Doublesex and mab-3–related transcription factor 5 promotes midbrain dopaminergic identity in pluripotent stem cells by enforcing a ventral-medial progenitor fate

Restricted Expression of Dmrt5 in the Developing Ventral-Medial Mesencephalon.

We identified Dmrt5 from a differential microarray expression screen of ventral-medial midbrain cell populations. Whole-mount in situ hybridization revealed precise spatial and temporal regulation of Dmrt5 expression in both mouse and chicken embryos. In the ventral midbrain, Dmrt5 is first expressed at the midbrain/forebrain border at embryonic day 9.5 (E9.5) in mice and Hamilton and Hamburger stage 20 (st.20) in chickens. Expression extends posteriorly to encompass the entire medial ventral midbrain (Fig. 1 A and C). Outside the ventral midbrain, Dmrt5 expression was found in the dorsal telencephalon as well as in the optic stalk and early olfactory bulb. Combining Dmrt5 in situ hybridization with antibody staining for regional markers of the ventral-medial and lateral midbrain neuroepithelium, we found that the Dmrt5⁺ region contains the Lmx1a⁺ cell population (Fig. 1A). Like Foxa2 and Lmx1a, Dmrt5⁺ cells occupy almost the entire thickness of the neural epithelium at E10.5 (6, 12). The lateral boundaries of Dmrt5 expression lie immediately ventral to Nkx2.2⁺ neuroepithelium (Fig. 1A), forming two mutually exclusive expression domains. At E12.5, Dmrt5 expression was found primarily in the ventricular zone of the FP and adjacent ventral midbrain (Fig. 1B). With the exception of a small number of cells in the most medial intermediate zone that coexpress Dmrt5 and TH, the vast majority of Dmrt5-stained cells lay dorsal to the differentiated TH⁺ dopamine neurons in the marginal layer (Fig. 1D). Together, these expression studies indentified Dmrt5 as a marker for mesencephalic dopamine neural progenitors.

Dmrt5 Regulates Ventral Mesencephalic Gene Expression in Naïve Neuroepithelium.

Given this highly specific expression profile, we used the chick electroporation system to examine the patterning effects of Dmrt5 in vivo. A mouse Dmrt5 expression plasmid was electroporated into naïve chick midbrain neuroepithelium at st.10. The embryos were allowed to develop for 24 h until st.17 and were analyzed for expression of Foxa2, a medial midbrain marker, and Mab21l2, a gene shown to maintain dorsal identities in early neural development (8). Both genes are highly conserved between the mouse and chick and show a nonoverlapping, spatially restricted expression pattern in the developing midbrain (8, 12).

As can be seen in the nonelectroporated side of the midbrain, Mab21L2 is expressed in the dorsal midbrain and extends ventrally to end in a ventral-lateral position (Fig. 2A). This expression was abolished in the Dmrt5 electroporated neuroepithelium in all embryos (4/4). In contrast, Foxa2 expression was ectopically induced in ventral-lateral cells by Dmrt5 (4/5) (Fig. 2B). No alteration in normal gene expression was seen in nonelectroporated, contralateral side of the embryo (Fig. 2). Our data provide proof that Dmrt5 is capable of regulating ventral-lateral and medial patterning genes in vivo. The ability of Dmrt5 to override endogenous gene induction pathways of these patterning genes points to a direct regulatory role for Dmrt5 in specification of the ventral-medial midbrain cell fate.

Dmrt5 Promotes mDA Neural Progenitor Marker Expression in Neuralized ESCs.

