Stem Cells, Phenotypic Inversion, and Differentiation

Abstract

Stem cells possess the potential to cure a myriad of ailments ranging from congenital diseases to illnesses acquired through the physiological process of aging. In the adult, these cells are extremely rare and often difficult to isolate in numbers sufficient to apply to medical treatment. Ex vivo expansion of these cells will be required for most meaningful interventions. The discovery of stem/progenitor cell inversion offers a new avenue for obtaining sufficient numbers of stem cells. Adult progenitor cells are much more common than quiescent stem cells and can be isolated with minimal interventions; therefore, inversion of progenitors to stem cells may become a feasible approach for therapeutic purposes. Stem cells are known to possess few mitochondria, and mitochondrial biogenesis is required for stem cell differentiation. The microtubule cytoskeleton is a major regulator for mitochondrial biogenesis. Investigations in the area of controlling cell differentiation and inducing phenotypic inversion, possibly through manipulation of mitochondrial biogenesis, may contribute to stem cell-based therapies.

Introduction

Stem cells possess a great potential to transdifferentiate [1-3]. Bone marrow stem cells have been shown to form bone, skeletal muscle, cardiac muscle, endothelium, liver and lung parenchyma, and neural tissue [4-10]. Recently, stem cells have emerged as a promising therapy for a variety of disorders, including neurodegenerative diseases, spinal injuries, and infarcted myocardium [4, 11, 12]. At the present time, however, knowledge about the fundamental properties of stem cells, including understanding their differentiation potential and plasticity, as well as how to direct stem cells to differentiate to a desired lineage, remains limited. This review will discuss the evolution of stem cell research with emphasis on the concept of a stem cell continuum, phenotypic inversion, and influence of mitochondrial and microtubule properties on stem cell differentiation.

Historical Overview

Virchow first proposed an undifferentiated cell population in the human body in 1855. Shortly thereafter, Julius Conheim expanded Virchow's concept while engaged in research to determine the origin of cancer [13,14]. Conheim proposed that a vestigial pool of cells, which he termed “embryonic rest,” remained intact from the developing fetus into adulthood. While this idea alludes to stem cells, the origin of our modern concept of a primitive pool of cells regulating tissue function would not become realized for another 50 years. In 1917, Pappenheim speculated that an undifferentiated cell gave rise to progenitor and precursor cells of all the blood lineages [15]. This theory guided research until 1961 when McCulloch and Till published their definitive study describing colony-forming units (CFUs) in the mouse spleen after irradiation and whole bone marrow transplantation [16, 17]. In their studies, McCulloch and Till put forward two important caveats to define stem cells: the ability of self-renewal and multipotentiality. Without a thorough understanding of the surface phenotype, stem cells could only be identified certainly by their ability to form colonies in the spleens of irradiated, recipient mice. Nevertheless, this was the first report to unequivocally identify hematopoietic stem cells (HSCs).

Studies in stem cell biology remained focused on hematopoiesis in this period until Gail Martin discovered embryonic stem cells (ESCs) in 1981 [18]. This report described cells isolated from preimplantation mice, that when injected into adult mice, could produce teratocarcinomas. This property is indicative of the pluripotent nature of ESCs. Human ESCs were discovered 17 years later, when James Thomson isolated a cell derived from the early blastocyst that had high telomerase activity and lacked phenotypic markers of any known cell lineages [19]. In 1990, Potten and Loeffler refined the definition of stem cells as: “undifferentiated cells capable of (a) proliferation, (b) self maintenance, (c) the production of a large number of differentiated, functional progeny, (d) regenerating the tissue after injury, and (e) a flexibility in the use of these options [20].” The one component missing from Potten's and Loeffler's definition is the concept that stem cells must reside in niches. In 1978, Schofield first theorized the niche concept as the cellular environment which retains the stem cell and prevents its further maturation [21]. In 2000, Xie and Spradling proved the histological existence and the function of the niche for germ line stem cells in the ovary of Drosophila melanogaster [22]. Three years later, the niches for hematopoietic stem cells were also identified [23, 24].

Endogenous adult stem cells (ASCs) from tissues other than the bone marrow have been identified since 1990. ASCs were discovered first by Potten and Loeffler in the intestinal crypt [20]. A major breakthrough occurred in 1999, with two independent reports describing a resident pool of ASCs in the brain, previously thought to be a post-mitotic organ [25, 26]. In 2001, a multipotential cell was discovered in the dermis of adult mice that was capable of differentiating into neurons, glia, smooth muscle cells, and adipocytes [27]. In 2003, the existence of a multipotential progenitor cell was observed in the mouse heart, another organ thought to be post-mitotic [28]. Lately, a previously overlooked cell from the bone marrow was described as being a very small embryonic-like stem cell (VSEL-SC) that displayed characteristics of totipotent, zygote cells [29]. This discovery might confirm Virchow's and Conheim's original concept of the embryonic rest. These studies, performed in a variety of tissues representing all germ layers and containing resident stem cell pools, have created a new dogma that all adult tissues likely possess a regulatory stem cell population for homeostatic turnover and tissue maintenance.

