Cochlea-Specific Deletion of Cav1.3 Calcium Channels Arrests Inner Hair Cell Differentiation and Unravels Pitfalls of Conditional Mouse Models

Inner hair cell (IHC) Cav1.3 Ca2+ channels are multifunctional channels mediating Ca2+ influx for exocytosis at ribbon synapses, the generation of Ca2+ action potentials in pre-hearing IHCs and gene expression. IHCs of deaf systemic Cav1.3-deficient (Cav1.3-/-) mice stay immature because they fail to up-regulate voltage- and Ca2+-activated K+ (BK) channels but persistently express small conductance Ca2+-activated K+ (SK2) channels. In pre-hearing wildtype mice, cholinergic neurons from the superior olivary complex (SOC) exert efferent inhibition onto spontaneously active immature IHCs by activating their SK2 channels. Because Cav1.3 plays an important role for survival, health and function of SOC neurons, SK2 channel persistence and lack of BK channels in systemic Cav1.3-/- IHCs may result from malfunctioning neurons of the SOC. Here we analyze cochlea-specific Cav1.3 knockout mice with green fluorescent protein (GFP) switch reporter function, Pax2::cre;Cacna1d-eGFPflex/flex and Pax2::cre;Cacna1d-eGFPflex/-. Profound hearing loss, lack of BK channels and persistence of SK2 channels in Pax2::cre;Cacna1d-eGFPflex/- mice recapitulated the phenotype of systemic Cav1.3-/- mice, indicating that in wildtype mice, regulation of SK2 and BK channel expression is independent of Cav1.3 expression in SOC neurons. In addition, we noticed dose-dependent GFP toxicity leading to death of basal coil IHCs of Pax2::cre;Cacna1d-eGFPflex/flex mice, likely because of high GFP concentration and small repair capacity. This and the slower time course of Pax2-driven Cre recombinase in switching two rather than one Cacna1d-eGFPflex allele lead us to study Pax2::cre;Cacna1d-eGFPflex/- mice. Notably, control Cacna1d-eGFPflex/- IHCs showed a significant reduction in Cav1.3 channel cluster sizes and currents, suggesting that the intronic construct interfered with gene translation or splicing. These pitfalls are likely to be a frequent problem of many genetically modified mice with complex or multiple gene-targeting constructs or fluorescent proteins. Great caution and appropriate controls are therefore required.


INTRODUCTION
The L-type calcium (Ca 2+ ) channel Ca v 1.3 is the main voltagegated Ca 2+ channel (VGCC) in inner hair cells (IHCs) and essential for hearing (Platzer et al., 2000;Baig et al., 2011). In both pre-hearing and mature IHCs, voltage-activated Ca v 1.3 channels trigger glutamate release resulting in signal transmission to the auditory nerve (Brandt et al., 2003). Before the onset of hearing at postnatal day 12 in mice, IHCs produce spontaneous Ca 2+action potentials (Kros et al., 1998;Platzer et al., 2000;Marcotti et al., 2003) required for the terminal differentiation of IHCs (Brandt et al., 2003;Nemzou et al., 2006;Johnson et al., 2013a) and maturation of the auditory brainstem (Tritsch and Bergles, 2010;Clause et al., 2014;Babola et al., 2018). IHC spontaneous activity is modulated by transient efferent input originating in the superior olivary complex (SOC), which activates smallconductance SK2 potassium (K + ) channels and thereby causes hyperpolarization of the IHC membrane potential (Glowatzki and Fuchs, 2000;Oliver et al., 2000). Around the onset of hearing, IHCs loose their efferent input (Simmons, 2002), SK2 channels are down-regulated (Marcotti et al., 2004) and spontaneous activity ends with the up-regulation of BK and KCNQ4 K + channels (Kros et al., 1998;Oliver et al., 2003). IHCs of systemic Ca v 1.3 −/− mice fail to acquire a mature composition of K + channels (Brandt et al., 2003;Nemzou et al., 2006), which is likely caused by lack of spontaneous activity and impaired Ca 2+dependent transcriptional regulation. However, altered efferent modulation due to lack of Ca v 1.3 in brainstem nuclei might add to the phenotype. Ca v 1.3 plays an intrinsic role for development and function of SOC neurons (Hirtz et al., 2011(Hirtz et al., , 2012Satheesh et al., 2012) and is therefore regarded not only as a peripheral but also a central deafness gene (Willaredt et al., 2014).
Here, the effects of cochlea-specific ablation of Ca v 1.3 channels before birth on the electrophysiological and molecular phenotype of IHCs as well as hearing function were investigated. To this end, Cacna1d-eGFP flex mice were used, in which the ablation of Cacna1d encoding Ca v 1.3 channels is directly coupled to the expression of the reporter gene eGFP via Cre-induced inversion ("switch") of the floxed allele (Satheesh et al., 2012). They were crossed with Pax2::cre mice (Ohyama and Groves, 2004), where Cre expression is initiated at E9.5 in the otic vesicle (Lawoko-Kerali et al., 2001;Burton et al., 2004) and found in the mature organ of Corti and spiral ganglion neurons (SGN) but not in the nuclei that are part of the afferent-efferent feedback loop onto hair cells, i.e., ventral cochlear nucleus and the SOC .

