Prognostic role and treatment tips of FHL1 expression in cytogenetically intermediate- and poor-risk acute myeloid leukemia CURRENT STATUS: POSTED

Background: Acute myeloid leukemia (AML) is a heterogeneous disorder of hematopoietic system, 35 to 40% of younger patients (<60 years) are recurrent, refractory or drug-resistant, which are canonically called as cytogenetically intermediate- and poor-risk AML (IP-AML). Identifying novel biomarkers is an urgent clinical need and research hotspot for treatment optimization. Methods: Bioinformatics analysis were carried out for RNA-Seq data derived from drug resistance-associated cell lines and The Cancer Genome Atlas (TCGA). Varieties of comparison, visualization and functional enrichments were performed to Four-and-a-half LIM domain 1 (FHL1) and related genes. FHL1's expression changes were confirmed by ‘wet’ experiments. Kaplan-Meier method, log-rank test and multivariate Cox proportional hazards models were utilized to evaluate the associations between FHL1 expression and Overall Survival (OS), Event Free Survival (EFS), Relapse Free Survival (RFS). Prognostic significance of FHL1 expression was further validated in another independent larger IP-AML cohort (GSE6891). Moreover, construction of nomogram and validation of prognostic model. Results: High expression of FHL1 (FHL1 high ) was a potentially effective biomarker of poor prognosis for IP-AML. Compared to FHL1 low group, FHL1 high was associated with short OS and EFS (145 patients, OS, P < 0.001; EFS, P < 0.001), which was further validated in GSE6891 (284 patients, OS, P < 0.001; EFS, P < 0.001). Multivariate analysis also confirmed the adverse prognosis of FHL1 high (HR = 2.2339, P = 0.000156). ROC indicated an ideal predictive accuracy of the outcome model (AUC was 0.773). In addition, to understand the inherent mechanisms FHL1 involved, genome-wide characteristics were investigated to find that FHL1 might be involved in several important carcinogenic signal pathways.


Background
AML is a heterogeneous malignancy characterized by clonal expansion, infiltration of the bone marrow, blood and differentiation arrest of myeloid progenitors at various stages of incomplete maturation [1]. A major cause of treatment failure in patients with AML is the occurrence of multidrug resistance, which can can arise during chemotherapy or at relapse [2]. Recurrent and refractory disease status is an obstacle [1]. Especially, dysregulation of transcription mechanisms related to the pattern of mutation acquisition can affect apoptosis susceptibility and lead to drug resistance [3].
Drug resistance mechanism is unique to individuals drugs, different patients, and clinical risk factors such as white blood cell count, age, and other factors, which have been reported to affect drug resistance singly or in combination [4]. Especially, the treatment choice based on risk stratification, and targeted therapy may become less toxic and more effective when well defined genes are available [4]. Studies have already demonstrated several crucial molecule factors associated with favorable or adverse outcomes, which includes a variety of cytogenetic changes, such as mutations of TP53 [5], and/or ASXL1 [6], double CEBPA mutation [7], FLT3-ITD [8] and high expression of RUNX1 [9], as well as low expression of NPM1 [10]. These biomarkers can be useful indicators for risk stratification, but not cover all the AML patients. What's more, some genes that commonly involved in malignancy genetic regulation can not only make cells multilineage differentiation, but also help them to survive chemotherapy and expand during remission. These processes could eventually facilitate reducing relapse and curing the disease. Therefore, it is necessary in order to more thoroughly understand the underlying mechanism and to identify more target genes as potential biomarkers for the individualized IP-AML diagnosis and treatment. FHL1 belongs to the FHL protein family, which are related to carcinogenesis and composed of four members, FHL1, FHL2, FHL3, and FHL5 in humans [11,12]. Though some studies found FHL1 played important roles in embryonic development [13], suppression of cell differentiation [14] and oncogenesis [15], But, up to now, the role of FHL1 in clinical prognosis, risk stratification of IP-AML remains unclear [16].
This study confirmed the prognostic value of FHL1 expression in IP-AML patients, further explored the distinctive genomic patterns associated with FHL1 expression and explored the potential mechanisms resolving FHL1 via combining 'wet' experiment and bioinformatics analysis in IP-AML patients. Results provide direct evidence for using FHL1 expression as a new prognostic biomarker in risk stratification or a potential target for the precision diagnosis and treatment of IP-AML.

Cell lines and cell culture
The human leukemia cell line K562, and its anthracycline-resistant counterpart K562/A02 were obtained from Tianjin Institute of Hematology. Cells were all cultured in RPMI-1640 medium supplemented with 10% fetal calf serum (Gibco BRL, Grand Island, USA) in a humidified atmosphere containing 5% CO 2 at 37 ˚C. To maintain the multidrug resistance phenotype, doxorubicin was added to the culture media for K562/A02 cells at the final concentration of 0.5 µg/mL. The cells were cultured for 2 weeks in drug-free medium prior to their use in the experiments.

