Clonal analysis by tunable CRISPR-mediated excision

Clonal analysis originates from efforts to track the fate of individual cells and their progeny during development (reviewed by Kretzschmar and Watt, 2012). Early studies used cell transplantation from genetically distinct donors or labelling using vital dyes to track the fate of marked cells. Spontaneous somatic mutations that produce a visible cellular phenotype, such as the loss of pigmentation, were also used to observe the territories occupied by cell clones in mosaic animals and plants. In principle, any heritable somatic change that is associated with a visible marker can be used for this purpose, and different types of genetic events have been exploited for this, including chromosome loss, transposon excision, genetic rearrangement and mitotic recombination (Garcia-Bellido et al., 1973; Janning, 1978; Nicolas et al., 1996; Stern, 1936; Vincent et al., 1995).

Clonal analysis received a great boost with the introduction of recombinases and fluorescent proteins in the 1990s: recombinases made the task of inducing genetic rearrangements more efficient and easier to engineer (Golic and Lindquist, 1989; Orban et al., 1992), and fluorescent proteins provided versatile markers for visualising the cell clones. The most common method for marking cell clones, used until today, consists of expressing a recombinase (bacteriophage-derived Cre or yeast-derived Flp) in the cells of interest, in which it induces recombination among pairs of the respective target site (lox or FRT). These target pairs can be placed either in tandem, to excise the intervening DNA fragment containing a marker (flip-out), or on homologous chromosomes, to exchange chromosome arms carrying markers and/or mutations. Many elaborations of those two configurations have been used, varying in the number and complexity of genetic rearrangements (reviewed by Griffin et al., 2014; Richier and Salecker, 2014).

Today, clonal marking techniques are widely applied to visualise the morphology and behaviour of single cells within complex tissues, and to investigate lineage relationships, stem cell dynamics and tissue growth (e.g. Blanpain and Simons, 2013; Chen et al., 2016; Kuchen et al., 2012; Livet et al., 2007; Morante and Desplan, 2008; Snippert et al., 2010). The same tools have also been adapted to create genetic mosaics in which specific genetic changes are introduced in some cells of the body. Such genetic mosaics have been invaluable for discovering the cells in which a gene function is needed, for distinguishing the cell autonomous and non-autonomous functions of a gene, and for studying the effects of mutations that would be lethal in the entire organism (e.g. Hotta and Benzer, 1972; Lee and Luo, 1999; Rajewsky et al., 1996). By allowing us to juxtapose cells of different genotypes in a tissue, mosaics have led to the discovery of genes that underpin key cell-cell interactions during development and tissue homeostasis (e.g. Brook et al., 1996; Meyer et al., 2014).

Clonal analysis techniques require good control over the frequency of clone induction. If the frequency is too low, a very large number of individuals must be screened in order to find suitable cell clones to analyse and experiments become very laborious. On the other hand, if the frequency is too high, independent cell clones frequently merge with each other and information on the clonal relationships of cells is lost. In established model organisms, such as Drosophila and mice, the research community has developed and fine-tuned these techniques over decades, generating tools with clone induction frequencies that are well suited for the purpose. When establishing these tools in new experimental models, however, adjusting these frequencies can be laborious and frustrating. Here, we present the difficulties that we faced while establishing clonal analysis tools in the beetle Tribolium castaneum. We show that the Cre/lox system has such high activity in Tribolium that a large number of marked cell clones are generated even by very low (undetectable) levels of Cre expression, driven by the leaky activity of a heat-shock promoter.

To overcome this problem and to optimise the frequency of induction of marked clones, we developed an alternative method for generating marked clones, based on the excision of a marker cassette by the CRISPR gene editing system (Cong et al., 2013; Gilles and Averof, 2014; Jinek et al., 2012; Mali et al., 2013), rather than by Cre- or Flp-mediated recombination (Fig. 1). This approach, which we name Valcyrie (for Versatile and adjustable labelling of clones by random CRISPR-induced excision), offers the major advantage that the efficiency of clone induction can be adjusted by manipulating the degree of sequence complementarity between the CRISPR guide RNA (gRNA) and its targets. We demonstrate the effectiveness of this approach in Tribolium. In established model organisms, this tunable method can be readily applied on existing flip-out lines that have been established with constructs carrying lox, FRT or attP/B recombinase sites.

