Primer and interviews: Advances in targeted gene modification


  • Julie C. Kiefer

    Corresponding author
    1. Department of Neurobiology and Anatomy, University of Utah, Salt Lake City, Utah
    • Department of Neurobiology and Anatomy, 20 North 1900 East, 401 MREB, University of Utah, Salt Lake City, UT 84132
    Search for more papers by this author


Gene targeting in mice, first reported 25 years ago, has led to monumental advances in the understanding of basic biology and human disease. The ability to employ a similarly straightforward method for gene manipulation in other experimental organisms would make their already significant contributions all the more powerful. Here, we briefly outline the strengths and weaknesses of reverse genetics techniques in non-murine model organisms, ending with a more detailed description of two that promise to bring targeted gene modification to the masses: zinc finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs). Dana Caroll, a forefather of zinc finger technology, and Bo Zhang, among the first to introduce TALEN-targeted mutagenesis to zebrafish, discuss their experience with these techniques, and speculate about the future of the field. Developmental Dynamics 240:2688–2696, 2011. © 2011 Wiley Periodicals, Inc.


Random Mutagenesis Screens

A gene of interest has a tantalizing expression pattern suggestive of a pivotal role in a biological process. What's the best way to figure out what it does? The dawn of the molecular age and availability of complete genome sequences have made possible reverse genetics, the ability to glean a gene's function by examining phenotypes wrought by perturbing gene sequence or expression.

For genetically tractable organisms (e.g., Arabidopsis thaliana, Drosophila melanogaster, Caenorhabditis elegans, Danio rerio), the most obvious reverse genetics strategies are extensions of the forward genetics techniques—random insertional, deletion, and point mutagenesis—that established them as valuable model organisms in the first place. Insertional mutagenesis is a product of viral DNA or transposons that randomly integrate into the genome (Amsterdam et al., 2004; Hardy et al., 2010). Deletion mutations can be triggered by mutagens, or by exposing transposon insertional mutants to transposase, which sometimes cause transposons to excise imprecisely, taking adjacent genomic DNA along with it (Jansen et al., 1997). Identification of a hit within a gene of interest can usually be accomplished through PCR-based screens, or by gene or enhancer trap techniques in the case of transposon insertional mutagenesis (Hardy et al., 2010). Targeting induced local lesions in genomes (TILLING) is a high-throughput, sensitive screening method that identifies chemical-induced point mutations in a specific gene. Originally, the method utilized a single-stranded nuclease to cleave heteroduplexes formed between treated and untreated DNA at sites with single nucleotide mismatches (McCallum et al., 2000; Moens et al., 2008). Cleaved products were then separated on a gel and sequenced. But now that deep sequencing is affordable, it has become the preferred method for TILLING (Kurowska et al., 2011). Random mutagenesis-based reverse genetics has proven invaluable to the research community, producing large collections of heritable mutations that are often shared between labs.

The “random” descriptor of random mutagenesis hints that there is no guarantee that a gene of interest will be hit, or that a resulting mutation will be a null. This is partly because mutagens are actually somewhat biased. Retroviruses and transposons trend toward inserting into transcriptionally active genes and the 5′ end of genes, with resulting mutations generating hypomorphs as well as nulls (Nakai et al., 2005; Amsterdam and Hopkins, 2006). Chemical mutagens causing point mutations will be less likely to affect the function of genes with small coding regions or those in which the encoded protein can tolerate missense mutations. They also create a range of phenotypes, including gain of function and dominant negatives. These uncertainties, combined with the generation time of the organism used, means it may take anywhere from months to years to uncover and outcross a useful mutation in a gene of interest.

Antisense RNA Knockdown

RNA interference (RNAi), a targeted method for knocking down gene expression, first came onto the scene in the 1990s, when it was discovered that introduction of double-stranded RNA (dsRNA) into C. elegans triggers systemic and specific inhibition of gene expression (Fire et al., 1998). The technique was quickly modified for use in Drosophila and Arabidopsis. However, it was found to be ineffective in mammalian cells because it invokes interferon-mediated, non-specific gene inhibition (Sledz et al., 2003).

