Mutational landscape and gene-expression patterns in adult acute myeloid leukemias with monosomy 7 as a sole abnormality

Abstract

Monosomy of chromosome 7 is the most frequent autosomal monosomy in acute myeloid leukemia (AML), where it associates with poor clinical outcomes. However, molecular features associated with this sole monosomy subtype (-7 AML) which may give insights into the basis for its poor prognosis have not been characterized. In this study, we analyzed 36 cases of -7 AML for mutations in 81 leukemia/cancer-associated genes using a customized targeted next-generation sequencing panel (Miseq). Global gene and microRNA expression profiles were also determined using paired RNA and small RNA sequencing data. Notably, gene mutations were detected in all the major AML-associated functional groups, which include activated signaling, chromatin remodeling, cohesin complex, methylation, NPM1, spliceosome, transcription factors and tumor suppressors. Gene mutations in the activated signaling and chromatin remodeling groups were relatively more frequent in patients <60 years of age, who also had more mutations in the methylation and spliceosome groups compared to patients {greater than or equal to} 60 years of age. Novel recurrent mutational events in AML were identified in the SMARCA2 gene. In patients {greater than or equal to} 60 years of age, the presence of spliceosome mutations associated with a lower complete remission rate (p=0.03). RNA sequencing revealed distinct gene and microRNA expression patterns between the sole -7 and non-7 AML cases, with reduced expression as expected of many genes and microRNAs mapped to chromosome 7, and overexpression of ID1, MECOM, and PTPRM, among others. Overall, our findings illuminate a number of molecular features of the underlying aggressive pathobiology in -7 AML patients.

Introduction

The loss of chromosome 7 (monosomy 7, -7) is the most frequent autosomal monosomy in acute myeloid leukemia (AML), and the most common chromosomal abnormality observed in high-risk AML patients (1). However, it has been rarely studied as a sole abnormality. The adverse prognostic impact of -7, both in combination with other chromosomal aberrations and as a sole abnormality, has been well known for decades. Patients with -7 are characterized by low initial complete remission (CR) rates, short remissions, and inferior survival compared with those of patients in most other cytogenetic groups (1–5). The elucidation of the underlying biology and identification of candidate genes responsible for the adverse clinical outcome of -7 AML have been the focus of numerous studies (6–14), but a comprehensive study using next-generation sequencing of a cohort of sole -7 AML patients without (possibly confounding) other cytogenetic abnormalities has not been reported to date.

The aim of our study was to characterize the mutational landscape and the gene- and microRNA-expression profiles of adult AML patients with -7 as a sole cytogenetic abnormality using next-generation sequencing techniques. Since, to our knowledge, all recent molecular studies characterizing -7 AML also included patients with additional chromosomal aberrations (15, 16), we hypothesized that restricting our study to sole -7 patients would enable us to find the molecular features associated with chromosome 7 loss, without the potential bias introduced by the presence of additional cytogenetic abnormalities. A more complete understanding of molecular events could provide insight into the pathobiology of chromosome 7 loss in AML.

Materials and Methods

Patients, treatment, and cytogenetic studies

Pretreatment bone marrow (BM) or peripheral blood (PB) samples containing ≥20% of leukemic blasts were obtained from 36 adult AML patients with -7 as a sole chromosomal abnormality. Cytogenetic analyses of pretreatment BM and/or PB samples were performed by institutional laboratories approved by Cancer and Leukemia Group B (CALGB)/Alliance for Clinical Trials in Oncology (Alliance). Cytogenetic results were confirmed by central karyotype review (17). All patients were treated with cytarabine and an anthracycline-based cytotoxic chemotherapy on CALGB/Alliance trials, the details of which are provided in the Supplementary data. Study protocols were in accordance with the Declaration of Helsinki and approved by the institutional review boards at each center, and all patients provided written informed consent.

Statistical analysis

Baseline characteristics were compared between younger and older -7 AML patients using Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables (18). Definitions of the clinical endpoints–CR, disease-free (DFS) and overall survival (OS)–are provided in the Supplementary data. For time-to-event analyses, we calculated survival estimates using the Kaplan-Meier method, and compared groups using the Cox proportional hazard regression models. The dataset was locked on June 10^th, 2015. Data collection and statistical analyses were performed by the Alliance Statistics and Data Center.

Molecular analyses

Patients provided written informed consent to participate in protocols CALGB 8461 (cytogenetic studies), CALGB 9665 (leukemia tissue bank) and CALGB 20202 (molecular studies), which involved collection of pretreatment BM aspirates and PB samples. Mononuclear cells were enriched through Ficoll-Hypaque gradient centrifugation and cryopreserved until use. Genomic DNA was extracted using the DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany). The mutational status of 80 protein coding genes was determined centrally at The Ohio State University by targeted amplicon sequencing using the MiSeq platform (Illumina, San Diego, CA; see Supplementary data for details). Briefly, DNA library preparations were performed according to the manufacturer’s instructions. Samples were pooled and run on the MiSeq machine using the Illumina MiSeq Reagent Kit v3. Sequenced reads were aligned to the hg19 genome build using the Illumina Isis Banded Smith-Waterman aligner. Single nucleotide variant and indel calling were performed using MuTect and VarScan, respectively (19, 20). All called variants underwent visual inspection of the aligned reads using the Integrative Genomics Viewer (Broad Institute, Cambridge, MA; ref. 21). Testing for the presence or absence of FLT3 internal tandem duplication (FLT3-ITD) and CEBPA mutations were performed as previously described (22, 23). Thus, since FLT3 was also included in the 80-gene Miseq panel, the mutation status of 81 genes (80 genes plus CEBPA) was assessed in our study. Gene mutations were assigned to functional groups as previously described by Ley et al. (activated signaling, chromatin remodeling, cohesin complex, methylation, NPM1, spliceosome, transcription factors, tumor suppressors; ref. 24). Mutations detected in the targeted sequencing approach at the DNA level were also analyzed for their presence at the RNA level by visual inspection of the BAM files using the same criteria as for DNA variant calls (Supplementary Table S1). Of the 140 detected mutations, 99 were also evaluable at the RNA level. RNAseq analysis identified 93 (94%) of those 99 mutations. All six mutations not seen in the RNAseq data had Variant Allele Fractions (VAFs) <15%. Ten patients with available buccal swab material (n=9) or BM obtained during morphologic remission (n=1) were additionally analyzed for the germline status of the genes found mutated in leukemic samples using the same 80 gene targeted amplicon sequencing panel (Supplementary Table S2).

