Author contributions: J.C.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; A.D.: collection and/or assembly of data, data analysis and interpretation, approval of manuscript draft; M.Y.: conception and design, collection and/or assembly of data, data analysis and interpretation, approval of manuscript draft; M.S.G. and G.E.: provision of study material, approval of manuscript draft; L.I.: conception and design, financial support, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS October 2, 2008.
Recent studies have provided important insight into the homeoprotein LIM homeobox transcription factor 1α (Lmx1a) and its role in the commitment of cells to a midbrain dopamine (mDA) fate in the developing mouse. We show here that Lmx1a also plays a pivotal role in the mDA differentiation of human embryonic stem (hES) cells. Thus, as indicated by small interfering RNA experiments, the transient early expression of Lmx1a is necessary for the coordinated expression of all other dopamine (DA)-specific phenotypic traits as hES cells move from multipotent human neural progenitor cells (hNPs) to more restricted precursor cells in vitro. Moreover, only Lmx1a-specified hNPs have the potential to differentiate into bona fide mDA neurons after transplantation into the 6-hydroxydopamine-treated rat striatum. In contrast, cortical human neuronal precursor cells (HNPCs) and mouse subventricular zone cells do not express Lmx1a or become mDA neurons even when placed in an environment that fosters their DA differentiation in vitro or in vivo. These findings suggest that Lmx1a may be critical to the development of mDA neurons from hES cells and that, along with other key early DA markers (i.e., Aldh1a1), may prove to be extremely useful for the selection of appropriately staged and suitably mDA-specified hES cells for cell replacement in Parkinson's disease. STEM CELLS2009;27:220–229
Parkinson's disease (PD) is a neurodegenerative disorder whose symptoms of tremor, rigidity, and bradykinesia are caused by the degeneration of dopamine (DA) neurons in the midbrain. Although the symptoms of early-stage PD can be alleviated by a variety of palliative treatments, efficacy declines and treatment-related side effects emerge with the inexorable progression of the illness. Consequently, cell replacement remains an important potential therapy, particularly at later stages of the disease (reviewed in [1, 2]). One promising source of renewable cells for transplantation in PD are human neural progenitor cells (hNPs) derived from pluripotent human embryonic stem (hES) cell lines. Their success, however, greatly depends on the discovery of ways to promote their proper differentiation into midbrain dopamine (mDA) neurons.
Toward this end, a number of protocols have been devised in which hES-derived hNPs can be induced to express traits of a DA phenotype in culture [3–11]. Like other ex vivo-bred DA neurons, these cells do not survive harvest and transplantation into the brain [12, 13]. Importantly, however, if hES cells are implanted while still hNPs, many cells are able to continue the DA differentiation process in situ [3, 5, 9]. Since this has not been observed with transplanted hNPs derived from other sources (i.e., fetal cerebral cortex, adult bone marrow), it raises the intriguing possibility that a subset of hES-derived hNPs are uniquely fated to become DA neurons prior to their engraftment. In this study, we examined whether these neurons are further specified to become midbrain-specific DA neurons.
Although the process by which cell fate is determined in mDA neurons is incompletely understood, several key molecular events have recently been elucidated in the developing mouse midbrain [14–20]. Of critical importance is the expression of the gene for LIM homeobox transcription factor 1α (Lmx1a), which is both necessary and sufficient for the induction of an mDA phenotype in midbrain neuroepithelial and Shh-treated mouse embryonic stem (ES) cells . Acting as a transcriptional activator, Lmx1a further induces msh homeobox 1 (Msx1), which both promotes neuronal differentiation (i.e., inducing the such as protein neurogenin 2 [Ngn2]) and suppresses alternative ventral cell fates (i.e., downregulating the floor plate marker Nkx6.1) . In addition to Lmx1a, the forkhead transcription Foxa2, through its regulation of Ngn2, has also recently been implicated in the specification of mouse mDA neurons during development and in their continued survival in the adult [18, 19].
Another known early marker of mDA progenitors in the mouse is the enzyme that catalyzes the oxidation of retinaldehyde to retinoic acid, retinaldehyde dehydrogenase (Aldh1a1). Aldh1a1 is first observed in the mouse midbrain at embryonic day (E) 9.5, but its expression persists in postnatal and adult midbrain neurons [21, 22], where it is thought to play a crucial role in the survival of A9 dopamine neurons . Soon after Aldh1a1 is detected, a number of transcription factors (Pitx3, Nurr1) implicated in the survival and maintenance of a mature DA phenotype are also expressed [16, 22, 24, 25].