Examination of intrinsic transcriptional regulators and extrinsic niche factors critical for mDA development in ESC differentiation has proved to be a powerful means for illuminating or validating their functions in dopaminergic neuron fate specification (6, 16–20). Notably, engineered expression of mDA transcription factors such as Lmx1a and Pitx3 induced midbrain regional character in derived dopamine neurons (6, 19). Overexpression of Nurr1 increased the number of TH⁺ neurons but without apparent impact on midbrain regional identity (18, 21). To explore the specification potential of Dmrt5 in greater detail, we generated a number of tetracycline-inducible Dmrt5 transgenic mouse ESC lines using ESCs, which harbour the reverse tetracycline-controlled transactivator (rtTA ESCs). (Fig. S1). After verifying the ability of Dmrt5-ESCs to produce significant levels of Dmrt5 protein, we selected two lines to investigate the effects of Dmrt5 expression on dopamine neuron differentiation after a monolayer-based neural differentiation method (22, 23).

Dmrt5 expression in the ventral mesencephalon is restricted to mitotic neuroepithelial cells and precedes the onset of dopaminergic neurogenesis. Thus, it is likely to function at neural progenitor stage. We therefore induced the Dmrt5 transgene at the peak of neural progenitor production and determined the effect of Dmrt5 induction on ventral-medial midbrain neural progenitor marker expression by quantitative PCR (qPCR) from day 6 to day 12 monolayer differentiation (d6–12 MD). We found that the induction of Dmrt5 did not affect pan-neuroepithelial marker gene level but resulted in a four- to fivefold increase in transcript levels of Foxa2, Lmx1a, and Msx1, three transcription factors required for the specification of ventral-medial cell identities and normal dopamine neuronal development (Fig. 3A and Fig. S2A). The up-regulation of these marker genes was detected concurrent with the induced transgenic Dmrt5 at d7 MD, and they remained at a higher level than in control cultures at all time points analyzed.

We then examined the expression at the protein level of two key mDA markers, Foxa2 and Lmx1a, by immunocytochemical staining (Fig. 3B). After 4 d of doxycycline treatment (i.e., at d9 MD) more than 30% of Nestin⁺ neural progenitors expressed Lmx1a (30.3 ± 5%) and 15.6 ± 2.6% became Foxa2⁺ in the Dmrt5 cultures. In contrast, only a small proportion of Nestin⁺ cells in the doxycycline-treated rtTA control cultures became Lmx1a⁺ (4.6 ± 0.8%) or Foxa2⁺ (1.2 ± 0.4%) (Fig. 3 C and D). Similar numbers of Foxa2⁺ neural progenitor cells were obtained in Dmrt5-ESC and rtTA cultures not treated with doxycycline, although we detected an increased number of Lmx1a⁺ cells in noninduced Dmrt5-ESC cultures; the latter could be attributable to the leakage of Dmrt5 transgene expression in the absence of doxycycline (Fig. S1D). Our data suggest that Dmrt5 is capable of promoting the Lmx1a⁺ and Foxa2⁺ neural progenitor phenotype.

Dmrt5 Restricts Shh-Induced Neural Progenitors to Ventral-Medial Identity by Suppressing Ventral-Lateral Neuroepithelial Characteristics.

Dmrt family proteins have been shown to function as transcription repressors as well as activators (14). The mutually exclusive expression between Dmrt5 and Nkx2.2 in the midbrain raised the possibility that Dmrt5 inhibits ESC-derived neural progenitors from adopting alternative cell fates, such as those associated with the ventral-lateral midbrain. To test this hypothesis, we examined the effect of Dmrt5 on the expression of nondopaminergic ventral-lateral marker genes, including Nkx2.2, Isl1, Lhx1, Helt, Brn3a, Meis2, and Mab21l2 (7–9, 24). By qPCR, we found that Dmrt5 overexpression resulted in a striking down-regulation of all these marker genes (Fig. 4A and Fig. S2 B and C).

A number of ventral-laterally expressed transcription factors, such as Nkx2.2 and Isl1, are induced by Shh (13). Therefore, Dmrt5-mediated suppression might involve Shh signaling. We repeated the above experiments in cultures treated with Shh or cyclopamine, an inhibitor of Shh signaling. Shh added to the cultures for 2 d (d3–5 MD) or 6 d (d3–9 MD) induced an increase in the transcript levels of Nkx2.2 and Isl1 in ESC-derived neural cultures, as Shh signaling does in the developing neuroepithelium (13). The presence of cyclopamine reduced their expression (Fig. 4A). However, conditional Dmrt5 transgene expression in the presence of Shh or cyclopamine still was able to cause a prominent repression of the two genes (Fig. 4A).

