Genome and epigenome engineering CRISPR toolkit for in vivo modulation of cis-regulatory interactions and gene expression in the chicken embryo

CRISPR/Cas9 genome editing has been successfully used in a number of model organisms including Drosophila (Port et al., 2014), mouse (Wang et al., 2013), Xenopus (Guo et al., 2014), zebrafish (Hwang et al., 2013), lamprey (Square et al., 2015) and chick (Véron et al., 2015). CRISPR/Cas9 technology requires two components, a single guide RNA (sgRNA) and a Cas9 endonuclease. The sgRNA includes a user-defined target-specific 20 bp spacer fused directly to a trans-activating RNA (tracrRNA), which is necessary for efficient Cas9 loading. Cas9 generates a double-stranded break (DSB), most frequently repaired by the non-homologous end-joining (NHEJ) pathway, resulting in disruption of the targeted genomic region by introduction of indels. In a previous study using CRISPR in the chicken embryo, the authors employed a tetracycline-inducible Cas9 and Tol2-mediated integration of CRISPR components to knock out Pax7 at later stages, allowing sufficient time for somatic genome editing (Véron et al., 2015). CRISPR-mediated germ-line editing using cultured primordial germ cells (PGCs) has recently been employed to develop transgenic chicken lines (Dimitrov et al., 2016; Oishi et al., 2016).

Here, we have established an in vivo genome and epigenome engineering toolkit for studying gene regulatory interactions in the early chicken embryo. Our optimised methods not only establish bi-allelic CRISPR-mediated gene knockouts, but also enable use of RNA-guided nuclease-deficient dCas9-effector fusion proteins to directly target and modulate the chromatin landscape and, as a consequence, gene expression in a developing embryo. To this end, we have generated and optimised a novel mini-vector system that uses a chick U6 promoter to mediate sgRNA expression in vivo. We have also created a wild-type Cas9 expression vector with a Citrine reporter, as well as several dCas9-effector constructs that enable epigenome manipulation of endogenous enhancers and promoters.

As a proof of principle, we use this novel toolkit to confirm gene regulatory interactions during early neural crest (NC) development. The chicken embryo is an ideal model for probing gene regulatory circuits, as it is amenable to in vivo perturbation using highly efficient electroporation methods. Moreover, ex ovo bilateral electroporation, in which each side of the embryo receives a separate set of plasmids, provides an excellent internal control for each experiment. Our analysis pipeline enabled (1) knockout of upstream transcription factors (TFs) and assessment of their effects on enhancer activity, (2) deletion of endogenous enhancers, (3) epigenetic modulations of endogenous enhancers to assess their role in regulation of the endogenous gene, and (4) premature activation of endogenous gene loci. The toolkit and optimised protocols developed in this study provide a comprehensive resource for studying gene regulatory interactions in the early chicken embryo.

For the purpose of this study, we have generated new plasmids encoding NLS-tethered Cas9 with two different fluorescent reporters (Cas9-2A-Citrine and Cas9-IRES-mRFP1) under the control of chicken β-actin promoter. The efficient delivery and broad even distribution of genome-editing components (Cas9 and guide RNA expression vectors) in the targeted cells is an essential pre-requisite for achieving a highly penetrant effect. Given the large size of Cas9 expression plasmids (∼10 kb), we first sought to quantify the proportion of cells that receive the plasmids using epiblast electroporation at Hamburger–Hamilton (HH) stage 4 followed by modified EC (early chick) culture (Sauka-Spengler and Barembaum, 2008) (Fig. 1A). Using this approach, all the derivatives of epiblast [ectodermal, neural, epidermal and NC] receive the constructs, but mesoderm and endoderm derivatives, which ingress earlier, do not. To assess electroporation efficiency and mosaicity, we performed immunohistochemistry on cryosections from three individual embryos electroporated with Cas9-2A-Citrine using an anti-GFP antibody. Image analysis and cell quantification using DAPI as counterstain revealed that ∼60% of cells in a wide region of interest (ROI), which included the entire neural tube and surrounding ectoderm as well as underlying mesoderm and endoderm tissues, were Citrine positive. However, when quantifying Citrine expression within a more focussed ROI including only the dorsal neural tube (dNT) region, we found that ∼80% of cells expressed the Cas9 construct (Fig. 1B,C), which is comparable to genome engineering (GE) experiments performed in routine cell culture transfections. We therefore conducted all downstream analysis of genome and epigenome editing effect on dissected dNTs (Fig. 1A).

