klf2a sh317 Mutant Zebrafish Do Not Recapitulate Morpholino-Induced Vascular and Haematopoietic Phenotypes

Introduction and Objectives The zinc-finger transcription factor Krϋppel-like factor 2 (KLF2) transduces blood flow into molecular signals responsible for a wide range of responses within the vasculature. KLF2 maintains a healthy, quiescent endothelial phenotype. Previous studies report a range of phenotypes following morpholino antisense oligonucleotide-induced klf2a knockdown in zebrafish. Targeted genome editing is an increasingly applied method for functional assessment of candidate genes. We therefore generated a stable klf2a mutant zebrafish and characterised its cardiovascular and haematopoietic development. Methods and Results Using Transcription Activator-Like Effector Nucleases (TALEN) we generated a klf2a mutant (klf2a sh317) with a 14bp deletion leading to a premature stop codon in exon 2. Western blotting confirmed loss of wild type Klf2a protein and the presence of a truncated protein in klf2a sh317 mutants. Homozygous klf2a sh317 mutants exhibit no defects in vascular patterning, survive to adulthood and are fertile, without displaying previously described morphant phenotypes such as high-output cardiac failure, reduced haematopoetic stem cell (HSC) development or impaired formation of the 5th accessory aortic arch. Homozygous klf2a sh317 mutation did not reduce angiogenesis in zebrafish with homozygous mutations in von Hippel Lindau (vhl), a form of angiogenesis that is dependent on blood flow. We examined expression of three klf family members in wildtype and klf2a sh317 zebrafish. We detected vascular expression of klf2b (but not klf4a or biklf/klf4b/klf17) in wildtypes but found no differences in expression that might account for the lack of phenotype in klf2a sh317 mutants. klf2b morpholino knockdown did not affect heart rate or impair formation of the 5th accessory aortic arch in either wildtypes or klf2a sh317 mutants. Conclusions The klf2a sh317 mutation produces a truncated Klf2a protein but, unlike morpholino induced klf2a knockdown, does not affect cardiovascular development.


Methods and Results
Using Transcription Activator-Like Effector Nucleases (TALEN) we generated a klf2a mutant (klf2a sh317 ) with a 14bp deletion leading to a premature stop codon in exon 2. Western blotting confirmed loss of wild type Klf2a protein and the presence of a truncated protein in klf2a sh317 mutants. Homozygous klf2a sh317 mutants exhibit no defects in vascular patterning, survive to adulthood and are fertile, without displaying previously described morphant phenotypes such as high-output cardiac failure, reduced haematopoetic stem cell (HSC) development or impaired formation of the 5 th accessory aortic arch. Homozygous klf2a sh317 mutation did not reduce angiogenesis in zebrafish with homozygous mutations in von Hippel Lindau (vhl), a form of angiogenesis that is dependent on blood flow. We examined expression of three klf family members in wildtype and klf2a sh317 zebrafish. We detected vascular expression of klf2b (but not klf4a or biklf/klf4b/klf17) in wildtypes but found no differences in expression that might account for the lack of phenotype in klf2a sh317 mutants. klf2b morpholino knockdown did not affect heart rate or impair formation of the 5 th accessory aortic arch in either wildtypes or klf2a sh317 mutants.

Introduction
KLF2 belongs to the KLF zinc finger transcription factor family, characterised by three tandem C 2 H 2 zinc fingers that serve as a DNA binding motif [1]. Since its characterisation in mice [1] and humans [2] it is clear that mammalian KLF2 plays important roles in vascular development and homeostasis [3]. KLF2 modulates expression of genes critical in regulating vascular tone [4,5], thrombosis [6], inflammation [3,4] barrier function [7] and antioxidative properties of endothelium [8]. Within the vascular wall, KLF2 is expressed exclusively in endothelium [3,5,9]. Endothelial KLF2 expression is induced by blood flow via mechanotransduction [3,10], as well as via non flow-dependent mechanisms such as inflammatory cytokines or statins [4,6,11]. Homozygous Klf2 knockout mice die between E12.5-14.5 from intra-embryonic and intra-amniotic haemorrhaging associated with defective blood vessel morphology. Impaired vascular smooth muscle cell and pericyte migration resulting in failure to organize into a compact tunica media has been reported [12,13], although while another group confirmed the stage of lethality (E11. 5-13.5) and abdominal bleeding, they but did not detect vessel wall abnormalities [14]. Conditional Klf2 deletion in murine embryonic tissues confirms that endothelial Klf2 deletion is responsible for embryonic mortality [15].
