A remote GATA2 hematopoietic enhancer drives leukemogenesis in inv(3)(q21;q26) by activating EVI1 expression

Abstract

Chromosomal inversion between 3q21 and 3q26 results in high-risk acute myeloid leukemia (AML). In this study, we identified a mechanism whereby a GATA2 distal hematopoietic enhancer (G2DHE or -77-kb enhancer) is brought into close proximity to the EVI1 gene in inv(3)(q21;q26) inversions, leading to leukemogenesis. We examined the contribution of G2DHE to leukemogenesis by creating a bacterial artificial chromosome (BAC) transgenic model that recapitulates the inv(3)(q21;q26) allele. Transgenic mice harboring a linked BAC developed leukemia accompanied by EVI1 overexpression-neoplasia that was not detected in mice bearing the same transgene but that was missing the GATA2 enhancer. These results establish the mechanistic basis underlying the pathogenesis of a severe form of leukemia through aberrant expression of the EVI1 proto-oncogene.

Figure 1. Identification of the Gata2 gene distal hematopoietic enhancer (G2DHE)

(A) Schematic structure of human 3q21 region. Arrowheads indicate documented breakpoints of translocation and inversion between 3q21 and 3q26. The breakpoints clustered in an approximately 25-kb region (yellow region). A blue circle indicates Gata2 distal hematopoietic enhancer (G2DHE). (B) Schematic structure of the G2BAC-GFP transgenes. The mouse Gata2 locus is depicted at the top. G2BAC-GFP transgene (G2BAC-GFP-flox or -delta) are shown beneath the endogenous Gata2 locus. White triangles represent loxP sequences. (C) Representative GFP histogram of bone marrow LSK, LK, TER-119⁺, B220⁺ and Mac-1⁺ cells from G2BAC-GFP-flox (red lines) and -delta (blue lines) mice. The gray shaded histograms represent GFP fluorescence in wild-type mice without a transgene (Tg (−)). (D) Representative GFP histogram in LT-HSC, ST-HSC, MPP, CMP, GMP and MEP cells from G2BAC-GFP-flox (red lines) and -delta (blue lines) mice. The gray shaded histograms represent GFP fluorescence in Tg (−) mice. (E) Percentages of GFP-positive cells in the indicated fractions of G2BAC-GFP-flox (red bars) and -delta (blue bars) mice (n=4). Data are represented as mean +/− SD. **, p<0.01 compared with G2BAC-GFP-flox mice. See also Figure S1.

Figure 2. Generation of 3q21q26 mice by linking two BACs

(A) Chromosomal inversion of the region between breakpoints (yellow arrowheads) in EVI1 (brown box) and C3ORF27 (orange box) genes in MOLM-1 cells. The boxed sequence shows the breakpoint between 3q21 (red) and 3q26 (black). (B) Strategy for linking two BAC clones. Human BACs RP11-94J18 (94J18) and RP11-662G11 (662G11) contain 3q26 and 3q21, respectively. We used site-specific Cre-mediated recombination to link the two BAC clones by simultaneous intermolecular homologous recombination between the loxP514 (black triangles) and loxP511 (white triangles) sites. Neo, Cm and Amp indicate the neomycin-, chloramphenicol- and ampicillin-resistant genes, respectively. (C) Pulse-field gel electrophoresis of BAC clones. We found two types of 662G11 clones; full-length (asterisk) and partially deleted (arrowhead) clones. (D) Schematic representations of the inverted allele, 3q21, 3q26, 3q21q26 and 3q21q26ΔG2DHE BAC transgenes. Red and black lines represent contributions from 3q21 and 3q26, respectively. (E) Copy numbers of integrated transgenes. Copy numbers were determined by q-PCR using Tg (−), 3q26 (lines A and B), 3q21q26 (lines A, B and C) and 3q21q26ΔG2DHE (lines A and B) mice (n=3-6). Red numbers depicted above the bar graphs represent copy numbers of transgenes. See also Figure S2, Table S1 and Supplemental Experimental Procedures.

Figure 3. Expression profiles of the human EVI1 transgene in 3q21q26 mice

(A-C) Expression levels of EVI1 gene in hematopoietic and non-hematopoietic tissues. Expression levels of human EVI1 mRNA (A), mouse endogenous Evi1 mRNA (B) and total (human + mouse) EVI1 mRNA (C) in hematopoietic tissues (spleen and thymus) and non-hematopoietic tissues (lung, kidney, liver and brain) of Tg (−) (n=10), 3q21q26 mouse lines (line A, n=3; line B, n=3; line C, n=3) and 3q21q26ΔG2DHE line B (n=4) were determined. The level of each mRNA was normalized to rRNA abundance. In panels B and C, average values for Tg (−) spleen cells were set to 1, while in panel A spleen cells of 3q21q26 line A was set to 1. (D, E) Abundance of EVI1 gene expression in hematopoietic cells. Expression levels of human EVI1 mRNA (D) and total EVI1 mRNA (E) in hematopoietic stem and progenitor cells and differentiated cells in Tg (−) (n=19), 3q21q26 (line A, n=3-6; line B, n=3-7; line C, n=4) and 3q21q26ΔG2DHE line B (n=4) bone marrows were determined. The abundance of each mRNA was normalized to the rRNA abundance. Average values for 3q21q26 line A LSK cells (D) and Tg (−) LSK cells (E) were set to 1. Bar graphs are represented as mean +/− SD. Arrows indicate undetectable expression levels. *, p<0.05; **, p<0.01 compared with Tg (−). ^#, p<0.05; ^##, p<0.01.

