Targeting aurora kinases as a potential prognostic and therapeutical biomarkers in pediatric acute lymphoblastic leukaemia

Aurora kinases (AURKA and AURKB) are mitotic kinases with an important role in the regulation of several mitotic events, and in hematological malignancies, AURKA and AURKB hyperexpression are found in patients with cytogenetic abnormalities presenting a unfavorable prognosis. The aim of this study was evaluated the mRNA expression profile of pediatric Acute Lymphoblastic Leukaemia (ALL) patients and the efficacy of two AURKA and AURKB designed inhibitors (GW809897X and GW806742X) in a leukemia cell line as a potential novel therapy for ALL patients. Cellular experiments demonstrated that both inhibitors induced cell death with caspase activation and cell cycle arrest, however only the GW806742X inhibitor decreased with more efficacy AURKA and AURKB expression in K-562 leukemia cells. In ALL patients both AURKA and AURKB showed a significant overexpression, when compared to health controls. Moreover, AURKB expression level was significant higher than AURKA in patients, and predicted a poorer prognosis with significantly lower survival rates. No differences were found in AURKA and AURKB expression between gene fusions, immunophenotypic groups, white blood cells count, gender or age. In summary, the results in this study indicates that the AURKA and AURKB overexpression are important findings in pediatric ALL, and designed inhibitor, GW806742X tested in vitro were able to effectively inhibit the gene expression of both aurora kinases and induce apoptosis in K-562 cells, however our data clearly shown that AURKB proves to be a singular finding and potential prognostic biomarker that may be used as a promising therapeutic target to those patients.

www.nature.com/scientificreports/ The aurora kinase family are constituted of three serine/threonine kinases (AURKA, AURKB and AURKC) that acts like mitotic kinases with an important role in the regulation of the G2 / M phase of the cell cycle and several mitotic events, including centrosome duplication, mitotic spindle formation, chromosomal segregation, and cytokinesis occurring at the end of telophase, events that are essential for cell division, pointing out the importance of studying the role of these genes and their involvement in the maintenance mechanism of cell cycle stability 5,6 .
Overexpression of AURKA and AURKB genes can disrupt the normal development of cell division, increasing genetic instability, triggering the development of tumors. This abnormal expression have been well characterized in several types of aggressive cancers [7][8][9][10] .
The aim of this study was evaluate the AURKA and AURKB mRNA expression profile of pediatric ALL patients, to construct a protein-protein interaction network to evaluate the possible role of those targets in leukemogenesis pathway and the efficacy of two designed aurora kinase inhibitors in a leukemia cell line as a potential novel treatment to ALL pediatric patients.

Results
Non-selective Aurora Kinase inhibitors decreased cell proliferation provoking cell death and cell cycle progression modulating AURKA, AURKB and BCR-ABL1 gene expression. Firstly, it was proved that non-selective AURKA and AURKB inhibitors GW809897X and GW806742X treatment reduced cell proliferation of leukemia cell line (K-562) with a potency constant IC50 of > 5 µM and 1.47 µM, respectively. In order to determine whether cell proliferation reduction was due to cell death process, we performed caspase 3 and 7 activity and morphological changes analysis by flow cytometry.
Results shown in Fig. 1 demonstrated that 1 µM GW809897X and GW806742X inhibitors significantly increased the number of shrunk cells (P < 0.05) which indicate a possible early stage of cell death. Caspase 3 and 7 activation was also observed after the treatment (P < 0.0001), which confirm the apoptosis as a cell death pathway (Fig. 1A,B).
Cell cycle distribution also showed that 1 µM of both aurora kinase inhibitors induced significant G2/M phase arrest (P < 0.001) in k562 cell line (Fig. 1C). Compared with control group, the K-562 exposure to GW809897X and GW806742X lead to the accumulation of cells in the G2/M phase from 16.20 ± 4.24 to 33.20 ± 1.41 and 37.50 ± 0.70, respectively (Fig. 1C). Additionally, GW809897X and GW806742X also significant induced a decreased of cells in G0/G1 (P < 0.05 and P < 0.01). These data proves that GW809897X and GW806742X inhibits the cellular proliferation through G2/M cell cycle arrest.

