The Serine-Threonine Protein Kinase PAK4 Is Dispensable in Zebrafish: Identification of a Morpholino-Generated Pseudophenotype

TALEN-based inactivation of the zebrafish pak4 gene resulted in embryos and adult fish that appear normal and fertile. This is in contrast to our previously published studies which were based on the use of antisense morpholino oligonucleotides (MOs). We have excluded potential explanations such as gene duplication, alternate splicing, cryptic initiation of translation, and translation-independent RNA function. Our conclusion is that pak4 is dispensable in zebrafish, and that even when corroborated by robust controls, such as RNA rescue, MOs may elicit misleading pseudophenotypes that do not correspond to results obtained by genetic mutations, and should thus be used with caution.


Introduction
The optical clarity, rapid, in vitro embryonic development and relatively short generation time of zebrafish (Danio rerio) have led to the extensive use of this species for the characterization of gene function in vertebrate development. Classical genetic screens of mutagenized zebrafish have been highly successful in the identification of genes required for primary embryonic axis formation and numerous aspects of tissue and tissue differentiation [1,2], yielding important insights with broad significance for developmental biology and human health. However, the majority of gene function studies in zebrafish have been based on the use of antisense oligonucleotides with degradation-resistant morpholino phosphorodiamidate backbones (MOs). These have been shown to interact specifically with complementary RNA targets, resulting in inhibition of translation, RNA splicing, and other steps in posttranscriptional gene regulation (e.g Lim et al., 2012 [3]). However, it has become evident that in addition to specific inhibition of the cognate target genes, MOs can also have side effects. Ekker and Larson [4] reported that 15-20% of tested MOs show non-specific toxicity on developing embryos, with the most common artifact being induction of apoptosis in one or more tissues, frequently but not exclusively in the brain (reviewed by Eisen and Smith, 2008 [5]). Ekker and colleagues showed this was due at least in part to activated expression of an N-terminal truncated isoform of the transcription factor p53, and that the MO-induced apoptosis artifact could be suppressed by adding an antisense p53 MO to the injection cocktail [6]. In addition to cell death, MO-induced p53 can also elicit patterned gene expression via activation of proapoptotic factors, that can be mistaken for developmental regulation [7]. In many cases, legitimate effects by MOs on gene expression have been distinguished from artifacts by using a rescue control in which a synthetic mRNA, engineered to lack the MO recognition sequence, is co-injected. If the MO-induced phenotype can be reverted to the control pattern, the knockdown has been deemed to result specifically from the loss of target gene expression. We used MOs, in combination with both of these types of controls, to analyze the developmental function of the serine/ threonine protein kinase PAK4 in zebrafish [8].
PAK4 is one of six related proteins that mediate signal transduction via Rho-class GTPases to control a plethora of biological processes, including cytoskeletal dynamics, cell polarity and migration [9]. The family can be subdivided into two groups, PAK1-3, which are tightly regulated by interaction with GTPases, and PAK4-6, which bind specifically to GTPases, but have kinase activity that does not depend on this binding. In addition, some functions of the Group II PAKs may be independent of protein kinase activity [10,11]. In mouse, loss of pak4 function via gene targeting results in a complex lethal phenotype including disruption of placental and epiblast vasculature, heart failure, defective neurons and folding of caudal neural tube. Some aspects of this phenotype might be secondary effects of vascular failure [12,13]. The loss of function phenotype we observed in zebrafish is also lethal, but differs in many respects from that found in mouse. The zebrafish egg contains high levels of pak4 mRNA, but PAK4 protein is not detectable until the high stage of cleavage, approximately three hours after fertilization. Consistent with this maternal mRNA storage, we found that loss of pak4 expression in zebrafish behaved as a maternal effect: disruption of zygotic expression alone, via splice-inhibitory MOs, had no effect on development, whereas inhibition of translation using a MO targeting the 59-end of the open reading frame, alone or in conjunction with splice MOs, resulted in complex defects on the formation of primary myelopoietic cells, vasculature, and somite formation. All of these defects were shown to be independent of the p53-mediated interference noted above, and could be efficiently rescued by the injection of synthetic mRNA lacking the MO target sequence [8].
Because of the uncertainties associated with MOs, and to provide a genetically defined platform for further analysis of pak4 function in zebrafish, we used TALEN-based gene targeting to establish germ-line loss-of-function mutations. TALENs (transcription activator-like effector nucleases) are recombinant proteins based on the TALE transcription factors of Xanthomonas bacteria. These have been used to generate site-specific double strand breaks in the genomes of several metazoan species, including zebrafish. These breaks are usually repaired by nonhomologous end joining, which is error prone, frequently leading to small insertions or deletions (indels) at the repair site. When such indels result in a translational frame shift, loss of function for the encoded protein may be inferred. This strategy was used to generate frame shifts just downstream from the N-terminal methionine codon in the pak4 transcript, resulting in complete loss of this protein. In contrast to the results with MOs, pak4 null zebrafish were found to be normal and fertile, in both F 2 and F 3 homozygous and compound heterozygous mutant individuals. We conclude that pak4 function is dispensable in zebrafish, and that the antisense MO approach, even with robust RNA rescue controls, may be subject to misleading results.

