Previously, a mosaic transgenic model of RAS-induced rhabdomyosarcoma (RMS) was developed, in which the rag2 promoter drives expression of activated human kRASG12D within mononuclear muscle cell populations (Langenau et al., 2007). In these studies, co-injection strategies were utilized to label tumor cell populations based on muscle differentiation status. Specifically, rag2-kRASG12D and rag2-dsRED2 were co-injected into alpha-actin-GFP transgenic animals. Cells were isolated from the externally visible portion of the tumor, and a single cell suspension was obtained by mincing RMS in the presence of liberase, a collagenase blend used for tissue dissociation (Langenau et al., 2007). Cells were then filtered (40 μm mesh) and analysed by FACS (fluorescence activated cell sorting). Zebrafish RMS contained mononuclear dsRED2 + cells that express early muscle markers and comprise the tumor-initiating stem cell, whereas alpha-actin-GFP + cells comprised more differentiated tumor cell types. Moreover, the double negative cell population, which comprises a bulk of the RMS tumor mass (75–80% of all mononuclear cells), predominantly comprised blood cells and was non-tumorigenic in transplantation assays (Langenau et al., 2007). In fact, co-injection of rag2-kRASG12D and rag2-dsRED2 or rag2-GFP into one-cell stage embryos leads to highly efficient co-segregation and co-integration of transgenes within a majority of developing RMSs (green fluroscent protein (GFP) expressing tumors, n = 14 of 14; dsRED, n = 67 of 69; Figures 1a–d, Supplementary Figure 1 and Supplementary Methods and Text), providing a novel method to express multiple transgenes within a developing tumor. Taken together, these experiments establish that transgenes co-segregate within RMS and can be used to label tumor cells. Moreover, they suggest that it may be possible to inject two or more transgenes that act synergistically to alter the kinetics of malignant transformation.

To assess if co-segregation of transgenes is a common feature of mosaic transgenic tumors, transgene co-segregation was also analysed in Myc-induced T-cell acute lymphoblastic leukemia (T-ALL) (Langenau et al., 2003). Specifically, the rag2-mouse cMyc (rag2-mMyc) transgene was co-injected into one-cell stage embryos along with either rag2-GFP or rag2-dsRED2. All T-ALLs that developed in injected animals were fluorescently-labeled (GFP, n = 12 of 12; dsRED2, n = 13 of 13; Figures 1g–j and Supplementary Table 1). Previously, we established that injection of rag2-mMyc into stable transgenic rag2-GFP animals leads to GFP-labeled tumors (Langenau et al., 2003). To determine the efficiency of co-segregation within tumor cell populations, rag2-mMyc and rag2-dsRED2 were co-injected into stable transgenic rag2-GFP animals. T-ALLs that developed were both GFP + and dsRED2 + by microscopic and FACS analysis (n = 3, 42–49 days old, Supplementary Figure 2). In fact, FACS analysis of whole tumor fish revealed that nearly all GFP + cells also express dsRED2 (n = 3, percent co-localization ranged from 98.6–99.6%, Supplementary Figure 2a–d). If the rag2-mMyc transgene integrated independently of rag2-dsRED2 within a portion of tumor cells, then we would have observed a discernable population of tumor cells that are only GFP-labeled; however, this was not observed. Additional morphological analysis showed that blast cells are largely confined to the dsRED2 + sorted cell populations, whereas the dsRED2-negative cells are predominantly red blood cells (Supplementary Figure 2e–h), further demonstrating that T-ALL tumor cells are confined to the dsRED2 + cell population. These data suggest that transgenes segregate together in T-ALL and are coexpressed in established tumors.

Limiting dilution analysis and cell transplantation independently confirmed that dsRED2 + T-ALL cells from animals co-injected with rag2-mMyc and rag2-dsRED2 were tumorigenic ((Tropepe et al., 2000), Supplementary Figure 2i–n and Supplementary Table 2). FACS was used to isolate dsRED2 + and dsRED2-negative cells from T-ALLs that developed in rag2-GFP stable transgenic animals co-injected with rag2-mMyc and rag2-dsRED2 (n = 3, dsRED2 + sort purity 98.2–99.7%, negative sort purity 88.8–96.7%, Supplementary Figure 2e–f and Supplementary Table 2). 1 × 10⁴, 1 × 10³, 1 × 10² and 1 × 10¹ sorted cells were supplemented with red blood cell carrier from non-tumorigeneic AB-strain zebrafish up to 1 × 10⁴ cells per 5-μl volume and introduced into irradiated recipients (n = 11 fish per dilution). Transplant animals were analysed for engraftment at 11 and 21 days post-transplantation under a fluorescent dissecting microscope for both GFP and dsRED2 expression. From this analysis, we found that the numbers of tumor-initiating cells vary widely between Myc-induced T-ALLs, comprising between 1 in 1258 and 1 in 32 046 of the dsRED2 + cells. By contrast, the negative cells contained within these same T-ALLs do not transplant as efficiently (Supplementary Table 2). Tumors that developed in transplant recipients were positive for both GFP and dsRED2, whereas no single positive GFP + tumors were observed (n = 41), again strongly suggesting that rag2-dsRED2 and rag2-mMyc transgenes co-segregate efficiently in established T-ALLs and that transgenes are not globally silenced over time. Finally, integration of both transgenes into developing T-ALL or RMS tumor cells was independent of the enzyme used to linearize the transgenes (either XhoI or NotI; Supplementary Methods and Text). Additional experiments in the mitfa-BRAF transgenic zebrafish models of melanoma established that co-segregation of transgenes is not specific to rag2 promoter-containing constructs (CJC, CB and LIZ, unpublished data).

