MAFB Promotes Cancer Stemness and Tumorigenesis in Osteosarcoma Through a Sox9-Mediated Positive Feedback Loop

Home , MAFB (gene), MAFG, NRL (gene), SATB2, SOX9

Author Manuscript Published OnlineFirst on March 31, 2020; DOI: 10.1158/0008-5472.CAN-19-1764 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

MAFB promotes cancer stemness and tumorigenesis in osteosarcoma through a Sox9-mediated positive feedback loop

Yanyan Chen1,†, Bin Wang2, †, Mengxi Huang1, †, Tao Wang 2, †, Chao Hu 3,†, Qin Liu 2, Dong Han 4, Cheng Chen1, Junliang Zhang5, Zhiping Li4, Chao Liu 6, Wenbin Lei 7,Yue Chang1, Meijuan Wu1,Dan Xiang1, Yitian Chen1, Rui Wang1, Weiqian Huang5, Zengjie Lei1 ,* and Xiaoyuan Chu1, *

1 Department of Medical Oncology, Affiliated Jinling Hospital, Medical School of Nanjing University, Nanjing, Jiangsu Province, People’s Republic of China 2 Department of Gastroenterology, Daping Hospital, Third Military Medical University (Ar- my Medical University), Chongqing, People’s Republic of China 3 Department of Orthopedics, 904 Hospital of PLA, North Xingyuan Road, Beitang Dis- trict, Wuxi, Jiangsu, People’s Republic of China 4 Department of Medical Oncology, Jinling Hospital, Nanjing Clinical School of Southern Medical University, Nanjing, Jiangsu Province, People’s Republic of China 5 Department of Orthopedics, Affiliated Jinling Hospital, Medical School of Nanjing Univer- sity, Nanjing, Jiangsu Province, People’s Republic of China 6 Department of Medical Oncology, Jinling Hospital, Nanjing Clinical School of Nanjing Medical University, Nanjing, Jiangsu Province, People’s Republic of China 7Department of Orthopedics, Tianshui Cooperation of Chinese and Western Medicine Hospi- tal, Tianshui, Gansu Province, People’s Republic of China

* To whom correspondence should be addressed. Xiaoyuan Chu, Tel: +86-25-80860131, Fax: +86-25-80860131, Email: [email protected]; Zengjie Lei, Email: leiz- [email protected] ; †The authors wish it to be known that, in their opinion, the first five authors should be regard- ed as joint First Authors.

Running title: MAFB-Sox9 positive feedback loop promotes osteosarcoma cell stemness

Keywords: osteosarcoma, transcriptional feedback loop, cell stemness, MAFB, Sox9

Downloaded from cancerres.aacrjournals.org on October 3, 2021. © 2020 American Association for Cancer Research. Author Manuscript Published OnlineFirst on March 31, 2020; DOI: 10.1158/0008-5472.CAN-19-1764 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Financial support: This work was supported by the National Natural Science Foundation of China [81572457 and 81872042 to Xiaoyuan Chu, 81702442 to Zengjie Lei, 81822032 to Bin Wang], the Natural Science Foundation of Jiangsu province [BK20170623 to Zengjie Lei], and the funding from the Army Medical University [Nos. 2019XQY19, 2018XLC2023 and 2019CXLCA001 to Bin Wang].

Conflicts of interest: The authors declare no conflicts of interest.

Word count: 4798 6 figures, 5 Supplementary Figures, 3 Supplementary Tables.

Significance

Transcription factors MAFB and Sox9 form a positive feedback loop to maintain cell stemness and tumor growth in vitro and in vivo, revealing a potential target pathway for therapeutic intervention in osteosarcoma..

ABSTRACT

Despite the fact that osteosarcoma is one of the most common primary bone malignancies

with poor prognosis, the mechanism behind the pathogenesis of osteosarcoma is only partial-

ly known. Here we characterized differentially expressed genes (DEG) by extensive analysis

of several publicly available gene expression profile datasets and identified MAFB as a key

transcriptional regulator in osteosarcoma progression. MAFB was highly expressed in tumor

tissues and required for proliferation and tumorigenicity of osteosarcoma cells. MAFB ex-

pression was elevated in osteosarcoma stem cells to maintain their self-renewal potential in

vitro and in vivo through upregulation of stem cell regulator Sox9 at the transcriptional level.

Sox9 in turn activated MAFB expression via direct recognition of its sequence binding en-

richment (SBE) motif on the MAFB locus, thereby forming a positive feedback regulatory

loop. Sox9-mediated feedback activation of MAFB was pivotal to tumorsphere-forming and

tumor-initiating capacities of osteosarcoma stem cells. Moreover, expression of MAFB and

Sox9 was highly correlated in osteosarcoma and associated with disease progression. Com-

bined detection of both MAFB and Sox9 represented a promising prognostic biomarker that

stratified a subset of osteosarcoma patients with shortest overall survival. Taken together,

these findings reveal a MAFB-Sox9 reciprocal regulatory axis driving cancer stemness and

malignancy in osteosarcoma and identify novel molecular targets that might be therapeutical- ly applicable in clinical settings.

Introduction

Osteosarcoma is one of the most common primary malignant bone tumors with two incidence

peaks, one in adolescence and another in the elderly population, especially those above 75 years of age (1). Chemotherapy combined with surgery has greatly improved clinical out-

comes of osteosarcoma patients, with the 5-year survival rate up to 60-70%, while ~20% of

osteosarcoma patients with metastasis continue to have poor prognosis (2).

Increasing clinical and experimental evidence indicates that osteosarcoma stem cells,

which derive from mesenchymal stem cells, may be the cellular origin of osteosarcomas (3).

Cancer stem cells (CSCs) share many similar properties with normal stem cells and are pri-

marily responsible for tumorigenesis in many cancers (4, 5). For example, a subpopulation of

self-renewing osteosarcoma cells, namely CSCs, are endowed with intrinsic capacities for

tumor initiation and drug resistance (6, 7). These CSCs are regulated by several key tran-

scription factors and signal pathways, such as Oct3/4, Sox2, Nanog, and Notch (8). Com-

pared to differentiated cancer cells, CSCs are generally more malignant and are critical de-

terminants of the response to chemotherapy and radiotherapy, and therefore the eradication of

osteosarcoma stem cells may be an effective treatment strategy (9, 10).

V-maf avian musculoaponeurotic fibrosarcoma oncogene homolog B (MAFB) is a

member of the MAF transcription factor family, containing basic leucine zipper domains

which bind to specific DNA elements (11). The seven MAF members are separated into two classes (small and large MAFs). Small MAF proteins (MAFF, MAFG, and MAFK) have

been shown to regulate antioxidant responses (12) whereas large MAFs (MAFA, c-MAF, and

MAFB) each contain a similar transactivation domain and are strongly oncogenic (13, 14). In

the hematopoietic system, MAFB induces myelomonocytic differentiation in immortalized

myoblasts and macrophage differentiation and maturation in mice (15). In podocyte differen-

tiation and the maintenance of progression, aberrant podocyte foot process formation is ob-

served in a MAFB-mutant zebrafish embryo model (15). There is also evidence that MAFB

regulates osteoclast genesis and epidermal keratinocyte differentiation (16, 17). Up- regulation of MAFB increases the risk of various human pathologies, such as diabetes and

atherosclerotic disorders (18, 19). In addition, MAFB has been shown to promote tumorigen-

esis, especially in the transformation of pancreatic β cells (14, 20). MAFB chromosomal

translocations occur more frequently in human myeloma cells, and high expression of MAFB

is observed in acute leukemia blast, hepatocellular carcinoma, and colorectal carcinoma pa-

tients (15, 18). Finally, MAFB has been shown to promote nasopharyngeal carcinoma cell

proliferation and migration (21).

