Molecularly distinct models of zebrafish Myc-induced B cell leukemia

Zebrafish are a valuable leukemia model due to highly conserved hematopoietic and oncogenic pathways, facile genetics, and ease of use in chemical genetic screens. However, until recently, robust zebrafish B-cell leukemia models had not been described [1, 2]. The first transgenic zebrafish leukemia model was created 15 years ago and targeted murine c-Myc (mMyc) to thymocytes of AB strain zebrafish, leading to the rapid development of T-cell acute lymphoblastic leukemia (T-ALL) [3]. Additional genetic models were subsequently developed that result in induction of T-ALL [4], but B-cell leukemia models lagged behind [5, 6].

In two recent reports published in Leukemia, our groups independently demonstrated the development of zebrafish B-ALL using transgenic expression of mMyc or human c-MYC (hMYC) controlled by the zebrafish recombination activating gene 2 (rag2) promoter [1, 2]. Both models shared genetic and phenotypic features, but there were also key differences including strain background and species differences in the MYC transgene that was used to generate each model. Here, we compare and contrast these models, making the important finding that zebrafish develop at least four molecularly-distinct ALL types, including cortical thymocyte-arrested cd4⁺/cd8⁺ T-ALL, ighm⁺ B-ALL, ighz⁺ B-ALL, and biphenotypic T/B-ALL.

Thirteen ALLs were purified from rag2:hMYC;lck:eGFP double-transgenic fish [2]. As previously reported, these leukemias had heterogeneous GFP expression, with T-ALL being exclusively GFP^hi, B-ALL exclusively GFP^lo, and other fish harboring mixed-ALL with both GFP^hi and GFP^lo cells, representing simultaneous T- and B-ALL, respectively [2]. Notably, simultaneous B- and T-ALL in single hMYC fish were frequently observed by Borga et al. [2]. In contrast, mMyc ALL were propagated by single cell allo-transplantation and then assessed by single cell transcript expression, confirming the existence of a single biphenotypic ALL in the mMyc cohort [1]. hMYC ALL were subjected to RNA-seq transcriptomic profiling and compared to transplanted leukemias generated from single mMyc ALL clones described by Garcia et al. [1]. Principal Component Analysis clearly distinguished mMyc-induced T-ALL from B-ALL, with the single mMyc-induced biphenotypic B/T-ALL clustering between these samples (Fig. 1A). The eight hMYC-induced ALLs that were largely GFP^hi clustered with known T-ALLs (hMYC 2, 6, 8–11, 13, 14), while three primarily GFP^lo ALL clustered near the mMyc B-ALLs (hMYC 3–5). Two hMYC-induced ALLs (hMYC1, 12) with substantial populations of both GFP^lo and GFP^hi cells grouped near the mMyc biphenotypic B/T leukemia. Hierarchical clustering using the top 100 positively- and negatively-correlated genes from PC2 confirmed that these genes defined B and T lymphocytes, respectively (Fig. 1B), with B-ALLs expressing cd79b, syk, pax5, blnk and ebf1, while T-ALLs expressed cd8b, lck, runx3 and gata3. As previously reported, the biphenotypic B/T-ALL and mixed hMYC-induced ALLs expressed both T- and B-cell lineage genes [1, 2]. PC2 up-regulated genes were enriched for B cell signaling pathways when independently assessed by GSEAsig (Supplemental Tables 1 and 2).

Fig. 1

Gene expression differences define T- vs. B-lineages in ALL of both the rag2:mMyc and rag2:hMYC;lck:eGFP models. A. Principal component analysis of RNAseq expression profiles of previously-classified mMyc-induced T- (blue, n = 8), B- (red, n = 2), or biphenotypic ALL (pink, n = 1) compared with 13 unknown hMYC-induced ALL (black). B. Heatmap and hierarchical clustering using the top 100 positively- and negatively-correlated genes from PC2. B cell-specific genes are denoted in blue and T cell-specific genes in red at the right