To establish the Cre/lox system for clonal analysis in T. castaneum, we generated stable transgenic lines bearing conventional Cre and lox flip-out constructs. The Cre-expressing line (Tc-hsp68:Cre) carries the coding sequence of the Cre recombinase that is driven by the heat-inducible promoter of the Tc-hsp68 gene (hereafter hsp68) (Schinko et al., 2012). The flip-out construct, pUb:lox-mYFP-lox-H2BmCherry, carries a flip-out cassette that consists of the coding sequence of membrane-localised Yellow Fluorescent Protein (mYFP) flanked by loxN sites (Livet et al., 2007), followed by a fusion of coding sequences for the fluorescent protein mCherry (Shaner et al., 2004) and histone H2B of Drosophila (H2B-mCherry); the flip-out construct is transcribed by the ubiquitous pUb promoter (see Materials and Methods). In the absence of a Cre transgene, embryos that carry the flip-out construct show ubiquitous expression of mYFP in late blastoderm and early germband stages (Fig. 2A).

By crossing these transgenic lines, we obtained embryos that carry both the heat-inducible Cre and the flip-out construct. In these embryos, a large number of cells lacked mYFP expression and instead expressed H2B-mCherry, indicating that the lox-mYFP-lox cassette was efficiently excised in the presence of Cre (Fig. 2B,C). We expected the excision to depend on the induction of Cre by a heat shock, but in fact we observed that embryos carry a large number of H2B-mCherry-expressing cells both with and without heat shock. These results suggest that sufficient Cre activity is present even in the absence of a heat shock, because of leaky activity of the hsp68 promoter. In the absence of heat shock, Cre mRNA was not detectable by in situ hybridisation, which suggests that very low levels of Cre are sufficient to mobilise the lox-mYFP-lox cassette. This Cre activity was present when Cre was provided either maternally or zygotically. Leaky activity of Cre has also been reported in similar experiments carried out in Drosophila (Frickenhaus et al., 2015).

In standard Cre/lox flip-out constructs, the expression of Cre drives recombination of paired lox sites, excising a marker gene and bringing a sequence encoding a second marker or gene of interest under the control of a ubiquitous promoter. In the Valcyrie approach, the same flip-out construct can be used, but excision is mediated by a pair of double-strand breaks targeted to the lox sites by CRISPR (Fig. 1). The induction of double-strand breaks is dependent on the expression of the Cas9 protein and a specific gRNA, which together constitute the CRISPR endonuclease (Hwang et al., 2013; Jinek et al., 2012). Once the CRISPR endonuclease has cleaved the two lox sites, the DNA repair mechanisms of the cell join the broken DNA ends; the cleaved sites may be repaired precisely, carry small mutations, or lack the entire fragment between the two targeted sites.

This design provides the opportunity to modulate the production of marked clones in different ways. First, the promoters used to express the Cas9 and the gRNA can influence when, where and at what frequency the marked clones are induced. Here, to test this approach, we used the hsp68 promoter to drive strong expression of Cas9 upon heat shock, and two constitutive RNA polymerase III promoters (Tribolium U6a or U6b; Gilles et al., 2015) to drive the expression of gRNAs.

Second, the level of CRISPR activity can be tuned down by manipulating the degree of sequence complementarity between the gRNA and its targets on the flip-out construct. The CRISPR system is known to tolerate mismatches, especially towards the 5′ end of the gRNA (Hsu et al., 2013; Jiang et al., 2013; Fu et al., 2013). With this in mind, we designed and tested two sets of gRNAs that were expected to target the lox sites of our flip-out construct with different efficiencies. First, we designed a pair of gRNAs to target each lox site as efficiently as possible (given the sequence constraints of the flip-out construct, see Materials and Methods): gRNA Lox.R fully matches one of the lox sites, and gRNA Lox.L matches the second lox site with a single nucleotide mismatch at the 5′ end of the gRNA (Fig. 3B). Second, we designed a single gRNA that is expected to target both lox sites with a lower efficiency: gRNA Lox.Uni carries two mismatches with each target, located within the first six nucleotides of the gRNA (Fig. 3B).

