Therapeutic targeting of preleukemia cells in a mouse model of NPM1 mutant acute myeloid leukemia

See allHide authors and affiliations

Science  31 Jan 2020:
Vol. 367, Issue 6477, pp. 586-590
DOI: 10.1126/science.aax5863

Taking preventive measures

Recent technological advances have made it possible to detect, in healthy individuals, premalignant blood cells that are likely to progress to hematologic cancer. These advances in early detection have fueled interest in “cancer interception,” the idea that drugs designed to treat advanced cancer might also be useful for cancer prevention. Uckelmann et al. now provide support for this concept in a study of mice genetically predisposed to develop acute myeloid leukemia. Early administration of an epigenetic therapy that had previously been shown to have anticancer activity in advanced leukemia models was able to eliminate preleukemia cells and extend survival of the mice.

Science, this issue p. 586


The initiating mutations that contribute to cancer development are sometimes present in premalignant cells. Whether therapies targeting these mutations can eradicate premalignant cells is unclear. Acute myeloid leukemia (AML) is an attractive system for investigating the effect of preventative treatment because this disease is often preceded by a premalignant state (clonal hematopoiesis or myelodysplastic syndrome). In Npm1c/Dnmt3a mutant knock-in mice, a model of AML development, leukemia is preceded by a period of extended myeloid progenitor cell proliferation and self-renewal. We found that this self-renewal can be reversed by oral administration of a small molecule (VTP-50469) that targets the MLL1-Menin chromatin complex. These preclinical results support the hypothesis that individuals at high risk of developing AML might benefit from targeted epigenetic therapy in a preventative setting.

Nucleophosmin (NPM1) mutant acute myeloid leukemia (AML) is one of the most common types of AML (13). Despite its high prevalence, the mechanism of leukemogenesis is still poorly understood, and targeted therapy options are lacking (4). NPM1 gene mutations (NPM1c) induce cytoplasmic localization of NPM1 and often co-occur with other mutations in genes such as DNA methyltransferase 3A (DNMT3AR882H). NPM1c leukemias express a distinctive stem cell–like gene expression pattern that includes homeobox cluster A and B (HOXA/B) genes and their DNA-binding cofactor MEIS1 (58). In humans, DNMT3A mutations are detected in the most primitive hematopoietic stem cell compartment, often long before the development of leukemia, a condition often referred to as clonal hematopoiesis of indeterminate potential (CHIP) (9). NPM1 mutations are found in committed progenitors and differentiated myeloid cells in AML but are absent from the stem cell and lymphoid compartments (9, 10). This suggests that NPM1c may induce self-renewal in myeloid progenitors as a critical step in the development of AML and that this aberrant progenitor self-renewal may represent a critical step in the progression from CHIP to AML.

To identify the leukemia-initiating cellular population in NPM1c AML, we used previously developed mouse models with an inducible Cre recombinase (MxCre) and heterozygous conditional knock-in of the humanized Npm1 mutation (Npm1flox-cA/+; hererafter called Npm1c mutant mice), alone or in combination with Dnmt3aR878H mutation (Dnmt3aR878H/+; hererafter called Dnmt3a mutant mice) (5, 11). We confirmed Hox gene up-regulation in different hematopoietic stem and progenitor populations of Npm1c, Dnmt3a, and Npm1c/Dnmt3a mutant mice 16 weeks after induction of the knock-in allele by polyinosinic:polycytidylic acid (pIpC) injection (Fig. 1A). At this time, mutant mice showed no signs of leukemia and had normal blood counts, and only the double mutant showed a slight increase in granulocyte-macrophage progenitor (GMP) frequencies (fig. S1, A and B). Sorted wild-type (WT) and Dnmt3a single-mutant cells showed a stepwise decrease of Hoxa9 mRNA expression from long-term hematopoietic stem cells (LT-HSCs) to GMPs, which coincides with their loss of self-renewal properties (Fig. 1A). Npm1c or Npm1c/Dnmt3a mutant cells maintained inappropriately high levels of Hoxa9 across the different progenitor cell types (Fig. 1A). RNA sequencing (RNA-seq) analysis 4 weeks after activation of the Npm1c allele revealed that half of the top 20 up-regulated genes in Npm1c GMPs were Hoxa/b genes. The HSC-enriched Lin, Sca1+, Kit+ population (LSK) showed much lower fold changes owing to their high baseline expression of Hoxa/b genes (Fig. 1B and table S1). The gene expression programs induced in Npm1c mutant GMPs were also enriched for LT-HSC and human NPM1c mutant AML signatures, which include Hoxa/b genes and Meis1 (fig. S1, C to I). On the basis of these gene expression data, we conclude that Npm1c supports the inappropriate expression of genes associated with normal stem cell self-renewal, such as Hoxa/b cluster genes, throughout myeloid differentiation.

