Wnt/BMP Mediated Metabolic Reprogramming Preserves Multipotency of Neural Crest-Like Stem Cells

Abstract

Neural crest-like stem cells resembling embryonic neural crest cells (NCs) can be derived from adult human tissues such as the epidermis. However, these cells lose their multipotency rapidly in culture limiting their expansion for clinical use. Here, we show that the multipotency of keratinocyte-derived NCs (KC-NCs) can be preserved by activating the Wnt and BMP signaling axis, promoting expression of key NC-specifier genes and ultimately enhancing their differentiation potential. We also show that transcriptional changes leading to multipotency are linked to metabolic reprogramming of KC-NCs to a highly glycolytic state. Specifically, KC-NCs treated with CHIR and BMP2 rely almost exclusively on glycolysis for their energy needs, as seen by increased lactate production, glucose uptake, and glycolytic enzyme activities. This was accompanied by mitochondrial depolarization and decreased mitochondrial ATP production. Interestingly, the glycolytic end-product lactate stabilized β-catenin and further augmented NC-gene expression. Taken together, our study shows that activation of the Wnt/BMP signaling coordinates the metabolic demands of neural crest-like stem cells governing decisions regarding multipotency and differentiation, with possible implications for regenerative medicine.

Graphical Abstract

Graphical Abstract.

Significance Statement.

In this study, we have shown that human skin-derived neural crest like stem cells can be expanded in culture, while retaining their multipotency by transiently treating them with the Wnt agonist CHIR99021 and BMP2. We also show that these signaling events cause global metabolic changes by increasing glycolysis and decreasing oxidative phosphorylation to confer multipotency. Our results are impactful, because we show for the first time that signaling events that are active during embryonic development also govern carbon metabolism of adult stem cells affecting their multipotency.

Introduction

Neural crest stem cells (NCSCs) are a vital pool of stem cells that arise during vertebrate embryo development. The contribution of these cells to the formation of developed tissues is significant, including osteoblasts in the craniofacial cartilage, smooth muscle cells in arterial heart valves, melanocytes in the skin, and neurons and Schwann cells throughout the peripheral nervous system (PNS).¹ Because of their multipotency, they are often considered as the “fourth germ-layer” in the developing embryo,² and studying the signaling pathways involved in induction, migration, and differentiation of NCSCs is important to understand embryo development, the molecular basis of developmental disorders, and explore their potential for regenerative medicine.

NCSC induction in the embryo occurs as a result of a series of well-coordinated signaling events that control the transcriptional network responsible for determining their fate. The most notable fate-determining signals include fibroblast growth factor (FGF) from the underlying mesoderm, Wnt ligands from the mesoderm and non-neural ectoderm, and intermediate levels of bone morphogenetic protein (BMP).^3,4 These signals have been widely accepted to contribute to the multipotency of the transient NCSC phenotype. As NCSCs undergo epithelial to mesenchymal transition (EMT) and migrate away from this signaling hub, they encounter differential levels of Wnt, BMP, and FGF signals guiding cell fate determination away from the original multipotent niche.^5,6 In addition to signaling, cellular metabolism was shown to be crucial for EMT and migration of NCSCs.⁷ Specifically, impaired glucose metabolism was shown to lead to impaired NCSC development,^8-10 resulting in neurocristopathies. Mitochondrial dysfunction was reported to affect NCSC differentiation in zebrafish¹¹ and folate metabolism was found to be crucial for orofacial development.¹² While metabolic dysregulation affects NCSC fate decisions, the relationship between the signaling pathways affecting stemness and cell fate decisions and the metabolic state of NCSC is not well understood.

NSCS-like cells have been isolated from adult murine and human tissues, and have similar transcriptomic signatures as embryonic NCSCs.¹³ In our lab, we have isolated NCSC-like cells from the epidermis of human neonatal foreskin¹⁴ and adult¹⁵ skin tissues (henceforth referred to as KC-NCs, short for keratinocyte-derived neural crest-like cells) and established that their global transcriptional profile and differentiation potential resemble those of embryonic NCSCs. Specifically, KC-NCs could differentiate into functional neurons, Schwann cells, melanocytes, and smooth muscle cells in vitro, and most notably, KC-NC migrated along stereotypical pathways and gave rise to multiple NC derivatives upon transplantation into chicken embryos. Given the abundance of the skin tissue and the ease of isolation in a chemically defined medium without the need for genetic modification, KC-NCs represent a highly accessible cell source for regenerative medicine and potential treatment of neurodegenerative disorders. However, just like their embryonic counterparts, KC-NCs lose their multipotency quickly in culture, which limits their expansion and use for cell-therapy applications.

In this study, we addressed this challenge by mimicking the signaling events regulating NCSC induction during development. Specifically, we show that treating KC-NCs with Wnt activator CHIR99021 (CHIR) and BMP2 in FGF2-rich media restored their multipotency and improved their differentiation potential. Further, we show that Wnt and BMP signaling work synergistically to rewire cellular metabolism to a highly glycolytic state with low mitochondrial activity. The glycolytic metabolite lactate further augmented Wnt signaling leading to transcriptional changes affecting KC-NC stemness and providing evidence of crosstalk between metabolic and signaling events in establishing a multipotent state.

Materials and Methods

Isolation and Expansion of Keratinocytes From Human Skin Tissue

Keratinocytes (KCs) were obtained from human skin tissue as described previously.¹⁴ Glabrous foreskin tissue from 1 to 3-day-old neonates was obtained from John R. Oishei Children’s Hospital, Buffalo according to the institutional guidelines. Each sample was washed thrice with sterile PBS, dissected into pieces (~5 mm × 5 mm), and enzymatically digested with dispase II protease (Sigma, St. Louis, MO) for 15-20 h at 4 °C. Next, the epidermis was separated from the dermis manually using fine forceps. The epidermis was then treated with Trypsin-EDTA (0.25%) (Life Technologies, Carlsbad, CA) for 10-15 min at 37 °C, filtered through 70 μm cell strainer (BD Biosciences, Franklin Lakes, NJ) and centrifuged at 300 × g and 4 °C for 5 min. After removal of the supernatant, cell pellet was resuspended and plated on a confluent monolayer of growth-arrested 3T3/J2 mouse fibroblast feeder cells (ATCC, Manassas, VA) in keratinocyte culture medium (KCM), which consisted of a 3:1 mixture of high glucose Dulbecco’s Modified Eagle’s Medium (DMEM, Catalog # 11995-040, Gibco, Grand Island, NY) and Ham’s F-12 medium (Life Technologies) supplemented with 10% (v/v) fetal bovine serum (FBS, Atlanta Biologicals, Flowery Branch, GA), 100 nM cholera toxin (Vibrio Cholerae, Type Inaba 569 B; Millipore, Burlington, MA), 5 μg/mL transferrin (Life Technologies), 0.4 μg/mL hydrocortisone (Sigma), 0.13 U/mL insulin (Sigma), 1.4 × 10⁻⁴ M adenine (Sigma), 2 × 10⁻⁹ M triiodo-l-thyronine thyronine (Sigma), antibiotic-antimycotic (Life Technologies), and 10 ng/mL epidermal growth factor (EGF, BD Biosciences). Cells were cultured in KCM for 8-10 days or until a monolayer of keratinocytes was obtained. Hereafter, the 3T3/J2 feeder layer was detached after a 10 min versene treatment. The remaining cells were treated with trypsin-EDTA (0.25%), which was then neutralized by a solution containing 10% FBS in PBS and plated in KC serum free growth medium (KSFM, Epilife medium with Human Keratinocyte Growth Supplement; Life Technologies). Cells were further expanded in KSFM prior to NC induction. Passages 1-3 keratinocytes were used in all experiments.