Understanding the molecular mechanisms leading to gene knockdown enabled researchers to find other ways to jump-start the RNAi pathway, an evolutionarily conserved immune response to viruses. Briefly, dsRNA or short hairpin RNA (shRNA) precursors are cleaved to ∼20 nucleotide-length small interfering RNAs (siRNAs), which, once unwound, bind and target endogenous complementary mRNA for transcriptional or post-transcriptional inhibition (Shan, 2010). Therefore, one can circumvent the mammalian interferon response by triggering the RNAi pathway with shRNA or siRNA instead of long dsRNA (Elbashir et al., 2001; Szulc et al., 2006). RNAi is commonly used in C. elegans, Drosophila, and cell culture because in these systems it is a relatively straightforward and fast way to achieve gene knockdown. However, no method is without its drawbacks. Effects on gene expression are usually not heritable, genes are sometimes incompletely silenced, and the extent of gene silencing is not necessarily reproducible (Alcazar et al., 2008). Further, siRNAs bind sequences with partial complementarity, leading to off-target effects (Ma et al., 2006).

The fact that RNAi is ineffective in two popular model organisms, Xenopus and zebrafish, led to the rise of an alternative knockdown strategy, morpholino antisense oligonucleotide (MO) technology (Nasevicius and Ekker, 2000). MOs are nucleic acids bound to morpholine rings and linked together by phosphorodiamidate groups. Upon delivery into cells, they stably duplex with RNA and are resistant to cleavage, preventing target gene transcription or translation (Eisen and Smith, 2008). When injected into early embryos, knockdown is achieved system-wide, but effects only last through larval development. Caged MOs and electroporation of MOs are newer techniques that allow for finer spatiotemporal control, including knockdown in adults (Falk et al., 2007; Shestopalov and Chen, 2011). Their ease of use has made MO technology the knockdown method of choice in Xenopus and zebrafish, and has won converts among scientists utilizing other organisms such as chick and sea urchin. However, non-specific effects that call MO results into question are general toxicity and off-target effects, including neural apoptosis mediated by a p53-dependent pathway (Bedell et al., 2011).

Non-specific effects triggered by RNAi and MOs warrant carefully controlled experiments (Hardy et al., 2010; Bedell et al., 2011). First, appropriate steps should be taken to ensure that the target of choice is knocked down at the mRNA or protein level, and that off-targets with a closely related sequence that might hybridize to MOs or siRNAs are not knocked down. Second, results should be validated with a second set of MOs or siRNAs directed toward the same gene. Third, phenotypes should be rescuable by co-delivery of target RNA engineered with silent mutations that render it immune to RNAi or MO knockdown. Comparison with a null mutant can also be valuable, although phenotypes may differ if RNAi/MO knockdown is incomplete, or knockdown blocks maternal gene activity that is unaffected by zygotic mutants.

The extensive controls that should accompany RNAi or MO gene knockdown experiments clearly indicate that there is room for improvement. What's more, neither random mutagenesis, RNAi, nor MOs achieve the holy grail that is homologous recombination (HR). Both of these objectives could be met using a method that alters genomic DNA directly. And so targeted nucleases were created.


Zinc Finger Nucleases

ZFNs are manmade “DNA scissors” that induce double strand breaks (DSBs) at a desired sequence, providing a jumping off point for further genetic modification (Smith et al., 2000). They are engineered by combining a series of Cys2His2 ZF DNA recognition domains with the FokI restriction endonuclease catalytic domain (Kim et al., 1996). ZF proteins are a class of eukaryotic transcription factors that primarily binds 3 bp of DNA per finger; binding site specificity depends upon key residues within the ∼30aa ZF (Pavletich and Pabo, 1991). The FokI cleavage domain can render DSBs at any sequence, but must dimerize to be functional (Smith et al., 2000). Therefore, one can create scissors that cut within a unique genomic sequence of 18–24 nucleotides by designing a dimerizing pair of ZFNs, each containing a set of three to four strategically chosen ZFs joined to a FokI catalytic domain (up to six fingers per monomer have been used) (Bibikova et al., 2002) (Fig. 1).

Figure 1.

Engineered zinc finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs). A: Custom-engineered ZFNs contain the catalytic domain of FokI endonuclease (grey shape) fused to an array of ZF DNA binding modules (circles), which each primarily recognize a three-nucleotide sequence. Two ZFN arrays function as obligate heterodimers, catalyzing double strand breaks (DSBs) at a chosen unique genomic sequence. B: Custom-engineered TALENs work in a manner identical to ZFNs except that the TALE DNA-binding modules (ovals) each recognize a single nucleotide.

