Interconnecting molecular pathways in the pathogenesis and drug sensitivity of T-cell acute lymphoblastic leukemia

Abstract

To identify dysregulated pathways in distinct phases of NOTCH1-mediated T-cell leukemogenesis, as well as small-molecule inhibitors that could synergize with or substitute for gamma-secretase inhibitors (GSIs) in T-cell acute lymphoblastic leukemia (T-ALL) therapy, we compared gene expression profiles in a Notch1-induced mouse model of T-ALL with those in human T-ALL. The overall patterns of NOTCH1-mediated gene expression in human and mouse T-ALLs were remarkably similar, as defined early in transformation in the mouse by the regulation of MYC and its target genes and activation of nuclear factor-kappaB and PI3K/AKT pathways. Later events in murine Notch1-mediated leukemogenesis included down-regulation of genes encoding tumor suppressors and negative cell cycle regulators. Gene set enrichment analysis and connectivity map algorithm predicted that small-molecule inhibitors, including heat-shock protein 90, histone deacetylase, PI3K/AKT, and proteasome inhibitors, could reverse the gene expression changes induced by NOTCH1. When tested in vitro, histone deacetylase, PI3K and proteasome inhibitors synergized with GSI in suppressing T-ALL cell growth in GSI-sensitive cells. Interestingly, alvespimycin, a potent inhibitor of the heat-shock protein 90 molecular chaperone, markedly inhibited the growth of both GSI-sensitive and -resistant T-ALL cells, suggesting that its loss disrupts signal transduction pathways crucial for the growth and survival of T-ALL cells.

Figure 1

Differentially expressed genes in a Notch1-induced mouse model of T-ALL. (A-B) SAM was applied in triplicate in 2 comparisons: polyclonal DP cells versus normal control (A) and leukemic DP cells versus polyclonal DP cells (B), with 3 different mice per group. Significant genes were selected by the following criteria: fold change ≥ 1.5 and δ value of 0.37 (polyclonal DP cells vs normal control) and 0.55 (leukemic DP cells vs polyclonal DP cells) to adjust the false discovery rate to ∼ 10%. Red or green dots represent genes that were significantly up-regulated or down-regulated in each comparison, respectively. (C) Heat-map images for the differentially expressed genes. The selected genes were classified into groups 1 to 8, based on significant levels and fold change between each of cells. Normal control (N), polyclonal DP cells (P), and leukemic DP cells (L).

Figure 2

Up-regulation of the Notch1 pathway at early phase of leukemogenesis and up-regulation of the Myc pathway at early and late phases of leukemogenesis. (A) GSEA histograms for the gene set “NOTCH signaling pathway.” The enrichment score (ES; y-axis) reflects the degree to which a gene set is overrepresented in normal control (N), polyclonal DP cells (P), and leukemic DP cells (L) at the extreme left or right of the entire ranked list. Each solid bar represents 1 gene within a gene set. Heat-map image illustrates gene expression levels of the leading edge subset. The normalized enrichment score (NES) and the nominal P value are indicated. (B) Expression of representative genes (Dtx, Hes1, Ptcra, and Snw1) in the Notch signaling pathway. Significance was evaluated by the Student t test: **P < .01. Data are mean ± SD values of triplicate experiments. (C) Expression of Myc and its targets (Apex1, Nme2, and Cdk4) in each cell. Significance was evaluated by the Student t test: *P < .05, **P < .01. Data are mean ± SD values of triplicate experiments. The NES and the nominal P value are indicated. (D) GSEA histograms for the gene set representing “MYC oncogenic signature.” See panel A description for details of the GSEA histogram.

Figure 3

Down-regulation of tumor suppressors and cell cycle regulators at the late phase of leukemogenesis. (A) GSEA histograms for the gene sets “p53 signaling pathway” and “tumor suppressor.” See Figure 2A description for details on the GSEA histogram. The NES and the nominal P value are indicated. Normal control (N), polyclonal DP cells (P), and leukemic DP cells (L). (B) Expression of Pten, Apc, Fbxw7, and Ep300 at the normal (control), polyclonal DP cells, and leukemic DP cells. Significance was evaluated by the Student t test: *P < .05, **P < .01. Data are mean ± SD values of triplicate experiments. (C) Expression levels of the cell cycle regulators Rb1, Cdkn2c, Cdkn2d, and Cdkn3. Significance was evaluated by the Student t test: *P < .05, **P < .01. Data are mean ± SD values of triplicate experiments. (D) GSEA histogram for the gene set “cell cycle arrest.”