Currently, little is known about the role of these molecules in the commitment of human cells to an mDA fate. In this study, we examined whether Lmx1a is also instrumental in the differentiation of a mDA phenotype in hES cells in vitro and in vivo. By following expression of Lmx1a, Aldh1a1, and other mDA and cell stage markers, we describe individual steps in the mDA fate restriction process as hES cells move from mDA-specified hNPs to precursors to terminally differentiated mDA neurons. Understanding these lineage relationships and developing ways to identify cells at each of these stages may be key to the selection of suitable hES cells for transplantation in PD.
MATERIALS AND METHODS
H9 hES cells were purchased from WiCell Research Institute (Madison, WI, http://www.wicell.org) and maintained as detailed in Iacovitti et al. . Details can also be found in supporting information data. Human neuroblastoma cell line BE(2)-C-16 (CRL-2268; American Type Culture Collection, Manassas, VA, http://www.atcc.org) were maintained in Dulbecco's modified Eagle's medium/Ham's F-12 medium with 10% fetal bovine serum. Primary hNPs (derived from 19–21 week fetuses) were purchased from Clonexpress (Gaithersburg, MD, http://www.clonexpress.com) or were generously supplied by ScienCell Research Laboratories (San Diego, CA, http://www.sciencellonline.com) and were grown as described in . Adult mouse subventricular zone (SVZ) cells were dissected from transgenic nestin-green fluorescent protein (GFP) mice [26, 27].
Cultures were fixed with 4% paraformaldehyde. Embryonic brains (E10.5–E18.5) from tyrosine hydroxylase (TH)-GFP mice () and C57BL6 mice (Taconic Farms, Germantown, NY, http://www.taconic.com) were immersion-fixed in 4% paraformaldehyde overnight. Transplanted rat brains and adult nestin-GFP mouse brains were perfused with cold periodate-lysine-paraformaldehyde (4%). Adult brains were sectioned at 30 μm and embryonic brains at 15 μm on a freezing microtome and processed for immunocytochemistry as described previously . Details are listed in supporting information data.
RNA Isolation and cDNA Synthesis
Total RNA was isolated directly from freshly collected five differentiation stages of H9 cells or after small interfering RNA (siRNA) treatment using TRIzol LS (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). cDNA was synthesized by using 100 ng of total RNA in a 20-μl reaction with Superscript III (Invitrogen). One microliter of RNaseH (Invitrogen) was added to each tube and incubated for 20 minutes at 37°C before proceeding to polymerase chain reaction (PCR).
Real-Time PCR Analysis
Real-time PCR was carried out using a 7500 Real-Time PCR System with SYBR Green PCR master mix (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) in triplicate for each sample. All PCR products were checked first on agarose gel and then by dissociation assay each time to exclude the possibility of multiple products. Primers are listed in supporting information Table 1.
Overexpression of Lmx1a and Lmx1a siRNA
Full-length cDNA encoding human Lmx1a and enhanced green fluorescent protein (EGFP) were amplified and cloned into pIRES (Clontech, Palo Alto, CA, http://www.clontech.com), respectively. The construct pLmx1a-IRES-GFP was sequence-verified and transfected into BE(2)-C-16 cells.
Three pairs of verified Stealth RNAis for human Lmx1a and Negative Control Duplex (Invitrogen) were transfected twice into stage IV H9. Additional details are given in supporting information data.
6-Hydroxydopamine Lesions and Transplantation
Animals were maintained in accord with the Office of Animal Resources at Thomas Jefferson University and Institutional Animal Care and Use Committee policies. As described previously , nine Fischer-344 rats (Taconic Farms) were made parkinsonian and implanted with 1 million cells for these studies. Additional details are given in supporting information data.