Gene- and microRNA-expression profiling

Total RNA was extracted from BM or PB cells from 31 patients using the TRIzol extraction method kit (QIAGEN). For five patients, the quantity and/or quality of RNA were insufficient to perform expression analyses. The RNA was used to make two sets of libraries. Gene expression libraries were constructed with the Illumina TruSeq stranded mRNA library preparation kit, while small RNA libraries were constructed with the NEBNext small RNA library prep kit (both New England Biolabs, Ipswich, MA). Both libraries were sequenced on an Illumina HiSeq 2500 targeting 40×10⁶ and 25×10⁶ reads for mRNA and small RNA libraries, respectively.

Quantification of polyA+ RNA sequences was performed as reported by TCGA (24). The ‘rnaseqv2’ University of North Carolina mRNA-seq Pipeline described at the TCGA Data Coordinating Center (DCC) was implemented in order to compare the -7 AML samples to the TCGA cohort (24). Essentially, the fastq files were aligned against hg19 using MAPSPLICE (25) and quantified with RSEM (26). The quantification of miRNAseq data was obtained using NovoAlign and miRBase 21 (www.mirbase.org). Batch effects were removed using the ComBat function (27) within the sva package in R (28, 29). The unsupervised clustering of sole -7 and TCGA samples (including patients with chromosome 7 abnormalities and/or complex karyotype) is depicted in Supplementary Figure S1. TCGA samples with chromosome 7 abnormalities and/or complex karyotype preferentially clustered with our sole -7 AML cohort.

Results

Clinical characteristics and outcome of AML patients with sole -7

Clinical characteristics of patients with sole -7, both as a total cohort and separated into younger (aged <60 years) and older (≥60 years) patients, are provided in Table 1. Only 21% of the sole -7 patients had extramedullary involvement of their leukemia, and there were no significant differences in the clinical characteristics between the age groups (Table 1).

Table 1.

Pretreatment clinical characteristics of patients with acute myeloid leukemia (AML) and sole -7, and comparison by age group (<60 years vs. ≥60 years)

Characteristic	Sole -7 AML All patients (n=36)	Sole -7 AML <60y (n=11)	Sole -7 AML ≥60y (n=25)	P valuea
Age, years				--
Median	63	52	68
Range	35–78	35–59	61–78
Sex, n (%) of females	15 (42)	3 (37)	12 (48)	0.30
Race, n (%)				0.29
White	34 (97)	9 (90)	25 (100)
Non-white	1 (3)	1 (10)	0 (0)
Hemoglobin, g/dl				0.90
Median	9.1	9.1	9.1
Range	7.3–12.3	7.9–10.6	7.3–12.3
Platelet count, × 109/l				0.27
Median	65	51	69
Range	9–989	9–274	10–989
WBC, × 109/l				0.75
Median	12.3	12.2	13.2
Range	1.6–212.7	5.2–111.0	1.6–212.7
Bone marrow blasts, %				0.47
Median	48	47	52
Range	23–93	23–93	28–91
Blood blasts, %				0.31
Median	30	37	25
Range	2–99	16–89	2–99
s-AML, n (%)				0.29
Yes	4 (11)	0 (0)	4 (16)
No	32 (89)	11 (100)	21 (84)
t-AML, n (%)				1.00
Yes	1 (3)	0 (0)	1 (4)
No	35 (97)	11 (100)	24 (96)
Extramedullary involvement, n (%)				1.00
Total	7 (21)	2 (20)	5 (22)
Central nervous system	0 (0)	0 (0)	0 (0)
Gum hypertrophy	0 (0)	0 (0)	0 (0)
Hepatomegaly	2 (6)	1 (9)	1 (4)
Lymphadenopathy	2 (6)	1 (9)	1 (4)
Mediastinal mass	0 (0)	0 (0)	0 (0)
Skin Infiltrates	0 (0)	0 (0)	0 (0)
Splenomegaly	5 (15)	2 (18)	3 (13)
Transplantation, n (%)				0.52
Allo in 1st CR	2 (6)	1 (9)	1 (4)
No Allo in 1st CR	34 (94)	10 (91)	24 (96)

Abbreviations: Allo, allogeneic hematopoietic stem-cell transplantation (HSCT); CR, complete remission, n, number; s-AML, secondary AML; t-AML, therapy-related AML; WBC, white blood count; y, years.