Besides generating pairs of double-strand breaks, leading to the excision of the lox-mYFP-lox cassette, we expect that, in some cells, CRISPR activity will generate point mutations (indels) in individual lox sites. The majority of these mutations will produce modified lox sequences that can no longer be targeted by CRISPR (Fig. 1). As these mutations accumulate, they should lead to a decline and eventually an arrest of clone-marking excisions, producing mosaic animals in which only a small number of cells have been marked, and the majority of cells remain unlabelled (carrying mutated lox sites). This would be beneficial for clonal analysis.

To facilitate the genetic crosses bringing the Cas9, gRNA and flip-out elements together, we cloned the Cas9 transgene and each set of gRNAs in the same vector (see Fig. 3A) and generated two independent transgenic lines for each construct, termed Valcyrie.LR #22 and #39 (expressing the Lox.L and Lox.R gRNAs) and Valcyrie.Uni #6 and #11 (expressing the Lox.Uni gRNA). The lines carry transgene insertions that segregate as single Mendelian loci. Valcyrie.LR #22, Valcyrie.Uni #6 and Valcyrie.Uni #11 are kept as viable homozygous lines. The insertion present in Valcyrie.LR #39 is homozygous lethal and it is therefore kept in heterozygous condition.

To test whether CRISPR-mediated excision is effective in generating marked clones, we crossed beetles from each of the Valcyrie.LR and Valcyrie.Uni lines to beetles carrying the flip-out construct pUb:lox-mYFP-lox-H2BmCherry (the same flip-out line that we previously tested for Cre-mediated excision). We collected mixed embryonic stages, from early to late germband stages, and subjected them to a heat shock (46°C for 10 min, see Schinko et al., 2012) to induce the expression of Cas9. We let these embryos develop for an additional 12 h at 32°C before fixation and screened them for H2B-mCherry expression. As a control, we screened embryos from identical crosses that were not heat shocked, but were collected and fixed as described above. Ten to 25 embryos from each cross were screened by confocal microscopy.

We detected H2B-mCherry-positive cells in embryos from all four crosses with variable frequency after heat shock (Figs 4 and 5, Table S1). A higher proportion of embryos with visible clones was recorded with the Valcyrie.LR #22 and #39 lines (>80% of embryos carrying the construct) than with the Valcyrie.Uni #6 and #11 lines (20-70% of embryos carrying the construct). This result is in line with our expectation that nucleotide mismatches of increasing severity in the gRNAs will generate flip-out clones at lower frequencies. We did not find any H2B-mCherry-positive cells in the controls for the Valcyrie.LR #22, Valcyrie.Uni #6 and Valcyrie.Uni #11 crosses, but we detected some marked cells in the control for Valcyrie.LR #39, indicating leaky expression of Cas9 in this line, in the absence of a heat shock (data not shown). These results indicate that most Valcyrie lines give us control over the induction of marked clones.

In addition to scoring the presence or absence of H2B-mCherry-expressing cells, we also counted how many distinct clusters of H2B-mCherry-positive cells could be found in each embryo (Fig. 4D, Table S1). We used the number of these clusters as a proxy of the number of independent cell clones marked in each embryo. As expected, embryos from the Valcyrie.LR lines have more clusters than embryos from the Valcyrie.Uni lines (Fig. 4D), which we attribute to the different efficiency of the two sets of gRNAs. However, there is also a considerable difference between the two lines carrying each construct, which is likely due to different levels of Cas9 and gRNAs produced from each transgene insertion site.

These results show that introducing mismatches between the gRNAs and the lox target sites is an effective strategy for reducing the frequency of marked clones in a predictable fashion.

Based on previous work (Schinko et al., 2012), we expected that the temperature and/or the duration of the heat shock could influence the expression level of Cas9, and therefore provide an additional means for tuning the frequency of clone labelling to desired levels. To investigate this possibility, we repeated the experiment described above with a milder heat shock, at 44°C instead of 46°C. As anticipated, we found that the number of marked cell clusters was greatly reduced with all four Valcyrie lines (Fig. 4D, Table S1B).

