RNA interference (RNAi) is the process by which dsRNA silences gene expression, either by inducing the sequence-specific degradation of complementary mRNA or by inhibiting translation [5]. In a wide range of organisms, double-stranded RNA triggers posttranscriptional gene silencing or RNA interference (RNAi). Genetic and biochemical investigations of the mechanisms guiding RNAi in different organisms revealed the conservation of cellular machinery that cleaves long dsRNA into duplexes of 21- to 28-nucleotide siRNAs, which guide the sequence-specific degradation of mRNAs. siRNA, shRNA and miRNA elicit RNAi through common biochemical pathways involving complexes of enzymes including Dicer [6,7,8,9]. The siRNAs are then incorporated into the RNA-induced silencing complex (RISC), which unwinds the duplex siRNA into single-stranded siRNA [10]. The antisense strand of the duplex siRNA guides the RISC to the homologous mRNA, where the RISC-associated endoribonuclease cleaves the target mRNA at a single site in the center, which results in the silencing of the target gene [6,11]. One caveat of siRNA design is that not all 19–22 base RNA duplexes will cleave their target with efficacy. A key finding, which improved the design was when it was found that the RISC complex is asymmetric and favors the strand of the siRNA duplex with the least thermodynamically stable 5' terminus [12,13]. This information was used in the design of several algorithms to better select an effective siRNA target site within a gene [14,15]. Small interfering RNA (siRNA) and short hairpin RNA (shRNA) have been initially proven effective in reducing gene expression in cultured mammalian cells [6,16,17]. Later, gene silencing has been shown to be effective in rodents and is being considered for a therapeutic strategy in humans (reviews [5,18]).

In mammals, DCX is part of a small superfamily of proteins, defined by the presence of a conserved microtubule-binding domain, the DCX domain [19,20]. Mutations in the founding member of this gene family were found to cause X-linked lissencephaly (“smooth brain”) in males and double cortex syndrome in females [21,22]. Further studies revealed that additional members of this gene family are involved in an array of neurological disorders. Its closest family member is Doublecortin like kinase-1, DCLK1, which has recently been implicated in cognitive function in humans [23]. Rare copy number variations of Doublecortin like kinase-2, DCLK2, were detected in a study investigating attention deficit hyperactivity disorder (ADHD) [24]. Doublecortin domain containing 2, DCDC2, has been suggested to play a causative role in reading disabilities and has been implicated as a susceptibility gene for Dyslexia [25,26]. Retinitis Pigmentosa 1, RP1, and Retinitis Pigmentosa like 1, RP1L1, are associated with inherited blindness [27,28]. The known CNS expression pattern and possible function of the DCX gene superfamily are summarized in Table 1.

Despite the severe brain malformation observed in male patients carrying mutations in the DCX gene, the mouse knockout did not exhibit the expected inhibition in the migration of pyramidal neurons in the developing brain [53]. However, knockdown of Dcx using in utero electroporation inhibited migration of cortical excitatory neurons both in rat and in mouse brains [4,54]. It has then been suggested that the lack of the radial migration phenotype in the knockout models may be due to gene redundancy. It is postulated that whereas in case of the knockout there is plenty of time for developmental redundancy mechanisms to become operative; the knockdown involves an acute gene reduction, which may not allow sufficient time for redundancy mechanisms to evolve. This notion received additional support following findings that the knockout of Dclk1 did not result in an observable phenotype in the migration of pyramidal neurons in the developing brain [39,55]. Similar to Dcx, the knockdown of Dclk1 impaired the migration of pyramidal neurons [55]. Nevertheless, the double knockout of Dcx and Dclk1 had a clear effect on cortical development. More specifically, the double mutant mice demonstrated perinatal lethality, disorganized neocortical layering and profound cytoarchitectural defects of the hippocampus caused by the disruption of radial neuronal migration. In the adult Dcx mutant mice some deficits were noted in the migration of neurons to the olfactory bulb in the rostral migratory stream [56]. The possibility of gene redundancy was investigated in Dcx mutant mice, where the expression of transcripts and proteins, which are products of the Dclk1 and Dclk2 gene, were analyzed [46]. A minor change in the expression of one of the DCLK1 proteins was detected in this study. In addition, more severe phenotypes were noted in the combination of mutant alleles for Dcx and Dclk2 [47]. In particular, in the absence of Dcx and Dclk2 there was a dosage-dependent phenotype in the hippocampus, where hippocampal lamination was disrupted and it was accompanied with simplification of pyramidal dendritic arborizations. However, as mentioned above, the DCX protein family includes additional family members, which may participate in gene redundancy mechanisms. Dcdc2, an additional family member, was suggested to exhibit functional redundancy with DCX. Knockdown of Dcdc2 inhibited neuronal migration, whereas Dcdc2^−/− mice revealed no obvious phenotypes in neuronal migration, neocortical lamination, neuronal cilliogenesis or dendritic differentiation [26,57]. The investigators tested whether decreasing Dcx expression by RNAi would differentially impair cortical development in Dcdc2 knockouts and wild-type mice. Consistent with this hypothesis, they found that deficits in neuronal migration, and dendritic growth caused by RNAi of Dcx were more severe in Dcdc2 knockouts than in wild-type mice. These results suggest that Dcdc2 exhibits partial functional redundancy with Dcx. These conclusions could not have been reached relying on mouse knockout data exclusively. Thus, the implementation of knockdown technology in the developing brain and the combination of knockout and knockdown approaches allowed for better understanding of gene function.

