Fanconi Anemia Mesenchymal Stromal Cells‐Derived Glycerophospholipids Skew Hematopoietic Stem Cell Differentiation Through Toll‐Like Receptor Signaling

Fanconi anemia (FA) patients develop bone marrow (BM) failure or leukemia. One standard care for these devastating complications is hematopoietic stem cell transplantation. We identified a group of mesenchymal stromal cells (MSCs)‐derived metabolites, glycerophospholipids, and their endogenous inhibitor, 5‐(tetradecyloxy)−2‐furoic acid (TOFA), as regulators of donor hematopoietic stem and progenitor cells. We provided two pieces of evidence that TOFA could improve hematopoiesis‐supporting function of FA MSCs: (a) limiting‐dilution cobblestone area‐forming cell assay revealed that TOFA significantly increased cobblestone colonies in Fanca−/− or Fancd2−/− cocultures compared to untreated cocultures. (b) Competitive repopulating assay using output cells collected from cocultures showed that TOFA greatly alleviated the abnormal expansion of the donor myeloid (CD45.2+Gr1+Mac1+) compartment in both peripheral blood and BM of recipient mice transplanted with cells from Fanca−/− or Fancd2−/− cocultures. Furthermore, mechanistic studies identified Tlr4 signaling as the responsible pathway mediating the effect of glycerophospholipids. Thus, targeting glycerophospholipid biosynthesis in FA MSCs could be a therapeutic strategy to improve hematopoiesis and stem cell transplantation. Stem Cells 2015;33:3382–3396


INTRODUCTION
Fanconi anemia (FA) is an inherited disorder associated with hematopoietic aplasia and cancer predisposition [1][2][3]. FA is genetically heterogeneous and the clinical phenotypes associated with FA are the result of deficiency of any of the 16 FA genes (FANCA-Q) [4][5][6][7]. Although physical signs appear from birth and early childhood, bone marrow (BM) failure is typically seen between ages 5 and 15 and in later ages leading to myelodysplastic syndrome and acute myeloid leukemia [8][9][10]. One standard care for these devastating complications is hematopoietic stem cell transplantation (HSCT). However, little is known about the interaction between healthy donor HSCs and FA BM microenvironment (niche). Recent HSC-BM niche interaction studies have demonstrated that nestin-expressing mesenchymal stromal cells (MSCs) constitute an essential HSC niche component [11,12]. Adipocytes, one of the niche compartments, act as predominantly negative regulators of HSCs [13]; while osteoblasts and chondroblasts are known to support HSCs [14]. Although the role of majority of these cellular constituents forming the niche in the BM is becoming clear, the metabolism of these cell types in the context of hematopoietic support during disease state is still unclear.
To address this question, we used an untargeted metabolomics approach that provides a comprehensive platform to identify metabolites whose levels are altered between wild-type (WT) and FA MSCs. Metabolomics has become a powerful technique for understanding the small-molecule basis of biological processes either in physiological or pathological conditions [15]. We show here that a group of MSCs-derived metabolites, glycerophospholipids, and their endogenous inhibitor, 5-(Tetradecyloxy)22-furoic acid (TOFA), are aberrantly produced by FA MSCs. To investigate the effect of these metabolites on hematopoietic-supporting function, we have modeled FA HSCT using ex vivo coculture followed by cobblestone area-forming cell (CAFC) and BM transplantation (BMT) assays and demonstrated that suppression of glycerophospholipid biosynthesis by TOFA or Lipin1 knockdown rescued differentiation skew of donor HSC and progenitor cells (HSPCs).

CAFC Assay
Confluent WT, Fanca2/2, and Fancd22/2 MSCs in 35 mm culture dish (BD Falcon, San Jose, CA) were overlaid with WT BMMCs to allow the precursor cells forming hematopoietic clones under the stromal layers. The cells were cocultured at 378C, 5% CO 2 , and were fed weekly by changing half of the medium. Phase-dark hematopoietic clone was imaged under phase-contrast images were taken at 320 objective and the area was analyzed with image J software.

