Perturbed development of calb2b expressing dI6 interneurons and motor neurons underlies locomotor defects observed in calretinin knock-down zebrafish larvae

Calcium binding proteins are essential for neural development and cellular activity. Calretinin, encoded by calb2a and calb2b , plays a role during early zebrafish development and has been proposed as a marker for distinct neuronal populations within the locomotor network. We generated a calb2b :hs:eGFP transgenic reporter line to characterize calretinin expressing cells in the developing spinal cord and describe morphological and behavioral defects in calretinin knock-down larvae. eGFP was detected in primary and secondary motor neurons, as well as in dI6 and V0v interneurons. Knock-down of calretinin lead to disturbed development of motor neurons and dI6 interneurons, revealing a crucial role during early development of the locomotor network. Primary motor neurons showed delayed axon outgrowth and the distinct inhibitory CoLo neurons, originating from the dI6 lineage, were absent. These observations explain the locomotor defects we observed in calretinin knock-down animals where the velocity, acceleration and coordination were affected during escapes. Altogether, our analysis suggests an essential role for calretinin during the development of the circuits regulating escape responses and fast movements within the locomotor network.


Introduction
Calcium binding proteins are well conserved across vertebrates (Kretsinger et al., 1972;Castro et al., 2006;Yokoi et al., 2009) and have two distinct functions: as sensors, translating calcium levels into signalling cascades, or as buffers, decreasing the concentration of free Ca 2+ in the cytoplasm (Skelton et al., 1994).Calretinin, calbindin, and parvalbumin are three calcium buffering proteins that have evolved distinct functional and biophysical properties (Schwaller et al., 2002).
Calretinin is expressed in the central nervous system, intestine, and kidney of various animal groups, including teleost, lamprey, and mammals (Parmentier and Lefort, 1991;Diaz-Regueira and Anadon, 2000;Megías et al., 2003).In developing and adult zebrafish, in situ hybridization against calb2a and calb2b and immunohistochemical analysis of calretinin revealed localization in the peripheral nervous system as well as in the brain and spinal cord (Castro et al., 2006;Levanti et al., 2008;Bhoyar et al., 2017).
Morpholino based knock-down in zebrafish larvae, targeting calb2a or calb2b individually, did not produce any distinct phenotype, suggesting intervention by compensatory mechanisms (Bhoyar et al., 2019).However, targeting both calb2a and calb2b to knock down calretinin, gave rise to a morphological phenotype that included malformation of the midbrain-hindbrain boundary, severe hydrocephalus, edemas, and curvature defects of the tail (Bhoyar et al., 2019).In addition, both swimming and touch-provoked escape responses were affected at 5 days post fertilization (dpf) after which the larvae partially recover from their physical phenotype (Bhoyar et al., 2019).There was no conclusion whether these locomotor defects were due to malformations per se or to disturbances within the spinal locomotor network.
In the zebrafish spinal cord, adjustment of swimming speed relies on a gradual recruitment of a slow, intermediate, and fast locomotor modules.These modules consist of functional subpopulations of interneurons and motor neurons that are active at their respective module's locomotor speed (Ampatzis et al., 2013).Calretinin has been indicated to be a marker for distinct cell populations within the lamprey locomotor network (Megías et al., 2003).In adult zebrafish, calretinin has been suggested to mark the fast module of the locomotor network, being expressed in both fast interneurons and fast motor neurons (Berg et al., 2018).Using single-cell RNA sequencing, we recently revealed two non-overlapping subpopulations within the dmrt3a-dI6 lineage, marked by wt1a and calb2b respectively (Iglesias González et al., 2021).
The inhibitory Dmrt3 interneurons have been shown to play a critical role in the coordination of locomotion in horses, mice and zebrafish (Andersson et al., 2012;Perry et al., 2019;Del Pozo et al., 2020).Absence of dmrt3a in zebrafish larvae gives rise to aberrant escape responses, a similar phenotype to what has been described for calretinin knock-down larvae (Bhoyar et al., 2019;Del Pozo et al., 2020).There is thus need to examine dmrt3a expressing interneurons and fast motor neurons in calretinin knock-down larvae to reveal if aberrant development may underlie the behavioral defects observed.
Here we set out to characterize calb2b expressing neurons of the locomotor network in the zebrafish spinal cord.Through the generation of a transgenic reporter line we provide a detailed description of the calb2b expression pattern.We correlate calb2b expressing cells to markers for V0v and dI6 interneurons and motor neurons during development.Behavioral analysis of calretinin knock-down morphants revealed that larvae swam shorter distances and had reduced maximum velocity and acceleration during free swimming and escape responses.We correlate this behavioral phenotype with aberrant morphology of both dmrt3a-dI6 interneurons and fast motor neurons following calretinin knock-down, highlighting the importance of calretinin for proper development of the locomotor network.