Figure 4

Genes and pathways deregulated in human T-ALL. (A) Heat-map image of differentially expressed genes in human T-ALL. Microarray gene expression profiling was performed on 4 human T-ALL cell lines and 34 primary T-ALL samples. Genes attaining statistical significance (P < .05, fold change ≥ 1.2) were selected and classified into 8 groups based on criteria used to classify genes in the mouse model (Table 1). The selected genes were also analyzed for significant differences in NOTCH1 mutant (MUT) versus wild-type (WT) primary T-ALL samples (20 MUTs vs 14 WTs). Genes that showed significant differences in expression levels in this comparison are indicated: *P < .05, **P < .01. (B) GSEA histograms for human T-ALL cells (cell lines, top; and primary clinical samples, bottom) representing “MYC oncogenic signature.” See the Figure 2A description for details of GSEA. The NES and the nominal P value are indicated. GSI-treated T-ALL cell lines (G), DMSO-treated T-ALL cell lines (D), wild-type primary T-ALL samples (W), and NOTCH1-mutant primary samples (M).

Figure 5

Up-regulation of proteasome pathway and the growth inhibitory effect of proteasome inhibitors on T-ALL cells. (A) GSEA histograms for human T-ALL cells (cell lines or primary clinical samples) representing the “proteasome pathway.” The NES and the nominal P value are indicated. GSI-treated T-ALL cell lines (G), DMSO-treated T-ALL cell lines (D), wild-type primary T-ALL samples (W), and NOTCH1-mutant primary samples (M). (B) IC₅₀ values with the proteasome inhibitors MG-132 and bortezomib. Mouse leukemic DP cells (mouse T-ALL) were obtained from the mouse after 8 weeks after transplantation of Lin⁻ hematopoietic cells that had been transduced with the ICN1 gene. Mouse T-ALL cells and various types of human leukemia/cancer cell lines were treated for 3 days with a proteasome inhibitor MG-132 (1 mouse T-ALL, 4 GSI-sensitive T-ALL, 4 GSI-resistant T-ALL, 3 acute myeloid leukemia [AML], 4 B-cell non-Hodgkin lymphoma [B-NHL]/multiple myeloma [MM]/chronic lymphocytic leukemia [CLL], 4 neuroblastoma [NBL], and 4 other cancer cell lines) or bortezomib (4 GSI-sensitive T-ALL and 4 GSI-resistant T-ALL). The number of viable cells was measured by the MTT assay, and the IC₅₀ values were calculated for each cell. (C) Synergistic effect of MG-132 in combination with GSI. Three GSI-sensitive and 3 GSI-resistant T-ALL cell lines were treated with or without MG-132 (0.3μM) in the presence or absence of GSI MRK-003 (1μM) for 6 days. The number of viable cells was measured by the MTT assay.

Figure 6

Growth inhibitory effect of the HDAC inhibitor vorinostat and the PI3K inhibitor LY-294002 on T-ALL cell lines. (A) Negative correlation of the T-ALL gene expression signature with vorinostat and LY-294002. CMAP analysis was performed on a T-ALL gene expression signature with 6100 instances of compounds. Green and red represent positive and negative correlation, respectively. Each dataset is indicated by a black bar. (B) IC₅₀ values with the HDAC inhibitor vorinostat and the PI3K inhibitor LY-294002. Mouse leukemic DP cells (mouse T-ALL) and various types of human leukemia/cancer cell lines were treated for 3 days with vorinostat or LY-294002 (1 mouse T-ALL, 4 GSI-sensitive T-ALL, 4 GSI-resistant T-ALL, 3 AML, 4 B-NHL/MM/CLL, 4 NBL, and 4 other cancer cell lines). The number of viable cells was measured by the MTT assay, and IC₅₀ values were calculated for each cell. (C-D) Synergistic effects of LY-294002 and vorinostat in combination with GSI. Three GSI-sensitive and 3 GSI-resistant T-ALL cell lines were treated with or without LY-294002 (10μM; C) or vorinostat (0.3μM; D) in the presence or absence of GSI MRK-003 (1μM) for 6 days. The number of viable cells was measured by MTT assay.

Figure 7

Growth inhibitory effect of the HSP90 inhibitor alvespimycin on T-ALL cell lines. (A) Negative correlation of the T-ALL gene expression signature with alvespimycin. CMAP analysis was performed on a T-ALL gene expression signature with 6100 instances of compounds. Green and red represent positive and negative correlation, respectively. Each dataset is indicated by a black bar. (B) IC₅₀ values with the HSP90 inhibitor alvespimycin. Mouse leukemic DP cells (mouse T-ALL) and various types of human leukemia/cancer cell lines were treated for 3 days with vorinostat or LY-294002 (1 mouse T-ALL, 4 GSI-sensitive T-ALL, 4 GSI-resistant T-ALL, 3 AML, 4 B-NHL/MM/CLL, 4 NBL, and 4 other cancer cell lines). The number of viable cells was measured by the MTT assay, and IC₅₀ values were calculated for each cell. (C) Combination treatment of alvespimycin with GSI. Three GSI-sensitive and 3 GSI-resistant T-ALL cell lines were treated with or without alvespimycin (0.1μM) in the presence or absence of GSI MRK-003 (1μM) for 6 days. The number of viable cells was measured by MTT assay.