Comparative Transcriptome Analysis of Bone Marrow Resident vs. Culture Expanded Mouse Mesenchymal Stem/Stromal Cells

Abstract

Introduction:

Mesenchymal stem/stromal cells (MSCs) are defined as culture expanded populations, and while these cells recapitulate many properties of bone marrow (BM) resident skeletal stem/progenitors, few studies have directly compared these populations to evaluate how culture adaptation and expansion impact critical quality attributes.

Methods:

We performed RNA-Seq on Lin⁻SCA1⁺ MSCs enriched from BM by immuno-depletion (ID) and after subsequent culture expansion (Ex) and Lin⁻LEPR⁺ MSCs sorted (S) directly from BM. Pairwise comparisons were used to identify differentially expressed genes (DEGs) between populations and gene set enrichment analysis was employed to identify biological pathways/processes unique to each population. K-means cluster analysis resolved isolation status dependent changes in transcription in pseudotime.

Results:

Hierarchical clustering segregated populations by isolation process and principal component analysis identified transcripts related to vasculature development, ossification, and inflammatory/cytokine signaling as key drivers of population variance. Pairwise comparisons identified 3849 DEGs between ID vs. S BM-MSCs mapping to gene ontology (GO) terms related to immune and metabolic processes, and 334 DEGs in Ex vs. ID BM-MSCs mapping to GO terms related to tissue development, cell growth and replication, and organelle organization. K-means cluster analysis revealed significant differences in transcripts encoding stemness and differentiation markers, extracellular matrix structural consituents and remodeling enzymes, and paracrine acting factors between populations.

Conclusions:

These comparative analyses reveal significant differences in gene expression signatures between BM resident vs. culture expanded MSC thereby providing new insight into how culture adaptation/expansion endows the latter with unique quality attributes.

Introduction

Bone marrow (BM) derived mesenchymal stem/stromal cells (MSCs) are culture expanded populations with stem/progenitor^1–3 and paracrine properties^4–6 that have been evaluated clinically for treating numerous human diseases. MSCs are routinely isolated from BM by plastic adherence and culture expansion and while protocols exist to enriched subpopulations directly from BM that give rise to skeletally restricted progenitors^7–9 these subpopulations are not used clinically. Several candidate MSC populations have been identified in BM^10–13including a population of leptin receptor (LEPR) expressing sinusoidal perivascular cells that function as bone and fat precursors and are the principle source of chemokines regulating retention of hematopoietic stem cells in marrow.^14–17 These cells also possess all BM colony forming unit-fibroblast (CFU-F) activity, a defining property of MSCs. Herein, we catalogued the transcriptomes of Lin⁻LEPR⁺ populations sorted (S) directly from BM and Lin⁻SCA1⁺ cells enriched from BM by immuno-depletion (ID) and following subsequent culture expansion (Ex). Comparative analyses revealed pseudo-time dependent changes in gene expression that resolve intrinsic and isolation process dependent differences in the profile of expressed cell surface markers, stem cell maintenance factors, matrix and matrix remodeling proteins, and paracrine acting factors. Inter-population differences in gene expression identified herein demarcate a transition from niche resident to culture adapted populations.