Obtaining Neural Crest Stem Cells (KC-NCs) from Keratinocyte Cultures

To obtain KC-NCs from keratinocyte cultures, KCs were seeded on collagen type 1 coated dishes (10 μg/mL collagen type I, rat tail; BD Biosciences) at a density of 5000-7000 cells/cm². On the next day, NC induction was initiated by treating KCs with FI induction medium (EBM2 basal medium (Catalog # CC4176, Lonza, Basel, Switzerland) supplemented with 2% (v/v) FBS (Lonza), 10 μg/mL heparin (Lonza), 100 μg/mL ascorbic acid (Lonza), and 0.5 μg/mL hydrocortisone (Lonza), 1× gentamicin/amphotericin-B (Lonza), 10 ng/mL FGF 2 (BD Biosciences), and 10 ng/mL insulin-like growth factor 1 (IGF1, Lonza)). The medium was changed every alternate day. KC-NCs were then separated from keratinocyte cultures between days 7 and 9 by short treatment with trypsin (1 min, 37 °C) and seeded on collagen I coated dishes for further expansion in FI medium.

For investigation of the effect of Wnt and BMP2 signaling pathways on KC-NCs, KCs were induced in the FI induction medium for 9 days. On day 9, KC-NCs obtained were seeded at 80% confluence (20 000 cells per cm²) and treated with 5 µM CHIR99021 (Sigma) and 10 ng/mL BMP2 (R&D Systems, Minneapolis, MN, USA) in FI induction medium for 5 days. Medium change was performed on every alternate day. Untreated cells cultured in FI for 5 days served as controls.

Quantitative Real Time PCR

KC-NCs (20 000 cells per cm²) were seeded on collagen I coated 6-well plates and treated with CHIR99021 or BMP2 alone, and in combination. Untreated cells were used as control. After 5 days treatment, total RNA was isolated using RNeasy Mini Kit (Qiagen, Valencia, CA) as per manufacturer’s specified protocol. Superscript III cDNA Synthesis Kit (Invitrogen, Waltham, MA) was used to obtain cDNA from 1 µg of isolated RNA per sample. To assess gene expression, real-time PCR was performed using SYBR Select Master Mix (Applied Biosystems, Waltham, MA) using the primer pairs listed in Supplementary Table S1.

Immunocytochemistry

Cells were washed with cold PBS (4 °C) once and fixed for 10 min with 4% paraformaldehyde (Sigma), followed by permeabilization with 0.1% (v/v) TritonX-100 (Sigma) in PBS for 10 min at RT. Next, samples were blocked using 5% (v/v) goat serum (Life Technologies) in PBS with 0.01% (v/v) Triton X-100 for 1 h at room temperature. Immunostaining was performed using primary antibodies listed in Supplementary Table S2, at 4 °C overnight. Subsequently, cells were incubated with secondary antibodies: Alexa 488- or Alexa 594-conjugated anti-IgG antibody (Life Technologies) for 1 h at room temperature. Nuclei were stained using Hoechst 33342 (Thermo Fisher Scientific, Grand Island, NY). Cells stained with only secondary antibodies served as controls.

Imaging and Image Analysis

Immunocytochemistry and immunohistochemistry images were acquired using the Zeiss Axio Observer Z1 inverted microscope (LSM 510; Zeiss, Oberkochen, Germany) with an ORCA-ER CCD camera (Hamamatsu, Japan). Number of cells staining for a protein and mean fluorescence intensity per cell were quantified using the NIH ImageJ software. Briefly, images were converted to 8-bit and positively stained cells were manually marked. Cell counting was performed using the cell counter plugin.

Differentiation of KC-NCs to NC Derivatives

Untreated KC-NCs and those treated with CHIR99021 and BMP2 for 5 days were seeded on collagen-I coated 48-well plates at 10 000 cells/cm² in FI. Next day, CHIR99021- and BMP2-containing medium were removed and cells were differentiated into various lineages as described previously.¹⁴ Briefly, smooth muscle differentiation was initiated by treating cells with Dulbecco’s modified Eagle Medium (DMEM) supplemented with 10% FBS and 10 ng/mL TGF-β1 (BioLegend, San Diego, CA) for 2 weeks. Schwann cell differentiation was initiated by treating cells with EBM2 basal medium containing 1% gentamycin, 1% FBS (Lonza), 50 ng/mL NRG1 (Catalog # H7660, Sigma), 10 µg/mL ascorbic acid, 1:100 B27 supplement (Invitrogen), 10 ng/mL BDNF (Catalog # PHC7074, Invitrogen) and 0.5× Glutamax (Invitrogen) for 2 weeks. Melanocyte differentiation was initiated by treating cells with EBM2 basal medium containing 5% FBS, 50 mg/mL SCF (Life Technologies), 100 nM Endothelin-3 (Sigma), 10-10 M cholera toxin (Sigma), 10 nM 12-O-tetra-decanoylphorbol-13-acetate (TPA; Sigma), 10 µM SB431542 (Sigma), 50 ng/mL Wnt1 (Life Technologies), 10 ng/mL bFGF (Thermofisher Scientific) and 5 µg/mL insulin (Sigma) for 2 weeks. After 2 weeks, melanin production was examined by fixing cells with 4% (w/v) paraformaldehyde for 10 min at room temperature, followed by washing with PBS thrice. Next, cells were incubated with freshly prepared 5 mM 3,4-dihydroxy-l-phenylalanine (l-DOPA; Sigma) overnight at 37 °C followed by washing with PBS thrice. Staining was analyzed using brightfield and phase contrast images. Melanocyte and smooth muscle differentiation media contained 1% (v/v) antibiotic antimycotic (Thermo Fisher Scientific).

Metabolic Assays

Untreated KC-NCs and KC-NCs treated with BMP2 for 2 days were seeded at a density of 5000 cell/cm² while those treated with CHIR99021 and both CHIR99021+BMP2 for 2 days were seeded at a density of 10 000 cell/cm² on 96-well white bottom plates (Catalog No. 3903, Corning, NY). The next day, cells were washed with warm PBS followed by starvation for 20 min. Next, the cells were incubated in a high glucose Dulbecco’s modified Eagle Medium without l-glutamine, sodium pyruvate, and phenol red (DMEM, Catalog # 31053-028, Gibco), and was supplemented with various inhibitors as indicated overnight. Intracellular ATP, extracellular lactate, intracellular pyruvate, and enzyme kinetics of pyruvate kinase (PK), phosphofructokinase (PFK), lactate dehydrogenase (LDH), and hexokinase (HK) were measured using commercial assay kits (Supplementary Table S3) on the next day as per manufacturer’s instructions.

Flow Cytometry Analysis for Glut4 Surface Expression and Cell Size Analysis

Untreated KC-NCs and those treated with BMP2 for 5 days were seeded on collagen-I coated 6-well plates at a density of 20 000 cell/cm². KC-NCs treated with CHIR99021 and CHIR99021+BMP2 for 5 days were seeded on collagen-I coated 6-well plates at a density of 40 000 cells/cm². Ninety minutes before the experiment, cells were washed once with PBS and then cultured in medium containing 5 mM glucose, with or without 100 nM insulin in the presence of an Alexa Fluor 594 conjugated antibodies against Glut4 (Clone No. IF8, Catalog No: sc-53566, 1:150 dilution, Santa Cruz Biotechnology, TX) for 90 min. Untreated cells without incubation with antibody served as a control. Next, cells were detached with TrypLE express (Invitrogen) and centrifuged at 300 × g and 4 °C for 5 min. This was followed by washing of cell pellet with cold PBS twice by gently breaking down the pellet, and centrifuging the cells after each wash. The supernatant was then discarded, and cells were resuspended in 200 µL of cold PBS. Glut-4 expression on live cells was assessed by flow cytometry analysis of 10 000 cells per sample using a BD Fortessa X20, four-laser, 16-color analyzer (BD Biosciences, Franklin, NJ). Cell size was plotted based on forward scatter values. The results were analyzed using FCS Express 6 (De Novo Software, Pasadena, CA).