A regulatory role for Shh on Lhx1, Helt, Brn3a, and Meis2 expression has not been previously reported. Unlike the induction of Nkx2.2 and Isl1, we found that Shh treatment results in a marked inhibition in the expression levels of Lhx1, Helt, Brn3a, and Meis2 (Fig. 4A), which are further reduced by the overexpression of Dmrt5. Cyclopamine treatment had no significant effect on the levels of Lhx1, Helt, Brnl3a, and Meis2, but Dmrt5 continued to reduce the expression of these genes under this condition. Foxa2 has been suggested to suppress Helt and Lhx1 expression in the midbrain (12); however, the barely detectible levels of Foxa2 in the presence of cyclopamine make a Foxa2 role in the observed decrease of Helt and Lhx1 unlikely (Fig. S3). Together, these results indicate that, independently, Dmrt5 negatively regulates the ventral-lateral midbrain gene-expression profile irrespective of its responsiveness to Shh.

To ask whether the observed transcript changes correlates with protein expression, we performed immunocytochemical staining for Nkx2.2, Isl1, and Lhx1, for which reliable commercial antibodies are available. In standard MD cultures, Nkx2.2⁺ and Isl1⁺ cells were present in small numbers, whereas Lhx1⁺ cells were readily detectable in most microscopic fields. Dmrt5 overexpression greatly reduced the production of Lhx1⁺ cells, whereas Nkx2.2⁺ and Isl1⁺ cells became almost nondetectable. To obtain appreciable numbers of Nkx2.2⁺, Isl1⁺, and Lhx1⁺ cells in the Dmrt5-overexpressing condition, and thus a reliable quantitative analysis, we used a 6-d Shh treatment from d3 to d9 MD. Again, we observed a clear reduction in the number of Nkx2.2⁺, Isl1⁺, and Lhx1⁺ cells in Dmrt5-overexpressing cultures compared with the no-doxycycline controls (Fig. 4B).

We found that a significant proportion of neural progenitors in Shh-treated cultures are Foxa2⁺Nkx2.2⁺ and Foxa2⁺Lhx1⁺ (Fig. 4C). However, expression of Dmrt5 resulted in a 60% and 75% reduction, respectively, of Foxa2⁺ cells coexpressing Nkx2.2 (from 52.2 ± 7.5% to 21.2 ± 5.2%) and Lhx1(from 17.2 ± 6.4% to 4.2 ± 2.7%) (Fig. 4 C and D), exposing a role for Dmrt5 in fine-tuning the ventral-medial identity. Consistent with the suppression of the nondopaminergic progenitor profile, we observed a reduction in the number of GABAergic neurons as demonstrated by GABA and GAD67 antibody staining, which was preceded by decreased transcript levels of the GABAergic regional and subtype specification genes, Helt and Gad1 (Fig. 4 E–G).

The observed decrease and increase of ventral-lateral and ventral-medial phenotype is unlikely caused by toxicity of doxycycline because it had no effect on the number of Pax6⁺ and Nestin⁺ neural progenitors or RNA expression levels for Foxa2, Lmx1a, or Nkx2.2 in the rtTA control cultures (Fig. S4). Furthermore, we observed no increase in progenitor apoptosis or selective expansion of Lmx1a⁺ subpopulation as determined by activated caspase3 staining and double immunostaining for Ki67 and Lmx1a (Fig. S5). Collectively, our data suggest that Dmrt5 promotes ventral-medial mesencephalic fate choice by suppressing transcription programs associated with alternative cell fates.

Dmrt5 Deficiency Compromises ESC Differentiation Toward Ventral-Medial Cell Fate.

