Introduction
MDS is a group of clonal disorders characterized by morphologic dysplasia, ineffective hematopoiesis, and peripheral cytopenia [
1], with a high risk of developing acute myeloid leukemia (AML) [
2,
3]. The incidence rate of MDS in the general population is 4.5 per 100,000 people per year, but it is higher in males than females (6.2 vs. 3.3 per 100,000 people per year) and substantially increases with age. In addition to gender and age, other risk factors such as chemotherapy drugs, radiation therapy, long-term workplace exposure to benzene, or familial forms can also induce MDS to different degrees [
4]. In clinical practice, epigenetic therapy is the main drug that can reverse the repressive state of DNA hypermethylation to relieve symptom, including azacitidine, decitabine, and lenalidomide [
5], but sometimes it is still impossible to achieve a proper response to the therapy. Allogeneic HSC transplantation remains the only potentially curative option, but it is so strict that many people cannot be effectively cured through this approach [
6]. Some studies indicate that the limited success of HSC transplantation was attributed to the altered BM microenvironment in MDS patients [
7].
As one of the main cellular components of the BM microenvironment, MSCs that characterized by the expression of CD73 and CD90 are a group of stem cells that can repair themselves and have the ability to form bone, adipose and nerve cells [
8‐
10]. Functionally, these cells can contribute to reprograming the BM microenvironment by dysregulating the proinflammatory cytokines and inducing the hypoxia, leading to abnormalities in supportive hematopoietic niches [
11]. Although some reports suggest that there are no differences in phenotype and growth characteristics of MSC between MDS patients and healthy donors (HD) [
12,
13], others believe that even though in the same phenotype MSC can also induce MDS appearing as different characteristics in diverse incubation circumstance [
14,
15]. The controversy indicated that there must be some differences in the secreted molecules or cellular effects of MSC which are closely related to the gene expression profile of the MDS, playing an important role in the progression of MDS. However, the mechanism of MSC promoting the progression of MDS needs to be further explored.
Recently, gene chip technology has developed rapidly and been widely used in gene detection. Through it, we have learned that the pathogenesis and progression of heterogeneity in MDS are closely related to the MSC genetic landscape. For instance, Kim Me et al. have reported that MSC could regulate MDS pathogenesis through inflammation and immune dysregulation responses that involve the interferon signaling pathway [
16], inducing an immune-suppressive microenvironment in MDS by an indirect mechanism involving monocytes or abnormal transforming growth factor β1, a relevant trigger causing MDS to progress to AML [
17]. All of these have shown the feasibility and reliability of exploring MDS-MSC from the perspective of bioinformation. Therefore, a better understanding of the gene expression, developing a comprehensive list, or more consistent testing may help us acquire more useful information to improve the management of patients with MDS.
Driven by the need for effective biomarkers to improve the diagnosis and treatment of MDS, we specifically focused on screening persistently altered genes involved in MDS-MSC. We discovered the novel function of HOXB3 and HOXB7 where gene overexpression is closely associated with MDS progression. Simultaneously, blocking these genes can repair cell proliferation, differentiation, apoptosis and the ability of cells to promote HSC hematopoietic differentiation. Our work identifies HOXB3 and HOXB7 as potential targets for future interventions in MDS.
Materials and methods
Study population
Seventeen MDS patients and two healthy controls (They were diagnosed with nonhematologic diseases) were enrolled in this study. Experiments were approved by the ethics committee of the Second Hospital of Dalian Medical University. All study subjects signed a written informed consent before participating in the study.
The gene expression data of MSC was obtained from the GEO database (
GSE140101, GSE107490 and GSE61853) (
http://www.ncbi.nlm.nih.gov/geo/). All the datasets included healthy donors as control and patients diagnosed with MDS. However, we didn’t analyze the differences in gene expression profiles of BM MSC between MDS subtypes. Total RNA was isolated from BM MSC for gene expression analysis comparing MDS vs. control. Databases were drawn through their portal for analysis [
15‐
17]. The data sets which include the HD and MDS groups were screened and severally analyzed based on GPL10558 (Illumina HumanHT-12 V4.0 expression beadchip), GPL11154 (Illumina HiSeq 2000) and GPL16791 (Illumina HiSeq 2500). Meanwhile, the two samples with missing results in GSE61853 HD group were removed.