Further, we suspected that the number of marked clones might also depend on the incubation temperature after the heat shock. To probe this, we collected embryos from the Valcyrie.LR #22 line at 32°C and subjected them to a heat shock for 10 min at 46°C. Then we let one batch of these embryos develop at 32°C for 20 h, and another at 25°C for 40 h. Tribolium development is approximately twice as fast at 32°C compared with at 25°C, so these batches developed to comparable developmental stages. We found that the embryos kept at 32°C had more marked cell clusters than the embryos kept at 25°C (Fig. 4D, Table S1C). We suggest that, after the heat shock, CRISPR activity is higher or persists for longer at 32°C than 25°C.

In these experiments we found that we were able to generate marked clones in all regions of the Tribolium embryo, including the anterior head, gnathal, thoracic and abdominal segments. Clones were visible in the ectoderm and the mesoderm of the germband, in developing appendages and in the hindgut (Fig. 5), demonstrating that Valcyrie can be used to generate marked clones in most, if not all, cell populations of the Tribolium embryo.

We were particularly interested to determine whether we could induce flip-outs as early as in late blastoderm or early germband stages, because these are stages when important patterning events, such as establishing segmental boundaries and organ primordia, take place in Tribolium embryos. To address that issue, we subjected late blastoderm embryos (6-7 h after egg laying, at 32°C) to a heat shock, and determined the onset of H2B-mCherry expression. We could detect the earliest H2B-mCherry-positive nuclei in early germband stages (after 13 h at 25°C), showing that Valcyrie constructs can be used to mark clones during these stages. However, we did not detect any H2B-mCherry expression in embryos that were subjected to a heat shock during earlier stages (5-6 h after egg laying at 32°C, or younger). The timing of clone induction should correspond to the onset of heat-inducible expression (Schinko et al., 2012).

To molecularly characterise the excision events that take place in the flip-out construct, we extracted genomic DNA from a mix of embryos carrying flip-out clones, used PCR to amplify the region that lies between the paired lox sites, and sequenced several of these DNA fragments. We could detect deletions of variable length spanning that region (Fig. 3C), as is expected for sites that are repaired by non-homologous end joining. In two out of 11 sequenced clones, the deletions extended into the coding sequence of H2B-mCherry. In one other case, part of the pUb promoter was deleted. These results show that, whereas the majority of flip-outs generate marked clones, some Valcyrie excisions can also lead to the loss of both mYFP and H2B-mCherry marker expression.

We have presented a novel method for generating marked cell clones, which relies on CRISPR. Compared with conventional recombinase-mediated clonal marking, our method provides additional means to control the frequency of clonal labelling, by manipulating CRISPR activity; this can be tuned in a predictable fashion by manipulating the degree of sequence complementarity between the CRISPR gRNAs and their targets. The method is highly versatile; the flexible design of gRNAs allows it to be applied on any target construct, including existing constructs with paired lox, FRT or attP/B sites, but also on new sites that are introduced using transgenes or on sites found naturally in the genome (see below).

We demonstrated the Valcyrie approach in the beetle T. castaneum, an experimental model that is used for comparative studies of development and for studies on pest control. The genetic tools and resources that are available in this species have expanded significantly in recent years, to inducible gene expression, the GAL4/UAS system, a complete genome sequence, large RNAi and enhancer trapping screens, and CRISPR-mediated genome editing (Gilles et al., 2015; Richards et al., 2008; Schinko et al., 2012, 2010; Schmitt-Engel et al., 2015; Trauner et al., 2009). However, clonal marking tools, which would allow us to mark and to track individual cell clones, have been lacking. Our initial effort to establish clonal analysis using a conventional Cre/lox flip-out approach was hampered by the leaky activity of the Cre recombinase, which led to uncontrolled clonal labelling. The Valcyrie approach allows us to overcome this problem, giving us control over the timing and frequency of clonal labelling.

Using Valcyrie, we have shown that the frequency of excision events can be tuned in several ways: by designing gRNAs that have optimal or sub-optimal matches with their target sites (i.e. the Valcyrie.LR or the Valcyrie.Uni constructs), by choosing different transgene insertions, and by modulating the temperature of the heat shock and post-heat-shock treatments. Using the Valcyrie.Uni #6 and #11 lines, we have found conditions in which we have reliably generated embryos that carry a single or a few marked clones. Additional levels of control can be envisaged by employing different promoters to drive the expression of Cas9 and/or gRNAs.