Assessment of gene function in zebrafish has often used morpholino antisense oligonucleotide-mediated gene knockdown, and several studies have reported the effect of klf2a knockdown on zebrafish cardiovascular and haematopoietic development. klf2a knockdown has been reported to cause high-output cardiac failure with pericardial oedema and high aortic blood flow velocity attributed to loss of smooth muscle tone [15]. klf2a knockdown induced failure to form the 5 th accessory aortic arch (AA5x) although with otherwise preserved vascular patterning [18]. In relation to haematopoietic development klf2a knockdown impaired maintenance of haematopoietic stem cell (HSC) gene expression via a blood flow dependent signalling cascade involving nitric oxide (NO) [19].
Morpholino oligonucleotides (MO) remain commonly used in zebrafish reverse genetic studies. However, their reliability has been called into question [20]. The advantages of MOs in terms of cost and speed are counterbalanced by their temporary and incomplete action and potential for off-target effects and toxicity [21]. The development of targeted genome editing to create stable, heritable mutations in genes of interest avoids the disadvantages of morpholino use. Several techniques are available, including Zinc Finger Nucleases (ZFN) [22], Transcription Activator-like Effector Nucleases (TALEN) [23] and most recently Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR-associated systems (CRISPR/Cas9) [24]. These induce site-specific double stranded DNA breaks that are repaired by error-prone nonhomologous end joining DNA repair resulting in mutations that can disrupt the reading frame.
Targeted mutation has been proposed to be a more reliable way of assessing gene function [20]. Recent publications have shown that several published morphants phenotypes are not reproduced in mutants [25]. Very recently, Rossi et al. found that the severe vascular defect induced by egfl7 morpholino knockdown was not seen in egfl7 mutants, due to a compensatory upregulation of Emilin 2 and 3 not seen in morphants [26].
The existence of mechanisms that compensate for the effects of loss-of-function mutations would explain the low rate of overt phenotypes seen in the Sanger zebrafish mutagenesis project [27]. However, as zebrafish were first established as a model for forward mutagenesis screens [28][29][30], clearly not all mutations can be so "rescued". There is therefore a pressing need to rapidly establish which mutations can be compensated for, and which cannot.
We used TALENs to generate a stable klf2a mutant line (named klf2a sh317 ). This mutation produces an aberrant mRNA transcript that is predicted to produce a significantly truncated klf2a protein without the essential DNA binding motifs. However, homozygous klf2a sh317 mutants exhibit no phenotypic differences compared to wildtypes (WT) and did not reproduce previously described klf2a morphant phenotypes, nor did they show any effect in miR-126 levels. We examined the expression patterns of three evolutionary closest klf family members; klf2b, klf4a and biklf/klf4b/klf17 in the search for a gene that might compensate for the mutation in klf2a. Although we detected klf2b expression in the vasculature, which has not been described previously, we found no differences in expression of these genes between klf2a sh317 mutants and wildtypes, and klf2b knockdown did not reveal any phenotype in klf2a sh317 mutants.

Zebrafish Husbandry
Zebrafish were raised in the Bateson Centre aquaria and fed Artemia nauplii. Zebrafish were kept on a constant 14

RNA Extraction, Reverse Transcription and Polymerase Chain Reaction
Total RNA was extracted from pools of 30 embryos or unfertilised eggs using TRIzol (LIFE TECHNOLOGIES) according to manufacturer`s instructions. Total RNA was reversely transcribed into complementary DNA using the VERSO cDNA Synthesis Kit (THERMO SCIENTIFIC). Primers for amplification of klf2a, klf2b, and gapdh were: Quantitative PCR (qPCR) of zebrafish mature miR-126-5p was performed according to the manufacturer's instructions using the TaqMan MicroRNA Assay. Total RNA from 72hpf zebrafish embryos was purified using Trizol. cDNA was prepared using the TaqMan microRNA Reverse Transcription kit (4366596, APPLIED BIOSYSTEMS) and qPCR performed using miR-126-5p and U6 TaqMan Assays (4427975 and 4427975, APPLIED BIOSYSTEMS). The primers used for reverse transcription and qPCR were supplied with each assay. The qPCR primer is a stem-loop oligonucleotide containing the sequence of the mature miR-126-5p (CAUUAUUACUUUUGGUACGCG). qPCR was performed in duplicate using an ABI 7900HT Fast Real-Time PCR System. Expression of miR-126-5p was normalized to U6 for relative quantification. Individual points represent relative quantification of miR-126-5p in a pool of embryos relative to mean expression in the wild-type group. Summary data is displayed as mean ± standard error of the mean (SEM). Statistical comparison was made using a two-tailed Student's t-test on log-transformed data.

