To generate putative tbx5a null alleles in zebrafish, we employed Cas9 ribonucleoprotein complex (RNP)-mediated mutagenesis using our established sgRNA[tbx5ccA] that targets the first coding exon (Fig. 1A). This sgRNA targets the coding sequence in the first coding exon downstream of the conserved translation initiation codon. We targeted the first exon to introduce frameshift and subsequent stop codons early in the open reading frame to avoid potential translation of N-terminal Tbx5a protein remnants that could retain function. Further indicating that targeting this region could result in loss-of-function alleles, the corresponding amino acid sequence is highly conserved between zebrafish and humans (indicating functional conservation) and human HOS patients have been identified with frameshift-introducing nucleotide insertions at similar positions within TBX5.
Maximized mutagenesis using Cas9 RNPs with sgRNA[tbx5ccA] cause recognizable tbx5a loss-of-function phenotypes in F0 crispants. We injected the sgRNA complexed with Cas9 protein as solubilized RNPs at sub-optimal concentration to achieve viable mosaicism (see Methods for details) in the multicolor Tg(lmo2:dsRED2;drl:EGFP;myl7:mCyan) reporter background, subsequently abbreviated as RGB. In RGB embryos, dsRED2 labels endothelial, hematopoietic, and endocardial progenitors (lmo2) in red, EGFP marks all lateral plate mesoderm lineages (drl) including pectoral fins in green, and AmCyan reveals the differentiated cardiomyocytes (myl7) in blue; consequently, RGB enables in vivo imaging of all cardiovascular and additional LPM lineages over the first 3 days of development.
From F0 outcrosses that transmitted mutant tbx5a alleles, we genotyped adult F1 zebrafish for the presence of mutated tbx5a alleles by tail clipping, PCR, sequencing, and CrispRVariants analysis. From the recovered germline alleles, we kept heterozygous strains for the lesions c.21_25del and c.22_31del (hence forward abbreviated as tbx5aΔ5 and tbx5aΔ10) (Fig. 1A, B). These alleles generate out-of-frame mutations starting from base +21 or base +22, respectively, and result in premature stop codons shortly after the conserved initiation codon.
We in-crossed adult F1 heterozygotes for tbx5aΔ5 and tbx5aΔ10 and inter-crossed parents for each allele to assess F2 homozygous and trans-heterozygous embryos for developmental phenotypes at 3 dpf. We found that all combinations of the alleles resulted in Mendelian ratios of heart defects (Fig. 1E-H,U) and concomitant, completely penetrant loss of pectoral fins (Fig. 1I-L, U). The cardiac defects for the allele combinations included: cardiac edema with blood accumulation at the inflow tract region (Fig. 1F,H, white asterisks), heart mis-looping (Fig. 1G), and misshapen atrial and ventricular chambers (Fig. 1F-H, N-P, R-T), with n=49/306 for tbx5aΔ5, n=55/231 for tbx5aΔ10, n=108/417 for tbx5aΔ5/Δ10 (Fig. 1U). Clutch values including mortality (from now on abbreviated as death rate D.R.) were: for tbx5aΔ5 clutch 1, n=29 D.R.=20.7%; clutch 2, n=144 D.R.=11.1%; clutch 3, n=133, D.R.=24.1%. For tbx5aΔ10 clutch 1, n=78 D.R.=7.7%; clutch 2, n= 65, D.R.=12.3%; clutch 3, n=88, D.R.=0%. For tbx5aΔ5/Δ10 clutch 1, n=58, D.R.=22.4%; clutch 2, n=170, D.R.=3.5%, clutch 3, n=147, D.R.=7.5%.