Limited Dilution Assay
Limiting dilution assay (LDA) of a LSK cells included the use of five dilutions (0, 10, 30, 90, 270, and 810) differing with a factor of 3, and 10 wells per cell concentration. Three different LDA experiments were performed with independently derived MSCs from WT, Fanca2/2, and Fancd22/2 mice. A well was scored as "positive" if contained one or more cobblestone areas and "negative" if contained no cobblestone areas. Cobblestone area is at least six cells (in proximity of each other) growing underneath the stroma. Although cobblestone-like cells appear as phase dark, these cells appear as nonrefractile in 96-well plates because of the deflection of light. Only dilutions with both negative and positive wells are informative for frequency analysis.

Metabolome Profiling
Metabolites were extracted from three independently derived, 98% pure Fanca2/2, Fancd22/2, and WT MSCs. MSCs were washed with ice cold Dulbecco's Phosphate Buffered Saline (DPBS) twice to remove any culture media. Cells were collected into 300 ml LC/MS-grade H 2 O containing 1 mM HEPES and 1 mM EDTA (pH 7.2). Samples were vortexed for 30 seconds, and incubated 1-2 minutes in boiling water and subsequently in LN 2 for 1 minute. Samples were then thawed on ice and normalized based on the protein content. Two milliliters of 2208C metabolite extraction solution containing Methanol, Acetonitrile, and H 2 O at a ratio of 2:2:1 was added to each sample and vortexed for 1 minute. Samples were then incubated at 48C for 30 minutes. Samples were centrifuged at 1,500g for 10 minutes. The supernatants (2 ml total) were pooled in an HPLC vial (Sigma# 27115-U, St. Louis, MO) and dried under forced N 2 at room temp before reconstituted for LC-Q-TOF-MS analysis. All pure standards were  from Sigma Aldrich. Samples were resuspended in 50 ml of 50:50 water/acetonitrile solutions for mass spectrometry analysis. Untargeted metabolomics was performed on the MSCs extract to identify metabolites whose levels are altered in Fanca2/2 and Fancd22/2 compared to WT. Samples were analyzed at Scripps Center for Metabolomics and Mass Spectrometry, La Jolla, CA. Using liquid chromatography quadrupole time-of-flight mass spectrometry (LC-Q-TOF-MS), hundreds of peaks with a unique m/z ratio and retention time were detected in Fanca2/2, Fancd22/2, and WT MSCs. Each peak, termed a metabolomic feature, is characterized on the basis of its accurate mass, retention time, and tandem mass spectral fragmentation pattern using the METLIN metabolite database. The data were then analyzed with the bioinformatics program XCMS Online [21], widely used XCMS software that is freely available at https://xcmsonline.scripps.edu.

TOFA Suppresses Lipid Biosynthesis in Fanca2/2 and Fancd22/2 MSCs
To examine the effect of the elevated glycerophospholipids on HSC function, we hypothesized that targeted reduction of these lipids might improve hematopoietic-supporting function of FA MSCs. In searching for utility to reduce glycerophospholipids in FA MSCs, we identified TOFA, one of downregulated metabolites in both Fanca2/2 and Fancd22/2 MSCs (12-fold in Fanca2/ 2 and 10-fold in Fancd22/2 MSCs) (Fig. 2D). TOFA inhibits the activity of acetyl-CoA carboxylase (ACC) [25,26], a rate-limiting enzyme in lipid synthesis that catalyzes the conversion of acetyl-CoA to malony-CoA. We performed five independent assays to confirm the effectiveness of TOFA in suppressing lipid biosynthesis in Fanca2/2 and Fancd22/2 MSCs: (a) TOFA suppressed ACC activity by enzyme assay (Fig. 3A); (b) TOFA inhibited the expression of genes involved in glycerophospholipid biosynthesis by q-PCR (Fig. 3B); (c) TOFA reduced the levels of Fasn but not Acsl1, Fasn is the major enzyme involved in glycerophospholipid biosynthesis by Western blot (Fig. 3C); (d) TOFA repressed the biosynthesis of total lipids by oil red O staining (Fig. 3D); and (e) TOFA decreased the levels of Fasn and Phosphoethanolamine by immunofluorescence staining (Fig. 3E).