Measurement of Glucose Uptake

Untreated KC-NCs and those treated with BMP2 for 2 days were seeded at a density of 5000 cells/cm² while those treated with CHIR99021 and CHIR99021+BMP2 for 2 days were seeded at a density of 10 000 cells/cm² on 96-well white bottom plates. The next day, cells were starved for 20 min followed by incubation in the medium containing 10 mM glucose supplemented with 1% FBS overnight. Glucose uptake measurements were performed as per manufacturer’s specifications using the glucose uptake‐Glo Assay kit (Promega Corporation, Madison, WI). 2NBDG assay for glucose uptake was performed by incubating cells with 150 µM 2NBDG (Invitrogen) for 40 min at 37 °C, followed by staining with DAPI for 2 min. Cells were washed thrice with PBS and imaged using Zeiss Axio Observer Z1 equipped with a digital camera (ORCA-ER C4742-80).

Western Blot Analysis

For analysis of protein expression levels, cells were lysed using a mixture of Halt Protease Inhibitor (10×; Thermo Fisher Scientific), 41.67 mM dithiothreitol (DTT; Cell Signaling Technology), and blue loading buffer (3×; Cell Signaling Technology) in DI water. The lysates were then centrifuged for 10 min at 14 000 × g and supernatants were collected in separate tubes. Next, samples were denatured at 95 °C for 5 min and loaded at equal volumes onto 4-20% Tris-Glycine SDS-PAGE gels (Thermo Fisher Scientific). After electrophoresis (90 min, 120 V), proteins were transferred to PVDF membrane (Bio-Rad Laboratories, Hercules, CA), which was blocked for 1 h using 5% (w/v) nonfat dry milk in Tris-buffered saline (20 mM Tris, 150 mM NaCl) with 0.1% (v/v) Tween 20 detergent (TBST) buffer. Expression level of proteins was then detected by incubating the membrane overnight at 4 °C with the antibodies listed in Supplementary Table S2. The presence of proteins was detected using horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology) and SuperSignal West Pico PLUS chemiluminescence substrate (Thermo Fisher Scientific), followed by visualization using the ChemiDoc MP imaging system (Bio-Rad). Images were analyzed using the NIH ImageJ software.

Seahorse Assay to Measure ECAR and OCR

The Seahorse extracellular flux (XFe96) analyzer (Agilent Technologies, Santa Clara, CA) was used to measure (1) the extracellular acidification rate (ECAR), which is an indicator of glycolysis, and (2) oxygen consumption rate (OCR), which is an indicator of mitochondrial respiration. Untreated KC-NCs (day 14 post induction) or KC-NCs treated with CHIR99021 and BMP2 for 5 days were seeded on the XFe96 seahorse culture plates at a density of 20 000 cell/well and 50 000 cell/well, respectively. After 24 h of seeding, cells were rinsed with warm PBS and switched to the Seahorse Base Medium (XF DMEM medium, #103575, Agilent Technologies). For ECAR measurements, the base medium was supplemented with 1 mM glutamine. For OCR and ATP rate assay, medium was supplemented with 1 mM pyruvate (Agilent Technologies), 2 mM glutamine (Sigma), and 10 mM glucose (Agilent technologies) to maintain cell viability. Glycolysis and glycolytic reserve were measured from ECAR by sequential injections of 20 mM glucose, 1 µM oligomycin (Sigma), and 50 mM 2-DG (Sigma) in seahorse base medium. Maximal respiration, basal respiration, and mitochondrial ATP production were measured from OCR by sequential injections of 1 µM oligomycin, 1.5 µM FCCP (Sigma), and 0.5 µM rotenone (Sigma), and antimycin A (Sigma). ATP production rate was calculated from ATP rate assay by sequential injections of 1µM oligomycin and 0.5µM rotenone and antimycin A. All the calculations were done according to the manufacturer’s protocol.

Live Staining of Cellular Mitochondria

Untreated and CHIR99021/BMP2 treated KC-NCs were seeded on 48-well plates at a density of 10 000 cell/cm² and 20 000 cell/cm², respectively. Live mitochondrial staining was carried out by washing cells once with PBS followed by incubation with 100 nM tetramethylrhodamine methyl ester (TMRM, Thermo Fisher Scientific), 100 nM MitoTracker Red CMXRos (Thermo Fisher Scientific) or 100 nM MitoTracker Green FM (Thermo Fisher Scientific) in FI medium for 30 min at 37 °C. Cells were then stained with the Hoechst 33342 nuclear dye at 1:500 dilution in PBS, washed with PBS, and visualized using a Zeiss Axio Observer Z1 equipped with a digital camera (ORCA-ER C4742-80).

Measurement of Wnt Activity Using TOP-Flash Assay

The 7-TFP plasmid was purchased from Addgene (Catalog number 24308; Cambridge, MA). Recombinant lentivirus carrying the 7-TFP construct was produced using the protocol detailed previously.¹⁶ KC-NCs were transduced with the 7-TFP lentivirus in FI in the presence of 8 μg/mL polybrene (Sigma) for 16 h. At 48 h post-transduction, cells were selected with puromycin at 1 μg/mL for 5 days. Transduced cells were seeded onto 96-well white bottom plates at a density of 10 000 cell/well. After seeding cells for 24 h, they were treated with chemicals: 5 µM CHIR99021, 10 ng/mL BMP2, or 4 mM sodium lactate (Sigma) for 24 h. To record Wnt activity, cells were washed with 100 µL warm PBS and then lysed using 20 µL lysis buffer with shaking for 5-10 min. One microliter of lysate was used to determine the total protein content by incubating with 99 µL Bradford reagent (Invitrogen) and reading absorbance at 595 nm (Synergy 4; BioTek Instruments, Winooski, VT). Remaining lysate was used to determine Wnt activity using the Illumination Lyophilized Firefly Luciferase Enhanced Assay Kit (Catalog number I-935-150, GoldBio, St Louis, MO) as per manufacturer’s instructions. Luminescence intensity for every sample was measured by Synergy 4 and was normalized to their respective total protein content.

RNA Sequencing and Pathway Analysis

The global gene expression profiles of KC-NCs isolated after days 7 and 14 of induction, and those treated with CHIR and BMP2 individually and in combination for 5 days were characterized by next generation RNA sequencing using Illumina platform. To this end, total RNA was isolated for all conditions for 3 donors using RNeasy Mini Kit and quality control analysis was performed by RNA gel and Agilent Fragment Analyzer. Sequencing libraries were prepared as per standard Illumina protocols (Illumina Stranded Total RNA Prep with Ribo-Zero Plus), quality checked, and quantified by Kapa Biosystems qPCR. The multiplexed libraries were sequenced in pair-end (2 × 50 bp) on the NovaSeq 6000 at 300 pM with 1% loading control. Sequencing reads passed quality filter from Illumina RTA were first processed using FASTQC (v0.10.1) for sequencing base quality control. Then sample reads were aligned to the human reference genome (GRCh38) and GENCODE (version 38) annotation database using STAR2.¹⁷ Second round of quality control using RSeQC¹⁸ was applied to the mapped bam files to identify potential RNASeq library preparation problem. Gene level raw counts were obtained using Subread¹⁹ package. Differential gene expression analysis was performed using DESeq2²⁰ and pathway analysis was performed with the Gene Set Enrichment Analysis (GSEA) method (4.0).²¹ The GSEA tool was chosen to run the analysis using the normalized gene count data that pre-filtered the low count genes. Pathway analysis was run against MsigDB, a collection of annotated and curated gene set repositories offered by the developer of GSEA (Broad Institute MIT and Harvard). This particular run used C2 of version 7.4 collection, containing 2307 gene sets from various well-known and up-to-date pathway databases such as BioCarta, KEGG, and Reactome among others. To compare KC-NC transcriptome of untreated cells and those treated with CHIR/BMP2 to the transcriptome of iPSCs, ESCs, and NCs differentiated from them, we used the published datasets from GEO database, referred to in Supplementary Table S4.

Statistical Analysis

Statistical significance was assessed by one-way or two-way ANOVA and Tukey’s multiple comparison tests was performed following that for every experiment. All analyses were done using the GraphPad Prism 8 software. Values were considered statistically significant for P < .05. Data were represented as mean ± SD for one representative experiment or multiple independent experiments. All experiments were repeated with at least 3 biological donors to ensure reproducibility.