To achieve efficient GE using the CRISPR/Cas9 system, it is also essential to maintain high expression of sgRNAs, which are rapidly degraded when not incorporated into Cas9 protein (Hendel et al., 2015). Most current RNA Pol III-dependent systems for expression of sgRNAs in amniotes employ a human RNU6-1 promoter, inherited from siRNA expression vectors (Miyagishi and Taira, 2002). However, the optimised Drosophila GE toolkit makes use of an alternative Pol III promoter (U6.3), which exhibits much higher activity in that organism (Port et al., 2014).

To build an optimal sgRNA expression system in the chicken embryo, we generated four sgRNA mini-vectors, each harbouring a different chick U6 promoter (U6.1, U6.2, U6.3 and U6.4) (Kudo and Sutou, 2005), tracrRNA and BsmBI-flanked cloning cassette (Fig. 2A, Fig. S1A). Using a modified Golden Gate assembly, we cloned the same spacer targeting the coding region of the FoxD3 gene into four U6-mini-vectors. We co-electroporated each U6-mini-vector with our plasmid ubiquitously expressing Cas9 (Cas9-2A-Citrine; Fig. S1B) into the entire epiblast of stage HH4 chicken embryos (Hamburger and Hamilton, 1951) and allowed embryos to develop to stage HH10. To measure the efficiency of sgRNA transcription from different U6 promoters, we analysed genome-editing events caused by different U6-mini-vector/Cas9 co-electroporations. To this end, we dissected cranial dNTs of four individual embryos for each U6-mini-vector/Cas9 and Cas9-only controls and assessed the presence of DNA hetero-duplexes within the target region using high-resolution melt analysis (HRMA) (Bassett et al., 2013; Dahlem et al., 2012). Normalised, temperature-shifted melt curves for U6.1 and U6.3 promoters showed consistent, reproducible evidence of Cas9-mediated mutations, whereas we observed more variable effects from U6.2 and U6.4 promoters, compared with Cas9-only controls (Fig. 2B). Thus, the U6.3 promoter-driven sgRNA expression mini-vector (pcU6.3) was used in all subsequent experiments.

To further quantify genome-editing efficiency, we profiled CRISPR-mediated somatic indels using a targeted next-generation sequencing (NGS) approach. We generated NGS libraries from six individual samples by amplifying the sgRNA-targeted region with primers that include sequencing adaptors and custom indexes (Fig. 2C; adapted from Gagnon et al., 2014). Sequencing reads were mapped to the FoxD3 amplicon and analysed using the CRISPResso tool (Pinello et al., 2016) to determine the indel frequency. The analysis showed that all experimental embryos had a higher percentage of NHEJ events than the controls (Fig. 2D), as well as multiple deletion variants (Fig. 2E,F). Controls showed minimal modification as a consequence of five recurrent single nucleotide polymorphisms (SNPs) (Fig. 2F).

Having established an efficient sgRNA delivery protocol, we next used CRISPR/Cas9-mediated gene knockout to probe input-enhancer interactions, focussing on targeting TFs implicated in NC development. To this end, we designed sgRNAs targeting Msx1, Pax7, Sox9, c-Myb and Ets1, previously demonstrated to act upstream of the FoxD3 enhancer NC1 (Simões-Costa et al., 2012) or the Sox10 enhancer 10E2 (Betancur et al., 2010) (Fig. 3A). sgRNAs were targeted to the intron-exon boundaries preceding the essential exons encoding the functional DNA-binding domain within each TF. This strategy, intended to destroy the splice acceptor site, would result in exclusion of the TF-binding domain from the spliced transcript and would thus ensure a robust loss of TF function (Fig. 3A). sgRNA spacers were predicted manually by scanning regions of interest for proximal protospacer adjacent motif (PAM) sequences and final choices were based on low self-complementarity and unique alignment to the chick genome. For each TF, three different sgRNAs were selected, cloned into the pcU6.3 mini-vector and tested individually by HRMA as described above (Fig. 3B). The guides mediating the most robust CRISPR/Cas9 cutting of the targeted region were chosen for further experiments.