Measurement of Cardiovascular Performance
Heart Rate. Embryos were kept in individual dishes and individually removed from the 28°C incubator immediately prior to measurement to minimize the effect of environmental temperature on heart rate [37]. Heart rates were measured directly by eye under a stereomicroscope for 30 seconds.
Blood flow velocity. Particle image velocimetry (PIV) analysis using erythrocytes as tracer particles was used to measure aortic blood velocity in the mid aorta immediately above the cloaca. Zebrafish embryos were imaged at 10x magnification and at 300 frames/second using a high speed camera (OLYMPUS IX81) and Video Savant 4.0 digital video recording software (IO INDUSTRIES). 400 images (frames) in.tiff format per each embryo were recorded. Images were analysed using ImageJ (version 1.45s public domain software http://imagej.nih.gov/ij) to obtain a kymograph with lines representing movements of individual erythrocytes in real time. Kymographs were uploaded to custom made software CORRELATOR which calculated instantaneous erythrocyte velocities throughout the cardiac cycle. Knowing the ratios of pixel/ μm (0.8pixel/μm), pixel/frame (1:1) and the rate of imaging (300frames/second), velocity (v) was calculated as follows: v = distance/time [μm/sec]. Average erythrocyte velocity per cardiac cycle was calculated by averaging all instantaneous velocities measured during the cardiac cycle. Additionally, cardiac cycles from all embryos of the same group were put in phase, and average instantaneous erythrocyte velocities for each frame were calculated and plotted as a single velocity curve. developmental stage: 0min (24hpf), 15min (36hpf), 40min (48hpf), 80min (72hpf), 100min (4dpf) and 150min (5dpf). Maleic acid buffer was used instead of phosphate buffered saline for day 2 and day 3 washes. Expression was detected with digoxigenin (ROCHE) labelled riboprobes and resolved using BM Purple (ROCHE). EST clone IMAGp998A0510285Q (IMA-GENES) containing full-length klf2a cDNA sequence was digested with XBaI and EcoRI (both NEB) resulting in a 1144bp fragment which was inserted into Bluescript KS expression plasmid and subsequently used for klf2a antisense riboprobe synthesis (EcoRI (NEB) and T7 RNA polymerase (ROCHE)). Total embryonic RNA at 48hpf was used as a template for klf2b, klf4a and biklf/klf4b/klf17 riboprobes synthesis. The following primers were designed to amplify approx. 1000bp cDNA fragments: klf2b F 5`CGTGGACATGGCTTTACCTT 3`, klf2b R 5`ATGGGAG CTTTTGGTGTACG 3`, klf4a F 5`TTGATAGCATGGCACTGAGC 3`, klf4a R 5`CCTGC GGAAATCCAGAATAA 3`, biklf/klf4b/klf17 F 5`ACCCCGGACATGAATTATCA 3`, biklf/ klf4b/klf17 R 5`TGTCCGGTGTGTTTCCTGTA 3`. T3 promoter sequence 5`CATTAACCCT CACTAAAGGGAA 3`was added to 5`end of each of the F primers and T7 promoter sequence 5`TAATACGACTCACTATAGGG 3`was added to 5`end of each of the R primers. RT-PCR and subsequent PCR amplification of corresponding cDNA was performed in a single step using Superscript III One-Step RT-PCR kit (INVITROGEN) following the manufacturer`s instructions. Antisense riboprobes were synthesized using T7 RNA polymerase (ROCHE). The remaining riboprobes used in this study have been kindly donated by other research groups: dll4 [40] runx1 [41] and cmyb [42].

Microscopy
Visual assessment of embryos, monitoring of WISH staining and heart rate counting were performed using a Leica S6E stereo microscope (LEICA MICROSYSTEMS). Images of live embryos and WISH images were taken using a Leica M165 FC fluorescent stereo microscope with Leica DFC310 FX camera (both by LEICA MICROSYSTEMS). Fluorescence microscopy was performed using UltraVIEW Vox spinning disc confocal microscope and Volocity v5.3.2 imaging software (both by PERKIN ELMER). Zebrafish embryos were anaesthetised with Tricaine and mounted on a cover slip in 1% low-melting point agarose prior to imaging. Images were taken in 2μm slices across the region of interest and digitally processed by focal plane merging (Z stacking). Merged images were further analysed using ImageJ.

