tp53 deficiency causes a wide tumor spectrum and increases embryonal rhabdomyosarcoma metastasis in zebrafish.

The TP53 tumor-suppressor gene is mutated in >50% of human tumors and Li-Fraumeni patients with germ line inactivation are predisposed to developing cancer. Here, we generated tp53 deleted zebrafish that spontaneously develop malignant peripheral nerve-sheath tumors, angiosarcomas, germ cell tumors, and an aggressive Natural Killer cell-like leukemia for which no animal model has been developed. Because the tp53 deletion was generated in syngeneic zebrafish, engraftment of fluorescent-labeled tumors could be dynamically visualized over time. Importantly, engrafted tumors shared gene expression signatures with predicted cells of origin in human tissue. Finally, we showed that tp53del/del enhanced invasion and metastasis in kRASG12D-induced embryonal rhabdomyosarcoma (ERMS), but did not alter the overall frequency of cancer stem cells, suggesting novel pro-metastatic roles for TP53 loss-of-function in human muscle tumors. In summary, we have developed a Li-Fraumeni zebrafish model that is amenable to large-scale transplantation and direct visualization of tumor growth in live animals.

Introduction

TP53 is a tumor suppressor protein that is mutated or functionally disrupted in more than 50% of all human tumors (Kastenhuber and Lowe, 2017; Muller and Vousden, 2014). Moreover, genetic mutation of TP53 in Li-Fraumeni patients leads to cancer predisposition early in life and is associated with transformation in a broad range of target tissues (Malkin, 2011). TP53 is commonly inactivated by single amino acid mutations that create dominant-negative forms of the protein that inhibit efficient tetramer formation and block transcriptional activity (Muller and Vousden, 2014). In this setting, TP53 alleles likely alter transcriptional activity of TP53 and its related transcription factor family members, TP63 and TP73 (Lang et al., 2004; Olive et al., 2004). By contrast, TP53 deletion is expected to have less wide-ranging transcriptional effects that are confined to tetrameric transcription factor function. Regardless of the genetic alteration, TP53 transcriptional inactivation can lead to genomic instability and impaired apoptotic responses that often are predisposing to a wide array of cancers (Kastenhuber and Lowe, 2017; Muller and Vousden, 2014).

To date, several murine genetic models have been developed to assess the effects of both loss- and gain-of-function Tp53 mutations in cancer (Donehower et al., 1992; Harvey et al., 1993; Jacks et al., 1994; Lang et al., 2004; Lavigueur et al., 1989; Lee et al., 1994; Olive et al., 2004). Both Tp53 mutant and null alleles spontaneously develop cancer. However, similar to human Li-Fraumeni patients, the spectrum varies among different alleles, suggesting that the mode of Tp53 inactivation has important implications in regulating the types of cancer that develop, the time to onset, and the overall propensity for tumor progression (Lavigueur et al., 1989; Lee et al., 1994). For example, mice heterozygous for the 172His point mutation are predisposed to developing osteosarcoma while animals harboring the 270His mutation develop hemangiosarcoma and carcinoma (Olive et al., 2004). By contrast, mice with homozygous Tp53 deletion mainly develop lymphoma, with rare cases of angiosarcoma, undifferentiated sarcoma, osteosarcoma, rhabdomyosarcoma, testicular tumors, nervous system tumors, teratoma, and mammary carcinoma being reported (Donehower et al., 1992; Harvey et al., 1993; Jacks et al., 1994). Together, these data suggest that differences in gain- and loss-of-function alleles have profound effects on tumor onset and spectrum in genetically engineered mice and yet, largely recapitulate the wide array of cancers observed in Li-Fraumeni patients. Importantly, a small subset of Li-Fraumeni syndrome patients harbor genomic deletions in the TP53 locus and cancers that develop in dominant-negative, heterozygous point-mutation carriers often display deletion of the second TP53 allele (Malkin, 2011). Thus, modeling complete TP53 loss-of-function in different animal models will likely provide novel insights into human disease.

TP53 is also commonly mutated in human sarcomas and is predictive of poor outcome (Taubert et al., 1996). For example, the TP53 locus is mutated in 16% of human embryonal rhabdomyosarcoma (ERMS), a common pediatric cancer of muscle and transcriptional activity is altered in >30% of human ERMS through TP53 locus disruption or MDM2 amplification (Taylor et al., 2000). Interestingly, TP53 mutations are also acquired at ERMS relapse (Chen et al., 2013), suggesting a likely role for TP53 in ERMS progression and therapy resistance. Finally, Li-Fraumeni patients with germline TP53 mutations commonly develop ERMS (Malkin, 2011), suggesting important roles for TP53 loss in the genesis of this disease. Yet, to date, the effect of TP53 pathway inactivation on cancer stem cell number, tumor progression, and metastasis in ERMS is not fully understood. Moreover, because genetically engineered mouse and human xenograft models of ERMS do not metastasize in vivo, assessing TP53 loss-of-function in the context of rhabdomyosarcoma metastasis has not been possible. Finally, to date, no tp53 deletion models have been generated in syngeneic zebrafish, precluding large-scale transplantation studies to assess how deletion regulates cancer stem cells and tumor invasion in vivo for a wide array of cancers.

To better study tp53 biology in vivo, we generated a complete loss-of-function tp53 deletion allele in syngeneic CG1-strain zebrafish using TALEN endonucleases. tp53 animals spontaneously developed a wide range of tumors including malignant peripheral nerve-sheath tumors (MPNSTs), angiosarcomas, germ cell tumors, and an aggressive natural killer cell-like leukemia not previously described in any animal model. This model contrasts with currently available point-mutation alleles for zebrafish tp53 that predominantly develop MPNSTs (Berghmans et al., 2005). Moreover, because the tp53 mutant was generated in CG1-strain syngeneic zebrafish (Mizgireuv and Revskoy, 2006), tumors efficiently transplanted into recipient fish enabling expansion of unlabeled and GFP-labeled tumors, dynamic live animal imaging of metastatic progression, and analysis of transcriptional differences between tumors using RNA sequencing approaches. Roles for tp53 were also assessed in kRAS-induced ERMS using large-scale cell transplantation assays and live fluorescent imaging over time. Using these approaches, we showed that the overall frequency of ERMS self-renewing cancer stem cells was unaffected by loss of tp53. In contrast, tp53 ERMS were more invasive, providing a potential explanation for increased aggression associated with TP53 disruption in human ERMS (Seki et al., 2015). Taken together, our work has uncovered novel roles for Tp53 loss in the onset of a wide array of cancers and has provided new insights into how tp53 affects ERMS progression in vivo.