Efficacy and specificity are two important issues that need to be addressed in a siRNA/shRNA system. Despite the fact that very sophisticated algorithms are being used to design the sequences, experimentally, only a portion of the sequences used in shRNA constructs is efficient in reducing the desired mRNA. Part of the problem may be due to the relative inaccessibility of the short sequence to form a duplex with the mRNA, due to the complex three-dimensional structure of the mRNA. This shortcoming may yield false negative results; however, usually it is possible to verify whether a gene has been successfully knocked down. Common assays include assaying the levels of the mRNA using real-time PCR (qPCR) of the target gene or in situ hybridization. If antibodies are available, then it is possible to test the reduction in the expression of the target protein by western blot analysis and/or immunostaining. However, it needs to be considered as a drawback when gene knockdown verification is not practiced, for example, in many high-throughput screens.

Target specificity is an additional critical step required for successful gene silencing. Off-target effects are driven by three main mechanisms: activation of the nonspecific immune response (such as the interferon response), saturating cell machinery and reduction of non-targeted mRNA (reviews [58,59]). Several practical steps are used to minimize unintended effects. Most design programs examine the existence of identical or near identical sequences in the genome using BLAST-based analysis. However, expression-profiling studies have demonstrated that a small degree of similarity is sufficient to result in off-target gene regulation [60]. In some cases, the off-target event may be due to converting a siRNA sequence into a miRNA (micro RNA), which requires less bases of exact homology for inhibition of translation through a pathway closely related to siRNA [59,61]. Therefore, siRNA sequences that are short and contain perfect matches of 11 bp to other genes are usually to be avoided. In each experiment, several sequences are tested for their specificity and efficacy. Usually, several concentrations of the siRNA are tested and experiments should be conducted with the lowest effective concentration. Sometimes, a combination of several sequences targeting a common mRNA is used to increase efficacy. Negative controls, which include either sequences with multiple mismatches, or sequences with no known homologies, or sequences with confirmed minimal effect assayed by high throughput approaches are used. Other types of controls include occasionally scrambled shRNAs or mismatch shRNA sequences. These controls are not always recommended, since in both cases it is possible that these controls may affect the expression of the target endogenous gene or off-target genes via microRNA activity, which require much less sequence identity, as described above. Finally, the ultimate control is a rescue experiment, which will be discussed below. A summary of the items discussed above is found in Table 2.

The rescue experiment involves expression of a construct of the relevant cDNA of the knocked down gene, which is resistant to the shRNA. The cDNA is introduced with the shRNA, and in case of a successful rescue, the functional phenotype is ameliorated. Resistance of cDNA to the shRNA can be achieved either when the complementary sequence is absent (for instance when it resides in the UTR), or when it contains silent mutations corresponding to the shRNA sequence. This is an excellent control, yet in some cases it may be very difficult to attain. Many of the genes involved in regulation of neuronal migration are dosage sensitive, thus reduced expression or elevated expression, which might be caused by the expression of the resistant mRNA, results in an apparent phenotype. For example, overexpression or reduced expression of the microtubule regulating kinase, MARK2/Par-1, affected neuronal migration [62]. A rescue experiment was successful only after tittering down the expression levels and controlling the temporal expression pattern using different promoters [62]. In the case of the DCX protein family, both gain and loss of function of Dclk1 disrupt the cell cycle progression of neuronal precursors, suggesting that very precise levels of Dclk1 are required for completion of mitosis [38]. Both knockdown and overexpression of Dcx impair neuronal migration [63]. Thus, very subtle changes in the concentration of DCX are required to achieve rescue. Rescue experiments can provide more information than just confirmation of specificity. The re-expression of Dcx at the postnatal stage restored developmental defects caused by the introduction of Dcx shRNA in the developing brain of the embryo [64]. Thus, this study demonstrated the existence of a time window in which the neurodevelopmental deficiency could have been reversed (review [65]). In addition, rescue experiments using other genes are very powerful to demonstrate genetic interactions. For example, DCX is phosphorylated by MARK2/Par-1, which resulted in its reduced affinity to microtubules [66]. We have shown that the reduction in DCX destabilized microtubules, while reduction in the cellular levels of MARK2 stabilized microtubules by alleviating its effect on DCX [67]. Subsequently, co-reduction of both proteins resulted in partial restoration of a normal migration pattern in vivo, and in partial restoration of centrosomal motility [67]. Thus, combinatorial additions of different shRNA may provide useful insights regarding genetic interactions.