Here, through high throughput bioinformatic analysis of publicly available transcriptome

datasets we identified MAFB as a novel regulator of osteosarcoma tumorigenesis. We

demonstrate that MAFB promotes tumorigenesis and self-renewal of osteosarcoma stem cells via a Sox9-mediated feedback activation loop, which could be exploited to eliminate the cul-

prit cells in osteosarcoma.

Materials and Methods

Collection and processing of microarray gene expression data

Microarray gene expression data of 10 normal and 107 osteosarcoma tissue samples were ex-

tracted from six datasets in the GEO (gene expression omnibus) database: GSE14359,

GSE16088, GSE16091, GSE14827, GSE12865, and GSE73166. The intersection genes of

these three platforms were identified (11,808 genes). Combat function was used for removing

batch effects in high-throughput experiments, the RMA (robust multiarray average) method

was used for data normalization using R software (R Foundation, Vienna, Austria), and the

limma package (R Foundation) was used to analyze differently expressed genes.

Patient specimens and Immunohistochemistry (IHC) staining

All osteosarcomas and adjacent tissues were collected from patients who were diagnosed as

OS at the department of Department of Surgery of Daping Hospital and Jinling Hospital. All

OS tissue was obtained from individual patients with written informed consent. The protocols

used in this study conformed to the ethical guidelines of the 1975 Declaration of Helsinki and

were approved by the Ethical Review Committees of Daping Hospital and Jinling Hospital.

Blocked tissue samples were cut into 5 μm sections and stained with hematoxylin and eosin,

then incubated with primary antibodies. Tumor staging was estimated according to the crite-

ria for histological classification proposed by the International Union Against Cancer. We followed a previously described protocol to quantify staining intensity(22). Five representa-

tive fields of a section were evaluated by two double-blinded pathologists. Final score of

MAFB and Sox9 in each sample was obtained by multiplying the strength score by the distri- bution score. The cutoff score in various analyses was 8 for both anti- MAFB and anti-

MAFB staining intensity (high expression, IHC scoring ≥ 8; low expression, IHC scoring <

8).

Cell culture

Human osteoblast cell lines (hFOB1.19) and osteosarcoma cell lines (MG63, U-2OS, 143B,

HOS) were purchased from the Shanghai Cell Collection (Shanghai, China). All cell lines were characterized by short tandem repeat profiling (STR) within 6 months. 143B and MG63 were maintained in MEM medium, while HOS and U-2OS was maintained in RPMI-1640 and MoCoy’s 5a media respectively. hFOB1.19 was maintained in DMEM medium. All medium (HyClone, Waltham, MA, USA) contained high glucose, 10% fetal bovine serum (FBS;

HyClone) and 1% antibiotic/antimycotic solution (Sigma-Aldrich, St. Louis, MO, USA). All cells were cultured at 37°C in a humidified atmosphere containing 5% CO2.

Multiple early-passage frozen stocks of each cell line in this study were stored under liquid nitrogen: all cell lines were used in experimentation for no longer than 10 passages from thaw. Cells were used in the described experiments for approximately 6 months. All cell lines were routinely tested for Mycoplasma contamination using MycoProbe Mycoplasma Detec- tion Kit (R&D Systems, Inc.), and any contaminated cell line was treated with Plasmocin treatment (InvivoGen), confirmed by negative detection of Mycoplasma before being used again. The most recent testing was 3 months ago.

Lentivirus vectors and cell infection

The target gene sequence of small hairpin (sh) RNA (shMAFB-a: TACTGGATGGCGAG-

CAACTACCAGCAGAT, shMAFB-b: TCACCAAGGACGAGGTGATCCGCCTGAAG, shSox9-a: GTGCGCGTCAACGGCTCCAGCAAGAACAA, shSox9-b: CAGCGAAC-

GCACATCAAGACGGAGCAGCT) were cloned into the pLVT-Vector (SBO Medical Bio- technology, Shanghai, China), and used for the generation of lentiviruses which were mixed with lentiviral transfection enhancer (Sigma-Aldrich) and applied to indicated cell lines.

Puromycin was then used to select and establish stable expression or knockdown cell lines.

Western blotting

Total protein was extracted from osteosarcoma cell lines and clinical samples using M-PER

Mammalian Protein Extraction Reagent (Thermo Scientific, Waltham, MA). Whole cell ly-

sates were separated by sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS-

PAGE) and transferred to nitrocellulose membrane (Bio-Rad, Hercules, CA, USA). Mem- branes were blocked with 5% skim milk or bovine serum albumin (BSA) and then incubated

with primary antibodies (see Supplementary Table 1A) and subsequently with secondary an-

tibodies (see Supplementary Table 1A). Protein bands were visualized using the Immobilon

Western Chemiluminescent HRP Substrate detection reagent (Millipore, Billerica, MA,

USA). GraphPad Prism 5.0 was used to measure the gray intensity of protein bands and nor-

malized the expression of control protein ( -actin).

RT-PCR assays

Total RNA was isolated from osteosarcoma cell lines and clinical samples using TRIzol rea-

gent (Invitrogen, Carlsbad, CA, USA) and then treated with DNase. Using a PrimeScript™

RT kit (Takara, Japan), RNA was reverse transcribed into cDNA to detect relative mRNA

levels using qPCR (BioRad). The fold changes of target genes were normalized to β-actin.

The primer sequences for qPCR are provided in Supplementary Table 1B. All experiments

were repeated three times.

Cell Counting Kit-8 (CCK8) assay

Cells were seeded in a 96-well plate with a density of 3,000 cells per well and treated with

indicated conditions. At 24 h intervals, the medium was removed and then incubated with

CCK8 (Dojindo Molecular Technologies, Rockville, MD, USA) for 60 min. The absorbance

was determined at 450 nm using a Varioskan Flash (Thermo Scientific).

Colony formation assay

For colony formation assays, approximately 500 cells were plated into 60 mm culture dishes and incubated for 14 days. Colonies were stained with 0.5% Crystal Violet in 6% glutaralde- hyde solution. The diameter of colonies consisting of ≥ 300 μm were counted.

Soft agar colony formation assay

Cells were harvested and suspended as single cells in culture medium (105 cells per mL). 60- mm dishes were precoated with 0.75% agar in culture medium, and single cell suspensions were mixed with 0.35% agarose in culture medium and seeded into precoated 60-mm dishes.

After 12–30 days of incubation at 37°C, in 5% CO2, colonies were stained, photographed, and counted.

