Acidic leucine-rich nuclear phosphoprotein-32A expression contributes to adverse outcome in acute myeloid leukemia.

BACKGROUND: Acidic leucine-rich nuclear phosphoprotein-32A (ANP32A) is a novel regulator of histone H3 acetylation and promotes leukemogenesis in acute myeloid leukemia (AML). However, its prognostic value in AML remains unclear.
METHODS: In this study, we evaluated the prognostic significance of ANP32A expression using two independent large cohorts of cytogenetically normal AML (CN-AML) patients. Multivariable analysis in CN-AML group was also presented. Based on the ANP32A expression, its related genes, dysregulation of pathways, interaction network analysis between microRNAs and target genes, as well as methylation analysis were performed to unveil the complex functions behind ANP32A.
RESULTS: Here we demonstrated overexpression of ANP32A was notably associated with unfavorable outcome in two independent cohorts of CN-AML patients (OS: P=0.012, EFS: P=0.005, n=185; OS: P=0.041, n=232), as well as in European Leukemia Net (ELN) Intermediate-I group (OS: P=0.018, EFS: P=0.045, n=115), National Comprehensive Cancer Network (NCCN) Intermediate Risk AML group (OS: P=0.048, EFS: P=0.039, n=225), and non-M3 AML group (OS: P=0.034, EFS: P=0.011, n=435). Multivariable analysis further validated ANP32A as a high-risk factor in CN-AML group. Multi-omics analysis presented overexpression of ANP32A was associated with aberrant expression of oncogenes and tumor suppressor, up/down-regulation of metabolic and immune-related pathways, dysregulation of microRNAs, and hypomethylation on CpG island and 1st Exon regions.
CONCLUSIONS: We proved ANP32A as a novel, potential unfavorable prognosticator and therapeutic target for AML. 2020 Annals of Translational Medicine. All rights reserved.

Introduction

Acute myeloid leukemia (AML) is a clinically disparate hematopoietic malignancy, characterized by predominant fatality and high heterogeneity (1). It is the most common type of adult acute leukemia with an annual incidence of 3–8/105 (2). To date, cytogenetic analysis has been one of the most useful prognostic methods to classify AML patients into different prognostic risk groups (favorable, intermediate, unfavorable) (3). The cytogenetically normal AML (CN-AML), representing 40–50% AML patients, is the largest subset of de-novo AML cases (4,5). Although CN-AML constitutes the main body of the intermediate-risk group, yet different molecular subtypes of CN-AML diverged in transcription and DNA methylation profiles (6). Moreover, due to the wide variety of clinical outcome, the prognosis of CN-AML cannot be predicted solely by cytogenetics (7). Intriguingly, without any microscopically detectable chromosome abnormalities in its leukemic blast, CN-AML harbour aberrantly expressed genes and microRNAs, mutations, epigenetic changes that can be used as potential prognostic markers (8). It is now recognized that WT1, ASXL1, DNMT3A, FLT3-ITD and TET2 mutations presence in CN-AML represent a subgroup of patients with unfavorable prognosis, while NPM1 and CEBPA are associated with favorable prognosis (9-15). Furthermore, unfavorable prognostic factors include high expression of CPT1A, ATP1B1, RUNX1, MAPKBP1, MAP7, ERG, DNMT3B, miR-3151, miR-155, etc. (16,17). As leukemogenic mechanisms of CN-AML still remains unclear, it is vital to continue identifying new molecular biomarkers of clinical significance.

Acidic leucine-rich nuclear phosphoprotein-32A (ANP32A), also known as PP32, is an acidic leucine-rich nuclear phosphoprotein which has shown to be divergently expressed in normal tissues as well as in certain types of tumor cells (18). On one hand, ANP32A correlates with cancer progression and metastasis, demonstrated to be a candidate biomarker in patients with pancreatic tumor, colorectal cancer, hepatocellular carcinoma, oral squamous cell carcinoma, etc. (19,20). On the other hand, ANP32A is a tumor suppressor in prostate cancer, non-small cell lung cancer and breast cancer by stimulating apoptosis (18,21,22). Recent study indicates that ANP32A is a novel regulator of histone H3 acetylation and promotes leukemogenesis in AML, suggesting the expression of ANP32A might be related to the prognosis of CN-AML patients (23). Therefore, the focus on the exact functions of ANP32A in hematopoietic malignancy might lead us to the identification of ANP32A as a potential molecular marker of clinical significance in CN-AML.