Materials and Methods

Cell isolation

Lin⁻LEPR⁺ cells (S) were enriched directly from BM using a previously published method with modifications.¹⁶ Briefly, BM pooled from femurs/tibiae of male, 3-month-old C57BL/6 mice (n=3) (Jackson Laboratories) was digested for 30 min at 37°C with 500 μg/mL Liberase DL (Sigma) and 400 μg/mL DNAseI (Sigma) followed by red blood cell lysis using ACK lysis buffer. Liberated cells were suspended in HBSS, stained with biotinylated anti-CD11b (BD, #553309), anti-CD34 (BioLegend, #119304), and anti-CD45 (BD, #553078) antibodies (10 μg/mg beads) and antibody bound cells removed using M280 Streptavidin Magnetic beads. Remaining cells were labeled with CD16/32 Fc block (BD, #553142, 1:40), FITC anti-Ter119 (Tonbo #35–5921, 1:100), FITC anti-CD31 (BD, #553372, 1:50), FITC anti-CD45 (BD, #553078, 1:100), biotinylated anti-LEPR (R&D, #BAF497, 1:33), and donkey-anti-goat Alexa 647 (Invitrogen, #A21447, 1:40) antibodies and propidium iodide (PI) (BD, #51–66211E, 1:100) and the Lin⁻LEPR⁺ fraction isolated by fluorescence activated cell sorting (FACS) using a BD FACS Aria at a flow rate of <10,000 cells/sec and a 100 μM nozzle. Alternatively, Lin⁻SCA1⁺ cells (ID) were enriched from BM pooled from the femurs/tibiae of male, 2–3 month-old C57BL/6 mice (n=5) (Jackson Laboratories) by immuno-depletion as previously described.^{18, 19} ID BM-MSCs were subject to RNA extraction post-depletion or plated at 5000 cells/cm² and expanded for 7d in complete culture media (CCM; MEM-alpha GlutaMax, 10% FBS, 100 U/mL of Penicillin, 100μg/mL Streptomycin) to generate Ex BM-MSCs, which have been extensively characterized.^20–23 All cell manipulations were performed in a modular airtight chamber (BioSpherix Ltd.) flushed with 5% O₂ and balanced with N₂ and cells were incubated in a humidified chamber at 37°C in 5% O₂/CO₂. Total RNA was isolated using the Quick-RNA MiniPrep Kit (Zymo Research, #R1054). All media components were purchased from GIBCO.

Flow cytometry

Ex BM-MSCs (2.5×10⁵ cells/ml) suspended in HBSS were incubated with antibodies against CD11b, CD19, CD29, CD31, CD34, CD44, CD45, CD73, CD105, CD106, CD146, CD166, SCA1, F4/80, SSEA-1 and SSEA-4 (R&D Systems Inc.) for 30 minutes at 4°C. Cells were washed twice with HBSS, incubated with the appropriate fluorochrome-conjugated secondary antibody for 30 minutes at 4°C and analyzed using an Epics FC 500 FACS scanner equipped with CXP software (Beckman Coulter, Inc.). Data were analyzed using FlowJo software (TreeStar Inc).

RNA Sequencing

RNA was run on an Agilent 2100 Bioanalyzer RNA pico chip (Agilent Technologies 5067–1513) for quality assessment and quantification, and samples with an RNA Integrity Number (RIN) > 8.0 were advanced for further analysis. Libraries were prepared in an RNase-free working environment. Poly-adenylated RNAs were selectively isolated from total RNA (~10 ng) using poly-T oligos attached to magnetic beads according to the manufacturer‟s guidelines in the NEBNext poly(A) mRNA magnetic isolation module (NEB E7490). Library preparation from enriched mRNA and total RNA was completed with the NEBNext Ultra II Directional RNA kit (NEB E7760). Briefly, RNA samples were chemically fragmented in buffer containing divalent cations at 94°C for 8 minutes, primed with random hexamers, and reverse transcribed to generate the first strand cDNA The second strand was synthesized after removing the RNA template and incorporating dUTP in place of dTTP. Incorporation of dUTP quenched the second strand during PCR amplification and therefore the strand information was preserved. The double-stranded (ds) cDNA was purified with 2.0X Agencourt AMPure XP bead (Beckman Coulter A63881) ratio to retain insert fragments >80 nt. The ds cDNA was end repaired and adenylated at the 3’ end. A corresponding ‘T’ nucleotide on the adaptors was utilized for ligating the adaptor sequences to the ds cDNA. Adaptor-ligated DNA was purified using magnetic Ampure XP beads and PCR amplified using 13 cycles to generate the final libraries, which were size selected and purified using 1.0 × Ampure XP beads to remove any primer dimers. The final library size was typically 200–600bp with insert sizes ranging from 80–450bp. Libraries were validated using bioanalyzer DNA chips and qPCR using primers that recognize the adaptors, pooled at equimolar ratios, loaded onto a NextSeq 500 flow cell at 1.8pM final concentration, and sequenced using 2 × 75bp paired-end chemistry. On average 18–20 million reads are generated (base quality score >Q30 suggesting less than 1 error in 1000bp).