When ZFNs are introduced into cells, the DSB acts as a beacon that evokes non-homologous end-joining (NHEJ) activity. The evolutionarily conserved repair machinery imprecisely rejoins the severed ends, incorporating small insertions or deletions (indels) that generally cause frameshift knockout mutations (Bibikova et al., 2002). When cleavage and repair are achieved in the germ line, the mutations are heritable. Heritable, ZFN-mediated site-specific mutagenesis has proven to be effective in over one dozen organisms including Arabidopsis, Drosophila, zebrafish, and rat, where frequency of target modification can reach 10% or higher (Carroll, 2011). The technique also works robustly in embryonic stem (ES) and induced pluripotent (iPS) cells without perturbing stem or differentiation potential (Hockemeyer et al., 2011). The success of ZFNs has led to testing in clinical trials, including treating HIV/AIDS by using it to knock out the CCR5 co-receptor required for HIV-1 infection (Perez et al., 2008).

The beauty of targeted nucleases is that they can also be used to achieve HR, significantly enriching their experimental potential (Bibikova et al., 2003). Gene editing, gene replacement, knock-in, or knock-out is triggered by co-delivering mRNA encoding ZFNs with donor DNA that is partially homologous to a genomic target but outfitted with appropriate modifications. Yet compared to ZFN-induced NHEJ knockouts, ZFN-induced HR has remained elusive in most experimental systems. One problem is that when ZFNs induce DSBs, NHEJ overwhelms HR machinery, making HR relatively rare. Dana Carroll's group increased HR efficiency 10-fold in Drosophila by injecting into mutants for Ligase IV, a component of the canonical NHEJ pathway (Beumer et al., 2008). It is a clever trick, but intractable in organisms where such mutants are not readily available. A second problem that hinders HR is a small probability of mosaic P0 animals transmitting this rare event to the germ line, which comprises a small proportion of the organism's cells. These obstacles can be compounded by other species-specific barriers. For example, ZFNs are inhibited in the C. elegans germ line, which harbors mechanisms that suppress transgene expression (Morton et al., 2006). Increasing ZFN efficiency will overcome some of these challenges, but others will require more specialized approaches.

There are other notable limitations to ZFN synthesis and performance that have prevented its widespread use (Carroll, 2011). Due to inherent non-specificity in sequence binding, ZFs bind off-target sites, and there are many sequences that ZFNs simply cannot recognize. In addition, binding affinity is context dependent, and may be affected by epigenetic modifications and chromatin packing. These issues are mitigated by pre-screening for binding-site specificity using phage display, one-hybrid, two-hybrid, context-dependent assembly (CoDa), or other strategies, but the process is time consuming and labor intensive. The limitations of ZFNs has led to a search for better alternatives.

Transcription Activator-Like Effector Nucleases

A new class of engineered nucleases, transcription activator-like effector nucleases (TALENs), was first reported in mid 2010 (Christian et al., 2010). Indigenous to Xanthomonas plant pathogens, TAL effectors (TALEs) bind effector-specific DNA sequences and transcriptionally activate plant gene expression. TALEs harbor ∼34 amino acid repeat domains that each bind one target base in the host genome with base specificity determined by two variable amino acids, the repeat variable di-residue (RVD). Similar to ZFNs, TALENs are made by fusing the catalytic domain of FokI to TALE repeat modules that are custom combined to target a unique stretch of genomic nucleotide sequence. The most significant advantage they have over ZFNs is that the rules governing recognition site specificity are more straightforward. There are fewer restrictions to the combination of bases recognized by TALENs, and recognition site binding is more predictable.

In 2011 alone, TALENs have successfully rendered site-specific mutations, some heritable, in yeast, ESCs, iPSCs, C. elegans, rats, plants, and zebrafish. Compared to ZFNs, efficiency and off-target effects occur at similar, or slightly improved, rates (Hockemeyer et al., 2011). Because TALEN-mediated HR faces the same challenges that ZFN-mediated HR does, the former has only been reported in cultured cells and yeast. Perhaps the most troublesome downside, however, is that the reasons behind failure of some custom TALENs remain unknown. Speculations include unknown effects of repeat composition on protein stability, and uncharacterized interactions between repeat domains that may affect DNA binding (Cermak et al., 2011). The questions raised speak to the fact that there is still much to learn about TALENs. Nevertheless, the technology's rapid achievements, plus its solid footing grounded in years of ZFN research, means that TALENs look promising.


A pioneer of zinc finger nuclease technology, Dana Carroll, PhD, discusses the history of how the synthetic clippers came into being, and their future prospects. Bo Zhang, PhD, one of the first investigators to achieve heritable TALEN-mediated gene editing in vertebrates, relates technical hurdles that have yet to be overcome, and compares TALENs with other reverse genetics technologies (Fig. 2).

Figure 2.