Here, we presented ANP32Ahigh as an unfavorable prognostic biomarker for CN-AML. We also explored ANP32A-associated genomic and epigenomic patterns to further elucidate its function. Our study represented direct evidences for ANP32A being a prognostic biomarker in AML risk stratification and a potential therapeutic target for patients with AML.

Methods

Patients and treatment

The study was approved by the local institutional review boards. In accordance with the Declaration of Helsinki, all patients provided written informed consent. The treatment of all patients was uniformly under the protocols of Dutch-Belgian Cooperative Trial Group of Hematology-Oncology (details see in http://www.hovon.nl).

The first cohort of this study was derived from a whole AML cohort, containing 185 primarily untreated CN-AML patients, all diagnosed at Erasmus University Medical Center in Rotterdam from 1990 to 2008. The cohort age ranged from 16 to 60 years old and the median age was 47 years old. All samples were collected at the time of diagnosis and contained more than 80% blast cells after thawing (24). At least 20 metaphases from bone marrow (BM) samples were examined by conventional cytogenetic methods to diagnose normal karyotype. The presence of FLT3-ITD, FLT3-TKD and the mutations of IDH1, IDH2, NPM1, CEBPA, K-RAS, N-RAS were all assessed by reverse transcription-polymerase chain reaction (RT-PCR) assays. The clinical, cytogenetic, molecular and gene expression data of these AML cases can be downloaded from the Gene Expression Omnibus (accession number: GSE6891, details see in http://www.ncbi.nlm.nih.gov/geo) (25).

The second cohort of this study contained 162 CN-AML patients with uniform treatment was used to validate findings in the first cohort. All patients received intensive double-induction and consolidation-chemotherapy in multicenter AMLCG-1999 trial from 1999 to 2003. The cohort age ranged from 17 to 83 years old and the median age was 57.5 years old. The gene expression data can be downloaded publicly (accession number: GSE12417) (26).

Microarray analysis

Gene expression was obtained from published microarray data using Affymetrix Human Genome 133 plus 2.0 as well as U133A Gene Chips (accession number: GSE9476, GSE1159, GSE6891, GSE12417) (24,25,27,28). Seventy-three CN-AML patients with mRNA, microRNA and methylation data were derived from The Cancer Genome Atlas (TCGA) (29). Design, data quality control and normalization of microarray experiments were in accordance with the standard Affymetrix protocols. Expression of mRNA and microRNA were obtained by high throughout transcriptome sequencing (RNA-seq), while methylation data was obtained by Illumina Infinium 450K BeadChips. The expression level of ANP32A was standardized in normally distributed. Then the appropriate cut-off subdivision value was compared by the 4 quartiles of 185 CN-AML patients, which the median value showed evident distinction (Figure S1A,B,C). Therefore, median value of ANP32A expression was used to classify patients into ANP32Ahigh and ANP32Alow groups. The expression levels of ERG, DNMT3A, BAALC, WT1 and other genes were obtained using the same strategy.

Statistical analysis

Overall survival (OS) was defined as the time from diagnosis date to death by any causes. Event free survival (EFS) was defined as the time from diagnosis date to the removal from the study because of the end of complete remission (CR), relapse or death (the censorship equals to 1 when death event was observed). The Kaplan-Meier method and the log-rank test were used to evaluate and validate the association between ANP32A expression and OS, EFS. The fisher exact and the Wilcoxon rank-sum test, analyzing categorical and continuous variables respectively, were used to estimate the association between ANP32A expression levels and the patients’ clinical, molecular characteristics. Multivariate Cox proportional hazard models were used to assess the effect of ANP32A expression on OS and EFS with other known risk factors in presence. Except ANP32A and other 3 clinical & demographical variables were chosen, two gene mutations were further brought into our investigation which were shown to have a strong association with AML (30,31). Student’s t-test as well as multiple hypothesis correction (False Discovery Rate, FDR) were used to identify discrepancy in gene expression, microRNA expression, and DNA methylation profiles between ANP32Ahigh and ANP32Alow groups. All analyses were performed using R3.2.3 and its related software packages.