Fig. 1 Npm1c induces self-renewal properties in myeloid progenitor cells.

(A) Hoxa9 gene expression in Npm1c, Dnmt3a, and Npm1c/Dnmt3a mutant LT-HSCs, multipotential progenitors (MPPs), common myeloid progenitors (CMPs), and GMPs 16 weeks after pIpC injection (n ≥ 3 mice per group; error bars indicate mean ± SD). Rel., relative; GAPDH, glyceraldehyde 3-phosphate dehydrogenase. (B) Heatmap showing the top 25 up-regulated genes in Npm1c versus WT LSK cells and GMPs, 4 weeks after pIpC treatment (n = 3 mice per group). (C) Relative expression of Hoxa9 3 days after in vitro Cre transduction (n ≥ 4 mice per group; error bars indicate mean ± SEM). The dotted lines indicate Hoxa9 expression level from freshly isolated LSK cells and GMPs. (D) Peripheral blood percentage CD45.2 engraftment of WT and Npm1c, Dnmt3a, and Npm1c/Dnmt3a mutant GMPs sorted 4 weeks after pIpC induction, transplanted into lethally irradiated recipients. Error bars indicate mean ± SEM. (E) Summary table of GMP-transplanted mouse numbers and percentages engrafted ≥1% for >12 weeks. MIG-CRE, MSCV-CRE-IRES-GFP retrovirus. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

We next investigated whether Npm1c can induce stem cell–associated gene expression de novo in committed progenitor cells, which lack self-renewal and have low levels of Hox and Meis1 expression. For this, we sorted Cre-negative Npm1c, Dnmt3a, and Npm1c/Dnmt3a mutant GMPs and LSK cells and then used retroviral Cre overexpression to induce the mutant knock-in alleles in vitro (Fig. 1C). Npm1c expression induced Hoxa9 expression in GMPs in vitro, suggesting that the Npm1c-driven stem cell–associated program can be turned on at different stages of myeloid differentiation (Fig. 1C). Induction of Dnmt3aR878H knock-in alone did not induce or enhance the Hoxa9 induction activated by Npm1c in progenitors, indicating that mutant Npm1c and not Dnmt3a is responsible for the observed up-regulation of stem cell–associated genes.

Our gene expression data suggest that Npm1c induces stem cell properties in non–stem cells. To examine whether these transcriptional changes coincide with functional self-renewal properties in Npm1c progenitors, we first performed colony-forming unit (CFU) assays. Npm1c mutant GMPs displayed increased in vitro self-renewal capacity, as shown by their ability to replate up to four rounds in CFU assays (fig. S2, A and B). Transplantation experiments performed using in vivo pIpC-induced and in vitro Cre-transduced mutant GMPs demonstrated that Npm1c enhances engraftment and self-renewal of GMPs (Fig. 1D and fig. S2C). Although some of the initially engrafted GMPs were depleted over time, about half of the mice retained self-renewing GMPs for >12 weeks (Fig. 1E). These long-term engrafting GMPs (LT-GMPs) showed CD11b+Gr1+ peripheral blood engraftment, and recipient mice showed no signs of leukemia for more than 6 months (fig. S2D). These experiments demonstrate that self-renewal properties induced by Npm1c in myeloid progenitors are sufficient to give rise to a preleukemic population that stably engrafts long term.

