A complete Fxn knockout in mice is embryonic lethal (Cossée et al., 2000). As a result, several conditional Fxn knockdown mouse models have been developed (Kelekçi et al., 2022). Studies using these conditional mouse models have suggested that Fxn may have roles in development (Chandran et al., 2017; Jiralerspong et al., 1997; Koeppen et al., 2017; Santos et al., 2001). These mouse models as well as work in systems ranging from bacterial to FRDA patient derived cell lines have contributed greatly to our understanding of FRDA pathology, but they have a limited ability to answer the question of how FXN deficiency impacts early aspects of embryonic development such as organogenesis.

To address this question, we developed a fxn knockdown zebrafish model using an antisense oligomer morpholino strategy to specifically target zygotic fxn expression, thereby preserving expression from maternal fxn transcripts. This model allows us to model Fxn deficiency in the zebrafish, rather than a complete Fxn absence, similar to what is seen in human patients that experience a reduction of FXN (Filla et al., 1996; Gellera et al., 2007). Our morpholino (MO) targets the exon 3 splice donor site, thus was predicted to interfere with proper connection between the exon 3 and exon 4 coding sequence (Figure 1A). Wild-type zebrafish were microinjected with fxn MO at the 1 cell stage, and at the 28 somite stage (ss) we employed reverse transcriptase polymerase chain reaction (RT-PCR) followed by gel electrophoresis (Figure 1B) and Sanger sequencing (Figure 1C) to scrutinize the effect on fxn transcripts. While wild-type controls exhibited normal splicing, embryos injected with the fxn exon 3 splice donor MO exhibited exon skipping. Morphant transcripts had a complete deletion of exon 3 and exon 2 was spliced onto exon 4 (Figure 1C). This aberrant splicing event resulted in a predicted frameshift of the open reading frame and the presence of an in-frame premature stop codon near the beginning of exon 4 (Figure 1C).

Next, using the fxn transcript sequences expressed by wild-type and fxn morphants, we employed the protein folding prediction algorithms Alphafold 2 and Phyre 2 to predict the three-dimensional structure of the Fxn protein (Figure 1D; Kelley et al., 2015; Jumper et al., 2021). This approach was utilized because at this time there is currently no x-ray crystallography data for zebrafish Fxn, and we wanted to visualize how the fxn transcript alteration would disrupt the Fxn protein structure. The two algorithms predicted similar structures for both the endogenous Fxn protein and the product of our zygotic Fxn knockdown, but with critical differences (Figure 1D). The endogenous immature Fxn protein predictions consist of three alpha helices and five beta-sheets (Figure 1D). This is consistent with the x-ray crystallography data of mature human Fxn which has two alpha helices and five beta-sheets (Dhe-Paganon et al., 2000). In contrast, the zygotic Fxn protein knockdown was predicted by the algorithms to have only one alpha helix (Figure 1D). Therefore, the algorithm predictions suggest that the morpholino would not only disrupt the three-dimensional structure of zebrafish Fxn, but indeed abrogate the active site due to the location of the premature stop codon.

In human patients with FRDA, multiple cell types are affected. Over time, there is considerable neuronal degeneration in the central and peripheral nervous systems and many patients experience severe hypertrophic cardiomyopathy, which is the most common cause of death (Koeppen and Mazurkiewicz, 2013; Koeppen et al., 2015). These neural and cardiac phenotypes have been recapitulated in Fxn-deficient mouse lines as well (Chandran et al., 2017).

To explore and characterize development in fxn-deficient zebrafish, we first examined their live morphology at 24, 48, and 72 h post fertilization (hpf) (Figure 2A). We observed several striking phenotypes in fxn morphants compared to wild-type embryos. At 24 hpf, 96.3% of fxn-deficient embryos had a gray head pallor, a hallmark of cell death (Figures 2A,B). Also, 22.9% of the fxn-deficient embryos displayed hallmarks of malformed ear development, with three otoliths at 30 hpf compared to wild-type controls who formed otic vesicles with two otoliths (Whitfield, 2019) (Figures 2A,C). At 48 hpf, 76.4% of the fxn-deficient embryos had blood pooling in or near the pericardium which was often present with pericardial edema, the latter of which is one hallmark of renal dysfunction (Figures 2A,D). At 72 hpf, pericardial edema/blood pooling was present in 38.1% of the morphant animals (Figures 2A,D). Further, the majority of fxn-deficient embryos exhibited overt jaw malformations, where jaws were significantly smaller in size (Figure 2A). To better examine facial structure and jaw development, we performed Alcian blue staining to reveal the underlying cartilaginous cranial structures at 4 days post fertilization (dpf) (Figures 2F,G). We found that 85.7% of the morphants failed to develop ceratobranchials 1-5 by 4 dpf (Figures 2F,G). The mandibular prominence in these animals also had a convex curve when visualized laterally, rather than a concave curve that follows the profile of the maxillary prominence. Additionally, the Meckel’s cartilage also failed to fuse medially (Figure 2G). These data strongly suggest that fxn expression is required for the proper development of the pharyngeal arches and ceratobranchials.