Western Blotting
Total protein was extracted from genotyped zebrafish embryos at 5dpf (4 embryos per group) using RIPA lysis buffer (SIGMA-ALDRICH) with added Halt Protease and Phosphatase inhibitor cocktail (THERMO SCIENTIFIC). Embryos were homogenised using an Eppendorf pestle (SIGMA-ALDRICH) and left on ice for 30min. Lysate was centrifuged at 14000rpm for 10min at 4°C and the supernatant was separated and protein amounts quantified using a Bradford assay. Protein samples were then mixed with 1M dithiothreitol (DTT) (SIGMA-ALDRICH) and NuPAGE LDS Sample Buffer (LIFE TECHNOLOGIES) heated to 95°C for 5min and loaded onto a NuPAGE NOVEX 4-12% Bis-Tris gel (LIFE TECHNOLOGIES). After the electrophoretic separation, proteins were transferred to Immobilon-P PVDF transfer membranes (MILLIPORE) using the Xcell II Blot Module (LIFE TECHNOLOGIES). After blocking with 5% non-fat milk for 1 hour at RT, transfer membranes were incubated with rabbit anti-mouse polyclonal Klf2 antibody (raised against a polypeptide containing AAs 70-270 of mouse wildtype Klf2a protein) (1:500) (MILLIPORE, Cat.# 09-820) or with a rabbit anti-human β-actin antibody (1:10000) (CELL SIGNALLING, Cat.# 4967) overnight. Membranes were washed in TBSTw (Tris buffered saline plus Tween 20: 136.8mM NaCl, 24.8mM Tris, 0.1% Tween 20 (vol/vol) (all from SIGMA), pH 7.6) for 3x5min on a rocker and incubated with a secondary antibody (goat anti-rabbit polyclonal antibody with conjugated horseradish peroxidase (HRP), 1:1000) (DAKO) for 45min on a rocker. After thorough washing of the membranes, immobilised proteins were detected by the EZ-ECL Chemiluminiscence Detection Kit (BIOLOGICAL INDUSTRIES).

klf2a TALEN Mutagenesis
We followed a published protocol for targeted TALEN mutagenesis [23]. klf2a genomic sequence was submitted to the Old TALEN Targeter software at https://tale-nt.cac.cornell.edu/ node/add/talen-old. A target site at the 5`end of exon 2 5`TCAACCCATCACCACCTCCA CCG/AT[ACACCACCAGCCTACTGGC]AGAGCTTCTGCAGTCTG 3`(19bp spacer sequence between target sequences for L and R TALEN subunits is annotated with square brackets) containing the XcmI (NEB) restriction enzyme site 5`CCANNNNNNNNNTGG 3`was chosen for the mutagenesis. Primers flanking the target site were designed amplifying a 281bp PCR product; klf2a TAL XcmI F 5`CAGGCGACTACAGAATGCAA 3`and klf2a TAL XcmI R 5`GCCCTCTTGTTTGACTTTGG 3`. Genomic DNA extraction was performed using REDExtract-N-Amp™ Tissue PCR Kit (SIGMA-ALDRICH). XcmI (NEB) restriction enzyme digest (3 hours, 37°C) results in generation of 180bp and 101bp fragments. Successful mutagenesis is indicated by the loss of the restriction site and the presence of an undigested 281bp band after the XcmI (NEB) digest. Following the assembly and capped mRNA synthesis, 1.5ng of capped mRNA coding for the klf2a TALEN construct was injected per embryo at one cell stage.

Statistical Analysis
F1 klf2a mutant studies were blinded to whether embryos were homozygous mutants, heterozygotes or wildtype (after imaging and quantification embryos were euthanized and genotyped). Statistical analysis was carried out using GraphPad Prism v5.04 software. Quantitative data are presented as mean ± standard error of the mean (SEM). Statistical comparisons between two groups were made by Student's t-test. Statistical significance is indicated as: ns = non-significant, Ã = p<0.05, ÃÃ = p<0.01, ÃÃÃ = p<0.001, ÃÃÃÃ = p<0.0001.

The Expression of klf2a in Developing Zebrafish Embryos
We analysed klf2a expression during zebrafish embryonic development daily from 1 to 5 days post fertilisation (dpf) in wildtype embryos using WISH (Fig 1A). Prior to onset of blood flow (24-26hpf) we could detect only non-vascular klf2a expression notably in the cloaca (Fig 1A, grey arrowheads) and in cells abutting the most posterior notochord (Fig 1A, black arrows). After onset of blood flow, klf2a was expressed in the developing vasculature, in particular at 48hpf (Fig 1A, red arrow). Cross sections at 48hpf confirmed vascular expression of klf2a ( Fig  1B).These observations confirm previously published data on klf2a expression in early embryonic stages [17]. We next extended our studies to later stages of development to 5dpf and found klf2a expression in the trunk vasculature, intersegmental vessels and cardiac outflow tract (Fig 1A, red arrows and black arrowheads respectively). klf2a mRNA in the cardiac outflow tract was detectable up to 5dpf, whereas klf2a expression in the trunk vasculature and intersegmental vessels was only detectable up to 72hpf (Fig 1A). From 72hpf onwards, klf2a was expressed in the subintestinal veins that subsequently form the hepatic vein at 5dpf  Fig 1A, green arrows). We also detected klf2a expression in the neuromasts of the lateral line from 72hpf onwards (Fig 1A, white arrowheads).