With the development of investigation in the fields of ESCs and ASCs, methodology for isolating stem cells was rapidly evolving. Of particular interest, the use of vital dyes emerged as a tool for isolating ASCs into short-and long-term repopulating HSCs (STHSCs and LTHSCs). Rhodamine-123 (Rho-123) in 1985 and Hoechst 33342 in 1996 were used to isolate side population cells [30, 31]. LTHSCs were originally speculated to efflux dyes through the multidrug resistance (MDR) transporters. However, in 1998 Spangrude showed that low fluorescence observed in LTHSC populations stained with Rho-123 was indicative of low mitochondrial activity, and not dye efflux [32]. This elegant study set a precedent for identifying stem cells by examining the mitochondrial activity of these cells.

ESCs are considered pluripotent (capable of differentiating into all germ layers), while ASCs are postulated by many as being only multipotent (capable of differentiating into a limited number of lineages) [33, 34]. Transdifferentiation is the ability of a cell to differentiate into an entirely different cell lineage, regardless of the original state of lineage commitment. In 1974, Selman and Kafatos reported that cuticle cells in the larval silk gland of the silkmoth were capable of “metamorphosing” into salt-producing cells without dividing [35]. The transdifferentiated cell possessed two distinct mitochondrial networks, whereas the progenitor cell contained fewer mitochondria than the differentiated cell. In 1986, Eguchi proposed a mechanism of dedifferentiation, followed by redifferentiation to explain the conversion of vertebrate pigmented epithelium to lens cells in their studies [36]. In 1998, HSC plasticity was described by Ferrari and colleagues, who showed that bone marrow cells could differentiate into muscle fibers and regenerate damaged tissue [37]. While Ferrari's results expanded the understanding of HSC multipotentiality, pluripotent properties have since been discovered in HSCs. HSCs have been shown to give rise to neural and connective tissue [38]. Research, as highlighted in these examples, is beginning to alter our thinking on the nature of hierarchical organization in stem cell commitment.

Surface Markers, Phenotypic Inversion, and the Stem Cell Continuum

A variety of surface antigens are used to isolate stem cells. These include stem cell antigen-1 (Sca-1), ckit (stem cell factor receptor), and CD34. Sca-1 was discovered 20 years ago and is specific for mouse SCs, while CD34 was first reportedly used to isolate human HSCs in 1984 [39, 40]. ckit was discovered in 1986 and was shown to be expressed on both human and murine HSCs in 1989 [41, 42]. Isolation of cells based on these methods almost always requires depletion of cells committed to a specific lineage. In terms of HSCs, the surface antigen profile is dictated by the organism studied. Mouse HSCs are lineage negative (Lin^-), ckit positive (ckit⁺), and Sca-1 positive (Sca-1⁺); whereas in humans, HSCs have been shown to be Lin- and CD34 positive (CD34⁺) [43, 44]. In contrast, mouse HSCs are CD34 negative (CD34^-) [45]. This difference between organisms is surprising, but more interesting is the surface antigen expression within a particular organism's pool of HSCs and hematopoietic progenitor cells (HPCs). The specificity of surface phenotype suggests that the stem cell is a distinct functional unit.

Mouse HSCs lose Sca-1 expression as they begin to differentiate, but ckit remains highly expressed [46]. Analysis of the two populations shows that Lin-Sca-1⁺ckit⁺ (KSL) cells form three times as many colonies as Lin-Sca-1^-ckit⁺ cells in an in vitro CFU assay [43]. Analysis of HSCs from Sca-1 knock-out (KO) mice shows an impaired ability of these cells to repopulate the bone marrow of irradiated mice [47]. The cells isolated from these KO animals also have diminished multipotentiality. Compared to cells from wild-type (WT) animals, KO cells produced less in vitro granulocyte, erythroid, macrophage, and megakaryocyte CFU (GEMM-CFU). Weissman reported a Lin-Sca-1⁺ckit^- population isolated from the bone marrow that was incapable of responding to hematopoietic growth factors [48]. The cells in this study did not form colonies in irradiated mouse spleens, nor did they reconstitute the bone marrow. They concluded that this population of cells is either a resting stem cell, or an entirely new subset of HSCs. The first conclusion is supported by Norio Suzuki, who found cells bound in the niche express Sca-1 and GATA2 (via GATA2-directed GFP expression) [49]. The cells in these studies were all in the G0/G1 phase of the cell cycle. Weisman's second conclusion has been supported more recently with the discovery of the aforementioned VSEL-SC, although Ratajczak's group did not analyze ckit expression on their cells [29]. Another study using the identical Lin-Sca-1⁺ckit^- phenotype found these cells had the ability to differentiate into hepatic-like cells in the absence of cell fusion [50].

A phenomenon recently described by Quesenberry's laboratory contradicts the notion of a hierarchy between stem and progenitor cells [51]. They isolated stem cells on the basis of Hoechst 33342 efflux and mitochondrial potential, rather than surface phenotype. The benefit to this method is that all the isolated cells are quiescent and/or synchronized in the cell cycle, which allowed them to correlate the phenomenon to cell cycle stage. His group used in vitro assays to analyze high-proliferative potential colony forming cells (HPP-CFC) and colony-forming unit cultures (CFU-c). In vivo competitive repopulation assays were used in conjunction to the HPP-CFC and CFU-c assays to determine HSC versus HPC. The results showed an inverse relationship between HPP-CFCs and engraftment ability. The cell cycle status was also analyzed, and cells in the S/G₂ phase of the cell cycle where found to have a decreased ability to engraft when compared to cells in G₀/G₁. The results confirmed that this defect of engraftment was reversible with cell cycle transit. Quesenberry termed this phenomenon a progenitor/stem cell inversion.