Acute lymphoblastic leukemia patients clinical features. As shown in
Results demonstrated significant differences between age groups considering the type of translocation (P < 0.0001). In these, 25% of patients with less than 1-year-old were positive for BCR-ABL, 29.2% patients with 1-9 years-old were positive for E2A-PBX1 and 60.7% of patients with ≥ 10 years-old were positive for others types of chromosomal translocations.
AURKA and AURKB mRNA expression in ALL patients. Since the leukemia cell line K-562 was chemosensitive for both aurora kinase inhibitors (GW809897X and GW806742X), we decided to measure the AURKA and AURKB mRNA expression in the ALL patient's blood cells. AURKA was 6.20 fold change overexpressed in ALL samples compared to control samples (P < 0.05), while AURKB was 15.32 fold change overexpressed (P < 0.0001). Furthermore, the results demonstrated that the AURKB expression level was significantly higher than AURKA in ALL patients (P < 0.0001), Fig. 2.
The next question was whether the AURKA and AURKB gene expression levels differ among the clinical features of ALL patients. As shown in Fig. 3, chromosomal translocation groups (BCR-ABL1, E2A-PBX1, MLL-AF4, SIL-TAL, TEL-AML1, and non-identified chromosomal translocation) were not different in terms of AURKA (P = 0.416) and AURKB expression (P = 0.948), as well as the immunophenotypic groups (biphenotypic, T-cell ALL, B-cell ALL), with no significant differences observed for AURKA (P = 0.656) and AURKB (P = 0.404). White Blood Cells (WBC) count groups were also analyzed and no significant differences were found for AURKA (P = 0.390) and AURKB (P = 0.687) expression. Moreover, no correlation was observed among WBC count and AURKA (Spearman r = −0.097) and AURKB (Spearman r = −0.062).  Fig. 4, results showed that AURKA (P = 0.078) and AURKB (P = 0.880) expression were not different between male and female patients, as well as in the age groups which AURKA (P = 0.705) and AURKB (P = 0.635) expression levels were equally distributed.

Patients survival rate.
We also performed a patients survival overall rate to evaluate the role of AURKA and AURKB gene expression levels and its potential association with prognosis. A total of 64 patients were   www.nature.com/scientificreports/ included, and the median time of follow-up was 22.9 months (range 5-56 months).The data was normalized and categorized assuming Log2 expression levels, for AURKA the expression levels was 1,fivefold and for AURKB fivefold. In comparison between both genes, AURKA did not had any impact at the patients overall survival (P = 0.17). However, AURKB expression levels predicted a poorer prognosis with significantly lower survival rates (P < 0.0001), as shown in Fig. 5.

Protein-protein interaction network analysis and co-expression.
To identify the high-trust hub genes, we inserted aurora kinases AURKA, AURKB into the STRING database. A Protein-Protein Interaction Network (PPI) was generated with a confidence score of 0.70 composed of 102 nodes and 2751 edges. All PPI network was analyzed using MCODE. Only the best module was chosen with a score of 53.47 (84 knots and 2219 edges). Hub genes were identified from the results of combined analysis between MCODE and CytoHubba. The PPI network had an enrichment of 1.0E−16, indicating that the proteins of these genes are biologically linked as a group. The first 20 genes identified with the MCC method were chosen by CytoHubba : AURKA,  DLGAP5, CCNB1, KIF11, NDC80, CENPE, CCNA2, BUB1B, CCNB2, MAD2L1, PLK1, TPX2, BIRC5, BUB1, AURKB, CDC20, CDK8 , PBK. Furthermore, GO enrichment analysis showed these genes strongly associated in biological process, cellular component and molecular function, respectively for mitotic cell cycle (GO: 0000278), spindle (GO: 0005819), protein serine / threonine kinase activity (GO: 0004674), as shown in Fig. 6. The most enriched KEGG pathway was identified for cell cycle (hsa04110).