To examine clonality and maturation in mMyc- and hMYC-induced ALL, we next analyzed the expression of constant and variable regions of the T cell receptor β (tcrβ) and immunoglobulins μ and ζ (ighm, ighz; Fig. 2A). Every mMyc and hMYC T-ALL exhibited tcrβ expression, with V(D)J recombination occurring in most samples as determined by expression of specific variable regions [1]. Conversely, mMyc and hMYC B-ALL did not recombine or express tcrβ, but expressed constant regions of ighm or ighz. Ig variable regions were not detected, indicating V(D)J rearrangement likely had not occurred in these leukemias and suggesting B-ALLs arrest at the early pro-B cell stage. Ig constant regions without variable regions are termed ‘sterile transcripts', and these non-coding mRNAs are transiently expressed during both early V(D)J recombination and during Ig class switching. Such sterile transcripts are detectable in mammalian pro-B cells before V(D)J rearrangement is complete, supporting our interpretation that mMyc and hMYC induce pro-B ALL [7,8,9]. As expected, hMYC mixed-ALL contained distinct T- and B-ALL clones expressing both tcr and ig mRNAs, with their relative expression correlating well with the percentage of GFP^hi/T-ALL vs. GFP^lo/B-ALL cells found in each sample (Fig. 2A). Intriguingly, mMyc B-ALL expressed exclusively ighm while hMYC B-ALL favored ighz expression, indicating that mMyc and hMYC might be oncogenic in distinct B cell lineages.

Fig. 2

Identification of two molecularly-distinct B-ALL types arising independently in either the ighm⁺ or ighz⁺ B cell lineages. A. T cell receptor beta and Ig heavy chain expression in individual ALLs. tcrβ recombination is denoted by grey-shaded boxes (left column), with percentage of GFP^hi cells in each rag2:hMYC;lck:eGFP ALL noted in right column. Histograms depict expression of tcrβ and igh constant regions by each ALL. Not available (NA). B. Heatmap showing expression of genes differentially expressed in T-ALL, mMyc/ighm⁺ B-ALL, and hMYC/ighz⁺ B-ALL. C. Geneset enrichment analysis using genes positively correlated with each ALL molecular subtype. T-ALL (blue), mMyc/ighm⁺ B-ALL (red), and hMYC/ighz⁺ B-ALL (orange). Complete gene set and GSEAsig results are provided in Supp. Tables 3 and 4

To further explore differences between these models, we next identified genes uniquely-expressed by T-ALL, mMyc/ighm⁺ B-ALL, or hMYC/ighz⁺ B-ALL (Fig. 2B). As expected T-ALLs expressed known T cell lineage markers, yet mMyc/ighm⁺ and hMYC/ighz⁺ B-ALLs were transcriptionally distinct. mMyc/ighm⁺ B-ALL expressed gfi1ab, zfhx3, notch1a, nf1b, and gtf3aa. By contrast, hMYC/ighz⁺ B-ALL expressed higher cd79a, cd83, mef2cb, and jak2a levels. To further test for differences in these two molecular subtypes of B-ALL, we next performed GSEAsig using these same differentially-regulated genes. From this analysis, we uncovered that mMyc/ighm⁺ B-ALLs exhibited significant enrichment for pathways regulating ribosome biogenesis and RNA binding (Fig. 2C and Supplemental Tables 3 and 4). By contrast, hMYC/ighz⁺ B-ALLs were enriched for intracellular signaling, protein binding, and germinal center B cell maturation pathways. In support of our findings, Liu et al. recently reported the identification of molecularly and biologically distinct ighz⁺ and ighm⁺ B cell lineages using rag2:mCherry; cd79b:GFP transgenic zebrafish [10]. In the context of normal B cell development, ighz⁺ B cells were mCherry^hi/GFP^lo while ighm⁺ B cells were mCherry^hi/GFP^hi. Overall, these results demonstrate that mMyc/ighm⁺ and hMYC/ighz⁺ B-ALLs are not subtle B cell leukemia variants, but rather distinct malignancies that arise in different B cell types with vastly different molecular pathway signatures.

In summary, although zebrafish B cell leukemia models were lacking for many years, our analyses reveal two highly-divergent types of B-ALL. This is surprising, as both models utilize the same promoter (rag2) to regulate a near-identical oncoprotein, c-Myc/MYC, with the only differences being the MYC transgene species of origin and the genetic backgrounds upon which the models were developed. Yet, despite the high molecular similarity of both models, these B-ALL subtypes also show unique gene expression signatures when compared to one another, which likely reflects differences in both their lineage (ighm vs. ighz) and potential differences in MYC transcriptional targets expressed by the early developmental stages of these distinct pro-B cell populations. Our new analysis of these models reconciles the perceived differences in the manuscripts published by our groups, identifying four molecularly-distinct ALL subtypes in zebrafish: cortical cd4⁺/cd8⁺ T-ALL, biphenotypic B/T ALL, ighm⁺ B-ALL, and ighz⁺ B-ALL. Developing a wider array of leukemia models and refining mechanisms that drive their growth, aggression, and stem cell frequency will surely lead to new insights into human disease.