When Quesenberry examined the surface phenotype ex vivo of cells isolated from BrdU treated mice, he found that 22% of the KSL, Thy1.1^lo cells were in S-phase [52]. The Weissman laboratory sorted cells that were Lin-ckit⁺Sca-1⁺Thy1.1^lo and then resorted for cells in G0, G1, or S-G2/M [53]. They then tested engraftment capacity in limited dilution competitive reconstitution assays. Only cells in G0 were able to rescue lethally irradiated mice for >12 weeks. This conclusion argues strongly for cell cycle dependence. Additionally, Jocobsen showed that KSL cells lose their self-renewal capacity when they become positive for flt-3 [54]. In fact, KSL, flt³⁺ cells show an increased differentiation potential towards the lymphoid lineage, and only serve as STHSCs for the myeloid lineage. KSL cells are considered stem cells, but without further depletion of HPC markers, many cells in these studies might represent lineage-committed cells. In a similar study, Ogawa's laboratory isolated KSL cells and then looked at CD34 and CD38 expression in murine HSCs [55]. They discovered a reciprocal relationship between CD34 and CD38. In vitro, cytokine stimulation shifted expression towards the CD34⁺ phenotype, and in vivo reconstitution experiments with KSL, CD34⁺ cells, showed that when the recipient bone marrow achieved steady-state, the cells were predominantly KSL, CD34^-, CD38⁺ in mice (Figure 1). This study highlights the fact that surface phenotype of the HSC population should be considered heterogeneous and that progenitors have the potential for inversions.

Figure 1.

Murine Stem/Progenitor Cell Inversion in the Niche. (1) CD38+ stem cell attached to stromal cell in niche. (2) Stem cell undergoing asymmetric division. (3) CD38+ stem cell remains in the niche. (4) CD34+ progenitor cell receives signals from the microenvironment. (5) CD34+ progenitor cell may leave the niche and continue differentiating, or invert and remain in the niche as a new CD38+ stem cell.

Bernstein examined Sca-1 in C2C12 myoblastic cells in vitro [56]. These cells contain a side population subset that resembles stem cells. The results show that Sca-1 is transiently upregulated upon cell cycle withdrawal when C2C12 cells are induced to form myotubes. When Sca-1 mRNA expression was inhibited with either antisense nucleotides or protein with monoclonal antibodies, the C2C12 cells remained proliferative and did not form myotubes. If this characteristic of Sca-1 expression is the same in stem cells, then in order to amplify the stem cell pool, Sca-1 would be expected to be down regulated. As aforementioned in stem cells, Sca-1 appears to be the most primitive marker [43]. Together with Quesenberry's, Weissman's, and Bernstein's results, Sca-1 might be transiently down regulated in stem cells in a cell cycle-dependent manner [51, 54, 56]. In our recent studies, we have observed that Lin^-ckit⁺Sca-1^- progenitor cells can invert to a KSL stem cell phenotype both in vitro and in vivo in response to a variety of inflammatory stimulants (unpublished data).

Stem cells have been shown to dedifferentiate in other model systems, which also suggest a continuum exists. Spradling's group showed that cystocytes in both larvae and adult Drosophila can revert to stem-like cells [57]. In the larvae stage, these cells differentiated after depletion of the stem cell pool with cytotoxic agents or heat shock. In adult flies, overexpression of decapentaplegic (dpp), the Drosophila homologue to the known stem cell regulator bone morphogenic protein 4 (BMP-4), was able to stimulate a reversion to a more primitive germline stem cell phenotype.

The expression patterns of HPCs and HSCs are remarkably similar; therefore inversions should not be surprising. Catherine Laiosa, Matthias Stadtfeld, and Thomas Graf reviewed this aspect extensively, and showed that a single transcription factor (TF) can convert HPCs of one lineage into a separate lineage; however, the cells in these studies were not synchronized in the cell cycle [58]. Altering either PU1.1 or GATA1 expression can transdifferentiate the myeloid and lymphoid precursors. Quesenberry reported hotspots at defined points in the cell cycle of TF expression [59]. Early in S-phase, cells showed a reversible propensity to differentiate to the megakaryocyte lineage. While late in S-phase, the cells were predisposed to the granulocyte lineage—also reversibly. Quesenberry's group also showed that TFs of the megakaryocyte lineage were expressed during the differentiation hotspot.

Recent investigations have reported the creation of embryonic-like stem cells from fully differentiated cells. Two groups, Marius Wernig and Kazutosh Takahashi and Shinya Yamanaka reprogrammed mouse fibroblasts with ectopic expression of four transcription factors: POU domain, class 5, transcription factor (Oct3/4), sex determining region Y-box2 (Sox2), cellular myelocytomatosis oncogene (cMyc), and Kruppel-like factor 4 (Klf4) [60, 61]. In 2007, human fibroblasts were also reprogrammed to an ESC-like state, termed induced pluripotent stem (iPS) cells, by transfection with the previously mentioned TFs or with Sox2, Oct4, NANOG, and Lin28 [62,63]. These TFs are all highly enriched in pluripotent stem cells. Further, Wernig analyzed chromatin remodeling and found epigenetic markers similar to ESCs. This would suggest that genomic alterations are important for the dedifferentiated phenotype. Nevertheless, the cytosolic milieu appears to be more important in determination of stem cell phenotype. Cloning studies have been instrumental in furthering this hypothesis. Nuclear transfer of adult, differentiated nuclei into enucleated ova creates totipotent stem cells as seen in the early zygote. Wilmut and Campbell produced live lambs from adult mammary gland, fetus, and embryo to confirm that adult cells are capable of forming stem cells [64]. These results showed definitively that genetic modifications are not the sole determinant for a differentiated phenotype.

