Transgenic IDH2R172K and IDH2R140Q zebrafish models recapitulated features of human acute myeloid leukemia

Isocitrate dehydrogenase 2 (IDH2) mutations occur in more than 15% of cytogenetically normal acute myeloid leukemia (CN-AML) but comparative studies of their roles in leukemogenesis have been scarce. We generated zebrafish models of IDH2R172K and IDH2R140Q AML and reported their pathologic, functional and transcriptomic features and therapeutic responses to target therapies. Transgenic embryos co-expressing FLT3ITD and IDH2 mutations showed accentuation of myelopoiesis. As these embryos were raised to adulthood, full-blown leukemia ensued with multi-lineage dysplasia, increase in myeloblasts and marrow cellularity and splenomegaly. The leukemia cells were transplantable into primary and secondary recipients and resulted in more aggressive disease. Tg(Runx1:FLT3ITDIDH2R172K) but not Tg(Runx1:FLT3ITDIDH2R140Q) zebrafish showed an increase in T-cell development at embryonic and adult stages. Single-cell transcriptomic analysis revealed increased myeloid skewing, differentiation blockade and enrichment of leukemia-associated gene signatures in both zebrafish models. Tg(Runx1:FLT3ITDIDH2R172K) but not Tg(Runx1:FLT3ITDIDH2R140Q) zebrafish showed an increase in interferon signals at the adult stage. Leukemic phenotypes in both zebrafish could be ameliorated by quizartinib and enasidenib. In conclusion, the zebrafish models of IDH2 mutated AML recapitulated the morphologic, clinical, functional and transcriptomic characteristics of human diseases, and provided the prototype for developing zebrafish leukemia models of other genotypes that would become a platform for high throughput drug screening.


INTRODUCTION
Isocitrate dehydrogenases (IDH) are a group of enzymes that catalyze the conversion of isocitrate to α-ketoglutarate in the physiologic citrate acid cycle [1]. Mutations of IDH include IDH1 R132 , IDH2 R140 , and IDH2 R172 in which arginine is substituted and they occur in more than 25% cases of acute myeloid leukemia (AML) with normal cytogenetics [2]. IDH mutations confer novel substrate specificity to the enzyme and instead of converting isocitrate to α-ketoglutarate, mutated IDH2 convert the latter to 2-hydroxyglutarate (2-HG). 2-HG is an oncometabolite and is associated with epigenetic alteration, genetic instability, and malignant transformation of hematopoietic cells [3][4][5]. Transgenic and knock-in mouse models of IDH1 R132 and IDH2 R140Q mutated AML have been reported, showing that these mutations, singly or in combination with co-existing mutations, induced leukemogenesis [6][7][8][9]. However, animal models of mutated IDH2 R172K with clinicopathologic characteristics of human AML are scarce [6,7,10,11].
Zebrafish has emerged as a model organism to study human diseases, including leukemia [12]. The optical transparency and high fecundity are distinct advantages, and the zebrafish genome and hematopoietic system are remarkably similar to those in mice and human [13]. Moreover, recent advances in genome editing, transgenesis, and rapid embryonic development have made zebrafish a unique model for studying mutation combinations at high throughput [14]. Over-expression of human IDH1 R132H has been shown to induce myelopoiesis in zebrafish embryos, suggesting that the pathogenetic pathway in IDH mutation is conserved in zebrafish [15,16].
In this study, we established transgenic zebrafish models of IDH2 R172K that recapitulated clinicopathologic features of human AML. Comparative studies on IDH2 R140Q highlighted both similarities and differences between the two IDH2 mutants in leukemogenesis. These models may provide important platforms for high throughput drug screening targeting IDH mutations in AML.