RNA sequencing and analysis

Total RNA was extracted using using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and sequenced with the cBot Cluster Generation System using a TruSeq PE Cluster Kit v3-cBot-

HS (Illumina, San Diego, CA, USA). Raw reads were mapped to the aligned reference ge- nome using Hisat2, version 2.0.5 (https://ccb.jhu.edu/software/hisat2/index.shtml) and read counts for each transcript were calculated using featureCounts, version 1.5.0-p3

(https://www.rdocumentation.org/packages/Rsubread/versions/1.22.2/topics/featureCounts).

Differential gene expression analysis was performed using the DESeq2 R package. Gene On- tology analysis was performed using the clusterProfiler R package.

Flow cytometric analysis

Trypsin digestion was used to produce individual cell suspensions, followed by incubation in blocking solution [phosphate-buffered saline + 1% bovine serum albumin (BSA) + 10%

FBS]. Cells were then centrifuged, and pellets were resuspended and incubated with primary

antibody (see Supplementary Table 1A). Cells were then analyzed or sorted using a BDAria

II Sorter (BD Biosciences).

Sox9 and MAFB promoter and luciferase activity assays

Cells were transfected with a firefly reporter vector containing MAFB or Sox9 promoter, and

the Renilla luciferase reporter vector (pRL-TK) using Lipofectamine 2000 (Invitrogen). Cell

samples were collected after 48 h and both MAFB or Sox9 firefly luciferase and Renilla lu-

ciferase activities were detected using a dual-luciferase reporter assay kit (BMG Labtech,

Ortenberg, Germany). The Renilla luciferase fluorescence intensity was utilized as an internal

control.

Chromatin immunoprecipitation (ChIP)

Cells were cross-linked with 1% formaldehyde, lysed, and sonicated into DNA fragments

between 200‒1,000 bp. The DNA fragments were immunoprecipitated with antibodies

against MAFB or Sox9, and PCR with specific primers to detect the relative sequence bind-

ing enrichment (SBE) motifs were used (Supplementary Table 1B).

Tumor sphere formation assay

Osteosarcoma cells were isolated in ultra-low attachment 24-well plates (Corning) with 200 cells per well and cultured in stem cell medium [DMEM/F12 medium (Thermo Fisher) con-

taining 20 ng/mL of EGF, 20 ng/mL FGF, and 10 ng/mL of HGF (PeproTech, Rocky Hill,

NJ, USA), B27 supplement (Invitrogen, Grand Island, NY, USA), 4 µg/mL of insulin (Sig-

ma-Aldrich), and 1% methyl cellulose (Sigma-Aldrich)] for 2 weeks. Fresh medium was ex- changed every 4 days, and the average diameters of tumor spheres were determined using a microscope.

Tumorigenicity in vivo and limiting dilution assay (LDA)

Nude mice, 4 weeks of age, were injected subcutaneously with stable clones of cell suspensions in Matrigel™ (1:1) (106 cells/mouse). Xenografts were allowed to grow for 5 weeks before being harvested for further analysis. The tumor volumes were then measured and calculated using the formula: V = (length × width2)/2. The tumors were stored in 10% buffered formalin for further analysis.

For LDA of tumorigenicity in vivo, cells were diluted at defined doses (106 cells/100 µl to

103 cells/100 µl) and injected subcutaneously into nude mice. The tumor-initiating cell frequency was calculated using ELDA software (http://bioinf.wehi.edu.au/software/elda/)(23).

Data representing the number of cells in each culture, number of cultures tested and number of positive cultures was entered to determine the ‘Tumor-initiating cell frequency’. The cutoff value for determining tumor formation was 100 mm3 (24).

All animal experiments were approved by the Institutional Animal Care and Use Com- mittee (IACUC) of the Jinling Hospital and the Army Medical University.

Immunofluorescence and confocal microscopy

Approximately 104 cells were grown on cover glass and cultured overnight, and then fixed with 4% paraformaldehyde. After permeabilization of membranes using 0.5% Triton X-100, nonspecific antibody binding was blocked by pre-incubation with 1% BSA. Specific antibodies were then incubated overnight followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibodies and visualized using confocal microscopy.

Availability of data and materials

Microarray gene expression data of ten normal and 107 osteosarcoma tissue samples were

extracted from six datasets in the GEO (gene expression omnibus) database:

https://www.ncbi.nlm.nih.gov/geo/. R software (R Foundation, Vienna, Austria) is a free

software environment for statistical computing and graphics. (https://www.r-project.org). The

datasets and the computer code used and analyzed during the current study are available from

the corresponding author upon request.

Statistical analysis

All experiments were performed three or more times independently under similar conditions.

Statistical analyses were performed using SPSS statistical software for Windows, version

17.0 (SPSS, Chicago, IL, USA). Data were expressed as the mean ± SD or percentage, and

analyzed using Student’s t-test, χ2 test, or analysis of variance. Survival data were estimated

using the Kaplan-Meier method and analyzed using the log-rank test. The association be-

tween factors for survival was determined by univariate and multivariate analyses using Cox

regression analysis. A value of p < 0.05 was considered statistically significant.

Results

Elevated expression of MAFB in osteosarcoma

To better characterize the molecular underpinnings of osteosarcoma tumorigenesis, we sys-

temically analyzed mRNA expression profiles from osteosarcomas and normal tissues. Ex-

pression profiles of 10 normal and 107 osteosarcoma samples were extracted from six da-

tasets in the GEO database: GSE14359, GSE16088, GSE16091, GSE14827, GSE12865, and

GSE73166. Since these data were from three different platforms, we intersected the detected

genes and found 11,808 genes in common. Combat function was used for removing batch

effects in high-throughput experiments, and the RMA method was used for data normaliza-

tion using R software. Using the limma package, we identified 1,833 differentially expressed

genes (DEGs) with P values < 0.05 and fold changes > 1.5 (Supplementary Table 2A), and

shown in a volcano plot (Fig. 1A). We listed the top 100 upregulated and downregulated genes (Fig. S1A). Using GO (gene ontology) enrichment and pathway enrichment analysis of

these DEGs (Fig. S1B–C and Fig. 1B), we found that the DEGs were closely associated with

transcriptional activity. Among these DEGs, there were 112 transcription factors (Supple-

mentary Table 2B), with the top ten upregulated and downregulated transcription factors in

these DEGs shown in Fig. 1C.

Previous studies revealed that SATB2 and HEY1 were related to osteosarcoma migration

and metastasis (25, 26), TRPS1 was associated with multi-drug resistance of osteosarcoma,

and HMGB2 suppressed p53 transcriptional activity in osteosarcoma cells (27, 28). However,

less is known about the role of MAFB in osteosarcomas, therefore we focused on the role of

MAFB in osteosarcoma tumorigenesis. MAFB was highly expressed in osteosarcoma sam-

ples utilized in the GEO datasets analyzed above (Fig. 1D). From samples collected from os-

teosarcoma patients, we observed a similar pattern of elevated MAFB expression in tumor

tissue compared to adjacent normal tissue (Fig. 1E–F). Together, these findings demonstrate that MAFB is highly expressed in osteosarcoma tumors isolated from patients.

MAFB promotes growth and tumorigenicity of osteosarcoma cells.

