Clinical Responsiveness to All-trans Retinoic Acid Is Potentiated by LSD1 Inhibition and Associated with a Quiescent Transcriptome in Myeloid Malignancies

Abstract

Purpose: In preclinical studies, the lysine-specific histone demethylase 1A (LSD1) inhibitor tranylcypromine (TCP) combined with all-trans retinoic acid (ATRA) induces differentiation and impairs survival of myeloid blasts in non-acute promyelocytic leukemia acute myeloid leukemia (AML). We conducted a phase I clinical trial (NCT02273102) to evaluate the safety and activity of ATRA plus TCP in patients with relapsed/refractory AML and myelodysplasia (MDS).

Patients and methods: Seventeen patients were treated with ATRA and TCP (three dose levels: 10 mg twice daily, 20 mg twice daily, and 30 mg twice daily).

Results: ATRA-TCP had an acceptable safety profile. The MTD of TCP was 20 mg twice daily. Best responses included one morphologic leukemia-free state, one marrow complete remission with hematologic improvement, two stable disease with hematologic improvement, and two stable disease. By intention to treat, the overall response rate was 23.5% and clinical benefit rate was 35.3%. Gene expression profiling of patient blasts showed that responding patients had a more quiescent CD34⁺ cell phenotype at baseline, including decreased MYC and RARA expression, compared with nonresponders that exhibited a more proliferative CD34⁺ phenotype, with gene expression enrichment for cell growth signaling. Upon ATRA-TCP treatment, we observed significant induction of retinoic acid-target genes in responders but not nonresponders. We corroborated this in AML cell lines, showing that ATRA-TCP synergistically increased differentiation capacity and cell death by regulating the expression of key gene sets that segregate patients by their clinical response.

Conclusions: These data indicate that LSD1 inhibition sensitizes AML cells to ATRA and may restore ATRA responsiveness in subsets of patients with MDS and AML.

Figure 1.. Clinical course of the patients and disease type distributions.

A) Swimmer’s plot where horizontal bars represent the duration on study for a cohort of 18 patients (17 treated and 1 screen-failure [Patient #3]). Bars are color-coded by the patient’s clinical working group response (Screen failure and N/A, non-evaluable response; PD, progressive disease; SD, stable disease; SD HI, stable disease with hematologic improvement; Marrow CR, marrow complete remission; Marrow CR with HI, marrow complete remission with hematologic improvement; MLFS, morphologic leukemia-free state). The reason for study discontinuation is listed at the end of each bar. Among the response evaluable patients, patients are grouped as responders (response/SD) by the pink vertical bar, and non-responders (lack or response/PD) by the grey vertical bar. An additional vertical bar color-codes the diagnosed disease type (MDS in green, AML in orange). *Patient #12 had neutrophil recovery and blast reduction but did not qualify for a clinical response. This patient’s RNA-seq signature clustered with the other responders.

Figure 2.. The transcriptomes of CD34⁺ cells from AML and MDS patients at baseline correlate with response to ATRA-TCP treatment.

A) Heat map of normalized gene expression from RNA-seq of 8 patient samples before treatment (Z-score of the RPKM, per row). An unsupervised hierarchical clustering was applied to rows (all expressed genes with sum(expression/row>=50 RPKM)) and columns (patient samples). B) Boxplot distributions of the mean gene expression (RPKM) for cluster 4 showing consistent lower gene expression in responders vs. non-responders before treatment (Kolmogorov–Smirnov test, p<0.0001). C) Volcano plot showing the differential expressed genes between responders (n=3) and non-responders (n=5) before treatment (2-fold change cut-off, and FDR<0.05, blue=down-regulated genes, red=up-regulated genes). D) Heat map of normalized gene expression of the differential expressed genes from panel C, sorted by fold change. E) Bar plot of the normalized enrichment score (NES) of the Top20 negatively-enriched “Hallmark genes sets” by GSEA. F) GSEA plots of the top-ranked negatively-enriched gene sets from panel E. The enrichment score (ES) is indicated by a green line. Below, each vertical line of the barcode represents a gene. Genes are ranked by their differential expression between responders and non-responders at baseline (from left=up-regulated to right=down-regulated in responders compared to non-responders).

