Zebrafish provide an in vivo vertebrate animal model system for functional studies of genes of unknown function. Recent advances in targeted mutagenesis using ZFNs and TALENs have overcome the limitations of previously available reverse genetics techniques in zebrafish that relied on random mutagenesis with chemical or retroviral methods. In the last couple of years, a plethora of cheaper and faster methods to design and assemble ZFNs and TALENs have been developed, thus making the technology accessible to all zebrafish laboratories. However, the design and acquisition of the appropriate nuclease is only the first step in using this technology to generate mutant fish. It would be of great benefit to develop general guidelines for successful application of ZFNs and TALENs to large-scale zebrafish knockout projects.
One of the critical factors in successful targeting of desired loci using ZFNs is the amount of mRNA injected per embryo. While lower mRNA dose may not cause any mutations at the target site, higher dose can lead to increased mutagenesis at the off-target sites in the genome 
. The effective mRNA dose is often determined empirically by performing evaluation of injected embryos for toxicity at multiple doses. In the published reports, the effective mRNA dose varied by more than 1000-fold, ranging from 5 pg to 7 ng 
. We observed high toxicity at doses higher than 500 pg and therefore, tested most ZFNs at 25–500 pg range. Based on our data, at least five-fold higher dose of mRNA is required for CoDA ZFNs when compared to CompoZr ZFNs. One noticeable difference between them is the specificity provided by the length of their recognition sequences. CoDA ZFNs are limited to 3 zinc-fingers on either side, giving a specificity of 9 bp as monomers. On the other hand, in most cases CompoZr-ZFNs consisted of 4–6 zinc-fingers on either side, thus recognizing 12–18 bp sequences as monomers. It is possible that ZFNs with shorter recognition sequences have increased chances of binding to other sites in the genome as monomers, thus requiring increased mRNA dose to induce mutations at the target site. Another possible explanation for these differences could be the use of improved obligate heterodimeric Fok1
variants (ELD/KKR) for CompoZr ZFNs as opposed to EL/KK Fok1
variants used for CoDA ZFNs. Doyon and colleagues 
demonstrated that ELD/KKR Fok1
variants worked better in inducing somatic lesions than the EL/KK variants when used with the same ZFPs. For a better comparison, expression vectors with identical Fok1
variants with improved activity, such as Sharkey 
, or ELD/KKR 
should be used with both CompoZr and CoDA ZFPs. Similarly, ZFNs targeting the same genomic loci should be assembled using CompoZr and CoDA platforms, with subsequent evaluation of their targeting efficiencies. This would rule out any differences in their performance due to the differences in the chromatin structure and accessibility of the target sites.
It is critical to evaluate the designed ZFNs by somatic lesion analysis before proceeding to growing the injected embryos to adulthood for founder screening. This is often done by surveying 8–10 injected embryos by PCR, followed by cloning and sequencing of large number of clones. Based on our results, sequencing of 48 clones is sufficient for highly efficient ZFNs (efficiency of >25%). Therefore, one can start with sequencing of 48 clones and sequence more if no mutant clones are identified. This decision can also be made based on the source of ZFNs used, as CoDA ZFNs have lower efficiency compared to CompoZr ZFNs.
Previous reports have compared the performance of CoDA ZFNs with ZFNs assembled through OPEN protocol and TALENs 
. Roughly half of the CoDA ZFNs pre-selected by their in vitro
activity at the target site using bacterial two- hybrid (B2H) assay were found to be active in zebrafish 
. We used two CoDA pairs per gene without screening with B2H assay for their activity. Both pairs for stat3
failed to introduce somatic mutations in injected embryos. Since one of the 2 ZFN pairs worked for cmet
, we did not extensively test the other pair for these two genes. Therefore, in our study two out of six CoDA ZFNs (~33%) were successful at introducing somatic mutations in zebrafish. This success rate is higher than the observed (18%) success rate and comparable to the estimated success rate (28%) for non-selected CoDA ZFNs 
. However, one must be cautious in interpreting the success rates of different studies as the threshold of detection methods varies between these studies.
A major concern in using ZFNs and TALENs is their potential for binding to other sites in the genome that may have some sequence similarity with the target site, leading to background mutations 
. Based on our data, for most ZFNs with >1% somatic lesion efficiency, it is relatively easy to identify at least two independent founders transmitting different in/dels leading to loss-of-function mutations. Therefore, we recommend generation of at least two mutant alleles from two independent founders for each gene for phenotypic analysis. To save time and resources, compound heterozygotes can be investigated instead of homozygotes for the mutations separately.
In conclusion, we have generated multiple mutant zebrafish lines for seven genes that would help in delineating their function during development and disease. In parallel, we have developed methods that use equipment and reagents readily available to most zebrafish laboratories for performing targeted mutagenesis of their desired genes. Our improved method for somatic analysis using colony PCR can be employed for quick evaluation of activity of a given pair of ZFNs, thus determining whether to proceed with growing the injected fish for founder screening or inject at lower or higher mRNA dose or design another ZFN pair. Our fluorescent PCR protocol for founder screening is high throughput, robust, and cost-effective as it eliminates the need for labeled primers for each target and allows for easy identification of the 1 bp in/dels by elimination of the stutter peaks that were observed by Foley and colleagues 
. Although we have only used ZFNs in our study, these protocols should also be applicable to TALENs as the subsequent steps are identical in both cases. Thus our improved methods can be applied to streamline the process of generating mutant zebrafish lines for large-scale knockout projects using ZFNs and TALENs.