Animals and Ethics Statement
Zebrafish of EK background were used in this study. Embryo handling and care was carried out using standard procedures [14]. All experiments using zebrafish were performed according to the NICHD IACUC-approved animal study protocol # 12-039. This ASP is not used for other projects. Total protein from 100 embryos at shield stage was collected, immunoprecipitated and analyzed using a custom-made zebrafish pak4 anti-peptide antibody. wt, wild type EK. D2/D4, F 3 mutants of mixed genotypes pak4D2/D2, pak4D4/D4 and pak4D2/D4. pak4 protein with a predicted molecular weight of 72 kDa was detected in the wild type embryos but not in the mutants. Duplicate membrane probed with antibody to a-tubulin was used as loading control. doi:10.1371/journal.pone.0100268.g001

Design and Generation of TALENs
The TALEN reagents used in this study were designed and constructed by Cellectis (Paris, France). Briefly, candidate TALEN target sequences in the zebrafish pak4 gene were identified (TALEN Hit Search Report), and from these we selected one immediately downstream from the start methionine codon, in order to achieve maximal disruption of the pak4 coding sequence (see Figure 1A for positions and target sequences). TALEN expression constructs were generated in the pTAL.T7 vector backbone using the ''Unit Assembly'' method. Sequences of repeat-variable di-residues (RVDs) in the TAL Effector DNAbinding domains were NN-NG-NG-HD-NI-NN-HD-NI-NI-NN-NI-NI-NN-NI-NI-NG for the TALEN left arm and NG-HD-NN-NI-NI-NN-NN-NG-NN-HD-NG-NN-NI-NN-NI-NG for the TA-LEN right arm. Both TALENs had nuclear localization signal sequences at the N-termini. Initial validation test for cleavage activity were performed by using the single strand annealing (SSA) assay on episomal target sequences in yeast by Cellectis.
The TALEN arms were subcloned into the pCS2 vector for expression in zebrafish embryos. TALEN RNAs were synthesized in vitro by using a mMachine mMessage Transcription Kit (Ambion, USA) and injected into 1-cell stage embryos. Total protein at shield stage was then extracted for Western blot analysis to confirm successful translation (data not shown). TALEN mRNAs (25,50,125 and 250 pg of each arm) were injected at 1-cell stage. To detect indel mutations in the TALEN target region, High Resolution Melting Curve Analysis (HRMA) was performed. Crude extracts of genomic DNA were prepared from zebrafish embryos at 24 hpf by alkaline hydrolysis in 20 ml 50 mM NaOH at 95uC for 20 minutes followed by neutralization with 200 ml 10 mM Tris-HCl. 1 ml of this crude extract was used as template for subsequent PCR. A 106-bp amplicon flanking the TALEN target region was amplified with primer pair F13 and R17 using KAPA HRM FAST PCR Kit (KAPA BIOSYSTEMS, USA) in an Eco Real-Time PCR machine (Illumina, USA) according to the manufacturers' instructions. The reaction profile consisted of polymerase activation at 95uC for 3 min, followed by 50 cycles of denaturation at 95uC for 5 sec and a combined annealing-extension step at 63uC for 30 sec. This was immediately followed by the HRMA profile program: the reaction mix was incubated at 94uC for 30 sec, 40uC for 30 sec and then the temperature gradually increased to 94uC. Double-stranded DNA content was monitored continuously via EvaGreen dye fluorescence. The melting curve profile of each sample was compared to that of control zebrafish EK DNA. When a significant deflection of the melting curve was observed, indicating heterogeneity in the sample, an extended PCR amplicon (primers F14 and R3) was cloned into the pGEM-T Easy vector (Promega) according to the manufacturer's instructions, and sequenced (Macrogen, USA). The TALEN dosage showing the highest indel mutation efficiency, and less than 50% morphological toxicity at 24 hpf, was used for subsequent injection to generate TALEN mutant fish.

