PPM1D Mutations Drive Clonal Hematopoiesis in Response to Cytotoxic Chemotherapy

Abstract

Clonal hematopoiesis (CH), in which stem cell clones dominate blood production, becomes increasingly common with age and can presage malignancy development. The conditions that promote ascendancy of particular clones are unclear. We found that mutations in PPM1D (protein phosphatase Mn²⁺/Mg²⁺-dependent 1D), a DNA damage response regulator that is frequently mutated in CH, were present in one-fifth of patients with therapy-related acute myeloid leukemia or myelodysplastic syndrome and strongly correlated with cisplatin exposure. Cell lines with hyperactive PPM1D mutations expand to outcompete normal cells after exposure to cytotoxic DNA damaging agents including cisplatin, and this effect was predominantly mediated by increased resistance to apoptosis. Moreover, heterozygous mutant Ppm1d hematopoietic cells outcompeted their wild-type counterparts in vivo after exposure to cisplatin and doxorubicin, but not during recovery from bone marrow transplantation. These findings establish the clinical relevance of PPM1D mutations in CH and the importance of studying mutation-treatment interactions. VIDEO ABSTRACT.

Keywords: CHIP; DNA damage response; PPM1D; cisplatin; clonal hematopoiesis; doxorubicin; etoposide; t-AML; t-MDS; topoisomerase inhibitors.

Graphical abstract

Figure 1

Mutational Landscape of Myeloid Neoplasm (MN)-Associated Genes in the t-AML/t-MDS Cohort (A) The twenty most frequently mutated genes detected by targeted gene sequencing in the t-AML/t-MDS study cohort (n = 156) are shown. The red bars represent the mutation frequency in the t-MN (t-AML/t-MDS) cohort and the blue bars represent the mutation frequency in a matched de novo MN (AML/MDS) control cohort (n = 228). (B) Volcano plot of genes enriched in t-AML/t-MDS compared to de novo AML/MDS. The horizontal dotted line corresponds to a p value of 0.05. (C) Pairwise association plot of overall mutation co-occurrence or mutual exclusivity, adjusted for multiple comparisons. Blue represents a negative association (mutual exclusivity) while red represents a positive association (co-occurrence). The magnitude of association is represented by both the size of the square and color gradient, which corresponds to a range of log odds ratio values. The statistical significance of associations is represented by the false discovery rate (FDR). The asterisks indicate the level of significance (FDR 0.1, 0.5, and 0.01). “PPM1D clonal” refers to the subset of PPM1D mutated cases with VAF > 0.2. (D) Seven cases where PPM1D was the only detected somatic mutation out of the 295 sequenced genes. See also Figure S1 and Tables S1 and S2.

Figure 2

Features of PPM1D Mutated t-AML/t-MDS Cases (A) Lollipop plot showing the distribution of PPM1D truncating mutations across the final exon of the gene. Vertical dotted lines demarcate the coding exons of the gene, with the corresponding amino acids shown below. The phosphatase domain of the protein is denoted by the green segment. Frameshift mutations are depicted in red and nonsense mutations in orange. The number of patients with each mutation is indicated in the lollipops (circles without numbers represent one case). (B) Variant allele frequency distribution plot in PPM1D mutated cases (n = 31; range, 0.02–0.47; mean, 0.1). (C) In three cases, next-generation sequencing was performed to determine the variant allele frequency of PPM1D in lymphoid (CD3⁺/CD19⁺) and non-lymphoid (CD3⁻/CD19⁻) peripheral blood fractions. (D) Genome-wide copy number plots of two separate cases with PPM1D copy number gain. Chromosome position is shown on the x axis and copy number log2 ratio is shown on the y axis. (E) Copy number alterations of PPM1D in 162 de novo AML cases from (TCGA, 2013), plotted with the corresponding PPM1D mRNA expression level. (F) Forest plot showing the association of PPM1D mutations with prior exposure to specific genotoxic agents, per clinical chart review. Log odds ratio is depicted with the 95% confidence interval. Agents with favorable associations with PPM1D mutations trend to the right of the dotted line. These include cisplatin (p = 0.004 and FDR = 0.056) and etoposide (p = 0.02 and FDR = 0.148). The p value and total number of patients exposed to each agent are noted to the right. See also Table S3.