Data processing
The raw data downloaded from GEO were used for further analysis. Data processing mainly utilized a set of different R packages (R version 4.1.0 (2021-05-18)) in Rstudio. We downloaded Gene expression file GSE107490_all_count.txt.gz, GSE140101_ FPKM_GEO.txt.gz, GSE61853_non-normalized.txt.gz and their corresponding annotation platforms from the GEO database. The quality control of each data set was performed to minimize false detection rate (FDR) in original studies by using fastqcr package. Next, the expression matrix was normalized using the normalize Between Arrays function under limma package (version: 3.48.0). Gibberish was then removed. PCA under FactoMineR package was used to verify difference between samples and Reliability of data processing. The normalized data were further processed using the limma package to obtain differentially expressed genes (DEGs) between HD and MDS in these three gene sets. In our study, genes with a p-value less than 0.05 and fold-change greater than 2 were considered as DEGs. Each GSE has been analyzed statistically. Venn diagram tool (
http://bioinformatics.psb.ugent.be/webtools/Venn/) was used to help us find overlapped genes.
GO enrichment and KEGG pathway analysis
To identify the overall overlapped genes enrichment differences between the patients and the controls, we used the Database for Annotation, Visualization and Integrated Discovery (
https://david.ncifcrf.gov/, DAVID, version: 6.8) for the further functional enrichment analysis. This website can perform GO and KEGG analyses. The predicted BP (biological process), CC (cell composition), MF (molecular function) of the DEGs were analyzed. Furthermore, the pathways in which the DEGs participated were predicted and mapped using the KEGG database with DAVID. Visualization of the enriched GO terms and KEGG terms were conducted using the GO plot package (version: 1.0.2) in R studio. Terms with p values greater than 0.05 were considered statistically enriched.
PPI network establishment and further module analysis
To investigate protein-protein interaction function, the Search Tool for the Retrieval of Interacting Genes (
http://string-db.org/, STRING) online database was used to identify interactions between known genes and predicted genes at the protein level. Briefly, we input the overlapped genes in the website using the default condition and downloaded the file about these proteins’ interaction to input into Cytoscape software (3.8.2) to obtain gene clusters. The plugins CytoHubba and MCODE in Cytoscape were applied to identify the significant modules in the PPI networks and calculate the degree exhibited by every protein node. Additionally, we extracted PPI pairs based on the combined score over 0.4. The degree cutoff was set to 2, node score cutoff to 0.2, k-score to 2 and Max depth to100 in MCODE.
Hub genes selection and correlation analysis
The screening of hub genes was mainly conducted through the MCODE plugin of Cytoscape. To perform the correlation analysis between the Hub genes and those genes that had been shown to influence the process of MDS, correlation analysis was carried out using the “tidyr”, “dplyr”, “ggstatsplot”, package (3.14.3) in R. Briefly, selected genes’ expression matrix was imported. The Cor.test function was then executed with the default parameters (type="spearman”) setting. Gene sets with a p-value less than 0.05 were considered to have significantly correlated relations.
Real-time quantitative polymerase chain reaction
For quantification of gene expression, RNA was isolated from Ctrl (They were diagnosed with nonhematologic diseases) or patients with MDS using a RNeasy Micro Kit or Rneasy Mini Kit (Qiagen). cDNA was synthesized from 1 µg RNA using Superscript IV Reverse transcription (Thermo Fisher) (37 °C for 15 min, 65 °C for 10 min). Real-time PCR analysis was set up with the SYBR Green qPCR Supermix kit (Invitrogen, Carlsbad, CA) and carried out in the iCycler thermal cycler. β-actin was used for normalization. Data were analyzed by the 2
−ΔΔCT method [
18]. Each sample was analyzed in triplicate, and the analysis was repeated three times. The primers for target genes were as follows.
Primers for HOXB3: forward-5′TGCTGCTGGGAGACTCGTAA 3′.
reverse-5′GCATCCCCTTGCAGCTAAAC 3′,
HOXB5 forward-5′AACTCTCCCCTCCCC ATC 3′.
reverse-5′GGCACTACCCCACCTCAA 3′,
HOXB6 forward-5′TCC CCTCCCAATGAGTTC 3′.
reverse-5 GCATAGCCCGA CGAATAGA 3′,
HOXB7 forward-5′CGTCCCTGCCTACAAATC 3′.
reverse-5′GAAGCAAA CGCACAAGAAG 3′,
SCF forward-5′ACCCAATGCGTGGACTATCTG 3′.
reverse-5′GGCGACTCCGTTTAGCTGTT 3′,
TPO forward-5′CTTCACTGCCTCAGCCAGAAC 3′.
reverse-5′GAATCCCTGCTGCCACTTCA 3′,
IGF1 forward-5′CCTCTCAAGAGCCACAAATGC 3′.
reverse-5′TCCAGCAGCCAAGATTCAGA 3′,
IGFBP2 forward-5′TGACAAGCATGGCCTGTACAA 3′.
reverse-5′CACGCTGCCCGTTCAGA 3′,
CXCL12 forward-5′ATGTCGAAGCCCCATAGTGAA 3′.
reverse-5′TGGGTGGTGAATCAATGTCCA 3′,
β-ACTIN forward-5′CATGTACGTTGCTATCCAGGC 3′.
reverse-5′CTCCTTAATGTCACGCACGAT 3′.