^aP-values relate to the comparison of younger versus older sole -7 patients. P-values for categorical variables are from Fisher’s exact test, P-values for continuous variables are from the Wilcoxon rank sum test.

Consistent with previous reports (1–5), the outcomes of sole -7 AML patients were very poor, with only 42% of patients achieving a CR. The median DFS was 8.4 months and the median OS was 9.2 months (Table 2). Only two patients were alive three years after diagnosis; one of the surviving patients received allogeneic hematopoietic stem-cell transplantation (HSCT) in first CR. There were no significant differences in CR rates or OS between the younger and older patients (Table 2), despite differences in treatment intensity. Possible differences in DFS between the two age groups could not be statistically assessed due to the limited sample size.

Table 2.

Outcome of all patients with acute myeloid leukemia (AML) and sole -7, and comparison of outcomes by age group (<60 years vs. ≥60 years)

Endpoint	Sole -7 AML All patients (n=36)	Sole -7 AML <60 y (n=11)	Sole -7 AML ≥60 y (n=25)	P-valuea
Complete remission, n (%)	15 (42)	5 (45)	10 (40)	1.00
Disease-free survivalb				--
Median, months	8.4	6	13.2
Disease-free at 12 months, % (95% CI)	33 (10–59)	0	50 (15–77)
Disease-free at 36 months, % (95% CI)	8 (1–31)	0	13 (1–42)
Overall survival				0.63
Median, months	9.2	9.6	8.4
Alive at 12 months, % (95% CI)	44 (28–60)	45 (17–71)	44 (24–62)
Alive at 36 months, % (95% CI)	7 (1–19)	9 (1–33)	5 (0–21)

Abbreviations: CI, confidence interval, y, years.

^aP-values relate to the comparison of younger versus older sole -7 patients. P-value for complete remission is from Fisher’s exact test. P-value for overall survival is from the log-rank test.

^bPatients who received an allogeneic hematopoietic stem-cell transplantation in first complete remission were not included in the analysis of disease-free survival (DFS). The P-value for DFS could not be determined due to the limited sample size.

The mutational landscape of AML patients with sole -7

Results of mutational analyses performed in our patient cohort are shown in Table 3, Figure 1 and Supplementary Table S1. A total of 140 mutations were found in 41 of the 81 tested genes, with a median of four mutations detected per patient (range: 0–7 mutations). Six genes had a mutation frequency of >20%: the most frequently mutated gene was RUNX1 (n=10 patients, 28% of the cohort), followed by ASXL1 (n=9, 25%), NRAS (n=9, 25%), TET2 (n=9, 25%), DNMT3A (n=8, 22%) and SRSF2 (n=8, 22%). Other recurrently mutated genes included SETBP1 (n=4, 11%), mutations of which have been previously reported to be enriched in -7 patients (30), and NOTCH1 (n=4, 11%), a gene frequently mutated in chronic lymphocytic leukemia (CLL; ref. 31). Three patients had mutations in GATA2, which have been implicated in familial -7 syndromes (32, 33). Some of the detected mutational features were in line with previous findings in myelodysplastic syndromes (MDS) with loss of chromosome 7 material, including the detection of four mutations in the EZH2 gene, which is located on chromosome 7 and has been suggested to act as a tumor suppressor gene (10, 34). Interestingly, we did not detect any mutations in some of the major AML-associated genes, including CEBPA, FLT3-ITD, and TP53.

Table 3.

Frequencies of mutations in single genes observed in patients with acute myeloid leukemia (AML) and sole -7