We used our NGS-based indel validation approach to assess the extent of genomic perturbation at the targeted TF loci (Fig. 3C-E). To this end, we co-electroporated selected sgRNAs with ubiquitous Cas9-2A-Citrine using a bilateral electroporation assay at HH4 (Fig. 4A). The left side of the electroporated embryo received the target sgRNA+Cas9 and the right side a scrambled sgRNA+Cas9 as an internal control (Fig. 4A). Left (experimental) and right (control) dNTs from the same embryo were dissected at the 6 somite stage (ss) and analysed independently for presence of GE-induced indels. Three individual embryos per factor (Msx1, Ets1, Sox9, Pax7 and c-Myb) were analysed. Alignments to the corresponding reference amplicons for each factor showed a variety of indels and mutations occurring only on the experimental side (Fig. 3C-E, Fig. S2A,B), with control side amplicons showing no perturbations, but highlighting the position of recurrent SNPs. Sashimi plots (Fig. 3C′-E′, Fig. S2C,D) across the analysed amplicons suggested that indels predominantly occur close to the Cas9 cleavage site, with consistently higher prevalence of deletions than insertions.

Next, the selected sgRNAs were bilaterally co-electroporated with ubiquitous Cas9-2A-Citrine and either FoxD3 (NC1) or Sox10 (10E2) enhancer driving mCherry (Fig. 4B,C), to assess enhancer reporter activity in the condition of the CRISPR/Cas9-mediated TF knockout. Consistent with previous morpholino-based studies (Betancur et al., 2010; Simões-Costa et al., 2012), we found that knockout of either Ets1, Msx1 or Pax7 strongly reduces mCherry expression mediated by the NC1 enhancer at 5-7ss in 70%, 100% and 71.4% of cases, respectively (Fig. 4D,D′,E,E′,F,F′; n=10, 7 and 11, respectively). At later stages, however, only Pax7 knockout had a moderate effect (33.3%). Additional experiments with NC1-Cerulean corroborated these findings (Fig. S6). Similarly, knockout of Ets1, Sox9 and c-Myb led to a decrease in the activity of Sox10 enhancer, 10E2, with a 100% penetrance for all factors at 5-7ss (Fig. 4G,G′,H,H′,I,I′). However, we observed a decreased effect at later stages (8-10ss; Sox9 14.3% and Ets1 50%), in line with previous findings that Sox10 autoregulates to maintain 10E2 activity and Sox10 expression (Betancur et al., 2010; Wahlbuhl et al., 2012).

To quantify the observed effect at a cellular resolution and ascertain the cell-autonomous loss of enhancer activity upon GE-mediated disruption of upstream TFs, we used high-resolution confocal imaging (Fig. 4D″,D‴,E″,E‴,F″,F‴,G″,G‴,H″,H‴,I″,I‴). Image quantification software was used to demarcate and manually count Citrine-positive cells for all five TFs that were knocked out. mCherry intensity within the Citrine-positive cells on the left and right side of each embryo, previously electroporated with target and control sgRNA, respectively, was automatically measured. Histograms depicting distribution of mCherry intensity (Fig. 4J-O) show clear separation of average density profiles on the left (experimental) and right (control) side of the embryo. We observe a near-complete loss of mCherry fluorescence in Citrine-positive cells on the left, where TF is targeted, and a Gaussian-like distribution of mCherry fluorescence with most Citrine-positive cells displaying medium to high level of expression on the right side, which received scrambled sgRNA. We also observe Citrine-negative cells that lack mCherry, on both control and experimental sides of the embryo; this is consistent with initial reports of studies using NC1 and 10E2 in which enhancer activity was not ubiquitous in the dorsal neural tube. As noted above, the effect of targeting Pax7 is analysed at earlier stages, when the enhancer is active in a smaller number of NC cells, and thus the percentage of Citrine-positive/mCherry-negative cells on the control side is higher, resulting in a shift of the average profile (Fig. 4L).

We next used fluorescent hybridisation chain reaction (HCR) in situ (Choi et al., 2016) to assess the effect of TF knockout on the endogenous expression of the respective downstream target (FoxD3 or Sox10). Similar to findings in the original studies (Betancur et al., 2010; Simões-Costa et al., 2012), FoxD3 transcripts were reduced on the experimental side, compared with control, following Ets1, Msx1 or Pax7 knockout (Fig. 5A-C′), whereas Sox10 expression was reduced when Ets1, c-Myb or Sox9 were targeted (Fig. 5D-F′).

Finally, we sought to compare the efficiency of genome editing using a single sgRNA targeted to a splice acceptor site upstream of an essential Pax7 exon (this study) to a pair of optimised sgRNAs, each targeting a different splice donor site flanking the same essential exon (Véron et al., 2015). To this end, two sgRNA sequences from the Véron et al. study (1.7 and 2.17) were cloned in the new cU6.3 sgRNA expression vector and tested in our bilateral electroporation assay (Fig. S3A). NC1-mediated mCherry activity was reduced on the experimental compared with the control side of the embryos (Fig. S3B,B′). Endogenous Pax7 expression was also reduced, albeit with lower penetrance compared with our single sgRNA experiments (Fig. S3C), demonstrating that gene knockout can be effectively achieved using a single sgRNA astutely selected to target the essential splice acceptor site (Fig. S3A). It is of note that all sgRNAs targeting TFs in this study were chosen in this manner (Fig. 3A).