Results and discussion

tp53 deletion mutants spontaneously develop MPNSTs, angiosarcoma, germ cell tumors, and leukemia

Given the critical function of Tp53 as a tumor suppressor and the absence of a complete null allele in zebrafish, we created a tp53 deletion mutant using two pairs of TALENs (Transcription Activator-Like Effector Nucleases) that cleaved at the 5’ and 3’ end of the tp53 locus (Figure 1A). One-cell stage CG1 syngeneic embryos (Mizgireuv and Revskoy, 2006) were microinjected with mRNA encoding each TALEN pair and raised to adulthood. F1 embryos were screened by genomic PCR to identify a single founder line with deletion of 12.1 kb tp53 genomic sequence (Figure 1A). CG1 tp53 heterozygous fish were in-crossed and progeny assessed for Tp53 protein loss and the ability to undergo apoptosis following ionizing irradiation. As expected, homozygous tp53 embryos lacked protein expression and were resistant to radiation-induced apoptosis (Figure 1—figure supplement 1B–D).

Figure 1.

Homozygous tp53 zebrafish spontaneously develop a wide range of tumor types.

(A) tp53 genomic locus and CG1 tp53 allele. TALEN arms were designed to target the 5’ and 3’ genomic sequence of tp53 (red). (B–M) CG1 tp53 zebrafish develop leukemia (B–D), angiosarcoma (E–G), MPNSTs (H–J), and germ cell tumors (K–M). Whole animal images (B,E,H,K), hematoxylin/eosin (H and E) stained sections (C,D,F,G,I,L,M), and immunohistochemistry for Sox10 (J). Blast-like leukemia cells predominate in the kidney marrow and efface the renal tubules (black arrow, (D). (N) Tumor incidence in CG1 tp53 zebrafish (n = 134). (O) Quantitation of tumor types that form in CG1 tp53 mutant zebrafish by 55 weeks of life based on histology review (n = 51). (P–S) kRAS-induced embryonal rhabdomyosarcoma (ERMS) generated in CG1 tp53 zebrafish. Whole animal bright field and GFP-epifluorescence overlap images (P and Q, respectively). H and E stained sections revealed features consistent with human ERMS (R,S). Scale bars equal 12.5 mm in whole animal images and 100 μm in histology images.

(A) Survival of animals by genotype from a heterozygous tp53 in-cross. tp53 homozygous wild-type (wt/wt), heterozygous (wt/del), and homozygous (del/del) mutant fish. (B) Western blot analysis at 24 hr post-fertilization (hpf) whole embryos. Actin is used as a loading control. (C–D) TUNEL staining performed on tp53 and tp53 embryos following gamma-irradiation at 24hpf (16 Gray) and fixation at 30hpf. Whole embryos images are shown for representative animals of each genotype (C). Quantification of TUNEL-positive cells within 1000 micron2 area. Regions where cells were counted are outlined by the white boxes in panel C. p<0.001 by Student’s T-test.

Designation as assessed by histology review.

Figure 1—figure supplement 1.

tp53 zebrafish survive at expected Mendelian ratios, lack Tp53 protein expression, and are resistant to irradiation-induced cell death.

(A) Survival of animals by genotype from a heterozygous tp53 in-cross. tp53 homozygous wild-type (wt/wt), heterozygous (wt/del), and homozygous (del/del) mutant fish. (B) Western blot analysis at 24 hr post-fertilization (hpf) whole embryos. Actin is used as a loading control. (C–D) TUNEL staining performed on tp53 and tp53 embryos following gamma-irradiation at 24hpf (16 Gray) and fixation at 30hpf. Whole embryos images are shown for representative animals of each genotype (C). Quantification of TUNEL-positive cells within 1000 micron2 area. Regions where cells were counted are outlined by the white boxes in panel C. p<0.001 by Student’s T-test.

Progeny from tp53 crosses were also raised to adulthood and assessed for viability and tumor onset over time. Both heterozygous tp53 and homozygous tp53 fish survived until adulthood at expected ratios (Figure 1—figure supplement 1A). By 4 months of age, tp53 zebrafish began to spontaneously develop tumors. Phenotypes of the earliest malignant tp53 cohort were consistent with loss of osmoregulation and kidney damage. Histopathological analysis of these animals revealed features consistent with leukemia, including blast-like cells predominating in the kidney marrow and loss of kidney stromal architecture, including effacement of the renal tubules (Figure 1B–D, Figure 1—figure supplement 2). Beginning by 7 months of age, a subset of tp53 animals developed externally visible tumors and histology consistent with angiosarcoma and malignant peripheral nerve sheath tumors (MPNSTs) (Figure 1E–J, Figure 1—figure supplement 2) (Berghmans et al., 2005; Choorapoikayil et al., 2012; Parant et al., 2010). A small subset of tp53 fish also developed prominent externally visible abdominal masses that were diagnosed as germ cell tumors following histopathological analysis (n = 2, Figure 1K–M, Figure 1—figure supplement 2) (Neumann et al., 2011). MPNST assignment was validated using IHC staining for sox10, which is a well-established clinical marker for this tumor type (Figure 1J) (Shin et al., 2012). As expected, tp53 zebrafish infrequently developed tumors, which is consistent with studies in other tp53 deficiency mouse models (Donehower et al., 1992; Harvey et al., 1993; Jacks et al., 1994).

Figure 1—figure supplement 2.

Tumor latency in tp53 zebrafish.

Designation as assessed by histology review.

In total, 37% of tp53 animals developed externally visible tumors by 12 months of age with a wider tumor spectrum than previously reported in homozygous tp53 and tp53 mutant zebrafish (Figure 1N,O) (Berghmans et al., 2005; Parant et al., 2010). For example, these point mutation models predominantly developed MPNSTs with only a rare, single melanoma being detected in homozygous animals (Berghmans et al., 2005). Remarkably, the spectrum in tp53 zebrafish was more similar to that reported in Tp53-null mice, with angiosarcomas and germ cell tumors occurring in both models (Donehower et al., 1992; Harvey et al., 1993; Jacks et al., 1994). However, the predominance of T cell lymphomas seen in Tp53-null mice was not observed in tp53 zebrafish, likely reflecting species differences in sensitivity to Tp53 loss in target cells. It is also possible that tumor types seen in Tp53-null mice exhibit longer latency in our model and would not manifest in the short-lived CG1 strain zebrafish. Finally, we also generated ERMS in tp53 fish by microinjecting linearized human kRASG12D oncogene and GFP under the control of rag2 promoter (Figure 1P,Q) (Langenau et al., 2007). Histopathological analysis of transgene-induced tp53 ERMS showed consistent morphology with the spindle-variant of human ERMS (Figure 1R,S) (Langenau et al., 2007).