Patients and treatment
The first cohort was derived from TCGA (https://tcga-data.nci.nih.gov/tcga/) which was used as the training set, including 145 IP-AML in the 179 clinically annotated adult de novo AML patients. The validation data set included 284 IP-AML patients derived GSE6891 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse6891), inlucding all the clinical, cytogenetic and molecular information. All these data were publicly accessible from the TCGA and GEO website.
Some molecule markers were available about the genetic risk of FHL1 high and FHL1 low groups, which was referred to NCCN(National Comprehensive Cancer Network)Acute Myeloid Leukemia Risk Status Guidelines Version 1.2018.
RNA analyses and gene expression by quantitative real-time reverse transcription polymerase chain reaction Total RNA from 3 K562/AO2 cell lines and 3 K562 cell lines was extracted with TRIZOL reagent (Invitrogen, Carlsbad, USA) and cDNA was synthesized by PrimeScript™ RT reagent Kit (Takara) according to the manufacturer's instructions. The quantitative real-time polymerase chain reactions were then performed using KAPA SYBR FAST q-PCR Master Mix (2x) Kit. The following primers were used for quantitative PCR: FHL1, forward primer: 5' -TGCTGCCTGAAATGCTTTGAC − 3' and reverse primer: 5'-GCCAGAAGCGGTTCTTATAGTG − 3'. We used the 2 −ΔCt formula to examine the relative quantification of the target genes.

Definition of clinical end points
Descriptive statistics including frequency counts, median and range were appointed to describe patient characteristics, which were publicly accessible from the TCGA website. The main objective of this study was to explore the prognostic value of FHL1 expression in AML patients.
OS was defined as the time from date of diagnosis to death due to any cause or the last follow-up. EFS was the time from date of diagnosis to removal from the study because of the absence of complete remission, relapse or death. RFS was defined as the time from the date of diagnosis to removal from the study due to relapse.
To determine the best classification method, we subdivided AML patients into FHL1 high and FHL1 low groups according to the median based on FHL1 expression values. Patients with higher than median FHL1 expression values were classified as FHL1 high group, and those with lower were classified as FHL1 low group. High FLT3, NPM1, CEBPA, MLL-PTD, IDH1, IDH2, RUNX1, DNMT3A, TP53, ERG, BAALC, MN1, WT1, FLT3 expression levels were also determined based on the data. In the 179-patient TCGA cohort, FHL1 high patients were more frequent in AML FAB subtype M0 and less fell into FAB subtype M3 (P = 0.0001438, P = 0.002776, respectively). See supplementary Table S1. In view of the therapeutic effect of FAB-M3 patient was very effective, this subtype datasets (16-AML cohort) was also deleted. In cytogenetically risk, good risk AML patients were deleted, and IP-AML patients (145-AML cohort) were finally just considered.
Estimation of outcome signature for patients' prognosis, construction and assessment of the nomogram To investigate the associations between FHL1 expression levels and clinical, other molecular characteristics, the Fisher exact test was available for categories of variables and the Wilcoxon ranksum test was available for continuous variables, respectively. The Kaplan-Meier method and log-rank test were utilized to estimate the association between FHL1 expression and OS, EFS. The proportional hazards assumption was verified for each variable before fitting Cox models. Multivariate Cox proportional hazards models were utilized to study the association between FHL1 expression levels and OS, EFS and RFS in the presence of other known risk factors. The forest was used to show the P value, HR and 95% CI of each variable through 'forestplot' R package. We used R software to perform the nomogram and evaluate the performance of the 4-year, 5-year and 6-year OS nomogram. Then receiver operating characteristic curve (ROC) analysis was used to calculate the area under curve (AUC) and check the prediction accuracy for our model.

Bioinformatics analysis and statistical analyses
Between FHL1 high and FHL1 low groups, the statistical cutoff values were a fold-change of 2 and an adjusted P-value of < 0.05 (training data set). Hierarchical clustering based on expression levels of these mRNAs were performed and visualized by heatmap. Pathways were developed by Metascape, DAVID and Kyoto Encyclopedia of Genes and Genomes (KEGG). All analyses were performed using the R 3.4.4 software packages.