Figure 3

PPM1D Mutants Resist Cisplatin-Induced Apoptosis (A) Immunoblot of PPM1D WT and mutant HEK293 cells in the presence and absence of DNA damage. Cells were treated with 30 μM cisplatin, harvested at 4, 8, and 24 hr, and probed with the indicated antibodies. A composite of images is shown (see STAR Methods for details). (B) Top: dose-response curves for cell viability with cisplatin and a specific PPM1D inhibitor (GSK2830371) in WT and PPM1D mutant MOLM13 lines. Mean ± SD (n = 3) is shown along with a non-linear regression curve. All values are normalized to the baseline cell viability with vehicle, as measured by the WST-8 assay. The IC₅₀ of cisplatin was 1.2 μM and 2.8 μM for the PPM1D WT and mutant lines, respectively (p < 0.001). Bottom: cell viability measured with WST-8 under combination treatment with cisplatin (1 μM) and GSK2730371 (a PPM1D inhibitor; 250 nM). (C) Schematic of experimental strategy shown in (D). GFP-negative PPM1D mutant cells were mixed with GFP-positive control cells at a starting ratio of 20:80 and subjected to treatment with vehicle (water) or cisplatin (+/− 18 nM GSK2830371). Population dynamics were assessed by flow cytometry every 4 days for 15 days. (D) Each bar depicts the proportion of PPM1D WT cells (in gray) and mutant cells (in red) in culture, measured at the indicated time points. Data represent mean ± SD of triplicates (n = 3). At least three independent experiments were conducted for each experiment shown above, with similar findings. The corresponding flow cytometry plots are depicted in Figure S3A. (E) PPM1D mutations confer resistance to cisplatin-induced apoptosis. PPM1D WT and mutant cells were treated with 1 μM cisplatin (+/− 24nM GSK2830371) for 72 hr, incubated with annexin V-APC and 7-AAD, and analyzed using flow cytometry. The percentage of annexin V positive (late and early apoptotic) cells is represented in the histogram (mean +/− SD shown). The experiment was performed in triplicate (n = 3). (F) PPM1D WT and mutant cells were treated with 750 nM cisplatin (or vehicle) for 24 hr and fixed for BrdU cell-cycle analysis. Anti-BrdU FITC antibody and propidium iodide (PI) were used to distinguish cells with active synthesis and DNA content, respectively. Mean values and SD are shown (n = 6). Three independent experiments were performed with similar findings. See also Figures S2, S3, and S4.

Figure 4

Screen for Chemoresistance in PPM1D Mutants Various classes of chemotherapy agents were screened in the isogenic CRISPR-generated WT and PPM1D mutant cell lines (MOLM13, OCI-AML2, and OCI-AML3). The chemotherapy agents selected include key agents utilized in the treatment of primary tumors that t-AML/t-MDS patients in our cohort were previously exposed to. Dose response curves are shown, with red representing PPM1D mutants and gray representing WT. The data points represent mean ± SD of triplicates (n = 3).

Figure 5

PPM1D Mutants Demonstrate a Selective Advantage In Vitro with Certain Classes of Chemotherapeutics In vitro competition was performed with GFP-negative PPM1D mutant cells to GFP-positive control cells at a starting ratio of 15:85. The competing cells were treated with doxorubicin (20 nM), etoposide (250 nM), or vincristine (1 nM) every 4 days, and flow cytometry was performed every 2 days over 14 days to assess the change in percentage of PPM1D mutants. Representative flow plots are depicted at three time points: day 2, day 8, and day 14. The red gate denotes the GFP-negative PPM1D mutant population, and the gray gate denotes the GFP-positive control population. Mean ± SD are shown in the summary graphs (n = 3).