Figure 3.. The LSD1 inhibitor TCP sensitizes SKNO-1 cells to ATRA treatment.

A) LSD1 inhibitor TCP (10 μM) and ATRA dose response plot. AML cell lines were treated with either ATRA (1 μM), TCP (10 μM), or ATRA-TCP (1 μM-10 μM) for 4 days and measured for B) cell viability and C) cell-cycle percentages in U-937, and D-F) SKNO-1 cells. Data were analyzed by one-way analysis of variance (ANOVA), followed by Tukey’s post-hoc test (ATRA vs. ATRA-TCP). Data represent the mean and the error bars represent the standard deviation of the mean for three independent experiments. G) Synergy maps for U-937 (top) and SKNO-1 cells (bottom). The 2D synergy matrix was generated with SynergyFinder 2.0 using ZIP model. H) Volcano plots showing differential expressed genes in U-937 (top) and SKNO-1 (bottom) upon treatment compared to the untreated control cells (treatments from left to right: ATRA, TCP or ATRA-TCP) (2 fold-change cutoff, and FDR<0.05, blue=down-regulated genes, red=up-regulated genes). I) Venn diagrams of the significantly up-regulated genes in U-937 and SKNO-1 from panel H (treatments from left to right: ATRA, TCP, or ATRA-TCP). J) Bar plot of the 14 significantly up-regulated pathways upon ATRA-TCP treatment in U-937 that are also up-regulated in SKNO-1 upon ATRA-TCP. K) Top-ranked genes from the pathways in panel I, sorted by the cumulative count in both cell lines. L) Heat map of the Z-score normalized gene expression in SKNO-1 for a subset of genes obtained by unsupervised clustering of the up-regulated genes upon any of the treatments in the ATRA-resistant cell line (SKNO-1). M) Heat map of the Z-score normalized gene expression in SKNO-1 for a subset of genes from panel K with positive-enrichment in the ATRA-induction signature reported for HL-60 cells (15) showing ATRA-TCP enhances the expression of known ATRA-responsive genes. Heat maps are color-coded from blue (lower expression compared to the row’s average), to white (no change), to red (higher expression compared to the row’s average). N-O) Normalized mRNA expression of the differentiation markers ITGAM (CD11b) and CD86, respectively, for untreated control cells, ATRA, TCP, or ATRA-TCP treatment in U-937 (left), and in SKNO-1 (right). Same-color bars represent two independent experiments.

Figure 4.. ATRA-TCP induces the expression of a RA-responsive gene signature in patient responders.

A) Signature of RA-responsive genes: genes with RARA occupancy and RARE motif from (38) that are up-regulated upon any of the treatments in the AML cell lines (U-937 and/or SKNO-1), that further overlapped with the gene signature of response from patients at baseline (N=3212 genes). Cartoons were downloaded from https://smart.servier.com. B) Heat map of mean normalized gene expression in AML cell lines for control and treatments (Z-score of the RPKM, per row). Unsupervised hierarchical clustering was applied to rows and columns. C) Boxplot distributions of the mean gene expression (RPKM) showing significant gene induction of RA-genes after ATRA-TCP treatment (Kolmogorov–Smirnov test). D) Heat map of mean normalized gene expression in patients before treatment (Z-score of the RPKM, per row). Unsupervised hierarchical clustering was applied to rows and columns. E-F) Boxplot distributions of the mean gene expression (RPKM) showing significant gene induction of RA-genes after ATRA-TCP treatment in responders but not in non-responders (Kolmogorov–Smirnov test).See Figure S4 for related heat map. G-J) Examples of RA-responsive genes: Gene expression plots for patients (before/after treatment), and for U-937 cells (from left to right: untreated control cells, ATRA, TCP, or ATRA-TCP treatment. Same-color bars represent two independent experiments).