tp53 tumors are transplantable

One major advantage of generating tp53 mutations in the CG1 syngeneic stain of zebrafish is the ease with which tumors can be used in cell transplantation assays. Primary MPNSTs that arose in the eye were dissected from euthanized tp53 animals and orthotopically transplanted into the equivalent periocular space or into the peritoneum of CG1-strain recipient fish (n = 2 primary tumors, n = 9 recipient fish total, 2 × 104 cells/fish). All recipient animals engrafted tumor with histology similar to the primary disease (Figure 2A–E, Figure 2—figure supplement 1). To more easily track tumor cells in host animals, we also crossed tp53 animals into CG1 syngeneic ubi:GFP transgenic zebrafish. We successfully engrafted tp53 angiosarcomas into CG1-recipient animals and ubi:GFP+ tumor cells were easily traceable in non-fluorescent recipients (n = 3 primary tumors, n = 8 of 9 transplant fish developed tumors, Figure 2F–I, Figure 2—figure supplement 1). Finally, ubi:GFP+ leukemias were also assessed by cell transplantation. Specifically, blood cells were engrafted into non-irradiated CG1 strain fish (2.5 × 104 cells/intraperitoneal injection). In total, five of five primary leukemias engrafted into recipient fish with GFP+ cells disseminating widely throughout the animal by 60 days post-transplantation (n = 5 primary leukemias, n = 25 of 25 animals engrafted leukemia, Figure 2J–M, Figure 2—figure supplement 1). Whole animal imaging and flow cytometric analysis revealed that GFP+ cells also invaded the recipient kidney marrow, the site of hematopoiesis in adult zebrafish (Figure 2M,N). Leukemic cells consisted of as much as 45% of the reconstituted marrow in transplanted fish (Figure 2N, n = 3 independent primary leukemias analyzed). To more closely observe leukemia cell morphology, FACs sorted GFP+ leukemia cells were assessed by cytospin and Wright/Giemsa staining (Figure 2O–R). The leukemic cells were large with prominent nucleoli and abundant, vacuolated cytoplasm, consistent with a rare, high-grade aggressive NK cell leukemia. In the context of gene expression profiling (outlined below), these leukemias were also similarly classified as aggressive NK cell-like leukemia, suggesting important roles for Tp53 in initiation of leukemias of NK cell origin (Figure 2Q,R). Finally, GFP-labeled kRASG12D-induced tp53 ERMS were readily transplantable when engrafted into syngeneic recipient fish (n = 11 primary tumors analyzed, n = 47/49 fish engrafted, Figure 2S–V, Figure 2—figure supplement 1). Histology was similar between primary and transplanted MPNSTs, angiosarcomas, leukemias and ERMS (Figure 2B,C,E,I,M,V).

Figure 2.

tp53 tumors efficiently transplant into syngeneic CG1 strain zebrafish.

(A–E) A primary tp53 MPNSTs that formed in the eye transplanted orthotopically into the periocular space (A–C) or into the peritoneum of CG1-strain recipient fish (D–E). Intraperitoneal injection (i/p). (F–I) tp53 Tg(ubi:GFP)-positive angiosarcoma. Primary tumor-bearing fish (F–G) and transplanted animal (H–I). (J–R) tp53 Tg(ubi:GFP)-positive leukemia. Primary leukemia (J–K) and transplanted leukemia shown at 20 days post-transplantation (L–R). Whole kidney marrow was isolated from leukemia-engrafted fish and analyzed by FACS (N–O). (N) Forward and side scatter plot of whole kidney marrow of unlabeled CG1 host animal to assess ubi:GFP-positive tp53 leukemia cells following transplantation. (O) Analysis of GFP+ ubi:GFP-positive tp53 leukemia cells following FACS. Purity was ≥90%. (P–R) Cytospins and Wright/Giemsa staining of whole kidney marrow cells isolated from wildtype fish (P) compared with FACS sorted GFP+ cells from two representative aggressive NK cell-like leukemias, showing large blastic cells with abundant basophilic, vacuolated cytoplasm (Q–R). (S–V) Embryonal rhabdomyosarcoma arising in tp53 fish micro-injected at the one-cell stage with linearized rag2:kRASG12D + rag2:GFP. Primary (S), transplanted (2°) (T), and serially transplanted ERMS (3°) (U,V). Whole animal bright-field images (A,D,F,J) and merged GFP-fluorescence images (G,H,K,L,S–U). Hematoxylin and eosin stained sections of engrafted tumors (B–C,E, I, M,V). Scale bars are 5 mm in whole animal images and 100 μm for histology images.

Engraftment was scored at >20 days post transplantation. i.p. intraperitoneal.

Figure 2—figure supplement 1.

Engraftment of tp53 tumors into CG1 recipient zebrafish.

Engraftment was scored at >20 days post transplantation. i.p. intraperitoneal.

Gene expression analysis of tp53 tumors arising in transplant recipient fish

We next profiled the transcriptome of tumor cells isolated from fish transplanted with tp53 MPNSTs, angiosarcomas, leukemias, a germ cell tumor and kRASinduced ERMS by Poly(A)+ RNA-sequencing (RNAseq). MPNSTs and the germ cell tumor were analyzed from bulk tumor isolated from non-GFP labeled animals, whereas angiosarcomas, leukemias, and ERMS were FACS sorted from engrafted GFP-labeled tumors (purity and viability >85%). Bulk mRNA from three independent wild-type CG1 strain fish was also sequenced and used as a control. Principal component analysis identified six distinct clusters corresponding to whole CG1 syngeneic fish, leukemia, MPNSTs, angiosarcomas, germ cell tumor and ERMS (Figure 3A). By comparing gene expression among different tumor types, unique tumor-specific expression profiles were identified and each assessed for overlap with gene sets found in the Molecular Signatures Database (MSigDB, Figure 3B,C, Figure 3—source datas 1–3). For example, the upregulated leukemia gene set identified in zebrafish was significantly enriched for GO terms associated with immune system processes, leukocyte activation, immune response and lymphocyte activation. By contrast, angiosarcomas were enriched in GO gene sets associated with vasculature, blood vessel morphogenesis and cellular proliferation. The germ cell tumor showed enrichment of GO gene sets associated with sexual reproduction and gamete formation. As expected, ERMS shared significant overlap of gene signatures with muscle structure, muscle contraction and muscle development.

Figure 3.

Gene expression analysis of tp53 tumors.

(A) Principal component analysis (PCA) of gene expression profiles from whole CG1 syngeneic fish, MPNSTs, a germ cell tumor, and FACS sorted GFP+ leukemia, angiosarcomas, and ERMS. All tumor samples were obtained following engraftment in CG1 syngeneic recipient fish. (B) Heat map of genes differentially expressed with respect to controls identifies molecularly defined tumor groups. (C) Upregulated genes identified within each tumor type are enriched for Molecular Signature Database (MSigDB) signatures consistent with the expected tissue of origin. (D) NK cell leukemias are enriched for gene signatures identified from normal NK and NKL cells in the kidney marrow and NK cells isolated from rag1 transgenic fish (NK cells*). For each analysis, enrichment is shown for the top 30 lineage-restricted genes identified from single-cell transcriptional profiling of transgenic cells using SMARTseq2 (denoted by asterisks) or unsorted cells using InDrops single-cell RNA sequencing approaches. (E) Heat map highlighting NK lineage genes significantly upregulated in tp53 leukemias when compared to all other tumor types analyzed. [log2(fold-change)]. Angiosarcoma (AS) and germ cell tumor (GC).