Genotype Analysis
Genomic DNA from tail fin biopsies or whole embryos was prepared by the hot alkaline lysis method described above. Wild type-and mutation-specific forward primers (wt-F, D2-F and D4-F; see Table 1 for sequences) were designed at the mutation site and used for PCR with reverse primer R3. Longer amplicons (primers F14 and R3) were cloned and sequenced for verification.

Whole Mount In Situ Hybridization
In situ hybridization of whole mount embryos at 7-somite and 24 hpf was carried out according to published procedures [15]. Antisense digoxigenin (DIG)-labeled riboprobes were synthesized using the DIG RNA Labeling kit (Roche) according to the manufacturer's instructions.

RNA Isolation and RT-PCR
Total RNA from whole embryos was extracted using guanidinium thiocyanate and phenol/chloroform procedures as previously published [16]. Total cDNA was synthesized by using the Superscript III Reverse Transcriptase (Life Technologies) and oligo (dT) 20 primer according to the manufacturer's instructions. Gene-specific primers were designed and used in PCR detection and amplification of pak4 or b-actin mRNA target regions (see Table 1 for primer sequences).

Cloning of PAK4-GFP Fusion Constructs
Forward and reverse primers F1 and R2, annealing to 59-UTR and translation stop codon of zebrafish pak4, respectively, were used to amplify pak4 cDNA by RT-PCR using RNA extracted from whole embryos at 60% epiboly from control EK, pak4D2/D2 and pak4D4/D4 mutant fish. PCR products were cloned upstream of eGFP, in frame, in pCS2 vector using the Gibson Assembly kit (New England Biolabs, USA) to generate mRNAs encoding PAK4-eGFP fusion proteins. All constructs were verified by DNA sequence analysis.

Immunoprecipitation and Western Blot Analysis
A custom-ordered anti-PAK4 peptides antibody raised in rabbit was used in immunoprecipitation and Western blotting as previously described [8]. A mouse monoclonal anti-GFP antibody (SC-9996, Santa Cruz, USA) was used for detecting enhanced GFP fusions of PAK4 in Western blotting.

Microscopy
Live and fixed embryos, for both fluorescence and bright field, were mounted in 3% methylcellulose and 100% glycerol, respectively, and imaged and documented with Leica MZ16F microscope and Leica DFC500 camera.