To determine whether these preleukemic Npm1c mutant clones would progress to leukemia, we performed secondary transplants of LT-GMPs (Fig. 2A). Secondary recipients of Npm1c single-mutant or Npm1c/Dnmt3a double-mutant LT-GMPs developed AML 3 to 5 months after secondary transplant similar to mice that received mutant LSK cells (Fig. 2B). LSK- and GMP-derived secondary transplanted mice presented with high white blood cell counts, enlarged spleens, and extramedullary hematopoiesis, suggesting that Npm1c is sufficient to give GMPs enough self-renewal capacity to ultimately generate AML (Fig. 2, C and D, and fig. S3, A and B). The long latency indicates that Npm1c mutant LT-GMPs may acquire further mutations over time, which has been shown to occur in this and other Npm1c knock-in mouse models (7, 12). Furthermore, mouse Npm1c/Dnmt3a mutant leukemia cells showed highly up-regulated Hoxa/b expression, which resembled expression patterns observed in human NPM1c AML and other HOX-associated AMLs such as MLL-AF9 AML (Fig. 2E and fig. S3, C and D). These results confirm that preleukemic Npm1c mutant LT-GMPs eventually give rise to leukemia.

Fig. 2 Myeloid progenitors are leukemia-initiating cells in Npm1c AML.

(A) Experimental overview of secondary transplantation of long-term engrafted mutant GMPs. (B to D) Kaplan-Meyer survival analysis (B), representative spleen images (C), and spleen weights (D) of secondary transplants of MIG-CRENpm1c/Dnmt3a LT-GMPs or LSK-derived cells and MxCreNpm1c LT-GMPs (n ≥ 4 mice per group). One-way analysis of variance (ANOVA) was performed. Error bars indicate mean ± SD. *P < 0.05; ns, not significant. (E) Top 10 up-regulated genes in MIG-CRENpm1c/Dnmt3a mutant leukemic GMPs compared with WT GMPs (n = 3 mice per group).

We have previously shown that inhibition of the interaction between the histone methyltransferase MLL1 and adaptor protein Menin reverses leukemogenic gene expression in the NPM1c AML cell line OCI-AML3 (13). Menin-MLL interaction inhibitors were originally developed to target the oncogenic MLL-fusion complexes by directly disrupting the oncogene complex from assembling on chromatin. Our findings, however, suggest that the WT MLL1-Menin interaction is essential to maintain NPM1c-driven leukemia. To test this, we used an orally bioavailable inhibitor of the Menin-MLL1 interaction (VTP-50469). This compound has been used to treat established disease in models of MLL-rearranged AML and B cell acute lymphoblastic leukemia [see (14) for details on the chemical synthesis of VTP-50469]. We assessed whether Npm1c mutant mouse cells respond to Menin inhibition in serial CFU replating assays of double- and single-mutant cell lines (Fig. 3A and fig. S4A). Menin inhibition led to a rapid loss of replating capacity and up-regulation of myeloid differentiation marker CD11b with no significant increase in apoptosis (fig. S4, B and C). Gene expression analysis of Npm1c/Dnmt3a mutant mouse cells after Menin inhibition revealed a rapid repression of stem cell genes, including Meis1 and Pbx3 (fig. S5A, left). Meis1 and Pbx3 are important cofactors of Hoxa/b transcription factors and play essential roles in Hoxa9-driven leukemogenesis and maintenance of leukemic stem cell gene expression programs (1517). Even though many Npm1c-induced genes, including Hoxa/b, remained highly expressed, the loss of essential cofactors such as Meis1 could account for the loss of self-renewal observed upon Menin inhibition.

Fig. 3 Meis1, Menin, and MLL1 are essential for maintaining self-renewal program.

(A) CFU serial replating assay of mouse MIG-CRENpm1c/Dnmt3a mutant cell line (SIIIL12) with dimethyl sulfoxide (DMSO) or VTP-50469. Data represent the mean of three independent experiments. d7, day 7. (B) CFU assay of mouse SIIIL12 cells transduced with MSCV-PURO control (left) or Meis1-PURO (right) virus and grown in the presence of 10 nM VTP-50469. Data represent the mean of three independent experiments. (C) CFU assay of SIIIL12 cells electroporated with control or Meis1- or Rpa3-targeting single guide RNAs (sgRNAs) for Cas9-mediated KO. Data represent the mean of three independent experiments. (D) Chromatin immunoprecipitation sequencing (ChIP-seq) density plots showing changes in chromatin occupancy of Menin, MLL, and H3K4me3 and changes in mRNA expression in response to VTP-50469 in OCI-AML3 cells at the MEIS1 and HOXA loci. (E and F) MEIS1 (E) and HOXB5 (F) gene expression in OCI-AML3 cells transduced with control, sgMLL1, sgMLL2, and sgMenin. Data represent the mean of three independent experiments. One-way ANOVA was performed. Error bars indicate mean ± SD. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