Finally, we noted that as fxn-deficient zebrafish developed between 24 and 72 hpf, they displayed a shorter tip to tail length compared to wild-type embryos, suggestive that their growth was being compromised wherein they displayed a general failure to thrive (Figure 2A). Given this observation, we examined whether fxn deficiency would impact survival. Over the course of 12 days, fxn-deficient embryos exhibited a 64% chance of survival as opposed to wild-type controls that by comparison had a 94% chance (Figure 2H). These data strongly suggest that fxn deficiency disrupts the proper development of zebrafish embryos and their transition to a healthy larval stage of ontogeny.

The nervous system, heart, and kidneys have high metabolic rates. This would cause one to hypothesize that all of these tissues would be affected in FRDA patients as well as FXN-deficient mammalian models. Further, FXN is abundantly expressed in the embryonic and adult mouse kidney (Yue et al., 2014). However, kidney function has not been reported to be affected in the lifespan of FRDA patients (Reetz et al., 2018; Shinnick et al., 2016; Watters et al., 1981) and has similarly not been observed in animal models. This paradox, even more puzzling given the compromised fluid balance with severe edema observed in our fxn-deficient zebrafish embryos which are hallmarks of compromised renal physiology (Drummond et al., 1998; Wingert and Davidson, 2008; Gerlach and Wingert, 2013; Desgrange and Cereghini, 2015; Poureetezadi and Wingert, 2016; Fatma et al., 2021), led us to investigate next the formation and function of the zebrafish pronephros.

To do this, we utilized whole mount in situ hybridization (WISH) to visualize the different domains of the zebrafish pronephros at the 28 ss with previously established specific markers of each individual cell type (Figure 3). First, we used the nphs1 probe to mark the podocytes, which are essential to form the blood filtration unit used by the embryonic nephrons (Kramer-Zucker et al., 2005). The area occupied by nphs1 positive cells was decreased in fxn-deficient embryos compared to wild-type embryos (Figures 3A,B). Next, cells in the proximal convoluted tubule were marked with the probe for slc20a1a (Wingert et al., 2007). There was no statistically significant difference in the slc20a1a domain between the two groups (Figures 3A,B). Additionally, there was not a statistically significant difference in the proximal straight tubule domain, as marked with trpm7 (Figures 3A,B) (Wingert et al., 2007). Interestingly, while there was not a statistically significant difference in the domain of trpm7 positive cells, the fxn-deficient embryos appeared to be expressing less of this transcript, as the stain was fainter. Next, multiciliated cells (MCC) were marked with cetn2 (Ma and Jiang, 2007). The number of cetn2 positive cells was significantly decreased in fxn-deficient embryos compared to wild-types, suggesting that were fewer MCCs developed in the pronephros. Lastly, we examined the distal tubule, which is comprised of distal early cells that express slc12a1 and distal late cells that express slc12a3 (Wingert et al., 2007). Both distal segments were significantly shorter in length in the fxn-deficient embryos compared to wild-type controls (Figures 3A,B).

To further assess each of the renal populations that were significantly altered in the fxn-deficient embryos, WISH was performed using a panel of independent molecular markers for the affected cell type (Figure 4). Specifically, we used wt1b, trim35-30-201, kcnj1a.1, and gata3 to examine the podocytes, MCCs, distal early, and distal late, respectively (Figure 4). In each case, fxn-deficient embryos displayed significant reductions in the domain occupied by these pronephros populations compared to wild-type embryos, confirming our previous conclusions about the changes to each cell type based on our initial characterization of unique markers for each lineage (Figure 3).

Next, we utilized WISH to examine the emergence of the podocyte, distal early and distal late lineages at 22 ss and 26 ss, timepoints which precede the completion of nephron segmentation (Wingert et al., 2007; Wingert and Davidson, 2011). Additionally, these lineages were examined at 32 hpf to examine if the appearance of each population was developmentally delayed. The podocyte lineage was significantly reduced in fxn-deficient embryos compared to wild-types at the 22 ss, 26 ss and 32 hpf (Supplemental Figure S1). Likewise, the distal early lineage as marked by slc12a1 and distal late lineage as marked by slc12a3 was significantly reduced in fxn-deficient embryos compared to wild-type controls (Supplemental Figures S2, S3). Taken together, these data suggest that fxn is critical for the proper emergence of multiple lineages within the zebrafish pronephros.