Lmx1a Expression in hES-Derived mDA Neurons in Culture
We began our studies by first investigating whether Lmx1a was expressed in cells in our hES (H9; WiCell) cultures. To do so, expression of the mDA fate gene was tracked in hES cells as they progressed from the undifferentiated state (stage I) through the mDA differentiation process (stages II–V) (as described previously in ). Low levels of Lmx1a were first detected by reverse transcription (RT)-PCR as early as the embryoid body stage (stage II), but they continued to be expressed at all stages in hES differentiation (Fig. 1A). To determine in which cells Lmx1a was expressed, hES cultures were simultaneously stained with antibodies to Lmx1a protein and the developmental-specific markers Sox2 (present in hNPs) and nestin (present in hNPs and more restricted precursors) . Regardless of stage, Lmx1a was overwhelmingly expressed by nestin+Sox2+ hNPs (Fig. 1B, inset) in rosette-like colonies (Fig. 1B, 1C, arrowhead) and was strikingly absent from surrounding nestin+Sox2− precursors (Fig. 1C, arrow). Thus, the mDA-specifying gene/protein was transiently and exclusively expressed in hES cells at an early stage of multipotency (hNPs) but vanished as cells differentiated further into more restricted precursor cells.
To verify that Lmx1a-expressing hNPs indeed represented prospective DA neurons, additional markers of a DA phenotype were examined by RT-PCR and immunocytochemistry. Similar to Lmx1a, expression of its downstream mediator Msx1 and mRNAs for the DA transcription factors Foxa2, Pitx3, and the enzyme Aldh1a1 were also detected throughout the hES differentiation process (Fig. 2 A). In contrast, transcripts for the transcription factor Nurr1 and the DA biosynthetic enzyme TH were not detected until stage V (Fig. 2A), when fully mature process-bearing β-tubulin III+ DA neurons were first observed in these cultures . Colocalization of these markers in the Lmx1a+ cell population was not always possible, as in the case of Pitx3 and Nurr1, since only antibodies of the same species (rabbit) yielded reproducible staining, leaving open the question of whether these factors were expressed only in DA neurons or in other cells as well. However, double-label immunocytochemistry of rabbit anti-Lmx1a and goat anti-Aldh1a1 revealed distinct populations of singly and dually labeled cells (Fig. 2) that were quantified in stage IV and stage V cultures (Fig. 2G, 2H). Although there was some variability from one plating to another, approximately half of the colonies in stage IV cultures contained Lmx1a+ cells (data not shown). When these cultures were costained for Aldh1a1, we found a few clusters of Lmx1a+-only cells (no Aldh1a1) (Fig. 2B, white arrowhead; Fig. 2G), but the vast majority of Lmx1a+ cells coexpressed Aldh1a1 at this stage (Fig. 2C, yellow arrowhead; Fig. 3). The converse, however, was not true, as many Aldh1a1+ cells did not express Lmx1a (Fig. 2C, yellow arrow; Fig. 2G). In general, these Lmx1a−Aldh1a1+ cells were found surrounding colonies of Lmx1a+Aldh1a1+ cells, consistent with the former arising from the latter. In contrast, Foxa2, another potentially important transcription factor in mDA specification, was present in a relatively small number of stage IV cells, and surprisingly few of those coexpressed Lmx1a or Alh1a1 (supporting information Fig. 1A, 1B) suggesting a potentially different fate for Foxa2+ cells.
At Stage V, despite the robust development of TH+ neurons, there continued to be copious numbers of cells at earlier steps in the fate restriction process. Thus, Lmx1a+Aldh1a1− hNPs (Fig. 2D, white arrowhead), Lmx1a+Aldh1a1+ hNPs (Fig. 2D, yellow arrowhead), and Lmx1a−Aldh1a1+ precursors (Fig. 2D, yellow arrow) were all represented in stage V cultures, although in considerably lower proportions than in stage IV (i.e., Lmx1a+Aldh1a1a+ cells: 54% in stage IV vs. 4% in stage V; Lmx1−Aldh1a1+ precursors: 37% in stage IV vs. 26% in stage V) (Fig. 2G, 2H). Whether the change in cell composition in stage V cultures represented a selective loss of the aforementioned cells or proliferation of other cells remains unclear as yet. Regardless, 20% of stage V cells progressed down the DA differentiation pathway to express TH. These clusters of TH+ neurons (Fig. 2E, white arrow) were almost always found lying immediately adjacent to Lmx1a+ cell colonies (Fig. 2E, white arrowhead). On very rare occasion, a cell of transitional phenotype, which expressed TH and residual Lmx1a (Fig. 2E, orange arrow, inset), could be found straddling the two cell populations, consistent with the notion that the latter arose from the former. Significantly, all TH+ cells were also Aldh1a1+ (Fig. 2F, blue arrow), confirming that Aldh1a1+ stage IV cells indeed gave rise to the 20% Aldh1a1+TH+ mDA neurons seen in stage V cultures. Thus, Aldh1a1 serves as an early marker of DA-specified cells that, because of its enduring expression, allows cells to be tracked as Lmx1a is downregulated in hNPs and TH expression is upregulated during the DA differentiation process.