Generation and Identification of pak4 Mutant Alleles
Two TALEN arms were designed targeting exon 2 of the zebrafish pak4 gene, centered just downstream of the ATG start codon ( Figure 1A). TALEN RNAs (50 pg left +50 pg right) were injected to produce pak4 mutant alleles as described in Materials and Methods. Bacterial clones of PCR-amplified genomic DNA isolated from pooled (5) 24 hpf TALEN-injected embryos revealed indel mutations in the target region, ranging from 1 to 9 nucleotides deletion or insertion. Sibling embryos were raised to 2 months and screened individually by HRMA on fin biopsy genomic DNA, which identified a total of 18 out of 64 fish carrying somatic indel mutations. These carriers were out-crossed with wild type EK fish and the offspring (F 1 ) were raised at a reduced culturing density (6 fish per 1.8 liter) in order to accelerate growth, allowing breeding as early as 6 weeks. Genotype analysis of the 1month-old F 1 fish revealed an average of 10% germ-line transmission of mutant alleles, with the same mutant allele present in all individuals derived from a single G 0 parent. A total of four mutations were identified: deletion of 2, 4 and 5 nucleotides (D2, D4 and D5) and insertion of 6 nucleotides (ins6; Figure 1A). The D2 and D4 mutant alleles were identified first, and for this reason we focused on these lines for subsequent analyses. The pak4D2 and pak4D4 mutant alleles both generate frame shifts at amino acid residue #9, followed by missense polypeptides of 9 and 83 amino acids, respectively, neither of which shows significant similarity to any putative protein in the NCBI database. Wild type zebrafish PAK4 is 663 amino acids, 72 kDa [8].
In the F 1 generation we identified 7 pak4D2 heterozygous carriers, all of which turned out to be female, while only a single male was identified as carrying the pak4D4 allele. We crossed the single D4 male with the D2 females to obtain a pool of F 2 offspring yielding the expected Mendelian proportion of 25% transheterozygous pak4 mutants (pak4D2/pak4D4), determined by DNA sequence genotyping. Juvenile pak4D2/D4 individuals were raised and in-crossed to obtain an F 3 generation comprising a mix of homozygous and trans-heterozygous mutant genotypes, which was used in most experiments. For brevity, we refer to these populations as ''pak4D2/D4''.

