CD9 is a leukemic stem cell-specic marker in human acute myeloid leukemia

Background: Leukemia stem cells (LSCs) are responsible for the initiation, progressing and relapse of acute myeloid leukemia (AML). Therefore, the therapy strategy of targeting LSCs is hopeful to eradicate AML. In this study, we aimed to identify LSCs-specic surface markers and uncover the underlying mechanism of AML LSCs. Methods: Microarray gene expression data were used to investigate the candidate AML-LSCs specic markers. CD9 expression was evaluated by ow cytometry (FC) in AML cell lines, patients with AML and normal donors. The biological characteristics of CD9-positive (CD9 + ) cells were analyzed by in vitro proliferation, chemotherapeutic drug resistance, migration and in vivo xenotransplantation assays. The molecular mechanism involved in CD9 + cell function was investigated by gene expression proling. Effects of alpha-2-macroglobulin (A2M) on CD9 + cells were analyzed in the aspects of proliferation, drug resistance and migration. Results (cid:0) CD9, as a cell surface protein, is specically expressed on AML LSCs, but barely can be detected on normal hematopoietic stem cells (HSCs). CD9 + cells exhibits more resistance to chemotherapy drugs and higher migration potential than CD9-negative (CD9 - ) cells. More importantly, CD9 + cells possess the ability to reconstitute human AML in immunocompromised mice and promote leukemia growth, suggesting CD9 + cells dene the LSCs population. Furthermore, we identied A2M as an important role in CD9 + LSCs stemness maintenance. Knock down of A2M impairs drug-resistance and migration of CD9 + cells. Conclusion: Our ndings suggested that CD9 is a new biomarker of AML LSCs and may serve as a promising therapeutic target. cells;


Background
Acute myeloid leukemia (AML) is the most common acute leukemia in adults, accounting for approximately 80% of cases in this group that is characterized by in ltration of the bone marrow, blood, and other tissues by proliferative, clonal, abnormally differentiated, and occasionally poorly differentiated cells of the hematopoietic system [1,2]. AML is caused by the disorders of the hematopoietic system which is commonly treated by chemotherapeutic and hematopoietic stem cell transplantation [3].
However 43% of patients eventually end up with relapse after complete remission in young adult patients [4]. Residual rare leukemia stem cells (LSCs) are the major cause of recurrence of AML, which possess chemoresistant and the ability to self-renew and differentiate thus reconstitute AML. Therefore, the LSCs concept has inspired the design of innovative treatment strategies for AML aiming at targeting LSCs hidden in cancers.
To identify the potential therapeutic targets, many groups have reported cell surface proteins preferentially expressed on AML LSCs, including CD47 [5], CD44 [6], CD96 [7], CD123 [8,9] CD99 [10] and TIM-3 [11,12]. During the past few years, some strategies for targeting LSCs markers antibody or immune cells have already been tested in patients, but still face the problems of toxicity and LSCs resistance. Therefore, more speci c LSCs markers still need to be explored.
Here we identi ed a potential AML LSCs speci c molecule CD9 by analyzing three microarray data of AML LSCs and a minimal residual disease (MRD) expression pro ling. As one member of the tetraspanins family, CD9 is the third most abundant protein on the platelet surface and is required for the release of microparticles from coated-platelets [13,14]. Furthermore, it was reported that CD9 plays an important role in cell adhesion, movement, differentiation, proliferation, apoptosis and resistance to chemotherapy [15][16][17][18][19]. Although CD9 has been reported to identify cancer stem cells in several types of cancers including pancreatic cancer, glioblastoma and B-acute lymphoblastic leukemia, and is related to the prognosis of AML [15,[20][21][22]. However, biological characteristic and regulatory mechanism of CD9 + AML LSCs remains to be elucidated.
In this study, we found the high-expression of CD9 in AML patient LSCs and extremely-low-expression in normal hematopoietic stem cells (HSCs). CD9 + cells exhibited stem cell characteristics, including drug resistance, migration ability and remodel human AML in immuneocompromised mice. Mechanically, we identi ed that A2M plays a crucial role in CD9 + LSCs maintenance by transcription pro ling analysis.
Down-regulation of A2M impairs drug-resistance and migration ability of CD9 + cells. In summary, our data suggested CD9 is a potential new target for AML therapy and A2M controls stemness characteristics of CD9 + AML LSCs.