In addition to DA phenotypic traits, we also assessed these cultures for markers of other neurotransmitters, such as gamma-aminobutyric acid (GABA). Importantly, no TH+ stage V hES cells colabeled for GABA (Fig. 2H; supporting information Fig. 2A) as compared with cortex-derived human neuronal precursor cells (HNPCs) (Clonexpress), which contained numerous TH+GABA+ cells (supporting information Fig. 2B). Taken together, these findings suggest that Lmx1a indeed may serve to specify hES cells to follow a differentiation pathway leading not only to the acquisition of DA traits but to the development of a full bona fide mDA phenotype.
Lmx1a Expression in Transplants of hES-Derived mDA Neurons
Similarly, when hES cells (stage IV) were transplanted into the 6-hydroxydopamine (6-OHDA)-lesioned rat striatum and analyzed by confocal microscopy, a mix of mDA-specified cell phenotypes were found in various parts of the graft. Thus, at 3–5 weeks after transplantation, colonies of Lmx1a+Aldh1a1− (Fig. 3A, white arrowhead) and Lmx1a+Aldh1a1+ (Fig. 3A, yellow arrowhead) Sox2+ (Fig. 3A, inset) hNPs and Lmx1a−Aldh1a1+ precursors (Fig. 3B, yellow arrow) were present. As seen previously , TH was also observed in the graft, occasionally in a transitional Lmx1a+Aldh1a1+TH+ cell (Fig. 3A, orange arrow) but more often in mature Lmx1a−Aldh1a1+TH+ DA neurons (Fig. 3A, 3C, blue arrow). Importantly, none of these TH+ cells colabeled for GABA (although GABA+ neurons were indeed present in the grafts) (supporting information Fig. 3); indicating that, as in culture, engrafted cells may have differentiated into authentic mDA neurons. In addition to mDA and GABA neurons, however, many other cells types, including dividing cells [3, 5, 9], were also present in these grafts, further emphasizing the need for purification of appropriately committed cells before transplantation.
Requirement of Lmx1a for the mDA Differentiation of hES Cells
The expression of the Lmx1a gene in hNPs at appropriate developmental times in hES cells, although correlative, does not establish its requirement for the mDA differentiation of hES cells. To test whether Lmx1a was in fact critical to the commitment of hES cells to a mDA fate, siRNA was used to substantially reduce the gene's expression in stage IV cells (when the number of Lmx1a+ hNPs was greatest in number), after which expression of Lmx1a and other DA neuron-specific mRNAs were analyzed and compared with non-DA markers using semiquantitative PCR. When untreated and negative (Stealth RNAi Negative Control Duplexes; Invitrogen) control cultures (C1–C4) (supporting information Fig. 4) were compared with siRNA-treated cultures (S1–S4) (Stealth RNAs for human Lmx1a; Invitrogen) (Fig. 4), we found comparable amounts of mRNAs for the metabolic enzyme glyceraldehyde-3-phosphate dehydrogenase and the neuronal cytoskeletal protein β-tubulin III. However, siRNA treatment substantially reduced the expression of mRNAs for Lmx1a (∼16-fold) and, concomitantly, the full profile of DA-specific genes, such as Aldh1a1 (∼4-fold), Nurr1 (∼4-fold), Pitx3 (∼64-fold), TH (∼16-fold), dopamine transporter (DAT) (∼16-fold), and Girk2 (∼4-fold; data not shown), compared with controls (PCR cycles in two-cycle increments; in Fig. 4, each lane represents an increase of approximately fourfold in PCR product compared with the previous lane). Because Lmx1a gene expression could not be entirely eliminated in stage IV cells by this method, TH protein continued to be detected by immunocytochemistry in stage V cells (data not shown). These data suggest that Lmx1a was necessary for the coordinated expression of other downstream mDA phenotypic markers in hNPs.
