For luciferase experiments, Gga-miR-9-1, Gga-miR-9-1-Mut, and FoxP1 3′-UTR Luciferase constructs were used as described (Otaegi et al., 2011). Briefly, a 289 nt genomic sequence including the gga-miR-9-1 precursor was cloned into the pCAGIG expression vector (mature miR-9 sequence: 5′-TCTTTGGTTATCTAGCTGTATGA). The mutated version is identical except for three altered nucleotides in the seed sequences of mature miR-9 and three matching nucleotides in the opposite side of the precursor stem-loop (mutant miR-9 sequence: 5′-TCCTTAGGTATCTAGCTGTATGA). The FoxP1 3′-UTR construct contains ~1.8 kb of the chick FoxP1 3′-UTR including two central predicted miR-9 binding sites.

For sponge oligo design, binding sites were designed as the reverse complement of the full length mature miRNA sequence with one deletion and two mismatches to create a bulge as described in Gentner et al. (2008). Spacer regions of different size were designed using the siRNA Wizard scrambled design tool (http://www.sirnawizard.com/scrambled.php). Forward and reverse oligos were ordered with one or two binding sites and spacers of different lengths (Table 1). Oligos were annealed and ligated together in pBSIIKS+ (Stratagene) at SpeI/XbaI sites. Constructs were sub-cloned into XbaI to finally contain 6, 12, or 24 binding sites. These sponges were then cloned into pCDNA3.1+ (Invitrogen) or attached to iCre as a 3′-UTR and then cloned into pCDNA3.1+ for luciferase experiments. For in vivo electroporation experiments, all constructs were cloned into the pCAGIG expression vector.

Neuro2a cells were transfected using Lipofectamine 2000 (Invitrogen) using the manufacturer’s protocol. Plasmids were quantified by UV spectrophotometry and used for transfection in a 3:1 ratio (miRNA:target luciferase constructs; Figure 2C); 9:3:1 ratio (sponge:miRNA:target luciferase constructs, Figures 2D and 3B,D); 1.5:3:1 ratio (sponge:miRNA:target luciferase constructs in Figure 3F). pGL4.13 firefly luciferase (Promega) was used for 3′-UTRs of targets. pGL4.73 renilla luciferase (Promega) was used as a transfection control. Luciferase was measured using the Dual-Luciferase Reporter Assay kit (Promega) using the manufacturer’s protocol and read on a Victor3 1420 multilabel counter (Perkin Elmer). Results are expressed as Firefly luciferase normalized to Renilla.

Electroporation was performed as previously described (Otaegi et al., 2011). Briefly, Hamburger and Hamilton (HH) stage 10–12 chick embryos were electroporated unilaterally (five 50-ms pulses at 25 V) with 1–3 μg/μl DNA or as indicated, using an ECM830 electro-squareporator (BTX). Embryos were analyzed 24 h later at HH20, or 48 h later at HH24.

Spinal cord tissues from the chick embryos were fixed in 4% paraformaldehyde (PFA) in PBS overnight, incubated in 30% sucrose in PBS, embedded in OCT, and stored at −80°C until use. The spinal cords were sectioned (14–18 μm) using a cryostat.

For antigen recovery, sections were incubated in heated (95–100°C) antigen recovery solution (1 mM EDTA and 5 mM Tris, pH 8.0) for 20 min and cooled down for 20–30 min. Before applying antibodies, sections were blocked in 10% normal goat serum in PBS with 0.1% Tween 20 for 1 h. Sections were incubated with primary antibodies at 4°C overnight and visualized using goat anti-rabbit IgG Alexa Fluor 488, goat anti-mouse IgG Alexa Fluor 594, or anti-guinea pig IgG Alexa Fluor 594 (1:350; Invitrogen) for 1 h at room temperature. Images were captured using a Leica digital camera under a fluorescent microscope (Leica). Primary antibodies against the following antigens were used: Lhx3 (1:20, Developmental Studies Hybridoma Bank), FoxP1 (1:100; kindly provided by Dr. B. Novitch, University of California, Los Angeles, Los Angeles, CA, USA), green fluorescent protein (GFP; 1:500; Rockland), and HB9 (1:20; Developmental Studies Hybridoma Bank).

The total number of labeled motor neurons on electroporated and non-electroporated sides of the spinal cord was quantified from at least four sections (14–16 μm) showing strong eGFP expression collected from at least five electroporated embryos. Data is shown as the difference in total number of a specific MN subtype between the electroporated and non-electroporated side of the spinal cord. Statistical significance between experimental and control groups was analyzed by two-tailed t-test. Data were shown as mean ± SEM.

