The Effects of METTL3-Mediated Regulations of the Adipogenic Differentiations of Bone Marrow MSCs Targeting PI3K/AKT Pathways on the Chemo-Resistance of AML CURRENT STATUS: POSTED

Background: The adipogenic differentiations of bone marrow mesenchymal stem cells (MSCs) result in the promotion of the chemo-resistance of AML cells. When compared with bone marrow MSCs from healthy donors (HD-MSCs), it has been observed that the MSCs from AML (AML-MSCs) displayed enhanced adipogenic differentiation abilities. However, the underlying mechanism remains elusive. METTL3-mediated modifications of m 6 A mRNA play key roles in the regulation of stem cell differentiation. Unfortunately, there has been no research to date which has reported the fact that the increases of adipogenic differentiation of AML-MSCs are regulated by the m 6 A. The goal of this study was to investigate the critical role of METTL3 in the adipogenesis of AML-MSCs for promoting the chemo-resistance of AML. Methods: This study detected the m6A levels of the MSCs total RNA, and then investigated METTL3 functions in the adipogenic differentiations of MSCs using knockdown and overexpression experimental methods. The effects of the adipogenic differentiations of HD-MSCs/AML-MSCs on the chemo-resistance to AML were also examined. In addition, the gene transcriptome and m 6 A epigenetic changes of AML-MSCs from four AML patients, and the HD-MSCs from three healthy donors, were analyzed in this study. The differentially expressed genes and pathways were successfully identified, and clinical specimens were used to adjust gene expressions in order to determine their effects on the adipogenic differentiations of MSC. Results: The results demonstrated that METTL3 could significantly inhibit the adipogenic differentiations of MSCs. It was found that when compared with the HD-MSCs, the expressions of the PI3K/AKT signaling pathways in the AML-MSCs were significantly increased. Meanwhile, the levels of m 6 A were significantly decreased. Therefore, it was indicated that by blocking the PI3K/AKT signaling pathways with AKT inhibitors, the adipogenic differentiations of the MSCs could be significantly inhibited. Moreover, the increased adipogenic differentiations induced by the METTL3 could significantly increase the chemo-resistance of the AML. Conclusions: It was found that


Introduction
Acute myeloid leukemia (AML) is the most common acute leukemia diagnosed in adults, and is characterized with a high recurrence rate [1,2] . Bone marrow microenvironments (BMM) and AML cells remodel to promote the development of the disease [3][4][5] . MSCs, which are important elements of the BMMs, are the precursors of osteoblasts and adipocytes in bone marrow [3] . The balance between the osteogenesis and adipogenesis is strictly regulated in order to ensure normal bone marrow functions [6,7] . The results of previous studies have shown that the MSCs from AML patients (AML-MSCs) have significant impairments of osteogenic differentiations, as well as remarkable enhancements of adipogenic differentiations [8] . The enhanced adipogenesis of the AML-MSCs has been found to promote the development of leukemia [9] .
The epigenetic regulation of mRNA is an important factor affecting stem cell differentiation [10] . N 6methyladenine (m 6 A) is the most abundant, dynamic, and reversible RNA modification for eukaryotic mRNA [11,12] . The m 6 A is carried out with the participation of methyltransferase complexes (writers, such as METTL3, METTL14, and WTAP); demethylase (erasers, such as ALKBH5 and FTO); and m 6 A reader proteins (readers, such as YTH family and IGF2BP family), which have the ability to dynamically and reversibly regulate RNA splicing, transport, translation, and stability [13][14][15] .
Both the methylase and demethylase of the m6A play important roles in AML. Recent studies have found that METTL3 is highly expressed in AML cells, which demonstrated the ability to induce the progression of leukemia in AML transplanted mice [16] . FTO can promote the malignant transformation and leukemogenesis mediated by oncogene by reducing the mRNA m 6 A levels of ASB2 and RARA [17] .
It has been found that R-2HG plays an anti-leukemia role by targeting the FTO by increasing m 6 A levels and inducing cycle arrest and apoptosis of the AML cells. However, the specific roles of m 6 A modifications in the adipogenic differentiations of AML-MSCs have not yet been reported. In this study, RNA sequencing and m 6 A microarrays, combined with clinical specimen validations, were used to analyze the mRNA and signal pathway differentially expressed in AML-MSCs [18] . The obtained results will provide a theoretical basis for determined new targets for AML treatments from the perspective of m 6 A modifications and bone marrow microenvironmental conditions, which have potential value in future clinical applications.