These gene lists were used to assess overlap with the GSEA signature database.

Significant overlap was observed for both commonly up-regulated (p=4e-321) and down-regulated (p=5e-182) genes. A fold change of log2(FC) ≥2 was considered differential and statistical significance was assessed as p≤0.05 with a one-sided Fisher’s exact test.

(A–C) GSEA analysis comparing zebrafish tp53 tumors to human counterparts. (A) Human angiosarcoma (FDR q-value = 0.001), (B) human MPNST (FDR q-value = 0.00433526), and (C) human ERMS (FDR q-value = 0) gene sets are significantly enriched in the corresponding zebrafish tp53 tumors. (D) Heat map depicting differential gene expression in zebrafish tp53 leukemias compared to whole CG1 control animals. Gene-expression defines tp53 leukemias based on lineage signatures previously generated using InDrop and SMARTseq (asterisk) sequencing approaches (Tang et al., 2017). (E) Expression of the top 200 tp53 ANKL-like genes assessed in the SMARTseq dataset from Tang et al. (2017). HSC/progenitors were isolated as cd41:GFPlow cells from transgenic zebrafish, T cells from tg(lck:GFP) transgenic zebrafish, NK cells from rag1-/-, tg(lck:GFP) transgenic zebrafish, myeloid cells from tg(mpx:EGFP) transgenic zebrafish, B cells from marrow-derived tg(rag2:GFP) transgenic zebrafish, and HSCs from tg(runx1 transgenic zebrafish. The tp53 leukemia gene expression signature was significantly enriched only in the NK cells (log2(TPM+ 1)≥2, p-value=0.015, one-sided binomial test).

Gene identifications correspond to SMARTseq and InDrop single cell sequencing from Tang et al. (2017), as indicated.

Top 500 transcripts differentially regulated in each tumor subtype identified by RNA sequencing analysis.

Differential gene expression for tp53 and tp53 MPNST.

Differential gene expression for tp53 leukemias with respect to blood cells and kidney cells shown in Figure 3—figure supplement 1D.

To assess similarities between tp53 tumors and human, we next assessed if zebrafish tumors express tumor-specific gene signatures identified from human angiosarcoma (Andersen et al., 2013), MPNST (Kolberg et al., 2015) and ERMS (experimentally determined using GEO:GSE108022) (Figure 3—figure supplement 1—source data 1). Using Gene set enrichment analysis (GSEA) (Mootha et al., 2003; Subramanian et al., 2005), we identified significant enrichment of signatures associated with human angiosarcoma (FDR q-value = 0.001, Figure 3—figure supplement 1A), MPNST (FDR q-value = 0.00433526, Figure 3—figure supplement 1B), and ERMS (FDR q-value = 0, Figure 3—figure supplement 1C) in the corresponding zebrafish tp53 tumors but not other tumor types (Figure 3—figure supplement 1—source data 1). Taken together, these data reveal conserved gene expression programs associated with both the predicted cells of origin and the corresponding human cancer counterpart.

Figure 3—figure supplement 1.

Zebrafish cancers share common gene expression with human tumors and confirmation of NK-cell linage derivation for tp53 leukemias.

(A–C) GSEA analysis comparing zebrafish tp53 tumors to human counterparts. (A) Human angiosarcoma (FDR q-value = 0.001), (B) human MPNST (FDR q-value = 0.00433526), and (C) human ERMS (FDR q-value = 0) gene sets are significantly enriched in the corresponding zebrafish tp53 tumors. (D) Heat map depicting differential gene expression in zebrafish tp53 leukemias compared to whole CG1 control animals. Gene-expression defines tp53 leukemias based on lineage signatures previously generated using InDrop and SMARTseq (asterisk) sequencing approaches (Tang et al., 2017). (E) Expression of the top 200 tp53 ANKL-like genes assessed in the SMARTseq dataset from Tang et al. (2017). HSC/progenitors were isolated as cd41:GFPlow cells from transgenic zebrafish, T cells from tg(lck:GFP) transgenic zebrafish, NK cells from rag1-/-, tg(lck:GFP) transgenic zebrafish, myeloid cells from tg(mpx:EGFP) transgenic zebrafish, B cells from marrow-derived tg(rag2:GFP) transgenic zebrafish, and HSCs from tg(runx1 transgenic zebrafish. The tp53 leukemia gene expression signature was significantly enriched only in the NK cells (log2(TPM+ 1)≥2, p-value=0.015, one-sided binomial test).

Gene identifications correspond to SMARTseq and InDrop single cell sequencing from Tang et al. (2017), as indicated.

Given that GSEAsig analysis failed to assign leukemias to a specific lineage, we next assessed if these tumors were enriched for signatures associated with normal blood cell lineages identified previously by our group using single-cell RNA sequencing of the zebrafish marrow (Tang et al., 2017). Using these lineage-specific gene sets, we found that tp53 leukemias expressed markers indicative of NK and NK-like cells but largely failed to express genes associated with other hematopoietic cell lineages (Figure 3D and Figure 3—figure supplement 1D,E). To independently confirm our results, we next identified the top 200 most differentially regulated genes in leukemias compared to all other tumor types and assessed if these genes were differentially expressed in each zebrafish blood lineage. Significant enrichment was only observed in NK cells (Figure 3—figure supplement 1E, p=0.015, one-sided binomial test), supporting a NK cell origin of tp53 leukemias. Importantly, tp53 NK cell-like leukemias also expressed well-known genes commonly associated with human NK cells, including il2ga and b, jak3, perforins 2,7, and 8, and these genes were highly up-regulated when compared to all other tumor types in our analysis (Figure 3E).

In humans, aggressive NK cell leukemias (ANKLs) have a very poor prognosis and often express perforins but lack markers of mature T- and B- cell lineages (Liang and Graham, 2008; Suzuki and Nakamura, 1999). In human disease, ANKLs are associated with Epstein-Barr virus infection, however, CD3-/CD4-/CD56+/CD13-/CD33- leukemias without EBV infection and intact germline configured T-cell receptor and immunoglobulin have been reported (Liang and Graham, 2008). Interestingly, both TP53 point mutations and deletions have been identified in human ANK cell leukemias, suggesting a role for TP53 in pathogenesis of this disease (Soliman et al., 2014; Yagita et al., 2000). Yet, to date no in vivo animal models of ANKL have been reported precluding direct functional assessment of TP53 loss in eliciting transformation of NK cells.