Genea	Sole -7 AML All patients (n=36)	Sole -7 AML <60y (n=11)	Sole -7 AML ≥60y (n=25)	P valueb
AKT1, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
ASXL1, n (%)				0.69
Mutated	9 (25)	3 (30)	6 (24)
Wild type	27 (74)	8 (70)	19 (76)
BCOR, n (%)				1.00
Mutated	2 (6)	0 (0)	2 (8)
Wild type	34 (94)	11 (100)	23 (92)
BCORL1, n (%)				0.52
Mutated	2 (6)	1 (9)	1 (4)
Wild type	34 (94)	10 (91)	24 (96)
BRAF, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
BRD4, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
BRINP3, n (%)				1.00
Mutated	2 (6)	0 (0)	2 (8)
Wild type	34 (94)	11 (100)	23 (92)
BTK, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
CBL, n (%)				0.22
Mutated	3 (8)	2 (18)	1 (4)
Wild type	33 (92)	9 (82)	24 (96)
DNMT3A, n (%)				1.00
Mutated	8 (22)	2 (18)	6 (24)
Wild type	28 (78)	9 (82)	19 (76)
ETV6, n (%)				0.54
Mutated	3 (8)	0 (0)	3 (12)
Wild type	33 (92)	11 (100)	22 (88)
EZH2, n (%)				0.08
Mutated	4 (11)	3 (27)	1 (4)
Wild type	32 (89)	8 (73)	24 (96)
FLT3-TKD, n (%)	2 unknown		2 unknown	1.00
Present	1 (3)	0 (0)	1 (4)
Absent	33 (97)	11 (100)	22 (96)
GATA2, n (%)				1.00
Mutated	3 (8)	1 (9)	2 (8)
Wild type	33 (92)	10 (91)	23 (92)
HIST1H1E, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
IDH1, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
IDH2, n (%)				0.64
Mutated	6 (17)	1 (9)	5 (20)
Wild type	30 (83)	10 (91)	20 (80)
KIT, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
KMT2A, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
KRAS, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
NOTCH1, n (%)				0.57
Mutated	4 (11)	2 (18)	2 (8)
Wild type	32 (89)	9 (82)	23 (92)
NPM1, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
NRAS, n (%)				0.41
Mutated	9 (25)	4 (36)	5 (20)
Wild type	27 (75)	7 (64)	20 (80)
PLEKHG5, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
PTPN11, n (%)				0.08
Mutated	4 (11)	3 (27)	1 (4)
Wild type	32 (89)	8 (73)	24 (96)
RAF1, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
RUNX1, n (%)				1.00
Mutated	10 (28)	3 (27)	7 (28)
Wild type	26 (72)	8 (73)	18 (72)
SETBP1, n (%)				0.57
Mutated	4 (11)	2 (18)	2 (8)
Wild type	32 (89)	9 (82)	23 (92)
SF1, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
SF3B1, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
SMARCA2, n (%)				0.63
Mutated	5 (14)	2 (18)	3 (12)
Wild type	31 (86)	9 (82)	22 (88)
SMC1A, n (%)				1.00
Mutated	3 (8)	1 (9)	2 (8)
Wild type	33 (92)	10 (91)	23 (92)
SMC3, n (%)				0.54
Mutated	3 (8)	0 (0)	3 (12)
Wild type	33 (92)	11 (100)	22 (88)
SRSF2, n (%)				0.08
Mutated	8 (22)	0 (0)	8 (32)
Wild type	28 (78)	11 (100)	17 (68)
STAG2, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
TET2, n (%)				0.03
Mutated	9 (25)	0 (0)	9 (36)
Wild type	27 (75)	11 (100)	16 (64)
TGM7, n (%)				1.00
Mutated	2 (6)	0 (0)	2 (8)
Wild-type	34 (94)	11 (100)	23 (92)
TYK2, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
U2AF1, n (%)				0.30
Mutated	5 (14)	0 (0)	5 (20)
Wild type	31 (86)	11 (100)	20 (80)
WT1, n (%)				1.00
Mutated	4 (11)	1 (9)	3 (12)
Wild type	32 (89)	10 (91)	22 (88)
ZRSR2, n (%)				0.31
Mutated	1 (3)	1 (9)	0 (0)
Wild type	35 (97)	10 (91)	25 (100)
Number of mutations per patient				0.43
Median (range)	4 (0–7)	3 (1–7)	4 (0–7)

Abbreviations: FLT3-TKD, tyrosine kinase domain mutation in the FLT3 gene; y, years.

^aOnly genes found mutated in at least one patient are listed in alphabetical order in this Table. No mutations were detected in the ARAF, ATM, AXL, BCL2, BIRC6, CCND1, CCND2, CEBPA (biallelic), CSNKN1A, CTNNB1, FBXW7, GATA1, GSK3B, HNRNPK, IKZF1, IKZF3, IL7R, JAK1, JAK2, JAK3, KLHL6, MAPK1, MAPK3, MED12, MYD88, PHF6, PIK3CD, PIK3CG, PIKD3, PLCG2, PRKCB, PTEN, RAD21, SAMHD1, SF3A1, SYK, TP53, U2AF2, XPO1, or ZMYM3 genes. No patient was found to harbor FLT3-ITD.

^bP-values relate to the comparison of younger versus older sole -7 patients. P-values are from Fisher’s exact test.

Figure 1.

Oncoprint of mutations found in functional groups (listed by descending frequencies; upper panel) and single gene mutations (lower panel) in patients with AML and sole -7. Patients (one per column) are depicted separately by age group (<60 years, green and ≥60 years, orange). Black highlights in the upper panels indicate the presence of ≥1 mutation in ≥1 gene assigned to the functional group. The lower panels list the respective gene mutations. Red highlights indicate the presence of a gene mutation, grey highlights indicate wild-type status (see Patients and Methods for details), white highlights indicate that the mutation status is not available (n.a.). s-AML denotes secondary AML; t-AML, therapy-related AML.

Five patients harbored mutations in the gene SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 2 (SMARCA2, Table 3, Figure 1), which encodes one of the two alternative catalytic subunits of the BAF chromatin remodeling complex (35). Non-truncating mutations in the ATPase region of SMARCA2 were previously identified in the rare Nicolaides-Baraitser syndrome (35, 36), which is characterized by intellectual disability and multiple congenital anomalies, but not increased risk of cancer. Recurrent SMARCA2 mutations are found in several different solid tumors, but in leukemia, SMARCA2 mutations were previously reported only in single cases of adult and pediatric AML (24) and in a patient with progressing CLL (37). In our cohort of sole -7 patients, four patients had a single SMARCA2 mutation and one patient had two SMARCA2 mutations. Some of the locations of mutations we discovered were in close proximity to mutation locations previously reported in Nicolaides-Baraitser syndrome patients (35, 36; Supplementary Figure S2). Interestingly, all three patients with a GATA2 mutation also had a mutation in SMARCA2 (Figure 1), suggesting there may be an undiscovered functional relationship between these genes. Four of the six SMARCA2 mutations had VAFs of <20%, suggesting that they likely represent later mutational events. While the mutation with the highest VAF (52%) could also be detected in the paired RNAseq data (37%, Supplementary Table S1), four of the remaining five SMARCA2 mutations could not be evaluated at the RNA level due to low SMARCA2 expression and thus inadequate coverage. The evaluable SMARCA2 mutation with a VAF of 13% at the DNA level was not detected at the RNA level (Supplementary Table S1).