The results presented here demonstrate that the CRISPR/Cas9 system is effective in the chicken embryo, as well as re-validating previously shown NC cis-regulatory circuits (Betancur et al., 2010; Simões-Costa et al., 2012).

To characterise enhancer function fully, it is important to study them in their endogenous genomic context. Thus, we used our newly developed CRISPR/Cas9 tools to remove the endogenous FoxD3-NC1 cranial enhancer (Fig. 6A) in vivo and assess the consequence of its deletion on the expression of FoxD3. We designed and tested sgRNAs flanking the NC1 core region and used these in conjunction with Cas9-2A-Citrine to remove the enhancer. To assay the effect of NC1 knockout, we used bilateral electroporation with target sgRNAs+Cas9 introduced on the left and scrambled control sgRNA+Cas9 on the right side of the same embryo (Fig. 6C). Embryos were reared to the desired stages, experimental (left) and control (right) dorsal neural tubes dissected, and the effect on endogenous FoxD3 expression was assessed using quantitative RT-PCR (qPCR). We observed a decrease in FoxD3 expression on the experimental side in 77.7% of embryos analysed at 5-8ss (n=9) (Fig. 6D). The threshold level of FoxD3 expression at its onset (∼0.35 arbitrary units on the control side) was established using the absolute quantification with the same standard curve cDNA across all experiments. This demonstrates that endogenous NC1 enhancer activity is essential for the onset of the FoxD3 expression.

To further refine our analysis of enhancer function in vivo, we adopted targeted epigenome engineering (EGE) approaches to alter the chromatin landscape associated with active enhancers. Hitherto, such techniques have only been used in vitro (Kearns et al., 2015; Mendenhall et al., 2013; Thakore et al., 2015). To bring EGE methodology into the developing chicken embryo, we generated chick expression constructs driving fusion proteins of the catalytically inactive Streptococcus pyogenes Cas9 (dCas9) with lysine specific demethylase 1 (LSD1) or the Krüppel-associated box (KRAB) domain (Fig. S5C). LSD1 is a lysine-specific demethylase that catalyses the removal of H3K4me1/2 and H3K9me2, which are associated with active and repressive chromatin, respectively (Shi et al., 2004) (Fig. 6G), whereas the KRAB domain is thought to recruit a chromatin compaction complex thus rendering enhancers inaccessible (Sripathy et al., 2006) (Fig. 6E). LSD1 was first used as a TALE-fused effector that demethylates enhancer-associated histone methylation leading to inactivation of targeted enhancers (Mendenhall et al., 2013). More recently, Neisseria meningitidis dCas9 fusions with either LSD1 or KRAB were used to inactivate known Oct4 (also known as Pou5f1) cis-regulatory elements in mouse embryonic stem cells (Kearns et al., 2015) and human codon-optimised S. pyogenes dCas9-KRAB was used to inactivate the HS2 enhancer within the globin locus control region in K562 erythroid leukaemia cells (Thakore et al., 2015).

Previously published work in cell culture systems suggests that dCas9-targeted gene activation requires use of multiple guide RNAs (Mali et al., 2013), whereas dCas9-KRAB- or dCas9-LSD1-mediated enhancer repression can be achieved with a single guide (Kearns et al., 2015). We initially conducted our targeted modification of the endogenous NC1 enhancer activity using a single sgRNA, but observed no effect on FoxD3 expression (Fig. S4A). We therefore designed multiple sgRNAs tiled across the core region of the NC1 enhancer, as assessed by ATAC-seq (R.M.W., U.S. and T.S.-S., unpublished). Guides were evenly distributed to avoid potential steric hindrance due to the presence of multiple effector molecules. We used a T7 endonuclease assay to select sgRNAs that would mediate efficient wild-type Cas9 endonuclease activity (Fig. S4B), reasoning that this would also be indicative of effective dCas9/genome interaction, thus allowing docking of the fusion proteins at the ROI. We established that using five tiled sgRNAs yielded optimal repression in the context of the NC1 enhancer. The necessity for use of multiple guides per enhancer will be influenced by factors such as length of the element and the presence of other enhancers acting in concert to regulate the same locus and therefore optimisation assays for different enhancers should be performed. To minimise the number of vectors electroporated into the embryo, we generated an expression system based on the Csy4 cleavage, allowing simultaneous expression of multiple sgRNAs from a single plasmid (Nissim et al., 2014). However, we found Csy4 to be highly toxic to chicken embryos, even when introduced at low concentrations. Thus, we proceeded with co-electroporation of multiple cU6.3 mini vectors, which proved to be effective due to their small size.