Figure 6

Cisplatin Treatment Confers a Survival Advantage on Ppm1d-Mutant Hematopoietic Cells In Vivo (A) Generation of the R451X knockin mouse model utilizing CRISPR-Cas9 and homology directed repair. The founder mouse was crossed with WT mice for the F1 generation, and a heterozygous line was subsequently maintained. (B) Representative flow plots showing baseline hematopoietic characterization of the R451X mutant mouse (Ppm1d 1^m/+) compared to WT (Ppm1d^+/+). We assessed lineage composition of the peripheral blood and examined the frequency of different progenitor compartments in the bone marrow by flow cytometry (n = 10 mice/group for peripheral blood, n = 6 mice/group for bone marrow). Long-term hematopoietic stem cells (LT-HSCs) were identified by c-kit⁺ lineage⁻ Sca-1⁺ (KLS), with either CD150⁺ CD48⁻ (“SLAM”) gating (n = 6 mice) or CD34⁻ Flk2⁻ gating (n = 3 mice). Mean +/− SD is shown. (C) Competitive whole bone marrow transplant scheme: 20% of either R451X (Ppm1d^m/+) or control (Ppm1d^+/+) CD45.1/45.2 bone marrow cells were mixed with 80% WT CD45.1 bone marrow cells. A total of 3 × 10⁶ whole bone marrow cells were transplanted into lethally irradiated 8-week-old recipient mice (n = 8 per group). Engraftment was assessed 4 weeks following transplant, and the recipient mice were treated with weekly doses of 4 mg/kg cisplatin (intraperitoneally [i.p.]) for 5 consecutive weeks. (D) Average peripheral blood chimerism 4 weeks following transplant was 22.8% (range 18.3%–27%) in the control cohort (Ppm1d^+/+) and 13.9% (range 8.2%–20.4%) in the mutant cohort (Ppm1d^m/+). Chimerism was monitored weekly by flow cytometry. The graph depicts the average of the fold change in chimerism for each mouse relative to the initial chimerism at 4 weeks post-transplant (n = 8 for each group, mean ± SD shown). ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, and ^∗∗∗∗p < 0.0001. (E) Bone marrow was harvested from the Ppm1d^m/+ competitively transplanted mice to assess chimerism on multiple levels of the hematopoietic hierarchy, including LT-HSC, KLS, and whole bone marrow (WBM) cells (cisplatin-treated in red and non-treated in gray, n = 5 mice per group). Data are represented by box-and-whisker plots, with the quartiles, minimum, and maximum values shown. See also Figures S5, S6, and S7.

Figure 7

Ppm1d Mutant Cells Lose Their Survival Advantage in the Context of Bone Marrow Transplantation (A) Competitive whole bone marrow transplant was performed, with 20% of either R451X (Ppm1d^m/+) or control (Ppm1d^+/+) bone marrow cells mixed with 80% WT bone marrow cells. The recipient mice (n = 8 per group) were not treated with chemotherapy following transplant. Chimerism was monitored weekly by flow cytometry. Normalized values are shown using the initial chimerism at 4 weeks as the baseline for calculation of fold change (mean ± SD shown). ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, and ^∗∗∗∗p < 0.0001. (B) Schematic of the serial bone marrow transplantation performed 13 weeks after the initial competitive transplant shown in (A). 3 × 10⁶ whole bone marrow (WBM) cells from the primary recipients of the transplant were serially transplanted into lethally irradiated secondary recipient mice (n = 8 per group). Peripheral blood chimerism was assessed in the secondary recipients 5 and 14 weeks following serial transplantation. (C) Graphs depicting the chimerism of the donor bone marrow (BM) at the time of serial transplantation, the chimerism in the peripheral blood (PB) of the secondary recipients at 5 and 14 weeks after transplantation (n = 8 per group), and the chimerism in LT-HSCs (n = 4 per group) at 18 weeks. The summary bar graph below shows chimerism at each time-point, normalized to the initial chimerism of transplanted bone marrow (red, Ppm1d mutant; gray, WT control). Mean ± SD is shown. (D) Model of clonal hematopoiesis, emphasizing that different genes have different fitness effects in different contexts. Intrinsic factors such as self-renewal have been shown to drive clonal expansion of DNMT3A and TET2 mutants, whereas aberrant differentiation and proliferation drive expansion of JAK2 mutations. With PPM1D, extrinsic stressors such as cisplatin promote expansion of the mutants. See also Figure S7.