Results

KC-NCs Lose Expression of Key NC-Specific Genes in Culture

KC-NCs express all key embryonic NC genes by day 7 after induction. However, the expression of these genes is rapidly lost after another week in culture. RNA sequencing (RNA-seq) analysis revealed that KC-NCs expressed Sox10, Sox9, HNK1, ID1, and ID4 after 7 days of induction (day 7), however the expression of these genes was significantly downregulated at 14 days post-induction (day 14; Fig. 1A-1E). To further confirm this, we analyzed the mRNA levels of the key NC-transcription factor Sox10, and genes associated with multipotency of embryonic NCs such as Pax7, HNK1, AP2A, and MycN by RT-PCR. Expression of all these genes was also downregulated in day 14 NCs as compared to day 7 NCs, indicating loss of self-renewal capacity of KC-NCs (Fig. 1F-1J). In fact, differential expression analysis revealed that genes relevant to embryonic NC development (SOX10, LINP1, OVOL2), ligands and proteins important to NC-specific signaling events (FGFBP1, BMP6, SMAD9, SMAD12), and NC differentiation (S100B, ADGRF1, ADGRG6) were among the Top 50 differentially upregulated genes in day 7 NCs, and were downregulated in day 14 NCs (Fig. 1K, 1M). In contrast, day 14 NCs expressed genes specific to fibroblastic/osteogenic phenotype such as COL14A1, KLHL41, TTN, FNDC5, MEGF10, ITGA11, MUSK, TNFRSF11B, BGN, STAC3, SGCD, COL5A3, JSRP1, MMP2 (Fig. 1L, 1M). Expression of myogenic genes DES, ACTA1, MYH3, MYH7B, ACTA2, and MYH7 was also upregulated by day 14 of KC-NC induction (Fig. 1N-1S). Hence, loss of self-renewal capacity of KC-NCs is accompanied by decreased expression of developmentally relevant NC genes and increased expression of fibroblastic/myogenic/osteogenic genes.

Figure 1.

KC-NCs lose expression of key NC-specific markers rapidly in culture. (A-E) Normalized read counts of key NC markers Sox10, Sox9, B3GAT1, ID1, and ID4 at days 7 and 14 of KC-NC culture obtained from RNA sequencing (n = 3). (F-J) Fold change in mRNA levels of key NC markers Sox10, Pax7, HNK-1, AP2A, and MycN normalized to day 14 NCs obtained by RT-PCR. (K) Heat map of Top 50 up-regulated genes in day 7 KC-NCs compared to day 14 KC-NCs, with NC specific genes relevant during development highlighted with green arrows. (L) Heat map of Top 50 down-regulated genes in day 7 KC-NCs compared to day 14 KC-NCs, with mesenchymal/osteogenic/fibroblast-associated genes highlighted with red arrows. (M) Volcano plot showing fold change in genes of interest upregulated and downregulated at day 7 (N-S) Fold change in mRNA levels of Des, Acta1, Myh3, Myh7B, Acta2, and Myh7 normalized to day 7 KC-NCs obtained from RNA sequencing. (T) GSEA analysis showing upregulation in Wnt signaling pathway at day 7 of KC-NC induction, which gets decreased by day 14. (U) Heat map of Top 50 down-regulated genes in KC-NCs treated with CHIR compared to day 14 KC-NCs, with mesenchymal/osteogenic/fibroblast-associated genes highlighted with red arrows, with fold change depicted in (V) volcano plot. (W) Heat map of Top 50 up-regulated genes in BMP2 treated KC-NCs compared to day 14 KC-NCs, with NC specific genes relevant during development and differentiation highlighted with green arrows, with fold change depicted in (X) volcano plot.

Interestingly, GSEA analysis also revealed increased Wnt signaling pathway in day 7 NCs as shown in the enrichment plot for formation of β-catenin/TCF transactivating complex from the reactome database (Fig. 1T). Since the Wnt signaling pathway has been implicated in NC induction during embryogenesis, we treated KC-NCs with the GSK3β inhibitor/Wnt agonist CHIR99021 (abbreviated as CHIR) for 5 days and compared the transcriptome CHIR-treated to untreated cells that had lost multipotency (day 14). Cells treated with CHIR downregulated the expression of fibroblastic/myogenic/osteogenic genes (Fig. 1U, 1V). Since the BMP signaling pathway has also been found essential for NC-induction during development, KC-NCs were treated with the BMP signaling ligand BMP2 for 5 days and compared with untreated day 14 cells. Interestingly, BMP2-treated cells upregulated the expression of several NC-specific genes including HOXB13, TFAP2B, FZD8, JAG1, PLXNA2, and MPZ as compared with day 14 NCs (Fig. 1W, 1X).

Treatment of KC-NCs with CHIR and BMP2 Upregulates Key NC Genes

These results prompted us to hypothesize that activating the Wnt and BMP pathways might preserve KC-NC multipotency. To this end, we expanded KC-NCs for 9 days in culture, followed by treatment with CHIR or BMP2 alone, or in combination. Untreated cells served as controls (Supplementary Fig. S1A). Interestingly, we found that the CHIR/BMP2 combination upregulated mRNA levels of bona fide NC markers Sox10, Pax7, Snail, MycN, HNK-1, TrkC, and AP2A significantly, as compared to treatment with each compound individually (Fig. 2A-2G). All these genes are responsible for NC-specification during development and are expressed in delaminating and migrating NCs.³ Of all the genes responsible for NC-gene identity, Sox10 has been identified as the master-regulator for NC multipotency and differentiation.^22-24 Protein levels of nuclear Sox10 in cells treated with CHIR and BMP2 were analyzed using immunostaining (Fig. 2H). Sox10 expression was upregulated (~50-fold) in cells treated with CHIR alone and was upregulated even further in cells treated with both CHIR and BMP2 (~150 fold; ~40% Sox10+ cells). BMP2 alone did not have any effect on Sox10 expression (Fig. 2I, 2J). Further, RNA sequencing revealed that the combination of CHIR and BMP2 upregulated the entire array of embryonic NC-genes, and in fact, expression levels of these genes were even greater than expression on day 7 (Fig. 2K), suggesting that CHIR and BMP2 might be dedifferentiating KC-NCs to more “embryonic-NC like state.” Principal component analysis (PCA) from the RNA-seq dataset showed that the samples clustered in distinct groups with CHIR/BMP2 and CHIR treated cells clustering close together and far from the BMP2 or day 7 or day 14 cells (Fig. 2L). We also found that genes pertaining to the GO Terms “Neural Crest Formation” and “Neural Crest Cell Development” were differentially expressed in cells treated with CHIR/BMP2 vs. untreated cells (day 14; Fig. 2M, 2N, P < .001), including Sox10, SEMA6A, SEMA6D, SEMA6E, and FOXC1 among others.

Figure 2.

Treatment of KC-NCs with CHIR and BMP2 upregulates key NC genes and is accompanied by decrease in cell size. (A-G) Fold change in mRNA levels of key neural crest specific genes Sox10, Pax7, SnaiI, MycN, HNK1, TrkC, and AP2A after treatment with either CHIR or BMP2 alone, or in combination normalized to control (day 14; untreated cells). (H) Immunostaining for nuclear Sox10 (red) without treatment, treatment with CHIR, BMP2, and in combination. Scale bar: 50 µm. Insets at higher magnification, scale bar: 20 µm. Nuclei were stained using Hoechst stain (blue). (I) Fold change in number of Sox10 positive cells obtained through immunostaining normalized to control (day 14; untreated cells). (J) Quantification for %Sox10 positive cells on treatment with either CHIR or BMP2 alone, or in combination. (K) Heat map of RNA-seq for NC specific genes at the indicated conditions. (L) PCA plot showing clustering of day 7, day 14 KC-NCs, and cells treated with CHIR, BMP2, and CHIR/BMP2 combination, n = 3 biological donors. Heat maps for differentially expressed genes comparing day 14 and CHIR/BMP2 treated KC-NCs belonging to the GO-term (M) neural crest formation and (N) neural crest cell maintenance (P < .05) (O) plot of forward scatter obtained using flow cytometry to quantify cell size. (P) Cell size measured as the geometric mean of forward scatter obtained using flow cytometry. (Q) Cell nuclei were stained with Hoechst and the area was quantified using ImageJ (n = 50).