The experts: (L) Dana Carroll, PhD, Professor of Biochemistry, University of Utah, (R) Bo Zhang, PhD, Professor, College of Life Sciences, Peking University, Beijing, P.R. China.

Dana Carroll Interview

Developmental Dynamics: How did you start working on zinc finger nucleases?

Dana Carroll: I'd been working on DNA repair and recombination methods for a long time. In my lab, we were working on reagents that we thought might improve the efficiency of gene targeting and we were already working with reagents that had a DNA recognition domain and a DNA damaging domain. We happened to see a paper published from Chandrasegaran's lab that described the zinc finger nucleases initially. He called them chimeric restriction enzymes; we renamed them zinc finger nucleases, but as soon as I saw that paper, I called Chandradegaran, whom I didn't know, and proposed a collaboration.

We didn't know when we started that these would actually work as well as they have, but we did know that zinc fingers bound DNA in a very modular fashion and even at that time, 15 years ago, there were fingers that had been identified that would bind a number of different DNA triplets. One of the things we discovered by working on the proteins with Chandrasegaran was that the cleavage domain has to dimerize in order to cut DNA. The way to enforce dimerization is to make two sets of zinc fingers, each one bound to a monomer of a cleavage domain. When both bind to the DNA target, the cleavage domain is at a high local concentration and becomes a nuclease. We also didn't know whether this bacterial enzyme would cut DNA that was wrapped up in chromatin. The first experiment we did involved injecting zinc finger nucleases into Xenopus oocytes after we injected a candidate target DNA that got wrapped up in chromatin. The question was, will they actually cut a target that was wrapped up in chromatin? The answer was a resounding “Yes!” That's basically how we got started.

Dev Dyn: Were labs pretty open to collaborating on this work?

DC: We also got some help from Carlos Barbas's lab in making new zinc finger combinations. Once we had shown that the ZFNs would work in a real genomic target in a real organism by attacking a gene in Drosophila, people working on other organisms became quite interested. So, we collaborated with quite a number of different labs to try to apply the ZFNs to different organisms.

Two companies were started based on zinc finger technology. One was Sangamo Biosciences, which now holds the intellectual property. The other was a company in England called Gendaq, which was started by Aaron Klug and Yen Choo. Eventually, Sangamo bought Gendaq and now everything's in one set of hands. Both those companies started with the idea of making synthetic transcription factors using novel zinc finger sets. They both set out to try to elaborate recognition specificity of zinc finger combinations, Gendaq even more enthusiastically than Sangamo. Neither one of them was interested at the onset in the nucleases but once we had shown that these things work, Sangamo got quite interested.

Although they still hold proprietary interest in materials related to zinc finger design, I think you have to give them a lot of credit. They publish the sequences and activity of all the zinc finger nucleases that they make, and they have collaborated with a lot of academic labs to apply zinc finger nucleases to different situations. They also drive the technology toward real-world applications, and have clinical trials going on with two different human genomic targets. They have done a lot that has really advanced the field.

Dev Dyn: What is your lab focusing on now?

DC: We've been working on defining what pathways are involved in the ZFN-induced mutagenesis and gene replacement. We've been working on aspects of the donor DNA for homologous recombination, how much homology is involved, what structure of the donors is effective, and how broad are the conversion tracts. What that means is how much donor sequence actually gets incorporated into the target during any gene targeting event. This is very important because if somebody working on whatever organism has a pair of zinc finger nucleases that will cut the target effectively, they want to know over how broad a range can I introduce sequence changes into that target using this one entry site? In Drosophila, those conversion tracts are very broad, several thousand base pairs on each side of the cut. In other cells and organisms, the tracts are much narrower, so one of the things that I would like to know is what determines the length of the conversion tract and is there a way we can manipulate that. The other thing we are working on right now is off-target cleavage by the ZFNs. This is a critical issue, particularly for any human gene therapy trials. We don't want to cure somebody's disease and create another at the same time. We are approaching this in a very general sort of way. Basically, we are trying to capture all the ends that any set of zinc fingers makes inside a real cell. We are also working on the TALENs.

Dev Dyn: What do you think are the biggest remaining challenges?

DC: Zinc finger design is still a challenge. If you're working on your favorite organism and find a gene you want to target, how do you design the zinc finger sets with the highest probability of actually successfully cutting your target, i.e., choose a target within the gene, and then actually design the zinc finger set? Sangamo has the best technology for doing that and it has been licensed to Sigma-Aldrich. Sigma is charging $25,000 for a validated pair. The good news is that they do all the hard work and they won't sell you a pair of nucleases unless they have demonstrated that they work in cells.