We confirmed that klf2a expression in the zebrafish vasculature is dependent on blood flow using three distinct approaches. These were by preventing cardiac contraction by morpholino antisense knockdown of cardiac troponin T2; preventing aortic blood flow to the trunk due to the aortic occlusion caused by homozygous mutation in hey2/gridlock; and finally by chemically-induced cessation of established blood flow between 46-65hpf by incubation in tricaine. When we examined vascular klf2a expression, this was reduced in all three models of hindered flow, in keeping with previous studies using tnnt2 knockdown [3] (S1 Fig).

Generation of a Stable klf2a Mutant Line
To examine the role of klf2a on cardiovascular and haematopoietic development we generated a klf2a mutant zebrafish using TALEN. The site for mutagenesis was chosen at the 5`end of exon 2 (Fig 2A). Mutations that produce a premature stop codon in this region are predicted to result in translation of a truncated Klf2a protein lacking the 3 tandem zinc fingers DNA binding motif located at the C terminus of the wildtype Klf2a protein (Fig 2B), rendering it unlikely to function as a transcription factor. We raised TALEN injected embryos and sequenced their genomic DNA and identified a founder carrying an allele with a 14 base pair (bp) deletion that disrupted the reading frame at the target site and generated a premature stop codon 14bp downstream. This allele was named klf2a sh317 and is predicted to generate a truncated Klf2a protein with molecular weight of 13.15 kilodalton (kDa) (Fig 2B). All mutant fish from the F1, F2 and F3 generation used in subsequent experiments were genotyped and the klf2a sh317 mutation confirmed by sequencing.
To determine the effect of the klf2a sh317 mutation on the mRNA transcript, we extracted whole embryonic RNA from 30 pooled embryos from the F3 generation of klf2a sh317 mutants. klf2a coding sequence was reversely transcribed, PCR amplified, cloned into an expression plasmid and transformed into bacterial cells. 23 separate colonies were sequenced. All confirmed the expected klf2a sh317 sequence without any additional transcripts (S2 Fig). We did not therefore find any evidence that the klf2a sh317 mutation induces alternative splicing.
Targeted mutation can induce gene loss-of-function by a number of mechanisms. As well as production of a truncated or non-functional protein, the mRNA can be targeted for nonsense mediated decay (NMD) [43]. We examined wither the klf2a sh317 mutation induced NMD by performing in situ hybridisation for klf2a in wildtypes and homozygous klf2a sh317 mutants, but found no difference in mRNA expression (S3 Fig). This, together with the fact that we could amplify and clone the expected mRNA transcript by rt-PCR (S2 Fig) strongly suggests that the klf2a sh317 mutation does not induce nonsense-mediated decay.
We next attempted to examine the effect of the klf2a sh317 mutation on Klf2a protein expression and size by western blotting in wildtype and klf2a sh317 mutant embryos (Fig 2C). Although there are no monoclonal antibodies available against zebrafish Klf2a we were able to obtain and use an anti-mouse polyclonal Klf2a antibody previously shown to bind to zebrafish Klf2a [19]. As might be expected from a polyclonal antibody raised against a largely conserved region of several Klf genes this blot shows multiple bands consistent with the presence of Klf2a, in addition to other bands which are consistent with Klf4a and Biklf/Klf4b/Klf17 expression (predicted molecular weights 38kDa and 47kDa respectively; indicated by asterixes in Fig 2C). However, a band likely to represent wildtype Klf2a protein with the correct size of 43kDa was detected in either wildtype or heterozygous klf2a sh317 embryos (Fig 2C). This band was greatly reduced in homozygous klf2a sh31 mutants, with the residual band migrating with a similar molecular weight likely to reflect the expression of Klf2b (predicted molecular weight 41.2kDa). Additionally, two bands of approximately 31-33kDa were detected in klf2a sh317 homozygous mutants, probably representing truncated Klf2a protein (Fig 2C). Although larger that the predicted molecular mass of Klf2a sh317 protein, the truncated protein would be expected to be very acidic (ph~4, versus~7.8 for full-length protein) and contain a significant proportion of proline residues, both features that contribute to aberrant behaviour on SDS-PAGE.