Interestingly, we observed increased mRNA levels of the pluripotency genes Sox2, Klf4, and NANOG, with no significant changes in Oct4 on treatment of KC-NCs with CHIR and BMP2 (Supplementary Fig. S1B-S1E). While we did not observe NANOG protein expression (Supplementary Fig. S1F), all KC-NCs expressed Sox2 even without CHIR/BMP2 treatment (Supplementary Fig. S1G, S1J). Interestingly, CHIR/BMP2 treatment upregulated Oct4 protein levels (Supplementary Fig. S1H), with ~80% Sox10+ cells expressing Oct4 (Supplementary Fig. S1K). In addition, there was a significant increase in Klf4 protein expression with CHIR/BMP2 treatment (Supplementary Fig. S1I, S1L). However, the protein levels of Sox2, Klf4, and Oct4 were significantly lower than ESCs.

These results prompted us to examine whether KC-NCs were reprogrammed to iPSCs or ESCs upon CHIR/BMP2 treatment. To this end, we plotted Venn diagram using embryonic stem cell specific genes, that have been published by Maguire et al.²⁵ and compared the common set of genes upregulated in the samples as compared to day14 NCs that had lost multipotency. We found that 322 genes were common in all these comparisons (Supplementary Fig. S1M, Table S5). Next, we compared the transcriptome of KC-NCs with or without treatment (CHIR+BMP2 and day 14, respectively) to that of keratinocytes (KCs), day 7 NCs, iPSCs, ESCs, NCs derived from iPSCs (iPSC_NC), and NCs derived from ESCs (ESC_NCs). As evident in the PCA plot, CHIR+BMP2 treated KC-NCs clustered closer to days 7 and 14 KC-NCs and not with iPSC or ESC populations (Supplementary Fig. S1N). We also constructed PCA plots by clustering for stem cell specific genes published by Maguire et al.²⁵ As evident, KC-NCs treated with CHIR/BMP2 clustered closely with iPSC_NCs and not with iPSCs or ESCs (Supplementary Fig. S1O). This analysis suggested that while KC-NCs de-differentiate to a more multipotent state and upregulate expression of pluripotency genes on treatment with CHIR/BMP2, they did not lose their neural crest identity and did not reprogram to pluripotency.

It has been established that dedifferentiation of cells to a stem-like fate is accompanied by a decrease in cell size.²⁶ Hence, we analyzed cell size using flow cytometry (Fig. 2O). There was a significant decrease in forward scatter after treatment with CHIR, BMP2, or the CHIR/BMP2 combination (Fig. 2P). The size of DAPI stained cell nuclei also decreased significantly with treatment with CHIR and the CHIR/BMP2, while BMP2 alone had no effect on nuclear size (Fig. 2Q).

To evaluate whether CHIR/BMP2 treatment could extend KC-NC multipotency for more than 2 weeks, we extended treatment for 15 days, instead of 5 days, and evaluated NC-specific gene expression (Supplementary Fig. S2A). Extended treatment with CHIR/BMP2 preserved transcription of key NC genes Sox10, Pax7, MycN, AP2A, HNK1, TrkC, and Snail (Supplementary Fig. S2B-S2H). Interestingly, cells expanded in FI medium for 25 days followed by 5 day CHIR/BMP2 treatment (Supplementary Fig. S2I) could still upregulate key NC genes, especially Sox10 by ~100 fold (Supplementary Fig. S2J-S2P), suggesting that CHIR/BMP2 treatment could preserve KC-NC multipotency even after extended cell expansion for a month. Interestingly, during CHIR/BMP2 treatment cell proliferation decreased as seen by immunostaining for Ki67 (Supplementary Fig. S2Q, S2R) and quantification of cell numbers over time (Supplementary Fig. S2S). Hence, CHIR/BMP2 treatment rendered KC-NCs to a quiescent-like state.

Treatment of KC-NCs with CHIR/BMP2 Rescues Their Differentiation Potential

Interestingly, treatment of KC-NCs with CHIR and BMP2 for 5 days upregulated a few genes related to melanocytes and Schwann cell lineages, downregulated genes related to neuronal lineage, and had no effect on genes associated with smooth muscle cells (Supplementary Fig. S2T). To access if treatment of KC-NCs with the Wnt and BMP agonists could rescue the loss of differentiation potential, we pre-treated KC-NCs with CHIR and BMP2 for 5 days and then coaxed them to differentiate into NC-derivatives. Indeed, differentiation to smooth muscle cells improved (Fig. 3A), as depicted by increased expression of mature markers ASMA (Fig. 3B) and Caldesmon (Fig. 3C) after 2 weeks of differentiation. Cells pre-treated with CHIR and BMP2 differentiated to more mature Schwann cells within 2 weeks of differentiation (Fig. 3D), as seen by increased expression of key glial markers KROX20 (Fig. 3E) and PLP1 (Fig. 3F). They also had an accelerated differentiation to melanocytic lineage within 2 weeks of differentiation (Fig. 3G), as seen by increased numbers of MITF+ (Fig. 3H) and PMEL+ (Fig. 3I) cells and increased L-DOPA staining (Fig. 3J).

Figure 3.

Treatment of KC-NCs with CHIR and BMP2 enhances their differentiation potential. (A) Immunostaining for ASMA (green) or Caldesmon (green). Mean fluorescence intensity per cell for (B) ASMA; (C) Caldesmon; (n = 50) (D) Immunostaining for KROX20 (red) or PLP1 (red). (E) Percentage of KROX20+ CHIR/BMP2 treated cells normalized to control and plotted as fold-increase. (F) PLP1 mean fluorescence intensity per cell (n = 50) (G) Immunostaining for MITF (green) or PMEL (green). Percentage of fold-increase in (H) MITF+; or (I) PMEL+ cells in CHIR/BMP2 treated cells normalized to control. Scale bar: 100 µm. Insets at higher magnification, scale bar: 50 µm. Nuclei were stained using Hoechst stain (blue). (J) l-DOPA assay of CHIR/BMP2 treated or control KCNC that were coaxed to differentiate towards melanocytes. Scale bar: 100 µm.

Wnt and BMP2 Signaling Induce Metabolic Rewiring to Anaerobic Glycolysis

To understand the metabolic changes that accompany loss of multipotency of KC-NCs, we first evaluated glycolytic activity for KC-NCs at days 7 and 14 post-induction. Multipotent KC-NCs at day 7 exhibited significantly higher glucose uptake as seen by increased staining for 2NBDG (Supplementary Fig. S3A, S3B). Glycolytic activity was also monitored using the Seahorse XF analyzer. Similar to glucose uptake, day 7 cells exhibited higher glycolysis (albeit not statistically significant) (Supplementary Fig. S3C, S3D) and significantly enhanced glycolytic capacity, almost two-fold, as compared to day 14 cells (Supplementary Fig. S3C,S3E).

To assess whether CHIR/BMP2 treatment affected KC-NC metabolic activity, we first measured ATP levels in cells after treatment with CHIR and BMP2 alone, or in combination, and compared this to the bioenergetics of untreated cells. We observed that the combined treatment doubled ATP production; CHIR alone increased ATP to much lower extent, while BMP2 slightly decreased ATP levels (Fig. 4A). The glycolysis inhibitor, 2DG decreased ATP levels in a dose-dependent manner (Fig. 4B); the change in ATP levels with 12.5 mM 2DG was around 3-fold greater in CHIR/BMP2 treated as compared to control cells (Fig. 4C), suggesting that glycolysis is the major source of ATP in treated cells. There was a smaller change with CHIR alone and almost no change with BMP2 treatment (Fig. 4C). To further confirm this, we measured lactate secreted by cells and found that the combination of CHIR and BMP2 doubled the production of lactate as compared to untreated cells (Fig. 4D). The glycolytic inhibitor, 2DG decreased lactate under all conditions (Fig. 4E) but to a higher extent in CHIR/BMP2 treated cells (Fig. 4F).