Data sources
Three AML LSCs sequencing chips and one AML minimal disease residue (MDR) sequencing chip were sourced from the publicly available database.

Cell lines and leukemic cells from patients
Leukemia cells including THP-1 and KG-1α were obtained from America Type Culture Collection (ATCC); U937 was obtained from the Chinese Academy of Sciences, Shanghai, China; MV-4-11 was obtained from the Query Network for Microbial Species of China; MOLM-13 was obtained from the COBIOER, Nanjing, China; HL-60 was obtained from the JOINN Labs, Suzhou, China.
Primary AML cells were obtained from bone marrows of patients with AML who have all signed the informed consent according to the protocols approved by the Institutional Review Board of the Southwest Hospital, Army Medical University. All the patients' information was in supplementary table S1.
Antibodies, Cell Staining, and Sorting All the antibodies for FC were purchased from BioLegend. For analyses of CD9 expression in AML cell lines, cells were stained with PE anti-human CD9 Antibody (312106). For primary AML cells, cells were stained with FITC anti-human CD3/CD19 antibody (300306, 392508), PerCP anti-human CD45 antibody (368506), APC/Cyanine7 anti-human CD34 antibody (343614), APC anti-human CD38 antibody (356606) and PE anti-human CD9 antibody (312106). Brie y, cells were harvested and suspended with 50µl staining/washing buffer (PBS including 1% FBS), then stained with antibodies and incubated for 30 minutes at 4℃. Cells were washed with staining/washing buffer and suspended in buffer for ow cytometry or cell sorting.

Migration assay
The migration AML cells were tested by Falcon® Permeable Support for 24-well Plate with 8.0 µm Transparent PET Membrane (Corning). 2x10 5 CD9 + and CD9cells were suspended in 200μl RPMI 1640 medium (without FBS) and seeded in the upper chambers, respectively. 900μl medium with 20% FBS was added to bottom chamber of each well. After the 6-hour incubation, migrated cells were counted by trypan blue at the indicated time points [23].

Drug resistance assay
The drug resistances of AML cells were assessed by MTS assay. 1x10 5 of AML cells were seeded in 96well microtiter plates (NEST) with different concentrations of cytarabine (10, 100μg/ml) in 100μl medium [24]. After 24-hour incubation, 10μl MTS (Promega) was added to each well. After 2 hours' incubation, the plate was measured at wavelength of 490nm with microplate reader (Thermo Fisher Scienti c).

Cell proliferation assay
The AML cells were seeded in 96 well plates with 5x10 3 cells/well. At the indicated time, 10μl MTS (Promega) was added to cells and then detected the absorbance at 490 nm after 2 hours by the microplate reader (Thermo Fisher Scienti c). The medium without cells was used as a negative control [25].

Transplantation of AML cells into immunode cient mice
The animal study was performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee of Southwest Hospital, Army Medical University. NOG mice (Vital River Laboratories) at age of 6 to 8 weeks were used for xenogeneic transplantation assays. THP-1 cells were infected with lentivirus co-expressing luciferase and GFP according to the previous method [26]. Sorted CD9 + and CD9 -THP-1 cells were transplanted into NOG mice via tail vein injection. The progression of leukemia was monitored by bioluminescence imaging with In Vivo Imaging System (IVIS) Spectrum (Perkin Elmer, USA) and Living Image Software for IVIS (Perkin Elmer).

Microarray analysis
Three pairs of sorted CD9 + and CD9primary AML cells (patient 6, patient 7, and patient 8) were investigated by the BGISEU-500 platform for gene expression. In brief, total RNA was extracted with TRIzol (Invitrogen) according to the manufacturer's instructions from sorted CD9 + and CD9primary AML cells, respectively. Then all samples are submitted to the BGISEQ-500 platform for RNA sequencing (RNAseq).

Real time PCR
Total RNA was extracted from CD9 + and CD9 -THP-1 cells using RNAiso Plus (Takara), and then Singlestranded cDNA was synthesized with the PrimeScript RT Reagent Kit (Takara) according to the manufacturer's instruction. Quantitative-PCR (Q-PCR) analysis was performed using SYBR premix Ex Taq (Takara). The sequence of the primers and the housekeeping gene GAPDH were all from Primer Bank (https://pga.mgh.harvard.edu/primerbank/).