Discussion
The advances achieved in the treatment of ALL during the last decades are a successful model in the practice of modern medicine based on translational research and clinical trials 1 . Despite the increase in cure rates, exceeding 80%, up to a quarter of the patients still present a relapse, which leads to a poor prognosis, leading to death, where there is still a high frequency rate 11 . These events are especially observed in patients who present genomic alterations that compromise the effective treatment of available chemotherapy protocols 12,13 . In this study, expression of AURKA and AURKB in ALL pediatric patients was evaluated, as well as, the efficacy of two new potential aurora kinase inhibitors as therapeutic options. The chromosomal translocations identified in the 104 patients analyzed were E2A-PBX1 (20.19%), BCR-ABL1 (14.42%), TEL-AML1 (8.65%), SIL-TAL (3.8%), MLL-AF4 (5.8%), these gene alterations were previously described in other studies in Brazilian ALL pediatric population [14][15][16][17] . However, in our study it was noted a higher prevalence, among the identified gene fusions, the types E2A-PBX1 and BCR-ABL. Is widely known in the literature that gene fusions as E2A-PBX1, BCR-ABL1 appears to have the worst outcome and prognosis compared to the others types, pointing an emerging need to new therapeutic approaches and targets specially to these patients [18][19][20][21][22][23] .
The ALL pediatric patients analyzed in this study showed overexpression of both, AURKA and AURKB genes, showing that these genes seems to have a prominent importance in this disease model, regardless of clinical   www.nature.com/scientificreports/ findings and variables of risk management. However, the AURKB stands out as significantly relevant when compared to AURKA. This is the first study in Brazil to show AURKA and AURKB expression data in pediatric patients with acute lymphoid leukemia, as well as to present them in comparison to clinical parameters. It was important to notice that no clinical features, presented by patients, was able to differentially influence analysis of AURKA and AURKB gene expression, showing the importance of aurora kinases as biomarkers in pediatric ALL pathogenesis. The overall survival rate demonstrated that AURKA did not show significant results at the patient's overall survival (P = 0.17). Nevertheless, AURKB expression levels predicted a poorer prognosis with significantly lower survival rates (P < 0.0001). In hematological malignancies, AURKA and AURKB overexpression are found in patients with cytogenetic abnormalities that are unfavorable to the prognosis and which compromise patients' survival 24 . These changes have been well described in acute myeloid leukemia 25,26 , Myelodysplastic Syndrome 27-29 , Chronic Myeloid Leukemia 30 and also in the ALL 31 .
In K-562 cells, both designed inhibitors, GW809897X and GW806742X, reduced cell proliferation, cell cycle progression and potentially induced apoptosis. However, only GW806742X inhibitor was able to strongly reduced AURKA and AURKB activity, and the molecular signature, of K-562 cells, BCR-ABL as well. As described in the literature, the overall rates and prognosis of ALL patients with BCR-ABL (Ph+) are very poor 32,33 , and aurora kinase inhibitors provided a new option for future clinical and therapeutic options for these individuals 34 . Ikezoe et al. (2007) showed that the K-562 cell line, from a diverse panel of 15 leukaemia strains, presented the highest expression of the AURKA and AURKB, proving the efficacy of using this particular experimental model to study potential inhibitors against these targets, as shown in this study 35 .
Cells with overexpression of AURKA overlap mitotic spindle control and enter anaphase despite abnormal spindle formation 36 . The overexpression of an inactive form of AURKB in cells would also compromise cell cycle control points and spindle formation because AURKB activity is required for recruitment of cycle check proteins 24,37 . Studies have shown that in murine models the overexpression of AURKB induce tetraploidy 38,39 , and cytokinesis failure in the absence of proper functioning of AURKB activity is a good explanation of why downregulation causes an increase in ploidy, and many studies supported the classification of AURKB as a cancer promoting gene 40,41 .
Cell cycle proteins have been show as potential anticancer targets 42 . To identify AURKA and AURKB interaction pathways in ALL pathogenesis, a PPI network was constructed, which could evidence the presence of both, and their participation in protein expression modulation of essential pathways for cellular organization during division, highlighting the mitotic cell cycle pathway and spindle control that appears with high correlation in this interaction pathway, demonstrating the importance of these proteins in mitotic organization. It's important to emphasize that further studies to comprehend the role of AURKA and AURKB modulation in mitotic cell cycle pathway and spindle control might be performed for a better understanding of AURKA and AURKB in this scenario of ALL pathogenesis, which are highly related to key activities in the pathogenesis and origin of chromosomal translocations that are the molecular signature presented in this disease [43][44][45] .
The presence of AURKB overexpression in ALL patients described in this study, proves to be an important target to inhibition by specific target molecules, as GW806742X. It was shown, that AURKB inhibitors interferes with normal chromosome alignment during mitosis and induces endoreduplication, leading cells to death through catastrophic mitosis, becoming a suitable anticancer strategy 37,46,47 . In specific cases, inhibitors of disease-related aurora-kinases have been used experimentally with some success and mark a major advance in the treatment of patients with ALL 6,48 .
The aurora kinase inhibitors tested in vitro, were able to effectively induce apoptosis of K562 cells, but only GW806742X was able to inhibit the gene expression of the AURKA and AURKB as well. The results indicate that the overexpression of the AURKA and AURKB genes is an important finding in childhood ALL in this study, despite the clinical features. However is clearly shown that AURKB proves to be a singular molecular finding in this population, strongly supported by its expression associated with poorer survival rates, pointing out that AURKB may be used potentially as a prognostic biomarker and therapeutic target for pediatric ALL patients.