Since -7 AML has been previously associated with familial AML syndromes (32, 33), we also tested germline samples of patients with available material (n=10) for the presence or absence of the variants detected in the leukemic samples (Supplementary Table S2). The analysis of the germline material of seven patients validated all variants detected in the diagnostic samples as somatic changes, whereas the germline material of two patients revealed one variant in each patient to be of germline origin. The CBL P510A variant found in patient no. 6 was detected with a VAF of 56% in the leukemic sample and with a VAF of 44% in the paired germline DNA (Supplementary Table S2). Further data mining revealed that this variant was also detected in a single individual in the 1000 genome project [minor allele frequency <0.0001%, amino acid change predicted not to affect the protein function (Scale-invariant feature transform; SIFT score: −0.62)], suggesting that CBL P510A represents an ultra-rare and probably benign germline variant. In contrast, the RAF1 R318W variant in patient no. 9, detected in both leukemic and germline material (with VAFs of 47% and 46%, respectively; Supplementary Table S2), is listed in the COSMIC database as a validated somatic variant, which was previously detected in a patient with endometrial cancer (COSM1037798, TCGA-B5-A0K6-01). Furthermore, the resulting amino acid change is predicted to be deleterious for RAF1’s protein function (SIFT scores: −3.84 to −4.52). Thus, RAF1 R318W could indeed represent a novel variant predisposing to -7 AML. Finally, the remission material of patient no. 10 could not be evaluated for possible germline variants, because all mutations in the remission sample had high VAFs suggesting the presence of residual disease (Supplementary Table S2).

Functional groups of gene mutations in AML patients with sole -7

To detect mutation patterns in sole -7 AML, we next assigned mutations to previously reported functional groups (24). All but a single patient had one or more mutations in at least one of the major functional groups comprised of the genes included in our mutation panel (median, 2 affected groups/patient; range, 0–5; Figure 1). The most frequently observed mutations were those in genes involved in methylation (DNMT3A, IDH1, IDH2 and TET2) that were found in 19 (53%) patients, and in genes leading to activated signaling (AKT1, BRAF, BRD4, BTK, CBL, FLT3-TKD, KIT, KRAS, NRAS, PTPN11, RAF1 and TYK2) that were present in 18 (50%) patients (Table 4 and Figure 1). Notably, the group of genes leading to activated signaling was dominated by mutations of genes affecting the RAS pathway, namely CBL, KRAS, NRAS, and PTPN11, which were detected in 15 (42%) patients. This is consistent with previous reports of the frequent involvement of RAS pathway members in myeloid malignancies, including AML, with unbalanced abnormalities involving chromosome 7 (15, 38). Other functional groups comprising frequently mutated genes were: chromatin remodeling (ASXL1, BCOR, BCORL1, EZH2 and SMARCA2; n=15 patients, 42%), spliceosome (SF1, SF3B1, SRSF2, U2AF1 and ZRSR2; n=14 patients, 39%) and transcription factors (ETV6, GATA2 and RUNX1; n=13 patients, 36%). In contrast, fewer mutations were detected in cohesin complex genes (SMC1A, SMC3 and STAG2; n=7 patients, 19%), and tumor suppressors (n=4 patients, 11%), with WT1 being the only tumor suppressor gene that was found mutated in the sole -7 cohort. Only one patient (3%) harbored an NPM1 mutation (Table 4 and Figure 1).

Table 4.

Frequencies of gene mutations in functional groups observed in patients with acute myeloid leukemia (AML) and sole -7

Functional groupa	Sole -7 AML All patients (n=36)	Sole -7 AML <60y (n=11)	Sole -7 AML ≥60y (n=25)	P valueb
Methylation, n (%)				0.07
Mutated	19 (53)	3 (27)	16 (64)
Wild type	17 (47)	8 (73)	9 (36)
Activated signaling, n (%)				0.47
Mutated	18 (50)	7 (64)	11 (44)
Wild type	18 (50)	4 (36)	14 (56)
Chromatin remodeling, n (%)				0.14
Mutated	15 (42)	7 (64)	8 (32)
Wild type	21 (58)	4 (36)	17 (68)
Spliceosome, n (%)				0.14
Mutated	14 (39)	2 (18)	12 (48)
Wild type	22 (61)	9 (82)	13 (52)
Transcription factors, n (%)				1.00
Mutated	13 (36)	4 (36)	9 (36)
Wild type	23 (64)	7 (64)	16 (64)
Cohesin complex, n (%)				0.40
Mutated	7 (19)	1 (9)	6 (24)
Wild type	29 (81)	10 (91)	19 (76)
Tumor suppressors, n (%)				1.00
Mutated	4 (11)	1 (9)	3 (12)
Wild type	32 (89)	10 (91)	22 (88)
NPM1, n (%)				1.00
Mutated	1 (3)	0 (0)	1 (4)
Wild type	35 (97)	11 (100)	24 (96)
Methylation and/or spliceosome and/or cohesin complex, n (%)				0.04
Mutated	28 (78)	6 (55)	22 (88)
Wild type	8 (22)	5 (45)	3 (12)

Abbreviation: y, years.