Co-injection of three constructs (rag2-oncogene, rag2-GFP and rag2-dsRED2) also led to tumors that expressed all three transgenes (RMS, n = 14 of 17; TALL, n = 22 of 23; Figures 1e, f, k, and l). In RMS, two distinct sub-populations of tumor cells were identified in individual triple injected animals, GFP + and GFP +/dsRED2 +. When each cell population was sorted to relative purity and transplanted into irradiated recipient animals at low cell number (n = 1000–1500 tumor cells along with 20 000 red blood cell carrier cells per animal, GFP + sort, >86% pure and 97.8% viable; GFP +/dsRED2 + sort, >81.2% pure and 89% viable), each cell population led to the development of dsRED2 +/GFP + RMS in irradiated recipient animals by 11 days post-transplantation (data not shown). Similar results were also obtained with additional tumors. Moreover, analysis of multiple rag2-kRASG12D/rag2-dsRED2/rag2-GFP tumors revealed that single positive dsRED2 + cell types were never observed by FACS analysis of total tumor (n = 6). If rag2-dsRED2 and rag2-GFP segregated independently into different tumor cell populations or were transcriptionally silenced, one would expect to find that a large subset of RMSs should have single positive dsRED2 + cell types but not single positive GFP + cells. However, this is never seen. Together these data suggest that injection of three constructs can lead to co-segregation in tumors and suggest that GFP and dsRED2 can differentially label cell populations based on the kinetics of the protein folding and maturation when expressed in rapidly dividing cell populations (GFP maturation time is 8–12 h, whereas for dsRED2 it is 24 h; Clontech protocol PT3404-1, clontech.com).

Three transgenes can also efficiently co-segregate into Myc-induced T-ALL (Figures 1i and k). In contrast to RMS, most T-ALL cells coexpress both GFP and dsRED2 when triple injected at the one-cell stage with rag2-mMyc, rag2-dsRED2 and rag2-GFP. Interestingly, unlike RMS cells that are rapidly dividing and undergoing limited differentiation (Langenau et al., 2007), a majority of T-ALL tumor cells comprise non-dividing blasts (Langenau et al., 2005a). These results suggest that coexpression of fluorescent protein containing transgenes will vary based on the tumor type, the kinetics of fluorophore protein folding, the proliferation of the target cell population and the differentiation status of the tumor of interest. In the end, our results show that multiple transgenes can be introduced into one-cell stage embryos very efficiently and eventually become coexpressed in the developing malignancy.

Stable transgenic approaches have shown that Myc-induced T-ALLs regress following radiation treatment, while transgenic overexpression of bcl-2 suppresses cell death within irradiated T-ALLs (Langenau et al., 2005b). To assess whether co-injection strategies can be utilized to modify radiation sensitivity and recapitulate what was observed in stable transgenic animals, rag2-mMyc and rag2-dsRED2 were co-injected into AB-strain zebrafish. Tumor cells were harvested from primary dsRED2-labeled T-ALL (Figure 2a) and introduced by intraperitoneal injection into irradiated recipients (1 × 10⁶ cells). Transplant recipients developed tumors within 10 days after injection (Figure 2b), which were photographed and then again irradiated (23 Gy). By 4 days after irradiation treatment, dsRED2 expressing tumors were absent in transplant animals (Figure 2c). By contrast, transplanted T-ALLs from animals co-injected with rag2-mMyc and rag2-EGFP-bcl-2 were resistant to irradiation-induced cell death (Figures 2d–f, n = 3 tumors analysed per condition).

Next, we wanted to establish that co-injection methodologies could be used to rapidly introduce multiple transgenes into diverse genetic backgrounds. Specifically, rag2-mMyc and rag2-GFP were co-injected into p53 loss-of-function animals (p53-LOF) (Berghmans et al., 2005) and assessed as above (Figure 2g–i, n = 3). As may be expected based on findings in P53-deficient T cells (Oda et al., 2000), p53-LOF T-ALLs were insensitive to irradiation. Taken together, these experiments highlight the power of this co-injection approach to identify modulators of radiation sensitivity. Moreover, they illustrate the ability to rapidly perform these analyses without the generation of stable transgenic animals that would require lengthy analysis involving the production and identification of transgenic founders; followed by complex breeding schemes and analysis over multiple generations to assess compound genetic effects on selected cancer phenotypes.