Generation and characterization of a calb2b transgenic reporter line
To characterize calb2b expressing neurons in the zebrafish spinal cord, we generated a Tg[calb2b:hs:eGFP] (from here onwards referred to as calb2b eGFP ) through CRISPR/Cas9-mediated knock-in (Fig. 1A; Auer et al., 2014;Ota et al., 2016).Expression of eGFP was observed in primary motor neurons and interneurons in the spinal cord, brain, retina, and the sensory ganglia of the lateral line (Fig. 1B and C).This is largely in line with in situ data of calb2b mRNA expression that described expression in the eye, brain, and spinal cord, but not in the lateral line ganglia (Bhoyar et al., 2017).We verified the reporter line expression by performing whole-mount immunohistochemistry against calretinin.At 2 dpf, 74% of calb2b eGFP cells (147/199 cells, 6 larvae) were also calretinin positive (Fig. S1).At 3 dpf, 60% of calb2b eGFP cells (183/307 cells, 6 larvae) were also calretinin positive (Fig. 1D).We also found that 61% of calretinin positive cells were calb2b eGFP negative at 3 dpf (281/464, 6 larvae).These likely represent calb2a expressing cells (Bhoyar et al., 2017), as the polyclonal antibody used recognized both variants (Berg et al., 2018).Quantification during spinal cord development revealed 43 ± 6.0 calb2b eGFP positive cells at 2 dpf, 68 ± 2.4 at 3 dpf and 53.3 ± 1.5 at 4 dpf (mean ± SD, cells per segment, n = 4 larvae; Fig. 1E).It should be noted that we found spinal eGFP positive cells at 3 and 4 dpf, in contrast to the in situ study that reported a lack of calb2b expression at 72 hpf (Bhoyar et al., 2017).However, a recent single cell RNA sequencing dataset revealed clear expression of calb2b mRNA in the zebrafish spinal cord at 3, 4 and 5 dpf, corroborating our finding (Sur et al., 2023).Moreover, immunohistochemistry at 6 dpf revealed calretinin positive cells in the zebrafish spinal cord (Fig. S2).

Co-localization of calb2b with markers for interneurons and motor neurons within the locomotor network
To characterize the spinal neurons that express calb2b, we crossed calb2b eGFP to Tg[dmrt3a:Gal4,UAS:RFP] (referred to as dmrt3a RFP ) and performed immunohistochemistry against Wt1a, marking interneurons part of the dI6 lineage.These 3 dpf larvae revealed an overlap between calb2b eGFP and dmrt3a RFP of 6 ± 0.5 cells per segment (Fig. 2A-C, S3A; mean ± SD, n = 4 larvae).Moreover, eGFP expression did not overlap with Wt1a-positive cells (0 ± 0; Fig. S3B).These observations regarding the dI6-lineage are in line with what we previously have shown through single cell RNA sequencing (Iglesias González et al., 2021), confirming calb2b expression in a subpopulation of dmrt3a expressing dI6 neurons.More detailed analysis showed that expression of eGFP was observed in the distinct axons of commissural local (CoLo) neurons, marked by dmrt3a RFP , which are involved in fast escape responses by inhibiting contralateral primary motor neurons (pMN; Fig. 2A; Satou et al., 2009).The V0v interneuron population, uniquely marked by Evx2 in zebrafish, is excitatory and help drive motor output (McLean et al., 2007;Satou et al., 2020).Immunoreactivity showed co-expression between Evx2 and calb2b eGFP positive cells (7.8 ± 1.2, cells per segment, n = 4 larvae, Fig. 2C, S3C).
Our results support findings by Berg et al. (2018), who proposed calretinin as a potential marker for motor neurons involved in escape responses and fast swims and interneurons part of the fast speed module.Our data show that among motor neurons, calb2b is primarily expressed in pMNs, which contribute to the escape response (Fig. 2D).The calb2b eGFP reporter line also marked the dmrt3a expressing CoLo neurons, further indicating involvement of the calretinin expressing population in escape behaviors.However, as there is only a single CoLo neuron per hemisegment (Satou et al., 2009), calb2b is also likely expressed in inhibitory dmrt3a neurons recruited during fast swim frequencies, while no expression was observed in the later born Wt1a positive cells involved in slower swims (Kishore et al., 2020;Iglesias González et al., 2021).In addition, we observed calb2b expression in the excitatory interneuron population V0v, which is reportedly involved in the locomotor output during slower movements (McLean et al., 2007).Thus, calb2b appears to be expressed in a range of spinal neurons involved during different motor behaviors.