Primary bone marrow MSCs cultures
In the current study, bone marrow specimens of AML-MSCs were obtained from the hematology department of Fujian Medical University Union Hospital. The collection of the bone marrow specimens was approved by the hospital's ethics review board, and obtained using an informed written consent process. Mononuclear cells (MNC) were isolated using a Ficoll-Hypaque Density Centrifugation Method. Then, the MNC were seeded in low-glucose Dulbecco's Modified Eagle Medium (LG-DMEM, Hyclone), supplemented with 20% Fetal Bovine Serum (FBS, Gibco) and 1% Penicillin-Streptomycin (Hyclone). The cells were plated in 100 mm 2 culture dishes at a density of 10 6 cells/ml and incubated at 37°C with 5% CO 2 . Then, following a culturing period of one week, the non-adherent cells were removed, adherent MSCs were passaged with a confluence of 80 to 90%, and the cells were expanded to passage (P) four. The bone marrow specimens of HD-MSCs were purchased from the Cyagen Biosciences Company. The cultured MSCs were identified as CD34 -/CD44+/CD45 -/CD73+/CD90+/CD105+, whose positive rates were all greater than 95%. In addition, the MSCs differentiations into osteocytes and adipocytes were conducted and had demonstrated multidirectional differentiation potential.

MSC differentiations
The MSCs were cultured in advance with 2 x 10 5 cells/well in six-well plates with 0.1% gelatin. Then, when the cultured cell fusion degree reached between 60% and 70%, the cells were carefully sucked out in order the complete the culture medium, and 2 ml of osteogenic differentiation culture medium were added. The induced differentiation was stained on the 21st day [19,20] .
During the process of the MSCs culturing with 2 x 10 5 cells/well in six-well plates, the solution was changed every three days until the cell fusion reaches 100% or over fusion occurred. Then, the cell complete culture medium was carefully sucked out, and 2 ml of adipogenic differentiation medium A were added to the six-well plates. After a period of three days, Solution A was sucked out and 2 ml of differentiation medium solution B were added. Then, after 24 hours, Solution B was replaced with Solution A. The aforementioned steps were alternately repeated 3 to 5 times (12 to 20 days). The cultures with Solution B were maintained for three days until the lipid drops became large and round enough for the subsequent experiments. During the maintenance of Solution B, fresh Solution B was used every three days. Then, the specimens were stained using oil red O [20,21] .

Alizarin red staining and oil red O staining methods
For the alizarin red staining (ARS) conducted in this study, the cells were fixed after a 21-day period of induction of osteogenic differentiation using 4% paraformaldehyde for fifteen minutes. The cells were then stained with 1% ARS (pH 4.2, Sigma-Aldrich) for five minutes. Then, the staining solution was dissolved with a 10% cetylpyridinium chloride solution and quantified using an enzyme-labeled instrument (Thermo Fisher Scientific) at 562 nm.
In regard to this study's oil red O staining process, after a 14-day period of induction of adipogenic differentiation, the cells were subjected to oil red O staining (Sigma). During the process, the cells were first fixed with 4% paraformaldehyde for thirty minutes, and then stained with an oil red O staining working solution (saturated oil red O and distilled water at a ratio of 3:2) for ten minutes. The lipid drops were shown as bright red under microscopy. Following the observational process, the staining solutions were dissolved with isopropanol and quantified using an enzyme-labeled instrument (Thermo Fisher Scientific) at 450 nm.

Quantifications of the global m 6 A levels
The global m 6 A levels of the total RNA were measured using an EpiQuik m 6 A RNA Methylation Quantification Kit (Colorimetric) (Epigentek) [22] , following the manufacturer's protocol. In this study, quantified using a colorimetric method, and the absorbance of each well at a wavelength of 450 nm was measured. The calculations were completed in accordance with the standard curve reflected by the m 6 A levels [23] . In addition, in order to ensure the accuracy of the results, each sample was run in triplicate.