Figure 4.

CHIR/BMP2 treatment reprograms KC-NCs to a glycolytic state. (A) ATP measurement of KC-NCs under the indicated conditions. (B) Inhibition with the indicated concentration of the glycolysis inhibitor 2-DG under the indicated conditions. (C) Fold decrease of ATP after treatment with 12.5 mM 2DG normalized to untreated cells (control). (D) Extracellular lactate measurement of KCNCs under the indicated conditions. (E) Inhibition with the indicated concentration of the glycolysis inhibitor 2-DG under the indicated conditions. (F) Fold decrease in lactate after treatment with 12.5 mM 2DG normalized to control. Enrichment plots using GSEA analysis for the pathways (G) insulin receptor recycling and (H) signaling by insulin receptor from the reactome database in cells treated with CHIR/BMP2 vs. untreated cells (day 14 control). (I-L) Fold change in mRNA levels of glucose transporters Glut1, Glut3, and Glut4, and INSR, respectively, normalized to control. (M) Flow cytometry histograms for Glut4 expression in cells treated at the indicated conditions and stimulated with insulin. (N) Geometric mean of fluorescence intensity under the conditions described in (M). (O) Fold-change of intracellular glucose levels depicting glucose uptake in KC-NCs under the indicated conditions.

GSEA analysis of the RNA-seq data suggested increased reliance on insulin signaling pathway in cells treated with CHIR and BMP2, and we observed enrichment of genes in the reactome database for insulin receptor recycling (Fig. 4G) and signaling by insulin receptor (Fig. 4H). We also found that the expression of V-ATPase family of genes, which have been shown to be activated by glycolysis,²⁷ is significantly increased in cells treated with CHIR/BMP2 (Supplementary Fig. S3F). CHIR alone increased mRNA levels of Glut1 (1.99 ± 0.27 fold; Fig. 4I), but not Glut3 or Glut4 (Fig. 4J, 4K); while the combination of CHIR and BMP2 increased expression of Glut3 (5.46 ± 0.67 fold) and Glut4 (8.79 ± 1.17 fold), and to a lesser extent also Glut1 and the insulin receptor INSR (Fig. 4I-4L). Glut2 was not expressed by KC-NCs under any of the conditions tested.

Interestingly, surface expression of Glut4 decreased after treatment with either CHIR or BMP2 but increased with the CHIR/BMP2 combination (Fig. 4M). Since Glut4 translocates to the cell surface in the presence of insulin, the level of Glut4 on the cell surface was measured after insulin treatment using flow cytometry. Interestingly, Glut4 surface expression increased significantly upon insulin stimulation of cells treated with CHIR alone, or in combination with BMP2; and to a lesser extent with BMP2 (Fig. 4M, 4N), suggesting increased insulin sensitivity. In agreement, we observed an increase in glucose uptake, as measured by intracellular glucose levels after starvation. CHIR increased intracellular glucose by 77-fold, BMP2 by about 15-fold, and in combination by ~30-fold (Fig. 4O). Interestingly, increasing d-glucose concentration in the medium from 5 mM (low) to 25 mM (high) increased the number of Sox10-expressing cells by 5.27 ± 0.6 fold; while l-glutamine concentration (1 mM vs. 10 mM) had a small effect under high glucose and no effect under low glucose concentration (Supplementary Fig. S3G, S3H). This observation was accompanied by a 4-fold increase in Sox10 mRNA levels in 25 mM (high) glucose media (Supplementary Fig. S3I), suggesting that increased glycolysis might be necessary to maintain the NC phenotype.

Enhanced Glycolysis Is Accompanied by Increased Glycolytic Enzyme Activities

Increased glycolytic flux in stem cells is often associated with increased glycolytic enzyme activities,²⁸ prompting us to measure the activity of key glycolytic enzymes after CHIR/BMP treatment. Interestingly, the activity of two key glycolytic enzymes, pyruvate kinase (PK) and phosphofructokinase (PFK) increased by treatment with CHIR and BMP2. PK catalyzes the last step of glycolysis, converting phosphoenolpyruvate (PEP) to pyruvate and generating ATP. The combined treatment increased PK activity by 6-fold (at t = 120 min) as compared to untreated cells (Supplementary Fig. S4A, S4B). The enzymatic activity of PFK, another rate limiting enzyme in glycolysis, was also increased after treatment with CHIR and BMP2 by as much as 20-fold (t = 120 min) as compared to control cells (Supplementary Fig. S4C, S4D). CHIR alone also increases PK and PFK activity, albeit to a lesser extent than the CHIR/BMP2 combination, while BMP2 had no effect on either enzyme (Supplementary Fig. S4A-S4D). In agreement, the level of pyruvate also increased by 6-fold in treated cells (Supplementary Fig. S4E) and decreased by the glycolysis inhibitor, 2DG (Supplementary Fig. S4F). Activities of other glycolytic enzymes such as hexokinase (HK) (Supplementary Fig. S4G), and lactate dehydrogenase (LDH) (Supplementary Fig. S4H) remained unaffected. Finally, the total protein level of these enzymes was not affected by any of the treatments (Supplementary Fig. S4I-S4M), suggesting that increased activity of PK and PFK was not due to increased protein, but due to increased catalytic efficiency per unit of enzyme.

CHIR and BMP2 Results in a Metabolic Fate Switch from Oxidative Phosphorylation (OxPhos) to Anaerobic Glycolysis

To measure metabolic activity, we used the seahorse XF analyzer to record extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) of cells with or without combinatorial treatment of CHIR and BMP2. ECAR was measured using the glycolysis stress rest (Fig. 5A). Cells treated with CHIR and BMP2 increased ECAR levels more than two-fold more upon injection of saturating levels of glucose as compared to untreated cells, and hence had greater glycolysis (Fig. 5B). Treated cells did not respond to the oligomycin injection, suggesting that they might be operating at maximum glycolytic capacity. Untreated cells on the other hand increased ECAR with oligomycin injection, indicating increased dependence on mitochondrial respiration and greater glycolytic reserve (Fig. 5C). The final injection of 2-DG decreased ECAR levels to the level recorded at the start of the experiment confirming that measured values were truly due to glycolytic function.

Figure 5.

Treatment of KC-NCs with CHIR and BMP2 decreases mitochondrial activity. (A) extracellular acidification rate (ECAR) per cell over time, used to calculate (B) glycolysis and (C) glycolytic reserve in control of CHIR/BMP2 treated KC-NCs. (D) Normalized oxygen consumption rate (OCR) per cell over time, used to calculate (E) maximal respiration (F) basal respiration and (G) mitochondrial ATP production in control of CHIR/BMP2 treated KC-NCs. (H) % Contribution to ATP from glycolysis and OxPhos in control and CHIR/BMP2 treated cells. Staining for (I) mitotracker red and (J) quantification of mitotracker red intensity per cell. (K) staining for mitotracker green and (L) Quantification of mitotracker green intensity per cell. (M) Staining for TMRM and (N) Quantification of mean fluorescence intensity (MFI) per cell; (J, L, N, n = 50). Scale bar: 50 µm. Nuclei were stained using Hoechst stain (blue).

Mitochondrial function was measured using the mitochondrial stress test, which measures the OCR after sequential addition of modulators of mitochondrial activity. Control cells had significantly higher basal mitochondrial respiration rates. As expected, oligomycin (ATP synthase complex V inhibitor) decreased OCR, indicating ATP production via mitochondrial respiration; injection of FCCP, increased OCR to the maximum level, while addition of rotenone and antimycin eliminated it, as expected. Surprisingly, cells treated with CHIR and BMP2 maintained very low OCR levels throughout the assay and did not respond to any mitochondrial modulators (Fig. 5D). Specifically, CHIR and BMP2 diminished the basal and maximal respiratory capacity as well as ATP production from mitochondrial respiration (Fig. 5E-5G). The contribution of glycolysis vs. OxPhos to total ATP produced in cells was measured using the Seahorse ATP rate assay. Untreated cells derived almost 40% of ATP from OxPhos, which decreased to 25% in cells treated with CHIR and BMP2. On the contrary, the treatment increased glycolytic ATP to 75% when compared with just 60% in untreated cells (Fig. 5H).