To investigate a Dmrt5 requirement in mDA progenitor fate choice, we generated shRNA-based conditional Dmrt5 knockdown ESC lines by using a pSico-based vector (25) (Fig. S6A). The parental ESCs (R26) constitutively express CreERT2, a Cre recombinase (Cre) fused to a mutant estrogen ligand-binding domain (ERT2), under the control of the constitutive and ubiquitous Rosa26 promoter. Thus, in the presence of tamoxifen, the puromycin-resistance cassette that functions as a transcription stop will be excised, leading to the (nonreversible) expression of Dmrt5 shRNA. Two independent ESC lines, C859 and C1021, which carry distinct shRNA target sequences, consistently showed 40–60% reduction of the Dmrt5 transcript in ESC-derived neural cultures when Cre activity was induced by tamoxifen in undifferentiated ESCs or during differentiation consistently (Fig. 5A and Fig. S6).

The noninduced and induced C859 and C1021 (referred to hereafter as Dmrt5KD) showed comparable levels of Sox1 and Nestin transcripts as well as the numbers of Nestin⁺ cells at d7 and d9 MD (Fig. 5 A–C and Fig. S6), suggesting that the lack of Dmrt5 does not affect the neuroectoderm lineage commitment of ESCs. However, we found that Dmrt5KD resulted in a more than twofold reduction in the levels of Foxa2 and Msx1 and a fivefold decrease of Lmx1a at d7 and d9 MD (Fig. 5A). This effect was specific to Dmrt5 knockdown because tamoxifen treatment on the parental R26 ESCs did not affect the levels of Foxa2, Lmx1a, or other neural markers tested (Fig. S6). Conversely, we observed an increased level of Helt, Meis2, Lhx1, and Isl1 in Dmrt5KD neural cultures compared with the noninduced controls (Fig. 5A). The above finding was further supported by a significant reduction in the number of Foxa2⁺ (60%) and Lmx1a⁺ (80%) neural progenitors in Dmrt5 KD cultures treated with Shh (Fig. 5 B and C). Taken as a whole, our data reveal a requirement for Dmrt5 in ventral midbrain neural progenitor fate specification.

Genome-Wide Analysis Supports a Role for Dmrt5 in Regulating Ventral Mesencephalic Neuroepithelial Cell Fates.

Further insight into the functionality of Dmrt5 and transcription networks potentially regulated by Dmrt5 is derived from Affymetrix expression profiling using neural progenitors derived from the Dmrt5-overexpressing and parental control ESCs with or without doxycycline treatment (Fig. 6A). Results showed 987 entities with qualified fold change > 2 and P < 0.05 by one-way ANOVA. Of these, 485 were down-regulated and 502 were up-regulated by Dmrt5 (Fig. 6B and Dataset S1). These changes are specific to Dmrt5 because the effect of doxycycline on parental rtTA cell gene expression was negligible.

We were particularly interested in investigating whether Dmrt5-responsive genes correlate with Dmrt5 region-specific gene expression in the developing midbrain. To this end, we performed another microarray analysis to identify genes differentially expressed in the mesencephalic FP and the adjacent, nonoverlapping ventral-lateral midbrain region (VL) of 10.5-d postcoitum (dpc) mouse embryos (Fig. 6C). Using a cutoff of fold change > 2 and P < 0.05, we identified 653 FP-enriched entities and 451 VL-specific entities (Dataset S1). The FP- and VL-specific entity lists contain all genes known to be differentially expressed in the respective regions of midgestation mouse midbrain (Fig. 6D), demonstrating the high quality of this dataset. Cross-comparison between the Dmrt5 up- and down-regulated lists with the FP- and VL-specific entities revealed remarkably parallel enrichments of the Dmrt5-regulated genes. The Dmrt5 down-regulated list matched significantly more entities in the VL list than in the FP list (97 vs. 16 entities or 21.5% vs. 2.2% of all FP or VL entities, respectively) (Fig. 6 E and F). Interestingly, the Dmrt5 down-regulated/VL-specific list is rich in transcription factors important for lateral and dorsal midbrain patterning, such as Pax3, -6, and -7 and Lhx9 (Fig. 6G and Dataset S2), as well as those already shown with the candidate gene approach to be suppressed by Dmrt5 (Fig. 4). Conversely, Dmrt5 up-regulated entities matched more of the FP list than those in the VL list (65 vs. 10 entities or 10% vs. 2.5%) (Fig. 6 E and F). Hypergeometric distribution analysis confirmed the strongest significance for the Dmrt5 down-regulated/VL and Dmrt5 up-regulated/FP entities (P value was almost 0) (Fig. 6F). Thus, our global gene-expression analysis provides strong independent support that Dmrt5 promotes ventral-medial midbrain neuroepithelial fate.