Analysis results of the RNA isolations and quantitative reverse transcription polymerase chain reactions (RT-qPCR)
In the current study, the total RNAs were extracted from the HD-MSCs and AML-MSCs using a Trizol reagent (Invitrogen, Carlsbad, CA), during which the operational instructions were strictly followed.
The purity levels and concentrations were measured using a NanoDrop 2000 spectrometer (Thermo Fisher Scientific, Waltham, MA). The analyses of the reverse transcription reactions were carried out using a cDNA synthesis kit (Roche, Mannheim, Germany). Furthermore, quantitative real-time PCR was performed according to the manufacturer's instructions, utilizing ABI 7500 Real-Time System (Life Technologies) and SYBR Green Fluorescence (Roche, Mannheim, Germany) instruments. The qPCR reaction components comprised 1 μl of cDNA (with a 2.5-fold dilution); 5 μl of thunderbird SYBR Green qPCR mix; 0.3 μl of PCR primers; and 3.4 μl of RNase-free water. Next, the following protocol was undertaken: 95℃ for 1 minute, and 35 cycles of 95℃ for 15 seconds, 60℃ for 30 seconds, and 72℃ for 30 seconds, respectively. For the in-vitro experiments, the relative expressions were analyzed using a 2 -ΔΔCt method. Also, in regard to the clinical data, the relative expressions were analyzed using a 2 -ΔCt method and the GAPDH were used as references [24] . The primers used are listed in Table 1, and each sample was run in triplicate.

Western Blotting Results
The MSCs were collected to extract the proteins and the protein concentrations were determined using a BSA method. The denatured proteins were loaded and separated by 10% SDS-PAGE and transferred to polyvinylidene fluoride membranes. The membranes were blocked with 5% no-fat milk (Bio-rad) and then incubated with primary and secondary antibodies. The Western Blotting substrates were detected using an immuno-blotting process with Pierce ECL (Thermo Fisher Scientific). The antibodies used for the Western Blotting were as follows: METTL3 (86132, Cell Signaling Technology); AKT (12319, Cell Signaling Technology); p-AKT (13987, Cell Signaling Technology); and GAPDH (2118, Cell Signaling Technology), among which the GAPDH were used as references.

Plasmid construction processes
The wild-type METTL3-CDS were PCR-amplified from the HD-MSCs cDNA using the primers described in Table 1. The PCR products were purified and then cloned into modified pLJM1-EGFP (#19319, Addgene) lentiviral vectors. The sequences of the shRNAs targeting human METTL3 were shRNA #9 and shRNA #12, and the control scramble shRNA (Addgene #1864) were purchased from Addgene [16] .

Virus preparation and infection processes
The recombinant lentivirus plasmids were extracted using an Endotoxin-free Plasmid Extraction Kit (QIAGEN) and co-transfected into 293 T cells with lentivirus packaging plasmids (psPAX2 and pMD2.G) for virus particles. Then, the virus particles were tested at 48 and 72 hours after transfection and concentrated with titer≥1×10 8 TU/ml. Subsequently, the virus was added to the MSCs and purinomycin for three days. Finally, the genomic DNA was extracted for verification purposes.

RNA sequencing and data analyses of the MSCs
RNA samples were extracted, and the concentrations were measured using NanoDrop 2000. Then, 1 to 2 ug of the total RNA from each sample were selected for the purpose of constructing an RNA sequencing library. The specific process was as follows: 1. Total RNA samples which were enriched in oligo dT (without rRNA) were selected; 2. The samples were constructed using a KAPA Stranded RNA Sequencing Library Prep Kit (Illumina); 3. During the library construction processes, double-strand cDNA was synthesized and combined with the subsequent high-fidelity PCR polymerase using a dUTP method, resulting in the RNA sequencing library having chain specificity; 4. The constructed library was identified using an Alient 2100 Bioanalyzer and quantified by qPCR; 5. The mixed sample libraries were sequenced using Illumina X-ten/NovaSeq. The image processing and base recognition were carried out in this study by adopting the software of Solexa Pipeline Version 1.8. In addition, FastQ software was utilized to evaluate the sequencing quality of the reads (using Cutsapt to remove 3' and 5' splices). Then, R software Balltown [25,26] was used to calculate the FPKM (fragments per kilobase of gene/transcription model per million mapped fragments) at the transcriptional level, as well as calculate and screen the differentially expressed genes. The threshold was set to a 1.5-fold change, p-value ≤ 0.05, and mean value of the FPKM in the group ≥ 0.5, in order to screen the differentially expressed genes. Finally, the results of the GO (gene ontology) functional significance enrichment analysis processes, as well as the results of the KEGG (Kyoto Encyclopedia of genes and genomes) pathway significance enrichment analysis processes, were further analyzed in this study.