Transfer of adult nuclei into oocytes and the subsequent generation of stem cells suggests that other mechanisms, besides chromatin and TF expression, are responsible for stem cell plasticity. Isolation of HSCs based on cell surface antigen expression invariably gives a heterogeneous population of HSCs and HPCs. Using mitochondrial energetics, with Rho-123 fluorescence, and Hoechst 33342 efflux isolates a population known to be quiescent. Intriguingly, the same percentage of HSCs is recovered when cells are sorted for low Rho-123 fluorescence or Sca-1 positivity [65]. Either group is equally capable of hematopoietic reconstitution. Other markers of energy metabolism have also been used in the isolation of HSCs and stem cells in general. Daniel Pearce examined human cord blood surface phenotype and showed a strong correlation with aldehyde dehydrogenase (AHD) activity [66]. Greater than 90% of CD34⁺CD38^- cells and CD34⁺CD133⁺ cells were also positive for high AHD activity. These two surface phenotypes are considered HSC populations. Michael Kastan and John Hilton compared AHD activity between HPCs and lymphocytes, and found high AHD activity in HPCs [67]. High activity levels of AHD protect the stem cells from retinaldehyde and oxidative stress during the transition from anaerobic glycolysis to oxidative phosphorylation (OXPHOS) for energy production [68]. This, with the data discussed earlier regarding Sca-1 as the most primitive marker, should lead research to more closely examine energy metabolism in Sca-1⁺ and KSL cells.

Mitochondrial Properties and Stem Cells

Mitochondria are as heterogeneous as the tissues in which they are examined. Each cell type possesses a certain range of mitochondria per cell [69]. Proteomes vary dramatically between cell types, and mitochondrial DNA (mtDNA) copy number may be one to ten per organelle. Fundamental properties, such as subcellular localization, mitochondrial membrane potential (ΔΨ_m), and reticular networks are also different between cell types [70, 71]. These facts suggest that a particular cell type might be identified based on these mitochondrial parameters.

Jankovic examined the role of Id1, a helix-loop-helix transcription factor, in myeloid differentiation [72]. Using Id1 KO mice, they discovered that HSCs prematurely entered into myeloid commitment, and also inappropriately expressed myeloerythroid genes. Tanaka and Finkel also examined this inhibitor of differentiation and found that Id1 was able to produce apoptosis in neonatal and adult cardiomyocytes, but not in endothelial cells (ECs) or fibroblasts [73]. In seeking to determine the mechanism of action, they discovered that Id1 functions through a redox-dependent mechanism. Cardiomyocytes are almost entirely dependent on mitochondria for ATP production, whereas ECs and fibroblasts rely more heavily on glycolysis [75]. These results offer clues as to the mechanisms of differentiation as they pertain to energy metabolism.

The environments of the uterus before placentation and the marrow surrounding HSC niches are anaerobic [75]. Feasibly, this prevents oxidative damage to the early zygote. To produce ATP in this environment, stem cells, from the fertilized oocytes to the adult stem cells, rely heavily on glycolysis for cellular ATP production, and thus do not require a large number of mitochondria [76]. As such, any lineage commitment requires an increase in OXPHOS and, subsequently, an increase in mitochondrial ATP production. This is achieved through two processes: mitochondrial differentiation and mitochondrial biogenesis [77]. Mitochondrial differentiation is simply the activation of mitochondria. This is most often equated with an increase in the ΔΨ_m. Mitochondrial activation correlates to many processes like subcellular localization, but Ca⁺⁺ is probably the main effector [78, 79]. In fact, isolation of stem cells based on Rho-123 is indicative of undifferentiated mitochondria. Biogenesis involves replication of mitochondrial genomes, as well as an increase in all proteins necessary to form a new mitochondrion from an existing organelle.

Mitochondrial biogenesis occurs through the process of binary fission, similar to their bacterial progenitors [80]. The regulation of this process is tissue specific, but all tissues examined require the expression of the peroxisome proliferator-activated receptor gamma coactivator-1 alpha (PGC-1α), beta (PGC-1β), or PPAR-γ, coactivator-related 1 (PPRC) [81-83]. This protein family regulates many of the transcription factors for both mitochondrial-encoded genes and nuclear-encoded genes, in particular mitochondrial transcription factor A (tFAM) and nuclear respiratory factor 1 and 2 (NRF-1 and NRF-2) [82]. Studies have shown that mitochondrial biogenesis is particularly important to stem cell differentiation [84-88]. This complex process is known to encompass about 20% of total cellular protein [89].