The non-commercial methods for designing zinc finger sets are a little more hit-and-miss and people are still working on that. The TALENs have the promise of being even more designable because each module binds one base pair of DNA instead of three, and the TALEN specificity seems to be less influenced by neighboring DNA sequences. We have much less experience with the TALENs so far, so I don't know that we have hit all the bumps in the road yet but they are looking very good.

The third challenge is off-target cleavage. How do we monitor and minimize that? Sangamo has done some very nice work in minimizing off-target cleavage but it can still be an issue. The TALENs seem to be better at being restricted to their designed target but again experience is pretty limited.

The fourth challenge for any new application is delivery. How do you deliver the ZFNs and a donor DNA if you're trying to actually do a gene replacement in a new organism? It's not just getting the ZFNs expressed, which can be a challenge by itself, but also how do you get the desired event to occur?

I'll give you an example. We've been working on nematodes for about 10 years. We published a paper in 2006 showing that we could get high levels of ZFN cleavage, both on synthetic target and on natural genomic targets in the worms. One problem: we could never get germ line cleavage. So that's a delivery issue. How do you deliver ZFNs so that they are expressed in the germ line, so that you make transmissible genomic alterations? There has been a paper published recently that has demonstrated this using messenger RNA injection. I have also heard that people are having trouble reproducing it, not because the paper was incorrect, but just because RNA injections into worms was the way RNAi was discovered. So you are fighting against the worm's natural tendency to shut down translation in germ lines. The approaches are limited by the biology of the organism.

Dev Dyn: You have achieved heritable homologous recombination with ZFNs in Drosophila, but that method hasn't been adopted by the community.

DC: No, it hasn't. I think a lot of it has to do with the uncertainty of ZFN designs. People are willing to take a shot at it if the probability of success is high but not so willing to devote time and energy to it if the probability is low or uncertain. It may be because the genetic tools for Drosophila are already pretty elaborate and powerful. What has been true is that people working in other organisms have adopted the technology with some success. These include zebrafish, sea urchin, frog, mouse, rat, and Platynereis, a marine worm that is manipulatable and in a phylogenetically distant branch from some of the other experimental organisms.

Dev Dyn: Do you think that TALENs might overcome many of the problems that ZFNs have?

DC: Yes, I do. We've started to work on them in collaboration with Dan Voytas' group in Minnesota. I think they look very promising, but it's early days yet. One of the issues is that the constructs for the TALENs are bigger than the ones we use for ZFNs, largely because each module of 34 amino acids recognizes only one base pair in the TALENs, whereas each ZF module of 30 amino acids recognizes three. So you are 3× bigger right at the start and it turns out that you need a little bit more of the natural TAL effector protein outside of the simple recognition module in order to have effective activity. I don't think anybody knows whether that is due to a folding problem or a stability problem or just what.

Dev Dyn: Do you think that TALENs might make ZFNs obsolete?

DC: Off the top of my head, I can't think of any particular application in which the TALENs wouldn't do just as well as ZFNs. Because the ZFN technology is ahead, there are ZFNs for particular targets that people probably wouldn't want to give up on. If you've got a pair that's working, and it's past the clinical trial hurdles, people might just want to continue using those since all the preliminary work has been done.

Dev Dyn: How would you compare nuclease-mediated gene targeting to the traditional ES cell gene targeting?

DC: Looking from one perspective, all we've done is to improve the efficiency of the Capecchi technology. The ZFN break-induced events are somewhere between 1,000- and 1,000,000-fold more efficient and you don't have to use selection to find the products. On the other hand, you do have to make the cleavage reagent, which can be challenging.

Dev Dyn: Where do you see the field going?

DC: One thing I think that we can definitely see happening is that these cleavage reagents, whether ZFNs or TALENs, will be used much more extensively in some of these out-of-the-mainstream model organisms. It's also pretty clear that some of these commercial applications are going to take off. Dow AgroSciences has established a division devoted to gene targeting using ZFN technology and they've licensed it at Sangamo. The thinking behind using this technology for crop improvement is that they will be targeting endogenous genes to improve drought tolerance or herbicide resistance; there will be less of a public objection if no exogenous genes are being put in.

Bo Zhang Interview

Developmental Dynamics: What is your lab's research focus?