We next investigated whether the mere expression of Lmx1a was sufficient to induce an mDA phenotype de novo in human cells. Using full-length cDNA encoding human Lmx1a, we generated a pLmx1a-IRES-GFP construct (pIRES; Clontech) (supporting information Fig. 5A) for transfection studies. Because of the inherent difficulty of transfecting cell aggregates present in stage III and IV hES cultures, we used monolayer cultures of a clonal line of hNPs (BE(2)-C-16) or HNPCs (Clonexpress) for these experiments. Even using these cells, the test pLmx1a-IRES-GFP construct was transfected with only 8% efficiency (3% double-labeled yellow nuclei and 5% Lmx1a-only red nuclei, the latter resulting from the incomplete translational efficiency of IRES with a bicistronic vector ). In comparison, it was possible to transfect 20%–30% of cloned hNPs with the smaller control construct (pEGFPN3) (data not shown). Despite low transfection efficiency, we found that an average 9,700-fold (p < .05) increase in Lmx1a mRNA in test cultures by real-time PCR (supporting information Fig. 5C), as compared with no change in Msx1 (1.05 ± 0.13), TH (1.30 ± 0.17), Pitx3, and Nurr1 mRNA (data not shown). Similar results were also obtained by immunocytochemistry; thus, despite Lmx1a being readily detected in the pLmx1a-IRES-GFP-transfected cells (supporting information Fig. 5B) compared with pEGFPN3-transfected cells (data not shown), there was no increase in TH expression or TH+ cell number (supporting information Fig. 5D, 5E). Likewise, no increase in TH expression was seen by real-time PCR or immunocytochemistry (data not shown) in comparable experiments with Lmx1a-transfected HNPCs. These results suggest that Lmx1a expression alone may not be sufficient to drive DA differentiation de novo in clonally derived or primary hNPs.
Transient Lmx1a Expression in the Mouse Midbrain
To ascertain whether the mDA differentiation process observed in hES cell cultures and transplants correlated with that seen in vivo, the expression of early markers was tracked in the developing midbrain of transgenic mice carrying the human TH promoter driving EGFP  or wild-type (C57BL6) mice. As shown in Figure 6A, Lmx1a was observed in nestin+Sox2+ midbrain NPs (Fig. 5 A, inset) in the floor plate of the embryonic day (E) 10.5 embryo (earliest time point examined). Importantly, these Lmx1a+ cells coexpressed Foxa2 (although there were also many Lmx1a−Foxa2+ in adjacent regions of the floor plate (supporting information Fig. 6A, 6B). By E11.5, TH-GFP expression was also detected in the midbrain (Fig. 5B) in a subset of Lmx1a+ cells (Fig. 5C). Even though some Sox2+ midbrain NPs continued to be present at this stage, they did not express TH-GFP (Fig. 5C, inset). Thus, in contradistinction to hES cultures and grafts, in the E11.5–E12.5 TH-GFP (Fig. 5 B–5D) and wild-type (supporting information Figs. 6C, 6D, 7A) mouse brain, expression of Lmx1a (and Foxa2) was not restricted to NPs and was not simultaneously downregulated with the first appearance of TH in differentiating ventral mDA neurons. Nonetheless, over time, expression of these early genes declined (Fig. 5E), and by E15.5, Lmx1a (Fig. 5F) and Foxa2 (supporting information Fig. 6E, 6F) expression was extinguished in mDA neurons (although Lmx1a+ and Foxa2+ cells continued to be present elsewhere in the brain). Importantly, by E18.5, Lmx1a, and to a lesser extent Foxa2 reappeared in ventral midbrain neurons (supporting information Fig. 8A, 8B), and both continued to be expressed in the adult (supporting information Fig. 8C, 8D).
Absence of Lmx1a Expression in Nonmidbrain-Derived Progenitor Cells
Since mouse and human mDA neurons are derived from cells that express Lmx1a, presumably during the mDA-specifying process, it follows that the progenitors of non-mDA neurons should not express Lmx1a. To test this notion, we tracked Lmx1a expression in primary hNPs (HNPCs from Clonexpress) derived from nonmidbrain regions of the embryonic human brain (predominantly cerebral cortex) and adult neural stem cells in the mouse SVZ. We found that undifferentiated nestin+ HNPCs did not constitutively express Lmx1a in culture (Fig. 6A) and did not acquire detectable levels of Lmx1a after treatment with TH-inducing factors  (Fig. 6B). Moreover, no Lmx1a+ expression was observed in engrafted human nuclear antigen+ HNPCs (Fig. 6C). These results indicate that neural progenitors that give rise to neurons of other transmitter types do not express Lmx1a and are not mDA-specified even when placed in an in vitro or in vivo environment that promotes the development of DA traits (reviewed in ).