The peroxisome proliferators-activated receptor (PPAR) family of TFs has been shown to be crucial in the control of cellular metabolism, by regulating genes of the different metabolic pathways [90, 91]. PGC family members differentially recruit the PPAR TFs under different conditions. PPAR-alpha (PPAR-α) has been shown to be crucially important in fatty acid metabolism, while PPAR-gamma (PPAR-γ) has been extensively studied in gluconeogenesis [92]. PPAR-γ, in conjunction with PGC-1α increases mitochondrial biogenesis [93]. PPAR-γ has also been shown to inhibit canonical wingless homologue (Wnt) signaling, which is a crucial pathway in maintaining an undifferentiated state [94]. Notch signaling has been shown to upregulate PPAR-γ [95]. Notch has been studied mostly in its ability to direct differentiation towards specific lineages [96-98]. Wnt and Notch signaling antagonize each other, and this might be due to influences of PPAR-γ [99].

The many molecular pathways governing mitochondrial biogenesis have varying importance depending on the tissue. In skeletal muscle and granulosa cells, calcium (Ca⁺⁺) is a major initiator of biogenesis [100, 101]. In adipose tissue, nitric oxide (NO), thyroid hormone, and rosiglitazones have been the most intensively analyzed pathways for increased mitochondrial biogenesis [102-104]. NO and hydrogen peroxide (H₂O₂) increase biogenesis in B cell chronic lymphocytic leukemia (CLL) and osteosarcoma cells, respectively [105, 106]. All of these initiators, either directly or indirectly, lead to an increased expression of PGC-1α [107-110].

In vivo PGC-1α is essential for the increased expression of nuclear-derived transcription factors involved in mitochondrial biogenesis, which are primarily mitochondrial transcription factor A (TFAM) and nuclear respiratory factor 1 (NRF1) [82]. PPRC performs similar functions as PGC-1α, but this coactivator is not regulated by thermogenesis [111]. However, PPRC is expressed under in vitro conditions if PGC-1α is not. These two regulatory proteins have complementary functions.

The activation of PGC-1α expression in vivo is under the direct control of both the myocyte enhancer factor 2 (MEF2) and the forkhead box O (FOXO) families of transcription factors [112, 113]. In addition, cyclic adenosine monophosphate (cAMP) response element-binding (CREB) proteins interact with cAMP response elements (CRE) in the promoter of PGC-1α to activate transcription [114]. v-akt murine thymoma viral oncogene homolog (AKT) phosphorylation of FOXO sequesters the TF to the cytoplasm, leading to FOXO inactivation; however, PKA phosphorylation of MEF2 increases transcriptional activity [115, 116]. Members of the p38 mitogen-activated protein kinase (MAPK) family interact with AKT and are involved primarily in cellular responses to environmental stress [117]. p38 MAPK phosphorylates MEF2 family members. p38 increases both the transcription of AKT and the activation of AKT through phosphorylation [118, 119]. Additionally, p38 MAPK phosphorylates PGC-1α protein, increasing its nuclear localization [120].

PGC-1α can interact with both MEF2 and FOXO family members to increase its own expression in a feed-forward loop [119]. In skeletal muscle, Ca⁺⁺ acts as a potent activator of MEF2 transcription through calcium/calmodulin-dependent protein kinase IV (CaMKIV) and calcineurin. CaMKIV increases CREB activity through phosphorylation and activates calcineurin [121]. Calcineurin dephosphorylates nuclear factor of activated T-cells (NFAT), which translocates to the nucleus where it interacts with MEF2 to activate PGC-1α expression [122]. NFAT knock-out (KO) mice die at embryonic day 10.5 and present with diminished ventricular myocyte proliferation containing abnormal, swollen mitochondria [123]. NFAT transcription and translation are enhanced by p38 MAPK, but phosphorylation of NFAT by p38 sequesters it in the cytoplasm [124]. Activated calcineurin leads to p38 inactivation [125]. In ESCs, bone morphogenic protein 4 (BMP4) inhibits p38 MAPK allowing for SC self-renewal [126]. Conversely, reactive oxygen species (ROS) upregulate p38, leading to increased HSC turnover and SC exhaustion [127]. p38 and Ca⁺⁺ initiate the signals involved in PGC-1α regulation, thereby increasing mitochondrial biogenesis and causing differentiation in a level and time-dependent manner (Figure 2).

Figure 2.

Signal Transduction Pathway for Regulators of PGC-1α Expression and Mitochondrial Biogenesis. A) Regulators of PGC-1α are interconnected in a complex network of positive and negative feedback loops. This ensures that a prolonged signal favoring mitochondrial biogenesis is required for enlargement of the mitochondrial pool. B) Transcriptional control of PGC-1α expression and mitochondrial biogenesis.

Mitochondrial mass and activity increases with differentiation [128-131]. Inhibition of mitochondrial biogenesis prevents terminal differentiation [82, 132]. This suggests cells possessing a progenitor phenotype will maintain a progenitor phenotype without increases in mitochondrial mass. Similarly, mitosis in a progenitor without a concomitant increase in mitochondrial number could feasibly lead to the production of stem cells. Increased mitochondrial mass is seen in germline stem cells, ESCs, and adult stem cells when they are induced to differentiate [129, 131, 133].

The literature is replete with methods for the in vitro differentiation of stem cells. Three popular agents that have been used to induce stem cell differentiation in vitro are dexamethasone (DEX), retinoic acid (RA), and dimethyl sulfoxide (DMSO) [134-141]. DEX is a glucocorticoid receptor (GR) agonist, and RA activates the retinoic acid receptor (RAR/RXR) family. These agents have been shown to upregulate transcripts for both PGC-1α and PGC-1β, and RAR/RXR forms complexes with PGC-1α [142, 143]. DMSO activates mitochondrial Ca⁺⁺ signaling, affecting differentiation through a different mechanism [141, 144]. DEX has been shown to play a crucial role in mitochondrial biogenesis in a tissue-specific manner [145]. In adrenal-ectomized rats treated with DEX, liver mitochondrial content is returned to sham-operated control values, but heart mitochondrial mass was unchanged. DEX directly increases PPAR-γ expression.