RESULTS
Effects of human IDH2 mutations on myelopoiesis Zebrafish idh2 exhibited remarkable similarities in amino acid sequence and syntenic neighboring genes to those of humans ( Fig. S2A, B), suggesting orthologous relationships. To examine the effects of human IDH2 mutations on myelopoiesis at the embryonic stage, IDH2 R172K and IDH2 R140Q mRNA were microinjected up to 200 pg into wildtype zebrafish embryos at 1-cell stage. Expression of IDH2 mutations induced a marked increase in 2HG level, surrogate of mutant IDH2 expression (Fig. S3A). Intriguingly, the equivalent amount of mRNA induced a substantially higher increase in 2HG in IDH2 R172K than in IDH2 R140Q injected embryos. Transient expression of IDH2 R172K or IDH2 R140Q mRNA had little effect on primitive myeloid progenitor as shown by whole mount in-situ hybridization (WISH) of pu.1 (Fig. S3B, C); however, both IDH2 mutations induced a remarkable increase in definitive hematopoietic stem and progenitor cells (HSPC) (cmyb) (Fig. S3B, D) and neutrophils (myeloperoxidase, mpo and Sudan Black B, SBB staining) (Fig. S3B, E, F). The prominent changes in embryonic myelopoiesis induced by IDH2 mutations prompted us to generate transgenic zebrafish lines with stable expression of IDH2 mutations in a lineage-specific manner. The generation of stable transgenic lines was described in "Materials and Methods" and Supplementary Information. F1 transgenic embryos were examined for hematopoietic gene expression by WISH and genotyped individually afterward. Interestingly, lineage-specific expression of IDH2 mutations did not alter the abundance of HSPC or myeloid cells ( Fig. 1A-C) in the embryos. On the other hand, only Tg(Runx1:IDH2 R172K ) but not Tg(Runx1:IDH2 R140Q ) embryos showed a significant increase in rag1 (T-cell marker) expression compared with wildtype siblings (Fig. 1A, D). F2 double transgenic embryos were generated by crossing F1 Tg(Runx1:IDH2 R172K ) or Tg(Runx1:IDH2 R140Q ) to Tg(Runx1:FLT3 ITD ) [17] to investigate potential synergistic effects of these mutant genes which may co-exist in AML patients. Embryos co-expressing FLT3 ITD and IDH2 R140Q or IDH2 R172K showed a significant increase in markers associated with HSPC (cmyb, runx1) (Fig. 1A, B; Fig. S4A, B); neutrophils (mpo; SBB) ( Fig. 1A, C; Fig. S4A, C) and pan-leukocyte marker l-plastin (Fig. S4A, D), when compared with siblings carrying single or no mutation. An increase in T-cell marker (rag1) (Fig. 1A, D) in the developing thymus was only seen in embryos carrying IDH2 R172K but not IDH2 R140Q irrespective of co-existing FLT3-ITD. Markers associated with primitive erythropoiesis (gata1, hbae1.1) and early myeloid progenitor (pu.1) were unaffected (Fig. S4A, E, F, G).
To investigate the mechanisms of myeloid leukemogenesis in the transgenic fish, the HSPC-MPP populations were further examined for possible lineage skewing. Eight sub-clusters were identified, including three myeloid subclusters with upregulation of myeloid gene s100a10b, one HSC subcluster characterized by myb expression and four lymphoid subclusters with increased expression of igic1s1 ( Fig. 4E and Fig. S7B). Both mutant IDH2 transgenic fish showed a significant increase in the prevalence of myeloid subclusters and a significant decrease of lymphoid subclusters (Fig. 4F), suggestive of HSPC-MPP priming towards a myeloid fate.
To examine the effects of IDH2 mutations on lineage differentiation, lineage trajectory and pseudotime analyses were performed for erythroid and myeloid lineages (Fig. 5A-H). Lineage-specific progenitors and terminally differentiated mature cells were differentiated from HSPC-MPP, with increasing pseudotime values along differentiation (Fig. 5A, B, E and F). Pseudotime of lineage-specific progenitors and erythrocytes from Tg(Runx1:FLT3 ITD IDH2 R140Q ) and Tg(Runx1:FLT3 ITD IDH2 R172K ), and that of neutrophils from Tg(Runx1:FLT3 ITD IDH2 R140Q ), were significantly lower than those of their WT siblings (Fig. 5C, D, G and  H). Interestingly, while these mutation combinations induced myeloid priming of HSPC-MPP, as shown by a significant increase in the prevalence of myeloid subclusters and significant decrease of lymphoid subclusters within this population, they also induced differentiation blockage downstream of HSPC-MPP, as shown by the lower pseudotime in the downstream populations.