Results

Overexpression of ANP32A in AML patients

ANP32A expression were analyzed in two microarray assays, including BM and peripheral blood samples (PB) in AML patients and healthy donors. The expression of ANP32A was remarkably high in AML BM than in Normal BM (, P=0.049, 7 AML BM samples vs. 10 normal BM samples, GEO accession number GSE9476). Overexpression of ANP32A was further validated by PB samples in the same microarray assay (, P=0.013, 19 AML PB samples vs. 10 normal PB samples). Furthermore, overexpression of ANP32A was found in different subgroups in AML patients than normal BM, including 116 CN-AML, 22 AML1-ETO, 17 MLL-translocation, and 19 CBFβ-MYH11 (, GEO accession number GSE1159). These results indicated that ANP32A was evidently overexpressed in AML patients.

Figure 1

Different expression of ANP32A. (A) AML bone marrow samples (n=7) vs. normal bone marrow samples (n=10); (B) AML peripheral blood samples (n=19) vs. normal peripheral blood samples (n=10); (C) normal bone marrow samples (n=10), CN-AML bone marrow samples (n=116), AML1-ETO bone marrow samples (n=22), MLL-translocation bone marrow samples (n=17), CBFβ-MYH11 bone marrow samples (n=19), complex karyotype bone marrow samples (n=10), other karyotype bone marrow samples (n=100). *, 0.01

Clinical and molecular characteristics between ANP32Alow and ANP32Ahigh group in AML patients

In a cohort of 185 CN-AML patients derived from GSE6891 assay (435 AML patients, no M3), ANP32A group tends to have less FAB-M2 subtype patients (P=0.0007) and more FAB-M5 patients (P=0.0004). ANP32A was significantly overexpressed in patients with FLT3-ITD (P=0.029), while no significant differences between ANP32A and ANP32A group were observed in CEBPA, NPM1 and N-ras/K-ras mutation groups (P=0.206, P=0.211, and P=0.817, respectively). Besides, ANP32A group was more likely to have a higher expression of MAPKBP1, RUNX1, and TCF4, which were known adverse prognostic biomarkers in AML. Furthermore, patients in ANP32A group was more likely to relapse than those in ANP32A group ().

Table 1

Clinical and molecular characteristics of the patients in the 185 CN-AML patients’ cohort according to ANP32A expression

Characteristics	ANP32A		U/χ2	P
	Low expression group (n=92)	High expression group (n=93)
Median age, years [range]	48 [16–60]	46 [17–60]	3889	0.2851
Gender, No. (%)			0.2653	0.6065
Male	48 (52.2)	45 (48.4)
Female	44 (47.8)	48 (51.6)
FAB subtype, No. (%)
M0	1 (1.1)	2 (2.2)	0.5669	0.5669
M1	28 (30.4)	24 (25.8)	0.4903	0.4838
M2	27 (29.3)	9 (9.7)	11.42	0.0007
M4	10 (10.9)	18 (19.4)	2.592	0.107
M5	15 (16.3)	37 (39.8)	12.62	0.0004
M6	3 (3.3)	0 (0.0)	3.083	0.079
Other	8 (8.7)	3 (3.2)	2.474	0.116
CEBPA, No. (%)			3.164	0.206
Single mutation	4 (4.3)	2 (2.2)
Double mutation	12 (13.0)	6 (6.5)
Wild type	76 (82.6)	85 (91.4)
NPM1, No. (%)			1.566	0.211
Mutation	48 (52.2)	57 (61.3)
Wild type	44 (47.8)	36 (38.7)
FLT3-ITD, No. (%)			4.732	0.029
Presence	31 (33.7)	46 (49.5)
Absence	61 (66.3)	47 (50.5)
N-ras/K-ras			0.053	0.817
Mutation	8 (8.7)	9 (9.7)
Wild type	84 (91.3)	84 (90.3)
ERG expression, No. (%)			0.655	0.418
High	49 (53.3)	44 (47.3)
Low	43 (46.7)	49 (52.7)
BAALC expression, No. (%)			0.049	0.825
High	47 (51.1)	46 (49.5)
Low	45 (48.9)	47 (50.5)
WT1 expression, No. (%)			0.913	0.339
High	43 (46.7)	50 (53.8)
Low	49 (53.3)	43 (46.2)
DNMT3A expression, No. (%)			0.655	0.418
High	49 (53.3)	44 (47.3)
Low	43 (46.7)	49 (52.7)
DNMT3B expression, No. (%)			0.005	0.942
High	46 (50.0)	47 (50.5)
Low	46 (50.0)	46 (49.5)
MAPKBP1 expression, No. (%)			5.885	0.015
High	38 (41.3)	55 (59.1)
Low	54 (58.7)	38 (40.9)
ITPR2 expression, No. (%)			0.913	0.339
High	43 (46.7)	50 (53.8)
Low	49 (53.3)	43 (46.2)
ATP1B1 expression, No. (%)			0.135	0.714
High	45 (48.9)	48 (51.6)
Low	47 (51.1)	45 (48.4)
RUNX1 expression, No. (%)			12.98	0.0003
High	34 (37.0)	59 (63.4)
Low	58 (63.0)	34 (36.6)
TCF4 expression, No. (%)			4.545	0.033
High	39 (42.4)	54 (58.1)
Low	53 (57.6)	39 (41.9)
Relapse, No. (%)			5.342	0.021
Yes	58 (63.0)	73 (78.5)
No	34 (37.0)	20 (21.5)