To test whether MAFB plays a role in the pathophysiology of osteosarcoma, we first exam-

ined the expression of MAFB in osteosarcoma cell lines. We found that MAFB protein levels

were higher in 143B and HOS cells, but lower in U-2OS and MG63 cells (Fig. S2A). Deple-

tion of MAFB led to impaired colony formation capacity of cells suspended in soft agar (Figs.

2A, B and S2B-D), as well as cell proliferation in two-dimensional culture conditions (Figs.

2C, D and S2E). Consistently, depletion of MAFB in these cells induced smaller subcutane-

ous tumors in nude mice (Fig. 2E and S2F).

To further validate a potential oncogenic role of MAFB in osteosarcoma, we ectopically

expressed MAFB in MG63 and U-2OS cells (Figs. 2F and S2G). MAFB-overexpressing in

these cells led to increased colony formation in soft agar (Figs. 2G and S2H) and were more

proliferative in anchorage-dependent growth conditions (Figs. 2H-I and S2I). Consequently,

MAFB-overexpressing cells show an increased ability to induce the formation of solid tu-

mors in mice (Fig. 2J). Expression of MAFB in these subcutaneous tumors was validated by

immunohistochemistry (IHC) analysis of the dissected tumors (Fig.S2F and S2J). Together,

these results demonstrate that MAFB plays an important oncogenic role in osteosarcoma in

vitro and in vivo.

MAFB binds to, and activates, the Sox9 promoter.

CSCs have been identified as the “cell-of-origin” of osteosarcomas (29), thus we hypothe-

sized that a significant increase of tumorigenicity in MAFB-expressing cells may be due to a

higher ratio of CSCs. To assess the potential effects of MAFB on cancer cell stemness, we

performed RNA-seq analysis of MAFB-depleted and control cells. DEGs between these two

groups are shown in a volcano plot (Fig. 3A and Supplementary Table 2C). According to gene ontology analysis, we defined a list of markers related to CSC population maintenance

(Supplementary Table 2D). Six CSC markers were differentially expressed after depletion of

MAFB (Fig. 3B). In the analyses of GEO datasets, we found 18 CSC markers differentially

expressed among osteosarcoma and normal samples; with the top five upregulated markers

being CD24, SPARC, Sox9, KDR, and Sox4 (Fig. S3A). Since the only CSC marker showing

differential expression both in GEO datasets and RNA-seq analysis was Sox9 (Fig. 3C), we anticipated that Sox9 may be a key downstream target of MAFB to control cancer cell stemness.

Consistent with this notion, ectopic expression of MAFB in MG63 cells led to increased

mRNA levels of Sox9, as well as CD44 and CD24, but not other CSC markers of osteosarcoma stem cells (Fig. S3B, C). Western blotting confirmed that overexpresssion of MAFB

increased the protein levels of CD44, CD24, and Sox9 (Fig. S3D). Conversely, depleting

MAFB in HOS cells significantly downregulated both the mRNA and protein levels of CD44,

CD24, and Sox9 (Fig. S3E–F).

We next determined whether there were any binding motifs of MAFB in the promoter

sequence of the Sox9 gene. As shown in Fig. S3G, there were six potential sequence binding

enrichment (SBE) motifs for the MAFB transcription factor in the human Sox9 gene within

the first 2kb upstream to the transcription start site. To confirm which SBEs was directly rec-

ognized by MAFB, we generated six constructs with Sox9 modified promoter sequences

(P1‒P6) upstream of luciferase. We found that the relative luciferase activities of P1, P3, P4,

and P5 were increased in the presence of MAFB (Fig. 3D), while depletion of MAFB signifi-

cantly decreased the activities of the same promoter sequences (Fig. S3H). These four constructs shared each contained the SBE1 element. Using ChIP, we found that binding of

MAFB to SBE1 and SBE3 elements was decreased in HOS cells after depletion of MAFB

(Fig. 3E). This was further confirmed in Lv-MAFB MG63 cells (Fig. 3F). These data sug- gests that MAFB binds to SBE1 and SBE3 of the Sox9 promoter to transcriptionally activate

Sox9 expression.

Sox9 forms a feedback loop to enhance the transcriptional activity of MAFB.

Sox9 is also a transcription factor and upregulated in aggressive osteosarcoma tissues with

poor prognoses (30), therefore we speculated that MAFB and Sox9 may regulate one another

and form a reciprocal feedback loop, as shown in other similar circumstances such as GA-

TA3/ZEB2 or p53/SET (31, 32). To this end, we established cell lines with Sox9 either de-

pleted or overexpressed (Fig. S3I, J). Using these cell lines, we found that MAFB levels posi-

tively correlated with Sox9. MAFB protein and mRNA levels were decreased upon depletion

of Sox9 (Fig. S3K) and increased upon ectopic expression of Sox9 (Fig. S3L).

To examine whether MAFB was directly regulated by Sox9, we analyzed the Sox9

ChIP-seq dataset from HT29 human colorectal cancer cells (GEO dataset GSE63629) and

retrieved positive Sox9 binding peaks on the promoter of MAFB (Fig. 3G). Moreover, we al-

so found that there were six potential SBEs in the promoter domain of MAFB (Fig. S3M).

We generated six constructs with different promoter sequences of MAFB upstream of lucifer-

ase. and found that the luciferase activity of P1, P2, and P3 were increased in the presence of

ectopic Sox9 (Fig. 3H), but significantly decreased in the Sox9-depleted cells (Fig. S3N).

These results were further confirmed by ChIP assays showing that the binding of Sox9 to

SBE1 and SBE5 were significantly decreased following depletion of Sox9 (Fig. 3I) and in-

creased with ectopically expressed Sox9 (Fig. 3J). Thus, Sox9 bound to SBE1 and SBE5 of

the MAFB promoter sequence and promoted MAFB expression.

Furthermore, we detected transcriptional activity of both the MAFB and Sox9 promoters

by luciferase report assay with overexpression of either MAFB or Sox9. Not only the tran-

script activity of Sox9 promoter increased after MAFB overexpression, but also the expres-

sion levels of MAFB were induced by overexpression MAFB itself (Fig. 3K). Similarly, both

Sox9 and MAFB promoter activity were enhanced by Sox9 overexpression (Fig. 3K). Given that both proteins are able to bind the promoters of each other, we concluded that MAFB and

Sox9 constituted positive transcriptional regulatory loop in osteosarcoma cells.

The MAFB-Sox9 reciprocal regulatory axis maintains cancer stemness and tumorigen-

icity.

In light of the essential role of Sox9 in CSCs, we suspected that MAFB may regulate cancer

stemness through up-regulating Sox9 expression in osteosarcoma. To support this notion,

both MAFB and Sox9 were highly expressed and co-localized in the nuclei of CD44+CD24+

osteosarcoma stem cells compared to CD44-CD24- differentiated cells (Fig. 4A-B). Moreo- ver, ectopic expression of MAFB in CD44-CD24- cells promoted sphere formation, resulting

in an increase in both sphere number and size, while this phenotype was inhibited by deplet-

ing Sox9 (Fig. 4C-D and S4A-C). Consistently, ectopic expression of MAFB induced an ex-

pansion of CD44+CD24+ cells, while depleting Sox9 minimized the stem cell pool (Fig. 4E and S4D). Moreover, using xenograft tumor formation and limiting dilution assays in vivo,

we found that MAFB-expressing cells formed larger tumors with higher tumor-initiating cell

frequencies, a phenotype that was dependent, in part, on its downstream target Sox9 (Fig. 4F-

G and S4E).