Given that tumor onset and spectrum differ based on the nature of Tp53 mutation or deletion in mice, we next compared gene expression between MPNSTs arising in tp53 and tp53M214K/M214K mutant zebrafish (Figure 3—source data 4). As may be expected, we found significant overlap in expression between homozygous tp53 deletion and point-mutant MPNSTs when compared to whole fish (p=4e-321 for up-regulated genes and p=5e-182 for down-regulated genes, one-sided Fisher’s exact test (Figure 3—source data 4), confirming the previously described loss-of-function activity for tp53 (Berghmans et al., 2005). Interestingly, differences in gene expression were also noted when comparing these tumors, likely arising from differences in the underlying mutations (Figure 3—source data 4), which are known to differentially affect Tp53 function and tumor etiology in genetically engineered mouse models.

tp53 ERMS display increased metastasis but did not alter cancer stem cell number

TP53 has roles in regulating self-renewal of normal stem cells and human cancer cells, including acute myeloid leukemia and breast cancer (Cicalese et al., 2009; Meletis et al., 2006; TeKippe et al., 2003; Zhao et al., 2010). Thus, we predicted that tp53 loss may affect the overall frequency of self-renewing cancer stem cells in zebrafish ERMS. To test this hypothesis, GFP+ ERMS cells were isolated by FACS and injected at limiting dilution into the peritoneum of CG1 recipients (1 × 104–10 cells/recipient, Figure 2T). Animals were followed for 90 days for engraftment using whole animal epi-fluorescent imaging. Unexpectedly, kRASG12D-induced tumors harboring wild-type tp53 had similar frequency of tumor-propagating stem cells when compared with those of tp53 ERMS (n ≥ 3 tumors analyzed per genotype, p=0.647 EDLA analysis, Table 1). We concluded that Tp53 loss-of-function does not alter the overall frequency of tumor-sustaining, cancer stem cells in ERMS, which contrasts with previous studies that defined major roles for NOTCH1, MYF5/MYOD, and WNT signaling in regulating self-renewal and the overall number of tumor sustaining cell types in rhabdomyosarcoma (Chen et al., 2014; Hayes et al., 2018; Ignatius et al., 2017; Tenente et al., 2017).

Table 1.

Results from limiting dilution cell transplantation experiments comparing engraftment potential of tp53 and tp53 kRASG12D-induced ERMS.

tp53wt/wt + rag2:kRASG12D ERMS
Cell #	Tumor 1	Tumor 2	Tumor 3	Tumor 3
10000	7 of 7	5 of 6	6 of 6	6 of 6
1000	2 of 6	2 of 7	6 of 8	1 of 8
100	0 of 9	0 of 8	1 of 8	0 of 8
TPC#	1 in 2832	1 in 4810	1 in 726	1 in 7388
	1 in 3495 (2291–5333)
tp53del/del + rag2:kRASG12D ERMS
Cell #	Tumor 1	Tumor 2	Tumor 3
10000	3 of 5	5 of 6	6 of 6
1000	3 of 4	3 of 5	0 of 7
100	3 of 9	3 of 7	1 of 8
TPC#	1 in 3546	1 in 2228	1 in 3640
	1 in 3038 (1739–5307), p=0.647

TP53 loss is predictive of poor outcome in human ERMS (Seki et al., 2015); however, given that tp53 loss did not regulate the overall frequency of ERMS stem cells in zebrafish, we reasoned that loss of Tp53 might rather affect tumor invasion and metastasis. To test this hypothesis, we undertook tumor cell transplantation experiments whereby GFP-labeled tumor cells were injected into the dorsolateral musculature of recipient fish and animals monitored for spread into the viscera using epifluoresence whole animal imaging (n = 5 tp53 and n = 11 tp53 ERMS, n = 7–15 recipient fish per tumor,>2×104 cells/recipient) (Tang et al., 2016). As expected, all animals developed GFP+ masses at the site of primary injection (Figure 4A,D, n = 160). Tumor growth was followed for up to 30 days after cell transplantation and animals were assessed for 1) local infiltrative disease defined by growth well beyond but contiguous with the primary site, or 2) metastasis defined by growth at sites unconnected to the primary lesion and/or associated with infiltration into organs within the peritoneal cavity (Tang et al., 2016). Through serial imaging of engrafted fish over time, we identified only rare metastatic lesions in zebrafish engrafted with tp53 ERMS (n = 5 of 58 engrafted animals, Figure 4A–F). By contrast, tp53 ERMS were highly aggressive and displayed elevated local invasion and disseminated metastatic disease (n = 28 of 102 engrafted fish had metastatic ERMS, p=0.003, one-sided Fisher’s exact test, Figure 4I). These metastatic lesions were confirmed on paraffin-embedded sections using both hematoxylin/eosin staining and anti-GFP antibody IHC (Figure 4G,H). Thus, our in vivo experiments demonstrate an important consequence for tp53 loss in stimulating local infiltration and metastasis, revealing a property that may account for poor outcome in RMS patients with TP53 pathway deregulation.

Figure 4.

tp53induced ERMS have increased invasion and metastasis.

(A–F) Whole animal fluorescent images of CG1-strain fish engrafted into the dorsolateral musculature with non-disseminated (A–C) and disseminated ERMS (D–F). Days post transplantation (dpt). White lines demarcate GFP+ tumor area. White arrowheads show site of injection and yellow arrowheads denote metastatic lesions. (G) H and E and (H) GFP immunohistological staining of fish engrafted with metastatic tp53 kRASG12D-induced ERMS. (I) Quantification of growth confined to site of injection (green bars) and compared with animals that exhibited local invasion or metastatic ERMS following tumor engraftment until fish were moribund. X-axis identifies 5 tp53 and 11 tp53 ERMS primary tumors that were transplanted into wild-type CG1 syngeneic host zebrafish. p=0.003, one-sided Fisher’s exact test. Scale bars denote 5 mm.

Taken together, our work has defined syngeneic zebrafish as a novel model to assess Tp53 loss-of-function phenotypes and has generated a wide array of cancer types now available for study by the community. This is particularly important for modeling angiosarcoma and aggressive NK cell leukemias for which readily available tp53-deficient zebrafish models are lacking. To date, our tp53 zebrafish is the first description of any animal model of aggressive NK cell-like leukemia, highlighting the importance of Tp53 loss in the genesis of these leukemias and opening exciting new avenues of future study. Finally, our work in embryonal rhabdomyosarcoma revealed that Tp53 loss likely has major impacts on regulating ERMS invasion and metastasis, without altering the overall frequency of relapse driving cancer stem cells. Such findings likely account for why human RMS are more aggressive following TP53 pathway disruption (Seki et al., 2015). This work is important, because unlike available genetically engineered mouse models and human ERMS xenografts, zebrafish ERMS are metastatic, which can be readily quantified and visualized in vivo. Future experiments will likely utilize the tp53 model to study cancer stem cell self-renewal pathways and metastatic progression in a wider array of tumor types including angiosarcoma, MPNSTs, and available transgenic models that require tp53 loss.