Likewise, adult neural progenitors present in the SVZ of nestin-GFP mice [26, 27] did not constitutively express Lmx1a (Fig. 6D), nor did the DA periglomerular cells (Fig. 6E, inset) that they give rise to in the olfactory bulb . Furthermore, when nestin-GFP+ SVZ cells were grown as neurospheres in vitro, they did not express Lmx1a in culture (data not shown) or after transplantation in vivo (Fig. 6F). Taken together, these findings further support the notion that Lmx1a is uniquely expressed in those neural progenitor cells specified to become mDA neurons and not other brain DA neurons.
Recent studies have provided critical insight into the process through which mouse midbrain cells commit to an mDA neuronal fate during development and the role that genes/proteins such as Lmx1a play in its mediation [14–20]. The current study was designed to investigate analogous events in the development of human mDA neurons derived from hES cells. The results presented here suggest that Lmx1a may play an important role in the mDA differentiation process. Thus, the early and transient expression of the Lmx1a gene during the hNP stage is necessary for the expression of other mDA phenotypic traits. By further tracking the expression of more enduring early markers, such as Aldh1a1, in hNPs, it has been possible to follow the lineage steps as these cells move from early mDA fate-specified cells to restricted precursors and finally terminally differentiated neurons. That these neurons do not also express ectopic neurotransmitters such as GABA provides further proof of their bona fide mDA phenotype. Finally, we show that only Lmx1a-expressing hNPs have the potential to differentiate into mDA neurons after transplantation into the 6-OHDA rat striatum, a fact of critical importance for the development of suitable replacement tissue for the treatment of PD.
Lineage Steps in the mDA Fate Restriction Process of hES Cells
By following the expression of Lmx1a and other mDA-specific proteins along with cell stage-specific markers, we were able to delineate discrete steps in the mDA fate restriction process in hES cell cultures (Fig. 7). Although Lmx1a was observed in cells at all stages (II–V) of hES differentiation, it was always coupled to Sox2 expression, suggesting that the inception of mDA specification and multipotency may be inextricably linked in hNPs, as has been suggested in mouse systems [33, 34]. With the continuing development of hES cells along mDA differentiation lines in our cultures, Lmx1a+ hNPs further acquired expression of another early mDA marker, Aldh1a1. These cells, in turn, gave rise to more fate restricted precursors in which Lmx1a and Sox2 expression was extinguished but in which Aldh1a1 and nestin expression persisted. Finally, in stage V, approximately 20% of Aldh1a1+ cells differentiated further to become β-tubulin III+ neurons in which the DA enzyme TH and presumably other mDA phenotypic traits (i.e., Pitx3, Nurr1, AADC, DAT, Girk2) were also expressed. Importantly, these in vivo-generated TH+ neurons did not coexpress traits of other neurotransmitter systems (i.e., GABA), consonant with their mDA phenotype. In contrast, progenitors from nonmidbrain sources (i.e., HNPCs or mouse SVZ) did not express Lmx1a and did not differentiate into bona fide mDA neurons even when placed in an in vivo or in vitro environment that fostered the acquisition of other DA traits.
Importantly, the lineage steps in the mDA differentiation process observed in hES cultures were mirrored in hES cell grafts. Thus, after the transplantation of stage IV hNPs into the parkinsonian rat brain, cells at all stages of mDA fate restriction, including Lmx1a+Aldh1a1− and Lmx1a+Aldh1a1+ hNPs, as well as more advanced Lmx1a−Aldh1a1+ precursors and Aldh1a1+TH+ neurons, were found in 3–5-week-old grafts. Additional time course studies should make clear whether, with increasing time (>5 weeks), these variously staged cells all develop into mature mDA neurons or whether a pool of self-renewing hNPs/precursors is retained indefinitely in the graft. Although these in vitro and in vivo data strongly support the hypothesis that only Lmx1a-expressing hNPs have the potential to differentiate into bona fide mDA neurons, final proof awaits the purification and study of Lmx1a+ (vs. Lmx1a−) cells in culture and after transplantation.