The cerebral cortex is an intricate structure composed of different areas, layers and cell types; each of which originates at a distinct place and unique time during development. Studying the development of individual cell populations requires the ability to manipulate discrete groups of cells at different times. The profuse variety of knockdown methods allows for a refined regulation over the cortical population or the area that is being knocked-down. Regulation, both temporal and spatial, is possible by the use of conditional RNAi-constructs (i.e., Cre mediated conditional RNAi activation or inactivation and drug-inducible RNAi) and by the RNAi-construct transfection system.

Currently, two systems for transfection of RNAi-constructs into the rodent brain are utilized: viral transduction (for example [38]) and in utero electroporation [68,69], where the latter is more commonly used. It should be noted that in both methods involve injection into the brain, thus involving extremely skillful experimentation of well-trained persons. Mechanical damage of the tissue may evoke the response of the immune system. In case of in utero electroporation, the uterine horns are exposed and shRNA constructs, usually together with a fluorescent-protein, are microinjected into the lateral ventricle of the embryo’s brain (Figure 1a). The DNA is incorporated into the cells at the vicinity of the ventricular surface by the application of a series of square-pulse waves. Depending on the position of the positive pole of the electrode, different cortical areas can be targeted. By targeting the ventricular zone of the dorsal telencephalon, populations of the cortical projection neurons can be affected (Figure 1b) [68,69]. Most of the abovementioned experiments, targeted at investigating radial migration in the developing cortex, used this configuration [4,54,62,67]. However, slight changes in the position of the electrodes can result in labeling of other cell populations. For example, successful transfection of neurons in the lateral cortical stream in the rat brain was achieved [33]. These cells contribute neurons to structures in the ventral telencephalon including the amygdala and piriform cortex. The authors noted that they were able to transfect cells at the corticostriatal junction without significant contamination of cells in the ganglionic eminence despite the proximity of these structures. RNAi of either Dcx or Lis1 significantly affected migration of neurons to these structures, yet the knockdown of the two different genes did not result in the exact same phenotype. Two additional populations of cortical neurons that can be manipulated are the GABAergic interneurons, which originate in the ganglionic eminences of the ventral telencephalon (Figure 1c), and the hippocampal neurons, which originate at the hippocampal neuroepithelium (Figure 1d) [70,71]. Furthermore, small-diameter electrodes enable expression in a rather small, localized area [65]. Projection neurons compose the majority of telencephalon neurons; thus, most of the research employing in utero electroporation is aimed at this population. Nonetheless, knockdown of Dcx and Dclk1 in the ganglionic eminences showed inhibited migration of cortical interneurons [32]. Moreover, the knockdown of Dcx in interneurons, but not the knockdown of Dclk1, increased branching, suggesting that DCX and DCLK1 act through different mechanisms.

Cortical neurons are organized in layers according to their birthdates [72], therefore, choice of the timing of electroporation can label different cortical layers [73]. Thus, manipulating both the position and timing of the in utero electroporation can target distinct cortical populations.

Another technique to manipulate the population being knocked-down is the use of Cre-loxP mediated recombination [74,75,76]. The technique utilizes the Cre recombinase, which catalyzes the site-specific recombination events between two loxP sites. Several approaches use the Cre-mediated recombination technique (reviewed by [77]). In the conditional RNAi activation approach, Cre-mediated recombination excises an inhibitory sequence and activates the shRNA expression. To this end, the shRNA is separated from its promoter by a loxP-flunked stuffer sequence (Figure 2a). By the use of a Cre recombinase under the control of a cell-specific promoter, the shRNA expression is restricted to the cell population in which Cre recombinase will be expressed (see Table 3 and [65] for a review of cell-specific promoters used in the in utero electroporation system). Alternatively, in the conditional RNAi inactivation approach, the expression of the RNAi-construct can be expressed in all electroporated cells and specifically inactivated in a particular cell population. Here, the loxP sites flank the shRNA and its promoter and following Cre activity the shRNA will be lost (Figure 2b). Both approaches, conditional activation and inactivation, will usually require an additional fluorescent reporter which will mark the Cre expressing cells. Knockdown experiments usually allow easy visualization of the treated cells, which are coexpressing a fluorescent protein. Some analyses, for example, detailed neuronal morphology, require the labeling of sparse cells. Labeling of sparse cells can be achieved using in utero infection with low-titer of lentivirus, however, in case of in utero electroporation, usually multiple cells are transfected. To overcome this drawback, the expression of a fluorescent marker to label the cells can be titrated either by the amount of the plasmid introduced and/or the activity of Cre-loxP mediated recombination. In this case many of the cells in the electroporated area will express the shRNA; however, only a fraction of them will receive the Cre construct and will be GFP positive (Figure 2c,d). If a Cre-dependent shRNA is used with the Cre-dependent fluorescent marker and different concentrations of Cre recombinase, then only the labeled cells will express the GFP and the shRNA.