Methods and materials

Animals

Zebrafish used in this work included: CG1 strain zebrafish (Mizgireuv and Revskoy, 2006), CG1-strain Tg(ubi:GFP) animals that were generated using Tol2-mediated transgenesis (Kawakami et al., 2000; Mosimann et al., 2011), and rag2 strain zebrafish that were used in a subset of ERMS metastasis assays. All animal studies were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care under the protocol #2011 N-000127.

Generation of tp53 zebrafish using TALENS

Four TALEN pairs (two pairs each flanking the tp53 gene locus in D. rerio) were designed to generate a ~12.1 kb deletion encompassing the entire tp53 coding sequence. Mojo Hand (http://talendesign.org) was used to design eight 15-mer repeat variable di-residue (RVD) TALENs with a 15 to 18 bp spacer (Ma et al., 2013; Neff et al., 2013). Each TALEN pair was designed to target a unique restriction site that could be used to determine TALEN cutting efficacy by restriction fragment length polymorphism (RFLP) analysis. All TALEN constructs were synthesized with the Golden Gate method using the RCIscript-GoldyTALEN scaffold (Addgene, https://www.addgene.org/Stephen_Ekker/, ID# 38142) (Ma et al., 2013; Neff et al., 2013). The RVDs NI, HD, NG and NN (recognizing A, C, T and G bases, respectively) were used to construct TALENs. Intermediate constructs containing RVDs for positions 1 to 10 were synthesized in the pFUS_A receiver plasmid in the first reaction. Pre-synthesized pFUS_B4 plasmids were then selected based on the target sequence. The library of 256 pFUS_B4 plasmids is available through Addgene (https://www.addgene.org/Stephen_Ekker/, Kit # 1000000038). The completed pFUS_A and pFUS_B4 as well as the last half-repeat plasmid (pLR-NI, -HD, -NN or -NG) were combined in the second Golden Gate reaction in the RCIscript-GoldyTALEN expression vector that has T3 promoter. The completed constructs were linearized using SacI, and mRNA was in vitro transcribed using the mMESSAGE mMACHINE T3 Transcription Kit (Thermo Fisher Scientific, cat. no. AM1348). Large deletions encompassing the tp53 locus were engineered through co-injection of TALEN pairs targeting the tp53 5’UTR and 3’UTR. Genotyping was performed using standard PCR: tp53 forward 5’-CACAGCAAGGACACATCTGC-3’, tp53 reverse 5’-AGATCAGTGCTTGTATTGTATCAGTTT-3’, tp53 reverse 5’-GATCGCTCAGAGTCGCAAA-3’

Embryonic protein extraction and western blotting

24 hpf embryos of the respective genotypes were dissociated in PBS, spun down at 1000xg to de-yolk samples, and lysed in 10% SDS buffer. Western blot was performed using anti-tp53 (ab77813, Abcam) and anti-actin (A2066, Sigma) antibodies.

Apoptosis assay

Embryos were raised at 28°C and gamma-irradiated at 24 hpf. At 30 hpf embryos were fixed overnight in 4% paraformaldehyde followed by staining using the In Situ Cell Death Detection Kit, TMR Red (Roche Applied Bioscience) as per manufacturer protocol.

Histology and immunohistochemistry

Paraffin embedding, sectioning and immunohistochemical analysis of zebrafish sections were performed as previously described (Chen et al., 2014; Ignatius et al., 2012). Anti-human SOX10 was performed at the MGH and BWH DF/HCC Research Pathology Cores. Slides were imaged using a transmitted light Olympus BX41 microscope and a Motic Easy Scan Pro slide scanner. Pathology review and staging were completed by board-certified sarcoma (G.P.N and E.Y.C) and hematology pathologist (R.P.H).

Micro-injection and ERMS generation

rag2:kRASG12D and rag2:GFP constructs were described previously (Langenau et al., 2008; Langenau et al., 2007). DNA plasmids were linearized with Xho1, phenol:chloroform-extracted, ethanol-precipitated, resuspended in 0.5 × Tris EDTA + 0.1 M KCl, and injected into one-cell CG1 strain embryos.

FACS and tumor cell transplantation

FACS analysis and RMS cell transplantation were completed essentially as previously described (Chen et al., 2014; Ignatius et al., 2012; Langenau et al., 2007; Smith et al., 2010). tp53 angiosarcomas, leukemias and ERMS tumor cells were stained with DAPI to exclude dead cells and sorted twice using a Laser BD FACSAria II Cell Sorter. Sort purity and viability were assessed after two rounds of sorting, exceeding 85% and 90%, respectively. GFP+ ERMS tumors were transplanted at limiting dilution and monitored for tumor engraftment under a fluorescent dissecting microscope from 10 to 90 days post-transplantation. Tumor-propagating cell frequency was quantified using the Extreme Limiting Dilution Analysis software package (http://bioinf.wehi.edu.au/software/elda/). GFP+ tumor cells were isolated by FACS from a subset of transplanted fish and RNA isolated for RNA sequencing. Subsets of tumors were fixed in 4% PFA and embedded in paraffin blocks, sectioned and stained with Hematoxylin and Eosin. Sorted GFP+ tp53 del/del leukemia cells were spun down onto a cytospin slide and processed by Wright/Giemsa staining.

RNA sequencing and analysis

Paired-end reads from poly(A)+ RNA-seq were aligned to the GRCz10 reference zebrafish genome with STAR v2.4.0 (Dobin et al., 2013) using GRCz10v85 Ensembl annotations. PCR duplicates were removed with Picard v1.95 [http://broadinstitute.github.io/picard/] and reads aligning to ribosomal RNA were removed with RSeQC (Wang et al., 2012) Gene counts were obtained from reads with an alignment quality of at least 10 using featureCounts (Liao et al., 2014) and transformed to transcript per million (TPM) units. Human orthologues of zebrafish genes were obtained from the Beagle database (Tang et al., 2017; available at: http://chgr.mgh.harvard.edu/genomesrus/). Differential expression analysis was performed with DESeq2 (Love et al., 2014), requiring log2(FC) ≥2 and an FDR < 0.05 was required. Each tumor type was individually compared to the control samples. Clustering of differentially expressed genes used the partitioning around medoids (PAM) method in the cluster R package and the Pearson correlation was used as distance. The number of clusters was optimized with the silhouette function from the same cluster R package.