Results
Expression of FHL1 in K562 and K562/A02 AML cell lines RNA-Seq was carried out and compared in the drug resistance cell lines (K562/A02) and the nonresistant cell lines (K562) (each has 3 biological replicates) (Fig. 1). Up-regulation genes (drug resistance-associated genes) were filtered out (Fig. 1a). Remarkably higher mRNA expression of FHL1 was evident in K562/A02 cell lines than K562 cell lines (P = 0.03) (Fig. 1b). Real-time quantitative PCR further confirmed that FHL1 was expressed at high mRNA expression levels in drug resistant group (P = 0.02427, 3 K562/A02 VS 3 K562) (Fig. 1c). All these results showed significant mRNA overexpression of FHL1. RNA study of 2 human leukemia cell lines found SPARC and FHL1 were two eminent rank up-regulated genes (Fig. 1a). However, SPARC has previously been in-depth studied in acute myeloid leukemia, so we decide to study gene FHL1 [17].
Differences of clinical and molecular characteristics between FHL1 high and FHL1 low groups Clinical characteristics of 145 IP-AML patients in TCGA dataset with FHL1 high were summarized in Table 1. No significant differences were noted in clinical information including age, WBC count, BM blasts and PB blasts between the two groups. FHL1 high patients were more common in AML FAB M0 (P = 0.0001212). FHL1 high patients seemed likely to have a lower prevalence of FLT3-ITD, NPM1 and biallelic CEBPA mutation than FHL1 low patients (P < 0.05). (See Table 1) With regard to cytogenetic characteristics, FHL1 high patients had a tendency for higher frequency of complex cytogenetics (P = 0.01282), and seemed more likely to have a higher prevalence mutation of RUNX1, and accompanied with higher expression levels of BGR, BAALC and MN1 than FHL1 low patients (P < 0.05). (See Table 1 In addition, the increase was linked to worse OS of patients in the Kaplan-Meier survival analysis after controlling for multiple hypothesis tests (Fig. 2d). FHL1 expression and age, WBC, FLT3, TP53, CEBPA mutations were significantly different in Hazard ratio (Fig. 2d, P < 0.01). A larger cohort of 284 AML samples (GSE6891, validation data set) was further studied, results also showed that FHL1 high was significantly associated with shorter OS and EFS (Fig. 3

Construction of nomogram and validation of prognostic model
According to the results of cox regression analysis and forest plot, we further construct a nomogram combining independent prognostic factors, including potential factor patient's age, WBC and gene NPM1, CEBPA, TP53, FLT3, FHL1 to provide a quantitative method for the clinicians to predict the probability of 4-year, 5-year and 6-year OS in IP-AML patients (Fig. 4a). Every patient would get a total point by plus each prognostic parameters point, and the higher total points mean a worse outcome for that patient. Moreover, ROC analysis showed that the prediction accuracy of the outcome model and AUC was 0.773 (Fig. 4b).

Associations between genome-wide gene-expression profiles and FHL1 expression
To further assess the biological role of FHL1 in leukemogenesis, we derived FHL1-associated geneexpression differential analysis based on the comparison of RNA expression between FHL1 high and FHL1 low patients group (training data set). A total of 172 up-regulated and 87 down-regulated genes were considered to be significantly associated with FHL1 expression (Fig. 5a). Further, these aberrant genes were presented as an expression heatmap (Fig. 5b). These up-regulated genes included genes known as leukemia-associated biomarkers, such as: (1) CCND1, which played a crucial role in the regulation of cell proliferation and hematopoietic differentiation, and CCND1-BCL2 gene network was a direct target of amifostine treating human acute megakaryocytic leukemia [18]; (2) MLLT3, which was a proto-oncogene, was first reported in leukemia and involved in many different cellular processes, such as cell differentiation [19];(3) AKT3, was a key regulator of signal transduction pathways. Its tight control over cell proliferation and cell viability is manifold [20]. However, some tumor suppressors were down-regulated, such as: (1) DDIT3, which emerged as a regulatory node with positive linkage to erythroid regulators and negative association with myeloid determinants [21]; (2) CEBPA, a decrease of CEBPA-dependent HK3 expression promoted primary AML. In K562/A02 cell lines, we found the top rank of differences pathways PI3K-Akt signaling pathway and JAK-STAT signaling pathway (Fig. 5c). In KEGG, it was interesting that cell signaling pathways of FHL1 expression were involved in "PI3K-Akt signaling pathway" and "MAPK signaling pathway", both of which were included in "JAK-STAT signaling pathways" (CCND1, FHL1, IL2RA, IL6ST, AKT3) (Fig. 5d).
These dysregulated genes and pathways were consistent with their known understanding about leukemogenesis, which might explain the involvement between FHL1 and the prognosis of IP-AML.