To determine the presence and potential contribution of maternal klf2a mRNA we analysed klf2a expression by RT-PCR in unfertilised zebrafish embryos. We found no evidence of maternal klf2a mRNA in unfertilised zebrafish eggs (Fig 2D). Taken together, these data strongly suggest that the klf2a sh317 mutation induces a significant truncation of the klf2a protein that is highly likely to perturb its function as a transcription factor.

Characterisation of the klf2a sh317 Mutant Line
Homozygous klf2a sh317 mutants displayed no morphological abnormalities compared to wildtypes by at least 5dpf (Fig 3). In contrast to previously described klf2a morphants [15], no pericardial oedema was observed at any stage. klf2a sh317 mutants showed no difference in heart rate or blood flow velocity in the dorsal aorta at 48hpf compared to wildtypes, in contrast to the high-output cardiac failure described in klf2a morphants [15] (Fig 4). Furthermore, we raised homozygous klf2a sh317 mutants to adulthood and both these (F2 generation) and incrossed offspring (F3) were viable and fertile, making late-onset cardiac failure or major cardiovascular defects unlikely and excluding maternal effects that compensate for the mutation when homozygous mutants are derived from heterozygous females.
Next we examined the vascular anatomy of klf2a sh317 mutant embryos. We crossed klf2a sh317 mutants with the double transgenic line Tg(kdrl:HRAS-mCherry;flk1:EGFP-nls) that differentially labels endothelial cell cytoplasm and nuclei. We found no differences in vascular patterning in klf2a sh317 mutants ( Fig 5A). Specifically, we examined the trunk vasculature and found no abnormal formation of the axial or intersegmental vessels, nor any difference in endothelial cell number in these regions (Fig 5B and 5C).Formation of the 5 th accessory aortic arch (AA5x) has been shown to be impaired either in klf2a morphants or in the absence of blood flow [18]. We confirmed that in our hands, formation of the AA5x is blood flow dependent (S4 Fig) but formation of the AA5x was unaffected in klf2a sh317 mutants (Fig 5A).
We recently showed that the excessive hypoxic signalling-driven trunk vessel angiogenesis in homozygous vhl mutants (vhl hu2117 ) is blood flow dependent [44]. Since klf2a is a major endothelial flow responsive transcription factor, we determined the effect of the klf2a sh317 mutation on angiogenesis in vhl hu2117 mutants. Heterozygotes for klf2a sh317 and vhl hu2117 in the Tg(fli1:eGFP) background were incrossed. Embryos homozygous for both klf2a sh317 and vhl hu2117 still display aberrant intersegmental vessel development (Fig 6) indicating that the klf2a sh317 mutation does not alter this process.
Since maintenance of haematopoetic stem cell (HSC) gene expression is impaired in klf2a morphants, or in the absence of blood flow [19] we examined this in klf2a sh317 mutants. While we confirmed that runx1 and cmyb expression was reduced in the absence of blood flow (S5 Fig), expression of both genes was unaffected in homozygous klf2a sh317 mutants (Fig 7).
Since homozygous klf2a sh317 mutants (unlike klf2a morphants) displayed no cardiovascular phenotype, we sought to determine whether this could be explained by redundancy between other klf family members. We examined expression of the three genes phylogenetically closest to klf2a, namely the paralog klf2b and both KLF4 orthologs klf4a and biklf/klf4b/klf17, to assess whether these might be compensating for the klf2a sh317 mutation. In addition to the close molecular phylogeny to KLF2, KLF4 has been identified as another endothelial blood flow dependent transcription factor in human and mouse [45,46]. Vascular expression of klf2b, klf4a or biklf/klf4b/klf17 in wildtype zebrafish embryos has not been described [17,47].
At 48hpf klf2b mRNA was detected in the cleithrum and the mesenchyme of the pectoral fin bud (Fig 8, black arrow) and in the epidermal cells (Fig 8, red arrowhead) in keeping with previously published data [17,48]. Additionally, klf2b mRNA was detected in the ISVs of some wildtype and klf2a sh317 mutant embryos (Fig 8, red arrows). At 72hpf klf2b expression persisted in the pectoral fin and epidermis. As in 48hpf embryos, a small proportion of both wildtype and klf2a sh317 mutants at 72hpf exhibited vascular expression of klf2b in the ISVs and the  (Fig 8, red arrows). No differences in klf2b expression were apparent between wildtype and klf2a sh317 mutants.