Lmx1a Is Necessary for mDA Differentiation of hNPs
Although the full elucidation of mDA-specifying events awaits further investigation, our results suggest that Lmx1a plays a critical role in the development of a mDA phenotype in hES cells. Thus, knockdown of the Lmx1a gene expression in hES-derived hNPs by siRNA resulted in a corresponding reduction in the expression of all other downstream DA-related genes (including Aldh1a1, Nurr1, Pitx3, TH, and DAT), indicating that Lmx1a was essential for the development of a mature mDA phenotype in hNPs. Whether Lmx1a acted to initiate, coordinate, and/or maintain the expression of other mDA traits in these cells and the pathways via which it accomplished these effects remain to be seen.
Although expression of Lmx1a was required for mDA differentiation, it is not yet clear whether the Lmx1a gene alone was sufficient to induce the DA phenotype de novo in hES-derived hNPs. In the absence of efficient methods for the transfection of Lmx1a expression vectors into hES cell colonies, we instead used clonal and primary HNPC monolayers to study the effects of Lmx1a overexpression. Unlike mouse ES and mouse midbrain cells [14, 15], we did not find de novo expression of DA traits in transfected HNPCs as detected by either real-time PCR or immunocytochemistry. This discrepancy may be explained in one of several ways. First, the difference seen in mouse and human systems may be cell stage-specific. As shown here and by others , cells are responsive to mDA specifying cues only during a brief window in their development, while they are still early progenitor cells. Since the hNPs used in our overexpression experiments were generated from postnatal tissue (BE(2) line) or from the midgestational fetal brain (HNPC), cells may have been too developmentally advanced to respond appropriately to mDA specification signals. Second, the difference may be species-specific. As reported previously, when compared side by side, human and mouse ES cells are fundamentally different in many aspects, including morphology, pattern of antigen markers, and expression profiles of signal pathway molecules . Indeed, Vinogradov and Anatskaya , on the basis of the findings of others, recently showed that in human as compared with mouse, the proportion of tissue-specific genes to housekeeping genes is greater in every tissue, reflecting a higher degree of regulation in the differentiation process. For example, the human Lmx1a gene (GeneID: 4009, http://www.ncbi.nlm.nih.gov/sites/entrez) has three isoforms, whereas mouse Lmx1a gene has only one (GeneID: 110648). Since we did not detect an increase in the expression in Msx1 after Lmx1a overexpression, it is possible that Lmx1a requires the cooperativity of other transcription factors to initiate expression of its downstream effector Msx1 in human cells, as has been recently suggested for the differentiation of mouse ventral midbrain progenitors . Finally, it is possible that these differences are unrelated to stage or species but merely reflect variations in the differentiation protocols leading to the activation of distinct mDA differentiation pathways. Going forward, it will be important to overexpress Lmx1a directly in hES cells, a goal that can be achieved only with the development of new technologies for high efficiency gene transfer in cells growing in colony formation or by introducing genes in undifferentiated hES cells without phenotypic compromise .
Other Differences in mDA Differentiation of Human and Mouse Progenitor Cells
The present study has uncovered several other potentially important differences in the mDA differentiation of human and mouse progenitors, in terms of both the type and the timing of specifying factor expression. In our hES cultures, we observed Lmx1a expression in Sox2+ hNPs. As cells moved to the next stages of mDA fate restriction, there was a marked downregulation in Lmx1a expression in Sox2−Aldh1a1+ and Aldh1a1+TH+ neurons in culture or in transplants. Surprisingly few of these cells coexpressed the forkhead transcription factor Foxa2 at any of these developmental stages, suggesting a limited role for Foxa2 in the hNP mDA specification process. In contrast, in the mouse midbrain, Lmx1a and Foxa2 were coexpressed by all midline E10.5 Sox2+ progenitors (i.e., those destined to become mDA neurons), suggesting a potentially significant role for Foxa2 in the differentiation of mouse mDA neurons.