To further validate the potential roles of the MAFB and its downstream target Sox9 in

CSCs, we depleted MAFB in CD44+CD24+ cells using two independent shRNAs (Fig. S4F-

G). Loss of MAFB in CSCs led to impaired sphere formation capacity while re-introducing

Sox9 rescued this stem cell function (Fig. 4H–I and S4H). Moreover, flow cytometry analysis

revealed that the pool of CD44+CD24+ cells were reduced upon depletion of MAFB and re-

covered upon overexpression of Sox9 (Fig. 4J and S4I). Using in vivo xenograft tumor

growth and limiting dilution tumor formation assays, we found that depleting MAFB in CSCs

inhibited tumor formation and self-renewing potential, which could be restored by reintro-

ducing Sox9 (Fig. 4K and S4J-K). IHC staining of xenograft tumor tissues confirmed that

MAFB and Sox9 enhanced expression of both MAFB and Sox9 in vivo (Fig. S4L-M), which

was consistent with the model that MAFB and Sox9 form a functional regulatory loop in os-

teosarcoma stem cells.

We next asked how Sox9-mediated feedback activation of MAFB contributed to CSC

maintenance. To this end, introducing Sox9 in CD44-CD24- cells enhanced sphere formation

and self-renewal in a limiting dilution assay, as well as expansion of the CSC pool, while de-

pleting MAFB attenuated these stemness properties (Fig. 5A-C, Fig. S5A–D). Moreover, de-

pleting endogenous Sox9 in CD44+CD24+ cells inhibited tumorsphere formation and self-

renewing capacity of CSCs, which was partially recovered by reintroducing MAFB (Fig. 5D–

F), suggesting that Sox9 functions in a MAFB-dependent manner to sustain stemness proper-

ties of osteosarcoma cells.

Furthermore, we examined whether the MAFB-Sox9 feedback regulatory axis contributed to chemoresistance, a hallmark of CSCs. We found that enforced expression of either

MAFB or Sox9 in CD44-CD24- cells conferred resistance to Cisplatin, a commonly used

chemotherapy agent (Fig. S5E). Consistently, Cisplatin treatment enriched CD24+CD44+

CSCs that expressed high levels of MAFB and Sox9 (Fig. 6A, Fig. S5F). Together, these data

demonstrate that reciprocal regulation between MAFB and Sox9 in osteosarcoma stem cells

to maintain their stemness potential, chemoresistance and tumorigenicity.

MAFB expression levels correlates with Sox9 in a subgroup of human osteosarcoma tis-

sues.

To further validate the correlation between MAFB and Sox9 expression in CSCs, immuno-

fluorescence staining revealed that both MAFB and Sox9 were highly expressed and co- localized in a subset of human osteosarcoma cells in vivo (Fig. 6A). Analysis of datasets from the GEO database also demonstrated that the mRNA levels of Sox9 was higher in tumors

compared to normal tissue (Fig. 6B). Consistently, the protein levels of both Sox9 and MAFB

were co-elevated in osteosarcoma compared to adjacent normal tissues. Moreover, expression

of MAFB positively correlated with Sox9 in osteosarcoma specimens (Fig.6C-D). Immuno-

fluorescence staining of primary tumor tissues also revealed that MAFB and Sox9 were co- expressed by a subset of tumor cells (Fig. S5G).

Clinical pathological analyses revealed that both the expression of MAFB and Sox9 were

correlated with higher histological grades. Moreover, increased expression of MAFB and

Sox9 was observed in patients with high grade tumors (G3 and G4 patients) compared to

those with low grade tumors (G1 and G2 patients) (Supplementary Table 3A). Furthermore,

higher expression levels of either MAFB or Sox9 in osteosarcoma is associated with shorter

patient survival (Fig. 6E–F and Supplementary Table 3B). For subgroup analysis, patients

were divided into four subgroups according to the expression levels of MAFB and Sox9. Ow-

ing to the limit numbers of patient samples with MAFBlowSox9high and MAFBhighSox9low, we

were unable to draw statistically significant conclusions from these two subgroups (Fig. 6G,

Fig. S5H). However, patients with MAFBhighSox9high tumors show poor survival compared

with patients with MAFBlowSox9low tumors (Fig. 6H). Taken together, MAFB and Sox9 form

a positive feedback loop fueling CSC self-renewal to support malignant growth and tumor-

igenesis (Fig. 6H), suggesting that this regulatory axis maybe a unique therapeutic vulnerabil-

ity for a subgroup of osteosarcoma.

Discussion

Osteosarcoma remains a major challenge due to poorly defined mechanisms of tumorigenesis

and limited treatment options. Here, we identified a critical role of MAFB in osteosarcoma

tumorigenesis via a Sox9-mediated positive feedback loop, which reciprocally promote ex-

pression of MAFB to induce cell proliferation, maintenance of CSCs, and tumorigenesis.

MAFB was highly expressed and positively correlated with osteosarcoma progression, as re-

vealed by analysis of GEO datasets and paired clinical osteosarcoma samples. Previous stud-

ies demonstrated that other large MAF family members, such as MAFA, c-MAF, and NRL,

also regulate tumorigenesis in humans (11). For example, c-MAF enhances cell proliferation and metastasis in lymphoma and breast cancer (18, 33) . Moreover, MAFB promotes cell cy-

cle progression and tumor growth in multiple myeloma, nasopharyngeal carcinoma, hepato-

cellular carcinoma, and colorectal carcinoma (15, 21, 34, 35). Consistent with these studies,

MAFB supports proliferation and colony formation of osteosarcoma cells in vitro and tumor

growth in vivo in nude mice, suggesting a previously unrecognized oncogenic role in human

osteosarcoma.

Our study identifies Sox9 as a functional downstream transcriptional target of MAFB.

Sox9 is known to play versatile roles in tumorigenesis. It is up-regulated in lung, breast, and

liver cancers and its expression levels are frequently associated with poor clinical outcomes

(36-39). However, Sox9 also inhibits tumorigenesis in melanoma (40). For osteosarcomas,

Sox9 expression is up-regulated and associated with disease progression and poor prognosis

(30). However, it remains unclear how Sox9 expression is regulated by upstream mecha-

nisms. In the present study, we identified that Sox9 is transcriptionally activated by MAFB

through direct binding to the SBE1 and SBE3 elements on the Sox9 promoter. Moreover,

Sox9 could also promote MAFB expression by directly binding to the SBE4 element on the

MAFB promoter, thereby generating a reciprocally regulated positive feedback loop in osteo-

sarcoma cells.