Dmrt5 Enhances the Generation of DA Neurons Exhibiting Key Midbrain Complements.

We next determined the effect of Dmrt5 on postmitotic dopamine neuron production with particular attention on their midbrain regional marker expression. To temporally mimic neural progenitor expression of endogenous Dmrt5, we induced Dmrt5 transgene expression for 4 d during the peak of neural progenitor production between d5 and d9 MD. We omitted exogenous dopamine-inducing factors to reveal Dmrt5 activity with minimum input from other signals. As previously reported (6, 23), few TH⁺ neurons in the parental or noninduced control cultures expressed Pitx3, Lmx1a, or Foxa2 (Fig. 7 A and B). The majority of TH⁺ cells in these cultures were dispersed and had an immature neuronal morphology as demonstrated by fewer and shorter processes. In contrast, a significant proportion of TH⁺ neurons in doxycycline-induced Dmrt5 cultures coexpressed Pitx3, Lmx1a, and Foxa2 (Fig. 7A): 40% of TH⁺ neurons were Pitx3⁺, 37% were Lmx1a⁺, and 59% were Foxa2⁺ (Fig. 7B). Additionally, most of the TH⁺ cells were found in large clusters and had a mature neuronal morphology with long and elaborate processes and small cell bodies (Fig. 7A). Furthermore, Dmrt5 induction led to a 63 ± 15% increase in the total number of TH⁺ neurons compared with the parental control cultures, leading, significantly, to an increase in the total number of TH⁺ cells coexpressing Pitx3, Lmx1a, and Foxa2 (Fig. 7C). Interestingly, although the vast majority of TH⁺ neurons coexpressed an A9 (substantia nigra) DA marker, Girk2, only a small proportion of TH⁺ cells expressed calbindin, which is preferentially expressed in the A10 (ventral tegmental area) subpopulation (Fig. 7D). Dmrt5 induction had little impact on coexpression of TH with Nurr1 (Fig. 7 A and B). This observation is not surprising because Nurr1 controls neural transmitter phenotype by regulating Th gene expression, and this activity does not appear to be coupled to regional or even neuronal phenotype (18). Therefore, our data demonstrate that Dmrt5 is capable of inducing mDA neuronal character in ESCs. This activity distinguishes Dmrt5 from Lmx1a, which requires cooperation with Shh to promote mDA neuronal phenotype.

ESC Lines.

The generation of ESC lines that conditionally overexpress Dmrt5 or Dmrt5 shRNA are described in SI Materials and Methods. Two independent Dmrt5-overexpressing and shRNA knockdown lines were used in this study and gave similar results. PCR primer list is provided in Table S1.

Neural Differentiation, Immunocytochemistry, and In Situ Hybridization.

Routine ESC culture and monolayer neural differentiation was performed as described in refs. 22 and 23. The schemes for conditional Dmrt5 overexpression and knockdown, detailed antibody staining, and in situ hybridization are described in SI Materials and Methods.

Microarray Analysis.

Detailed sample preparation and analysis of the microarray data are provided in SI Materials and Methods. Three biological replicates of the paired samples were generated. Labeled cRNAs were generated from 10 ng of total RNA by the T7-based one-cycle protocol and hybridized to Affymetrix MOE430.2 GeneChips per the manufacturer's instructions. Microarray data in CEL format were analyzed with the GeneSpring GX11 program.