Cell transfection
MSC cells were seeded 2 × 105 cells per well in 6-well culture plates with DMEM/F-12 medium containing 10% FBS for one day. When cells were grown to a concentration of 70%, transient transfection was performed using the transfection reagent GP-transfectMate (GenePharma, China) according to the manufacturer’s protocol. The MSC were then transfected with HOXB3 or HOXB7 small interfering RNA (siRNA) (GenePharma, China) and a non-speciffc control siRNA (NCsiRNA). SiRNA was mixed with GP-transfect-Mate transfection reagent in serum-free medium according to the manufacturer’s instructions incubated for 6 h, then the cells were incubated in a Growth Medium for following analysis. The siRNA sequences are as follow:
siRNA HOXB3: sense-5′GAUGAAAG AGUCGAGGCAATT 3′
antisense-5′UUGCCUCGACUCUUUC AUCTT 3′,
siRNA HOXB7: sense-5′GCUAUUGUAAGGUCUUUGUTT 3′
antisense-5′ACAAAGACCUUACAAUAGCTT 3′
CCK-8 assay
MSCs were seeded in 96-well cell culture plates. Cells were transfected when they reached a density of 40–60%. Cell proliferation was assessed every 24 h after transfection by measuring absorbance at 450 nm with a multiskan-FC (Thermo Fisher) according to the instructions of the CCK-8 kit (Vazyme).
apoptosis assay
After 24 h of transfection, the cells were treated according to the instructions of the eBioscience Annexin V-FITC Apop Kit (Thermo Fisher) and detected by flow cytometry. Finally, the data were statistically analyzed using Graphpad Prism software to plot the cell distribution. Three replicates were set for each group.
Assay of adipogenic differentiation and osteogenic differentiation
After transfecting with HOXB3 or HOXB7, MSC were collected and cultured with adipogenesis induction medium (α-MEM containing 10% FBS, 5 µg/mL insulin, 0.5 mmol/L 3-isobutyl-1-methylxanthine, and 1 µmol/L dexamethasone) in 6-well plates with 1 × 105 cells per well for 14 days. Every three days, we changed culture medium. We used Oil Red O staining to distinguish mature adipocytes from preadipocyte during the process of culture. For osteogenic differentiation, MSC were inoculated in 6-well plates with 1 × 105 cells per well and cultured in freshly formed osteogenic medium (OM) for 21 days. Alizarin Red staining was used to detect bone mineralization.
MDS-MSC (with or without HOXB3/7 treatment) were plated at a near confluent density of 1.0 × 104 cells per well in 48-well plates. 24 h later, healthy CD34 + HSC were seeded in contact with the MSC feeder layer at a density of 5.0 × 105 cells per well in hematopoietic media and cultured for up to 5 days. Then, HSC were cultured in methylcellulose media (MethoCult™ H4434, STEMCELL) for 14 days and colonies were counted using an inverted microscope.
Survival analysis
The overall survival (OS) analysis of hub genes was conducted using the Kaplan-Meier curve in cBioPortal. cBioPortal (
http://www.cbioportal.org) is an online analysis platform for multidimensional cancer genomic data that visualizes genes, samples and data types. Survival analysis was performed by alternating the Hub genes on line. And P-value less than 0.05 was considered statistically significant.
Statistical analysis
The data are displayed as histograms and line charts. The parameters were compared using one- or two- way analysis of variance. Data are representative of at least three independent experiments. Statistical analysis was performed using Prism 8.4.3 (GraphPad Software, La Jolla, California). A p-value lower than 0.05 was considered statistically significant.
Discussion
MDS is a clonal hematopoietic system disease that is difficult to diagnose, characterized by reduced hematopoietic function, peripheral blood cytopenia and morphogenesis [
22]. According to the revised version of the International Prognostic Scoring System (IPSS-R), MDS can be divided into different subtypes. The treatment for MDS in the risk group included component blood infusion, hematopoietic factor therapy, immunomodulator and epigenetic drug therapy. But only a few drugs are currently available for treatment, more drugs are now under clinical investigation [
23], and overcoming MDS remains a challenge for us.
In recent years, targeted therapies are emerging for small subsets of MDS patients with specific somatic mutations (such as TP53, IDH1/2, FLT3). But currently, they have not been approved widely for use as mutation-directed medications of treating MDS [
23]. At the same time, accumulated data indicate that MSC in MDS model display aberrant characteristics contributing to disease initiation and transformation into AML [
24]. Hence, it is very urgent to identify potential markers, especially in MSC, to promote the diagnosis and prognosis of MDS. Therefore, in the research we analyzed the potential therapeutic targets of MSC in MDS patients based on bioinformatics, to find potential therapeutic targets of MDS.