Pools of five selected sgRNAs per enhancer were co-electroporated bilaterally with the ubiquitous dCas9-effector fusion construct on the left side of the embryo. The right (control) side was electroporated with equal molar quantity of scrambled control sgRNA mini-vector combined with the same dCas9-effector (Fig. 6C). Embryos were reared to the desired stages, their dNTs dissected and the effect of chromatin EGE modification on the targeted enhancer was assessed by comparing the expression levels of the endogenous target gene on the experimental and control side using qPCR. The time point for analysis of each tested enhancer was set to approximately 3 h before the fluorescent reporter can first be detected, to account for the maturation time of the fluorophore.

We first targeted dCas9-KRAB to the FoxD3 enhancer NC1, fluorescent reporter activity of which is detectable from 5ss (Fig. 6A), When analysed at 4-6ss, we observed a decrease in endogenous FoxD3 expression on the experimental side in 57% of embryos (n=14) (Fig. 6F), suggesting that although NC1 is required for proper FoxD3 expression, a different, potentially earlier acting enhancer(s) might co-regulate the onset of FoxD3 expression. Interestingly, targeting dCas9-LSD1 to the NC1 enhancer had a weaker effect on FoxD3 expression (Fig. S5C), with only 44.4% embryos (n=9) showing FoxD3 downregulation. Targeted enhancer repression was then applied to the Sox10 enhancer, 10E2. As no changes were observed in the endogenous level of Sox10 on the experimental versus control side at any of the stages tested (4-8ss), using dCas9-LSD1 or dCas9-KRAB (Fig. S5A,B), we reasoned that 10E2 element might not be the earliest enhancer controlling Sox10 expression. This assumption is supported by the observation that its fluorescent reporter activity only starts to be detected at ∼8-9ss. Using epigenomic profiling we have identified a novel Sox10 enhancer (enhancer 99, enh-99; Fig. 6B), yielding Sox10-like fluorescent reporter activity from 4ss. We hypothesised that this element might be important for the onset of endogenous Sox10 expression. Indeed when dCas9-LSD1 activity was targeted to the enh-99, we achieved a reproducible (100%, n=9, **P<0.01) knockdown of endogenous Sox10 expression on the experimental versus control side, in embryos ranging from 5 to 7ss (Fig. 6H). When we targeted dCas9-KRAB to Sox10 enh-99, we observed downregulation of Sox10 in 62.5% of embryos (n=11) (Fig. S5D), but only 33.3% of embryos (n=15) showed a decrease in Sox10 when the 10E2 enhancer was targeted with dCas9-KRAB (Fig. S5B). Our results suggest that the two presented EGE approaches have different mechanistic modes of action, with one efficiently stopping the initiation of the enhancer activity (effect of dCas9-LSD1 on enh-99), and the other (dCas9-KRAB) allowing the repression of active elements that are already engaged in enhancing gene expression (NC1 and enh-99). Previous work in cell culture shows that enhancer-targeted LSD1 activity results in depletion of the active enhancer marks H3K4me2 and H3K27Ac, but causes no change in the histone landscape of the promoter (Kearns et al., 2015; Mendenhall et al., 2013). In contrast, KRAB activity targeted to the enhancers does not cause any changes to the chromatin modification status at the element itself, but does lead to an increase of the repressor marks (H3K27me3/H3K9me3) at the cognate promoter (Kearns et al., 2015). Consistently, our results suggest that efficient LSD1-mediated enhancer decommissioning can only be achieved at early steps of gene activation, whereas KRAB-mediated regulatory repression might be more potent at genes already engaged in transcription. Thus, the choice of the approach to target any given enhancer will be dictated by the timing of its activity. Applying these two different approaches not only shows that specific enhancers can be successfully decommissioned in vivo, but also demonstrates their functional importance to the expression of the cognate downstream target gene.