To further evaluate the effects of IDH2 mutations on the initiation and promotion of leukemogenesis, gene set enrichment analysis (GSEA) was performed for early hematopoietic cell populations, including HSC, HSPC, myeloid, and erythroid progenitor cell clusters, based on differentially expressed genes between Tg(Runx1:FLT3 ITD IDH2 R172K ), Tg(Runx1:FLT3 ITD IDH2 R140Q ) and their WT siblings (Fig. 5I). In both double transgenic fish, genes associated with MTORC, MYC and RAS signaling, were positively enriched in most cell clusters, consistent with their leukemia phenotypes. Intriguingly, the Tg(Runx1:FLT3 ITD IDH2 R172K ) zebrafish showed positive enrichment of genes associated with interferon responses and signaling whereas the Tg(Runx1:FLT3 IT-DIDH2R140Q ) zebrafish showed positive enrichment of genes associated with interleukin 1-related signaling.

Use of transgenic fish in therapeutic evaluation
The clinical relevance of the zebrafish models was tested at both embryonic and adult stages. Tg(Runx1:FLT3 ITD IDH2 R172K ) and Tg(Runx1:FLT3 ITD IDH2 R140Q ) embryos and their WT siblings were treated with gilteritinib and quizartinib (FLT3 inhibitors) as well as enasidenib (IDH2 inhibitor), which have been shown to induce clinical response and confer a survival advantage to patients with FLT3 ITD and IDH2 mutations (Fig. 6A). Using cytochemical staining with SBB as a surrogate for embryonic myelopoiesis, these inhibitors ameliorated the increase in myelopoiesis in the double transgenic embryos (Fig. 6B-E). The therapeutic responses were also tested in adult fish (Fig. 6F). Initial dose-finding studies showed that daily gavage of quizartinib at 10 mg/kg and enasidenib at 100 mg/kg were compatible with normal fish survival (Fig. S8A, B). Double transgenic leukemic fish and their WT siblings were treated with 14 days of quizartinib, enasidenib, or their combination, and their KM cellularity, blast, neutrophil, and erythrocyte counts were enumerated (Fig. 6G-K). Quizartinib but not enasidenib monotherapy significantly reduced cellularity in the KM of the double transgenic fish (Fig. 6G). Enasidenib but not quizartinb monotherapy significantly increased the percentage of neutrophil and erythrocyte and slightly decreased the blast cell population in the KM of the double transgenic fish (Fig. S8C-I). Combination treatment significantly reduced blast population (Fig. 6H, I), increased neutrophil abundance (Fig. 6H, J), and restored erythropoiesis in the KM (Fig. 6H, K). When the effects of these therapeutic agents on the spleen were examined, only the combination of quizartinib and enasidenib effectively reduced the size of the spleen (Fig S8J, K).

DISCUSSION
IDH2 R172K and IDH2 R140Q mutations occur in more than 15% of patients with cytogenetically normal AML. Despite their frequent occurrence, comparative studies of their roles in leukemogenesis have been scarce. In this study, we generated transgenic zebrafish models and demonstrated that transgenic expression of IDH2 R172K and IDH2 R140Q in hematopoietic stem/progenitor cells induced myeloid skewing and differentiation blockade at HSPC-MPP levels, resulting in expansion of KM, splenomegaly, myelodysplasia and increase in Mpo+ blasts capable of self-renewal in serial transplantations. The leukemic phenotypes responded to target therapies expected of their mechanisms of action, attesting to the clinical relevance of the zebrafish models and their potential application in the development of personalized medicine. These findings were consistent with mouse models of transgenic or knockin IDH2 R140Q [7,9,29,30], where IDH2 mutation was shown to induce differentiation block and leukemogenesis, either singly or in combination with other genetic perturbations. Moreover, observations arising from this study may shed important lights to our understanding of leukemogenesis pertinent to IDH2 mutation.