Characterization of the pak4 Knockout
Based on the abundance of pak4 mRNA in zebrafish eggs and the observation that a translation-inhibiting MO, but not spliceblocking MOs, disrupted development, we concluded that pak4 may function as a maternal-effect gene [8]. To eliminate the possibility of maternal PAK4 protein contribution to eggs from heterozygous mothers, we carried out our genetic analyses on F 3 offspring of F 2 compound heterozygous null individuals. To verify the loss of pak4 expression, total protein from pak4D2/D4 embryos (F 3 ) at shield stage was extracted for immunoprecipitation and Western blot analysis using the zebrafish PAK4 custom antibody as previously reported [8]. As shown in Figure 1B, the pak4D2/D4 embryos expressed no detectable PAK4 protein, indicating that both deletions represent null alleles for this gene. However, pak4D2/D4 embryos at 2 dpf (Figure 2A), as well as sexually mature adults (6 months old) exhibited no signs of physical abnormalities ( Figure 2B). This was in contrast to the morphology of maternal-zygotic(MZ) pak4 MO knockdown, which included reduced head and tail size, kinked body axis and lethality by 6-7 dpf [8].
In addition to physical defects, pak4 knockdown also resulted in the inhibition of primitive myelopoiesis. Primitive erythropoiesis was unaffected, as was definitive hematopoiesis [8]. It seemed possible that transient loss of myeloid cell types might not be evident from gross morphology, so we carried out whole mount in situ hybridization of the two hematopoietic markers stem cell leukemia (scl) and myeloperoxidase (mpo) in the pak4D2/D4 embryos. Figure 2 shows that scl (C) and mpo (D) were expressed in the pak4D2/D4 embryos in normal patterns as compared to the wild type counterparts at 7-somite and 30 hpf, respectively. We conclude that unlike in pak4 morphants, primitive myelopoiesis is normal in pak4 null zebrafish.
In gene-inactivation experiments it is important to verify that there exists only a single copy of the target gene in the relevant haploid genome. As previously described [8], both of the most recent versions of the EnsEMBL zebrafish genome builds (Zv8 and Zv9) were exhaustively searched for pak4 and pak4-like sequence using different BLAST search engines. Similar searches were repeated in the present study with the same result, i.e. only a single copy of pak4 could be identified in the sequenced zebrafish genome (ENSDARG00000018110; Zv9). We have also searched zebrafish cDNA/mRNA and expressed sequence tag (EST) databases, and again only a single pak4 transcript was identified. All other sequence matches obtained from either genomic or mRNA databases corresponded to other PAK family members: pak1a, pak1b, pak2a, pak2b, pak5, pak6a or pak6b (data not shown). Absence from sequence databases is strong evidence, but is not proof, that an additional copy or copies of the pak4 gene do not exist. We further investigated this question by performing RT-PCR using 4 primer combinations annealing to 59-untranslated and coding regions of pak4. Using total RNA isolated from both pak4D2/D4 and wild type EK embryos at shield stage as templates, an identical pattern was obtained, with only a single product evident for each primer pair ( Figure 3A, B). Cloning and sequencing of the PCR products gave only one result for each amplicon, in all cases matching perfectly to the database pak4 version, with the exception of the 2-and 4-base deletions in RNA isolated from the mutants. Similar results were obtained at reduced annealing stringency (50uC; data not shown). In this experiment, amplicon b is especially important, as the primer targets correspond to the MO sites. Thus a hypothetical alternatively spliced mRNA that somehow bypassed the frame shift mutations would still have to include these targets sequences, since the MO cocktail was shown to block expression of the protein [8]. We conclude from these results that there is only a single pak4 gene in the zebrafish genome, without alternative splicing within the open reading frame, and therefore neither gene duplication nor splice variants can account for the discrepancy between loss of function experiments with MOs and TALENs.
