Dear Editor,

Myelodysplastic syndrome (MDS) is a group of clonal hematopoietic disorders that frequently progress to acute myeloid leukemia (AML) [1]. However, mechanisms underlying such transformation are not yet fully understood. TET2 is one of the most frequently mutated genes in myeloid malignancies [2]. We previously demonstrated that post-translational modification of TET2 protein led to DNA hypermethylation and dysregulated gene expression in MDS hematopoietic stem and progenitor cells (HSPCs), conferring a clonal advantage [3]. TET2 down-regulation was also seen during MDS progression [4]. Herein, we retrospectively analyzed GEO datasets including AML or MDS sample cohort. We found that lower TET2 levels seen in high-risk MDS (HR-MDS) were closely associated with shorter survival (Fig. S1A–C) [5]. Moreover, relative to those with wild type (WT) TET2, AML patients harboring TET2 mutations exhibited a lower survival rate and a higher likelihood of AML secondary to MDS or MPN (Fig. S1D, E) [6]. Collectively, these observations prompted us to assess TET2 function in leukemia transformation of MDS.

To do so, we used a Nup98-HoxD13 (NHD13) transgenic mouse model, in which ~30% of mice develop AML. Interestingly, TET2 levels were lower in c-kit+ bone marrow (BM) cells of leukemia-transformed NHD13 mice relative to age-matched NHD13 mice, which developed MDS exclusively (Fig. S1F, G), suggesting transformation linked to TET2 downregulation. Thus, we crossed Tet2 conditional knockout (KO, Tet2fl/fl/Mx1-Cre) or corresponding control (Tet2fl/fl) mice with NHD13 mice and monitored leukemia development following poly(I:C) treatment on both genotypes (Fig. S1H, I). Notably, Tet2 deletion shortened median survival of NHD13 mice (Fig. 1A). Within 30 weeks, 5 of the 10 mice from the NHD13/Tet2-KO cohort developed AML, while none of the NHD13/Tet2-WT mice exhibited signs of leukemia (Table S1). Specifically, leukemic NHD13/Tet2-KO mice showed increased white blood cell (WBC) counts, splenomegaly and hyper-cellularity in BM, while age-matched NHD13/Tet2-WT mice exhibited only cytopenia (Figs. 1B and S1J–L). NHD13/Tet2-KO mice showed increases in the c-kit+ subset and blasts in BM compared to NHD13/Tet2-WT mice (Figs. 1C, D and S1M). Moreover, secondary recipients also developed AML following transplant of leukemic NHD13/Tet2-KO BM cells (Fig. S1N, O).

Fig. 1: Tet2 deficiency expands the stem/progenitor pool and accelerates leukemia transformation in a murine model of MDS.
figure 1

A Survival of NHD13/Tet2-WT (n = 20; median survival, 347 days) and NHD13/Tet2-KO (n = 16; median survival, 199 days) mice. B WBC count of WT, Tet2-KO, NHD13, and NHD13/Tet2-KO mice (30-weeks-old). C Frequencies of BM c-kit+ cells from indicated mice at 30-weeks-old. D Wright–Giemsa staining of BM cells from indicated mice. Green arrows: dysplastic cells; red arrows: blast cells; scale bars, 20 μm. E Total cell number and percentage of LK subsets in BM of indicated primary mice at a pre-leukemic stage (20-weeks-old). F Apoptosis of LK population in the BM of indicated mice based on Annexin V staining. G Lethally-irradiated mice were transplanted with 2 × 105 LK cells from pre-leukemic NHD13/Tet2-WT (n = 10) or NHD13/Tet2-KO (n = 18) mice plus 2 × 105 unfractionated WT support cells. Shown is chimerism of donor-derived cells (CD45.2+) in PB of recipient mice at different time points. *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001.

Full size image

NHD13 transgenic mice were characterized by HoxA9 elevation [7]. Thus we evaluated TET2 function in MDS or AML patients that showed differences in HOXA9 expression. While HOXA9 or TET2 levels alone did not predict prognosis of the entire MDS population, the HOXA9high/TET2low combination predicted shorter survival relative to those with HOXA9low/TET2high (Fig. S1P–R). Moreover, TET2 mutation also predicted shorter overall survival in the HOXA9high AML population (Fig. S1S, T). We then transduced Tet2-WT or Tet2-KO BM cells with HoxA9 and transplanted the cells into recipient mice to monitor leukemia development. We observed that 6 of 8 HoxA9/Tet2-WT recipients survived up to 200 days, while all 8 recipients of HoxA9/Tet2-KO cells developed lethal AML, starting at day 62 (Fig. S1U, V), suggesting that Tet2 deletion promotes AML transformation and in agreement with outcomes seen in NHD13/Tet2-KO mice.