To further explore this notion, we performed several control studies. First, we examined whether the fxn morpholino might have off target effects, which can be associated with edema and grey cranial pallor. The p53 morpholino is well known to attenuate common off target effects of morpholinos (Robu et al., 2007). We observed that co-injection of the p53 morpholino with the fxn morpholino led to similar phenotypes as fxn morpholino alone (Supplemental Figure S4), namely gray pallor in the cranium at 24 h post fertilization and edema between 48 and 72 hpf. In light of these results, we conclude that off target effects are not a significant component of the fxn-deficient morphant phenotypes. Additionally, wild-type zebrafish were microinjected with a standard control morpholino, raised to 24 hpf, and fixed for nephron segment analysis. WISH revealed that formation of the podocytes and tubule segments were not altered due to the standard morpholino control (Supplemental Figure S5). Taken together, these studies lead us to conclude that fxn deficiency causes specific defects in development including pronephros formation.

Defects in MCC development have been linked to embryonic fluid imbalance such as pericardial edema (Marra and Wingert, 2016; Marra et al., 2019; Chambers et al., 2020b). Therefore, we next sought to assess MCC specification in fxn-deficient embryos and examined whether the MCC populace was diminished at later timepoints. To characterize MCC progenitors, we examined the expression of pax2a and jag2b at the 22 ss using WISH and quantified these populations (Marra et al., 2019). MCC progenitor numbers in fxn-deficient animals were significantly decreased as compared to wild-type controls (Figures 5A,B). To examine whether the MCC populations might be diminished due to developmental delay, WISH was also performed several hours subsequent to this time point to survey the number of pax2a and jag2b expressing MCC progenitors. Fxn-deficient embryos at the 26 ss had significantly decreased pax2a+ and jag2b + MCCs compared to wild-type controls (Figures 5A,B). These data further support the conclusion that MCCs are not specified correctly in the absence of Fxn function.

To examine the progression of MCC development in the context of fxn deficiency, we conducted WISH analysis on 32 hpf animals using the trim35-30-201 and cetn2 riboprobes, which mark differentiating MCCs (Ma and Jiang, 2007; Liu et al., 2007; Marra and Wingert, 2016). While both the wild-types and fxn-deficient embryos formed MCCs that expressed these more mature markers, consistent with the ability of this lineage to differentiate, the number of trim35-30-201+ and cent2+ MCCs were significantly reduced in number in fxn-deficient embryos compared to wild-types (Figures 5C,D). In sum, these findings suggest that fewer MCCs are patterned in the fxn-deficient pronephros.

Because we observed a variety of renal cell populations being affected by fxn-deficiency, we hypothesized that pronephros dysfunction may indeed underlie the edema exhibited by fxn-deficient embryos. To test this hypothesis, we injected 36 hpf wild-type controls and fxn-deficient embryos and with 40 kD FITC-Dextran, as this compound can be utilized as a proxy to visualize and measure fluid clearance within live animals in real time (Weaver et al., 2022) (Figure 6). After introduction of this tracer, embryos were imaged at 6 h post injection (hpi) and 48 hpi to assess fluorescent intensity (Figure 6A). As observed previously, fxn-deficient embryos developed pronounced pericardial edema over this time period (Figure 6A). Upon analysis of the percentage fluorescent change between 6 and 48 hpi, fxn-deficient embryos displayed edema a significantly decreased ability to clear the FITC-dextran compared to wild-types (Figure 6B).

We also used FITC-Dextran to visualize proximal convoluted tubule (PCT) endocytosis activity. (Figure 6C). Receptor mediated endocytosis is utilized extensively by the proximal tubule to reabsorb materials in the renal filtrate (Anzenberger et al., 2006). While wild-type embryos exhibited robust endocytosis in the PCT of both nephrons, the fxn morphants displayed a range of reduced PCT function. Approximately 35% of fxn morphants had PCT endocytosis activity in both nephrons. However, about 25% of fxn morphants had PCT endocytosis activity in nephron only and about 40% of the fxn morphants had no PCT endocytosis in either nephron (Figure 6D). From this, we conclude that nephron function is compromised, and thus concluded that Fxn function is necessary to achieve proper pronephros function during development.