RNA-Seq Data Analysis

All raw RNA-Seq data processing was conducted on the Galaxy workbench. Collected reads were downloaded as raw FASTQ files and quality checks were performed using the FASTQC application. CutAdapt was used to remove adapter sequences and short reads. Reads were mapped against the NCBI build 37 mm9 mouse reference genome using the HISAT2 and featureCounts application packages. Raw read counts were then exported for downstream analysis in the R software package environment. Differential gene expression analysis was carried out in R using the edgeR package. Briefly, low expressing raw reads across samples were filtered out such that the sum of counts was greater than 2 across any given transcript. Raw read counts were normalized with the edgeR calcNormFactors function to generate a model of normalized reads (cpm), which were then filtered to only include protein coding transcripts in subsequent analyses. A dispersion value of 0.2 was used for differential expression exact test analysis. Differentially expressed genes (DEGs) were used to create ranked ordered lists for gene ontology (GO), and gene set enrichment analysis (GSEA) was done using the clusterProfiler R package. PCA analysis was carried out using the DESeq2 software package. Heatmaps were generated using the Pheatmaps R package. Heatmaps are displayed as a calculated Z score across each row of transcripts. The Enhanced Volcano package was used to generate volcano plots of DEGs. K-means gene expression clustering was performed in R with the cluster package. The number of significant clusters was determined using the sum of squares method. Significance level was set at p ≤ 0.05. Significance in RNA-Seq data analysis was calculated using a pre-defined biological dispersion value set at 0.2.

Quantitative real-time PCR

Total RNA was isolated using the RNeasy Kit (Qiagen), converted to cDNA using the Transcriptor First Strand cDNA Synthesis Kit (F. Hoffmann-La Roche), and amplified by PCR using the TaqMan^® EZ RT-PCR kit (Applied Biosystems) according to the manufacturer’s instructions. Reactions were performed on a 7900 HT sequence detector (Applied Biosystems) using the FastStart Universal SYBR Green Master Mix (F. Hoffman-La Roche) and transcript levels quantified using the relative Ct method with Gapdh as an internal control. Primers used for amplification were as follows: Clec11a, 5′- AATCGCAGTTAGTCTTACAA-3′ and 3′- GTCCATCTACTTCCTCATTC-5′; Hmox1, 5′- CCAGTTCTACCAGAGTAA-3′ and 3′- ACAGAAGTTAGAGACCAA-5′; Timp2, 5′- GTCTCTTAATGTCGCTTA-3′ and 3′- TGGCTGTCTATATTGTTAG-5′; Aspn, 5′- GTCTTGAGGATACTTACATTG-3′ and 3′- AAGCAGTAGCATCACATA-5′; Itga5, 5′- GTTCACTGTTCCTCATCT-3′ and 3′- GAGTTGTTCAGGTTCTTG-5′; Alpl, 5′- AAGGACATCGCATATCAG-3′ and 3′- TTCGTATTCCACATCAGTT-5′; Gas6, 5′- AAGATGCTGTAATGAAGAT-3′ and 3′- TGAGATTCAGGTGATAGA-5′; Mmp2, 5′- TGATGTCCAGCAAGTAGA-3′ and 3′- CTTCATTGTATCTCCAGAACT-5′; Itga11, 5′- ACACTTCCTCATCAATATCC-3′ and 3′- GGCATCTTCTCCATACTG-5′; Ptgs2, 5′- TCAGCATTCATTCCTCTACATA-3′ and 3′- ACAACACTCACATATTCATACATT-5′; Mif, 5′- TCTACATCAACTATTACG-3′ and 3′- GATAAACACAGAACACTA-5′; Tnn, 5′- CGGAGTTATGTAGAAGGTT-3′ and 3′- CAGTGGTGAGATTATGTAGT-5; Ccn3, 5′- TATAATACCAAGGTTGACT-3′ and 3′- CTCTACTGCTGTAACTATA-5′; Omd, 5′- AGGTTGTGCTAAGGAATG-3′ and 3′- AAGGTTGAGTTGCTGAAT-5′; Dmp1, 5′- TGTCATTCTCCTTGTGTT-3′ and 3′- TCAGATTCAGTATTGTGGTA-5′; Tnfrsf11b, 5′- TGTTCCTACCAAGATTATACCA-3′ and 3′- TACACTCTCGGCATTCAC-5′; Mmp14, 5′- AAGTGATGGATGGATACC-3′ and 3′- TCGTAGGCAGTATTGATG-5′; Aebp1, 5′- AGACCTATACAGAAGTAGTGA-3; and 3′- TCCTCCTCCTCTATCTCT-5′