Gene set enrichment analysis

Human tumor-specific gene signatures were assessed for enrichment in tp53 tumor types using GSEA 3.0. (Mootha et al., 2003; Subramanian et al., 2005). Gene signatures were assessed for anigosarcoma (Andersen et al., 2013), MPNST (Kolberg et al., 2015), and ERMS (GEO:GSE108022). The ERMS signature was defined by genes up-regulated in both human and zebrafish kRAS-induced tp53 ERMS when compared with normal muscle (log2(FC) ≥2). GSEA was completed in comparing individual tumor types to all other tumors using the default parameters and 1000 permutations of the data.

Comparison of tp53 and tp53 MPNST

RNA sequencing data from four tp53 homozygous mutant MPNST samples were processed as described above. Differential expression analysis was performed as described above, comparing tp53 and tp53 MPNSTs to whole CG1 controls. Statistical significance was assessed with a one-sided Fisher’s exact test on a background of genes that were expressed in at least 4 out of 15 samples.

Molecular signature database (MSigDB) analysis and leukemia similarities with NK cells

The top 500 most differentially regulated genes within each tumor type were identified and assigned human gene IDs using the Beagle database. These humanized gene lists were then queried for overlaps with molecular signatures from MSigDB. Only the top 50 enriched gene sets were analyzed and representative examples of enriched data sets are shown in Figure 3C. For blood cell analysis in Figure 3D,a 30 gene signature was defined for each of the major blood cell lineages and then cumulative gene expression analyzed across tumor types as described by Tang et al. (2017). The top 200 genes up-regulated in tp53 leukemias compared to all other tumors where assessed relative to gene sets generated using SMARTseq as described in Tang et al. (2017).

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "tp53 deficiency causes a wide tumor spectrum and elevates embryonal rhabdomyosarcoma metastasis in zebrafish" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Marianne Bronner as the Senior Editor. The following individual involved in review of your submission has agreed to reveal his identity: Yariv Houvras (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

In this work, the authors have constructed a new tp53 deficient zebrafish strain in a CG1 syngeneic background. This will enable the study of tumors without the need for immunocompromised status. They then go on to show that this loss correlates with invasive and metastatic potential. This work further opens up the field for delineating how this important tumor suppressor can contribute to a variety of tumor progression models.

Essential revisions:

The major experiments asked for by the reviewers can be grouped into the following two categories, one focused on the point vs. del mutant, and the other on the methods used for gene expression analysis:

1) Comparison of the deletion mutant to the point mutant:a) Since Tp53 mutation in zebrafish only induces MPNST, does it promote RMS metastasis or not? This would be very interesting to know in order to further understand how Tp53 contributes to RMS progression.

b) Another interesting question is: Are MPNST gene expression profiles derived from Tp53 mutant fish similar to those of MPNST from Tp53 fish? This may help us to understand how P53 mutation and loss-of-function behave in regulating gene transcription.

c) The lack of change in tumor propagating stem cells is interesting in the tp53 ERMS. Is this also the case for tp53(point mutant) allele ERMS models? If not, this might help bolster the case that tp53 may be a valuable allele.

d) The increased metastasis observed in the tp53 ERMS model may be due to the increase number of spontaneous tumors in tp53. This doesn't mean that there isn't an increase in metastasis, but that the result might be conflated with the fact that the animal may be unhealthy due to already existing tumors. Have the authors looked at whether the GFP+ positive metastasized fish also have non GFP+ tumors?

2) The gene expression analyses: a) Assigning NK cell of origin to the leukemias. The authors have used their prior gene expression data from single cell studies and overlap with known markers in human NK cells to make this claim. Are there human gene expression data on NK cell leukemia that further supports this claim? Can they exclude the possibility that the leukemic blasts they identify are myeloid in origin but lack differentiated markers and represent a more primitive myeloid state. Can they examine myeloid markers on cytospin preparations to further evaluate this possibility? The authors' claim would be strengthened if they are able to relate the gene expression to published NK cell gene expression signatures, if possible. The authors should consider modifying the claim that they can identify the lineage with absolute certainty based on the available data.

b) The gene expression studies are elegant, yet I would like to see more detail included in the manuscript with regard to certain details. What controls did the authors use for each tissue type? The conclusion sentence (subsection “Gene expression analysis of tp53tumors arising in transplant recipient fish”, first paragraph) depends on understanding what the comparison is.

c) The GSEA results for tp53 tumors were not entirely surprising given the tissue origin (i.e. I would expect that the angiosarcoma would have enrichment of vascular genes). The data set would hold a lot more value if the tumor profile for each of the tumors were compared to "normal" tissue of the same type (i.e. leukemia tumor cells compared to normal fish blood cells), rather than whole syngeneic fish. Additionally, have the GO sets been compared to GO sets in human disease, and what is the overlap?

Essential revisions:

The major experiments asked for by the reviewers can be grouped into the following two categories, one focused on the point vs. del mutant, and the other on the methods used for gene expression analysis:

1) Comparison of the deletion mutant to the point mutant:a) Since Tp53 mutation in zebrafish only induces MPNST, does it promote RMS metastasis or not? This would be very interesting to know in order to further understand how Tp53 contributes to RMS progression.

We analyzed the small cohort of tp53 point mutant zebrafish that we maintain in the lab and did not have fish in the correct age class to find MPNSTs (they develop around 14 months of age). Moreover, the point mutant line is not in the CG1 strain background, confounding direct comparison to tp53tumors. Although we agree these experiments would be interesting, potential impacts of strain differences and lack of animals prevented us from being able to complete these experiments during the allotted 3 month resubmission period.

b) Another interesting question is: Are MPNST gene expression profiles derived from Tp53 mutant fish similar to those of MPNST from Tp53del/del fish? This may help us to understand how P53 mutation and loss-of-function behave in regulating gene transcription.

To address this question, we reached out to members of the zebrafish cancer community and have obtained RNAseq data from four tp53 MPNSTs and compared expression signatures to tp53MPNSTs. We identified significant overlap between p53 and p53 MPNST (Author response image 1, p=4e-321 for up-regulated genes and p=5e-182 for downregulated genes; one-sided Fisher’s exact test), suggesting functional commonalities between the different alleles. This is expected given the loss-of-function phenotypes previously reported for tp53(Berghmans et al., 2005). We have added these data to the revised manuscript and included gene lists for these comparisons in Supplementary file 5.

Author response image 1.

Venn diagram depicting overlap between up-regulated and down-regulated genes when comparing homozygous mutant and tp53MPNST to whole adult zebrafish.

c) The lack of change in tumor propagating stem cells is interesting in the tp53del/del ERMS. Is this also the case for tp53(point mutant) allele ERMS models? If not, this might help bolster the case that tp53del/del may be a valuable allele.