Calb2b knock-down larvae showed developmental changes in dmrt3a-dI6 and Evx2 interneurons but did not display a behavioral phenotype
We investigated the outcome of targeting calb2a and calb2b individually using splice blocking morpholinos, verified in Bhoyar et al. phenotype (Fig. 3A, S4A).We did not observe any significant differences in cell number or cellular morphology in calb2a morphants (Fig. S4B-E).For calb2b morphants, there was a reduction, but not a complete lack, of calretinin positive cells (11.2 ± 0.8, p < 0.0001, n = 4) compared to controls (29.7 ± 1.7, n = 4; Fig. 3B-C).The remaining calretinin positive cells are likely a consequence of continued calb2a expression.Knock-down of calb2b resulted in a small decrease in the amount of dmrt3a RFP cells (36.0 ± 1.4, p = 0.01) compared to controls (41.5 ± 2.6) but there was no effect on the number or morphology of mnx1 RFP cells (23.9 ± 1.8, p = 0.99) compared to controls (24.0 ± 1.5; Fig. 3D-G).Additionally, a reduction of Evx2 positive cells was observed in calb2b morphants (44.3 ± 5, p = 0.028) compared to controls (52.6 ± 2.5; Fig. 3H-I).Cells double labelled for Evx2 and eGFP were absent in calb2b morphants (0 ± 0; p < 0.0001) compared to controls (4 ± 0.4).To investigate if the knock-down of calb2a or calb2b affected behavioral output, we analyzed free swimming and escape response performance at 4 and 6 dpf.There were no differences in any of the parameters analyzed, such as mean velocity and acceleration, at either 4 or 6 dpf when compared to control larvae (Tables 1 and 2, Fig. S4F-G).
This indicates that targeting calb2a or calb2b individually does not generate a morphological phenotype in larvae or a behavioral phenotype, possibly due to compensatory effects from the remaining variant encoding for calretinin.This is in spite of the mild cellular effect on interneurons in calb2b morphant larvae.However, it has previously been reported that a combination, targeting both calb2a and calb2b, causes a strong phenotype with behavioral consequences (Bhoyar et al., 2019).

Lack of calretinin results in developmental defects in dmrt3a-dI6 interneurons
Knock-down of both calb2a and calb2b (hereafter referred to as calretinin knock-down) resulted in a gross morphological phenotype similar to that previously described (Bhoyar et al., 2019): hydrocephalus, a defect in the axial curvature of the body and pericardial edema (Fig. 4A).A curved or malformed back is a common feature when affecting the proliferation and differentiation of the spinal cord (Schier et al., 1997;Karlstrom et al., 2003;Odenthal et al., 1996).The calretinin knock-down successfully inhibited production of the calretinin protein as there were no calretinin positive cells in morphants (0 ± 0 cells per segment), compared to controls (24 ± 3.2 cells per segment) (p < 0.0001, n = 4 larvae; Fig. 4B).Calretinin knock-down in dmrt3a RFP larvae revealed a decrease in the number of RFP positive cells (31 ± 2.5, p < 0.0001) compared to controls (44 ± 3.3) at 3 dpf (Fig. 4C, E).Moreover, the axons of the distinct CoLo dmrt3a-subtype were absent in calretinin morphants (0 ± 0 cells per segment; Fig. 4C), where two per section are expected (Satou et al., 2020).
While the majority of knock-down animals retained the physiological and cellular defects throughout the experiment, a portion showed signs of recovery, which may be caused by the transient nature of morpholinos.At 6 dpf, 9% of the animals (15 of 160), which all displayed the characteristic phenotype at 3 dpf, had significantly recovered and by 10 dpf this number was 12% (19 of 160; Fig. 4D).In these physically recovered morphants, we found similar numbers of dmrt3a expressing interneurons compared to controls at 6 dpf (41.2 ± 0.7 vs 42 ± 1.8, p = 0.61, n = 4 larvae), indicating a cellular recovery (Fig. 4D, F).Moreover, the distinct axons of CoLo neurons were also clearly visible at 6 dpf (Fig. 4D).
To further support the morpholino knock-down and our finding that the CoLo neurons are missing in these animals we generated F0 crispants.Here CRISPR/Cas9 and gRNAs targeting calb2a and calb2b were injected at single cell stage to introduce indels in their coding regions.Just below 10% (at 2dpf) and 5% (at 3dpf) of larvae displayed a morphological phenotype, which was consistent between animals.The phenotype was more severe than the morpholino knock-down with a severely disturbed development of the spinal cord preventing the identification of CoLo neurons (Fig. S5A).However, some larvae displayed a milder phenotype, where only parts of the animal were affected, likely as a consequence of mosaic knock-out (Fig. S5A).Here, the spinal cord development appeared more normal but, despite there being a similar amount of Dmrt3 neurons present (36.5 ± 1.5 per segment), the distinct axons of the CoLo neurons were not visible (Fig. S5A).These results are in line with the morpholino knock-down, indicating the importance of calretinin for normal development, in particular regarding CoLo neurons.
Our morpholino calretinin knock-down replicated the morphological larval defects observed by Bhoyar et al. (2019), but we also found effects on the dmrt3a lineage, including the CoLo neurons, in these morphants, indicating a developmental effect on the locomotor network.
These data regarding motor neurons further exemplifies a perturbed development of the locomotor circuit in calretinin knock-down larvae.Combined, the effects on dI6 interneurons, with the notable lack of CoLo neurons, pMN and sMN in calretinin knock-down larvae could explain the aberrant touch induced escape responses previously described (Bhoyar et al., 2019).