Decreased expression of FXN leading to ROS and consequent cell death is hypothesized to be a major underlying pathogenic component of FRDA, commonly affecting the nervous system (Abeti et al., 2016; Koeppen et al., 2011; Koeppen and Mazurkiewicz, 2013). In our morphological studies, we had observed gray pallor in the developing central nervous system of Fxn-deficient embryos, which is a harbinger of cell death (Cheng et al., 2015). Thus, next we sought to characterize cell survival in fxn-deficient zebrafish to test the hypothesis that Fxn loss of function was indeed causing an increase in cell death across tissues during embryogenesis.

To test this, we used 24 and 48 hpf Fxn-deficient embryos to perform acridine orange (AO) staining (Tucker and Lardelli, 2007), which is a DNA intercalating fluorescent dye that rapidly and robustly labels apoptotic cells and readily penetrates deep tissues in whole mount embryos (van Ham et al., 2010). Next, we quantified cell death in several regions, focusing on the brain as well as the pronephros and the heart. Fxn-deficient zebrafish had significantly more AO + cells in the brain at both 24 and 48 hpf compared to wild-types (Figures 7A–D). Fxn-deficient embryos also had significantly more AO + cells in the pronephros at 24 hpf than wild-type controls (Figures 7A,B). At the 48 hpf time point, however, there was not a significant difference in AO + cell number between wild-types and Fxn-deficient embryos (Figures 7C,D). Interestingly, we did not observe AO + cells within the heart in either wild-type or fxn-deficient embryos at 24 or 48 hpf (Figure 7). These observations suggest that cell death is a component of the developmental defects which transpire during both pronephros and central nervous system formation, but not the heart, in fxn-deficient zebrafish.

As noted previously, we found that fxn-deficient embryos were typically smaller in tip to tail length than wild-types of a comparable developmental age, suggesting compromised growth over time (Figure 2). Thus, we wondered whether the decreases seen in the several pronephros populations like the distal tubules and MCCs were simply due to the animal being smaller. To examine this possibility, we conducted WISH using cdh17, which is a pronephros specific marker (Figure 8). cdh17 stains the length of the nephron tubules (Wingert et al., 2007), thus allowing us to compare the absolute length of the pronephros versus the embryo tail length from tip to tail (Figure 8A). By measuring and comparing these lengths in 24 hpf wild-types and fxn-deficient embryos, we found that the pronephros and the tail were both significantly shorter than in fxn-deficient embryos compared to wild-type control embryos at this age (Figure 8B). When we calculated the ratio of tail/cdh17 length, fxn-deficient embryos were found to have a reduction in pronephros size that was slightly less severe than the decrease in tail length (Figure 8B). This suggests that the pronephros undergoes a decrease in size that is very similar to the overall embryonic size decrease in fxn-deficient embryos. We also examined performed tail analysis to explore whether the decreases in distal tubule length were proportionate to the smaller pronephros size. Examination of two distal early markers (slc12a1, kcnj1a.1) and two distal late markers (slc12a3, ptgs1) revealed that the decreased lengths of these segments were approximately 25% while tail lengths were reduced by approximately 10%–12% (Supplemental Figure S6). This analysis reveals that the distal segments of fxn-deficient embryos are disproportionately reduced compared to the tail or the nephron overall (Figure 8C). Future studies are needed to ascertain whether changes in regions such as the cloaca, collecting duct or Corpuscles of Stannius, endocrine glands associated with the pronephros lineage, are affected in fxn-deficient embryos and how changes in these cell types relate to the distal segment reductions.

Zebrafish were cared for by the Freimann Life Sciences animal facility. All studies were approved by the University of Notre Dame Institutional Animal Care and Use Committee (IACUC), under protocol numbers 19–06-5,412 and 22–07-7,335. The Tübingen strain was used for all our zebrafish studies. Embryos were incubated at 28°C in E3 embryo media. Once the animals reached the relevant developmental stage (Westerfield, 1993), they were used for live imaging or anesthetized with 0.02% tricaine and then fixed with 4% paraformaldehyde/1x phosphate buffered saline for gene expression studies (Kimmel et al., 1995).

Morpholinos were obtained from Gene Tools, LLC. The morpholino sequence targeting the zebrafish fxn Exon 3/Exon4 splice site is 5′-CTA​AAT​TTA​CCT​GAG​GTG​ATG​TGC​C-3’. The morpholino was injected into zebrafish embryos at the 1 cell stage and uninjected wild-type siblings were raised as controls. The concentration and volume used are 3 g/L and ∼1.5 nL respectively. The standard control morpholino used was 5′-CCT​CTT​ACC​TCA​GTT​ACA​ATT​TAT​A-3’. The p53 morpholino used was 5’ –AGA​ATT​GAT​TTT​GCC​GAC​CTC​CTC​T-3’.