To define mechanisms underlying MDS progression, we evaluated the Tet2-KO vs. Tet2-WT NHD13 BM compartment at a pre-leukemic stage (20-weeks-old). At that time point, neither genotype showed signs of leukemia (Fig. S2A), but Tet2 deletion increased the number of monocytes in BM of NHD13 mice (Fig. S2B–D). Importantly, BM cells from NHD13/Tet2-KO mice showed an increase in the Lin-c-kit+Sca-1− (LK) population relative to those of Tet2-WT NHD13 mice, whereas the Lin-c-kit+Sca-1+ (LSK) population was unchanged by Tet2 deletion (Figs. 1E and S2E). Increases in the LK subset are likely due to decreased apoptosis following Tet2-KO (Fig. 1F). Within the LK subset, we observed increases in common myeloid progenitors (CMPs) and granulocyte-monocyte progenitors (GMPs) in NHD13/Tet2-KO mice (Fig. S2F, G). Moreover, Tet2 loss did not alter the cell cycle of c-kit+ cells (Fig. S2H) or that of the LK subset (not shown). Colony-forming cell (CFC) assays revealed that NHD13/Tet2-KO BM cells formed colonies in the absence of cytokines (Fig. S2I). In the presence of cytokines, Tet2-KO cells exhibited a slightly higher CFC number than did Tet2-WT cells, the difference was further amplified in the context of NHD13 (Fig. S2J, K). Tet2-KO cells also exhibited higher replating capacity than did Tet2-WT cells (Fig. S2L). We next transplanted LK cells (CD45.2+) from pre-leukemic NHD13/Tet2-KO or corresponding control NHD13 mice into lethally-irradiated secondary recipients to assess leukemogenicity. As expected, NHD13/Tet2-KO cell transplantation increased the percentage of CD45.2+ cells and WBCs in peripheral blood (PB) relative to NHD13/Tet2-WT cells (Figs. 1G and S2M, N). By 16 weeks post-transplantation, 14 of 18 NHD13/Tet2-KO recipients developed AML and exhibited increased numbers of c-kit+ cells and blasts in PB (Fig. S2O, P). Notably, NHD13/Tet2-KO transplants showed shorter survival than NHD13/Tet2-WT transplants (Fig. S2Q). Moreover, we also analyzed the transplants using donor c-kit+ cells (CD45.2+) from WT or Tet2-KO mice (Fig. S2R, S) and observed that recipients from both genotypes survived up to 24 weeks following transplantation, with no signs of leukemic transformation (data not shown). Collectively, these results indicate that Tet2-KO-mediated leukemogenesis is associated with expansion of the MDS HSPC (LK subset) pool.

Tet2 loss in HSPCs can lead to hypermutagenicity [8]. To evaluate these outcomes, we performed whole-exome sequencing of c-kit+ cells from pre-leukemic NHD13/Tet2-KO vs. matched NHD13/Tet2-WT mice. Relative to NHD13/Tet2-WT mice, we observed 271 newly acquired alterations and 199 alterations with increased variant allele frequency (VAF, fold-change >1.5) in NHD13/Tet2-KO mice (Fig. S3A and Table S2). KEGG analysis of these alterations (271 + 199) in NHD13/Tet2-KO mice revealed significant enrichment of genes related to cancer and signaling pathways (Fig. S3B). We next focused on the top 70 altered genes (VAF fold change>2, P < 0.05) and ranked them based on association with AML prognosis (http://precog.stanford.edu) (Table S3). Accordingly, we selected Setd2, Arih2 and Tet3 for further study due to their higher average VAF (VAF > 50%) and mutation ratio (mutation ratio >0.5) (Fig. 2A). Pairwise cooperativeness distribution of these genes showed that Tet3 mutation had the highest pairing associations with other mutations (Fig. S3C). To assess the impact of Setd2, Arih2, and Tet3 loss-of-function mutations, we transduced NHD13 c-kit+ cells with shRNA targeting each gene individually and performed a CFC assay (Fig. S3D). In the first plating, among all genes analyzed, Arih2 knockdown (KD) promoted the greatest increase in colony and cell number of NHD13 c-kit+ cells (Fig. 2B). However, Arih2-KD cells did not show higher replating capacity (Fig. S3E), indicating that Arih2 loss merely impacts proliferation.

Fig. 2: Vitamin C treatment mimics Tet2 restoration and blocks leukemogenesis.
figure 2

A Depicted are 37 genes identified from PRECOG based on two conditions (VAF > 50%, mutation ratio >0.5). Red dots represent preferential enrichment of mutations associated with AML poor prognosis. B c-kit+ BM cells from NHD13 mice were transduced with shRNA targeting indicated genes and then plated for CFC. Shown are colony number (left) and cell number (right) of shArih2 or shCtrl cells after first plating. C Overall survival of total MDS patients, stratified by ARIH2 expression levels. D 5hmC peaks around mutation sites (±1 kb) were counted in Tet2-WT or Tet2-KO NHD13 mice (P = 0.0265, paired t test). E Representative University of California Santa Cruz (UCSC) tracks showing 5hmC peaks (red boxes) associated with Arih2 (Chr9: 108607200–108607400) and Mtss1 (Chr15: 58979400–58979800) loci. Tracks showing 5hmC enrichment were from biological replicates of BM c-kit+ cells from NHD13/Tet2-WT (N) or NHD13/Tet2-KO (NT) mice. F Vitamin C treatment of mice reconstituted with NHD13/Tet2fl/fl and NHD13/Tet2fl/fl/Mx1-Cre BM. Tet2 was deleted by injection of poly(I:C) at 6 weeks post-transplant. Mice were intraperitoneally injected with normal saline (vehicle) or ascorbate (ASC, 4 g/kg), 5 days a week for 16 weeks. WBC counts were monitored at 24 weeks post-transplant. G NHD13/Tet2fl/fl, NHD13/Tet2-/-, and NHD13/Tet2-/-/shTet3 BM c-kit+ cells were treated with vehicle or ASC (0.25 mM) for 3 days and relative 5hmC levels in total DNA were measured by ELISA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Full size image