High expression of ERG, BAALC, WT1, DNMT3A, DNMT3B, MAPKBP1, ITPR2, ATP1B1, RUNX1 and TCF4 was defined as above the median expression of all samples, respectively. CN-AML, cytogenetically normal acute myeloid leukemia; FAB, French-American-British classification; ITD, internal tandem duplication.

Overexpression of ANP32A was associated with unfavorable outcome in AML patients

As patients in ANP32A group was more likely to relapse than those in ANP32A group in the cohort of 185 CN-AML patients (, P=0.021), the survival analysis was further carried out in a whole cohort of CN-AML patients (n=185) and ELN Intermediate-I patients (n=115). ANP32A group showed significantly shorter OS and EFS in CN-AML patients’ cohort (, ANP32A: median OS: 11.9 vs. 37.4 months, P=0.012; median EFS: 8.4 vs. 18.3 months, P=0.005) as well as in ELN Intermediate-I patients’ cohort (, ANP32A: median OS: 8.5 vs. 30.4 months, P=0.018; median EFS: 6.9 vs. 13.0 months, P=0.045). To further validate the prognostic significance of ANP32A expression, multivariable analysis was carried out in these two cohorts. It showed that ANP32A group had 0.47 times higher risks on EFS and 0.46 times higher risks on OS in CN-AML patients’ cohort (, P=0.0088 and P=0.016, respectively). The presence of FLT3-ITD was another adverse factor (P=0.009 in EFS and P=0.019 in OS). As for the ELN Intermediate-I patients’ cohort, ANP32A group had 0.45 times higher risks on EFS and 0.48 times higher risks on OS (, P=0.041 and P=0.051, respectively), indicating ANP32A expression had more prognostic significance in CN-AML cohort. In order to further investigate the prognostic value of ANP32A expression, survival analysis was also performed in 435 non-M3 AML patients and 255 NCCN Intermediate Risk AML patients. The results indicated a remarkable shorter OS and EFS of ANP32A group in 435 non-M3 AML patients’ cohort (, ANP32A: median OS: 16.9 vs. 24.2 months, P=0.034; median EFS: 9.0 vs. 13.1 months, P=0.011) and 255 NCCN Intermediate Risk AML patients’ cohort (, ANP32A: median OS: 16.5 vs. 24.7 months, P=0.048; median EFS: 9.0 vs. 14.1 months, P=0.039). However, the multivariate analysis in these two cohorts didn’t confirm the independent prognostic value of ANP32A (see ), further indicating the prognostic significance of ANP32A in CN-AML. To validate the unfavorable outcome associated with ANP32A in AML patients, another survival analysis was conducted in another independent cohort of 232 CN-AML patients derived from GSE12417 assay. ANP32A group showed significantly shorter OS in CN-AML patient cohort (, ANP32A: median OS: 9.4 vs. 16.7 months, P=0.041).

Figure 2

The prognostic significance of ANP32A expression in AML patients. Overall survival and event free survival in 185 CN-AML patients (A,B), 115 ELN Intermediate-I patients (C,D), 435 non-M3 AML patients (E,F), 255 NCCN-intermediate risk AML patients (G,H).