Next, we wanted to determine whether co-injection methodologies could be used to identify genetic modifiers of cancer initiation. Having previously established that p53-LOF collaborates with RAS to induce RMS (Langenau et al., 2007), we hypothesized that reestablishing p53 pathway activation in tumors would alter tumor initiation, ultimately affecting tumor penetrance and latency. Noxa is a potent pro-apoptotic BH3-only protein and is a direct transcriptional target of P53 (Oda et al., 2000). The zebrafish noxa ortholog has recently been identified and functionally tested for its ability to induce apoptosis during embryogenesis (Kratz et al., 2006). Given that RAS-induced RMS can be easily visualized in injected animals (Langenau et al., 2007), tumor onset in wild-type AB-strain animals co-injected with rag2-kRASG12D and rag2-noxa was compared to those injected with rag2-kRASG12D and rag2-dsRED2. In this setting, zebrafish noxa is a potent suppressor of tumor initiation (rag2-kRASG12D + rag2-noxa, n = 12 of 174 vs rag2-kRASG12D + rag2-dsRED2, n = 29 of 179; P = 0.007 using a log-rank analysis, Figure 3a). Tumor onset was also compared in p53-LOF animals co-injected with rag2-kRASG12D and either rag2-noxa or rag2-GFP. Because tumor penetrance is much higher in p53-LOF animals, fewer fish were required to show the effect of noxa expression on tumor initiation (n = 10 of 64 in the noxa cohort and n = 20 of 48 in the GFP cohort, P = 0.001, Figure 3b). To verify that rag2-noxa is expressed in established tumors and is not globally transcriptionally silenced, the presence and expression of the rag2-noxa transgene was confirmed by PCR of either genomic DNA isolated from whole tumor or complementary DNA, respectively (n = 8 tumors, Figures 3d and e, primers available in Supplementary Table 3).

P53-deficient tumors that express noxa do not have increased numbers of apoptotic cells when compared to control tumors (n = 6 tumors per group). Specifically, rag2-kRASG12D was co-injected with either rag2-noxa or rag2-GFP into P53-deficient animals at the one-cell stage of life. Animals were scored for tumors at 25 days of life, killed and fixed in 4% PFA. Subsequently, animals were sectioned and stained by TUNEL or hematoxylin/eosin as described previously (Traver et al., 2004). Tumor sections were photographed under ×1000 oil emersion and apoptotic cell number and total cell number were counted per field (n = 3 fields per sample, Supplementary Figure 3). Percentage of apoptotic cells did not differ between the noxa and GFP expressing tumors (0.43 ± 0.31 and 0.94 ± 0.96%, respectively; ± s.d.; P = 0.24; n = 428 ± 134 cells counted per sample), suggesting the existence of a mechanism by which tumor cells evade high levels of noxa transgene expression during tumor development. Taken together, these experiments establish noxa as a potent suppressor of tumor formation in RMS and establish a novel technology to rapidly assess collaborating genetic events in modulating tumor onset.

Heat-shock-inducible transgenic approaches have provided an invaluable tool through which gene expression in development can be modulated (Halloran et al., 2000; Xiao et al., 2003; Burns et al., 2005; Pyati et al., 2005). To establish if co-injection techniques along with heat-shock-inducible transgenesis could be used to elicite gene expression within established tumors, the rag2-kRASG12D, rag2-GFP and hsp70-dsRED2 constructs were co-injected into one-cell stage AB-strain animals. Triple transgenic RMS-affected animals developed GFP-labeled RMS that could induce dsRED2 expression after heat-shock at 37 °C for 45 min (n = 19 of 20, Figure 4). Expression of dsRED2 could be detected in heat-shocked animals from 2–10 days post treatment (Supplementary Figure 4). Importantly, dsRED and GFP expression largely co-localized within tumors that received heat-shock (Figures 4i–k), establishing that the hsp70 promoter is active in most cell types contained in zebrafish RMS (n = 4). In contrast to what is observed in RAS-induced RMS, co-injection of rag2-mMyc, rag2-GFP and hsp70-dsRED2 led to T-ALLs that were both GFP + and strongly dsRED2 + despite never receiving heat-shock treatment (n = 14 of 15, Supplementary Figure 5). Additional confocal analysis established that nearly all T-ALL blast cells expressed high levels of both GFP and dsRED2 in the absence of heat-shock treatment (n = 3, data not shown). In Myc-induced T-ALL, it is likely that the heat-shock response—a stress response that can be affected by other stimuli other than heat—is already active in these malignancies. Although the heat-shock-inducible transgenic approaches will be amenable for use in only a subset of tumor types, our experiments clearly establish that co-injection and inducible approaches can be utilized in zebrafish transgenic models of cancer to selectively turn on genes within established malignancies, providing a novel and rapid methodology to assess gene effects on tumor maintenance.