To confirm that reduced Meis1 expression is crucial for the drastic differentiation of Npm1c/Dnmt3a mutant cells observed after VTP-50469 treatment, we first attempted to rescue the VTP-50469–induced loss of leukemic stem cell gene expression by retroviral overexpression of Meis1. Maintaining Meis1 expression rescued the replating capacity of Npm1c/Dnmt3a mutant cells in the presence of Menin inhibitor and increased the median inhibitory concentration (IC50) values significantly (Fig. 3B and fig. S5B). Whereas control cells lost essential components of their self-renewal program in response to VTP-50469, Meis1-expressing cells showed increased expression of a group of stem cell–associated genes, including Mecom and Pbx3, and retained them in the presence of Menin inhibitor (fig. S5, B to G). Conversely, Cas9-mediated knockout (KO) of Meis1 led to a rapid loss of out-of-frame edited cells in culture as well as a reduction in CFU replating capacity, confirming Meis1 as a dependency in NPM1c mutant AML (Fig. 3C and fig. S5H). These data confirm the essential role of Meis1 in maintaining leukemic stem cell programs.

Next, we confirmed that human NPM1c mutant leukemia cell line OCI-AML3 also responds to VTP-50469. OCI-AML3 cells were highly sensitive to Menin-MLL inhibition, as demonstrated by their low IC50 value (3 nM on day 6) and rapid down-regulation of MEIS1 and PBX3 upon VTP-50469 treatment (fig. S6, A to C). In contrast to previously published Menin inhibitor molecules such as MI-2-2 and MI-503 that were shown to reduce expression of HOXA/B cluster genes as well as MEIS1, HOX genes were not repressed in response to VTP-50469 in OCI-AML3 cells (fig. S6, B and C). In mouse cells, a modest repressive effect on some Hox genes was observed, whereas others were up-regulated (figs. S7, D and E, and S10, A to D) (18, 19). Furthermore, we observed a reduction of Menin and MLL1 chromatin occupancy at the MEIS1 and PBX3 transcriptional start sites (TSSs), whereas MLL1 binding at HOXA/B TSSs was retained in regions where Menin was depleted (Fig. 3D, fig. S6E, and table S2). Globally, Menin chromatin occupancy was decreased, whereas MLL1 and trimethylated histone H3 lysine 4 (H3K4me3) were lost only at specific sites that were highly enriched for genes down-regulated in response to Menin inhibition (figs. S6, F to H, and S7, A to C). To verify that MLL1 loss is responsible for the observed loss of stem cell–associated gene expression, we generated Cas9-mediated OCI-AML3 KO cell lines of MLL1, MLL2, and Menin (fig. S8, A to C, and table S4). Menin KO mimicked the expression changes observed upon VTP-50469 treatment, with reduced MEIS1 and PBX3 expression and up-regulation of HOXB5 and HOXA5 (Fig. 3, E and F, and fig. S8, D and E). Loss of MLL1, however, also resulted in a reduction in HOX expression, whereas MLL2 disruption showed no or only minor effects on HOX and MEIS1 (Fig. 3F and fig. S8E). In agreement with this, Menin and MLL1 KO cells were rapidly depleted in competition assays, whereas MLL2 KO cells were not (fig. S8F). Our findings confirm MLL1 as the main driver of oncogenic HOX and MEIS1 gene expression in NPM1c mutant AML and show that only a subset of MLL1 target genes are also Menin dependent.