Western blotting
Cells were washed by ice PBS and lysed with RIPA buffer added with protease and phosphatase inhibitor cocktail (Roche). Primary antibody for GAPDH (2118), A2M (4929) were purchased from Abcam and EGR1 (100899-T32) from Sino Biological.
A2M network with CD9 utilizing the GeneMANIA database Datasets, including physical interactions, pathway, and genetic interactions, were collected from the public domain GeneMANIA database. The dataset relevant to A2M and CD9 network was produced from the GeneMANIA database (http://www.genemania.org).

Chromatin Immunoprecipitation
Chromatin immunoprecipitation (ChIP) assay was performed according to the manufacturer's instruction (Cell Signaling Tech-nology, 9005S). Anti-EGR1 antibody was purchased from Cell Signaling Tech-nology (4154S). PCR primers for the CD9 promoter were listed in Supplementary table S2.

Statistical analysis
All experiments were repeated at least in triplicate. Collected data were analyzed with GraphPad Prism 8.0 software (GraphPad Software, Inc., San Diego, CA) and estimated variation was taken into account for each group of data and indicated as SEM or SD in each gure legend. Comparison between two groups was carried out with unpaired Student's t test (two tailed), and differences among more than two groups were determined by a one-way ANOVA followed by Newman-Keuls test. Difference with p < 0.05 was considered statistically signi cant. Results 1. CD9 is highly expressed in CD34 + CD38cell population of AML patients almost not expressed in normal HSCs To identify cell surface markers selectively expressed on AML LSCs, we surveyed three sets of AML LSCs sequencing chips data [11,27,28]. Fifty-ve commonly up-regulated genes were detected by comparing the three datasets (Fig. 1A). MRD (Minimal Residual Disease) in AML patients refers to residual cancer cells after treatment, and it was reported to be a powerful and independent prognostic factor in treatment outcome [29][30][31][32][33][34][35][36][37][38]. Whether the expression level of the fty-ve LSCs up-regulated genes was also upregulated in the MRD microarray has lead us to a further investigation. By analyzing the MRD microarray data, we found that among these fty-ve genes, twenty-three genes were signi cantly up-regulated in the MRD microarray, including eight cell surface proteins CD33, CD96, HCK, C3AR1, TYROBP, FCER1G, LPXN and CD9 (Fig. 1A, 1B, supplementary table S3). Interestingly, CD96 has already been reported as a LSCsspeci c marker in human AML [7], which strongly con rmed the reliability of our guess.
Among these membrane proteins, CD9 was the most intensively up-regulated gene. We rstly analyzed the CD9 expression level in AML cell lines by ow cytometry (FC). The results showed that CD9 + cells account for 11.14% (1.9%-42.3%, Fig. 1C, S1). Then, we examined CD9 expression in the bone marrow of AML patients and normal donors. With the previously described gating strategy in ow cytometry, we rstly gated away CD3 and CD19 positive T cells and B cells, then focused on CD45 low SSC low population, and analyzed CD34 + CD38cells within CD45 low SSC low , then further investigated CD9 expression in CD3 -CD19 -CD45 low CD34 + CD38 - [10] (Fig. 1D). Our data revealed that the average percentage of CD9 + cells was 12.9% (5.84%-36%) in AML blasts. It has been shown that the AML LSCs mainly reside within the CD34 + CD38fraction of leukemic cells. CD9 + cells account for 62.76% (37.2%-87.1%) in CD34 + CD38cells ( Fig. 1E), suggesting that CD9 + cells were enriched in AML LSCs. To determine whether CD9 expression could distinguish LSCs from normal HSCs, we examined CD9 expression in the bone marrow of normal donors. The results showed that CD9 expression was very low in normal bone marrow cells (1.7%, 1.1%-2.4%) and normal HSCs (0.9%, 0.3%-1.3%) (Fig.1F). These data together demonstrate that the cell surface protein CD9 could be a promising marker for targeting AML LSCs.