Calretinin knock-down larvae display aberrant behaviors during swimming and escape responses
To study behaviors in calretinin knock-down larvae we performed free swimming and escape triggered behavioral tests.At 4 dpf, we did not find significant changes in free swimming activity, however, for tap evoked escape responses, we observed a reduction in distance moved, maximum acceleration and velocity (Tables 1 and 2, Fig. 6A-B).To reveal if the physiologically recovered morphants displayed any locomotor phenotype we also assessed locomotion at 6 dpf and 10 dpf.Calretinin morphant larvae that did not display physical recovery were unable to perform swims and were therefore not quantified (data not shown).At 6 dpf, there were significant reductions regarding total distance moved, maximum acceleration, mean velocity and mean time accelerating during free swimming (Table 1, Fig. 6A).For escape responses, parameters such total distance, maximum acceleration and maximum velocity were significantly lower than in controls (Table 2, Fig. 6B).It seems that despite physical and cellular signs of recovery at 6 dpf, a behavioral phenotype remains at this time-point.At 10 dpf there appeared to be extensive recovery in calretinin knock-down larvae as few parameters during free swimming and no parameters during escape responses differed from control larvae (Tables 1 and 2, Fig. 6A-B).
To complement our observations of free-swimming calretinin knockdown larvae after physical phenotype recovery, we performed fictive locomotion recordings at 5-6 dpf.We observed that in routine swims, i. e. spontaneous swims and swims elicited by optomotor response, calretinin knock-down larvae showed a shift in swim output frequency (Fig. 6C).When comparing the distribution of median swim frequency per animal per group, we observed a narrowing of the locomotor speed bandwidth in knock-down larvae (Fig. 6C).Calretinin knock-down larvae were less prone to swim at output frequencies over 22.5 Hz and instead showed an increased tendency to swim in the 17.5-22.5Hz range.Morphant swims manifested a number of qualitative deviations; Control animals generally display locomotor outputs with consistent counter-phase left-right activity and symmetry, meaning that the spike burst duration and amplitude in the left channel is somewhat predictive of spike burst duration and amplitude in the consecutive burst in the right channel (Fig. 6D).This was not the case for calretinin morphant swims, which displayed simultaneous contralateral bursts (Fig. 6E), consecutive bursts of varying durations and amplitude (Fig. 6F), and swims with a general low coherence in burst frequency and prevalence (Fig. 6G).Due to the heterogeneity of locomotor artefacts in the calretinin knock-down larvae we turned to the more stereotyped escape swim, also known as a C-start, elicited by electrical stimulation of tip of the larvae's tail.The escapes consist of an initial fast swim (>30 Hz) of roughly 100ms followed by slow phase (<30 Hz; Fig. 6H-I).We observed that in calretinin knock-down larvae, the output frequency during the fast phase was reduced to frequencies similar to those observed during the slow phase (P < 0.05; Fig. 6I).Additionally, escape swims of calretinin knock-down larvae were significantly shorter than those of control larvae (P < 0.05; Fig. 6J).Calretinin morphant C-start escapes also differed qualitatively from the controls.In controls, 90% of all recorded escapes matched the C-start profile as seen in Fig. 6H, while only 7% of recorded escapes matched this profile for calretinin knockdown larvae.The remaining calretinin knock-down larvae escapes instead consisted of what we classified as 'short lateralized' swims (43%) that displayed short duration and primarily had spiking on one side of the tail (Fig. 6K).Another 37% were categorized as 'less coherent' swims, which showed one or multiple output defects as described for regular swims, including instances where spiking activity in the supposed fast phase (<100 ms) was either absent (Fig. 6L) or vastly reduced (Fig. 6M).
Our analysis presents detailed information regarding the locomotor parameters affected in calretinin knock-down larvae.We also showed that the developmental recovery observed on a cellular level was trailed by a behavioral recovery regarding locomotor performance.The locomotor defects described by Bhoyar et al. (2019), an affected escape behavior, likely originates from the absence of the dmrt3a expressing CoLo neurons, responsible for contralateral inhibition during escape responses (Satou et al., 2009), and the delayed innervation of musculature by primary motor neurons.However, we found that the number of dmrt3a expressing cells, including CoLo neurons, and motor neurons in calretinin knock-down larvae recovered to match the numbers in control larvae at 6 dpf.A similar recovery was observed for axonal projections as CoLo axons were present and muscular innervation of motor neuron axons resembled controls in 6 dpf calretinin knock-down larvae.The recovery may be due to the transient nature of morpholinos, and we did detect calretinin again in 6 dpf recovered larvae (Fig. S6).However, calretinin was also detected in non-recovered larvae (Fig. S6), suggesting that the return of calretinin alone does not guarantee recovery.Instead phenotype severity may play a crucial role in chance of recovery.In addition, although motor neurons appeared recovered, we did note a lack of calb2b positive motor neuron projections in 6 dpf morpholino treated larvae, suggesting that possible compensatory mechanisms allow the developmentally delayed knock-down larvae to catch up to the control larvae.While analysis of mutant larvae could clarify this uncertainty, it should be noted that the majority of larvae did not fully recover from the morphological defects.Rather, due to the severely bent tails, these larvae had difficulties to perform any swims (data not shown) and where thusly not subjected to behavioral analysis.Despite the cellular recovery of calretinin knock-down larvae by 6 dpf, the behavioral phenotype was prominent both during free swimming and tap evoked escapes at this age, indicating that the locomotor network was not yet fully operational.Analysis at 10 dpf did not reveal any locomotor phenotype during escape responses, and only mild effects were detected during free swimming, indicating continued recovery.Further behavioral experiments, at juvenile stages, would reveal if locomotor performance is eventually completely restored.