To complement our gene and enhancer loss-of-function studies, we have adapted the dCas9-VP64-mediated activation of endogenous gene loci in vivo. VP64 is a well-established transcriptional activator domain consisting of a tetrameric repeat of the minimal activation domain found in herpes simplex protein VP16 (Seipel et al., 1992). dCas9-VP64 fusion has been successfully used in CRISPR-ON experiments to activate gene expression ectopically (Cheng et al., 2013; Guo et al., 2017). Here, we use a dCas9-VP64 fusion in conjunction with the synergistic activation mediator (SAM) system (Konermann et al., 2015). SAM is a three-component system employing (1) a modified pcU6.3_MS2 vector with a tracrRNA scaffold containing stem loops that associate with bacteriophage MS2 coat protein (MCP), (2) MCP-VP64 and (3) dCas9-VP64 expression construct (Fig. S1A,D). Co-expression of all three components results in saturation of the targeted site with effector molecules, thus enabling a ‘CRISPR-ON’ response using just one sgRNA (Fig. 7A). We screened five sgRNAs selected within the region of the Sox10 promoter upstream of the transcription start site using the T7 assay. We selected the most efficient sgRNA to activate Sox10 expression prematurely using the dCas9-VP64/SAM system, again using the same reasoning that efficient wild-type Cas9 endonuclease activity provides strong evidence for highly efficient dCas9-VP64 docking. Following bilateral electroporation at HH4 (Fig. 7B), we observed an increase in Sox10 expression at 3-4ss on the experimental versus control side of the embryo, indicating premature activation of the gene (50% embryos, n=12; Fig. 7C). This effect is further confirmed by fluorescent HCR in situ analysis of Sox10 expression on the experimental (activated) and control sides on the embryo (Fig. 7D). It is of note that although all the components of the activating machinery (sgRNA, dCas9-VP64, MCP-VP64 plasmids) are expressed ubiquitously in the epiblast/ectodermal derivatives, observed Sox10 activation that was not ectopic, but rather extemporaneous (premature) in the same NC cells that would normally express the gene. This suggests that the mechanisms controlling tissue-specific gene expression in a developmental context are tightly controlled not only by regulation at the gene promoter, but require competency at the level of distal cis-regulatory elements.

Having successfully used the SAM system to activate gene expression with a single sgRNA, we next applied this approach to our enhancer knockdown experiments with dCas9-KRAB and dCas9-LSD1 fusions. However, we recorded no evidence that this was effective, probably due to steric hindrance of multiple copies of large effector proteins.

Here, we have optimised CRISPR/Cas9 approaches to perform GE and EGE in the early chicken embryo, with a particular focus on probing gene regulatory interactions.

Previous in vivo CRISPR studies using chick (Gandhi et al., 2017; Véron et al., 2015) have focussed on sgRNA-mediated CRISPR/Cas9 disruption of the coding regions of targeted genes. In addition to improving the efficiency of the genome editing, this study focuses on building and using an EGE toolkit that enables in vivo deletion, repression and activation of endogenous enhancers and promoters using targeted transient modifications of the epigenomic landscape. Epigenome editing approaches have been predominantly used in in vitro systems (Kearns et al., 2015; Konermann et al., 2015; Mendenhall et al., 2013; Thakore et al., 2015); however, adaptions to in vivo models are beginning to emerge. A recent study used a dead Cas9 variant fused to the histone methyltransferase Ezh2 to achieve targeted gene repression in the developing mouse embryo (Albert et al., 2017). Also, CRISPR/dCas9-mediated gene activation and repression has been successfully implemented in zebrafish and worms using dCas9-KRAB and dCas9-VP160 fusion proteins (Long et al., 2015) and in Drosophila using the dCas9-VPR (VP64-p65-Rta) system (Lin et al., 2015). The present study introduces the chicken embryo as an excellent model for in vivo EGE building on previously established advantages for studying enhancer activity in the chick. The methods developed here will enable functional analyses of enhancer activity and function and further facilitate probing of gene regulatory interactions in the chick embryo, as well as provide an adaptable system for use in other amniotes.

In addition to the experimental procedures outlined below, full protocols pertaining specifically to the use of GE and EGE toolkit are available in supplementary Materials and Methods and at http://www.tsslab.co.uk/resources.