Material and methods
Patients and samples. A total of 104 ALL pediatric patients were included in this study, all diagnosed at Octávio Lobo Children's Hospital (Belém-Brazil) according to the French-American British (FAB) criteria 49 . The clinical data from patients were analyzed based on risk-stratification criteria of Berlin-Franklin-Münster (BFM) 50 . Blood samples were collected at the time of diagnosis and samples from 40 healthy volunteers were used as healthy control. The study was approved by the Ethics Committee of the Ophir Loyola Hospital (approval number: 119.649), informed written assent was obtained from the patient's legal guardians and all methods were carried out in accordance with Helsinki guidelines and regulations.
Cell culture. Leukemic cell line K-562 were kindly provided by Dr. Vivian Rumjanek from Federal University of Rio de Janeiro. Cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37 °C in a 5% CO 2 air-humidified atmosphere at 37ºC and cell confluence was observed in conventional microscope.
AURKA and AURKB drug inhibitors. Aurora Kinase (AURKA and AURKB) inhibitors GW809897X and GW806742X, belonging to the chemotype 16:2,4-diamino-pyrimidines, was obtained from a panel provided by Dr. Bill Zuercher from University of North Carolina and the Structural Genomics Consortium. Both inhibitors were well characterized as a molecular probe for targeted therapy studies by Elkins and co-workers (2016) 13 52 . The gene expression levels were based on absolute and relative analyses and calculated using the 2-ΔΔCT (delta-delta threshold cycle) method, the expression level of the gene of interest is reported relative to the reference gene for each sample 53 . PPI network building and module analysis. AURKA and AURKB genes were submitted to the STRING Protein Data Retrieval Search Tool (v11.0) (http://strin g-db.org/) 54 . The confidence score was 0.70, with the other parameters being used as standard. The maximum number of 100 interactions has been tested. The Molecular Complex Detection (MCODE) application was used to analyze the PPI (Protein-Protein Interaction) network module and MCODE scores > 50 and the number of nodes > 80 were set as cutoff criteria with the default parameters (Degree cutoff ≥ 2, Node score cutoff ≥ 2, K-core ≥ 2 and Max depth = 100) 55 . Finally, the CytoHubba plugin was used to explore PPI network hub genes with the Maximal Click Centrality (MCC) metric for better PPI network performance 56 . Program STRING enrichment was used for GO (Gene Ontology) and KEGG enrichment, and Cytoscape (v3.7.1) was used to view PPI and co-expression networks 57 .

Statistical analysis.
Three independent experiments were performed in triplicate. All data were expressed as Mean or median ± dispersion measures depending on the normality of samples. For it, the Shapiro-Wilk test was applied to determine if samples followed a normal distribution. Comparison tests were performed to compare two or more different groups using Analysis of Variance (ANOVA), Bonferroni posttest and t-test or their corresponding non-parametric tests. Frequency data were analyzed by Chi-square test and correlation analysis was performed using the Spearman test. To predict the overall survival rate from patients log-rank test was calculated from the date of diagnosis to the date of mortality or last follow-up, using R package software 58 . Significant differences were determined by setting a significant level in P < 0.05 (confidence interval of 95%).