^aThe functional groups (ref. 24; ≥1 mutation in ≥1 gene assigned to the functional group, listed are only the genes which were found mutated in our cohort) are listed by descending mutation frequencies in all -7 patients. They include: Methylation, DNMT3A, IDH1/2 or TET2; Activated signaling, AKT1, BRAF, BRD4, BTK, CBL, FLT3-TKD, KIT, KRAS, NRAS, PTPN11, RAF1 or TYK2; Chromatin remodeling, ASXL1, BCOR, BCORL1, EZH2 or SMARCA2; Spliceosome, SF1, SF3B1, SRSF2, U2AF1 or ZRSR2; Transcription factors, ETV6, GATA2 or RUNX1; Cohesin complex, SMC1A, SMC3 or STAG2; Tumor suppressors, WT1; NPM1, NPM1.

^bP-values relate to the comparison of younger versus older sole -7 patients. P-values are from Fisher’s exact test.

Age-associated mutation patterns in AML patients with sole -7

For cytogenetically normal AML (CN-AML), it is known that certain mutations have age-related differences in their frequencies (39, 40). Our comparison of the frequencies of single gene mutations (Table 3) and functional groups (Table 4) between patients younger than 60 years and those aged 60 years and older revealed preferential mutation patterns in both age groups. Whereas EZH2 and PTPN11 mutations tended to occur more frequently in younger patients with sole -7 AML (27% of younger vs. 4% of older patients), TET2 and SRSF2 mutations were detected exclusively in older patients (TET2 in 36% and SRSF2 in 32%). However, with the exception of TET2 (P=0.03), none of the observed differences in the mutation frequencies between the two age groups reached statistical significance (Table 3). The fact that TET2 mutations were found exclusively in sole -7 AML patients who are aged ≥60 years differs somewhat from the previous reports on both CN-AML (40) and an unselected patient cohort comprised of patients with various cytogenetic findings (41). Although the incidence of TET2 mutations was also the highest among older patients in both CN-AML (29%, ref. 40) and an unselected AML patient population (24%, ref. 41), in contrast to younger sole -7 AML patients who did not harbor any TET2 mutations, considerable proportions of younger patients with CN-AML (15%, ref. 40) and those with AML with various karyotypes (7%, ref. 41) carried TET2 mutations. Because of a relatively low number of younger patients with sole -7 AML in our study, the lack of TET2 mutations in sole -7 AML patients under the age of 60 years should be corroborated.

We also observed a preferred involvement of functional groups depending on age (Table 4). Mutations in genes leading to activated signaling were frequently found in both age groups (64% of younger and 44% of older patients). However, while younger patients frequently harbored mutations in chromatin remodeling genes (64%, compared with 32% of older patients), older patients were characterized by mutations in methylation genes, which were found in 64% of patients (compared with 27% of younger patients), followed by mutations in spliceosome genes (48%, compared with 18% of younger patients). Eighty-eight percent of older patients harbored at least one mutation in genes involved in methylation, the spliceosome and/or the cohesin complex (compared with only 55% of younger patients, P=0.04). Even though the patient cohorts are small, this is the first report to suggest possible age-related differences in single genes and functional groups affected by mutations in sole -7 AML.

Effects of gene mutations on the outcome of AML patients with sole -7

To test whether any of the clinical parameters, single gene mutations or mutational groups (as defined by the TCGA, ref. 24) impact on the outcome of sole -7 patients, we performed univariable analyses for associations with CR achievement, DFS and OS in the whole patient cohort. None of the tested parameters was significantly associated with the achievement of CR, DFS or OS (Table 5).

Table 5.

Univariable outcome analyses in patients with acute myeloid leukemia and sole -7

Endpoint	Variablesa	P valueb
Total cohort (n=36)
Complete remission	Spliceosome mutationc, ≥1 mutation vs. no mutation	0.06
	SMARCA2, mutated vs. wild-type	0.09
Overall survival	SMARCA2, mutated vs. wild-type	0.09
	U2AF1, mutated vs. wild-type	0.06
	WT1, mutated vs. wild-type	0.06
Patients ≥60 years (n=25)
Complete remission	Spliceosome mutationc, ≥1 mutation vs. no mutation	0.03d
Overall survival	U2AF1, mutated vs. wild-type	0.07

^aVariables tested in the univariable outcome analyses are listed in the Supplementary data. Listed in Table 5 are all variables with a P-value ≤0.10. Variables written in bold-type letters are those associated with unfavorable outcome. No variable tested for association with DFS in all patients or in patients ≥60 years had a P-value ≤0.10.

^bP-values for achievement of CR using logistic regression; P-values for disease-free and overall survival using Cox proportional hazards regression.

^cSpliceosome mutations include mutations in the SF1, SF3B1, SRSF2, U2AF1 and/or ZRSR2 genes.

^dOdds ratio for achievement of CR ≥1 spliceosome mutation versus no mutation: 0.13, 95% confidence interval: 0.02 to 0.82.

We next tested the older and younger patient groups separately for associations between clinical and molecular parameters and outcome. Again, no parameter was significantly associated with any of the outcome endpoints in younger sole -7 patients, which may be due to the small sample size. However, older patients with sole -7 harboring a mutation in a spliceosome gene had a lower CR rate compared with that of patients without a spliceosome mutation (17% vs. 62%, P=0.03; Table 5). No parameter was significantly associated with DFS or OS.