Concluding remarks
In this study we confirm that calb2b is a promising marker for distinct neuronal populations within the locomotor circuitry, as previously suggested, by generating a transgenic reporter line (Megías et al., 2003;Berg et al., 2018).We verified our previous transcriptomic analysis of zebrafish larvae, describing expression of calb2b within the inhibitory dmrt3a-dI6 lineage (Iglesias González et al., 2021), and investigated the distribution within primary and secondary motor neurons.Our data suggest that the lack of calretinin during early development has a strong impact on the formation of the locomotor network, resulting in aberrant motor performance in zebrafish larvae.During development, the circuitry responsible for escape responses is the first to establish and it is coordinated by primary motor neurons and CoLo neurons, both types that showed developmental defects in calretinin knock-down larvae.In conclusion, calretinin seems to play a vital role during the formation of the spinal circuitry and in its absence, there is developmental delay, resulting in disturbed motor output.

Generation of Tg[calb2b:hs:eGFP]
To generate Tg[calb2b:hs:eGFP] by CRISPR/Cas9 knock-in methodology, we used the following genomic sequences to target calb2b (GGCTTCAGCAGTTGCGGTGG).The targeted sequence is upstream of the UTR for calb2b and should not interfere with normal expression of calb2b, however, heterozygous fish were used to avoid this uncertainty.Fertilized zebrafish eggs from AB fish were obtained in natural crosses, injected at the one-cell stage into the cell with 110 pg of Cas9 mRNA, 250 pg Cas9 protein (TrueCut™ Cas9 Protein V2, Invitrogen), 50 pg of sgRNA for gene specific target and 50 pg of mbait sgRNA and 1,5 pg of the mbait-hs-eGFP plasmid (kindly provided by Professor Atsuo Kawahara (Ota et al., 2016).eGFP positive larvae were raised and outcrossed with AB confirm for germline transmission.Fluorescent larvae were raised to adulthood and offspring of founders were used to establish a stable transgenic line.