We detected klf4a expression in the epidermis (Fig 9, black arrows) and in the pectoral fin (Fig 9, red arrowheads) of 48hpf wildtype embryos in keeping with previously published data [47]. klf4a expression in klf2a sh317 mutants was not different to wildtype and no vascular expression of klf4a was detected (Fig 9). Similarly, in situ hybridisation for biklf/klf4b/klf17 at 48hpf confirmed previously published data [17] showing expression in the neuromasts of the lateral line organ (Fig 9, black arrowheads) and in the hatching gland (Fig 9, red arrows). Again, no differences in biklf/klf4b/klf17 expression were observed between wildytpe and klf2a sh317 mutants and no vascular biklf/klf4b/klf17 expression was detected (Fig 9).
The above detailed experiments found no evidence of compensatory upregulation of klf2b, klf4a, or biklf/klf4b/klf17 in klf2a sh317 mutants. However, even without evidence of upregulation, normal levels of expression of these genes could conceivable compensate for loss-of-function of klf2a. Since only klf2b was seen to be expressed in the vasculature, we examined whether morpholino knockdown of klf2b induced cardiovascular phenotypes in klf2a sh317 mutants. Morpholino-induced klf2b knockdown had no effect on either heart rate or AA5x formation in either heterozygous or homozygous klf2a sh317 mutants, strongly suggesting that redundancy between klf2a and klf2b does not explain the lack of a cardiovascular phenotype in klf2a sh317 mutants (Fig 10).  microRNA 126 (miR-126) has been shown to be upregulated by klf2a and morpholinoinduced klf2a knockdown resulted in reduced miR-126 expression levels in zebrafish [18]. To attempt to further explore the reason for the lack of phenotype in klf2a sh317 mutants, we quantified miR-126 levels in 72hpf klf2a sh317 mutants compared with wildtypes. We found no reduction in miR-126 levels in klf2a sh317 mutants compared to wild types (Fig 11). This suggests that although the mutation clearly induces a significant alteration in the klf2a transcript in association with a truncated Klf2a protein, the downstream functions of klf2a appear not to be perturbed.

Discussion
In this study we describe for the first time the generation and characterisation of a stable klf2a mutant zebrafish. We find that, in contrast to several studies describing a range of  klf4a and biklf/klf4b/klf17 expression patterns do not differ between wildtype and klf2a sh317 mutants at 48hpf. klf4a expression was detected in epidermis (black arrows) and pectoral fins (red arrowheads) and biklf/klf4b/klf17 expression was detected in neuromasts of lateral line organ (red arrows) and in hatching glands (black arrowheads) of wildtype embryos and klf2a sh317 mutants in keeping with previously published data [17,47] No vascular expression of klf4a or biklf/klf4b/klf17 could be detected in any of the wildtype embryos or klf2a sh317 mutants examined. Figures in bottom left corner of each image indicate the number of embryos with similar staining patterns out of total number of embryos examined. Scale bar = 500μm. cardiovascular phenotypes in klf2a morphant zebrafish, homozygous klf2a sh317 mutant embryos display no discernible cardiovascular abnormalities, survive into adulthood and are fertile. There are several possible explanations for the discrepancy between the morphant and klf2a sh317 mutant phenotype, a phenomenon that is increasingly being recognised as targeted mutants are generated for genes previously studied using morpholinos [25].
It is possible that some or all of the published klf2a morpholino studies are describing offtarget or toxic effects of the reagent used, rather than specific effects of klf2a knockdown. This is perhaps particularly possible in the case of klf2a, expression of which is regulated by blood flow. Non-specific effects that reduce cardiac output would indirectly reduce klf2a expression and confound interpretation of the phenotype. This could explain the findings that klf2a morpholino knockdown impairs AA5x [18] and HSC formation [19], both of which are impaired in the absence of blood flow. It is harder to ascribe the phenotype of high-output heart failure [15] to off-target effects or toxicity since this is a phenotype not commonly seen in morpholino studies, although the pericardial oedema also reported in this study is commonly induced by high-dose morpholino injections. However, high-output heart failure was not reported in the later klf2a morpholino studies, suggesting that it was not observed by these groups.