We applied Valcyrie to lox flip-out constructs, but the same approach could be used in any experiment that requires excision of a DNA fragment. To perform genetic manipulations within the marked clones, the H2BmCherry marker could be co-expressed with effector transgenes using ribosomal skipping peptides (Szymczak-Workman et al., 2012; the P2A peptide is functional in Tribolium, see Fig. S1). In established model organisms it can be used on existing FRT or lox flip-out constructs, including constructs that are used for mosaic/clonal analysis of gene functions, to fine-tune the frequency of excision. In principle, CRISPR-mediated excision could also be extended to perform clonal analysis in the absence of flip-out transgenes, by targeting sequences already present in the genome. For example, target sites flanking a cis-regulatory element or an exon could be used to excise that element in specific cell populations (using an appropriate driver for Cas9) or in random cell clones. The major challenges in that case will be finding ways to mark the excision events and identifying CRISPR target sites that have no phenotype when individually mutated (e.g. targeting intergenic or intronic regions).

Our method of CRISPR-mediated clonal analysis may be of particular interest in experimental models in which conventional Cre/lox- and Flp/FRT-based clonal analysis tools are not yet available, but in which transgenesis and CRISPR have been established, such as Nematostella vectensis and Parhyale hawaiensis (Ikmi et al., 2014; Renfer and Technau, 2017; Pavlopoulos and Averof, 2005; Martin et al., 2016). The tunability of Valcyrie provides a straightforward way of adapting and optimising clonal analysis tools in diverse experimental settings and organisms.

The Tc-hsp68:Cre construct was generated by cloning the Cre coding sequence under the heat-inducible promoter of the hsp68 gene, including the endoHSE upstream region, the core promoter region (bhsp68) and the 3′ UTR of hsp68 (Schinko et al., 2012).

The pUb:lox-mYFP-lox-H2BmCherry construct was generated by cloning the lox-mYFP-lox-H2BmCherry flip-out cassette under the Tribolium pUb element, a 940 bp DNA fragment that includes the promoter, the 5′ UTR and the first intron of the Tribolium polyubiquitin gene (see below). Each coding sequence is followed by the SV40 transcriptional termination and polyadenylation signal.

The heat-inducible Cas9 transgene (Tc-hsp68:Cas9) was generated by cloning the Streptococcus pyogenes Cas9 coding sequence under the heat-inducible promoter of the hsp68 gene, including the endoHSE upstream region, the bhsp68 core promoter region and the 3′ UTR of hsp68 (Gilles et al., 2015; Schinko et al., 2012).

To generate the three gRNA transgenes U6b:Lox.Uni, U6b:Lox.R and U6a:Lox.L, we synthesised oligonucleotides that corresponded to each gRNA targeting sequence and cloned these into p(U6b-BsaI-gRNA) or p(U6a-BsaI-gRNA), as described in Gilles et al., 2015. For Lox.Uni we used oligonucleotides TTCGTTCGTAATAACTTCGTATA and AAACTATACGAAGTTATTACGAA, for Lox.R we used oligonucleotides TTCGATCGCAATAACTTCGTATA and AAACTATACGAAGTTATTGCGAT, and for Lox.L we used oligonucleotides AGTGTACGTAATAACTTCGTATA and AAACTATACGAAGTTATTACGTA.

To generate the Valcyrie.Uni and Valcyrie.LR constructs, we amplified each gRNA transgene using PCR and cloned them into the plasmid carrying the Tc-hsp68:Cas9 transgene, as illustrated in Fig. 3A.

To generate transgenic lines, these constructs were cloned into the piggyBac transposon vector pBac[3xP3-gTc'v] (Addgene plasmid #86446, deposited by Gregor Bucher). All constructs were made using a combination of conventional cloning and gene synthesis. Plasmids and annotated sequences are available at Addgene (www.addgene.org).

