The recombination efficiency and specificity of Olig2^Cre/+ mice (Dessaud et al., 2007; Chen et al., 2011) were assessed by crossing them to the reporter mice ROSA26-loxp-STOP-loxp-YFP mice (Srinivas et al., 2001) and resulting embryos were analyzed at E13.5.

Olig2^Cre/+ mice were crossed with Dicer^flox/flox (Harfe et al., 2005) to generate Olig2^Cre/+; Dicer^flox/WT strain. Olig2^Cre/+; Dicer^flox/WT mice were then mated with either Dicer^flox/flox or Hb9::GFP;Dicer^flox/flox for experimental analysis. Embryos were analyzed at E11.5–E16.5. All mice were maintained in C57BL/6 background. All experiments were performed according to the approved Columbia University IACUC protocols.

Antibodies used in this study include: rabbit anti-mouse Olig2 (Millipore, AB9610), guinea pig anti-mouse Olig2, mouse anti-Lhx3 and Hb9, guinea pig anti-Isl1, FoxP1, and SCIP (gift from Tom Jessell), rabbit anti-Sox9 and GFAP, rabbit anti-pSmad (gift from Ed Laufer), mouse anti-Hoxc8 (gift from Tom Jessell), goat anti-Hoxc6 (Santa Cruz, sc-66925), and Alexa488-, FITC-, Cy3-, and Cy5-conjugated secondary antibodies were obtained from either Invitrogen or Jackson Immunoresearch. Images were collected on a Zeiss LSM510 confocal microscope.

In situ hybridizations were performed in 15 μm cryosections from E13.5 brachial spinal cord. Sections were fixed in 4% paraformaldehyde and acetylated in acetic anhydride/triethanolamine, followed by washes in PBS. Sections were then pre-hybridized in hybridization solution (50% formamide, 5× SSC, 0.5 mg/mL yeast tRNA, 1× Denhardt’s solution) at room temperature, then hybridized with 3′-DIG labeled LNA probes (3 pmol; LNA miRCURY probe; Exiqon) at 25°C below the predicted T_m value. After post-hybridization washes in 0.2× SSC at 55°C, the in situ hybridization signals were detected using the NBT/BCIP (Roche) according to the manufacturer’s instructions. Slides were mounted in Aqua-Poly/Mount (Polysciences, Inc.) and analyzed with a Zeiss LSM 510 confocal microscope. Detail protocols can be checked from Chen et al. (2005, 2007, 2011).

Motor neuron differentiation is driven by a transcriptional program that diversifies generic motor neurons into hundreds of distinct subtypes along the rostro-caudal axis of the spinal cord (Dasen, 2009). To investigate if miRNAs are involved in this process, we took advantage of Olig2^Cre/+ mouse line that expresses Cre recombinase in progenitors of all spinal motor neurons and oligodendrocytes (Dessaud et al., 2007). To confirm efficiency of recombination we crossed the Olig2^Cre/+ mouse to ROSA26-loxp-STOP-loxp-YFP reporter mouse in which all cells derived from Olig2 positive progenitors should be labeled by YFP expression. Consistent with previous studies (Masahira et al., 2006), we observed that all motor neurons and oligodendrocyte precursors (OPCs) on day 13.5 of development (E13.5) are marked by YFP expression demonstrating that they are derived from Olig2 expressing spinal progenitors (pMN; Figures 1A–C).

Next we examined the efficiency of Dicer deletion in pMN lineage in Olig2^Cre/+; Dicer^flox/flox mice (Figures 1A–C). In situ hybridization using a locked nucleic acid (LNA) probe revealed depletion of a neuronal enriched miRNA mir-100 in the ventral horn of the spinal cord as well as depletion of mir-9 enriched in postmitotic motor neuron, indicating loss of miRNAs in postmitotic motor neurons (Figures 1D–H).

Because deletion of Dicer function in early neural progenitors results in dorso-ventral patterning defect (Chen et al., 2011), we examined whether decreased levels of miRNAs in motor neuron progenitors affect spinal cord patterning or motor neuron production. Consistent with previous report (Zheng et al., 2010), we did not detect any changes in the position of dorso-ventral boundaries marked by Olig2, Nkx2.2, and Irx3 expression in E10.5 spinal cord and observed comparable number of Hb9^on and Isl1^on motor neurons generated in conditional KO and WT E11.5 embryos (data not shown and Figures 2A–E).

Previously it has been shown that oligodendrocyte differentiation is impaired in Olig1-Cre^+/−; Dicer^flox/flox mutants (Dugas et al., 2010). Staining E16.5 spinal cords of Olig2^Cre/+; Dicer^flox/flox mutant mice for oligodendroglial and astrocytic markers revealed nearly complete loss of oligodendroglial progenitor cells expressing Olig2 and Sox9 whereas astroglial cells (GFAP^on, Olig2^off, Sox9^on) remained unaffected (N = 3, Figures 2F–H). Taken together, our data are consistent with previous results, and indicate efficient deletion of Dicer function and loss of miRNAs in motor neuron and oligodendroglial lineages in the Olig2^Cre/+; Dicer^flox/flox mutants.