Also unlike hNPs, the expression of these specifying factors persisted beyond the progenitor stage to be expressed in newly differentiated TH+ neurons in the E12.5–E13.5 mouse midbrain before vanishing at E15.5. That their disappearance was not due to fixation/staining issues is suggested by the continued presence of specific staining in these and other E15.5 brain regions. Why Lmx1a and Foxa2 were expressed by both progenitors and early postmitotic mDA neurons before disappearing at E15.5 in mouse, whereas only Lmx1a was expressed, and only at the early progenitor stage, in hES cell cultures remains unclear but may relate to species differences in the differentiation process.
Importantly, by E18.5, there was a re-emergence of Lmx1a expression in ventral mDA neurons that persisted even in the adult. This same pattern of remitting and recurring expression was also observed for Foxa2 and suggests different roles for these factors early (i.e., fate specification) and later (i.e., phenotypic maintenance) in the mouse life cycle, consistent with the findings of others . Whether fully mature human mDA neurons also require the re-expression of factors such as Lmx1a for the maintenance of cell survival and the mDA phenotype has not yet been determined in long-term hES cultures and grafts.
Role of Patterning Molecules in mDA Differentiation of hNPs
The studies here did not investigate the role of patterning factors such as Wnt1, fibroblast growth factor 8, and Shh, which have been shown to be important in initiating the mDA specification process in mouse cells [14–17]. For example, in some cases, precise concentrations of Shh in the culture medium were required for the mDA specification of mouse ES cells . However, using other differentiation protocols , mouse ES cells, like our hES cells, developed into mature mDA neurons in the absence of exogenous Shh. Likewise, large numbers of mDA neurons were generated from cultures of primary E8.5 mouse ventral midbrain progenitors grown without added Shh, and their proportion was not increased by supplementation with Shh (0.5–1.0 pg/ml) . Possibly, in these latter instances, sufficient quantities of Shh were manufactured intrinsically in the culture (i.e., Shh mRNA can be detected in our stage II–V hES cultures; unpublished data), precluding the need for exogenous supplementation. Consistent with this possibility, Kittappa et al. found that inhibition of endogenous Shh by cyclopamine decreases the number of mDA neurons in mouse embryonic stem cell cultures . Additional studies will be needed to determine the precise role of Shh and other patterning molecules in the regulation of Lmx1a during the mDA specification process.
Implications for Cell Transplantation in PD
The discovery that Lmx1a may be a critical gene underlying the mDA differentiation process in hES-derived cells is important from several standpoints. First, by exploring events upstream and downstream of Lmx1a expression, it may be possible to unravel the signaling pathways through which Lmx1a regulates the expression of other DA phenotypic traits in midbrain neurons. Second, the implication that only Lmx1a-expressing hNPs developed into mDA neurons after transplantation in vivo has important implications as researchers seek to generate homogeneous grafts of mDA neurons. One approach currently under development is the synchronization of cultures to cultivate hNPs of the same developmental stage (i.e., all neuron restricted precursors) for transplantation. The results presented here suggest that even these seemingly homogeneous hNPs will ultimately give rise to phenotypically heterogeneous progeny in the graft unless they have been mDA-specified earlier in their development. By using early expressed genes/proteins such Lmx1a and Aldh1a1 as a basis for their selection, however, it may be possible to choose cells of the appropriate stage that are exclusively programmed to become mDA neurons, while simultaneously eliminating other, unsuitable cell types. Developing these approaches will be key in the quest to produce a purified population of human mDA neurons for use in cell replacement therapy in Parkinson's disease.
We conclude that the homeobox gene Lmx1a plays a key role in the mDA differentiation of hES cells. Thus, the transient early expression of Lmx1a is essential for the coordinated expression of all other mDA-specific phenotypic traits as hES cells progress from multipotent hNPs to more restricted precursors and finally mDA neurons. Importantly, only Lmx1a-specified hNPs have the capacity to differentiate into bona fide midbrain mDA neurons after transplantation into the 6-hydroxydopamine treated rat brain. Therefore, Lmx1a and other key early mDA markers like Aldh1a1, may be extremely useful for the selection of appropriately staged and suitably mDA-specified hES cells for cell replacement in Parkinson's Disease.
This work was supported by NIH Grants NS43309, NS32519, and NS488315; the Tilker Foundation; the Michael J. Fox Foundation; and the Hassel Foundation. This project is funded, in part, under Grant SAP4100026302 C.U.R.E. with the Pennsylvania Department of Health. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions. We thank Lena Wei for excellent technical support.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.