We demonstrate that the MAFB-Sox9 feedback loop as an essential signaling node con-

trolling self-renewal and tumorigenicity of osteosarcoma stem cells. Sox9 is a transcription

factor that plays important roles in normal stem cells and CSCs. Sox9, as well as two addi-

tional CSC markers, CD24 and CD44, were elevated upon ectopic expression of MAFB,

suggesting that the MAFB-Sox9 feedback loop may be an important regulatory pathway of

cancer stemness in osteosarcoma. To test this hypothesis, we purified osteosarcoma stem

cells using CD24 and CD44 as biomarkers (2, 41) and detected elevated expression levels of

MAFB in CD24+CD44+ CSCs. By overexpressing or knocking down MAFB expression, we found that MAFB maintains tumor sphere formation and tumor initiation capacity of CSCs in

vitro and in vivo through activating it downstream target Sox9. Moreover, we observed that

Sox9-associated self-renewal of osteosarcoma stem cells were largely dependent on MAFB.

Thus, coordinated reciprocal activation between MAFB and Sox9 represents an essential

mechanism underlying stemness maintenance in human osteosarcoma.

Furthermore, both MAFB and Sox9 may serve as potential biomarkers to inform clinical

outcomes of osteosarcoma patients. IHC analyses of osteosarcoma tissues revealed that

MAFB expression positively correlates with Sox9. More importantly, expression of both

MAFB and Sox9 were associated with poor outcomes of these patients. Strikingly, the pa-

tients with MAFBhigh Sox9high tumors displayed the shortest survival, as compared to those

with MAFBlowSox9low tumors. MAFBlowSox9high or MAFBhighSox9low tumors also show a

trend towards better survival compared to MAFBhigh Sox9high although the small sample size

precluded a statistically robust analysis. Thus, combined detection of the MAFB-Sox9 feed-

back loop represents a potential alternative method by which to stratify osteosarcoma patients

with increased prognostic ability of disease progression and mortality.

In conclusion, our study identifies a MAFB-Sox9 reciprocal regulatory loop that sustains stem cell properties of osteosarcoma cells in vitro and in vivo. Given their essential roles in maintaining osteosarcoma CSCs, this signaling axis could be exploited to design novel therapeutic strategies against this deadly bone malignancy. Moreover, simultaneous examination of MAFB and Sox9 expression status may help provide prognostic value in clinical settings.

ACKNOWLEDGEMENT Not applicable.

REFERENCES 1. Isakoff MS, Bielack SS, Meltzer P, Gorlick R. Osteosarcoma: Current Treatment and a Collaborative Pathway to Success. J Clin Oncol. 2015;33:3029-35. 2. Otoukesh B, Boddouhi B, Moghtadaei M, Kaghazian P, Kaghazian M. Novel molecular insights and new therapeutic strategies in osteosarcoma. Cancer Cell International. 2018;18. 3. Brown HK, Tellez-Gabriel M, Heymann D. Cancer stem cells in osteosarcoma. Cancer Lett. 2017;386:189-95. 4. Wang T WH, Liu S,Lei Z,Qin Z,Wen L,Liu K,Wang X,Guo Y,Liu Q,Liu L,Wang J,Lin L,Mao C,Zhu X,Xiao H,Bian X. SMYD3 controls a Wnt-responsive epigenetic switch for ASCL2 activation and cancer stem cell maintenance. Cancer Lett. 2018;430:11-24. 5. Lytle NK, Ferguson LP, Rajbhandari N, Gilroy K, Fox RG, Deshpande A, et al. A Multiscale Map of the Stem Cell State in Pancreatic Adenocarcinoma. Cell. 2019;177:572- 86.e22. 6. Brown HK, Tellez-Gabriel M, Heymann D. Cancer stem cells in osteosarcoma. Cancer Letters. 2017;386:189-95. 7. Takahashi N, Nobusue H, Shimizu T, Sugihara E, Yamaguchi-Iwai S, Onishi N, et al. ROCK Inhibition Induces Terminal Adipocyte Differentiation and Suppresses Tumorigenesis in Chemoresistant Osteosarcoma Cells. Cancer Research. 2019. 8. Cooper J, Giancotti FG. Integrin Signaling in Cancer: Mechanotransduction, Stemness, Epithelial Plasticity, and Therapeutic Resistance. Cancer Cell. 2019;35:347-67. 9. Yan G-N, Lv Y-F, Guo Q-N. Advances in osteosarcoma stem cell research and opportunities for novel therapeutic targets. Cancer Letters. 2016;370:268-74. 10. Miao Y, Zhang H, Pan Y, Ren J, Ye M, Xia F, et al. Single-walled carbon nanotube: One specific inhibitor of cancer stem cells in osteosarcoma upon downregulation of the TGFbeta1 signaling. Biomaterials. 2017;149:29-40. 11. Pai EL, Vogt D, Clemente-Perez A, McKinsey GL, Cho FS, Hu JS, et al. Mafb and c- Maf Have Prenatal Compensatory and Postnatal Antagonistic Roles in Cortical Interneuron Fate and Function. Cell Rep. 2019;26:1157-73 e5. 12. Blank V. Small Maf proteins in mammalian gene control: mere dimerization partners or dynamic transcriptional regulators? Journal of molecular biology. 2008;376:913-25. 13. Nishizawa M, Kataoka K, Vogt PK. MafA has strong cell transforming ability but is a weak transactivator. Oncogene. 2003;22:7882-90. 14. Pouponnot C, Sii-Felice K, Hmitou I, Rocques N, Lecoin L, Druillennec S, et al. Cell context reveals a dual role for Maf in oncogenesis. Oncogene. 2006;25:1299-310. 15. Yang L, Liu Y, Zhu L, Xiao M. DNMT3A R882 mutation is associated with elevated expression of MAFB and M4/M5 immunophenotype of acute myeloid leukemia blasts. Leukemia & lymphoma. 2015;56:2914-22. 16. Smink JJ, Begay V, Schoenmaker T, Sterneck E, de Vries TJ, Leutz A. Transcription factor C/EBPbeta isoform ratio regulates osteoclastogenesis through MafB. The EMBO journal. 2009;28:1769-81. 17. Miyai M, Hamada M, Moriguchi T, Hiruma J, Kamitani-Kawamoto A, Watanabe H, et al. Transcription Factor MafB Coordinates Epidermal Keratinocyte Differentiation. The Journal of investigative dermatology. 2016;136:1848-57. 18. Pettersson AM, Acosta JR, Bjork C, Kratzel J, Stenson B, Blomqvist L, et al. MAFB as a novel regulator of human adipose tissue inflammation. Diabetologia. 2015;58:2115-23. 19. Hamada M, Nakamura M, Tran MT, Moriguchi T, Hong C, Ohsumi T, et al. MafB promotes atherosclerosis by inhibiting foam-cell apoptosis. Nature communications. 2014;5:3147.