Furthermore, it is possible to regulate the timing of expression by using a Cre-recombinase, which is activated following the addition of 4-hydroxytamoxifen. The Cre recombinase has been fused to a mutated ligand-binding domain of the human estrogen receptor (ER) resulting in a tamoxifen-dependent Cre recombinase, Cre-ERT, which is activated by tamoxifen, but not by estradiol [80]. The Cre-ERT system is currently being used mainly in transgenic mouse experiments [81], but is very suitable for in utero electroporation [65]. Another system for temporal regulation of the RNAi expression is the drug-inducible RNAi. The logic behind this system relies on drug-mediated control of the shRNA promoter, where the addition of the drug (e.g., doxycycline) sequesters a repressor of the promoter and allows the shRNA to be expressed [82,83,84]. The fundamental difference between this system and the Cre-inducible RNAi is the reversibility. That is, Cre-inducible RNAi is a one-time excision that cannot be restored to its original state. In drug-inducible RNAi, whilst the drug is delivered, the RNAi is expressed, whereas its removal restores the system to its original state. Finally, it is possible to combine in utero electroporation experiments with different transgenic mice. The combination of in utero electroporation gene transfer with the growing set of conditional RNAi systems and regulated Cre expression constructs provides a toolbox to precisely control the RNAi expression at almost any time in development and in any neuronal population.

In transgenic mice cell autonomous features can be visualized using mosaic analysis with double markers (MADM) [85,86]. This system involves Cre-loxP-dependent recombination, however, the recombination is interchromosomal and mitosis dependent. The system involves formation of chimeric gene constructed from the N-terminal part of GFP together with the C-terminal part of DsRed (or another fluorescent protein) coding sequences. The parts of the fluorescent proteins are separated by an intron containing a loxP site, which are targeted to a chromosome containing the gene of interest to be mutated. The reciprocal chimeric gene, with the N-terminal part of DsRed and the C-terminal of GFP separated by an intron with a loxP site, is targeted to the reciprocal site of the WT chromosome. The resultant mice are then crossed to each other to generate the trans heterozygote mice. Cre-mediated recombination between the chromosomes results in green knockout cells, red WT cells and yellow heterozygous cells, the rest of the cells are colorless and heterozygous. Analyses of sparse and uniquely labeled mutant cells in mosaic animals reveal distinct cell-autonomous functions for the genes investigated. The system allows color-coding of the various genotypes and is very informative, nevertheless several drawbacks can be noted in comparison with the possible in utero electroporation methods. The MADM system is time-consuming and requires the generation of several knockin and knockout mice combined with a series of sequential mating of the respective mice. In addition, it should be noted that the background cells are heterozygous thus, may result in a phenotype different from the WT.

It is possible to visualize cell-autonomous and non-cell autonomous features using in utero electroporation. One possibility is to use sequential in utero electroporation, where initially the shRNA transfected cells are labeled with GFP and then, following a time delay, cells are electroporated with a red fluorescent protein. This scenario was tested first in the case of Dcx shRNA, where the time interval between consecutive electroporations was thirty minutes [4]. In this sequential electroporation the migration of the green cells, which received the shRNA of Dcx, was inhibited in a cell-autonomous manner. However, the position of the red cells, neighboring the green cells, was also disrupted, thus demonstrating a non cell-autonomous effect. An alternative possibility is using dual colors, which will indicate the cells in which the shRNA is active or inactive following Cre-dependent recombination (Figure 2). For example, in Figure 2b, the cells expressing shRNA are also green, but following the activity of Cre recombinase the red fluorescent marker will be expressed and the shRNA with the green fluorescent protein will be deleted. The relative concentration of Cre recombinase can result in different fractions of labeled cells, which reside very close to each other. Thus, it may be possible to study the cell autonomous features of the green, treated cells and their effect on the red, control neighboring cells.