FoxP1 fluorescence intensity was measured in at least three sections from each of at least three embryos. At HH stages 20 and 24, FoxP1 is expressed only in the ventral horn throughout the cervical, thoracic, and lumbar regions of the spinal cord. Fluorescence intensity was measured using ImageJ (NIH) in identically sized regions encompassing all FoxP1⁺ neurons in the ventral horn and recorded as the intensity of the electroporated side divided by that of the non-electroporated side of each sample.

Our previous work showed that miR-9 regulates MN subtypes by targeting chick FoxP1 (cFoxP1) via two miR-9 binding sites in its 3′-UTR (Figure 1A; Otaegi et al., 2011). Over-expression of miR-9 significantly decreased numbers of FoxP1⁺ MNs but increased Lhx3⁺ and HB9⁺ MNs in chick spinal cords, electroporated at HH stages 10–11, and analyzed at stage 24 (Figures 1B,C,F,G). As previously reported, we observed a reduction in the number of FoxP1⁺ cells, and corresponding increases in the number of Lhx3⁺ and Hb9⁺ cells on the electroporated side of the spinal cord. If FoxP1 is the critical target of miR-9, then restoring the FoxP1 coding sequence (CDS) alone, without its miR-9-sensitive 3′-UTR in miR-9 over-expressing spinal cords should rescue altered MN subtypes. Co-electroporating miR-9 with a small amount of FoxP1 CDS was able to rescue miR-9’s effects on MN production, as detected by a reduced increase in the number of Lhx3⁺ and HB9⁺ MNs, compared to miR-9 alone (Figures 1D,F,G). We next electroporated FoxP1 CDS alone. Consistent with previous results (Dasen et al., 2008; Rousso et al., 2008), over-expression of FoxP1 CDS reduced Lhx3⁺ and HB9⁺ MNs (Figures 1E,H). Our results indicate that miR-9 and FoxP1 have opposite effects on MN production and FoxP1 CDS specifically rescues miR-9 effects on MN subtypes.

MicroRNA sponges, transcripts with complementary sequence to mature miRNAs, have been used as a loss-of-function approach to study miRNA function in vitro and in vivo (Ebert et al., 2007; Gentner et al., 2008; Loya et al., 2009; Ebert and Sharp, 2010). Previous work has shown that sponges containing bulged miRNA binding sites can block a miRNA from silencing its endogenous targets (Figure 2A).

To block miR-9 silencing activities in vitro and in vivo, we designed a complementary sequence called miR-9 sponge, which contains no other predicted miRNA binding sequences (Figure 2B; Table 1). We first tested miR-9 targeting effects on FoxP1 3′-UTR using luciferase assays. The 3′-UTR of cFoxP1 was cloned into a luciferase construct. The activity of luciferase was greatly reduced by miR-9, but not by a miR-9 construct with three deactivating mutations in the seed sequence (miR-9-Mut), when the luciferase construct contained the cFoxP1 3′-UTR (Figure 2C). Luciferase constructs containing no 3′-UTR were unaffected by miR-9. Our results indicate that miR-9 specifically targets FoxP1 3′-UTR.

Since miRNA sponges are able to block miRNA repression of a target gene, we next co-transfected miR-9 and miR-9 sponge in a luciferase assay. The miR-9 sponge transcript containing six closely spaced (6 nt spacer) binding sites was able to rescue the knockdown effect of miR-9 on FoxP1 3′-UTR, while a similarly designed sponge construct with scrambled binding sites had no effect (Figure 2D). These results indicate that miRNA sponges can be used to block miRNA silencing activities on target genes.

In order to achieve the most effective and repeatable knockdown of endogenous miRNA activities in vitro and in vivo using miRNA sponges, we optimized sponge design using miR-9 sponge as an example.

Existing sponges have used 6–8 nt separation space between miRNA binding sites. A recent study has shown that the RNA induced silencing complex (RISC) provides RNase protection for 46–62 nt in the 3′-UTR of target mRNAs (Chi et al., 2009). This suggests that using increased spacing between miRNA binding sites in sponge design might allow more RISCs to bind and thus increase miRNA knockdown effects. To test this, we generated miR-9 sponge constructs with 6, 29, and 42 nt spacers separating miR-9 binding sites (the spacers contain no miRNA binding sites). These spacers created binding sites of different lengths for the RISC: 28, 51, and 64 nt respectively (Figure 3A). Luciferase assays demonstrated that all three sponge constructs were able to interfere with miR-9 induced silencing activity (Figure 3B). There was no discernable blockade of miR-9 silencing from sponges with scrambled binding sites. Opposite to our prediction, using a shorter spacer between binding sites in miR-9 sponges generated a stronger blockade of miR-9 activities.