Statistical analysis

Significance was estimated by unpaired Student’s t test in experiments where comparisons between two groups was warranted. Differences between treatment groups were considered significant with a p value of ≤0.05.

Results

Enrichment of BM-MSC populations.

We isolated Lin⁻LEPR⁺ cells (S) directly from BM via FACS using a published protocol¹⁶ with modifications and Lin⁻SCA1⁺ cells from plastic adherent cultures by immuno-depletion (ID)^{18, 19} and after a subsequent week of culture expansion (Ex) (Fig. 1a). Zhou et al., reported that Lin⁻LEPR⁺ cells account for 0.2–0.3% of total cells recovered from BM following enzymatic disaggregation (Fig. 1b, d).¹⁶ To improve yields, we subjected enzymatically disaggregated BM to immuno-depletion using anti-CD45, anti-CD11b, and anti-CD34 antibodies, then stained the remaining cells with antibodies against CD31, Ter119, CD45, and LEPR and isolated the Lin⁻LEPR⁺ fraction by FACS (Fig. 1c). Using this approach, the yield of Lin⁻LEPR⁺ cells was ~1% of the total BM cells representing an 11-fold enrichment (Fig. 1d). Alternatively, immuno-depletion of plastic adherent cultures afforded a ~5-fold enrichment of ID BM-MSCs based on CD44 and CD45 expression (Fig. 1e), which are devoid of hematopoietic contaminants and uniformly express SCA1 based on flow cytometric analysis (Fig. 1f).

Figure 1. Isolation of Lin⁻LEPR⁺ and Lin⁻SCA1⁺ BM-MSCs.

a, Schematic showing timeline and enrichment method used to generate S, ID, and Ex BM-MSCs. BM was pooled from n=3–5 mice per isolation. b, Representative dot plots of flow cytometric data showing gating strategy used to isolate the Lin⁻LEPR⁺ fraction from BM following red cell lysis; forward vs. side scatter (left, top), single cell gating (right, top), viability staining with PI (left, bottom), and final sort gate (right, bottom). c, Representative dot plot showing gating of Lin⁻LEPR⁺ cells after pre-depletion with anti-CD45, anti-CD11b, and anti-CD34 antibodies. d, Table of Lin⁻LEPR⁺ cell yields following isolation by RBC lysis vs. preimmuno-depletio n. e, Represe ntative dot plots of flow cytometr y data showing BM cells stained with anti-CD44 and anti-CD45 antibodi es pre (left panel) and post immuno-depletion (right panel). Bar graph shows fold enrichment of Lin⁻SCA1⁺ ID BM-MSCs based on CD44 and CD45 staining. Data are mean ± SD from three biological replicates. f, Representative dot plots of flow cytometry data showing surface phenotype characterization of Ex BM-MSCs.

Transcript profiles of BM-MSCs.