Table 2

Multivariable analysis of OS and EFS in 185 CN-AML and 115 ELN Intermediate-I patients

Variables in final model by end point	EFS				OS
	HR	95% CI	P value		HR	95% CI	P value
All CN-AML, n=185
ANP32A expression, high vs. low	1.60	1.1–2.3	0.0088		1.58	1.1–2.3	0.016
Risk, poor vs. non-poor	2.12	0.49–9.2	0.32		3.50	0.81–15.2	0.09
Age, per 10-y increase	1.04	0.90–1.2	0.58		1.08	0.93–1.3	0.32
Gender, male vs. female	1.18	0.83–1.7	0.34		1.37	0.95–1.9	0.1
FLT3-ITD, presented vs. others	1.64	1.1–2.4	0.009		1.59	1.1–2.3	0.019
N-ras, mutated vs. wild type	1.66	0.88–3.1	0.12		1.24	0.63–2.4	0.54
ELN Intermediate-I, n=115
ANP32A expression, high vs. low	1.57	1.02–2.4	0.041		1.59	0.99–2.5	0.051
Risk, poor vs. non-poor	1.43	0.31–6.6	0.64		1.92	0.42–8.7	0.4
Age, per 10-y increase	1.15	0.96–1.4	0.12		1.19	0.99–1.4	0.07
Gender, male vs. female	1.17	0.77–1.8	0.46		1.45	0.92–2.3	0.11
FLT3-ITD, presented vs. others	1.17	0.74–1.9	0.5		1.10	0.68–1.8	0.69
N-ras, mutated vs. wild type	1.59	0.62–4.0	0.33		1.43	0.56–3.6	0.45

EFS, event free survival; OS, overall survival; CN-AML, cytogenetically normal acute myeloid leukemia.

Table S1

Multivariable analysis of OS and EFS in 435 AML and 225 NCCN intermediate patients

Variables in final model by end point	EFS				OS
	HR	95% CI	P value		HR	95% CI	P value
All AML, n=435
ANP32A expression, high vs. low	1.09	0.90–1.32	0.358		1.14	0.93–1.39	0.201
Risk, poor vs. non-poor	2.24	1.71–2.93	5.54E-09		2.16	1.64–2.86	5.77E-08
Age, per 10-y increase	1.07	0.98–1.18	0.140		1.14	1.04–1.26	0.008
FLT3-ITD, presented vs. others	1.66	1.27–2.15	0.0002		1.73	1.32–2.27	8.32E-05
Gender, male vs. female	0.94	0.75–1.18	0.601		1.07	0.84–1.35	0.581
N-ras, mutated vs. wild type	1.20	0.83–1.74	0.340		1.06	0.70–1.58	0.796
NCCN intermediate, n=255
ANP32A expression, high vs. low	1.16	0.92–1.46	0.208		1.25	0.98–1.60	0.069
Risk, poor vs. non-poor	1.07	0.86–1.33	0.535		1.13	0.90–1.43	0.289
Age, per 10-y increase	1.19	0.71–2.01	0.507		1.17	0.68–2.02	0.572
FLT3-ITD, presented vs. others	1.42	0.83–2.41	0.201		1.46	0.83–2.56	0.188
Gender, male vs. female	0.87	0.50–1.51	0.613		0.93	0.52–1.66	0.803
N-ras, mutated vs. wild type	NA	NA	NA		NA	NA	NA

EFS, event free survival; OS, overall survival; AML, acute myeloid leukemia.

Figure S2

The validation of the unfavorable outcome associated with ANP32Ahigh in a cohort of 232 CN-AML patients.