When PPAR-γ is induced in mouse bone marrow (BM) or peripheral blood, angiogenic progenitor cells differentiated towards the endothelial progenitor cell (EPC) lineage [146]. These cells concomitantly showed a decreased potential to differentiate into the smooth muscle cell lineage. PPAR-γ induction also increases the multipotential murine cell line, C3H10T1/2, to differentiate towards the adipocytes lineage, while inhibiting osteoblast differentiation [128]. Conversely, treatment with RA led to differentiation into the osteogenic lineage [147]. Increased mitochondrial mass accompanied the different-tiation in these cells. In human bone marrow cells, DEX enhanced osteoblast differentiation [137]. In a separate study, mouse mesenchymal stem cells (MSCs) preferentially differentiated towards the adipocytes lineage, and were inhibited from the osteogenic lineage [138]. These results appear to be conflicting; however, the intracellular milieu and species of origin may determine the end effect of DEX treatment.

The selective differentiation of DEX is also seen in monomyelocytic HPCs [135]. DEX rapidly caused neutriphil differentiation, whereas vitamin D3 led to monocyte/macrophage lineage commitment. It is important to note that PGC-1α has been shown to coactivate the D3 receptor [143]. Different blood cell lineages display preferential expression of PPAR family members [148]. Myeloid cells predominantly express PPAR-γ, while lymphocytes mostly express PPAR-α. Intriguingly, DEX treatment of erythroid progenitors actually leads to proliferation, but RA treatment causes differentiation [136]. These results suggest that a finely tuned mechanism controlling the metabolic state of stem cells is active in determining different-tiation capacity.

RA is the active metabolite formed from retinaldehyde, a downstream product from dietary retinol [149]. PGC-1α is a coactivator for the RAR/RXR family of nuclear receptors [143]. As aforementioned, PGC-1α plays a major role in mitochondrial biogenesis. Aldehyde dehydrogenase (AHD) is a key enzyme in the production of RA [149]. Three AHD isoforms are largely responsible: AHD2, AHD7, and AHD-M1 [150]. The first two are cytoplasmic and the last one is mitochondrial. Stem cells and progenitor cells have been shown to express high levels of AHD activity, and this fact has been used in their isolation [66].

Similar to DEX treatment, RA displays differential effects on differentiation in a time-and tissue-dependent manner. This is likely due to the large number of different RAR/RXR receptors. Treatment of murine ESCs with RA during the first two days of development promoted neurogenesis, but inhibited cardiomyogenesis [144]. Treatment for two to five days increased skeletal muscle and adipocyte differentiation, in addition to neurogenesis. Treatment with RA after five days promoted cardiomyogenesis and smooth muscle cell differentiation, while inhibiting the myogenic and adipogenic lineages. Human ESCs could not be induced to differentiate into cardiomyocytes with RA, displaying species differences or ESC heterogeneity in RA's mechanism of action. In promyeloid progenitor cells, RA has been shown to drive commitment to the granulocyte lineage [151]. However, in KSL stem cells, RA appears to maintain the quiescent, undifferentiated state [152]. RA has also been shown to alter the lineage of differentiated cells. When white adipocytes are treated with RA, these cells display gene expression profiles and properties of brown adipose tissue (BAT) [153]. BAT is enriched for mitochondria as compared to white adipose tissue (WAT) [154]. Similar increases in mitochondria are seen in murine ESCs. RA consistently increased the level of transcripts for tFAM and polymerase G, the major polymerase for mtDNA replication [86].

ΔΨ_m, representing the mitochondrial proton-motive force, is responsible for controlling Ca⁺⁺ uptake, respiratory rate, mitochondrial ATP production, and reactive oxygen species (ROS) production [130, 155, 156]. Mitochondria within a single cell have varying ΔΨ_m, depending on the subcellular localization and the relative health of the mitochondrion [70]. Most often, researchers consider ΔΨ_m by averaging intensity for a single cell or multiple cells due to the prohibitive nature of studying single mitochondrion in situ. Analyzing the average ΔΨ_m offers an adequate appraisal of total cellular oxidative metabolism.

Rho-123 is commonly employed to analyze ΔΨ_m, since incorporation of the fluorophore increases with increasing ΔΨ_m [32]. As previously mentioned, LTHSCs can be isolated from the bone marrow using Rho-123. Morphologic analysis of Rho-123^low cells revealed small mitochondria with diffuse cytoplasmic distribution, whereas Rho-123^high cells had two-fold larger mitochondria and a perinuclear distribution. The modest increase in mitochondrion size would not account for the thirty-fold increase observed in Rho-123 fluorescence. Additionally, cell cycle analysis showed that 3% of Rho-123^low and 33% of Rho-123^high cells were in the S/G2 phase. These data suggest that cell cycle status alone is a poor indicator of engraftability. In accordance with this data, osteoprogenitor cells (OPCs) induced to differentiate show an increased retention of Rho-123 in a time- and concentration-dependent manner [157]. Two peaks in Rho-123 are seen at days 4 and 10 in both control and induced cultures, with the induced cultures retaining more dye at all time points. This fluctuation seen in the uninduced OPCs is similar to the cyclic changes observed in ΔΨ_m from single, isolated neural mitochondria [158].