Morpholino injections
To generate calb2a, calb2b, and calb2a, calb2b double knock-down larvae (morphants), we used splice blocking morpholino antisense oligonucleotides (0.3 mM, Gene Tools) as previously described in Bhoyar et al. (2019).These morpholinos were thoroughly validated in Bhoyar et al. (2019) through mispaired oligos, RT-PCR to visualize intron retention, and mRNA rescue.We used a standard control morpholino (5′-CCTCTTACCTCAGTTACAATTTATA-3′) to assess side-effects from the injections.Morpholinos were dissolved in MilliQ water to a final concentration of 1 mM and 1 nL was injected into the yolk of one-cell stage zebrafish eggs.As larvae injected with the calb2a, calb2b combination experienced a double load of morpholino, which may result in toxic effects, we also injected the individual morpholinos at double concentration as a control.There was no increase in phenotype for larvae subjected to double amount of the individual morpholinos.

CRISPR sgRNA design and synthesis
Two CRISPR targets were designed for both calb2a (NM_200718-1) and for calb2b (NM_200711), using ChopChop.We designed the following sgRNA sequences for calb2a: AGCTCCGCCAAGTGAAGGTA (ZNN-66, exon 1) and TGACAACTTCTTTAGTCAGC (ZNN-67, exon 2), and calb2b: GCCCATTTCTCTGCATTTGG (ZNN-68, exon 1) and GGAAAATTTCATCCGTGAGC (ZNN-69, exon 2).To enhance sgRNA synthesis by T7 polymerase used in the oligo assembly approach (Varshney et al., 2015;Habicher et al., 2022), we altered the underscored nucleotide to a G, so all guides start with 'GG'.Oligos were ordered with a T7 promoter at the 5′ of the target sequence and a DNA stretch overlapping with the guide core sequence at the 3 ′ end.These oligos were then annealed with a second fragment containing the guide core sequence (oligo B).Assembled oligos were then used as a template for RNA in vitro transcription (HiScribe T7 High Yield RNA Synthesis Kit, NEB).Generated RNA was then purified using the GeneJET RNA Cleanup and Concentration Micro Kit (Thermo Fisher Scientific, Waltham, MA, USA).

Injections of sgRNA
Fertilized zebrafish eggs from Tg[dmrt3:GAL4,UAS:GFP] fish were obtained in natural crosses, and injected at the one-cell stage into the cell with: 110 pg of Cas9 mRNA, 250 pg Cas9 protein (TrueCut™ Cas9 Protein V2, Invitrogen), and 50 pg for each of the four sgRNAs (200 pg sgRNA total).

Assessing sgRNA genome editing by fragment length analysis (FLA)
DNA of injected (phenotype displaying) and control embryos was extracted at 3 dpf by dissolving tissue in 30 μl NaOH (50 mM) for 5 min in 95 • C. Next, 6 μl Tris-HCl (500 mM) was added to neutralize the pH.
PCR products were obtained via a 3-primer-PCR using sgRNA specific forward primers and reverse primers (Fig. S5B), and an M13 forward primer (TGTAAAACGACGGCCAGT) fluorescently labelled with 6-FAM (Varshney et al., 2015;Habicher et al., 2022).PCR products were analyzed by fragment separation using capillary electrophoresis (known as FLA), as previously described (Sood et al., 2013).Size determination was carried out on a 3130XL ABI Genetic Analyzer (Applied Biosystems, Waltham, MA, USA) and the data was analyzed using the Peak Scanner Software (Thermo Fisher Scientific).Fragment lengths were compared between sgRNA injected larvae and non-injected control larvae and revealed multiple peaks in the crispants (Fig. S5C).