We analyzed all three BM-MSC populations by RNA-Seq and filtered data to include only processed transcripts mapping to annotated protein coding regions. Unsupervised hierarchical clustering of the 1000 most highly expressed transcripts revealed that S BM-MSCs were more distanly related to ID and Ex BM-MSCs, which showed highly similar transcript profiles (Fig. 2a). Principal component analysis (PCA) similarly discriminated all three populations with 93% of the variance accounted for by PC1 (Fig. 2b), which included Lepr and genes involed in blood vessel and vasculature development (Lox, Col1a1, Col8a2, Fn1, Cspg4, Mmp2), ossification (Col1a1, Tmem119), and inflammatory/cytokine signaling (Lurap1l, Cxcl10, S100a8) (Fig. 2c). Alternatively, transcripts contributing to PC2 function in biomineral tissue development (Bglap, Dmp1), ion transport (Kcnbm1, Slc26a7), and Notch (Dner) and GDNF (Gfra2) signaling. This analysis also identified 2775 and 1074 differentially expressed genes (DEGs) that were up or downregualted, respectively, by >Log₂ fold change (p<0.05) in ID vs. S BM-MSCs (Fig. 2d, e and Supplementary Table 1) and 2800 and 985 up and downregulated DEGs, respectively, in Ex vs. S BM-MSCs (Supplementary Table 2). Comparison of Ex vs. ID BM-MSCs identified 211 and 123 DEGs that were up and downregulated, respectively, (Fig. 2f, g and Supplementary Table 3) reflecting the high degree of relatedness between these populations.

Figure 2. BM-MSC populations exhibit discordant transcriptomes.

a, Heat map showing cluster analysis of the top 1000 most abundant transcripts in S (Lin⁻LEPR⁺) and ID and Ex (Lin⁻Sca1⁺) BM-MSCs. Color scale corresponds to per-gene z-score computed across each row. b, PCA analysis of data from (a). c, Genes contributing to PC1 and PC2. Y-axis represents the percent contribution to the corresponding principal component. d-g, Volcano (d, f) and violin (e, g) plots of DEGs in ID vs. S (d, e) and Ex vs ID (f, g) BM-MSCs. h & i, Top GO terms (biological processes) based on p-value identified by GSEA of DEGs for ID vs. S (h) and Ex vs. ID BM-MSCs (i). Arrows indicate gene sets corresponding to each GO term that show positive (up) or negative (down) enrichment.

Gene set enrichment analysis (GSEA) of DEGs identified in ID vs S BM-MSCs corresponded to GO terms related to immune processess including lymphocyte and mononuclear cell proliferation, reguation of myeloid cell homeostasis and differentiation, activation of immune responses, and lymphoctye mediated immunity (Fig. 2h). These terms included transcripts encoding various hematopoietins (Kitl, Clec2d) and cytokines (Il6, Il7, Il17ra, Cxcl1, Cxcl12, Csf1), which were all downregulated. GO terms corresponding to negative regulation of cellular amide metabolic process and peptidase activity were also downregulated while those related to ATP and carboxylic acid metobolic process were upregulated (Fig. 2h). Genes implicated in cellular redox balance including peroxidasin (Log₂FC = 3.21) and superoxide dismutase 3 (Log₂FC = 3.69) were also significantly upregulated in ID vs S BM-MSCs, and the GO term “skeletal system development” also showed a positive normlaized enrichment score while the term “fat cell differentiation” had a negative score. DEGs identified between Ex vs. ID BM-MSCs corresponded to GO terms related to cell division, cell cycle regulation, chromsome organization, regulation of organele organization, and nucleic acid metabolic process, which were downregulated, while terms corresponding to biotic external stimulus and tissue development were upregulated (Fig. 2i). The latter included transcripts encoding several ECM constituents (Dmp1, Ccn3, Mmp13), and proteins implicated in WNT signaling (Rspo3, Sfrp1, Dkk1, Lgr5, Nkd1) and bone development/mineralization (Ifitm5, Ibsp, Bmp4, Phex, Adrb2, Dlx3).