To define ARIH2 function in leukemogenesis, we retrospectively analyzed GEO datasets and observed lower ARIH2 expression associated with shorter survival in MDS and AML patients (Figs. 2C and S4A). Moreover, SETD2 or TET3 levels alone were not associated with MDS prognosis (Fig. S4B, C). Genetic deletion of endogenous ARIH2 significantly increased FGFR1 protein levels as well as downstream ERK, STAT5 and AKT activity in FGFR1-proficient K562 cells (Fig. S4D). Notably, in ARIH2-KD K562 cells, overexpression of ARIH2WT down-regulated FGFR1 protein and its downstream effectors, while overexpression of ARIH2K441N (analogous to the mouse Arih2K440N) did not, and cells harboring the mutant exhibited a growth advantage relative to ARIH2WT cells (Fig. S4E, F). Similarly, ARIH2 overexpression also inhibited growth of MDS-L/TET2-KD cells relative to MOCK controls (Fig. S4G–I). To extend this analysis in vivo, we injected ARIH2WT or MOCK MDS-L/TET2-KD cells into NSGS mice and observed that ARIH2WT overexpression significantly decreased MDS-L/TET2-KD cell engraftment in BM relative to MOCK controls (Fig. S4J).

To assess whether TET2 deficiency underlies increased mutation frequency, we analyzed GEO datasets and found that MDS or AML patients with TET2 mutations harbor a greater number of mutational events (excluding TET2 itself) compared to those with TET2-WT (Fig. S5A) [9,10,11]. Additionally, ASXL1 mutation also predicted a higher mutation frequency in MDS and AML patients, whereas other mutations analyzed did not (Fig. S5B–D). Further analysis of tumor mutational burden (TMB) profiles from The Cancer Genome Atlas (TCGA) leukemia cohort showed that patients with mutations in either TET2 or ASXL1 displayed significantly higher TMB levels than those with respective WT genes, whereas mutations in DNMT3A, NRAS or FLT3 alone did not predict higher TMB (Fig. S5E, F). An HPRT assay showed that TET2-deficient K562 cells exhibited an increase in HPRT spontaneous mutation frequency relative to controls (Fig. S5G).

The hMeDIP-seq analysis of murine MDS HSPCs indicated that Tet2 deletion decreased overall 5hmC levels (Fig. S5H). Moreover, correlation of 5hmC enrichment with mutation loci showed a greater number of 5hmC peaks at mutation sites in NHD13/Tet2-KO c-kit+ compared to NHD13 c-kit+ cells (Fig. 2D). Specifically, NHD13/Tet2-KO cells exhibited a 5hmC peak enriched at the Arih2 locus but decreased 5hmC enrichment at the enhancer of Mtss1, a reported Tet2 target locus (Fig. 2E). Moreover, hMeDIP-qPCR analyses also revealed that Tet2 deletion increased 5hmC enrichment at mutation sites of Arih2, Setd2 and Tet3 (Fig. S5I), in agreement with reports of others that TET2 loss may enrich 5hmC peaks at mutation loci [8].

Given that vitamin C treatment mimics effects of Tet2/Tet3 restoration [12], we treated transplant mice reconstituted NHD13 or NHD13/Tet2-KO BM with either normal saline (vehicle) or ascorbate (ASC, the dominant form of vitamin C). Relative to vehicle-treated NHD13/Tet2-KO recipients, ASC treatment in NHD13/Tet2-KO mice significantly decreased WBC counts and the frequency of c-kit+ BM cells (Figs. 2F and S6A). At 24 weeks post-transplant, 5 of the 7 NHD13/Tet2-KO recipients from the vehicle cohort developed AML, while none of the ASC-treated mice showed signs of leukemia (Fig. S6B). Notably, vitamin C effects in BM cells from NHD13/Tet2-KO mice were partially dependent on Tet3, as 5hmC increases seen in cells from NHD13/Tet2-KO mice were markedly attenuated upon Tet3 KD (Fig. 2G). Moreover, an HPRT assay in K562/TET2-KD cells confirmed that vitamin C treatment prevented mutagenicity induced by TET2 deficiency (Fig. S6C). Finally, CFC assays with purified blasts (CD34+) from two TET2 mutant high-risk MDS patients (Table S4) showed that vitamin C treatment decreased CFC of MDS HSPCs, while sparing healthy CD34+ cells (Fig. S6D).

In summary, our results indicate that TET2 activity prevents further transformation of MDS HSPCs by decreasing the occurrence of secondary mutations, and that pharmacological enhancement of TET activity may represent an optimal strategy to block MDS malignant transformation.