Genome-wide gene-expression profiles associated with ANP32A expression

In order to further investigate the biological value of ANP32A in leukemogenesis, we derived ANP32A-associated gene expression profiles from genome-wide microarray analysis in 73 CN-AML patients from TCGA dataset; 1,401 up-regulated and 543 down-regulated genes were found to be significantly associated with ANP32A group (FDR-adjusted P<0.001, ). The up-regulated genes included: (I) genes encoded transcription factor proteins: MYB, PHF10 (32); (II) independent unfavorable prognostic factor in CN-AML: ERG (33); (III) gene encoded cohesion complex: SMC3 (34); (IV) gene involved in phosphatases: PTPN11 (34); (V) genes regulated cell growth and signal transduction pathways: ANXA family genes (35); (VI) genes promoted tumorigenesis: CDK4 (36); ZPF91, involved in leukemogenesis and could activate HIF1a via NF-kB to promote tumorigenesis (37); SRSF family genes (SRSF1, SRSF3, SRSF10), which were oncogene and overexpressed in multiple cancers to promote cell proliferation and transformation (38); (VII) gene related to ribosome biogenesis: NPM1 (39); (VIII) gene involved in fine-tuning hematopoietic stem and progenitor cell homeostasis: PHF6 (40). The down-regulated genes included: (I) gene inactivated in solid tumor and acute lymphoblastic leukemia (ALL): CREBBP (41); (II) gene improved tumor immunogenicity and enhanced Fas-induced apoptosis: TNFRSF14 (42); (III) genes interacted with NAD metabolism and p53 signaling: SIRT family genes (SIRT2, SIRT7) (43); (IV) genes interacted with WNT-pathway and B-cell receptor signaling mediators: IRF9, BCL9L, SIT1 (44).

Figure 3

Genome-wide genes/cell signaling pathways associated with ANP32A expression. (A) Volcano plot of genome-wide gene profiles between ANP32A group and ANP32A group; (B) gene expression heatmap. The top curve indicates expression distribution of ANP32A in 73 CN-AML samples; (C) cell signaling pathways associated with ANP32A expression.

There is evidence that the uncovering of unique gene expression profiles with the identification of dysregulated signaling pathways is now providing new insights into the leukemogenesis of CN-AML (45). Therefore, data from MSigDB was used to investigate cell signaling pathways associated with ANP32A (46). A total of 6 down-regulated and 22 up-regulated pathways were found to be significantly associated with ANP32A group (P<0.05, ). The up-regulated pathways included: (I) important molecular and biological pathways, such as RNA transport, ribosome biogenesis in eukaryotes, spliceosome, ribosome, DNA replication, protein export, proteasome, cell cycle, and RNA polymerase; (II) metabolic pathways of proteins, amino acid and nucleic acid, such as basal transcription factors, carbon metabolism, pyrimidine metabolism, purine metabolism, and valine, leucine, isoleucine degradation; (III) lipid metabolic pathways, such as fatty acid degradation and fatty acid metabolism. On the contrary, the down-regulated pathways were mostly immune-related ones, such as T and B cell receptor signaling pathway, TNF signaling pathway, Toll-like receptor signaling pathway. These dysregulated signaling pathways might explain the involvement between ANP32A and the leukemogenesis of CN-AML.

Genome-wide microRNA profiles associated with ANP32A expression

MicroRNAs have been increasingly used to diagnose and assess prognosis in hematopoietic malignancy. Dysregulated microRNAs could promote aggressive tumor phenotype, identify risk classification, and reflect therapeutic effects (47). The genome-wide microRNA profiles were analyzed and 61 microRNAs were identified to be significantly correlated to ANP32A expression, including 51 positive and 10 negative microRNAs (P<0.05, ). Positively correlated microRNAs included miR-17, miR-19b-1, miR-20a, miR-25, miR-106a, miR-106b, miR-125b-1, miR-99a, miR-301b, miR-199b, miR-191, miR-629, etc. MiR-17, miR-19b-1 and miR-20a, the members of miR-17-92 cluster, were found to be upregulated in AML and helped apoptotic resistance in K562 cells (48). MiR-25 had statistically higher expression in c-kit subgroup and might be involved in leukemogenesis in pediatric AML (49). MiR-106a and miR-106b, paralogs of miR-17-92 cluster, were found to enhance proliferation of T-ALL cells and significantly upregulated in relapsed pediatric MLL-AML, respectively (50). By targeting genes associated with tumorigenesis, miR-125b-1 might act as an onco-microRNA and its overexpression alone in mice could induce tumors in multiple hematopoietic lineages (51). MiR-99a upregulation was proven to be associated with poor prognosis of AML, leading to cell expansion and progression of AML (52). MiR-301b and miR-199b were found upregulated in AML patients at diagnosis and trended to normalization after chemotherapy (53). Overexpression of miR-191 had significantly worse OS and EFS than those with low expression (54). MiR-629 was found upregulated in pediatric T-ALL, and might be used for distinguishing pediatric ALL subtypes (55). Negatively correlated microRNAs included miR-24-2, miR-26b, miR-197. MiR-24-2 was proved to be down-regulated in acute leukemia cell lines compared to hematopoietic stem progenitor cells (56). Decreased expression of miR-26b was found in T-ALL cells and it functioned as a tumor suppressor in T-ALL. Furthermore, it could inhibit the PI3K/AKT pathway, reduce proliferation and promote apoptosis of T-ALL cells (57). Overexpression of miR-197 inhibited tumor growth and prolonged OS in subcutaneous mouse xenograft model in multiple myeloma, revealing its novel role as tumor suppressor (58). These results indicated that microRNA expression might link to unfavorable outcome in ANP32A group.