CD9 + cells exhibited LSCs characteristics
To investigate the biological function of CD9 + cells, we isolated CD9 + cells and CD9cells from THP-1 and AML patients by uorescence-activated cell sorting. Cell proliferative assay showed that there was no signi cant difference in the proliferative capacity of CD9 + cells and CD9cells either from THP-1 cells or AML patients ( Fig. 2A, 2B, p=0.9669, p=0.9005), which was consistent with previous reports that stem cells do not exhibit superior proliferation capacity in the normal condition [39][40][41]. In addition, cell cycles were analyzed by FC, and the results demonstrated that there are no differences between CD9 + and CD9cells (Fig. S2). To assess the ability of drug resistance, we treated the sorted cells with different doses of Ara-C (10μg/ml and 100μg/ml) which is a commonly used as leukemia chemotherapy drug, and checked cell survival rate after 24 hours. The results showed that CD9 + cells were more resistant to Ara-C than CD9cells (Fig. 2C, 2D). Furthermore, the transwell migration assay showed CD9 + cells exhibited higher migration potential than CD9cells (Fig. 2E, 2F).
To study the function of CD9 + AML cells in vivo, THP-1 cells were stably infected with lentivirus coexpressing luciferase and GFP to facilitate subsequent observation of leukemia growth in vivo. 1×10 6 CD9 + cells and CD9cells from THP-1 were respectively injected into NOG mice via tail vein. Due to the severe development of leukemia, mice were sacri ced on day 50 and the results showed that CD9 + cells exhibited superior proliferation capacity than CD9cells in vivo (Fig. 3A, 3B). The proportion of CD9 + cells in the bone marrow of mice were tested by ow cytometry and the results showed that CD9 + cells injected mice bone marrow contained a large number of in ltrating CD9 + cells (50.9%, 17.5%-73.5%).
Unexpectedly, varied degrees of CD9 + cells were also contained in CD9cells-injected mice bone marrow (21.02%, 4.27%-33.9%) (Fig. 3C), which may explain why CD9mice can also form leukemia. In addition, the existence of the same phenomenon in the peripheral blood of mice was observed (Fig. 3D). Previous studies have shown that cancer stem cells and differentiated tumor cells can be transformed into each other in tumor microenvironment [42,43]. Futhermore, survival research showed that mice survived for a shorter period of time after being transplanted with THP-1 CD9 + cells compared with mice transplanted with THP-1 CD9cells (Fig. S3). In conclusion, CD9 + cells display LSCs characteristics, drug resistance, increased capacity of migration and promoting cancer progression.

A2M is expressed in CD9 + cells at high levels
To investigate the molecular mechanisms involved in CD9 + LSCs maintenance, we performed global gene expression pro les in CD9 + cells and CD9cells from three AML patients by cDNA microarray. The Venn diagram was used to analyze the differential genes that were up-regulated in these 3 gene sets, and 52 differential genes that were commonly up-regulated in the CD9 + population were detected (Fig. 4A, 4B).
We further performed gene ontology analysis of the 52 genes, among which the top cluster was genes involved in the extracellular matrix organization, including A2M, SULF2, TGFB1, LRP1, MMP9, SERPINE1 and CRISPLD2 (Fig. 4C). We then con rmed the expression levels of these extracellular matrix-associated genes by real-time PCR in THP-1 (Fig. 4D). We focused on A2M not only because it's expression level in CD9 + cells was at high level, but also because activation of A2M signals was reported to promote proliferation and survival of cancer cells [44]. We also con rmed the high expression of A2M protein in CD9 + cells of THP-1 and HL-60 by Western blotting (Fig. 4E).