Postmitotic motor neurons acquire diverse subtype identities and settle into distinct motor columns in the developing spinal cord (Figure 3A). We therefore inspected pattern of expression of column specific markers in brachial and thoracic spinal cord in E13.5 mutant embryos. Despite the normal initial production of generic motor neurons at E11.5 (Figure 2E), all examined Olig2^Cre/+; Dicer^flox/flox embryos exhibited a ∼50% decrease in the number of FoxP1 expressing LMC motor neurons innervating limb muscles at brachial levels of the spinal cord (Figure 3O, p < 0.01, n = 3). In contrast, the Lhx3 expressing MMC motor neurons innervating axial muscles were not affected (N = 3, Figures 3G–O). This result indicates that LMC and MMC motor neurons display differential requirements for Dicer function and miRNAs.

To determine whether motor neuron subtype specification was impaired at other spinal levels we examined PGC motor neurons innervating sympathetic ganglia and HMC neurons innervating intercostal and abdominal wall musculature in the thoracic spinal cord (Figures 3B–F). We observed ∼30% reduction of PGC (pSmad^on and FoxP1^low) motor neurons in the Olig2^Cre/+; Dicer^flox/flox embryos, whereas HMC (Lhx3^off and Hb9^high) and MMC (Lhx3^on) motor neuron numbers remained unchanged compared to their heterozygous or wild type littermates (N = 3, Figures 3B–F). These results demonstrate that specification or maintenance of LMC and PGC motor neurons depends on Dicer function.

Given the decrease in the number of FoxP1^on LMC neurons in the Olig2^Cre/+; Dicer^flox/flox embryos, we considered whether this might be due to the failed induction of FoxP1 expression or due to the selective death of LMC neurons. To mark all motor neurons in the mutant and control embryos we introduced Hb9::GFP transgene (Wichterle et al., 2002) into Olig2^Cre/+; Dicer^flox/flox mice. To examine whether LMC neurons might be susceptible to increased rate of apoptosis we analyzed apoptotic cells by TUNEL labeling in the mutant spinal cord on embryonic days E11.5, E13.5, and E14.5 (Figures 4A–F). We did not notice increase in TUNEL positive cells between mutant and control spinal cords on E11.5 and E13.5. By E14.5 we detected an increase in TUNEL positive cells within the LMC in mutant embryos, while no difference was observed in MMC (Figures 4E,F). Although the trend was consistent across many examined sections, the observed increase in the TUNEL labeling in LMC did not reach statistical significance, likely due to the overall small number of labeled cells. Importantly, loss of FoxP1 expressing cells was not compensated by an emergence of supernumerary FoxP1 negative motor neurons in the LMC, suggesting that LMC cells do not simply fail to express FoxP1 (Figures 5C,D). Collectively, these results indicate that the observed decrease in the number of FoxP1 positive neurons results from an increased rate of cell death in LMC.

Since not all LMC neurons are lost in Olig2^Cre/+; Dicer^flox/flox embryos, we asked whether remaining motor neurons acquired proper subtype identities. We examined expression of rostro-caudal determinants and motor pool subtype specific markers in the mutant spinal cord (Figure 5A). Despite the severe reduction of LMC motor neurons, the expression of appropriate set of brachial Hox genes was detected in the remaining LMC neurons. Hoxc6^on brachial LMC neurons were correctly subdivided into the rostral Hoxa5^on and caudal Hoxc8^on LMC domain (Figures 5C–H and data not shown; Dasen et al., 2005; Jung et al., 2010), indicating that the rostro-caudal intrasegmental division of the brachial LMC is not affected in the Olig2^Cre/+; Dicer^flox/flox embryos.

Combinatorial Hox transcription program results in the diversification of LMC neurons into multiple motor pool identities, several of which are marked by differential expression of TFs Nkx6.1, Pea3, and Scip (Dasen and Jessell, 2009). In contrast to the normal pattern of Hox gene expression, the expression of motor pool markers was significantly eroded in mutant LMC motor neurons (Figures 5B,I–N). To determine whether the decrease in pool specific marker expression is solely due to the loss of LMC neurons we quantified the neurons expressing individual pool markers and normalized them to the total number of FoxP1 positive LMC neurons (Figure 5B). This analysis revealed that while the same fraction of LMC neurons expressed caudal brachial marker Hoxc8, the fractions of LMC neurons expressing Pea3, Scip, and Nkx6.1 were significantly decreased. Thus, terminal differentiation of motor pools in LMC is sensitive to the loss of Dicer function, indicating that the subtype specific program requires intact miRNAs for proper execution.