Stable transgenic lines were generated as described previously (Berghammer et al., 2009) in a Tribolium stock of vermilion^white background (Lorenzen et al., 2002). Transgenic lines carrying single copies of pBac[3xP3-gTc'v; Tc-hsp68:Cre], pBac[3xP3-gTc'v; pUb:lox-mYFP-lox-H2BmCherry], pBac[3xP3-gTc'v; Valcyrie.LR] (transgenic line #22) and pBac[3xP3-gTc'v; Valcyrie.Uni] (transgenic lines #6 and #11) were established and kept as homozygous lines. The insertion of pBac[3xP3-gTc'v; Valcyrie.LR] in Valcyrie.LR transgenic line #39 is homozygous lethal and was therefore kept in a heterozygous state.

To identify cis-regulatory elements that are capable of driving uniform, ubiquitous expression of transgenes in early embryos of Tribolium, we cloned and tested the following DNA fragments: (1) a 440 bp fragment of alpha tubulin, cloned by PCR using the primers TGGCCGGCCTGCAGTGAACGGTTATGATGG (including an FseI restriction site) and GGTATACTATACGAAGTTATGGTAGTTGAGTTTTACAAATTAC (including a 20 bp adaptor sequence to facilitate cloning); (2) a 708 bp fragment of RPS3, cloned by PCR using the primers ATGGCCGGCCTCAACGTATGTTGTCAAACC (including an FseI restriction site) and ATGGATCCGACGTTCTAAATGGAAAAGG (including a BamHI restriction site); (3) a 269 bp fragment of RPS18, cloned by PCR using the primers TTGGCCGGCCTGTCAGCGGGACATTGAC (including an FseI restriction site) and TTAGATCTGCAACCAATGCACAAACAAG (including a BglII restriction site); (4) a 939 bp fragment of polyubiquitin (pUb), cloned by PCR using the primers TTGGCCGGCCTTTTCTTTGTCCCAAATGACC (including an FseI restriction site) and TAAGATCTGCAACGACACAAAAAATTAC (including a BamHI restriction site).

These fragments were cloned upstream of the lox-mYFP-lox-H2BmCherry flip-out cassette and the resulting constructs were introduced in the pBac[3xP3-gTc'v] transgenesis vector (see above). The activities of the RPS3, RPS18 and pUb constructs were compared by observing mYFP fluorescence in Tribolium embryos 24 h after injection. RPS3 and RPS18 gave weak fluorescence compared with pUb, which gave very strong fluorescence. Stable transgenic lines were then generated for the pUb and alpha tubulin constructs. pUb line 153#13 gave the strongest uniform expression and was thus chosen for subsequent experiments. We note that all the Tribolium ‘ubiquitous’ promoters that we have tested (including the previously published EFA fragment; Sarrazin et al., 2012) show patchy expression in late germband stages. As there are no consistent patterns, we ascribe this to noisy zygotic expression.

To generate clones using the Cre/lox approach, we crossed males homozygous for the Tc-hsp68:Cre transgene (transgenic line CR02) to females homozygous for pUb:lox-mYFP-lox-H2BmCherry (transgenic line 153#13). We collected embryos from various stages and subjected them to a heat shock for 5 min at 44°C. Embryos were then allowed to develop overnight before fixation.

To generate clones using the Valcyrie approach, we crossed males carrying the Valcyrie.LR or Valcyrie.Uni constructs to females homozygous for pUb:lox-mYFP-lox-H2BmCherry (transgenic line 153#13). We collected embryos from various stages (early germband to late germband elongation) and subjected them to a heat shock for 10 min at 46°C. Unless otherwise stated, embryos were allowed to develop for an additional 12 h at 32°C before fixation. We screened the fixed embryos for native H2B-mCherry expression using a Zeiss LSM 780 confocal microscope. As a control, we screened 20 embryos from the same crosses that were collected and fixed as described above but were not subjected to a heat shock.

We thank Manon Peleszezak and Benjamin Boumard for their work to characterise the Cre/lox system in Tribolium; Agnès Vallier for technical help; Peter Kitzmann and Gregor Bucher for communicating unpublished data, plasmids and transgenic lines; and Vera Hunnekuhl-Terblanche, Maura Strigini, Jordi Casanova, Luke Hayden and Marco Grillo for helpful comments on the manuscript.