Discussion
The majority of AML patients still succumb to this disease because of refractoriness to therapy or relapse, and new therapeutic approaches for AML are an urgent clinical need. Our study provided direct evidences that FHL1 high predicted adverse outcomes for AML. Firstly, FHL1 was widely expressed at high levels in drug resistant cell lines (Fig. 1b). Meanwhile, quantitative RT-PCR was confirmed consistently that FHL1 as an important gene was differed dramatically and expressed at high level in drug resistant group (Fig. 1c). All these supported that FHL1 was a vital drug resistance gene and significantly overexpressed.
Secondly, for better studying the correlation between FHL1 expression and the prognosis of IP-AML, including some pretreatment molecular characteristics, the prognostic value of FHL1 was further confirmed in the independent group. In these 145 IP-AML patients, we found that FHL1 high was linked to the presence of other adverse prognosticators. With regard to molecular factors, FHL1 high patients had a lower prevalence of FLT3-ITD, NPM1 and Biallelic CEBPA mutation and higher prevalence mutation of RUNX1, higher expression levels of BAALC and MN1 than FHL1 low patients (Table 1).
These findings indicated that FHL1 might play an active role in leukemogenesis just like other tumor markers in IP-AML patients.
Further, patients with FHL1 high were significantly more classified in the FAB M0 subgroups than with FHL1 low , suggesting that the leukemic FHL1 patients derived from relatively more minimally differentiated IP-AML patients (Table 1), which might indicate adverse malignancy. In addition, based on TCGA cohorts, FHL1 high was proved to have significant associations with adverse outcomes.
FHL1 high was associated with shorter OS, EFS and RFS, noteworthy difference in OS and EFS (Fig. 2ac, Fig. 3). This study demonstrated that FHL1 might play an oncogenic role in leading to adverse prognosis in the development of leukemia. Meanwhile, negative factors (Age, WBC and FLT3, TP53, CEBPA mutations) were associated with poor OS and EFS. After adjusting for known prognosticators by multivariable analyses, the association of FHL1 high with adverse OS and EFS still existed and FHL1 high expressers harbored poor OS and EFS. In addition, the nomogram based on the model exhibits an impressive performance and clinical applicability. Moreover, prognostic significance in FHL1 was further consistently validated in another larger 284 IP-AML cohort ( (Fig. 3). All these results proved FHL1 as a potential prognosticator and therapeutic target of IP-AML, which could promote further fine stratification of NCCN cytogenetically IP-group.
Disease recurrence occurs in most IP-AML patients within 3 years after diagnosis. A short duration of remission (i.e., < 6 months), adverse genetic factors, older age, and poor general health status are the major determinants of outcome after relapse [22]. Ignoring the cause of induction death by intense chemotherapy by AML patients whose survival was > 30 days, we also found that high expression of FHL1 was still independently associated with short OS and EFS. Taken  . This paper supported that FHL1 was up-regulated and acted as a potent oncogenic role in leukemia development. Here, we further studied functional pathway of FHL1.
In cell lines, this paper found top rank of differences pathways (PI3K-Akt signaling pathway and JAK-STAT signaling pathway), which might be connected with drug resistance (Fig. 5c), and FHL1 was involved in JAK-STAT signaling pathway (CCND1, FHL1, IL2RA, IL6ST, AKT3) (Fig. 5d), which had shown promising prognostic values. In KEGG, surprise was that cell signaling pathways of FHL1 expression were involved in "PI3K-Akt signaling pathway" and "MAPK signaling pathway", both of which were included in "JAK-STAT signaling pathways" (Fig. 5d). Gene CCND1 played a vital role in the regulation of cell proliferation and hematopoietic differentiation [18,30]. AKT3 control over cell proliferation and cell viability was manifold [20]. In the JAK-STAT pathway, following the binding of cytokines to their cognate receptor, STATs were enabled to modulate the expression of target genes by members of the JAK family of tyrosine kinases. In addition to the activation of STATs, JAKs mediate the recruitment of other molecules such as the MAP kinases, PI3 kinase etc (PI3K-Akt signaling pathway and MAPK signaling pathway existed in the JAK-STAT pathway). These molecules process downstream signals via the Ras-Raf-MAP kinase and PI3 kinase pathways, led to the activation of additional transcription factors. PI3K-Akt-mTOR pathway amplification was associated with reduced OS and DFS [20]. So we speculated that PI3K-Akt signaling pathway and MAPK signaling pathway played a pivotal role in JAK-STAT signaling pathway. In platelet activation pathway and another cancer pathway (VEGF signaling pathway also was found, including PI3K-Akt signaling pathway and MAPK signaling pathway) (Fig. 4d),

Conclusions
In summary, our study provided evidences that high FHL1 expression was associated with drug     The performance of prognostic model in predicting OS. a Nomogram construction. b ROC curves for OS.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.