blocking step is necessary before adding secondary antibody. After that, embryos were incubated with the Alexa 488 goat anti-rabbit secondary antibody (Invitrogen, #A-11008) at a 1:200 dilution for 2 hours at room temperature or overnight at 4°C. After secondary antibody incubation, embryos were washed four times for 15 minutes each with 0.1% PBST or PDT, respectively. For pH3-staining, embryos were stained with 4′,6-diamidino-2-phenylindole (DAPI) for 10 minutes after two washes, followed by two additional washed before storage. Stained embryos were stored in PBS at 4°C and imaged within one week.

Light, immunofluorescence and confocal imaging

Embryos were dechorionated at the specific developmental stages by incubating them in 0.03% protease (Sigma, P5147) for 6 minutes and anesthetized using 0.4% tricaine in E3 blue medium. For light and immunofluorescence imaging, embryos were placed in a 60 x 15 mm Falcon petri dish, and for confocal imaging, embryos were mounted in a glass-coverslip-bottomed dish using 1% low-melting agarose as required. Numbers were written on the bottom of the dish near the embryos to facilitate genotyping after imaging. For active Caspase 3-stained embryos, their images were captured using a SMZ-18 Zoom Stereo Microscope, ensuring consistent magnification, laser power, exposure time, and gain setting across all samples at the same timepoints. For pH3-stained embryos, images in Figure 3 were taken using a Nikon A1 inverted confocal microscope, capturing approximately 100-μm Z-stacks. For other figures, a Nikon AX-R Confocal Microscope was used to scan whole embryos in the lateral position (the limitation is 700-μm Z-stacks), with consistent laser power setting for the GFP channel applied across all genotypes for quantification under the same experimental conditions. After imaging, embryos with unknown genotypes were removed from the agarose for genomic DNA extraction using ELB buffer (as previously described). Genotyping was performed following imaging. Further figure processing and analysis was conducted using Nikon NIS Element and ImageJ (Fuji) software.

Quantification of pH3-positive cells

Fuji software was used to quantify pH3-expressing cells in each embryo. The “Freehand selections” tool was applied to outline the embryo body, excluding the yolk. Area measurements were calculated using the scale bar from the corresponding images. The number of mitotically active cells was quantified by counting pH3-stained cells within the outline area for each individual embryo. The same threshold and particle size settings were applied across all embryos in the same comparison panel. Each dot in the plots represents an individual embryo. All imaged embryos were collected from at least two independent experiments.

Bulk RNA sequencing and analysis

Total RNA was extracted using the QIAGEN RNeasy Plus Mini Kit (QIAGEN, #74134) following the manufacturer’s instructions. Two biological replicates were prepared for each condition. For zebrafish, RNA was pooled from approximately 30 randomly selected embryos per condition (IR-irradiated and untreated) and from approximately 50 embryos collected before 24 hpf. For mice, RNA was extracted individually from each embryo. Representative images of mouse embryos for each condition are shown in Figure S4B. Each replicate was used to generate one RNA library. Only RNA Samples with an RNA Integrity Number (RIN) greater than 9 were used for RNA library preparation. RNA libraries were prepared using the Illumina RNA kit with PolyA selection, and sequencing was performed on Illumina HiSeq platform, either with 2x150 bp paired-end reads (sequenced by Genewiz from Azenta Life Sciences or 2x75 bp pair-end reads (sequenced by UAB Heflin Center for Genomic Core). FastQC (version 0.11.7, Java-1.8.0_74) was used to assess the quality of RNA-sequencing samples before analysis. Raw sequencing reads were aligned to the Danio rerio GRCz11 (for zebrafish) or Mus musculus GRCm39 (for mouse) genome assemblies using STAR (version 2.7.3a, GCC 6.4.0-2.28) with default setting.⁶⁷ Following alignment, raw read counts were generated using HTSeq (version 0.12.3, foss-2018b, Python-3.6.6).⁶⁸ The corresponding gene annotation files used in HTSeq were Danio_rerio.GRCz11.102.chr.gtf for zebrafish samples and Mus_musculus.GRCm39.103.chr.gtf for mouse samples. Differentially expressed genes (DEGs) were identified using DESeq2 package (version 1.42.1) in R (version 4.3.3).⁷⁰ The specific cutoff criteria for significance are indicated in the figure legends for each analysis. Volcano plots were generated using GraphPad Prism 10. Gene Ontology (GO) analyses were conducted using DAVID Bioinformatics Resources 6.8, and the results were visualized with the GOplot package (version 1.0.2) in R, with gene ordered by logFC values.^71,72,76 Transcripts Per Million (TPM) values for heatmap generation were calculated using RSEM (version 1.3.3).⁶⁹ The Heatmap was created with the complexheatmap (version 2.22.0) package in R,⁷⁷ using TPM values scaled by z-scores across genes. Bidirectional clustering was performed using the complete linkage method and Euclidean distance metric. The Heatmap with box plots for each cluster was plotted via https://www.bioinformatics.com.cn, an online platform for data analysis and visualization using the ClusterGVis package (version 0.1.2) in R.⁷³ No original code was generated for the RNA sequencing, alignment and analysis described above. The code used to identify the orthologs and paralogs across zebrafish, mouse and human is publicly available via GitHub, as described in the key resources table. For zebrafish genes without defined orthologs in mouse or human through the Ensembl Genome Database, manual searches were conducted using the Alliance of Genome Resources database with the orthology stringency set to “No filter” to ensure accurate identification of gene relationships across species. If orthologs could not be defined through the Alliance of Genome Resources, manual searches for paralogs of these zebrafish genes were performed, and their corresponding orthologs were subsequently identified. Orthologous genes identified via zebrafish paralogs were categorized into the paralog list.