Microarray hybridization and data analysis results
The immuno-precipitated RNA samples from the HD-MSCs and AML-MSCs were labeled with Cy5 fluorescent dye using a Super RNA Labeling Kit (Arraystar), and then purified using an RNeasy Mini Kit. The Cy5 labeled cRNAs were fragmented and hybridized to human mRNA and lncRNA m 6 A epitranscriptomic microarray (8 × 60 K, Arraystar), which contained 44,122 mRNAs and 12,496 lncRNAs degenerate probes. The hybridized arrays were scanned using an Agilent Scanner G2505C [27] .
This study utilized Agilent feature extraction software (Version 11.0.1.1), and each spot on a microarray was evaluated and flagged as present, absent, or marginal, based on its uniformity above the background and saturation levels. The raw intensities of the immuno-precipitated RNAs were normalized with an average of log2-scaled spike-in RNA intensities, and the fold changes and P values were calculated for each transcript between the HD-MSCs and the AML-MSCs. A cutoff of 5-fold (P < 0.05) was used to identify the differentially m 6 A methylated RNAs. At that point in the experiment, both GO analysis and KEGG Pathway analysis were performed for the differentially m 6 A methylated mRNA transcripts using the DAVID (Database for Annotation, Visualization, and Integrated Discovery) bioinformatics database [28] .

SRAMP database
SRAMP (which is based on RNA adenosine methylation site predictions) is a sequence-based m 6 A modification site predictor [29] . It has been found that SRAMP can be a useful tool for predicting m 6 A modification sites on RNA sequences of interest. It was required to input the full-length RNA sequence of the AKT into the prediction website in order to predict the m 6 A modification site on the sequence according to the specific m 6 A modification sequence.

Statistical analysis results
In this study, the experiments were repeated three times for all of the examined specimens.
Statistical analysis was performed using GraphPad Prism8.0 software, and the obtained statistical differences between the groups were determined using a Student's t test. In the current study, Pvalues < 0.05 were considered to be statistically significant.

Significant increases in the adipogenic differentiations of the AML-MSCs
The detections of cell morphology, phenotype, and multidirectional differentiation potential of the MSCs indicated that there were no differences in the HD-MSCs and the AML-MSCs ( Supplementary Fig.   1A to 1D). During the cell culturing process, there were also no significant differences observed in proliferation between the HD-MSCs and AML-MSCs ( Supplementary Fig. 1E). The adipogenic differentiations of the HD-MSCs/AML-MSCs were induced, and the adipocyte was detected using an oil red O staining process, as shown in Fig. 1A. Then, after the isopropanol was dissolved, the OD450 was determined (Fig. 1B). The results indicated that the adipogenic differentiation ability of the AML-MSCs was stronger than that of the HD-MSCs, and the survival timeframes of the AML patients with high adipogenic differentiations would be significantly shorter than those of patients with low adipogenic differentiations, as illustrated in Fig. 1C.

AML-MSCs inhibit the chemo-resistance of AML cells more significantly than HD-MSCs following the induction of adipogenic differentiations
In the current study, the adipogenic differentiations of HD-MSCs and AML-MSCs were induced for fourteen days, and then co-cultured with AML cells in order to test the chemo-resistance of the AML cells. The results showed that the chemo-resistance of the AML-MSCs to AML cells (including HL-60, U937 and THP-1) were significantly higher than that of the HD-MSCs, as detailed in Fig. 2. Therefore, it was indicated that they may play important roles in affecting the chemo-resistance of AML.
Subsequently, it was particularly important in this study to identify the molecular mechanism affecting the adipogenic differentiations.