DNMT3A mutations are frequently found in patients with CHIP and are associated with increased risk for hematologic malignancies (9, 2023). By contrast, NPM1c mutations have not been reported in CHIP, suggesting that their acquisition is rapidly followed by leukemic progression. This was demonstrated in at least one patient with IDH2 mutant CHIP that developed AML shortly after NPM1c was detected (8, 24). Our mouse model of preleukemic Npm1c/Dnmt3a LT-GMPs allowed us to test whether we can interrupt leukemia progression by means of eradication of Npm1c mutant preleukemic clones. We evaluated the in vivo efficacy of VTP-50469 using secondary transplants of Npm1c single-mutant and Npm1c/Dnmt3a double-mutant LT-GMPs (fig. S9A). Engraftment was confirmed 3 weeks after transplant, and mice were treated with 0.1% VTP-50469–spiked chow for 9 weeks (fig. S9, B and C). In control animals, we observed an expansion of LT-GMP engraftment and eventually mice succumbed to AML (Fig. 4, A and B, and fig. S9D). After 3 weeks, Menin inhibitor–treated preleukemic mice showed a rapid decrease in engraftment (<1%) (Fig. 4A and fig. S9D). Notably, no relapse of LT-GMPs was observed more than 6 months after the treatment was discontinued, and VTP-50469–treated groups showed prolonged survival of more than 9 months versus an average of 5 months in the untreated groups (Fig. 4B and fig. S9E). Furthermore, when VTP-50469–treated mice were sacrificed 300 days after transplant, no Npm1c mutant cells were detected in bone marrow, spleen, or liver (fig. S9, F to H). WT stem cell self-renewal was not affected by VTP-50469 treatment, as demonstrated by stable engraftment of WT HSCs (fig. S9I). Repression of Meis1 and Pbx3 and other stem cell–associated genes was validated by RNA-seq analysis of sorted Npm1c/Dnmt3a LT-GMPs after 5 days of in vivo treatment (fig. S10, A to D). VTP-50469 was well tolerated even when administered for long periods (9 weeks continuously), which could potentially be extended to ensure complete clearance of NPM1c mutant cells if needed. These data indicate that we can specifically eradicate preleukemic Npm1c mutant self-renewing myeloid progenitor cells using targeted epigenetic therapy without having detrimental effects on either normal HSCs or hematopoiesis.

Fig. 4 Preleukemic Npm1c LT-GMPs and human AML cells can be eradicated by Menin inhibition.

(A and B) Percent engraftment of CD45.2 in peripheral blood (A) and Kaplan-Meier survival analysis (B) of mice transplanted with Npm1c/Dnmt3a LT-GMPs receiving control or 0.1% VTP-50469–spiked chow for 9 weeks (one-way ANOVA; n = 3 mice per group; error bars indicate mean ± SD). (C and D) Percent engraftment of hCD45 in peripheral blood (C) and Kaplan-Meier survival analysis (D) of NPM1c,FLT3ITD,FLT3TKD-transplanted PDX mice receiving control or 0.1% VTP-50469–spiked chow for 129 days (patient 1, table S5; n = 5 mice per group; error bars indicate mean ± SD). (E) Mutational screening of 49 paired MDS and sAML patient samples for RUNX1, TP53, NPM1, FLT3, ASXL1, DNMT3A, IDH1, and IDH2 mutations revealed six patients with persistent NPM1 mutations detected in MDS samples before AML development. IPSS, International Prognostic Scoring System; SNP, single-nucleotide polymorphism; CN-AML, cytogenetically normal AML; CNAs, copy number alterations. ns, not significant; **P < 0.01; ***P < 0.001.

We next investigated whether NPM1c mutant cells remained sensitive to Menin-MLL inhibition after progression to AML. Menin-MLL inhibitors have been shown to be effective targeting MLL-fusion leukemias in vivo, but whether they will be similarly effective in the more common NPM1c mutant AML was less clear. To this end, we used patient-derived xenograft (PDX) assays of untreated and relapsed NPM1c AML harboring FLT3, DNMT3a, and IDH1 co-mutations (table S5). Inhibiting MLL1-Menin dramatically reduced tumor burden in blood, spleen, and bone marrow of three different PDX models treated for 30 to 43 days (fig. S11, A to I). The few detectable human cells expressed high levels of the differentiation marker CD11b (fig. S11, J and K). VTP-50469 treatment significantly prolonged survival in two independent NPM1c PDX models (Fig. 4, C and D, and fig. S12, A and B). Gene expression analysis of NPM1c PDX cells isolated 10 days after in vivo Menin inhibitor treatment confirmed reduced expression of MEIS1 and PBX3, as observed in our mouse model, whereas HOX genes were slightly increased (fig. S12C). Furthermore, Menin inhibition was effective in PDX mice with high tumor burden [40 to 80% human CD45 (hCD45)] (fig. S13A). A reduction of blood leukemia burden was observed after 3 weeks of VTP-50469 treatment. Except for one mouse that expired after 10 days of treatment, the three remaining VTP-50469–treated mice survived more than 150 days after transplant with hCD45 engraftment of <1% (fig. S13, B and C). Our data suggest that Menin-MLL inhibition is highly effective not just in the preleukemic setting but also in fully developed aggressive human NPM1c mutant AMLs.