20. Tsuchiya M, Misaka R, Nitta K, Tsuchiya K. Transcriptional factors, Mafs and their biological roles. World journal of diabetes. 2015;6:175-83. 21. Li Y, Min D, Wang K, Yin S, Zheng H, Liu L. MicroRNA152 inhibits cell proliferation, migration and invasion by directly targeting MAFB in nasopharyngeal carcinoma. Molecular medicine reports. 2017;15:948-56. 22. Qin ZY, Wang T, Su S, Shen LT, Zhu GX, Liu Q, et al. BRD4 Promotes Gastric Cancer Progression and Metastasis through Acetylation-Dependent Stabilization of Snail. Cancer research. 2019;79:4869-81. 23. Hu Y, Smyth GK. ELDA: Extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. Journal of Immunological Methods. 2009;347:70-8. 24. Castaño Z, San Juan BP, Spiegel A, Pant A, DeCristo MJ, Laszewski T, et al. IL-1β inflammatory response driven by primary breast cancer prevents metastasis-initiating cell colonization. Nature Cell Biology. 2018;20:1084-97. 25. Seong BK, Lau J, Adderley T, Kee L, Chaukos D, Pienkowska M, et al. SATB2 enhances migration and invasion in osteosarcoma by regulating genes involved in cytoskeletal organization. Oncogene. 2015;34:3582-92. 26. Tsuru A, Setoguchi T, Matsunoshita Y, Nagao-Kitamoto H, Nagano S, Yokouchi M, et al. Hairy/enhancer-of-split related with YRPW motif protein 1 promotes osteosarcoma metastasis via matrix metallopeptidase 9 expression. British journal of cancer. 2015;112:1232-40. 27. Jia M, Hu J, Li W, Su P, Zhang H, Zhang X, et al. Trps1 is associated with the multidrug resistance of osteosarcoma by regulating MDR1 gene expression. FEBS letters. 2014;588:801-10. 28. Stros M, Ozaki T, Bacikova A, Kageyama H, Nakagawara A. HMGB1 and HMGB2 cell-specifically down-regulate the p53- and p73-dependent sequence-specific transactivation from the human Bax gene promoter. The Journal of biological chemistry. 2002;277:7157-64. 29. Dean M, Fojo T, Bates S. Tumour stem cells and drug resistance. Nature reviews Cancer. 2005;5:275-84. 30. Zhu H, Tang J, Tang M, Cai H. Upregulation of SOX9 in osteosarcoma and its association with tumor progression and patients' prognosis. Diagnostic pathology. 2013;8:183. 31. Si W, Huang W, Zheng Y, Yang Y, Liu X, Shan L, et al. Dysfunction of the Reciprocal Feedback Loop between GATA3- and ZEB2-Nucleated Repression Programs Contributes to Breast Cancer Metastasis. Cancer Cell. 2015;27:822-36. 32. Wang D, Kon N, Lasso G, Jiang L, Leng W, Zhu WG, et al. Acetylation-regulated interaction between p53 and SET reveals a widespread regulatory mode. Nature. 2016;538:118-22. 33. Pavlovic M, Arnal-Estape A, Rojo F, Bellmunt A, Tarragona M, Guiu M, et al. Enhanced MAF Oncogene Expression and Breast Cancer Bone Metastasis. Journal of the National Cancer Institute. 2015;107:djv256. 34. Yu H, Jiang HL, Xu D, Jin JZ, Zhao ZM, Ma YD, et al. Transcription Factor MafB Promotes Hepatocellular Carcinoma Cell Proliferation through Up-Regulation of Cyclin D1. Cellular physiology and biochemistry : international journal of experimental cellular physiology, biochemistry, and pharmacology. 2016;39:700-8. 35. Yang LS, Zhang XJ, Xie YY, Sun XJ, Zhao R, Huang QH. SUMOylated MAFB promotes colorectal cancer tumorigenesis. Oncotarget. 2016;7:83488-501. 36. Drivdahl R, Haugk KH, Sprenger CC, Nelson PS, Tennant MK, Plymate SR. Suppression of growth and tumorigenicity in the prostate tumor cell line M12 by overexpression of the transcription factor SOX9. Oncogene. 2004;23:4584-93.

37. Zhou CH, Ye LP, Ye SX, Li Y, Zhang XY, Xu XY, et al. Clinical significance of SOX9 in human non-small cell lung cancer progression and overall patient survival. Journal of experimental & clinical cancer research : CR. 2012;31:18. 38. Riemenschnitter C, Teleki I, Tischler V, Guo W, Varga Z. Stability and prognostic value of Slug, Sox9 and Sox10 expression in breast cancers treated with neoadjuvant chemotherapy. SpringerPlus. 2013;2:695. 39. Guo X, Xiong L, Sun T, Peng R, Zou L, Zhu H, et al. Expression features of SOX9 associate with tumor progression and poor prognosis of hepatocellular carcinoma. Diagnostic pathology. 2012;7:44. 40. Passeron T, Valencia JC, Namiki T, Vieira WD, Passeron H, Miyamura Y, et al. Upregulation of SOX9 inhibits the growth of human and mouse melanomas and restores their sensitivity to retinoic acid. The Journal of clinical investigation. 2009;119:954-63. 41. La Noce M, Paino F, Mele L, Papaccio G, Regad T, Lombardi A, et al. HDAC2 depletion promotes osteosarcoma's stemness both in vitro and in vivo: a study on a putative new target for CSCs directed therapy. Journal of experimental & clinical cancer research : CR. 2018;37:296.

FIGURES LEGENDS Figure 1. Expression levels of MAFB is higher in osteosarcoma tissues compared to normal tissues. A. Volcano plot of DEGs between 10 normal and 107 osteosarcoma samples from the GEO database. DEGs with P < 0.05 and fold changes > 1.5 shown in green. Not significant DEGs were shown in red, with P ≥ 0.05 or fold change ≤ 1.5 (DESeq2, R software). (DEGs: differentially expressed genes) B. GO enrichment analysis of the molecular functions of these 1,833 DEGs. C. Heatmap of the top 10 upregulated and downregulated transcription factors represented in the set of identified DEGs. D. The mRNA expression levels of MAFB in 10 normal and 107 osteosarcoma samples from the GEO database. (*P < 0.05) E–F. The mRNA levels (E) or the protein levels (F) of MAFB in 15 paired tumor and adjacent tissue samples from osteosarcoma patients (*P < 0.05). The western blotting bands were quantified and normalized to beta-actin to indicate relative levels of MAFB expression.

Figure 2. Overexpression of MAFB enhances malignant growth of osteosarcoma cells in vitro and in vivo. A. Protein levels of MAFB were detected in HOS and 143B cells expressing control shRNA (shGFP), shMAFB (shMAFB-a, shMAFB-b), or vehicle (Blank ctrl). Numbers above blots are quantification of MAFB protein expression normalized to beta-actin. B. The cell proliferation capacity detected by the soft agar assay in HOS cells expressing control shRNA (shGFP), shMAFB (shMAFB-a, shMAFB-b) or vehicle (Blank ctrl). (*P < 0.05, **P < 0.01) C. Cell Counting Kit-8 assays indicate proliferation of HOS and 143B cells was inhibited following MAFB depletion. (*P < 0.05) D. Colony formation assay in HOS cells transfected with control shRNA (shGFP), shMAFB (shMAFB-a, shMAFB-b) or vehicle (Blank ctrl). (*P < 0.05, **P < 0.01) E. In vivo subcutaneous xenograft tumor formation performed on 4-week nude mice with cells expressing control shRNA (shGFP) or shMAFB (shMAFB-b). Tumors were harvested after 5 weeks and representative images are shown (left) and tumor volumes were analyzed (right). (n=6 per group) (**P < 0.01)