A more remote possibility we considered is that, as a result of the introduced deletions, translation of the mutant pak4 mRNA might initiate downstream from the N-terminal methionine codon used in wild type individuals, conceivably resulting in a truncated but partially functional protein. There are three ATG codons between the deletion site and the beginning of the kinase domain (which would presumably be needed for at least some pak4 functions). The first (amino acid 91) is in a Kozak initiation consensus (ACCATG; Kozak, 1986 [17]), but is upstream from the epitope site, and would thus result in 64-kDa protein that would have been detected by the IP-Western experiment ( Figure 1B). The other two do not match the Kozak consensus (amino acid 345, CCCATG and amino acid 382, CAGATG), making this hypothesis unlikely. However, as a direct test, enhanced GPF fusions of wild type PAK4 (PAK4wt), PAK4D2 and PAK4D4 were cloned and expressed in fish embryos ( Figure 4A). As shown in Figure 4B, green fluorescence was detected in the PAK4wt-eGFP RNA-injected embryos at 24 hpf while no fluorescence was found in either PAK4D2-eGFP or PAK4D4-eGFP samples. This was confirmed by Western blot analysis of the injected embryos using anti-GFP antibody ( Figure 4C). These results indicate that the D2 and D4 mutations prevent expression of the downstream pak4 reading frame.
Finally, we asked if zebrafish pak4 mRNA might have a function in addition to serving as the template for pak4 protein synthesis, for example, some regulatory activity residing within the open reading frame that could rescue the MO phenotype, that might be inhibited by the MOs used in our knockdown experiments. Such a regulatory function has been reported for the squint (sqt) transcript in zebrafish, in which elements residing in the 59-and 39-noncoding regions cooperate in an RNA-dependent and MO-sensitive scaffolding function required for dorsoventral axis specification [3]. To address this possibility for pak4, the RNAs for the wild type and mutant PAK4-eGFP fusions were co-injected with the MO cocktail in an attempt to rescue the knockdown phenotypes. Myeloperoxidase (mpo), a marker gene for normal granulocytes, was used as an indicator for myelopoiesis. As shown in Figure 5A and B, morphants expressed a reduced level of mpo (MO) as we reported previously [8]. Co-injection of wild type pak4 RNA (MO+ wt) rescued mpo expression, whereas the mutant forms, PAK4D2-eGFP (MO+D2) and PAK4D4-eGFP (MO+D4), did not. It is therefore unlikely that the discrepancy between pak4 MO knockdown and gene inactivation by TALENs can be explained by non-coding RNA function. We conclude that the MOs used in these and previous experiments result in a highly specific but nevertheless spurious phenotype that is both independent of p53mediated apoptosis [8] and reversible by co-injection of pak4 mRNA.
What could be the basis of such an effect? One possibility is that certain MOs might induce some form of stress on embryos that could conceivably, in concert with gene-specific knockdown, result in a phenotype that might not appear in mutant embryos, since they would lack the second hypothetical stress from the MOs. We attempted to test this notion by subjecting the pak4 mutant embryos to conditions that might lead to such a stress. First, we injected ''standard control'' MO (Gene Tools, USA) into wild type and mutant embryos (pak4D2/D4, F 3 ) at the 1-cell stage, at the same dose as used for the pak4 MO cocktail (6 ng). As shown in Figure 6A, this had no effect: both sets of embryos developed normally into feeding larvae. Second, we tested thermal stress. Zebrafish embryos develop normally between 23uC and 33uC. The recommended optimal incubation temperature is 28.5uC [14]. Exposure to elevated temperature results in the induction of heat shock chaperone proteins [18], leading to increased mortality and developmental abnormalities. Wild type and mutant embryos (D2/D4, F 3 ) were incubated at 35uC from blastula stage to 24 hpf. The mutant samples exposed to high temperature developed more rapidly, but had no observable morphological phenotype at 24 hpf, compared to siblings raised at 28.5uC ( Figure 6B). Identical results were obtained with wild type embryos. By 3 dpf, 25-30% of both mutant and wild type embryos receiving the 35uC treatment showed abnormalities, such as curled tail, short body axis and edema (data not shown) indicating that thermal stress did in fact disrupt development. Thus neither MO injection per se, nor heat stress, was able to elicit the MO knockdown phenotype in pak4 mutant embryos. It is possible that some other form of stress, such as impurities in the original MO preparations, might have such an effect, but it is not obvious how this hypothesis could be tested.