First, our observations of IDH2 R172K AML in zebrafish were consistent with those in viral transduction studies where IDH2 R172K expression in hematopoietic cells induced leukemogenesis [6,7] but was different from the knock-in mouse model in which IDH2 R172K expression in hematopoietic cells led to perturbed lymphoid development but not leukemia [10,11]. Different experimental models and tissue promoters chosen for IDH2 R172K expression as well as cooperative mutation partners might account for the difference in phenotypes. In zebrafish, IDH2 R172K alone was found to induce leukemia-like phenotypes in a small percentage of the transgenic zebrafish, and full-blown leukemia was developed in those fish with IDH2 R172K in the combination of FLT3 ITD . Their cooperativity was also shown by the superior pharmacologic responses to the combination of enasidenib and quizartinib, inhibitors of IDH2 R172K and FLT3 ITD , underscoring the pathogenetic role of each mutant gene in this model.
Second, we demonstrated that in addition to IDH2 R140Q , transgenic expression of IDH2 R172K in HSPC induced differentiation blockade of hematopoiesis, as illustrated by trajectory analyses of the single-cell transcriptome. The lower pseudotime of lineagespecific progenitors and mature neutrophils and erythrocytes as they differentiated from HSPC-MPP supported the proposition of differentiation blockade in both IDH2 mutant AML. Furthermore, the rapid recovery of neutrophils and erythrocytes upon treatment with enasidenib and its combination with quizartinib also suggested a relief of differentiation blockade reminiscent of patient response to enasidenib.
Third, single-cell transcriptome analysis of the KM has empowered us to examine the lineage development and hematopoietic phenotypes of zebrafish AML in detail. Both Tg(Runx1:FLT3 ITD IDH2 R172K ) and Tg(Runx1:FLT3 ITD IDH2 R140Q ) showed multi-lineage hematopoietic expansion in KM as compared with wildtype siblings. Expression of these transgenes in HSPC as driven by the Runx1 enhancer and mouse β-globin minimal promoter could lead to clonal expansion of both lymphoid and myeloid progenitors, notably CLP and CMP. A close examination of HSPC-MPP showed full-blown lineage skewing typical of myeloid neoplasm, suggesting leukemogenesis might begin at the HSPC-MPP stage in the zebrafish model. Moreover, GSEA analysis revealed potential associations between Tg(Runx1:FLT3 ITD IDH2 R172K ), Tg(Runx1:FLT3 ITD IDH2 R140Q ) and immune activation with particular reference to interferon and IL-1 signal activation. Their mechanistic links would have to be further evaluated.
Finally, information arising from this study has shed important lights to the hitherto undescribed hematopoietic effects of IDH2 mutations. In particular, IDH2 R172K but not IDH2 R140Q , accentuated T-cell development at both embryonic and adult stages. The effects were cell-autonomous and IDH2 R172K mutation could be demonstrated in the thymus. It was unclear if the preferential effects of IDH2 R172K on T-cell development were related to a significantly higher level of oncometabolite 2HG, as demonstrated in zebrafish embryos and mice [10]. To our knowledge, IDH2 R172K has not been reported in precursor T-cell lymphoblastic leukemia. Whether the IDH2 R172K transgenic zebrafish would provide a disease model of mature T-cell neoplasm, including angioimmunoblastic T-cell lymphoma, in which IDH2 R172K mutation occurs in 20% of cases [31][32][33], would have to be further investigated [10].