Bo Zhang: Zebrafish (Danio rerio) is an excellent model animal to study vertebrate development with genetic approaches. Using zebrafish as a major animal model, we are interested in discovering genetic and developmental mechanisms of vertebrate embryogenesis and organogenesis, particularly those that are involved in hematopoiesis, cardiovascular and pancreas development, with a focus on how cell proliferation and differentiation are coordinated. To achieve this goal, part of our research interest has become developing new genetic tools.

Since 2005, we have conducted two genome-wide, large-scale genetic screens to both identify genes with tissue-specific expression patterns and to establish a fish library with genomic mutations. (1) Using GFP or RFP as the reporter genes, we performed a Tol2 transposon-mediated large-scale enhancer trap screen in zebrafish. A total of 1,670 individual F1 transgenic fish lines were isolated, including 30 pancreatic and more than 40 hematopoietic and/or cardiovascular-specific transgenic lines. One hundred eighteen insertion sites have been identified so far, most of which hit novel genes. (2) We have established a highly efficient zebrafish mutagenesis platform based on pseudo-typed retrovirus-mediated insertion, and are aiming at generating a mutant library via random insertion of this retrovirus. So far, we have obtained about 600 insertions that reside in genes, most of which are functionally unknown. These projects have been in close collaboration with Prof. Shuo Lin in the Department of Molecular, Cell and Developmental Biology at UCLA, and Prof. Shawn Burgess at the National Human Genome Research Institute (NHGRI) at NIH.

To develop new and better methods for genomic mutagenesis, recently we have established a method called “Unit Assembly” to assemble TALE nuclease (TALEN), with which we have efficiently generated heritable gene targeting in zebrafish (Huang et al., 2011). So far, seven different genes have been successfully targeted via TALENs in our lab, where more than half of the TALEN pairs showed more than 50% mutagenesis activity, with three pairs of them leading to nearly 100% target site disruption after the corresponding mRNA pairs were injected into the one-cell-stage zebrafish embryos.

The pancreas consists of endocrine and exocrine compartments both of which exert important physiological functions. The development of the pancreas is a dynamic process of cell proliferation and differentiation, which is controlled by extrinsic signals from the surrounding tissues and intrinsic transcriptional programs. Pancreatic cancer is one of the most poorly understood diseases. The majority of pancreatic cancers originate from uncontrolled expansion of exocrine pancreas. Understanding the mechanisms of cell proliferation during normal pancreas development is important for elucidating how pancreatic carcinogenesis occurs. It is known that both endocrine cells and exocrine cells derive from the same population of progenitor cells. However, how the progenitor cells adopt different cell fates in response to extrinsic signals as well as how these cells undergo rapid lineage-specific proliferation during embryogenesis is not fully understood. The 30 pancreas-specific transgenic fish lines from our previous Tol2 enhancer trap screen will give us a good opportunity to address these questions.

Hematopoiesis and blood vessel formation are highly conserved among vertebrates. It has been shown that blood cells and blood vessel cells, which are both derived from ventral mesoderm, share common ancestors called hemangioblasts. These hemangioblasts give rise to hematopoietic stem cells and angioblasts, which, through complicated cell proliferation and differentiation processes, will further differentiate into various types of mature blood cells and blood vessels. Recently, angioblasts and endothelial cells have been shown to be able to give rise to definitive hematopoietic stem cells, which makes the blood cells and blood vessels even more closely related to each other, and thus the determination and differentiation of these cell populations become more complicated than previously thought. Zebrafish is an ideal model for the analysis of hematopoiesis and cardiovascular development due to its rapid external embryonic development and transparency. These processes can be visualized in live embryos using transgenic fluorescent protein reporters and embryos can survive for several days without circulating blood cells. We aim to study how cell proliferation and differentiation are involved in hematopoiesis and cardiovascular development in a coupled fashion, using the hematopoietic- and/or cardiovascular-specific GFP trap lines and mutant fish that were obtained from our screens.

Dev Dyn: What initially triggered your interest in zebrafish development?

BZ: To be honest, I didn't know much about zebrafish until 2003, shortly after I came back to China from Switzerland where I had my postdoc training. Once I started thinking about switching my study from in vitro culture cells to an in vivo organism, I became interested in zebrafish. I was first attracted by its robust vitality and natural beauty, as well as how inexpensive it is to raise them in the lab and then, more importantly, the advantages as a vertebrate animal model to study embryonic development (transparent embryos, rapid ex vivo development, etc.) with genetic manipulations (large amount offspring, short generation time, etc).

Dev Dyn: Why did you pursue TALENs?