Sadly, the tp53 point mutations have not been generated in the syngeneic CG1 strain zebrafish, obviating our ability to complete these interesting analysis. Syngeneic models are required to accurately assess stem cell frequency following limiting dilution cell transplantation (Smith et al., 2010; Blackburn et al., 2012; Ignatius et al., 2012; Blackburn et al., 2014; Ignatius et al., 2017; Hayes et al., 2018; Garcia et al., 2018). Future studies, which are currently beyond the scope of this work, could generate patient-specific point mutations in syngeneic lines and assess for effects on ERMS stem cell frequency.

d) The increased metastasis observed in the tp53del/del ERMS model may be due to the increase number of spontaneous tumors in tp53del/del. This doesn't mean that there isn't an increase in metastasis, but that the result might be conflated with the fact that the animal may be unhealthy due to already existing tumors. Have the authors looked at whether the GFP+ positive metastasized fish also have non GFP+ tumors?

We are sorry that our initial presentation of our experimental design was confusing. Here, GFP-labeled kRAS-induced ERMS were generated in CG1 tp53and tp53zebrafish. Primary tumor cells were then harvested from both genotypes and transplanted into wild-type, non-GFP expressing CG1 recipient animals. These recipient fish do not transgenically express GFP endogenously, and thus any GFP mass must be derived from engrafted cells.

To rule out the possibility of other spontaneous tumors arising in CG1 syngeneic transplant fish, we looked at H&E stained sections of recipient animals engrafted with tp53ERMS and confirmed the presence of only ERMS (n=19). A subset of animals were also assessed by anti-GFP immunostaining on section, confirming that identified tumors were only derived from GFP+ ERMS engrafted cells and did not arise from recipient fish tissues.

2) The gene expression analyses:a) Assigning NK cell of origin to the leukemias. The authors have used their prior gene expression data from single cell studies and overlap with known markers in human NK cells to make this claim. Are there human gene expression data on NK cell leukemia that further supports this claim? The authors' claim would be strengthened if they are able to relate the gene expression to published NK cell gene expression signatures, if possible. The authors should consider modifying the claim that they can identify the lineage with absolute certainty based on the available data.

Unfortunately, expression data sets for human ANKL tumors are not currently available, likely due to the rarity of these tumors.

Yet to directly address this important reviewer comment, we have now completed additional analysis to support the similarity of zebrafish NK cell leukemias with normal NK and NK-like cells from zebrafish. For example, we have expanded the original analysis to show individual gene expression data in Figure 3D (see Figure 3—figure supplement 1D) and updated gene expression analysis in Figure 3E, showing a heat map for expression of well-known NK cell marker genes across tumor types and individuals.

We have also now completed additional experiments to support the assignment of these tumors to the NK cell lineage. Specifically, we identified the top 200 most differentially regulated genes in leukemias compared to all other tumor types and assessed if these genes were differentially expressed within defined blood cell lineages from the zebrafish. We specifically assessed expression using the SMARTseq single cell gene expression dataset from Tang et al., 2017 which included HSC/progenitors isolated as cd41:GFPlow cells from tg(cd41:GFP) transgenic zebrafish, T cells from tg(lck:GFP) transgenic zebrafish, NK cells from rag1-/-, tg(lck:GFP) transgenic zebrafish, myeloid cells from tg(mpx:EGFP) transgenic zebrafish, B cells from marrow-derived tg(rag2:GFP) transgenic zebrafish, and HSCs from tg(runx1 transgenic zebrafish. Significant enrichment was only observed in NK cells (Figure 3—figure supplement 1E, p=0.015, one-sided binomial test), supporting a NK cell origin of tp53 leukemias.

Finally, tp53 NK cell-like leukemias also expressed well-known genes commonly associated with human NK cells, including il2ga and b, jak3, perforins 2, 7, and 8, and these genes were highly up-regulated when compared to all other tumor types in our analysis (Figure 3E).

Can they exclude the possibility that the leukemic blasts they identify are myeloid in origin but lack differentiated markers and represent a more primitive myeloid state. Can they examine myeloid markers on cytospin preparations to further evaluate this possibility?

To address this reviewer comment, we have used our previously published single cell expression data (Tang et al., 2017) to show significant enrichment of NK cell gene expression in tp53leukemias and not other cell lineages (see above and Figure 3 and Figure 3—figure supplement 1).

We have also now included analysis of myeloid genes and confirm that they are not differentially upregulated in leukemias when compared with whole fish (see Figure 3—figure supplement 1D and Supplementary file 7).

b) The gene expression studies are elegant, yet I would like to see more detail included in the manuscript with regard to certain details. What controls did the authors use for each tissue type? The conclusion sentence (subsection “Gene expression analysis of tp53del/del tumors arising in transplant recipient fish”, first paragraph) depends on understanding what the comparison is.

We are sorry for this confusion and have amended the revised manuscript to clarify these comparisons throughout.

c) The GSEA results for tp53del/del tumors were not entirely surprising given the tissue origin (i.e. I would expect that the angiosarcoma would have enrichment of vascular genes). The data set would hold a lot more value if the tumor profile for each of the tumors were compared to "normal" tissue of the same type (i.e. leukemia tumor cells compared to normal fish blood cells), rather than whole syngeneic fish. Additionally, have the GO sets been compared to GO sets in human disease, and what is the overlap?

To validate our assigned tumor designations, we have now assessed if zebrafish tumors express tumor-specific gene signatures identified in human angiosarcoma (Andersen et al., 2013), MPNST (Kolberg et al., 2015) and ERMS (experimentally determined using GEO:GSE108022) (Supplementary file 4). Using Gene set enrichment analysis (GSEA), we report significant enrichment of signatures associated with human angiosarcoma (FDR q-value = 0.001, Figure 3—figure supplement 1A), MPNST (FDR q-value = 0.00433526, Figure 3—figure supplement 1B), and ERMS (FDR q-value = 0, Figure 3—figure supplement 1C) only in the corresponding tp53 zebrafish tumors (Supplementary file 4). Taken together, these data reveal conserved gene expression programs associated with both the predicted cells of origin and corresponding human cancer counterpart.

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Genetic reagent (danio rerio)	CG1	Mizgireuv and Revskoy (2006)
Genetic reagent (danio rerio)	rag2E450fs/E450fs	Tang et al. (2014)
Antibody	anti-tp53 (zebrafish)	Abcam	ab77813, RRID: AB_10864112	WB 5 ug/mL
Antibody	anti-Actin	Sigma	A2066, RRID: AB_476693	WB 1:200
Antibody	anti-mouse HRP	Sigma	NA931, RRID: AB_772210	WB 1:1000
Antibody	anti-rabbit HRP	Cell Signalling Technology	7074, RRID: AB2099233	WB 1:1000
Recombinant DNA construct	rag2:kRASG12D	Langenau et al. (2007)
Recombinant DNA construct	rag2:GFP	Langenau et al. (2007) Langenau et al. (2008)
Recombinant DNA construct	ubi:GFP	Mosimann et al. (2011)
Commercial assay or kit	In situ Cell Death Detection Kit, TMR red	Sigma	12156792910