The zebrafish model was of clinical relevance. As proof-ofprinciple, both zebrafish embryos and adults carrying FLT3 ITD IDH2 R172K and FLT3 ITD IDH2 R140Q mutations responded to quizartinib/gilteritinib, and enasidenib, which were effective agents for FLT3 ITD [34,35] and IDH2 mutant AML [36,37]. The optical transparency, high fecundity and relatively simple husbandry of zebrafish have made it uniquely suitable for high throughput drug screening. Furthermore, the availability of lineage-specific fluorescent reporter lines has also facilitated the evaluation of drug effects on specific lineages. Intriguingly, increased HSPC and enhanced primitive myelopoiesis were observed only in monogenic IDH2 R172K and IDH2 R140Q expression in the transient but not stable transgenic model. The apparent discrepancy might be explained by a ubiquitous and higher level of mutant gene expression in the transient system. Whether the transient expression system might provide a more robust high throughput model for zebrafish drug screening would have to be further tested. Data are mean ± s.e.m and statistical analysis was performed by Student's t test (treated vs. untreated for each genotype), *P < 0.05, **P < 0.01, ***P < 0.001.
In conclusion, the present study generated zebrafish models of AML carrying FLT3 ITD IDH2 R172K and FLT3 ITD IDH2 R140Q mutations, recapitulating the morphologic, clinical, transcriptomic and functional characteristics of the corresponding human diseases. These double transgenic fish will become prototypes of zebrafish AML models carrying mutation combinations and powerful tools for rapid drug discovery targeting specific drive mutations.

Generation of transgenic zebrafish lines
Tol2 transgenesis was used to generate single transgenic fish lines Tg(Runx1: IDH2 R140Q ) and Tg(Runx1: IDH2 R172K ) in which human IDH2 R140Q or IDH2 R172K were expressed under the control of the HSPC-specific Runx1 + 23 enhancer and mouse β-globin minimal promoter (Fig. S1A) [38]. Human sequence of IDH2 R140Q or IDH2 R172K was cloned into pDONR221 vector to generate the "middle" clone (pME-IDH2 R140Q or IDH2 R172K ) by Gateway BP reaction. The final Tol2 integrable construct was generated via multisite Gateway LR reaction in which three entry clones [p5E-Runx1 + 23 (Addgene #69602), pME-IDH2 R140Q or IDH2 R172K , and p3E-mCherrypA] [39] were incorporated into the destination vector pDest-Tol2CG [39] with EGFP fluorescent protein expressed under the control of the cardiomyocyte-specific cmlc2 promoter. The latter served as a marker of successful transgenesis. Single transgenic lines were generated by coinjecting 50-100 pg of the respective Tol2 construct and the Tol2 transposase mRNA into the wildtype (WT) TU embryos at the one-cell stage. Founders were identified by PCR and EGFP fluorescence of the heart. F1 single transgenic fish were generated by outcrossing the identified founder fish with WT TU fish. F1 embryos with positive heart EGFP fluorescence at 2dpf were raised to adulthood (Fig. S1B), and the genotype was further confirmed by genotyping (Fig. S1C). The generation of Tg(Runx1:FLT3 ITD ) transgenic line was described previously [17]. Double transgenic fish Tg(Runx1:FLT3 ITD IDH2 R140Q ) and Tg(Runx1:FLT3 ITD IDH2 R172K ) were generated by crossing the single transgenic lines of Tg(Runx1: IDH2 R140Q/R172K ) with Tg(Runx1:FLT3 ITD ). The expression of FLT3 and IDH2 mRNA in F1 adult fish was confirmed by q-PCR (Fig. S1D). F1 transgenic fish were also outcrossed with different transgenic reporter lines, including Tg(rag2:EGFP) and Tg(mpo: EGFP) for evaluation of thymus and kidney marrow size.

Statistical analysis
Data were assessed for normal distribution with a Shapiro-Wilk normality test using Prism9 (GraphPad, San Diego, CA) and presented as mean ± standard error of the mean (s.e.m) of at least three independent experiments. Sample sizes were determined by power analysis to provide sufficient statistical power to detect differences. Comparisons between group of data were evaluated by two-sided Student's t test, one-way ANOVA, or Wilcoxon Rank Sum test if the data were normally distributed and the variance was equal. Survival data were evaluated by Kaplan-Meier analyses and compared using Log-Rank test. P value less than 0.05 was considered statistically significant.
Additional methods and materials used in this study are provided in the Supplementary Information.