TOFA Partially Corrects the Effects of FA MSC Cells on Self-Renewal and Differentiation
We next determined whether TOFA could restore or improve the hematopoiesis-supporting function of FA MSCs. We conducted two sets of experiments to examine the effect of TOFA on MSC-dependent HSC function. First, we used limiting-dilution CAFC assay to evaluate the self-renewal capacity of the cocultured LSK cells on WT, Fanca2/2, or Fancd22/2 MSCs that had been treated with or without TOFA for 48 hours. TOFA increased the number of cobblestone colonies in Fanca2/2 and Fancd22/2 cocultures to WT levels (Fig. 4A). Second, we performed competitive repopulating assay using approximately 1 3 10 5 output cells collected from cocultures treated with or without TOFA. Four months later, we analyzed donor engraftment in both peripheral blood and the BM. The cells from Fanca2/2 and Fancd22/2 cocultures produced approximately two-to threefold more donor chimerism in peripheral blood compared to WT cocultures, and pretreatment of the Fanca2/2 and Fancd22/2 MSCs with TOFA resulted in 30%-50% reduction of donor chimerism (Fig.  4B, 4C). Therefore, reduction of donor chimerism by TOFA was likely due to its inhibitory effect on the expansion of the myeloid cells rather than on self-renewal capacity of the donor HSCs. Moreover, TOFA greatly alleviated the abnormal expansion of the donor myeloid (Gr11Mac11) compartment in both the peripheral blood and BM of recipient mice transplanted with cells from Fanca2/2 and Fancd22/2 cocultures (Fig. 4B, 4C). Consistently, TOFA significantly inhibited the proliferation of donor-derived (CD45.2) total (Fig. 4D) and myeloid (Fig. 4E) progenitors isolated from the BM of recipient mice. These data show that the endogenous ACC inhibitor TOFA corrects the defect of the Fanca2/2 and Fancd22/2 MSCs and support the notion that the aberrant myeloid explosion is resulted from elevated levels of lipid metabolites, including Phosphoethanolamine, in FA MSCs.
To determine whether overexpression of glycerophospholipids in normal MSCs could cause similar defect in hematopoiesis as with FA MSCs, we performed experiments in which WT MSCs were treated with or without adipogenic supplement to induce Fabp (fatty acid binding protein), which is known to activate PPARc leading to increased fatty acid metabolism and overproduction of glycerophospholipids [27][28][29] (Supporting Information Fig. S2). The expression levels of Fabp in adipogenic supplement-treated WT MSCs were compared to those in Fanca2/2 and Fancd22/2 MSCs without treatment by immunostaining using an antibody against the Fabp protein. The immunofluorescence levels of Fabp show that adipogenic supplement effectively induced Fabp expression in WT MSCs to a level that was comparable to untreated FA MSCs (Supporting Information Fig. S3A). Induction of Fabp in treated WT MSCs led to elevated production of glycerophospholipids, as determined by BODIPY-PE immunofluorescence staining (Supporting Information Fig. S3B), and decreased frequency of CAFC of cocultured LSK cells, as analyzed by the CAFC assay (Supporting Information Fig. S3C). These results indicate that overproduction of glycerophospholipids impairs hematopoiesissupporting function of WT MSCs, mimic of the phenotype observed in Fanca2/2 or Fancd22/2 MSCs.

Lipin 1 Knockdown Prevents Myeloid Expansion by Correcting the Defects of FA MSCs
To genetically demonstrate the role of glycerophospholipids in MSC function, we depleted Lipin1, a proximal enzyme that converts Diacyl glycerol to Phosphoethanolamine and other glycerophospholipids [30] (Fig. 5A), using lentiviral shRNA. Baseline Lipin1 was higher in both Fanca2/2 and Fancd22/ 2 MSCs, and the Lipin1 shRNA effectively reduced the levels Figure 5. of Lipin1 proteins, as analyzed by immunofluorescence (Fig.  5B) and Western blotting (Fig. 5C). Like TOFA treatment, Lipin1 knockdown in Fanca2/2 and Fancd22/2 MSCs also reduced Phosphoethanolamine biosynthesis to WT levels (data not shown). We next determined whether Lipin1 knockdown was capable of recapitulating the effect of TOFA on cocultured HSPCs. Indeed, Lipin1 knockdown effectively prevented myeloid expansion in transplant recipients of Fanca2/ 2 or Fancd22/2 MSC-supporting cells (Fig. 5D). These genetic data thus corroborate the notion that elevated lipid biosynthesis in Fanca2/2 and Fancd22/2 MSCs is associated with myeloid expansion observed in cocultured cells.