Cas9 ribonucleoprotein (RNP) preparation and microinjection for G0 knockouts screening

Alt-R crRNAs were designed with Integrated DNA technologies (IDT) predesigned CRSPR-Cas9 guide RNA tool (https://www.idtdna.com/site/order/designtool/index/CRISPR_PREDESIGN). Alt-R CRISPR-Cas9 crRNA, tracrRNA (IDT, 1072532) and Alt-R S.p. Cas9 Nuclease V3 (IDT, 1081058) were prepared following manufacturer’s instructions. Detailed procedures for RNA complex preparation were performed as described.³⁶ Final concentration of each guide is 1.5 μM. Microinjection was performed by injecting 1 nL of RNP complex into the yolk of one-cell stage embryos. The RNP complex was freshly prepared and left on ice until microinjection.⁷⁸ IDT crRNA used for multiple guide RNA injections are listed in Table S10. The sample harvest and analysis were performed as shown in the respective figures.

Quantification and statistical analysis

GraphPad Prism 9 was used in graph generation and all statistical analyses. Statistical significance was determined using an unpaired t-test unless otherwise stated in the figure legends. Error bars represent the standard error of the mean (SEM). The number of embryos analyzed (N) and corresponding significance values (p-values) are provided in the figure legends.

Published: May 2, 2025

Supplemental information

Table S1. 60 p53 upregulated DEGs identified in zebrafish at 1 hpi, related to Figure 4

16 grey-colored genes showed no significant difference between 1-hpi pnp^−/− vs. untreated pnp^−/− and 1-hpi p53^−/− vs. untreated p53^−/− datasets.

Table S2. 242 p53 upregulated DEGs identified in zebrafish at 3 hpi, related to Figure 4

82 grey-colored genes showed no significant difference between 3-hpi pnp^−/− vs. untreated pnp^−/− and 3-hpi p53^−/− vs. untreated p53^−/− datasets.

Table S3. 49 p53 upregulated DEGs identified in mouse at 1 hpi, related to Figure 5

6 grey-colored genes showed no significant difference between 1-hpi p53^+/+ vs. untreated p53^+/+ and 1-hpi p53^−/− vs. untreated p53^−/− datasets.

Table S4. 1,602 p53 upregulated DEGs identified in mouse at 3 hpi, related to Figure 5

83 grey-colored genes showed no significant difference between 3-hpi p53^+/+ vs. untreated p53^+/+ and 3-hpi p53^−/− vs. untreated p53^−/− datasets.

Table S5. Mouse orthologs of 264 p53 upregulated zebrafish DEGs at 1 or 3 hpi, related to Figure 5

Grey-colored genes represent zebrafish genes that are not protein-coding and lack defined mouse orthologs. Orange-colored genes lack direct mouse orthologs but were identified by first defining their zebrafish paralogs and then determining the mouse orthologs of those zebrafish paralogs. These genes are referred to as mouse paralogs.

Table S6. Mouse paralogs of 323 mouse orthologs, related to Figure 5

Orange-colored genes represent those that lack direct orthologs, as shown in Figure S5.