Conversely, it is possible that the klf2a sh317 mutation does not induce loss-of-function and this accounts for the lack of a mutant phenotype. Despite concerns about the reliability of morpholinos, and the increasing preference for using stable mutants, it remains challenging to confirm that a targeted mutation truly induces loss-of-function. The deletion in the genomic DNA sequence of klf2a sh317 mutants leads to a frame shift resulting in a premature stop codon that Fig 11. Comparison of miR-126 expression levels in klf2a sh317 mutants with wildtypes at 72hpf. Expression levels of miR-126 do not differ between klf2a sh317 and wildtypes at the observed time point. Expression of miR-126 was normalized to U6 for relative quantification. Individual points represent relative quantification of miR-126 in a pool of embryos relative to the mean expression in the wild-type group. Summary data is displayed as mean ± SEM. doi:10.1371/journal.pone.0141611.g011 would remove the DNA binding motifs necessary for klf2a to function as a transcription factor. Sequencing the cDNA from klf2a sh317 mutants found no evidence that the deletion leads to an alternatively spliced transcript that could remain functional. We were fortunate to obtain an antibody with at least some degree of cross-reactivity against zebrafish Klf2a, since a lack of suitable antibodies prevents assessment of the effect of targeted mutation on protein expression in the case of most genes. Western blotting with this antibody confirmed loss of wildtype Klf2a in klf2a sh317 mutants. However, we did not detect a band of the predicted size of the mutant transcript (13kDa) but instead detected a novel band of around 33kDa. Although it is possible that this represents the predicted mutant protein with additional post-translational modifications, it is also possible that some mechanism has allowed the ribosome to produce a truncated yet functional Klf2a protein in the face of the klf2a sh317 mutation.
After initial submission of this paper, Rossi et al. published a description of a compensatory mechanism that rescues the effect of egfl7 mutants but not of egfl7 morpholino knockdown, via upregulation of Emilins [49]. Since the klf2a sh317 mutation seems highly likely to perturb klf2a function, yet does not affect miR-126 levels, we consider it most likely that a similar mechanism exists for klf2a. We have shown that this compensation is not mediated by klf2b, but it is entirely possible that other genes may rescue the effect of the the klf2a sh317 mutation. Since forward mutagenesis screens in zebrafish have been highly successful, not least for cardiovascular phenotypes [50], such mechanisms appear not to compensate for many genes. However, the existence of such mechanisms may explain the observation that in a large-scale zebrafish mutagenesis study only 6% of disruptive mutations in protein-coding genes induced an observable phenotype [27].
In summary, our work shows that even in the face of clear evidence of a potentially disruptive mutation induced in a gene of interest, it is currently very difficult to be certain that this leads to loss-of-function, and hence to be confident about the role of the gene in embryonic development.
Supporting Information S1 Checklist. NC3Rs ARRIVE Guidelines Checklist. (PDF) S1 Fig. klf2a vascular expression is blood flow-dependent. klf2a is expressed in zebrafish embryonic vasculature of a WT embryo at 48hpf. Cessation of blood flow in the trunk vasculature by an occlusion of proximal aorta in the gridlock mutants (indicated by a green arrow) results in a complete loss of klf2a vascular expression distally to the occlusion. Blockage of embryonic heart contractions by tnnt2 MO results in significantly decreased klf2a vascular expression. Pharmacological inhibition of heart contractions by tricaine from 32 to 48hpf results in significantly decreased klf2a vascular expression. Interestingly tricaine also reduces klf2a expression in the heart region and in the cells lateral to the most posterior notochord. Red arrows indicate trunk vasculature, black arrows indicate the cells lateral to most posterior notochord, white arrows indicate pectoral fins, black arrowheads indicate the cardiac outflow tract and white arrowheads indicate cloaca. Numbers in top right corners indicate the number of embryos with similar staining pattern out of all embryos examined. klf2a riboprobe used. Scale bar = 500μm. (TIF) S2 Fig. klf2a wildtype and klf2a sh317 coding sequences, primary protein structures and klf2a sh317 cDNA sequencing. (A) klf2a wildtype and klf2a sh317 coding sequences. Exons are colour coded in alternating black and blue colours. Nucleotides within the 19bp spacer between the klf2a TALEN L and R subunit binding sites are underlined. Targeted TALEN-induced mutagenesis occurs around this region. Codons coding for amino acids that bind the zinc atoms within the 3 tandem C 2 H 2 zinc fingers at the C terminus of wildtype Klf2a protein are highlighted in red. 14bp deletion in the klf2a sh317 coding sequence causes disruption of the reading frame. (B) Primary protein structures of Klf2a WT and Klf2a sh317 proteins are aligned. Cysteine (C) and histidine (H) amino acids that bind zinc atoms within the 3 tandem C 2 H 2 zinc fingers at the C terminus of wildtype Klf2a protein are highlighted in red.  ) is significantly diminished in the tnnt2 morphants which lack blood flow confirming previously published data [19]. Black arrows indicate the aorta-gonad-mesonephros (AGM) region and red arrows indicate caudal haematopoietic tissue (CHT). Numbers in the top left corners indicate number of embryos with similar staining patterns out of total number of embryos examined. Scale bar = 500μm. (TIF)