Figure 4

Interaction network analysis of genome-wide microRNAs and their target genes associated with ANP32A expression. (A) Volcano plot of genome-wide microRNAs profiles between ANP32A group and ANP32A group; (B) microRNAs expression heatmap. The top curve indicates expression distribution of ANP32A in 73 CN-AML samples; (C) microRNA-mRNA interaction network analysis. (red: up-regulation; green: down-regulation; rectangle: microRNAs; ellipse: mRNAs).

To further validate the results of the microarray platform and microRNAs, interaction network analysis was used to explore the interaction and correlation between microRNAs and their target genes (). The results showed the tumor suppressors repressed by the up-regulated microRNAs, led to worse outcome. For example, SIRT7, targeted by miR-629, possessed tumor-suppressing properties in breast cancer, pancreatic cancer, head and neck squamous cell carcinoma, etc. (59). KLF13, targeted by miR-106b, inhibited the transcription of MYB and BCL2 in ALL (60). Additionally, many oncogenes were the targets of the down-regulated microRNAs. SUGT1, targeted by miR-24-2, contributed to cancer development by stabilizing oncoproteins (61). FBXO45, targeted by miR-26b, exerted anti-apoptotic effect by interfering the proteasome-dependent degradation of p73, a known pro-apoptotic factor (62). These results suggested why ANP32A acted as an unfavorable prognosticator and could help provide a comprehensive view of the molecular mechanisms of ANP32A.

Genome-wide methylation profiles associated with ANP32A expression

Differential methylated regions (DMR) were further analyzed to uncover the different methylation patterns between ANP32A group and ANP32A group of CN-AML. A total of 733 hypomethylation and 31 hypermethylation DMRs were discovered from the comparison between ANP32A group and ANP32A group (P<0.05, |Log2(FC)| >1, ). In addition, position distributions around CpG islands were compared in these aberrant DMRs. It showed that significantly more hypomethylation DMRs were found in the CpG islands (P=0.03, ). Furthermore, position distributions of different structural fragments of genes were analyzed. The results indicated relatively more hypomethylation DMRs lied in 1st Exon regions (P=0.08, ). Finally, the heatmaps showed ANP32A-associated DMRs enriched on islands and gene promoter regions ().

Figure 5

Genome-wide methylation profiles associated with ANP32A expression. (A) A total of 733 hypomethylation and 31 hypermethylation differential methylated regions were discovered from the comparison between ANP32A group and ANP32A group; (B) more hypomethylation differential methylated regions were found in the CpG islands and more hypomethylation differential methylated regions lied in 1st Exon regions; (C,D) the heatmaps showed ANP32A-associated differential methylated regions enriched on islands and gene promoter regions.

Discussion

Recent study provided strong evidence that ANP32A was a key factor to deregulate histone H3 acetylation, and played a role as an oncogene in leukemogenesis (23). It could interact with DNA-binding transcription factors, thus influencing gene expression (63). ANP32A acted as an oncogene in certain solid tumors, such as colorectal cancer, liver cancer, ovarian cancer, etc. (19,64). Besides, study showed ANP32A might repress ERK and subsequently inhibit RUNX1 and FLI1 to promote megakaryocyte differentiation in acute megakaryoblastic leukemia (65). In our study, we provided direct evidence that high expression of ANP32A predicted unfavorable outcomes for CN-AML. Firstly, high expression of ANP32A was shown in AML BM and PB than in normal BM and PB samples, as well as in CN-AML BM than in normal BM samples (). Secondly, survival prognosis with gene expression were well studied previously (66,67) as well as the various functionality of omics data (68-73). After validating in relatively large independent CN-AML cohort, high expression of ANP32A was shown to associate with unfavorable outcomes. Similar results were also shown in ELN Intermediate-I cohort, NCCN-Intermediate Risk cohort, and non-M3 AML cohort, all containing a variety of karyotypes of AML. These results were consistent with earlier findings of ANP32A, suggesting its potential role of a prognosticator and therapeutic target of AML. Furthermore, our results could provide evidence for further fine stratification of AML, especially in CN-AML, ELN Intermediate-I, NCCN-Intermediate Risk patients.