Global m 6 A levels and expressions of METTL3 decreased in the AML-MSCs
It has been determined that RNA epigenetic modifications play key roles in the differentiations of stem cells. The results of previous experimental studies have shown that m 6 A methylase METTL3 played a key role in the processes of adipogenic differentiation of bone MSCs in pigs [30] and mouse [31] . Therefore, confirming that m 6 A modifications play important roles in the differentiations of human bone marrow MSCs was a priority in the present study. An EpiQuik m 6 A RNA methylation quantitative detection kit (Epigentek) was adopted to detect the global m 6 A levels. The results showed that when compared with the HD-MSCs, the m 6 A levels of total RNA in the AML-MSCs were decreased (Fig. 3A). Meanwhile, the results of the RT-qPCR (Fig. 3B) and Western Blot (Fig. 3C) showed that the expressions of METTL3 were significantly decreased in the AML-MSCs. The next step was to determine if the increased adipogenic differentiation abilities of the AML-MSCs were related to the expressions of METTL3. In order to verify the effects of the METTL3 on the adipogenic differentiations of the MSCs, this study induced the adipogenic differentiations of the MSCs by overexpression and knockdown of the METTL3. Subsequently, the changes in the adipogenic differentiation abilities of the MSCs were successfully detected. The results indicated that following the infection of the AML-MSCs with METTL3 overexpression lentivirus, the RT-qPCR (Fig. 4A) and Western Blot (Fig. 4B) results showed significant overexpressions. In addition, the results revealed that the METTL3 overexpressions could significantly inhibit the adipogenic differentiations of the AML-MSCs (Fig. 4C). In contrast, two independent shRNAs targeting METTL3 were used to knockdown the METTL3 in the HD-MSCs, and the RT-qPCR (Fig. 4D) and Western Blot (Fig. 4E) results had verified the significant knockdown. The adipogenic differentiations of the HD-MSCs were found to be significantly enhanced (Fig.4F). Therefore, it was confirmed that the METTL3 was a key factor in the regulation of the adipogenic differentiations of the MSCs. However, the question still remained as to how the m6A regulates the expressions of target genes and affects the adipogenic differentiations of the MSCs. Therefore, the mRNA expressions and differential m6A expressions of the HD-MSCs and AML-MSCs were detected in this study in order to further explore the intrinsic mechanism of adipogenic differentiations of the MSCs.

High expressions of AKT in the AML-MSCs and their association with the adipogenic differentiations
In order to access the differentially expressed genes between the HD-MSCs and AML-MSCs, this study sequenced the MSCs RNA of three HD and four untreated samples of newly diagnosed AML. In this study, all of the subjects were ranged between 30 and 50 years of age, and the samples were genes, including 242 up-regulated genes and 828 down-regulated genes, were significantly different between the two groups (|fold change| ≥1.0; P-value < 0.05), and were displayed in the form of a volcanic map (Fig. 5B). Then, based on the differentially expressed genes, gene enrichment analysis was carried out. The results of the GO analysis showed that, when compared with the HD-MSCs, the AML-MSCs were significantly up-regulated in terms of certain biological processes, including the regulation of signaling receptor activities; blood vessel morphogenesis; extracellular matrix organization; extracellular structure organization; and so on (Fig. 5C), which indicated that the MSCs may change the microenvironmental conditions through signal exchanges between the extracellular matrix components and the extracellular substances. This study's KEGG analysis results indicated that the PI3K/AKT signal pathways were significantly up-regulated in the AML-MSCs, and may potentially be involved in the processes of enhancing the adipogenic differentiation abilities of the AML-MSCs (Fig. 5D). In addition, RT-qPCR and Western Blot were used to detect the expression levels of the AKT mRNA and proteins in the examined clinical samples. The results showed that the AKT activities and expression levels in the AML-MSCs were significantly up-regulated (Figs. 5E and 5F). It has been noted that the AKT inhibitor MK-2206 2HCL is a highly selective AKT1/2/3 inhibitor, which has the ability to inhibit AKT activities in AML-MSCs and induce the adipogenic differentiations of the MSCs. The results obtained in this study indicated that the adipogenic differentiations of the MSCs with MK-2206 2HCL were significantly inhibited, as illustrated in Fig. 5G.