Some sponge designs have put miRNA binding sites in the 3′-UTR of a coding gene (typically a reporter such as GFP), while others have simply generated a non-coding transcript. To test whether a coding gene is able to maximize sponge function, we generated miR-9 sponge constructs with and without a coding gene, here using improved Cre (iCre) as an example. Using the most effective 6 nt spacer sponges we tested relative luciferase activity using iCre-miR-9 sponge or miR-9 sponge alone (Figure 3C). miR-9 sponge alone was able to interfere with miR-9 silencing effects. But the presence of the coding gene iCre resulted in stronger rescue (Figure 3D).

Previous studies have shown blocking effects using different numbers of binding sites in miRNA sponges, typically ranging from 4 to 16 binding sites (Ebert and Sharp, 2010). To test whether the number of binding sites in a sponge may affect the sponge’s blocking ability, we generated constructs containing 6, 12, or 24 binding sites per sponge transcript (Figure 3E). miR-9 sponge with 12 binding sites showed higher rescue effect on miR-9 silencing activity than sponges with 6 binding sites. miR-9 sponge with 24 binding sites did not display the highest rescue effect (Figure 3F). Moreover, when a higher concentration of sponges was used, all three constructs were able to fully rescue miR-9 silencing activity (data not shown). These results suggest that the optimal number of binding sites is greater than 6, and somewhere around 12.

Taken together these results show that a miR-9 sponge optimized for the most efficient function in vitro contains more than 6 but fewer than 24 binding sites, short spacers between binding sites, and is expressed as the 3′-UTR of a coding gene.

To test whether miR-9 sponge functions in vivo, we examined the effect of miR-9 and miR-9 sponge on FoxP1 expression levels and MN subtype development in the chick spinal cord. A sponge containing six miR-9 binding sites with the shortest spacers or a similar scrambled control sponge were first co-electroporated with miR-9 in the chick neural tube at stage HH10–12 and analyzed at stage HH 24. Consistent with our previous results, over-expressing miR-9 with control sponges caused a significant reduction in the number of FoxP1⁺ MNs and subsequently a great increase of Lhx3⁺ and HB9⁺ MNs (Figures 4A,E). Interestingly, co-electroporation of miR-9 sponge and miR-9 was able to partially rescue the reduction of FoxP1⁺ MN numbers and resulted in milder changes of Lhx3⁺ and HB9⁺ MNs, suggesting that miR-9 sponge can block exogenous miR-9 silencing activities on the target gene FoxP1 (Figures 4B,E). miR-9 sponge and scrambled control sponge were next co-electroporated with miR-9-Mut construct in the spinal cord. Neither miR-9-Mut nor scrambled sponge displayed significant activity on FoxP1, Lhx3, or HB9 expression (Figures 4C,E). However, miR-9 sponge caused an increase in FoxP1 expression levels, as detected by FoxP1 fluorescent signal intensity on the electroporated side of the spinal cord compared to the non-electroporated side (Figure 4D). At the same time, slightly fewer Lhx3⁺ MNs were detected (Figure 4E). Furthermore, in situ hybridization for miR-9 showed that detectable miR-9 was reduced in the ventral horn on the electroporated side of the spinal cord in the presence of miR-9 sponge (Figure 4F). Together, these data suggest that the sponge may be leading to a reduced concentration of mature miR-9, permitting higher expression of FoxP1 and altering specification of Lhx3⁺ spinal MNs.

To further examine this finding, chick embryos were electroporated with mir-9 sponge alone or scrambled sponge alone at low (0.5 μg/μl) and high (1.5 μg/μl) concentrations, evaluated at HH 20, a time when FoxP1 expression is present, but endogenous miR-9 is weak or absent; and at HH 24, when both are expressed (Otaegi et al., 2011). FoxP1 fluorescent intensity was quantified in a region containing all FoxP1⁺ MNs in the ventral horn on the electroporated versus non-electroporated side. FoxP1 intensity was not changed by sponges on the electroporated side at HH 20, when miR-9 is weak or absent (Figures 5A,C,I). At HH 24, a low concentration of miR-9 Sponge was unable to create an effect (Figures 5E,I), but at high concentration, FoxP1 intensity was significantly increased (Figures 5G,I). Scrambled sponge showed no effect in any condition (Figures 5B,D,F,H,I). This demonstrates that ectopic expression of miR-9 sponge blocks endogenous miR-9 activity in a concentration dependent manner, and releases its silencing of target gene FoxP1.

Together, these results indicate that miR-9 sponge reduces miR-9 presence, effectively blocking its silencing activity in vivo and revealing the role of endogenous miR-9 in MN fate decisions (Figure 6).