Gene- and microRNA-expression profiling of AML patients with sole -7

The reduced expression of some genes mapped to chromosome 7 caused by the chromosome loss has been suggested as a major factor in the pathogenesis of -7 AML (6, 7), and several promising haploinsufficient genes located on chromosome 7 have already been identified as candidates contributing to the phenotype of -7 AML (8–13). Using our, the largest to date, series of patients with -7 as a sole abnormality, we comprehensively characterized the gene- and microRNA-(miR) expression changes caused by chromosome 7 loss. We compared the gene- and miR-expression profiles between sole -7 patients and AML patients with other cytogenetic findings in the TCGA AML cohort (24), excluding patients with chromosome 7 abnormalities and/or complex karyotype. We derived a signature comprising 284 genes that were significantly downregulated and 42 genes significantly upregulated in sole -7 AML (adjusted P-value <0.001; Supplementary Figure S3 and Supplementary Tables S3 and S4). Consistent with a gene dosage effect, 94% of the genes significantly downregulated in -7 AML were mapped to chromosome 7.

The most significant downregulation was observed for an open-reading-frame on chromosome 7 which likely encodes a transmembrane spanning protein C7orf42 (TMEM248). The mismatch repair gene PMS2 (42) was the second most significantly downregulated gene. We also confirmed downregulation of CUX1 (8), EZH2 (10) and MLL3 (11), all of which have been previously identified as haploinsufficient genes of chromosome 7 (Supplementary Table S3).

The most significantly upregulated gene in sole -7 AML was the protein tyrosine phosphatase receptor type M gene (PTPRM), which is an important regulator of cell growth, differentiation and oncogenic transformation (43). Other highly overexpressed genes in sole -7 AML included ID1, which is a common downstream target of oncogenic tyrosine kinases in leukemic cells (44), and MECOM, the MDS1/EVI1 complex locus, which has been previously associated with inv(3)(q21q26.2)/t(3;3)(q21;q26.2) and chromosome 7 abnormalities (38, 45) (Supplementary Table S4). The miR-expression profile consisted of 16 differentially expressed miRs, six of which were significantly downregulated and 10 significantly upregulated in sole -7 patients compared with non-sole -7 AML (Figure 2, Supplementary Tables S5 and S6). All but one of the six downregulated miRs are mapped to chromosome 7 (miR-25, miR-93, miR-106b, miR-339, and miR-671). MiR-25, miR-93 and miR-106b belong to a miR cluster located in different introns of the MCM7 gene, which itself did not belong to the most downregulated mRNAs in sole -7 AML. The most downregulated miR with respect to the observed fold-change was miR-9-1, which is mapped to chromosome 1 (Supplementary Table S5).

Figure 2.

Heatmap depicting the differential miR expression of AML patients with sole -7 (n=31, indicated by red bars) and AML patients with other cytogenetics of the TCGA AML cohort¹⁹ (excluding patients with chromosome 7 abnormalities and/or complex karyotype, n=136). Lower expression is shown in green, while higher expression is shown in red. The derived miR-expression signature comprised six miRs downregulated and 10 miRs upregulated in sole -7 AML, with an adjusted P-value of <0.001. The significantly down- and upregulated miRs and the corresponding P-values and fold changes can be found in Supplementary Tables S5 and S6, respectively.

The miRs upregulated in sole -7 patients mainly belong to two miR-clusters: the cluster comprising miR-20b, miR-363 and miR-106a located on the X chromosome and the cluster of miR-99b, miR-125a and miR-let7e located on chromosome 19 (Supplementary Table S6), suggesting a shared regulatory mechanism.

It has been recently shown that large-scale loss of chromosome material in cancer cells can disturb multiple signaling pathways and produce phenotypes distinct from those arising through loss of a single gene (46). We performed gene ontology (GO) analyses of the differentially expressed mRNAs and of the mRNA targets of the differentially expressed miRs to test whether genes involved in specific biologic processes are enriched in signatures associated with loss of the entire chromosome 7. Indeed, our GO analysis of downregulated genes located on chromosome 7 revealed a significant enrichment of genes involved in the molecular function of RNA binding (both Poly-A and non-Poly-A), and the biological process of small protein conjugation (Supplementary Table S7).

In addition, an analysis of the mRNA targets of the miRs which were upregulated in -7 AML showed a striking enrichment of target mRNAs involved in metabolic processes, especially in the cellular nitrogen compound (Supplementary Figure S4). In contrast, GO analysis of the mRNA targets of the downregulated miRs only resulted in weak clustering (Supplementary Figure S5). Taken together, the observed enrichments suggest that multiple gene networks might also be important in the pathogenesis of -7 AML.

Discussion

In this study, we analyzed the hitherto largest, to our knowledge, homogeneous group of adult AML patients with -7 as the sole abnormality to provide insights into the molecular characteristics associated with chromosome 7 loss using a comprehensive panel of 81 cancer- and/or leukemia-associated genes. The relatively large number of gene mutations detected in virtually all major functional groups indicates that -7 AML is molecularly heterogeneous. However, our data also suggest that mutation patterns we observed are partly associated with the patients’ age. Although both the younger and older patients frequently harbored mutations in genes leading to activated signaling that comprise kinases and RAS pathway members, younger patients with -7 AML often had mutations in chromatin modifiers. In contrast, older -7 AML patients often harbored mutations in genes involved in methylation and/or the spliceosome. Our finding that spliceosome mutations may confer a reduced probability of achieving a CR in patients receiving cytarabine and an anthracycline (7+3)-based induction chemotherapy might provide a first rationale for testing of spliceosome mutations at diagnosis in older -7 AML patients. Whether this molecular feature will also be predictive of CR attainment in patients receiving other therapeutic approaches, such as a demethylating agent therapy, should be assessed by future studies, as thus far, to our knowledge, no published data about molecular prognosticators in -7 AML in patients treated with demethylating agents are available.