Whole mount immunohistochemistry
Zebrafish embryos were housed at 28 • C until 3 dpf in embryo water, supplemented with PTU (0.003%) at 1dpf to prevent pigmentation.Embryos were anesthetized with 0.1% of MS-222 and fixed in 4% paraformaldehyde (PFA) for 15 min at room temperature (RT) followed by 3 × 10 min washes in 0.2% PBS-Triton (1xPBS at pH 7.3, 0.2% Triton X-100).Embryos were cryoprotected in 30% sucrose in PBS at RT for 2h and equilibrated in 0.2% PBS-Triton for 3 × 10 min and treated with acetone for 20 min at − 20 • C. Embryos were subsequently washed in 0.2% PBS-Triton 2 × 5 min, 2 × 5 min in Milli-Q water and in 0.2% PBS-Triton 2 × 5 min.Nonspecific protein binding sites were blocked with 1% BSA in 0.2% PBS-Triton.After blocking, embryos were incubated with the primary antibody in the blocking buffer (mouse anti-Wt1 1:100 (Dako), rabbit anti-Evx2 1:300 (Higashijima Lab), and rabbit anti-Calretinin 1:500 (Swant Lab)) at 4 • C. Following incubation, larvae were washed for 2 × 1h in PBS-T and incubated with their respective secondary antibodies (Alexa Fluor 647 donkey anti-mouse) in the dark at 4 • C overnight.Finally, the tissue was rinsed in PBS-T for 2 × 1h and imaged in a Leica SP8 DLS microscope.All incubations were performed in an incubation box (patent pending, application number: 2151279-3) allowing the samples to be washed and incubated horizontally and in darkness.

Confocal imaging and analysis
For imaging, embryos were anesthetized and mounted in 1% low melting agarose.Confocal (25× water objective) images from live animals and fixed tissue were collected using the Leica SP8 confocal microscope.Quantification of cells was performed using the ImageJ plugins CellCounter and Single Neurite tracer for tracing of motor neurons.Analysis of the time-lapses were performed in the Leica's LasX software.

Behavioral experiments
Behavioral experiments were performed as described previously (Del Pozo et al., 2020), using larvae at 4, 6 and 10 dpf.The protocol established for the free-swimming analysis consisted of a habituation period of 20 min followed by 50 min of free-swimming under constant conditions.This period was followed by a sequence of five taps at 3 min intervals to elicit escape responses.Behavior was assessed under darkness.

Fictive swimming
To assess subtle changes in motor output, 5-6 dpf calretinin morphants (n = 9) and control larvae (n = 4) were subjected to dual channel fictive locomotion recordings as described previously (Koning et al., 2022).Briefly, larvae were immobilized by submersion in 0.04% tricaine before injection of α-bungarotoxin (Almone labs) into the pericardium.Tricaine was washed out by bathing the larvae in embryo water for 15 min while α-bungarotoxin took effect.Larvae were then placed in the lid of a 35 mm Petri dish filled with extracellular recording solution (in mM; NaCl, 134; KCl, 2.9; CaCl 2 , 2.1; MgCl 2 , 1.2; HEPES,10; glucose,10, pH = 7.8, 290-300 mOsm/mL) (Drapeau et al., 1999).Immobilized larvae were mounted with tissue adhesive (Vetbond M3) at the lower jaw and tip of the tail.The skin covering segments 8-15 was removed with fine forceps to expose the axial trunk musculature.Electrodes with large tip opening (50-70 μM) were filled with extracellular recording solution and placed on the intermyotomal cleft on both sides of a muscle segment (segment 10-15) and slight negative pressure was applied, fictive motor signal appeared after 5-10 min.Optomotor response was used to elicit fictive swimming episodes where a black and red striped grating moved in an anterior-posterior direction at 5, 20 or 40 mm/s (alternating between 10 s static and 2 s moving), projected (Asus ZenBeam e1) onto a diffuser screen placed under the dish.For escape swim recordings an additional concentric bipolar electrode (World precision instruments) was placed at the tip of the larvae's tail.Nine sets of three 10 ms electric pulses (70-100 mA) were delivered at 6-16 s intervals through an analogue stimulus isolator (AM-systems model 2200).All electrophysiological activity was recorded in current clamp at 50 ks/s through an INTAN technologies CLAMP amplifier.
Electrophysiology recordings were processed in a custom Matlab script.Recording traces were filtered using a continuous wavelet transform and peak-finding was applied to find clusters of spikes to sort out periods of fictive swimming activity automatically.For escape swim recordings, the stimulation time points were used to extract swim episodes.The resulting swim libraries were used to extract fictive swim parameters.Additionally, each fictive swim trace was fitted with an envelope filter, and by combining the bilateral envelope traces a fictive movement trace could be formed (Koning et al., 2022).Said fictive movement trace was used to extrude time-frequency information with high temporal resolution.

Statistical analysis
Statistical analyses were performed using Prism 8 for Windows (GraphPad Software, La Jolla, USA).Firstly, data were tested for normality and then the appropriate parametric or non-parametric test was selected for analysis.Number

Statement of financial interest
The authors declare no competing financial interests.