To determine the feasibility of detecting preleukemic NPM1 mutant clones in patients, we screened 49 paired myelodysplastic syndrome (MDS) and secondary AML (sAML) samples for AML-associated mutations (NPM1, DNMT3A, RUNX1, TP53, NF1, ASXL1, IDH1, and IDH2). NPM1c was detected in six (12%) of MDS and paired sAML samples, whereas co-occurring signaling mutations NF1 and FLT3 were mostly acquired during progression to sAML in these samples (Fig. 4E). Half of these NPM1c mutant MDS patients rapidly developed leukemia within 1 to 2 months, whereas the remaining three patients (or the other half of patients) progressed more slowly (5 to 6.5 months) (table S6). NPM1c can therefore be detected in a preleukemic setting and may act as a marker for progression to AML, making it an ideal target for preventative therapy. In the context of screening and monitoring, this may plausibly be extended to individuals with large DNMT3A or IDH1/2 mutant CHIP clones, which is predictive of high AML risk (9).

In summary, this study shows that eliminating preleukemic cells with targeted therapy is a potentially promising approach; specifically, we present evidence in a mouse model of AML that early intervention is possible with molecules that target chromatin regulators. Combined with improved long-term monitoring of patients with high-risk CHIP or MDS for appearance of an NPM1c preleukemic clone, disease prevention could become a realistic possibility in the future.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Tables S1 to S6

References (2532)

References and Notes

Acknowledgments: We thank Z. Feng and all members of the Armstrong Lab for their help; A. Cremer and J. Perry for critically reading the manuscript; F. Perner for the Menin sgRNA constructs; and Y. Soto-Feliciano for the ipUSEPR sgRNA expression plasmid. Funding: S.A.A. was supported by NIH grants CA176745, CA204639, CA066996, and CA206963 and by grants from Wicked Good Cause and Cookies for Kids’ Cancer. K.D. and L.B. were supported by SFB 1074 project B3. H.J.U. was supported by the German Research Foundation (DFG, UC77/1-1). R.L.L. was supported by NIH grants P30 CA008748 and U54 OD020355-04. G.S.V. is funded by a Cancer Research UK Senior Fellowship (C22324/A23015). Author contributions: H.J.U. and S.A.A. conceived the study and wrote the manuscript; H.J.U., S.M.K., E.M.W., H.G., A.V.K., and J.Y.G. conducted experiments; C.H. analyzed RNA-seq and ChIP-seq data; G.M.M. provided MLL1-Menin inhibitor VTP-50469; R.L.L. and G.S.V. provided the Dnmt3a and Npm1 mutant knock-in mice used in this study; and F.G.R., K.D., and L.B. provided primary MDS and sAML data. Competing interests: S.A.A. has been a consultant and/or shareholder for Vitae/Allergan Pharmaceuticals, Epizyme Inc., Imago Biosciences, Cyteir Therapeutics, C4 Therapeutics, Syros Pharmaceuticals, OxStem Oncology, Accent Therapeutics, and Mana Therapeutics. S.A.A. has received research support from Janssen, Novartis, and AstraZeneca. R.L.L. is on the supervisory board of Qiagen and is a scientific advisor to Loxo, Imago, C4 Therapeutics, and Isoplexis, which each include an equity interest. He receives research support from and consulted for Celgene and Roche, he has received research support from Prelude Therapeutics, and he has consulted for Incyte, Novartis, Astellas, Morphosys, and Janssen. He has received honoraria from Lilly and Amgen for invited lectures and from Gilead for grant reviews. G.S.V. is a consultant for Oxstem and a consultant for and minor stockholder in Kyma. H.G. owns stock in Theravance Biopharma. Data and materials availability: VTP-50469 can be obtained by means of a MTA from G.M.M. through Syndax. All data of this study are deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE129638.

Stay Connected to Science

Navigate This Article