Tlr4 Signaling Mediates the Effect of Glycerophospholipids on HSPC Function
To understand the mechanism involved in myeloid skewing of HSCs induced by MSC-derived glycerophospholipids, we analyzed our microarray data obtained with freshly isolated phenotypic HSC (CD150 1 CD48 2 LSK; SLAM) cells from WT and Fancd22/2 mice (accession number GSE64215 at http:// www.ncbi.nlm.nih.gov/geo/). We used significance analysis of microarrays with the criteria of at least a 1.5-fold change in expression to identify genes as being upregulated or downregulated in Fancd22/2 SLAM population. Gene set enrichment analysis identified significant enhancement of Toll-like receptor (TLR) signaling in Fancd22/2 SLAM cells compared with WT cells (Fig. 6A). To evaluate the effect of glycerophospholipids on HSPC function, we focused on one of glycerophospholipids, Phosphoethanolamine, because it was one of the highly produced metabolites in FA MSCs (Fig. 2D, 2E). We first treated sorted LSK cells with or without 1 mM PE for 24 hours and performed qRT-PCR for major genes in the TLR signaling pathway. Enhancement of Tlr4 signaling and its downstream targets was evident in HSCs after PE treatment (Fig. 6B). In addition, Tlr2 and MyD88 were also upregulated in HSCs upon PE treatment. To genetically validate the involvement of TLR signaling pathway, we treated mice deficient for Tlr2, Tlr4, or MyD88 with PE. Peripheral blood was collected for analysis at 2 and 4 weeks postinjection. PE treatment induced a greater production of Gr11 and Mac11 cells in the blood of Tlr22/ 2 and WT mice compared to Tlr42/2 and MyD882/2 mice (Fig. 6C). To assess the effect of glycerophospholipids on HSC differentiating activity, we performed competitive repopulation assays using whole BM from WT, Tlr22/2, Tlr42/2, and Myd882/2 mice (CD45.2) either treated or untreated with 1 mM PE for 24 hours. Equal numbers of CD45.2 and untreated BM cells from Boy J mice (CD45.1) were mixed and transplanted into lethally irradiated Boy J mice. Analysis of lineage reconstitution at 8 weeks post-transplant showed that PE phenocopied the effect of Fanca2/2 and Fancd22/2 MSCs on WT and Tlr22/2 donor cells but not on Tlr42/2 or MyD882/2 cells. That is, PE significantly increased myeloid lineage repopulation in recipient mice transplanted with WT and Tlr22/2 donor cells compared to those transplanted with Tlr42/2 or MyD882/2 cells (Fig. 6D). These data indicate that Tlr4 signaling contributes to the glycerophospholipidmediated myeloid skew of HSC differentiation (Fig. 7).