In single mitochondrion analysis, the oscillations in ΔΨ_m from total depolarization to physiological levels occurred at a rate of one per minute, and they were shown to be Ca⁺⁺- dependent. In murine erythroleukemia cells (MEL) induced to differentiate with DMSO, ΔΨ_m remains relatively constant for the first 10 hours of treatment [141]. A total collapse in ΔΨ_m occurs prior to the appearance of differentiated progeny. This decay in potential corresponds to a release of mitochondrial Ca⁺⁺ stores. Depolarizing the mitochondria with an uncoupling agent before induction eliminated the 10 hour lag time for MEL differentiation. Again, this depolarization corresponded to increased cytosolic Ca⁺⁺ concentration. In P19 cells, DMSO treatment almost instantly causes an increase in cytoplasmic Ca⁺⁺ [144]. These cells express a very early gene of the mesoderm lineage, Brachyury T. DMSO also causes a similar Ca⁺⁺ rise in mouse embryonal carcinoma cells, which may lead to their differentiation. This is consistent with the role mitochondria fulfill as intracellular Ca⁺⁺ buffers, representing a possible feedback mechanism for PGC-1α regulation. Thus, ΔΨ_m controls differentiation by enhancing mitochondrial biogenesis through Ca⁺⁺ signaling and by increasing ATP supply through OXPHOS.

Since ΔΨ_m is responsible for the rate of ATP synthesis, cytosolic ATP levels would be expected to regulate membrane potential. In fact, the Ca⁺⁺-dependent fluctuations seen in ΔΨ_m from isolated mitochondria are stabilized by micromolar concentrations of ATP and ADP [158]. The mammalian target of rapamycin (mTOR), a downstream kinase in the AKT pathway and chemotherapy target, has recently been described as a cytosolic ATP sensor, independent of its ability to sense amino acid (AA) concentrations [159]. The classical functions of mTOR have been studied in terms of regulating ribosomal biogenesis and cell growth, and increasing protein translation rates by phosphorylation of S6 kinase 1 (S6K1) and initiation factor 4E binding protein 1 (4E-BP1). mTOR and mTOR/raptor complexes have recently been purified from mitochondrial fractions [160]. This colocalization with the mitochondria was abolished by treatment with rapamycin. Additionally, rapamycin treatment and small interfering RNA (siRNA) directed towards mTOR binding partners lead to a decrease in ΔΨ_m, oxygen consumption, and ATP production. Thus, mTOR integrates signals from nutrient concentration, ATP levels, and growth factor stimulation in order to modify the translational response of the cell, and functions in a feed-forward manner to increase mitochondrial activity.

Perhaps the best evidence to date suggesting a role for mitochondrial biogenesis, ΔΨ_m, and subcellular localization in affecting different-tiation is in the study of oocytes [161, 162]. Mitochondria are few in number and low in activity level in immature oocytes. The subcellular distribution is diffuse. As the oocytes mature, activity level increases, and distribution becomes pericytoplasmic. Mitochondrial ΔΨ_m becomes intermediate up until the 16-cell blastocyst stage, after which it decreases. The distribution of mitochondria in inner mass cells, from which ESCs arise, is perinuclear, and the ΔΨ_m remains low. Cultured ESCs retain these parameters, and loss of this subcellular localization is correlated to differentiation or quiescence [163].

Microtubules in the Regulation of Mitochondria and Stem Cell Differentiation

Microtubules (MTs) are the most important protein in the control of organelle subcellular localization, including the mitochondria [164]. This function is critical during interphase, but during the G₂/M phase of the cell cycle, an additional microtubule (MT) network forms called the mitotic spindle [165]. The mitotic spindle is regulated by the MT organizing center (MTOC). The MTOC is comprised of a centriole, which forms the centrosome. During mitosis, the location of the maternal centriole is responsible for determining the cleavage plane of the stem cell [166, 167].

Redistribution of the MT network corresponds to differentiation. MT distribution changes during development [168]. As the fertilized oocytes proceed through the 8-cell stage, MTs reorganize from perilamellar to a perinuclear arrangement. This correlates to the mitochondrial distribution early in development [161]. The centriole, as expected, also reorganizes in a similar manner. As epithelial cells mature, the typical non-polarized progenitor cell acquires an apico-basal organization [169]. Centrioles migrate to the nascent apical surface, and reorganize the MT cytoskeleton from a radial to an apico-basal distribution.

Maternal centrioles reproduce once during the course of the cell cycle [170]. After replication, the daughter centriole migrates to the opposite end of the cell from the maternal centriole. The mature maternal centriole is capable of forming a more stabilized MT network [171]. This is largely due to a larger number of MT associated proteins (MAPs) remaining with the maternal centriole. In adult germline stem cells in Drosophila, the stem cell is anchored to the supporting cells of the niche [170]. When these cells undergo asymmetric division to form a daughter progenitor cell, the stem cell retains the mature, original centriole. Two hypotheses exist for this phenomenon: the more organized MT network maintains the mitochondrial localization and prevents biogenesis, or the maternal centriole acts a basal body, creating a primary cilium that serves as a nurse cell communication domain [172]. A primary cilium has a large surface area-to-volume ratio, and could feasibly act as an amplifying center for any signals sent by the nurse cell to maintain quiescence.