Figure 4. 3q21q26 mice develop G2DHE-dependent leukemia

(A) Kaplan-Meier survival curve of Tg (−) (n=96), 3q26 (n=55), 3q21q26 (n=121) and 3q21q26ΔG2DHE (n=26) mice. Survival rates are calculated by compiling data from multiple Tg mouse lines. The inserted graph shows the overall survival of each line of 3q21q26 mice (lines A, B and C). (B, C) White blood cell counts (B) and hematocrits (C) in the peripheral blood of Tg (−) (n=11-46) and 3q21q26 line A (n=6-19), line B (n=5-15) and line C (n=3-9) mice. (D) Representative smears of peripheral blood taken from Tg (−) and leukemic 3q21q26 mice. Scale bar, 20 μm. (E) The numbers of white blood cells in leukemic 3q21q26 line A (n=6), line B (n=7) and line C (n=3) mice and Tg (−) (n=5) littermates. (F) Representative spleens from Tg (−) and leukemic 3q21q26 mice. Scale bar, 1 cm. (G) Average spleen weights from Tg (−) (n=5) and leukemic 3q21q26 line A (n=6), line B (n=7) and line C (n=3) mice. (H) Hematoxylin-eosin staining of tissues of Tg (−) (top panels) and leukemic 3q21q26 line B (bottom panels) mice. Scale bar, 100 μm. Bar graphs are represented as mean +/− SD. *, p<0.05; **, p<0.01 compared to Tg (−).

Figure 5. 3q21q26 mice suffer from three types of leukemia

(A) Expression profiles of CD34, c-kit, B220, CD3a, Gr-1, Mac-1 and TER-119 in peripheral blood (PB) leukemic cells of 3q21q26 line A (n=5), line B (n=5-6) and line C (n=3) mice as well as mononuclear cells of Tg (−) mice (n=4-5). *, p<0.05; **, p<0.01. (B) Flow cytometric analyses of leukemic cells in 3q21q26 mice. Representative flow cytometric patterns of B-type (left panel), B+GM-type (middle panel) and GM-type (right panel) leukemias are shown. (C) Wright-Giemsa staining of leukemic cells in the peripheral blood of 3q21q26 mice. Scale bar, 10 μm. As a control, mononuclear cells from Tg (−) mice are shown. (D) Frequencies of B-type (white), B+GM-type (gray) and GM-type (black) leukemia developed in the three 3q21q26 mouse lines. See also Figure S3.

Figure 6. Leukemic cells of 3q21q26 mice expand autonomously

(A) Kaplan-Meier survival curve of recipient nude mice (n=12) transplanted with 10⁷ mononuclear bone marrow cells from leukemic 3q21q26 mice. (B) Representative spleens from control or recipient nude mice 40 days after transplant. Scale bar, 1 cm. (C-E) Numbers of white blood cells (C), red blood cells (D) and platelets (E) in peripheral blood in control (n=3) and transplanted (n=11) nude mice. **, p<0.01 compared with control nude mice. (F) Representative flow cytometric analyses of peripheral blood cells from donor mice suffering from B-, B+GM-, and GM-type leukemia and comparison to their respective recipient nude mice. (G) Leukemia types of donor and recipient mice. Numbers in the table represent those of mice suffering from each type of leukemia. (H) Wright-Giemsa staining of leukemic cells in the peripheral blood of transplant recipient nude mice suffering from B-, B+GM-, and GM-type leukemia. Scale bar, 10 μm. Bar graphs are represented as mean +/− SD.

Figure 7. G2DHE drives EVI1 overexpression in HSPC

(A-F) Representative flow cytometric profiles of Lin⁻ cells (A), LSK cells (C) and LK cells (E) in 12-week-old Tg (−), 3q26 line B, 3q21q26 line B and 3q21q26ΔG2DHE line B mice. Cells in red boxes were analyzed in panels B, D and F. Absolute cell numbers of LSK and LK cells (B), LT-HSC, ST-HSC, MPP and CD34^highFlt3⁻ (CD34^high) cells (D), and CMP, MEP and GMP cells (F) in the Lin⁻ cell population in 12-week-old Tg (−) (white, n=4), 3q26 line B (blue, n=4), 3q21q26 line B (pink, n=4) and 3q21q26ΔG2DHE line B (green, n=4) mice are depicted. (G-I) Expression levels of human EVI1 mRNA (G), mouse endogenous Evi1 mRNA (H), and total EVI1 mRNA (I) in LT-HSC, ST-HSC, MPP, CD34^high, CMP, MEP and GMP fractions in 12-week-old Tg (−) (white, n=4-8), 3q26 line B (blue, n=4-8), 3q21q26 line B (pink, n=4-8) and 3q21q26ΔG2DHE line B (green, n=4-8) mice. The expression level of each mRNA was normalized to rRNA abundance. Average values for 3q26 LT-HSC cells (G) and Tg (−) LT-HSC cells (H, I) were set to 1. Bar graphs are represented as mean +/− SD. Arrows indicate undetectable or slight expression levels. N.D.; not determined. *, p<0.05; **, p<0.01 compared with Tg (−). ^#, p<0.05; ^##, p<0.01. See also Figure S4.

Figure 8. A model for the involvement of G2DHE in normal and 3q21q26 inverted/translocated hematopoietic cells

In normal hematopoietic progenitors, G2DHE enhances GATA2 transcription (top panel). In AML, G2DHE becomes localized downstream or upstream of the EVI1 gene in inv(3)(q21;q26) (middle panel) or t(3;3)(q21;q26) (bottom panel) alleles, respectively. In both cases, the chromosomal rearrangement-dependent juxtaposition of G2DHE induces EVI1 gene transcription instead of GATA2 transcription which subsequently leads to leukemia.