F. Protein levels of MAFB were detected in MG63 and U-20S cells stably expressing Lv-ctrl, Lv-MAFB, or vehicle (Blank). Numbers above blots are quantification of MAFB protein expression normalized to beta-actin. G. Cell proliferation capacity was detected using soft agar assay in MG63 cells expressing Lv-ctrl, Lv-MAFB, or vehicle (Blank). (**P < 0.01) H. Cell Counting Kit-8 assay showing proliferation of MG63 and U-20S cells were promoted by MAFB overexpression. (*P < 0.05) I. Colony formation assay showing cell proliferation capacity of MG63 cells expressing Lv- ctrl, Lv-MAFB, or vehicle (Blank). (**P < 0.01) J. In vivo subcutaneous xenograft tumor formation was conducted using cells expressing Lv- ctrl or Lv-MAFB. Representative images are shown (left) and the tumor volumes were analyzed (right). (n=6 per group). (**P < 0.01)

Figure 3. MAFB and Sox9 form a transcriptional feedback loop to co-elevate their expression. A. Volcano plot of DEGs analyzed by RNA-seq between cells expressing control shRNA (shGFP) or shMAFB (shMAFB-b). (n=3 per group) B. Heatmap of expression levels of CSC markers identified in DEGs between cells expressing control shRNA (shGFP) or shMAFB (shMAFB-b). (n=3 per group) C. Venn diagrams of the common CSC markers between GEO datasets and RNA-seq. D. Luciferase assays showing promoter activities of six Sox9 promoter sequences in MG63 cells expressing Lv-ctrl or Lv-MAFB. (*P < 0.05) E–F. ChIP assay measuring MAFB binding to Sox9 SBEs in HOS cells expressing control shRNA (shGFP), or shMAFB (shMAFB-b) (E), and MG63 cells expressing Lv-ctrl or Lv- MAFB (F). (*P < 0.05) G. ChIP-seq of Sox9 in HT29 human colorectal cancer cells from the GEO dataset, GSE63629. H. Luciferase assays showing different promoter activities of six MAFB promoter sequences in MG63 cells transfected with Lv-ctrl or Lv-Sox9. (*P < 0.05) I-J. ChIP assay of Sox9 binding to MAFB SBEs in HOS cells expressing control shRNA (Kd-ctrl) or shSox9 (shSox9-b) (I), and MG63 cells expressing Lv-ctrl or Lv-Sox9(J). (*P < 0.05) K. Luciferase assays of Sox9 promoter and MAFB promoter in the MAFB overexpressing cells (left) or Sox9 overexpressing cells (right). (*P < 0.05)

Figure 4. MAFB promote stemness and tumorigenicity of osteosarcoma cells in part through activating its downstream target Sox9. A. Immunofluorescence of MAFB and Sox9 in CD44+CD24+ or CD44-CD24- cells sorted MG63 cells. B. Protein levels of MAFB were detected in CD44+CD24+ or CD44-CD24- MG63, U-2OS, 143B, and HOS cells. Numbers above blots are quantification of MAFB protein expression normalized to beta-actin. C–D. The average sphere diameter (C) and number (D) generated by CD44-CD24- MG63 or U-2OS cells expressing Lv-ctrl, Lv-MAFB, and shSox9-b individually or together. (**P < 0.01) E. Flow cytometry analysis of the percentage of CD44+CD24+ cells in CD44-CD24- MG63 cells expressing Lv-ctrl, Lv-MAFB and shSox9-b individually or together. (**P < 0.01) F. Quantification of volumes of subcutaneous xenograft tumors formed by CD44-CD24- MG63 cells expressing Lv-ctrl, Lv-MAFB, and shSox9-b individually or together. (n=6 per group) (**P < 0.01) G. Quantification of tumor-initiating cell frequency generated by MG63 CD44-CD24- cells expressing Lv-ctrl, Lv-MAFB, shSox9-b individually or together. (n=6 per group) H-I. The average diameters (H) and numbers of spheres (I) generated by CD44+CD24+ MG63 or U-2OS cells expressing control shRNA (shGFP), shMAFB (shMAFB-b), Lv-Sox9 individually or together. (**P < 0.01) J. Flow cytometry analysis of the percentage of CD44+CD24+ cells in CD44+CD24+ MG63 or U-2OS cells expressing control shRNA(shGFP), shMAFB (shMAFB-b), Lv-Sox9 individually or together. (**P < 0.01) K. Quantification of volumes of subcutaneous xenograft tumors formed by CD44+CD24+ MG63 cells transfected with control shRNA (shGFP), shMAFB (shMAFB-b), Lv-Sox9 individually or together. (n=6 per group) (*P < 0.05)

Figure 5. Sox9-mediated feedback activation of MAFB maintains stem-like properties of osteosarcoma cells. A-B. Average sphere diameter (A) and number (B) generated by CD44-CD24- MG63 or U- 2OS cells expressing Lv-ctrl, Lv-Sox9, and shMAFB-b alone or both lv-Sox9, and shMAFB- b. Representative images of spheres are shown (A, left). (**P < 0.01)

C. Quantification of tumor-initiating cell frequency generated by CD44-CD24- MG63 or U- 2OS cells expressing Lv-ctrl, Lv-Sox9, and shMAFB-b individually or together. (n=6 per group) D–E. Average sphere diameter (D) and number (E) generated by CD44+CD24+ MG63 or U- 2OS cells expressing Lv-ctrl, Lv-MAFB, and shSox9-b individually or together. Representa- tive images of spheres are shown (D, left). (**P < 0.01) F. Quantification of tumor-initiating cell frequency generated by CD44+CD24+ MG63 or U- 2OS cells expressing Lv-ctrl, Lv-MAFB, shSox9-b individually or both. (n=6 per group) (Three repeats, *P < 0.05, **P < 0.01, ***P < 0.001).

Figure 6. MAFB expression levels were positively correlated with Sox9 in a subgroup of osteosarcoma tissues. A. Immunofluorescence analysis of MAFB and Sox9 in MG63 cells from subcutaneous xenografts. B. The mRNA levels of Sox9 in 10 normal and 107 osteosarcoma samples from the GEO database. (**P < 0.01) C. IHC analysis of MAFB and Sox9 in tumor tissue and adjacent normal tissue from patients. D. IHC score of Sox9 expression was positively correlated with MAFB levels. (n=83) (***P < 0.001) E. Kaplan-Meier survival analysis of overall survival in MAFB high-expressed patients and MAFB low-expressed patients. (n=83) F. Kaplan-Meier survival analysis of overall survival in Sox9 high-expressed patients and Sox9 low-expressed patients. (n=83) G. Kaplan-Meier survival analysis of overall survival in MAFBlowSox9low and MAFBhigh- Sox9high patients. H. Summary of the MAFB-Sox9 positive feedback loop and its effects in promoting cancer stem cell stemness and promoting xenograft tumor growth in vivo.

MAFB promotes cancer stemness and tumorigenesis in osteosarcoma through a Sox9-mediated positive feedback loop

Yanyan Chen, Bin Wang, Mengxi Huang, et al.

Cancer Res Published OnlineFirst March 31, 2020.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-19-1764

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2020/03/31/0008-5472.CAN-19-1764.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet Manuscript been edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2020/03/31/0008-5472.CAN-19-1764. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.