Regardless of the actual mechanism, this phenomenon of asymmetric division, where the stem cell retains the proximity of the niche, is seen in many tissues of multiple organisms. In skeletal muscle of mice, satellite cells are paired box gene 7 positive (Pax7⁺) and myogenic factor 5 negative (Myf5⁺) [173]. As these cells undergo asymmetric division, the basal cell retains the satellite cell phenotype, while the apical cell becomes Myf5 positive (Myf5⁺). When these two populations were transplanted into injured muscle, Pax7⁺Myf5⁺ cells contribute to the satellite cell pool, and the Pax7⁺Myf5⁺ cells undergo marked differentiation into new muscle cells.

Niche proximity and mitotic spindle orientation also plays a role in final progenitor cell commitment. In the developing mouse brain, cleavage plane determines the fate of the neural progenitor cell (NPC) [174]. While a cleavage plane parallel to the niche did not inhibit differentiation in all cases, a perpendicular cleavage plane always resulted in a symmetric division that produced two neurons. These studies taken together suggest that mitotic spindle stability is a function of the maternal centriole, and that disparities in centriole maturity can determine symmetric versus asymmetric division and commitment of progeny.

Mitochondrial biogenesis can be altered by MT stability. The mechanisms involved are not clear, but Wnt signaling and cytoplasmic Ca⁺⁺ concentration might be the modulator of this process. Wnt signaling increases MT stability through activation of janus kinase (JNK) and inhibition of glycogen synthase kinase 3 beta (Gsk3β). As mentioned above, Wnt signaling is downregulated by PPAR-γ [175]. This suggests a link between MT stability and major TFs involved in mitochondrial biogenesis. The adapter protein, myeloid differentiation primary response protein-5 (MyD88-5), has been demonstrated to link MTs and mitochondria and to coordinate JNK localization to the mitochondria [176]. When MTs are destabilized, MyD88-5 dissociates from tubulin. In cells with high MyD88-5 expression, mitochondria localize around the MTOC. MT organization is required for optimal store operated calcium release (SOCE) from the endoplasmic reticulum (ER), through regulating the location of the ubiquitously expressed Ca⁺⁺ channel activator, STIM1 [177]. The perinuclear arrangement of mitochondria in stem cells, therefore could allow for proper cytoplasmic-mitochondrial-ER Ca⁺⁺ homeostasis. As stated previously, Ca⁺⁺ has an essential role in mitochondrial biogenesis.

MT stability also appears to be regulated by p38 MAPK. Disruption of p38 attenuates MT disruption by tumor necrosis factor alpha (TNFα) and nocodazole, a pharmacological MT disrupter [178, 179]. Environmental stress, such as oxidative damage, activates p38; however the role of p38 in oxidant-induced MT disruption have not been examined [180]. Nevertheless, H2O2 and NO (from iNOS) cause MT disruption, and both of these oxidants lead to increased mitochondrial biogenesis [181, 182]. Interestingly, MT stability does appear to have opposing effects depending on the state of differentiation. When human keritinocytes (skin SCs) are treated with a MT disrupting agent, E-cadherin-catenin complexes are reorganized to the cell periphery [183]. These complexes anchor the SCs to the niche cells, maintaining quiescence. When the same treatment is performed in progenitor cells, no relocalization of this junctional-signalling complex is observed. These data allude to the fact that mitochondrial biogenesis is required for final commitment, and before biogenesis occurs, progenitor cells may invert to stem cells.

When MTs are destabilized by heterozygous deletion of survivin, a MAP, in mouse HPCs, decreased formation of enucleated erythroid cells is seen [184]. In human CD34⁺ HPCs, the peptide, shepherdin, causes survivin depletion. At low levels of shepherdin treatment, HPCs show diminished erythroid burst-forming units. At higher shepherdin concentrations, granulocyte-macrophage CFUs (GM-CFUs) and granulocyte erythrocyte macrophage megakaryocyte CFUs (GEMMCFUs) are also inhibited. This research, with the above centriole localization studies suggest a mechanism by which differentiation hotspots and inversions might correlate to cell cycle status. However, this cell cycle-dependency of differentiation potential might only relate to the maturity of the MT network and the level of mitochondrial biogenesis.

Perspectives

Understanding stem cell differentiation is critically important for the clinical application of these cells. At the present time, expansion of stem cells ex vivo is required for conducting potential treatments. Controlling the differentiation state of stem cells will increase the efficacy of these treatments. In addition, seemingly contradictory results from clinical trials with stem cells might be eliminated as we acquire a better understanding of their phenotypic variations, particularly the role of the mitochondria and their biophysical properties in regulating phenotypic changes [185, 186]. The phenomenon of stem cell inversions could represent a new mechanism in the regulation of adult stem cell turnover. Further studies on this mechanism may provide novel information for understanding how stem cells maintain tissue homeostasis. Advancement of knowledge in this area will form a foundation upon which new strategies of stem cell therapy will be developed. The progenitor pool of most stem cell compartments is considered rapidly dividing and greatly outnumbers the stem cell pool. Therefore isolation of progenitors could yield a sufficient number of inverted stem cells in order to support many new therapeutic protocols with minimal interventions in patients.