These results suggested that CHIR and BMP2 might have inactivated mitochondria. To examine this hypothesis, we stained live cells with mitotracker green dye which stains for total cellular mitochondria, as well as with mitotracker red dye, which accumulates in polarized (active) mitochondria (Fig. 5I-5L). Indeed, cells treated with CHIR and BMP2 had significantly lower mitotracker red staining, indicating very low mitochondrial potential as compared to untreated cells (Fig. 5I, 5J). On the other hand, the total mitochondria per cell remained unchanged, as shown by mitotracker green (Fig. 5K, 5L). Inactivation of mitochondria on treatment with CHIR and BMP2 was also confirmed by decreased staining for the cell permeable dye TMRM that stains for active mitochondria (Fig. 5M, 5N).

CHIR/BMP2 treatment increases phosphorylation of pyruvate dehydrogenase and shifts the fate of pyruvate from OxPhos to lactate production.

Increased glycolysis by CHIR and BMP2 prompted us to examine the fate of the end product of glycolysis, ie, pyruvate. Pyruvate generated from glycolysis enters the mitochondria through the mitochondrial pyruvate carrier (MPC), converts to acetyl CoA by the enzyme pyruvate dehydrogenase (PDH), and fuels the TCA cycle. It can also convert to lactate by the enzyme lactate dehydrogenase (LDH), resulting in anaerobic respiration. Phosphorylation of PDH by pyruvate dehydrogenase kinase (PDK) inhibits PDH activity (Fig. 6A). Immunostaining for PDH and pPDH showed that CHIR and CHIR/BMP2 treatments resulted in a significant increase in pPDH. CHIR treatment decreased total PDH levels as compared to control cells, resulting in significantly increased pPDH/PDH ratio in cells treated with CHIR and CHIR/BMP2. BMP2 treatment decreased PDH but the pPDH/PDH ratio remained similar (Fig. 6B-6E). At the same time, we observed increased expression of PDK3 (2.23 ± 0.16 fold; Fig. 6F) and especially PDK4 (26.68 ± 1.7 fold; Fig. 6G); as well as decreased MPC2 levels (1.66 ± 0.008 fold; Fig. 6H) following the combination CHIR/BMP2 treatment, indicating that pyruvate generated by increased glycolysis might not be able to enter the mitochondria and fuel the TCA cycle. Indeed, when the medium was supplemented with pyruvate, the amount of lactate produced by control cells was low, but increased by ~3-fold upon inhibition of MPC with UK5099, indicating that in control cells pyruvate was transported into the mitochondria. In contrast, CHIR/BMP2 treated cells produced significantly higher levels of lactate after supplementation with pyruvate, and MPC inhibition had no effect on lactate production (Fig. 6I), suggesting that the primary fate of pyruvate in CHIR/BMP2 treated cells was conversion to lactate.

Figure 6:

CHIR/BMP2 redirect pyruvate from TCA cycle to lactate production increasing Sox10 expression. (A) Schematic of metabolic fates of pyruvate; pyruvate can be converted to lactate by LDH or can enter the mitochondria via the mitochondrial pyruvate carrier (MPC) and converted to Acetyl CoA by PDH. UK5099 is a small chemical inhibitor of MPC. (B) Immunostaining for pPDH (red) and PDH (green) under the indicated conditions. Mean Fluorescence (MFI) per cell of (C) pPDH, (D) PDH, (E) pPDH/PDH ratio (n = 50). Scale bar: 20 µm. Nuclei were stained using Hoechst stain (blue). (F-H) Fold change in mRNA levels of PDK3, PDK4, and MPC2, respectively, normalized to control (untreated cells). (I) Fold change in extracellular lactate levels in cells cultured in media containing only pyruvate (4 mM) with or without inhibition of MPC using UK5099. Data normalized to untreated cells cultured in pyruvate only media. (J) Immunostaining for Sox10 of control and CHIR/BMP2 cells treated with or without pyruvate. (K) Fold change in number of Sox10 positive cells with or without exogenous pyruvate, normalized to Sox10 expressing cells cultured in control media. (L) Fold change in Sox10 mRNA levels with or without exogenous pyruvate, normalized to untreated cells cultured without pyruvate.

These results prompted us to examine whether supplementation with pyruvate (4mM) had an effect on Sox 10 expression. Interestingly, addition of pyruvate diminished Sox10 expression in control but enhanced Sox10 mRNA and protein expression even further in CHIR/BMP2 treated cells (Fig. 6J-6L). Since pyruvate is converted to lactate in CHIR/BMP2 treated cells, we hypothesized that lactate might be important for Sox10 expression, and hence multipotency of KC-NCs. Further, we coaxed untreated control cells to produce more lactate, by supplementing pyruvate and blocking its entry to the TCA cycle by inhibiting MPC using the small molecule MPC inhibitor UK5099 (UK) (Fig. 6I). Indeed, combination of both UK and pyruvate treatment in control cells increased Sox10 expression by 2-fold (Supplementary Fig. S5A, S5B). UK treatment alone had no effect on Sox10 expression, possibly because untreated cells produce low amounts of pyruvate (Supplementary Fig. S4E) and hence lactate (Fig. 4D).

The Glycolytic End Product Lactate Regulates NC Gene Expression by Stabilizing β-Catenin

To examine whether lactate had any effect on KC-NC multipotency, we investigated the effects of lactate on expression of the key NC-specific gene Sox10. Interestingly, day 7 KC-NCs treated with exogenous lactate (4 mM) for 5 days upregulated Sox10 expression by 2.9 ± 0.89 fold (Supplementary Fig. S5C-S5D). To test whether addition of lactate with CHIR/BMP2 could further upregulate expression of NC genes, we treated KC-NCs with 4 mM lactate either alone, or in combination with CHIR/BMP2 for 2 days. Surprisingly, while maximum transcriptional changes were observed after 5 days of CHIR/BMP2 treatment, addition of lactate upregulated expression of Sox10, Pax7, MycN, TrkC, and AP2A in just 2 days of treatment. Lactate alone however did not alter expression of these genes significantly (Fig. 7A-7E). RNA-seq also showed that addition of lactate to CHIR/BMP2 in day 14 KC-NCs resulted in greater increase of a subset of developmentally relevant NC genes, including SOX2, KLF4, HOXD4, SKIDA1, KLF13, HDAC4, HOXB2, and BMPR1A (Fig. 7F).

Figure 7.

The glycolytic metabolite lactate affects expression of NC genes by stabilizing β-catenin. (A-E) Fold change in mRNA levels of NC specific genes Pax7, MycN, AP2A, Sox10, and TrkC in cells treated for 2 days with lactate, in the presence or absence of CHIR/BMP2 treatment, normalized to untreated cells (no exogenous lactate). (F) Heatmap summarizing RNA sequencing data as fold-change in expression of NC-genes in CHIR2/BMP2 and CHIR/BMP2/lactate treated cells compared to day 14 (untreated) KC-NCs. P-values for each comparison are shown in each cell of the heatmap; ns: not significant (P > .05), 0: very significant (P < .001). (G) Fold-change in Wnt activity after 24 h of treatment with CHIR, BMP2, CHIR/BMP2, lactate, and CHIR/BMP2/lactate normalized to untreated cells. (H) Fold change in mRNA level of Axin2 in cells treated with CHIR and BMP2 individually and in combination normalized to untreated cells. (I) Fold change in mRNA level of Axin2 in cells treated for 2 days with lactate alone or in combination with CHIR/BMP2. Data were normalized to the levels in untreated cells with no lactate treatment. (J) Western blot showing total protein content of β-catenin and unphosphorylated (activated) β-catenin in cells without treatment, treatment with lactate alone, or in combination with CHIR/BMP2, quantified after normalization to GAPDH (K-M).