BZ: This is due to both frustration and inspiration. We had long been annoyed by the fact that there was not an easy method to target a gene of interest efficiently in fish. Since we had fished out quite a few interesting and novel genes through our large-scale enhancer trap screen, which were waiting to be functionally studied, we thought that a good reverse genetics approach was desperately urgent for us, especially when the morpholino (MO) knockdown experiments were inconclusive. In fact, before getting in touch with TALENs, we had been struggling with gene targeting via ZFNs for about two years, and had put a lot of effort into improving the existing methods for screening and evaluation of efficient ZFNs, which proved to be quite difficult. We realized that a more specific DNA-binding domain should be the key to improving efficiency. Luckily, at the end of last year, Miss Weiye Wang, one of the undergraduate students from our college who was doing an undergraduate research project in our lab, noticed and recommended to us a paper published online in Nature Biotechnology (Miller et al., 2011) showing that TALENs could be engineered to disrupt desired target genes in human culture cells. We were immediately excited about the possibility to introduce this new technique into zebrafish. Thanks to our experience with ZFNs, we were able to quickly adopt this new method to our system.

Dev Dyn: What papers have most impacted your research with TALENs?

BZ: Besides the 2011 paper by Miller et al., these two papers (Boch et al., 2009; Moscou and Bogdanove, 2009) described the discovery of the simple rules for TAL effectors to recognize and bind their target DNA, which opened up a new possibility for gene targeting.

Dev Dyn: Can you speculate on what the barriers are to achieving TALEN-mediated homologous recombination (HR) in zebrafish?

BZ: Regarding genetic toolbox collections, HR is almost the last and the biggest technical difficulty to be surmounted in zebrafish. In my opinion, the deciding/limiting factors seem to be the efficiency of HR as well as the balance between NHEJ and HR, rather than the efficiency of TALEN itself in zebrafish. Unfortunately, people have limited knowledge about the mechanism(s) of HR in zebrafish although the machinery is there, ready for service (Hagmann et al., 1998; Cui et al., 2003; Fan et al., 2006; Xie et al., 2007). It has been shown that NHEJ predominates over HR in zebrafish early embryos (Hagmann et al., 1998). Increasing the HR efficiency and controlling/manipulating the DNA double strand break repair system to preferentially activate the HR machinery could be a quick solution to this problem.

Dev Dyn: Are you attempting to get HR working in zebrafish?

BZ: Yes, we are. We could detect HR events in our preliminary experiments, though the efficiency was far below NHEJ. I hope that I can tell you more in the near future.

Dev Dyn: What other aspects of TALEN technology still need to be resolved or improved?

BZ: There are still many problems about TALE and TALEN waiting to be resolved. The most puzzling observation was that different TALEN pairs exhibited distinct mutagenesis activities towards their targets, ranging from zero to nearly 100% (our unpublished data). Unfortunately, it seems to be difficult to predict how a TALEN pair would work regarding its activity at this moment. This variation does not seem to be related to the differences between species since people observed both high and low efficiencies in most species tested so far, including humans (culture cells), yeast, and zebrafish. There is another level of variation of TALEN activities, namely, there also exist differences among different culture cells from the same species. ZFN has a similar problem in this respect, which means that this difference could be rather due to the cellular differences between different cell lines than due to the differences in TALEN or ZFN activity themselves.

I think one of the keys to this important problem is to solve the crystal structure of the TALE complex with its target DNA, to further understand the structural basis and mechanism(s) for TALE to recognize and bind to its DNA target. I believe the unique feature of TALE DNA-binding domain must have caught the attention of many structural biologists and they are trying to crystallize the protein and DNA complex. In fact, a colleague of mine has been working on the crystal structures of TALE and DNA complexes, and hopefully the 3-D structure of TALE will soon be exposed and we can learn more on how to use this wonderful tool (personal communication).

Other issues regarding the dissection/improvement of TALEN activity include but are not limited to: (1) Why position 0 in the TALE target site should be nucleotide T? (2) Why the last repeat in TALE should be a half unit? (3) Is there any preference in the nucleotide sequence for TALE to bind its target efficiently? (4) Is there a certain length and number of TALE repeat units that is optimal for efficient targeting? (5) Are there any context-dependent effects among TALE repeat units? (6) Are there RVDs (repeat variable di-residues) in TALE repeat units that are better than the naturally occurring ones in terms of target-binding activity and specificity? (7) Can we make each repeat unit shorter, i.e., is there an essential and/or optimal sequence or size/length for the repeat unit? (8) Does chromatin structure or DNA double-strand repair efficiency influence TALEN activity? (9) Could TALENs confer toxicity to its target cells or organisms? (10) Will TALENs induce immune responses in its target cells or organisms?