While cardiac defects were fully penetrant in homozygous and trans-heterozygous mutants, the expressivity of the cardiac phenotype was highly variable, ranging from inflow tract defects (Fig. 1F) to mis-looped chambers (Fig. 1G). Live imaging using selective plane illumination microscopy (SPIM) allowed optical sectioning (Fig. 1M-P) and imaging of the whole heart (Fig. 1Q-T, side view), revealing additional details of the chamber defects. We detected atrium mis-positioning (Fig. 1N,O), freely floating and rounded-up ventricles within the pericardial cavity (Fig. 1N-P, R-T), and thinner cardiac walls (Fig. 1P) compared to wildtype or heterozygous siblings that develop a regularly formed ventricle anchored within the pericardium (Fig. 1M,Q). mRNA expression of versican a (vcana) in homozygous tbx5aΔ5 mutants was expanded in tbx5a-mutant hearts (Fig. 1V,W). All these phenotypes are well-documented for both tbx5a morphants in which tbx5a mRNA is downregulated via morpholino injection and in embryos homozygous for the classic tbx5a allele hst.
Nonetheless, in contrast to the reported morpholino and hst mutant phenotypes, we never detected the most-severe form of the hst phenotype consisting of a string-shaped heart tube and a deformed head. We readily observed this phenotype using translation-blocking tbx5a morpholino injections (n=30/106) (Fig. 1D), in line with previous reports of variable expressivity. The presence of the hst phenotype itself has previously also been linked to the genetic background, suggesting that the hst phenotype is a variation of the tbx5a loss-of-function phenotype. Taken together, homozygous and trans-heterozygous combinations of our new tbx5a frameshift alleles recapitulate morphological and molecular phenotypes of tbx5a morphants and the classic hst mutant with exception of the heartstrings phenotype. This observation suggests that either our frameshift alleles are not null and possibly hypomorphs, or alternatively that the existing hst allele and morpholino injections result in hypomorphic or dominant-negative conditions arising from truncated residual protein or lower protein concentration.
The introduction of CRISPR-Cas9 for genome editing has provided the zebrafish field with an easily accessible tool for generating mutant alleles for any gene of choice. Targeted mutagenesis using CRISPR-Cas9 requires careful assessment of targeted candidate gene loci to generate loss-of-function alleles. In contrast to classic forward genetic screens that by definition start from a mutant phenotype linked to a molecular lesion, non-homologous end joining (NHEJ)-based mutagenesis of a candidate locus can result in non-phenotypic lesions. Potential causes of the lack of phenotypes in de novo generated mutants include i) translation from downstream start codons, leading to truncated protein products with retained functions that are difficult to assess beforehand; ii) the unpredictable efficiency of nonsense-mediated mRNA decay (NMD) activated in case of premature stop codons; iii) use of alternative, cryptic splice sites to generate functional, translatable mRNA; iv) gene compensation caused by activation of alternative pathways mitigating the phenotype severity. Compensatory mechanisms in mutants have recently been reported in zebrafish for the egfl7 gene, and the role of compensation in mutant phenotype expressivity and variability in a broader context remains to be assessed.
Of note, the classic hst mutant still features detectable tbx5a mRNA, and we also detect tbx5a transcript by mRNA ISH in tbx5aΔ5- and tbx5aΔ10-mutant embryos (Fig. 1X,Y, and data not shown).The tbx5aΔ5 and tbx5aΔ10 lesions are situated in close proximity to the tbx5a translation initiation codon; while several possibly initiating ATGs are situated downstream and before the T-box, the amino acid sequence at the N-terminus where our alleles are introduced show conservation from teleosts to mammals (E.C., C.M., data not shown). In addition, frameshift mutations in similar positions within human TBX5 have been recovered from HOS patients. The full penetrance of concomitant pectoral fin loss and cardiac defects further suggest that no efficient alternative starting codon downstream of the two mutations provides a fully compensating protein product, nor that tbx5b would functionally compensate for the function of tbx5a. We do acknowledge the possibility that tbx5b could act redundant or could compensate for the heartstrings phenotype, clarification of which will require double mutants for both Tbx5-encoding genes in zebrafish.