DISCUSSION
In this study, we used integrated metabolome, genetic, and functional approaches to identify and a group of FA MSCsderived metabolites, glycerophospholipids, and their endogenous inhibitor as regulators of donor HSPCs in an experimental transplant model. FA is a major inherited BM failure syndrome with extremely high risk of developing acute myeloid leukemia. The only curable treatment for this devastating disease is stem cell and gene therapies through HSCT. However, the effects of metabolic alterations of transplant recipient BM niche on donor HSCs have been underappreciated, and it remains unclear whether the metabolites released by the recipient niche into the BM are responsible for signaling directly to the mechanisms driving donor HSCs into abnormal differentiation and/or leukemia initiation. This study is aimed at identifying critical donor HSC-niche interaction regulators in a significant health-care setting, and thus would lead to an improved mechanistic understanding of donor HSC maintenance in the context of HSCT. It has been shown in several studies that the stromal feeder layer can be used to support HSC expansion and maintain quiescence both in vivo and in vitro [31][32][33]. To understand the hematopoiesis-supporting role of FA stromal cells, we modeled FA HSCT using ex vivo coculture followed by CAFC and BM transplantation assays. Although it is speculated that the environment beneath and/or niche atmosphere created by healthy MSC layer can keep HSCs in an immature state, it was demonstrated that human CD341CD382 HSCs prefer to migrate through the MSC layer [33]. We hypothesized that the HSC-MSC interaction in the ex vivo coculture model has an impact on HSC differentiation. Indeed, the true cobble stone formation (phase-dim cells) was reduced by MSCs derived from the Fanca2/2 or Fancd22/2 BM, indicating the loss of the knockdown in FA niche this myeloid-skewing phenotype was reversed, indicating a crucial role of phospholipids produced by FA MSCs in affecting the function of healthy HSCs. Although phospholipids have been traditionally considered as membrane lipids and their roles in cell signaling are yet to be discovered, our findings and emerging data [24] implicate a role of phospholipids in hematologic malignancies making the glycerophospholipid biosynthesis pathway potentially a novel therapeutic target in blood cancer that can be manipulated. In addition, we showed that genetic knockdown of Lipin1 could also ameliorate the myeloid-skewing phenotype induced by elevated glycerophospholipids. Lipin1 is a key enzyme having dual role in glycerophospholipid biosynthesis and adipocyte maturation and maintenance by modulating the C/EBPa (CCAAT/enhancer-binding protein a) and PPARc (peroxisome-proliferator-activated receptor c) network [34,35]. Our mechanistic study suggests the involvement of Toll-like receptor signaling in mediating the effect of FA MSC-derived glycerophospholipids on HSC differentiation. It has been reported that although the activation of Tlr signaling pathway did not affect the overall health of the mice, HSCs from the BM were unable to maintain quiescence and myeloid skewed upon BM transplantation [36]. TLR2 and TLR4 use TIRAP and MyD88 as adaptor proteins to engage in transducing the signal to downstream molecules and activate the NF-jB pathway [37,38]. Genome-wide chromatin immunoprecipitation-Seq analysis of H3K4me3 in BM CD341 cells derived from Myelodysplastic syndrome (MDS) patients identified a large majority of pathogenic genes involved in TLR-mediated innate immunity signaling and NF-kB activation [39]. Expression of many of TLRs (TLR1, 2, 6, 7, 9, 10, and RP105) in B cells has been identified that mediate proliferation, plasma cell differentiation, and antiapoptotic effects in B cells but only TLR4 and 8 were noted as a possibility [40]. TLR4-mediated signaling has been implicated in a variety of cancers responsible for tumor cell invasion, survival, and metastasis. Studies involving loss of TLR4 suggest several beneficial roles that could inhibit proliferation and survival of breast cancer cells [41], play a protective role in radiation-induced thymic lymphoma [42], and reduce the risk of acute Graft versus host disease (GVHD) [43]. TLR4 and TLR7/ 8 induced overproduction of p38 mitogen-activated protein kinase-dependent tumor necrosis factor a (TNFa) was also linked to certain extent to BM failure in FA [44,45]. We postulate that FA MSCs overproduce a group of glycerophospholipids including phosphocholine, phosphoethanolamine, and phospo-serine, which activates Tlr4 in HSCs. Tlr4 in turn signals through MyD88 to activate an NF-kB transcriptional program that leads to upregulation of myeloid-specific gene expression (Fig. 7). In support of this notion, 52 of 76 Acute myeloid leukemia (AML) cases from different studies have shown constitutive activation of NF-kB associated with permanent activation of the IkB kinase complex [46]. Furthermore, a combination of TLR4 and 7/8 agonists was used for the production of dendritic cells, which were derived from AML cells in order to use in the immunotherapy of AML patients [47]. Several studies were devoted in transforming inhibition of NF-kB as to model the  strategy of clinical trials [48][49][50][51]. Whether this constitutive activation of Tlr4 in HSCs is caused by a direct binding or by indirect effect of the FA MSC-derived glycerophospholipids requires further investigation.

CONCLUSIONS
In summary, our results show that the endogenous ACC inhibitor TOFA partially corrects the defects of FA MSCs and indicates that elevated levels of lipid metabolites produced by FA MSCs, including glycerophospholipids, are associated with aberrant myeloid expansion. Our studies suggest that targeting glycerophospholipid biosynthesis either by TOFA or modulating Lipin1 in FA MSCs could be a therapeutic strategy to improve hematopoiesis and stem cell transplantation for FA patients.