It is at present unclear about the pathogenesis of AML. However, previous studies showed the molecular biology of leukemia and identified favorable or unfavorable prognosticators, facilitating the understanding of leukemogenesis. In our study, associations between ANP32A expression and previously known prognosticators were explored in a cohort of 185 CN-AML patients. We found that ANP32A group contained significantly more M5 (P=0.004) and less M2 (P=0.007) FAB subtype, suggesting that ANP32A expressers carried more mature cells. Furthermore, ANP32A group had significantly more FLT3-ITD mutation (P=0.029), and remarkably high expression of MAPKBP1 (P=0.015), RUNX1 (P=0.0003) and TCF4 (P=0.033). all of which were known independently adverse prognosticators in AML (25). Besides, ANP32A group had more relapse patients than ANP32A group (P=0.021). These results indicated that ANP32A might be another independent biomarker of unfavorable outcome for CN-AML.

There is a growing number of evidences demonstrated that aberrant gene expression, microRNA post-transcriptional regulation and DNA methylation played significant role in leukemogenesis (74-77). Our study further discussed how overexpression of ANP32A affected prognosis from these three aspects. Firstly, we investigated genome-wide genes and cell signaling pathways that associated with ANP32A. Genes promoted tumorigenesis were up-regulated, such as CDK4, ZPF91, SRSF family genes, as well as ERG, an independent unfavorable prognosticator in CN-AML. On the contrary, genes inactivated in tumor and immunogenicity were down-regulated, such as CREBBP and TNFRSF14, respectively. Secondly, cell signaling pathway analysis results were in accordance with ANP32A’s unfavorable prognosticator role. Results showed up-regulation of cell cycle pathway, lipid metabolic pathway, metabolic pathways of proteins, amino acid and nucleic acid as well as down-regulation of immune-related pathways of T cell, B cell and Toll-like receptor signaling pathway and TNF signaling pathway. All these results further validated the previous study that ANP32A was a member in nuclear protein family implicated in multiple cellular pathways (78).

Furthermore, in our study, ANP32A was found to correlate positively or negatively to microRNAs, such as members of miR-17-92 cluster and miR-24-2, which played novel role in leukemogenesis or tumor suppression, respectively. Also, synchronized changes of microRNA-mRNA pairs along with ANP32A expression might contribute to the aggravation of CN-AML malignancy. For instance, the pair of positive correlated miR-629 and down regulated SIRT7, the pair of negative correlated miR-26b and up regulated FBXO45, etc. These genes, microRNAs, pathways, microRNA-mRNA pairs might be attributable to the adverse outcome of CN-AML.

Finally, genome-wide methylation profiles associated with ANP32A expression were analyzed. Results showed that more hypomethylation DMRs lied around CpG islands and in 1st Exon regions. Hypomethylation in different genic regions had been proven to exhibit a significant adverse effect on gene expression, resulting in malignant transformation and tumor progression (79). For example, in chronic lymphocytic leukemia (CLL), epigenomic analysis detected widespread DNA hypomethylation in the gene body, implicating functional and clinical roles of DNA hypomethylation in leukemogenesis (80). The hypomethylation DMRs detected in our study might possibly contribute to the unfavorable prognosis and provide deeper understanding in leukemogenesis.

In summary, we had performed a large-scale analysis of ANP32A in CN-AML. Combined with recent study that ANP32A could promote leukemogenesis, we further validated that overexpression of ANP32A was an unfavorable biomarker for CN-AML. We provided evidence that ANP32A patients had different kinds of adverse molecular characteristics. Furthermore, we highlighted the power of multi-omics analysis to provide new insights into the leukemogenic mechanisms of CN-AML and define their clinical implications.

Conclusions

We proved ANP32A is a novel, potential unfavorable prognosticator and therapeutic target for AML.

Median value of ANP32A expression as the cut-off. (A) Normally distribution of ANP32A expression. (B) OS and (C) EFS of CN-AML patients, the patients were subdivided into four quartiles based on the quartile of ANP32A expression.