Previous studies have employed similar conditional Dicer knockout approach to address miRNA functions in gliogenesis (Dugas et al., 2010; Zhao et al., 2010; Zheng et al., 2010; Table 1). Dugas et al. and Zhao et al. reported that miRNAs are essential for oligodendrocyte maturation, but not for OPC generation and astrocyte development in the brain, while Zheng et al. suggested that OPC generation and astrocyte development require Dicer functions in the developing spinal cord. Our study demonstrates that Sox9/Olig2 double positive glial progenitors are severely reduced in the mutant spinal cord, but numbers of Sox9 positive Olig2 negative glial progenitors are not changed. The discrepancy could be due to the differential developmental stage and CNS regions examined. Furthermore, different oligodendrocyte Cre divers used in the studies might give rise to differential recombination efficiencies (summarized in Table 1). Although Olig2^{tm2(TVA,cre)Rth}; Dicer^flox/flox mice display a notable tremor phenotype from postnatal day 9–10, more than 50% of oligodendrocytes are generated and recovered at older stage (Dugas et al., 2010). This phenotype contrasts with the postnatal lethality of Olig2^Cre/+; Dicer^flox/flox mice examined in this study, indicative of more efficient and complete deletion of Dicer function in motor neuron and oligodendroglial lineages (Dessaud et al., 2007, 2010; Chen et al., 2011).

Consistent with the results of Zheng et al., we did not observe an overt ventral pattering defect or a decrease in the generic postmitotic motor neuron production at E11.5. This is not surprising, given that it takes 3 ∼ 4 days to deplete existing miRNA reservoir in the cells following the conditional inactivation of the Dicer function (Chen et al., 2011). Therefore Olig2^Cre/+ recombination should not significantly affect miRNA function in motor neuron progenitors, resulting in the normal patterning and generic motor neuron production in the conditional Dicer mutant embryos.

Previous study has shown that inactivation of Dicer in the astroglial lineage results in non-cell autonomous neuronal degeneration (Tao et al., 2011), raising the question whether some of the observed phenotypes might be due to non-cell autonomous effects. We think this is unlikely because Olig2^Cre/+ mediated deletion of Dicer function did not markedly affect astroglial progenitors (Sox9^on; Olig2^off), and the observed motor neuron phenotypes preceded the wave of oligodendrocyte and astrocyte production. However, future studies targeting Dicer function in spinal glial compartment are needed to formally exclude the possibility of non-cell autonomous contribution to the observed motor neuron defects.

Dicer participates in a broad spectrum of biological processes, including miRNA biogenesis, heterochromatin maintenance, and fragmentation of chromosomal DNA during apoptosis (Nakagawa et al., 2010; Cernilogar et al., 2011). Although pro-apoptotic role of Dicer enzyme is inconsistent with the observed increase in motor neuron loss in conditional Dicer embryos, follow-up studies targeting specific microRNAs in the developing spinal cord are needed to understand the motor neuron defects on a mechanistic level.

Our study revealed that Dicer function requirements are motor neuron subtype specific. Brachial LMC and thoracic PGC display strong reliance on Dicer function, whereas MMC present in all spinal cord segments are relatively spared in the conditional Dicer mutants (Figure 6). The subtype specificity of motor neuron loss is unlikely due to a differential efficiency of Dicer deletion in MMC vs. LMC neurons, as both motor columns exhibit comparable degree of Olig2^Cre/+ driven recombination. Instead we suggest that the differences in motor neuron survival might reflect motor neuron subtype specific differences in the expression of miRNAs or their targets, or differential susceptibility of individual motor neuron subtypes to degeneration.

It is of interest that both affected populations of motor neurons in our study express FoxP1 TF (high level in LMC and low level in PGC). Recently it has been shown that mir-9 is critical for fine-tuning of FoxP1 expression levels in the chick spinal cord (Otaegi et al., 2011a,b). Knock-down of mir-9 alone is sufficient to increase FoxP1 expression levels and promote specification of PGC or LMC motor neurons. In contrast, our study revealed selective loss of LMC and PGC neurons, pointing to a more complex role of Dicer function in motor neuron specification and survival.

Despite the reduction of LMC motor neurons in the conditional Dicer mutants, the rostro-caudal (Hoxa5^on and Hoxc8^on) division of LMC remains patterned normally. However, the expression of motor neuron pool markers (Pea3, Scip, and Nkx6.1) was severely disrupted in the mutants. This suggests that besides their role in the regulation of LMC cell death, miRNA function is also important for proper motor pool specification and maturation. Our data do not establish whether the failure to properly specify motor pool identity is the primary defect, leading to the LMC neuron cell death. We think this is unlikely, as a complete failure to specify LMC pool identity in FoxP1 mutant animals results only in a ∼15–30% decrease in motor neuron numbers (Dasen et al., 2008; Surmeli et al., 2011), while conditional Dicer mutant animals exhibit a ∼50% loss of LMC neurons. Finally, once subtype specific markers for MMC and HMC neurons are identified, it will be of interest to examine whether diversification of non-LMC neurons is also dependent on Dicer function.

Motor neuron subtype specific degeneration has been described in several motor neuron degenerative diseases (Kanning et al., 2010). Our observation that a subset of LMC neurons degenerate in the Dicer null animals raises the possibility that the molecular mechanism underlying increase sensitivity of LMC and PGC motor neurons in the Dicer mutant mice might be implicated in motor neuron pathologies.