Inter-population differences in transcriptomes displayed in psuedo-time.

To further highlight inter-population differences in gene expression, we performed K-means clustering of our RNA-Seq data, which identified 5 unique gene expression clusters (Fig. 3a, b) depicting isolation status-resolved changes in pseudotime. Cluster 1 includes transcripts highly expressed in S and downregulated in ID and Ex BM-MSCs (Fig. 3c), while cluster 2 contains transcripts whose expression is lowest in S and highest in Ex BM-MSCs (Fig. 3d). Clusters 3 and 4 include transcripts that are down or up regulated, respectively, in ID relative to S and Ex BM-MSCs (Fig. 3e, f) and cluster 5 contains genes that are highly expressed in S and ID and downregulated in Ex BM-MSCs (Fig. 3g).

Figure 3. Isolation status-resolved changes in the BM-MSC transcriptome in pseudotime.

a, Plot of within sum of squares from K-means clustering of RNA-seq data from S, ID, and Ex BM-MSCs depicting the variance within the clusters. b, Gene expression patterns identified for the 5 most significant clusters from (a) displayed in pseudotime. c-g, Transcript profiles (left panel) for each cluster depicted in (b). Color scale corresponds to the fitness score for each gene to the cluster centroid line (black). Dot plot showing biological process GO terms generated by gene enrichment corresponding to each cluster (right panel).

Based on GSEA, transcripts in cluster 1 mapped to GO terms corresponding to positive regualtion of catabolic processes, response to oxidative and endoplasmic reticulum stress, RNA localization, processing and splicing, and regulation of hemopoiesis (Fig. 3c). These changes are consistent with early stages of culture adaptation involving rewiring of cellular catabolic processes in favor of anabolic processess required to support energy demands associated with cell proliferation, ECM protein secretion, and redox challenges resulting from increased metabolic activity and oxygen saturation levels.²⁴ DEGs from cluster 2 mapped to terms involving regulation of cell growth, mitochondrian organization, cell-cell signaling by WNT, ECM organization, endosomal, vacuolar, lysosomal and golgi vesicle transport, cell-substrate adhesion, collagen metaoblic process, and connective tissue development (Fig. 3d). These terms reflect changes in cellular protein transport/secretion needed to produce an ECM that supports cell attachment and spreading, and activation of signaling pathways that prime cells for ECM-directed lineage differentiation. Cluster 3 transcripts also mapped to terms related to WNT signaling and Golgi vessicle transport and also included cellular response to inteferon-alpha, inositol phosphate biosynthetic proces, and hematopoeitic stem cell migration (Fig. 3e) thereby reflecting cellular processes common to S and Ex BM-MSCs. Cluster 3 includes the vitamin D receptor (Vdr), a gene (Mn1) induced by vitamin D²⁵ that regulates calvarial osteoblast maturation and RANKL expression,²⁶ Eif2ak2 that modulates expression of pro-inflammatory cytokines and interferons²⁷ and genes involved in microtubule cytoskeleton organization (Bbs1, Bbs4, Cep120). Lastly, clusters 4 and 5 contain several common GO terms inlcuding ribosome and ribonucleoprotein complex biogenesis and ncRNA/rRNA processing. Terms unique to cluster 4 included those corresponding to bioenergetic processes, such as mRNA processing, mitochondrion organization, and generation of precursor metabolites (Fig. 3f) while terms unique to cluster 5 included organelle fission, negative regulation of cell cycle, cell cycle checkpoint, and DNA replication (Fig. 3g). These genes reflect impacts of culture adpation/expansion but that trend in the opposite diretion of cluster 3

Gene profiles reflect inter-population differences in phenotypic and functional attributes.