RNA was extracted from approximately 30 animals from wild-type and morphant groups using TRIZOL Invitrogen (15596026). Next, cDNA was generated using qScript cDNA Supermix (Quantabio). The primers used were: forward 5′- TCA​ATA​AGT​GGT​GGT​AGG​AGT​GTT​T-3′, and reverse 5′-CCA​GTG​AAG​TTT​TCA​TCT​GTG​AGA​T-3’. Our samples were sequenced at the University of Notre Dame Genomics and Bioinformatics Core Facility. The trace files were analyzed using 4Peaks (4Peaks by A. Griekspoor and Tom Groothuis, nucleobytes. com.)

WISH was performed in the same manner as described by previous publications (Chambers B. E. et al., 2020; Chambers J. M. et al., 2020; Marra et al., 2019). The antisense RNA, digoxigenin-labeled probes that were used are cdh17, wt1b, nphs1, slc20a1a, trpm7, trim35-30-201, cetn4, pax2a, jag2b, slc12a1, kcnj1a.1, slc12a3, and gata3. These probes were generated using in vitro transcription from IMAGE clone templates as described by previous publications (Gerlach and Wingert, 2014; Wingert et al., 2007).

Wild-type and fxn morphant animals were incubated in 0.003% phenylthiourea (Sigma-Aldrich, P7629) starting at 8 hpf. Animals were anesthetized at 36 hpf with 0.02% tricaine and subsequently injected with 3 nL of 40 kDa FITC-Dextran (Invitrogen, D-1845) (5 mg/mL). The injection site was around somite three in muscular tissue as described in a previous publication (Kroeger et al., 2017). The animals were then imaged at 6- and 48-hours post injection. The fluorescent change of the entire animal was measured using the FIJI software. The percent change of fluorescent intensity was calculated for each animal by keeping animals separate in a 96-well plate. The change in fluorescent intensity for both groups was analyzed using an unpaired t-test.

Cell death was approximated using AO (Sigma-Aldrich A6014) as described in previous publications (Kroeger et al., 2017; Weaver et al., 2022). 50 mg of AO were dissolved into 50 mL of MilliQ water as a stock solution (100x). The stock solution was stored at −20 C in a light-protected container. AO was diluted down to a 1x concentration in E3 embryo medium before being applied to 24 and 48 hpf animals which had previously been incubated in 0.003% phenylthiourea starting at 8 hpf. Animals were incubated for 30 min at room temperature in the AO/E3 solution and then washed three times over the course of 10 min in E3. Samples were then anesthetized in a tricaine/E3 solution and immediately imaged in methylcellulose.

AB (Thermofisher 05500-5G) staining was conducted as described in a previous publication (Neuhauss et al., 1996; Weaver et al., 2022). Wild-type and fxn morphants were fixed at 4 dpf using 4% paraformaldehyde at 4 C for 16 h. Dehydration was done using 100% methanol at −20 C. The rehydration was done with a series of methanol washes of decreasing concentration. Animals were bleached for 2 h, rinsed in PBST, and then digested using a proteinase K (Roche 3115836001) (10 mg/mL for 15 min). Animals were then rinsed in PBST again and incubated at room temperature on a rocker for 16 h in an Alcian Blue (0.1%), ethanol (70%), and HCl (5%) solution. The animals were then destained on a rocker in an ethanol (70%) and HCL (5%) solution for 8 h. After destaining, the animals were washed in PBST, dehydrated in a series of ethanol washes that increased in concentration, and stored in glycerol prior to imaging.

WISH, AB, AO, and live animal images were all taken on a Nikon Eclipse microscope with a DS-Fi2 camera. Methylcellulose with small amounts of tricaine was used to hold the animals in place for the live images and glycerol was used to hold the fixed animals.

All measurements were taken using the Nikon Elements imaging software on 10x images. Surface area measurements were taken using the auto-select tool. Length measurements were taken using the polyline tool. MCCs were counted manually using the binoculars on the microscope. All of the WISH data were entered into GraphPad Prism and analyzed using unpaired t-tests. AB data was analyzed using the FIJI software. AO images were analyzed by counting fluorescent cells manually. Live, brightfield images were analyzed by quantifying the phenotypes manually.