A2M regulates CD9 + LSCs maintenance
To further investigate the connection between CD9 and A2M, the GeneMANIA webserver were applied to predict their interactions in the network with the parameters limited to physical interactions, genetic interactions, and pathways to score nodes and source organism Homo sapiens as additional parameters (Fig. 5A). From the Gene MANIA network, we found that A2M has networked with CD9. To test whether A2M regulates CD9 expression, the expression of A2M in CD9 + cells was knocked down by short hairpin RNAs (shRNAs). The results showed that A2M knockdown signi cantly reduced the expression levels of EGR1 and CD9 (Fig. 5B, 5C). EGR1, as an important transcription factor, is a node in the network of A2M and CD9 (Fig. 5A). The results revealed that A2M possibly regulates CD9 expression by regulating its downstream protein EGR1 and this conclusion was con rmed by ChIP (Fig. S4). Functionally, even though knockdown of A2M had no effect on the proliferative of CD9 + cells (Fig.5D), but signi cantly increased the sensitivity of CD9 + cells to Ara-C treatment and attenuate CD9 + cells migration, compared with control groups (Fig.5E, 5F). Therefore, we concluded that A2M is an upstream gene that regulates CD9 gene expression through EGR1 and controls AML LSCs characteristics (Fig. 5G).

Discussion
Cancer stem cells (CSCs) drive tumor initiation, progression and metastasis. AML is a clonal malignant disorder derived from a small number of LSCs. LSCs could be the ultimate cellular target to cure human AML. Scientists are dedicated to searching speci c LSCs markers, which can effectively distinguish between LSCs and normal HSCs. Many molecules were reported to be differently expressed on AML LSCs, such as CD47, CD44, CD96, TIM3, CD99 and CD123 [5-7, [9][10][11], however, some of these markers are not speci c for AML LSCs. For example, targeting CD123 antibody impairs cytokine signaling and is toxic to common myeloid precursors (CMPs) [9], and targeting CD44 antibody disrupts blast-niche interactions [6].
To nd a more speci c marker of AML LSCs, three RNA-sequencing data of LSCs and an AML MRD microarray data were analyzed. CD9, the most intensively upregulated membrane molecule, was selected as a candidate marker for AML LSCs, and has been reported to be involved in several types of CSCs, including pancreatic cancer stem cells, breast cancer stem cells, ovarian cancer stem cells, glioblastoma stem cells, LSCs in B-acute lymphoblastic leukemia [15,20,22,45,46].
As a member of the tetraspanin superfamily, CD9 was rst identi ed by Kersey et al as the human hematopoietic progenitor cell surface antigen p24 using a monoclonal antibody that bound to acute lymphoblastic leukemia cells [47]. CD9 has been reported to be expressed in 40% of human AML samples and associated with clinical outcomes in AML [21]. In this study, we demonstrate that CD9 is highly expressed in the CD34 + CD38 -AML LSCs, and extremely low or no expression in normal HSCs, which could serve as a hopeful therapeutic target in AML. Nevertheless, the evaluation of targeting-CD9 therapy still requires further study.
Understanding the underlying mechanisms of CSCs maintenance also provides potential for patient care and improved prognosis. For example, Hedgehog (Hh), Notch and Wnt signaling exhibit signi cant crosstalk during embryogenesis. Inhibitors of Hh and Notch pathways have achieved considerable progress in early phase clinical trials [48]. To identify the mechanisms that regulate the characteristics of CD9 + LSCs, we performed RNA-sequencing of CD9 + and CD9cells from three AML patients. We veri ed A2M was involved in regulating stemness characteristics of CD9 + cells. Importantly, it has been reported that activated A2M signals promote proliferation and survival of cancer cells predominantly through cell surface GRP78 (CS-GRP78) [49]. {Misra, 2005 #2}Therefore, we believe that A2M signal play a crucial role in AML and the treatment of targeting-A2M signaling pathway will bring new hope to AML patients.

Conclusion
Our study demonstrated that CD9 was highly expressed in AML LSCs, but almost not expressed in normal HSCs, which allowed it to serve as a potential LSCs marker. CD9-positive cells possess CSCs characteristics, including drug-resistance, migration ability and promoting leukemia progression. Importantly, we found that A2M signal plays a crucial role in the stemness maintenance of CD9-positive cells in AML. Overall, our results found CD9 a new target for AML therapy.

Declarations
Chen, Gang Heng and Guiqin Wang: Acquisition of patient specimens, collection and assembly of data, nal approval of manuscript; Jun Chen and Yongchun Zhao: Collection and assembly of data, nal approval of manuscript; Huailong Xu and Yuanli Ni: Data analysis and interpretation, nal approval of manuscript; Jiatao Li and Yingzi Zhang: Data analysis and interpretation, nal approval of manuscript;