Fig. 1 .
Fig. 1.Overall characterization of calb2b eGFP transgenic line.A) Dorsal and lateral light-sheet images of the calb2b eGFP transgenic line at 3 days post fertilization (dpf).B) Dorsal and lateral view of the spinal cord at 3 dpf.calb2b eGFP positive primary and secondary motor neurons labelled by are indicated by white arrowheads.C) Confocal image with details of the lateral line from head to spinal cord at 3 dpf.D) Whole mount immunohistochemistry against calretinin in calb2b eGFP larvae at 3 dpf.E) Quantification of eGFP-positive cells in calb2b eGFP larvae from 2 to 4 dpf.Scale bar, 100 μm in A and B; 50 μm in C and D.
Fig. 2. Colocalization of calb2b eGFP with dI6-dmrt3a and motor neurons.A) Dorsal view of calb2b eGFP ; dmrt3a RFP cell bodies and fiber layer at 3 days post fertilization (dpf).Arrowheads indicate overlapping cells and arrows the axon of the CoLo subtype.B) Lateral view of calb2a eGFP ; mnx1 RFP at 3 dpf.Asterisks mark primary motor neurons and arrowheads mark secondary motor neurons.C) Percentage of calb2b eGFP that are colocalized with primary and secondary motor neurons, Evx2 and dI6-dmrt3a interneurons.D) Quantification of calb2a eGFP ; mnx1 RFP motor neurons by position in the dorso-ventral axis and soma size at 3 dpf.pMNprimary motor neuron, sMNsecondary motor neuron.Scale bar, 50 μm in A, B, C, D and G; 25 μm in zoom boxes in A, B and D.

Fig. 3 .
Fig. 3.A minor cellular phenotype in calb2b knock-down larvae.A) Brightfield images of control and calb2b knock-down larvae reveal no physical phenotype at 3 days post fertilization (dpf).B) Immunohistochemistry against calretinin at 3 dpf show a significant diminished number of calretinin positives cells.C) Quantification of calretinin positive cells in calb2b knock-down larvae compared to controls.D) Confocal images of dmrt3a RFP in control and calb2b knock-down larvae.E) Quantification of dmrt3a RFP cells in control larvae and calb2b knock-down larvae.F) Confocal images of mnx1 RFP in control and calb2b knock-down larvae.G) Quantification of mnx1 RFP cells in the control larvae and calb2b knock-down larvae.H) Confocal images of Evx2 positive cells in control and calb2b knock-down larvae.I) Quantification of Evx2 positive cells in control and calb2b knock-down larvae.Scale bar, 100 μm in all images.Statistical differences: >0.05 (ns), ≤0.05 (*), ≤0.01 (**), ≤0.001 (***) and ≤0.0001 (****).

Fig. 6 .
Fig. 6.Analysis of locomotor behavior in calretinin knock-down larvae during swims and escapes.A) Parameters indicating swimming performance, (total distance, maximum acceleration, and mean velocity, at 4, 6 and 10 days post fertilization (dpf) during the free swimming in controls and calretinin knock-down larvae.B) Distance moved, maximum acceleration and maximum velocity during escape responses at 4, 6 and 10 dpf in freely moving control and calretinin knock-down larvae.C) Probability density distributions of fictive swim frequency in spontaneous and optomotor response evoked swims in 5-6 dpf calretinin knockdown larvae (N = 9, n = 1323) and control larvae (N = 6, n = 788).D) Example of a regular swim of a control larvae.E) Example of simultaneous contralateral bursting in a calretinin knock-down larvae.F) Example of inconsistent burst duration and amplitude in a calretinin knock-down larvae.G) Example of a generally incoherent swim in a calretinin knock-down larvae.H) Representative example of a stereotyped C-start escape swim in a control larvae consisting of a 100ms fast phase (>30Hz) followed by a slow phase (<30Hz).I) Calretinin knock-down larvae (N = 7, n = 135) showed a lack of fast swim speeds during the initial phase of a Cstart escape compared to control larvae (N = 4, n = 86) (P < 0.005).J) Duration of C-start escape swims is shorter for calretinin knock-down larvae (P < 0.01).K) Example of qualitatively altered escape swim in calretinin knock-down larvae consisting of a short period of strongly lateralized activity.L-M) Examples of calretinin knock-down larvae escapes that show no (L) or reduced (M) activity during the supposed fast phase of the escape swim.Statistical differences: >0.05 (ns), ≤0.05 (*), ≤0.01 (**), ≤0.001 (***) and ≤0.0001 (****).