Discussion
Loss of pak4 in mouse by gene targeting resulted in an embryonic lethal phenotype with multiple tissue defects including placental and epiblast vasculature, myocardial walls, spinal cord motor neurons and interneurons, caudal neural tube [12,13]. It is surprising that in zebrafish this gene appears to be dispensable. It is reasonably clear that the locus we have studied in fish is the closest homolog to the mammalian pak4 [8], but are these genes actually functional homologs? The genetics would seem to indicate that they are not. If there is an essential function in zebrafish that correspond to what PAK4 does in mammals, it must be performed by some other gene, or genes. It may be that PAK4, and the other Group II PAK proteins (5, 6a, 6b) have evolved different functions, or even lost essential functions altogether, during the mammalianteleost divergence period. Such a notion is relevant to the practice of using zebrafish as a model system for the functional study of genes identified by mammalian genetics. As more loss-of-function data become available from gene targeting in fish, it will be interesting to see how common it is to find functional divergence compared to mammals, and if this might be predicted in some way a priori.
MOs have been in widespread use for the identification of primary gene function in zebrafish, Xenopus as well as other organisms [19,20]. While this approach affords rapid and relatively inexpensive assays, it comes with serious disadvantages that are becoming increasingly evident. Off-target effects on proapoptotic genes have been widely reported and reviewed [6,7]. Many studies have incorporated controls using MOs targeting the pro-apoptotic transcription factor p53 in an attempt to exclude such artifacts. Additional controls have included the use of multiple MOs targeting the same mRNA, or using injection of MO-resistant modifications of the targeted mRNA to achieve a phenotypic rescue [5].
However, as gene targeting methods have become established in zebrafish, some MO-induced phenotypes have been confirmed by loss of function alleles. For example AHR2 [21,22,23], WASp [24] and cas [25,26] mutant zebrafish closely mirror the MO phenotypes. However, other examples are beginning to emerge in which MO-induced phenotypes have been validated by stringent controls, but nevertheless do not coincide with stable loss-of-function genetic mutations. The first such report concerned Wild type and F 3 mutants were exposed to 35uC from blastula stage to 24 hpf. Control samples (28.5uC) were incubated for 5 hours longer than the heated embryos in order to compensate for different developmental rates. doi:10.1371/journal.pone.0100268.g006 Fmr1, the zebrafish homolog of FMR1, a highly-conserved gene that is mutated in human Fragile X Syndrome. Tucker et al. [27] used MOs to knock down Fmr1, reporting neuronal and craniofacial abnormalities, both of which could be rescued in RNA injection controls. Subsequently, den Broeder et al. [28] reported on two zebrafish lines independently isolated from a TILLING screen with inactivating mutations in Fmr1. Both lines yielded normal, fertile progeny, in particular lacking any sign of the defects noted in the earlier study using MOs. A second example came from analysis of barx1, a homeodomain gene homologous to the human BARX1 locus. Sperber and Dawid [29] reported MO knockdown of this gene resulted in reduced chondrogenesis in pharyngeal arch derivatives. Again, this phenotype could be rescued by co-injection of suitably modified barx1 mRNA. From another TILLING screen, Nichols et al. [30] identified two alleles with premature stop codons within the homeobox domain and thus likely to result in loss of function. These lines showed a number of defects in craniofacial morphology, but had none of the abnormalities reported with the MO knockdowns. A third case is the novel gene inka, which we identified using Xenopus and have also studied in mouse and zebrafish. MO knockdowns in frog and fish both resulted in severe craniofacial dysmorphology. These experiments were controlled by the use of multiple MOs, and parallel analysis in both species; mRNA injection controls were infeasible due to the highly disruptive nature of inka overexpression [31]. However, targeting the inka gene in mouse resulted in normal, fertile inka null adults [32]. Furthermore, inka null alleles in zebrafish have been identified by the TILLING procedure. Homozygous inka null zebrafish are, like the mouse, apparently normal and fertile (unpublished data). None of the TILLING-based studies carried out extensive tests for alternate expression modes that might account for the difference in phenotype compared to MOs, however the most likely interpretation in all three cases is that the MOs induced phenotypes unrelated to the target genes. It is interesting to note that craniofacial defects were prominent in all of these examples.
We have now described yet another example in which a MO phenotype, validated in this case by RNA rescue controls, fails to correspond to the phenotype of stable genetic mutations. In addition to the defects in myelopoiesis, we also observed other abnormalities in pak4 morphants, including somite morphology, which like the myelopoiesis defect was not p53-dependent and could be rescued with pak4 mRNA, and also defects in craniofacial cartilage, which did show a response to p53 inhibition ( [8] and unpublished observations). In the Fmr1 and barx1 studies noted above, p53-dependence was not investigated, so it is possible that some aspects of the knockdown phenotypes might have been promoted by apoptosis. There have been anti-apoptotic functions attributed to pak4 [33,34,35,36], so it is conceivable that the p53independent effects we observed could result from apoptosis mediated by other pathways nevertheless responsive to pak4 overexpression. Even if such a scenario could be supported, the fact remains, however, that mutating the pak4 gene did not elicit any of the defects found with pak4 MOs.
In conclusion, MOs can be useful in confirming the identity of genes responsible for mutant phenotypes, or other experiments in which the loss of function phenotype has been established independently. However, when these agents are used as a primary tool to determine gene function, extensive controls are essential. Ideally, these should include both rescue by mRNA injection, and the demonstration that non-overlapping MOs yield identical phenotypes. To date, few reports based on MO experiments have met these criteria. Hence, it is likely that additional instances of incongruity between MO and genetic mutant phenotypes will emerge.