To further discriminate differences between BM-MSC populations, we interrogated transcripts encoding cell-type specific surface makers, which revealed high expression of Lepr, Bmpr1a, and Pdgfra in S and Sca1 in ID and Ex BM-MSCs consistent with previous characterizations (Fig. 4a).^{16, 18} Transcripts encoding Tgfbr1, Tgfbr3, Egfr, Adrb2, Lifr, and CD73 were also highly expressed in S BM-MSCs while ID and Ex BM-MSCs expressed high levels of Lrp5, CD109, CD248. Ptprf and Fzd2. With respect to lineage markers, S BM-MSCs were characterized by high expression of the adipogenic markers Adipoq, Cebpa and Cebpbb and the BMP inhibitors Chrdl1 and Bmper while Ex BM-MSCs expressed high levels of transcripts implicated in osteoblast differentiation and matrix mineralization including Sp7, Alpl, Msx1, Bglap, Gdf, Dmp1 and the WNT modulator Kremen1 (Fig. 4b). ID BM-MSCs were further distinguished from S and Ex BM-MSCs by expression of Itga5, Twist2, Tnfrsf11b and the carboxypeptidase Aebp1. Additionally, S BM-MSCs expressed high levels of transcripts involved in collagen fibril formation (Dcn), ECM remodeling (Fap, Mmp8, Timp4), and cell adhesion (Icam1, Spon, Ecm2, Vit) while ID and Ex BM-MSCs expressed high levels of ECM structural constituents (Col1a1, Col3a1, Col4a1, Col4a2, Col5a1, Col5a2, Fn1) and remodeling factors (Mmp2, Mmp17, Mmp23, Timp3. Adamts2, Adamts6, Adamts12) (Fig. 4c). Ex BM-MSCs were further characterized by high expression of ECM remodeling enzymes (Timp2, Mmp14, Mmp19) and cell adhesion molecules (Lama1, Lamb2) that modulate cellular spatial organization via interaction with ECM components. With respect to paracrine acting factors, S BM-MSCs were characterized by high expression of several chemokines/cytokines (Cxcl1, Cxcl10, Cxcl12, Ccl7, Ccl9, Il1b, Il6), angiogenins (Angpt1, Angptl4, Vegfc, Esm1), the hematopoietin Kitl,²⁸ and the anti-inflammatory mediators Il1ra, Tnfaip6, and Ltf while ID BM-MSCs highly expressed Apln, Tgfb2 and Mif and Ex BM-MSCs highly expressed Hmox1, Celc11a, Bmp3, Nog, Hgf, and Gas6 (Fig. 4d). To validate these data, we analyzed RNA samples from ID and Ex BM-MSCs used for RNA-Seq analysis by qRT-PCR (Fig. 4e). This analysis confirmed that a subset of surface markers (Itga5, Itga11), lineage markers (Alpl, Aspn), ECM proteins (Mmp2, Timp2, Omd, Mmp14, Tnn) and paracrine factors (Clec11a, Hmox1) were expressed a significantly different levels and in a pattern consistent with results of our transcriptomics analysis.

Figure 4. Stem/progenitor and stromal markers are highly discordant between S, ID and Ex BM-MSCs.

a-d, Gene specific heatmaps of commonly used phenotypic markers (a), lineage differentiation markers (b), matrix remodeling genes (c), and paracrine acting factors (d) for S, ID, and Ex BM-MSCs. Color scale corresponds to per-gene z-score computed across each row. e, QPCR quantification of select genes from (a-d) in ID and Ex BM-MSCs. Data are mean ± SD and p values are by student‟s t test.