Embryos selected for immunostaining were fixed in 4% paraformaldehyde in PBS for 1 h at room temperature (RT) or at 4°C overnight (O/N). Embryos were washed three times (15 mins each) in PBS. Embryos were then cryoprotected in 15% sucrose in PBS (8 h at RT or O/N at 4°C), followed by incubation in 7.5% gelatine/15% sucrose in PBS O/N at 37°C and an equilibration in pre-warmed 20% gelatine in PBS (∼4 h). Embryos were embedded in 20% gelatine in PBS and cryosectioned at 10 µm. Prior to immunocytochemistry, gelatine was removed from the slides by a brief rinse in pre-warmed PBS (37°C). Sections were rinsed three times (5 mins each) in PBT (2% DMSO, 0.5% Triton X-100 in PBS), blocked in 10% donkey serum in PBT (block solution) for 1 h at RT, and incubated O/N at 4°C with primary antibody (1:200 in block solution). Sections were then washed in PBT at RT (three to five times times for 10 mins each), followed by incubation with secondary antibody (1:1000 in PBT) for 2 h at RT. Sections were washed six to eight times for 10 mins in PBT at RT then overnight at 4°C and mounted using Vectashield with DAPI (Vector Laboratories, H-1200). The primary antibody used to detect Citrine was rabbit anti-GFP (Torrey Pines Biolabs, TP401) and the secondary antibody used was AlexaFluor-488-conjugated anti-rabbit IgG (Thermo Fisher Scientific, A21206). Images of stained sections were taken on an upright Zeiss 780 confocal microscope.

Chick U6 sgRNA expression mini-vectors were cloned by replacing the BsaI-flanked cassette from pNG1-pNG4 vector backbones (Cermak et al., 2011) (Addgene plasmid #49043, deposited by Eric Mendenhall) with custom-synthesised gBlocks (IDT) containing chick cU6.1-4 promoters (Kudo and Sutou, 2005).

The pCAG_Cas9-2A-Citrine construct was generated by removing the IRES-H2B-RFP cassette from the pCI_H2B-RFP vector (Betancur et al., 2010) and inserting the Cas9-2A-Citrine fragment by In-fusion HD (Clontech, 638910) cloning.

To construct the pX330 dCas9-LSD1 vector, Cas9m4-VP64 was amplified from Addgene plasmid #47319 (deposited by George Church) and cloned into pX330 (Addgene plasmid #42230, deposited by Feng Zhang). VP64 was then removed by EcoRI digest and LSD1 was inserted from EMM67 (Addgene plasmid #49043, deposited by Bradley Bernstein) (Mendenhall et al., 2013). pX330 dCas9-KRAB was generated by In-fusion of a synthetic gBlock containing the KRAB sequence into the EcoRI-linearised pX330-dCas9-LSD vector.

pCAG-dCas9-KRAB-2A-GFP was generated by amplifying dCas9-KRAB-2A-GFP sequence from pLV hUbC dCas9 KRAB T2A GFP (Addgene plasmid #71237, deposited by Charles Gersbach) and cloning into the pCI-H2B-RFP vector linearised with NotI and XhoI.

The MCP-VP64 construct was generated by removing the IRES-H2B-RFP cassette from the pCI_H2B-RFP vector (Betancur et al., 2010) and inserting the MCP and VP64 fragments [amplified from Addgene plasmids #61423 (deposited by Feng Zhang) and #47319 (deposited by George Church), respectively] using two-fragment Infusion cloning.

Chosen spacer sgRNA sequences used in this study (see Table S1 for sequence information) were designed to include flanking BsmBI sites and corresponding overhangs for Golden Gate-based cloning into the U6 vector. Forward and reverse oligonucleotides were ordered from Integrated DNA Technologies and annealed by heating equal amounts of each to 94°C for 5 min, followed by cooling from 64°C to 34°C (at −1°C per minute). Annealed oligos are cloned into the U6 vector using BsmBI and T4 DNA ligase in modified, Golden Gate reaction; ten cycles of 37°C 5 min, 16°C for 10 min, followed by 50°C 5 min, 80°C 5 min. For detailed protocol, see supplementary Materials and Methods.

HRMA was used as selection criteria for efficient/functional sgRNAs. To assess the efficiency of sgRNAs rapidly, we electroporated each tested sgRNA (cloned into the cU6.3 vector) into four individual embryos and allowed them to develop to the desired stage. Genomic DNA was extracted from dissected dNTs following electroporation. Primers were designed to generate a ∼100 bp product spanning the sgRNA cut site. HRMA PCR was performed using Hotshot Diamond PCR Mastermix (Client Lifescience, HS002-TS) together with LC Green Plus dye (BioFire Diagnostics, BCHM-ASY-0005). Reactions were performed on a C1000 Touch Bio-Rad thermal cycler and Bio-Rad Precision Melt analysis software was used to visualise and analyse the data. The shift in normalised melt curve was used as evidence of the presence of heteroduplexes in the edited amplicons, compared with Cas9-only or Cas9-scrambled sgRNA controls. We used the penetrance of the effect as a measure of sgRNA efficiency and chose only guides that induced NHEJ mutations in 100% of analysed embryos for further experiments. For detailed protocol, see supplementary Materials and Methods.