Concordant with our results, enrichment in mutations affecting the RAS pathway has been previously suggested as an important biologic feature of -7 AML by McNerney et al. (15). However, while that study suggested a paucity of mutations in methylation genes (15), our data found the methylation group to be the most frequently affected functional group of sole -7 AML. This discrepancy may be related to the limited sample sizes in both studies, but also may stem from the fact that McNerney et al. (15) studied not only patients with -7 but also patients with del(7q) (~18%), and that two-thirds of patients analyzed in their study harbored chromosomal aberrations in addition to -7 or del(7q). Moreover, a recent study by Gröschel et al. (38) on 30 patients with AML, five patients with myelodysplastic syndromes, four cell lines and two patients with chronic myelogenous leukemia in blast crisis that harbored inv(3) or t(3;3), which in two thirds of the cases was accompanied by -7 as an additional cytogenetic abnormality, detected mutations in genes leading to activated signaling (kinases and RAS pathway members) in 98% of samples studied, but only 10% of their samples had mutations in methylation genes. We also frequently detected activated signaling mutations in our patient cohort, but our sole -7 AML patients had frequent mutations in methylation, chromatin remodeling and spliceosome genes. These apparently discordant results may be attributed not only to the fact that one-third of cases studied by Gröschel et al. (38) did not harbor -7, but also to the age-related differences in mutation distribution we report herein, since the majority of our patients with sole -7 were older than 60 years, whereas patients with -7 that accompanies inv(3) or t(3;3) are usually younger.

The use of large-scale next-generation sequencing for molecular profiling studies of leukemia/cancer patients is becoming increasingly common. However, the diversity of the analyzed cohorts makes it difficult to detect novel variants with potential importance in specific patient subgroups. The identification of six SMARCA2 mutations in five sole -7 AML patients in our cohort underlines the discovery potential of sequencing efforts in small but homogeneous patient populations. While a SMARCA2 mutation has been previously reported in a single adult patient with AML and a translocation involving the KMT2A (MLL) gene (24), the prevalence of SMARCA2 mutations and the potential of their enrichment in certain subsets of AML patients have likely been underestimated. It seems noteworthy that four of the six SMARCA2 mutations had VAFs <20%, providing a possible reason why they were not more frequently discovered in other sequencing studies. Efforts to assess the frequency of SMARCA2 mutations in other cytogenetic subgroups and to evaluate their possible functional consequences are already ongoing. Mutations in other genes belonging to the ATP-dependent mSWI/SNF chromatin remodeling complex have also been detected recently in solid tumors. Those studies and the detected mutation diversity not only highlight the pathophysiologic importance of these genes, but also suggest that specific genes encoding subunits of the mSWI/SNF complex appear to be mutated in specific cancers (47).

In addition, our finding of the RAF1 variant as a germline change, which may be a leukemia-predisposing mutation in the affected patient, highlights the usefulness of paired germline-leukemic DNA analyses, especially in AML with -7, which has been previously associated with familial AML syndromes. In fact, two other RAF1 germline variants have been previously detected both in the primary tumors and in the leukemic blasts of two patients with therapy-related AML after treatment for, respectively, embryonal cell carcinoma of the testis and colorectal cancer (48), suggesting that these RAF1 mutations may constitute possible cancer- and AML-predisposing events.

Furthermore, the results of our gene- and microRNA-expression analyses provide important insights into the biology of -7 AML. Besides the expected and previously described largely homogeneous downregulation of genes located on chromosome 7 (49), we observed an overexpression of the MECOM (EVI1/MDS1) locus in most sole -7 AML samples. A possible mechanistic link between MECOM overexpression and subsequent loss of chromosome 7 has been previously suggested in a case report of two patients with genomic instability and myelodysplasia with -7 consequent to MECOM activation after gene therapy for chronic granulomatous disease (50). Moreover, MECOM overexpression has been reported as a consistent feature of AML with the chromosome 3q rearrangements, most often inv(3) and t(3;3), which in most, but not all, patients occur together with -7 as a secondary abnormality (14, 38). Our study shows that high MECOM expression is a molecular characteristic of AML with -7 also in the absence of inv(3) or t(3;3). While high expression of MECOM is known to be associated with very poor outcome of AML patients, little is known how to target this deregulated expression. The observation that additional mechanisms other than the super-enhancer formation may lead to aberrantly increased MECOM may provide an additional challenge to target this important oncogene. However, the first in vitro data about the possible efficacy of the BET inhibitor JQ1 in cell lines with t(3;3) or inv(3;3) might provide a rationale to test this therapeutic approach also in patients with sole -7 AML with aberrantly expressed MECOM (51).

Although this is the largest study characterizing the mutational landscape of adult AML with sole -7 published to date, our findings should be interpreted with caution because of the still limited sample size of 36 patients. Furthermore, while the targeted sequencing approach enabled us to also identify subclonal mutations, it did not permit discovery of novel mutations in genes that have not hitherto been associated with leukemia and/or solid tumors. These limitations may also explain the paucity of mutations detected in the only patient with t-AML, who might harbor mutation(s) in less frequently affected gene(s).

In summary, the identification of age-associated gene mutation patterns and commonly deregulated gene- and microRNA-expression profiles sheds light on the molecular events in sole -7 AML.