Due to the strong heterogeneity of MDS patients and significant changes in the course of the disease, it is difficult to find common targets in limited sample size studies. Here, we used three databases containing MDS and HD samples, including
GSE140101, GSE107490, and
GSE61853 from GEO. Significantly, differences of PSG5 (a putative AF (Amniotic fluid)-MSC markers [
25]) and SLC5A3 (essential to support a myo-inositol auxotrophy in AML [
26]) were expressed in various stages of MDS in these data. In order to find more potential targets on MDS-MSC, we integrated the differential genes in three databases to obtain 62 differential genes, all of them appear more than twice in three Dataset. Furthermore, GO enrichment analysis indicated that the identified DEGs were mainly enriched in embryonic skeletal system morphogenesis, angiogenesis, anterior/posterior pattern specification, sequence-specific DNA binding and platelet degranulation. They were all related to growth and development, which showed abnormal biological processes associated with cellular phenotypes and transcriptional regulation in MSC. Actually, they had been taken as the more important cause of MDS, similar to the study in 2013 conducted by Geyh S et al., in which they reported that MSC are structurally, epigenetically and functionally altered, which leads to impaired stromal support and seems to contribute to deficient hematopoiesis in MDS [
27]. Then KEGG enrichment analysis shown the difference were enriched in P53 signaling pathway representing the tumor suppression and MAPK signaling pathway, playing a key role in the differentiation, proliferation and apoptosis of cells [
28,
29]. This clearly proved that inhibition of MSC played important roles in the transformation from MDS to AML, although some research had shown that there were no differences observed with respect to phenotype, differentiation capacity, immunomodulatory capacity or hematopoietic support in MSC between MDS and HD [
30].
By establishing the PPI network of DEGs, we picked out HOXB3, HOXB5, HOXB6, and HOXB7 as hub genes with the highest degrees. And, our experiment results manifest that HOXB3 and HOXB7 significantly regulates hematopoiesis capacity in MSC at the process of MDS and plays a key role. Homeotic (HOX) genes, a group of genes regulating the body shape, are developed in regulatory system and transcription factor that causes cell differentiation blocking and malignant self-renewal [
31,
32]. The role of the HOXB3 varies in different tumors. Some studies suggested that loss of HOXB3 correlates with the development of hormone receptor negative breast cancer [
31], or act as tumor suppressors through FLT3-ITD driver in AML [
33]. But some scholars believe that HOXB3 promotes prostate cancer cell progression by transactivating CDCA3 [
34]. We found that the gene expression of HOXB3 was increased significantly indicating that it may associate with malignant lesions of MDS. Different from HOXB3, HOXB5 was negatively correlated with myeloid cell differentiation signaling [
35], but promoted tumor aggression and progression of various tumors including AML [
36,
37]. According to our results, HOXB5 may promote the progression of MDS to AML. Abnormalities of the HOXB6 expression in granulopoiesis and monocytopoiesis may contribute to the development of the leukemic phenotype appearing as its overexpression in murine BM and generating a myelomonocytic precursor in vitro [
38], and causes HSC expansion and AML in vivo [
39]. As expected, HOXB6 increased significantly also in MDS-MSC in our results. HOXB7 that can induce activation of MAPK/ERK pathway which promotes tumor progression is also upregulated in MDS-MSC [
40]. The results implied that the HOX family, especially the overexpression of HOXB3, HOXB5, HOXB6, HOXB7, played important roles in normal and malignant hematopoiesis in MSC of MDS. Furthermore, to demonstrate the importance of hub genes, we screened a series of genes, including TET2, DNMT3A, ASXL1, EZH2, SF3B1, SRSF2, U2AF1, ZRSR2, RUNX1, TP53, STAG2, NRAS, CBL, NF1, which had been shown to influence the procedure of MDS, for correlation analysis. As predicted, significant correlations indicated the important roles of hub genes. At the same time, survival analysis showed that MDS patients having HOXB3, HOXB5, HOXB6, HOXB7 alterations showed worse OS. Furthermore, we explored the function of hub genes in MDS-MSC that obtained through bioinformatics analysis
in vitro. The results showed that compared with other genes, HOXB3 and HOXB7 could regulate the function of MSCs cells to a greater extent. But because of the size and source of sample, the results need to be more explored.
However, due to a shortfall of the underlying data acquisition technology, false positive results may occur, which is also the biggest flaw of this paper. In addition, recently, different views on the pathogenic role of MSC in MDS have been raised. The researchers believed that although MDS-MSC displayed higher mutational burdens compared to healthy MSCs, no evidence for acquired mutations as disease initiators for MDS was found [
41]. Next, more samples of MDS patients will be collected to verify the conclusions through experiments.
In conclusion, our findings predicted that dysplasia of MDS-MSC is closely related to the pathogenesis of MDS through altered HOXB family, providing potential targets for therapeutic and diagnostic applications in MDS.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.