Genomic DNA was extracted from single-electroporated embryos. Primers were designed to generate a ∼1 kb product spanning the sgRNA cut site. This fragment was gel-purified and incubated with NEB Buffer 2 at 95°C for 5 min. The sample was then cooled to 25°C. T7 endonuclease I (NEB, M0302) was added and samples were incubated at 37°C for 15 min. For detailed protocol, see supplementary Materials and Methods.

Primers including Illumina adapters were designed to amplify ∼120 bp region covering the sgRNA target site. Primary amplification was performed on gDNA from dissected half dNTs following electroporation. To maintain complexity, five replicates of two-step primary PCR reactions were performed. Following clean up, primary amplicons were amplified using Illumina Nextera primers including i5 and i7 indexes. Library integrity was assessed by Agilent Tapestation and Invitrogen Qubit assay, before sequencing on the Illumina Miseq platform using a v2 300-cycle kit. For detailed protocol, see supplementary Materials and Methods.

Fertilised wild-type chicken eggs were obtained from Henry Stewart & Co (Norfolk), staged according to Hamburger and Hamilton (1951) and electroporated as previously described (Sauka-Spengler and Barembaum, 2008; Simões-Costa et al., 2012). All experiments were performed on chicken embryos younger than 12 days of development, and as such were not regulated by the Animals (Scientific Procedures) Act 1986.

Detailed standardised protocols describing guide RNA selection, Golden Gate cloning, bilateral electroporation and validation by HRMA, T7 assays and NGS are available in supplementary Materials and Methods and at our resource page (http://www.tsslab.co.uk/resources). All plasmids are available from Addgene (https://www.addgene.org/Tatjana_Sauka-Spengler/).

RNA extractions of dissected embryonic tissue were carried out using the RNAqueous-Micro Kit (Life Technologies, AM1931). Oligo-dT-primed cDNA was synthesised using Superscript III reverse transcriptase (Invitrogen) and qPCRs performed using Fast SYBR Green reagent (Thermo Fisher, 4385612) on the Applied Biosystems 7500 Fast Real-Time PCR System. The standard curve method was used to quantify the gene expression. In all embryos, the contralateral side was used as an internal control. Statistical significance of bilateral electroporations was determined by generating a control embryo group that received scrambled sgRNAs on both sides and calculating P-values using χ² statistical test, based on contingency tables comparing control and experimental embryo groups. *P<0.05, **P<0.01.

Embryos were imaged on an Olympus MVX10 stereomicroscope with 2-2.5× objective using Axio Vision 4.8 software. A Zeiss 780 Upright confocal microscope was used for imaging at high cellular resolution and obtained images were used for quantification of the effect. For cryosection image analysis, cells were automatically counted using Fiji (ImageJ) (Schindelin et al., 2012), where Gaussian blur was set to 3 and the noise filter adjusted to 2. Three embryos were analysed, 15 consecutive 10 µm sections from each. From whole-embryo high-resolution images, Citrine-positive cells were manually counted and demarcated using ImageJ. This was overlaid onto the mCherry channel from which mCherry intensity within Citrine-positive cells was automatically quantified. Plots were generated in RStudio (http://www.rstudio.com/).

For in situ HCR, a kit containing a DNA probe set, a DNA HCR amplifier, and hybridisation, wash and amplification buffers were purchased from Molecular Instruments for each target mRNA. The Sox10 and FoxD3 probes initiate B3 (Alexa-546) and B4 (Alexa-546) amplifiers, respectively. In situ HCR protocol was previously detailed by Choi et al. (2016). The protocol provided by the manufacturer was followed. Briefly embryos were fixed with 4% paraformaldehyde for 1 h at room temperature, dehydrated in methanol and stored at −20°C. Following rehydration, embryos were treated with 10 mg/ml Proteinase K for 2.5 min and post-fixed in 4% paraformaldehyde for 20 min. Embryos were incubated with the probes in hybridisation solution overnight at 37°C and, following appropriate washes, incubated in the hairpin solution in amplification buffer overnight at room temperature protected from light.

We thank Professor Tudor Fulga for helpful advice on the project and manuscript, and Francesco Camera and Deniz Taşgin for technical assistance. Thanks are also due to Dominic Waithe for assistance with all the image analysis. High-resolution confocal imaging was conducted within the Wolfson Imaging Centre, at the Weatherall Institute of Molecular Medicine.