Discussion

Numerous studies have compared the transcriptomes of human MSCs sourced from different tissues,^29–33 and how it is impacted by culture density,³⁴ passage number,^35–37 and differentiation to connective tissue lineages.^38–42 Collectively, these studies indicate that MSCs express a core set of genes that reflect their identity,^{29, 30} many of which are implicated in cell adhesion and ECM production, organization and remodeling. However, few studies have compared conventional mouse MSCs to their BM resident counterparts even though these populations share several critical quality attributes including the ability to support hematopoiesis, form osseous tissue and generate CFU-Fs. Our comparative analysis of BM resident Lin⁻LEPR⁺ (S) and cultured expanded Lin⁻SCA1⁺ MSCs (ID and Ex) revealed significant differences in gene expression profiles. For example, S BM-MSCs express transcripts implicated in adipogenesis and lipid metabolism while ID and Ex BM-MSCs express transcripts implicated in osteogenesis. These results are consistent with studies showing that bifurcation of Lin⁻LEPR⁺ cells toward adipogenesis is regulated by nutritional status and induced by leptin signaling^{16, 17} and that ID and Ex BM-MSCs bifurcation favors osteogenesis at the expense of adipogenesis.^{22, 23, 43} Additionally, GSEA of ID vs. S BM-MSCs returned GO terms related to immune processes and included several hematopoietins and cytokines that were downregulated reflecting the disruption of niche interactions by S BM-MSCs that support hematopoiesis. Alternatively, genes implicated in cellular redox balance were upregulated thereby reflecting changes in the bioenergetic demands of culture adapated ID BM-MSCs and the accompanying increase in cellular oxidative stress. Comparison of Ex vs. ID MSCs returned GO terms involving cell division, cell cycle regulation, chromsome organization, regulation of organele organization, and nucleic acid metabolic process consistent with an observed decreased proliferative capacity of Ex vs. ID BM-MSCs due to increased oxidative stress.²³

While impacts of culture expansion on MSC gene transcription is well documented, these studies typically evaluate cells subjected to long-term culture expansion³⁷ or large-scale manufacturing.⁴⁴ Consequently, an important aspect of this study is that BM-MSCs analyzed herein were subjected to minimal culture expansion. Tissue disaggregation and cell enrichment protocols using similar lineage specific antibodies were also employed to ensure a high degree of integrity between samples. Source material used for RNA-Seq analysis was also pooled from BM donated from n=3–4 mice to eliminate inter-animal variation. Accordingly, inter-population differences in gene expression profiles reflect cell intrinsic and cell isolation dependent impacts. Based on studies showing that gene expression profiles of MSCs from different tissues are highly correlated (0.88–0.92) and biases in expression strongly reflect organ specific functions,³⁰ inter-population differences in gene expression described herein are anticipated to predominantly reflect impacts of culture adaptation on quality attributes.

A key element of this study is the choice of BM resident Lin⁻LEPR⁺ cells as a comparator to culture expanded Lin⁻SCA1⁺ populations. While we recognize that these two populations are not equivalent, i.e. it is undocumented whether the former directly gives rise to the latter, both populations are heterogeneous, function as osteoblasts, adipocytes, and chondrocytes precursors, secrete hematopoietins, contain all CFU-F activity in BM⁴⁵, and have been extensively studied in the literature. For example, recent studies have shown that the frequency and lineage trajectory of Lin⁻LEPR⁺ cells is altered in BM by nutritional status¹⁷ and aging.⁴⁶ Importantly, we recently reported that deletion of inositol hexakisphosphate kinase 1 (Ip6k1) in mice, which protects animals from diet-induced obesity and insulin resistance, enhanced the growth, survival, osteogenic potential, and hematopoiesis-supporting activity and reduced the adipogenic potential of Ex BM-MSCs.²² Moreover, culture expanded BM-MSCs from aged vs. young donors have been shown to exhibit decreased osteogenic and increased adipogenic potential,^47–49 which has been shown to be driven in part by age-induced changes in the BM microenvironment.⁵⁰ Consistent with these findings, we recently showed that alpha-2-macroglobulin (A2M) is highly downregulated in Lin-LEPR+ BM-MSCs obtained from mice with senile osteoporosis, which is characterized by bone loss and increased marrow adiposity, and that A2M loss of function also skews lineage bifurcation of culture expanded human BM-MSCs in favor of adipogenesis. Therefore, Lin-LEPR+ and Lin-SCA1+ populations evaluated herein exhibit comparable responses to pathophysiological changes induced by nutritional imbalances and aging, thereby buttressing comparisons made herein to read out impacts of culture adaptation on BM-MSC quality attributes.