Significantly reduced m 6 A levels of AKT mRNA in the AML-MSCs
This study analyzed the methylation RNAs of m 6 A in the HD-MSCs and AML-MSCs, and HD-MSCs served as the control. The methylation RNAs were profiled using the microarray with probes for 44,122 mRNAs and 12,496 lncRNAs. The results of microarray profiling showed that 127 mRNAs were lowly methylated between the HD-MSCs and AML-MSCs (Fig. 6A). In addition, this study's KEGG (Fig.   6B) and GSEA (Fig. 6C) analysis results showed that, when compared with the HD-MSCs, the m 6 A levels of the PI3K/AKT signaling pathways in the AML-MSCs were significantly reduced. This study also used the online prediction to mammal m 6 A modification site website SRAMP to predict the m 6 A modification sites of the AKT. The predicted results showed that for a total forecast of 27 m 6 A modification sites (Fig. 6D), eight were very high confidence m 6 A sites; seven were high confidence m 6 A sites; four were moderate confidence m 6 A sites; and three were low confidence m 6 A sites, as detailed in Table 2. Further experiments will be carried out on the aforementioned sites.

Discussion
It is known that m 6 A methylation modifications play a key role in regulating the pluripotency, proliferation, metastasis, and immunity of tumor stem cells. METTL3, which is a type of m 6 A methylase, has been observed to have two-way effects in many types of cancers. METTL3 is highly expressed in AML [16,32,33] , liver cancer [22] , glioblastoma [34] , and other tumors, and can promote the occurrence and development of tumors. In contrast, it has been reported that METTL3 can also be used as a tumor suppressor to inhibit the growth and invasion of ovarian cancer [35] and prostate cancer [36] . However, there have been few studies conducted regarding MSC. However, high and m6A methylation had been confirmed to be a regulator of the AKT pathways [37] . However, there have not yet been reports made available regarding the regulations of PI3K/AKT signaling pathways by METTL3 expressions in the adipogenic differentiations of MSCs.
The results of the differentially expressed genes in this study's RNA sequencing showed that the lower expressions of the AML-MSCs mRNA gene numbers were more than the high expressions. In addition, the results of the m 6 A microarray showed that the low m6A modified mRNA gene numbers were greater than that of the high m6A modified genes in the AML-MSCs. Therefore, based on the fact that the m 6 A functions mainly depend on the "readers", this study speculated that IGF2BP proteins [38] may play important roles in subsequent functions. Also, according to the KEGG analysis of the RNA sequencing and the m6A microarray, it was determined that only the PI3K/AKT signaling pathways were characterized by low levels of m 6 A, while the transcription levels were significantly increased. Therefore, the PI3K/AKT signaling pathways modified by m6A during the process of adipogenic differentiation may function by interacting with YTHDF2 [39] , which will be further verified in the following experiments. At the present time, there are many m6A modification site prediction websites, such as MeT-BD [40] and WISTLE [41] , which are based on SRAMP, and are not widely applied as SRAMP. The SRAMP prediction site predicted different confidence sites based on the characteristic motif sequence of the m 6 A. Of course, whether or not the predicted sites are accurate still requires verification by experimental data.

Conclusions
In summary, our research demonstrated that METTL3 is highly related to the adipogenic differentiations of bone marrow MSCs. It was observed that decreased expressions of METTL3 in the AML-MSCs could significantly reduce the level of m 6 A modification of the PI3K/AKT signaling pathways. Furthermore, consequent activations of the PI3K/AKT signaling pathways could promote the adipogenic differentiations of MSCs, which could potentially mediate the chemo-resistance of AML (Fig. 7). The results obtained in this research will provide a theoretical basis for determining new targets of AML treatment from the perspective of bone marrow microenvironments, which will potentially have important clinical application value in the future.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
All clinical specimens were conducted in accordance with standard operating procedures under the consent and guidance of the Ethics Committee of the Fujian Medical University Union Hospital.
Consent to participate is not applicable.

Consent for publication
Not applicable