Finally, we examined the activity of the Wnt/BMP signaling pathway using the TOP-Flash assay after supplementation of CHIR or BMP2 alone and in combination to the medium for 24 h. As expected, CHIR alone increased Wnt activity as did BMP2 albeit to a lesser extent. The combination of CHIR and BMP2 enhanced Wnt signaling more than CHIR alone, and surprisingly supplementation with lactate enhanced Wnt even further by 2.29 ± 0.62 fold (Fig. 7G). Further, we evaluated Wnt activity by checking the mRNA levels of Axin2 which is downstream of the TCF/Lef binding domain and is an indicator of enhanced Wnt activity.²⁹ Indeed, CHIR/BMP2 also increased Axin2 mRNA levels (Fig. 7H), which were further increased by supplementation of lactate (Fig. 7I). Interestingly, the levels of non-phosphorylated β-catenin increased by CHIR/BMP2 and further increased by supplementation with lactate (Fig. 7J-7M), suggesting that lactate might increase Wnt signaling by stabilizing β-catenin in synergy with CHIR/BMP2.

Discussion

In this study, we demonstrated the signaling pathways that are essential to maintain the stemness of adult tissue-derived NCSCs. We have shown that treatment of KC-NCs with a chemical cocktail of CHIR and BMP2 is sufficient to rewire cellular transcriptome by altering the metabolic requirements of these cells to preserve their multipotency. This treatment improved the differentiation capacity of KC-NCs, making them more suitable for applications in regenerative medicine. Furthermore, we demonstrated the close interactions between signaling and metabolic pathways and showed that the glycolytic metabolite lactate might have a role in promoting the expression of key-NC genes.

While NCSCs from various adult tissue sources as well as those differentiated from induced pluripotent stem cells (iPSCs) or embryonic stem cells (ESCs) have been successfully characterized and isolated over the years, their clinical applications have been limited by their notoriously transient phenotype.^13,30-32 In culture, these cells rapidly lose expression of key transcription factors along and their ability to differentiate into NC-specific lineages. This limits the extent to which they can be expanded after isolation before loss in plasticity, a criterion essential for their clinical application, given that conventional cell therapy procedures require large numbers of highly potent cells. As a result, it is important to identify strategies to prevent loss of multipotency of NCSCs, preferably without genetic manipulation. Previous studies showed that cultures of chick embryo-derived NCSCs can be maintained in culture as self-renewing crestospheres, though the culture conditions vary based on the axial-identity of NCSCs.^33,34 Mimicking the signaling events that induce NCSC formation during embryogenesis has also been found to be useful in maintaining these cells in culture. Of these, the Wnt pathway holds special relevance, as it has been shown to regulate the NCSC multipotency network.^35-37 While the combination of Wnt/BMP signaling has been shown to be essential for maintaining multipotent NCSC cultures from chick explants,³⁸ here we show that this signaling axis could also maintain multipotency of KC-NCs derived from human epidermis. Perhaps more importantly, we provide evidence that this may be at least in part due to metabolic reprogramming to a highly glycolytic state that may trigger transcriptional changes leading to expression of NC-specific genes and suppression of pro-fibrotic genes.

Recent studies aiming to understand the signals involved in maintaining stem cell plasticity found that the metabolic requirements of multi- and pluripotent stem cells differ from those of differentiated cells. Reprogramming of somatic cells to iPSCs is accompanied by a bioenergetic shift from oxidative phosphorylation (OxPhos) to glycolysis,^39-41 and a high glycolytic flux was found to be crucial for maintenance of stemness.^28,42 In this context, the transcription factor family of hypoxia-inducible factors (HIF), especially HIF-2α is critical in regulation of glycolysis^43,44 as well as for NCSC proliferation, migration, and self-renewal.^45-48 In this study, we find that Wnt/BMP signaling axis reprograms KC-NC multipotency by modulating cellular energetics and metabolism. Specifically, we observed increased glycolytic ATP, extracellular lactate, and glucose consumption in cells treated with both CHIR and BMP2 together, while individual treatments did not have such a dramatic effect. In fact, simply supplementing the media with glucose could increase Sox10 expression in KC-NCs, suggesting that increased glycolysis might regulate NC gene transcription. Increased glycolysis was accompanied by increased activities of glycolytic enzymes PFK and PK, possibly to enable breakdown of the increased amounts of glucose entering the cells. In addition, we observed increased expression of PDK3/4 and increased phosphorylation of PDH in CHIR/BMP2 treated cells, suggesting that pyruvate was inhibited from entering the TCA cycle and was rather converted to lactate. Interestingly, activation of PDK4 was shown to be crucial in regulating glycolysis, often functioning as a metabolic checkpoint for maintenance of stem cell quiescence.⁴⁹

The role of mitochondria in maintaining pluripotency has also been well studied. iPSCs were shown to have poorly developed mitochondria,⁵⁰ low mitochondrial content,^51,52 and low mtDNA copy number.⁵³ However, other studies showed that OxPhos is crucial for generating the majority of ATP in ESCs,⁵⁴ suggesting differences between iPSC and ESCs. Interestingly, a recent study by Hu et al. found that the metabolic requirements of primed iPSCs differ greatly from those of naïve iPSCs, the latter having higher dependence on OxPhos for energy generation and hence more active mitochondria.⁵⁵ These studies suggested that mitochondrial activity might depend on the exact state of pluripotency. In the case of adult KC-NCs, we observed that upon treatment with CHIR/BMP2, they upregulated NC-specific genes that were lost in culture and at the same time, they showed marked decrease in mitochondrial activity and ATP generation. Interestingly though, unlike iPSCs the mitochondria of KC-NCs did not seem to be degenerated, and in fact, they still possessed intact mitochondria as seen by mitotracker green staining.

To understand whether CHIR/BMP2-induced metabolic reprogramming was crucial to rewire KC-NC transcriptome, we examined the role of glycolytic metabolite lactate in regulating NC gene transcription. Lactate has been most widely studied and recently established as an important oncometabolite governing transcription of key oncogenes in several types of cancer.^56,57 Lactate can cause transcriptional changes in cells either by stimulating epigenetic modifications,^58,59 or binding to the orphan G-protein coupled receptor (GPCR) HCAR1, resulting in a cascade of signaling events having diverse implications in cellular metabolism, growth, and survival.^60-62 While the effects of lactate have been studied in cancer and endothelial cells, its potential role in stem cell multipotency remains elusive with some studies suggesting that lactate could influence the transcriptome of mesenchymal stem cells,^59,63,64 but did not contribute to pluripotency of iPSC.^65,66 In this study, we show that lactate could serve as a metabolic regulator of KC-NC phenotype. Our results showed that as glycolysis is activated by CHIR/BMP2, production of lactate stabilizes β-catenin, further increasing the activity of the Wnt signaling pathway and transcription of NC-specific genes. Interestingly, NC-induction during development takes place in an environment lacking oxygen supply, where lactate is present in abundance.⁷ However, the exact mechanism through which lactate may be regulating NC multipotency remains unknown.

Conclusion and Future Perspectives

We have shown that cell signaling pathways that are crucial during NC development also regulate the metabolic profile of KC-NCs. Specifically, we show that Wnt activation alters the metabolism of KC-NCs to confer multipotency. Our study provides simple means to preserve the multipotency of KC-NCs in culture without genetic manipulation. Given the accessibility of human skin and ease of isolation and expansion of human KC and KC-NCs, maintaining the multipotency of KC-NCs during expansion may have significant implications for their use in regenerative medicine and disease modeling.

Funding

This work was supported by grants from the National Institutes of Health R01 EB023114 (S.T.A.) and the New York Stem Cell Science NYSTEM (Contract #C30290GG, S.T.A.).

Conflict of Interest

The authors declared no potential conflicts of interest.

Author Contributions

P.M., S.T.A.: experiments planned and designed. P.M.: generated, collected, and analyzed in vitro experimental data. P.M., I.I.: performed metabolic assays. P.M., G.T.: collection of skin tissues and isolation of KCs. Y.Z., J.W., P.M., S.L.: RNA sequencing analysis. P.M., I.I., S.T.A.: data analysis and interpretation. P.M., P.L., M.B., S.T.A.: writing and critical revisions of the manuscript.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.