In addition to its own activity, the successful application of TALEN technology also relies on other factors, e.g., how efficient the TALEN pairs could be introduced into the cells or organisms, either as plasmids (expression vectors) or as mRNAs? How error-prone the cells or organisms are regarding their DNA repair system/machinery? How efficient the germ line transmission is? Also, one has to be cautious about the possible outcome after introducing mutations in the target genes. For example, if the target gene is haploinsufficient or a deleterious dominant allele is generated, one might not be able to get living cells or individual organisms that would carry the mutation(s), preventing further analysis.

Dev Dyn: How do you plan to use TALENs in your lab? Will they replace morpholinos?

BZ: We are trying to knock out all of the genes that are of our interest with this method. Meanwhile, we are interested in trying HR and even conditional mutagenesis using TALEN techniques as well. In our lab, TALEN will become a routine method for functional study of candidate genes. It will largely, but not completely, replace the use of MOs, though. MOs will still retain its own advantages for our study since it is faster and easier to perform than TALEN targeting. For example, it is a good and quick method to confirm the result (phenotype) from TALEN knock outs. Alternatively, MOs could be used in a pre-screening step to evaluate candidate genes as worthy for TALEN targeting. In addition, MOs are especially advantageous for the analysis of maternal influence of candidate genes since it can be designed to inhibit the translation of both maternal and zygotic mRNAs at the same time, while TALEN targeting needs one more generation to remove maternal effect. For the transient effect, the dosage-dependent knockdown effect, the localized (in certain group of cells or tissues) inhibition effect, the flexibilities in the applications of MOs are also advantageous over the TALEN technique under certain circumstances.

Dev Dyn: Are there reasons why the mouse community might choose TALEN-mediated HR over traditional ES cell-mediated HR?

BZ: In my opinion, there are certain advantages to using the TALEN technique for HR-based gene targeting over traditional ES (embryonic stem) cell–mediated HR in mice. TALEN targeting is quick, simple, and easy. It can at least bypass the tedious work of making big constructs, and culturing naughty ES cells. If the activities of TALENs are high enough, one can even target two (or more) genes/sites at the same time, or create large chromosomal fragment deletions between the two target sites. In addition, there also exist possibilities for two alleles to be disrupted at the same time with TALEN pairs with high efficiency. In fact, we have started collaborating with a professor in the United States on targeting mouse genes, though he already has a lot of experience and good success with the ES cell–based HR technique.

Dev Dyn: Do you predict that within five years TALENs will be used widely? How else do you think the field might change?

BZ: Yes. Actually, I would predict that the TALEN technique will be widely picked up in a short time because this method is really easy to establish. Any lab that is able to do molecular cloning is qualified to use this method. Another reason is its potential for wide usage; theoretically, it is applicable to any cell and any species as long as it is possible to get exogenous nucleic acids into the cells/organisms. An interesting thing I have experienced thus far is that it seems that the most excited researchers besides our fish colleagues are those who are working with human culture cells. In our college, these people were the first ones to jump on our TALEN technique. Our method has also been picked up by other labs, even mouse and Drosophila groups, whom I thought would be the least interested because they already have mature and effective gene-targeting technologies.

It is still early to say what the impact and range of applications of TALEN technology will be. But it is not difficult to imagine that all the strategies that have been adapted to ZFNs should now theoretically also be applicable to TALENs, including the creation of large chromosomal deletions, attempts to pursue gene therapy and gene corrections in clinic trials, and the potential to improve characteristics of domestic animals through genetic modifications/manipulations. In addition, the TALEN technique has also opened a new way to dissect the mechanisms of DNA repair and HR, very fundamental cellular functions, and other related activities.

All the other potential applications are based on the very basic characteristics of TALENs to generate DNA double-strand breaks, and thus the fundamental application of this technology is for site-specific mutagenesis. Knocking out a gene has never been so simple; one can easily make a knockout animal through easy transgenic approaches. The power of simplified molecular cloning plus transgenic techniques mean that TALENs hold promise for achieving gene targeting in potentially any species.


Many thanks to Dana Carroll and Bo Zhang for generously sharing their expertise, insights, and time, and to Dana Carroll and Richard Dorsky for their comments on the manuscript. I am also grateful to Courtney Montgomery for her expert assistance as transcriptionist. Bo Zhang thanks Professor Renjie Jiao and Professor Xiaodong Su for helpful discussions. Interviews were performed and edited by J.C.K. The author regrets that due to space constraints she was unable to reference many excellent papers.