Animals
Cacna1d-eGFP flex mice were generated within the CavNET consortium (EU-CAVNET MRTN-CT-2006-035367) by Katrin Bartels née Kunert, Kai Schönig and Dusan Bartsch, Central Institute of Mental Health, Mannheim, Germany (Satheesh et al., 2012). They were cross-bred with Cacna1d −/− mice (Platzer et al., 2000) and Pax2::cre mice (Ohyama and Groves, 2004;Zuccotti et al., 2012). To reduce the risk of unwanted effects caused by Cre expression, only mice heterozygous for Pax2::cre were used (Jae Huh et al., 2010;Janbandhu et al., 2014). Animals were housed with free access to food and water at an average temperature of 22 • C and a 12 h light-dark cycle. Mice of either sex were sacrificed by decapitation under isoflurane anesthesia and their cochleae were dissected from the temporal bones. All experiments were carried out in accordance with the European Communities Council Directive (86/609/EEC) and approved by the regional board for scientific animal experiments of the Saarland, Germany. Additional ethics approval was not required according to the local and national guidelines.

Hearing Measurements
Auditory brainstem responses (ABR) and distortion product otoacoustic emissions (DPOAE) were recorded in anesthetized mice aged 4-6 weeks as described in Fell et al. (2016). Growth functions of ABR waves I to IV in response to click stimuli were analyzed for peak-to-peak amplitudes and latencies between the click stimulus delivered at t = 0 and the time point of the negative peak of the respective wave.
Before performing whole-cell patch clamp recordings using an Optopatch (Cairn Research, United Kingdom) or an Axopatch 200B amplifier (Molecular Devices, United States), green fluorescent protein (GFP) fluorescence of the specimen was assessed with an epifluorescence system consisting of a UV lamp and FITC fluorescence filters attached to the patch microscope (Olympus BX51WI with a 40 x water immersion objective, Germany) and a CCD camera (Scientifica, United Kingdom). Ba 2+ currents were elicited by depolarizing the cells for 8 ms from -98 to +48 mV in 5 mV increments. Uncompensated series resistance was corrected by 70-80%. Analysis, including off-line linear leak subtraction and correction of the currents by subtracting the liquid junction potential of 8 mV, was performed using Igor Pro software (Wavemetrics, United States). I-V relations were calculated as the average current taken from the last ms of the voltage step as a function of the respective voltage.
I-V curves of Ba 2+ currents were fitted to a second-order Boltzmann function times Goldman-Hodgkin-Katz driving force to determine parameters of activation, the voltage of halfmaximum activation, V h , and the voltage sensitivity of activation determined by the slope factor k, according to (1) where I is I Ba at the time point the I-V was calculated (average over 7-8 ms after depolarization); P max the maximum permeability; ν = zFV/(RT), with z being 2, F the Faraday constant, R the universal gas constant, T the absolute temperature, V the membrane potential.
[Ba] i (set at 50 nM) and [Ba] o denote the intra-and extracellular Ba 2+ concentration, respectively.
For quantification of Ca v 1.3 clusters and RIBEYE-positive ribbons, images of 67.5 µm × 38.9 µm size covering eight IHCs were acquired at equal laser and gain settings, and maximum intensity projections (MIPs) were calculated. The channel of interest of a MIP image was background subtracted. A thresholded binary image was created (0 below threshold; 1 above threshold) with thresholds of 10% of the maximum intensity of the green color channel (RIBEYE) and 17% of the red color channel (Ca v 1.3). Fluorescent dots < 0.05 µm 2 were discarded. Size and number of clusters were analyzed using the particle count routine in Fiji and normalized to one IHC.

Statistics
Data are provided as mean ± SD, unless otherwise stated. Depending on the distribution of the data, Ba 2+ current properties, as well as size and number of Ca v 1.3 clusters and ribbons were statistically analyzed using Student's t-test or Mann-Whitney U test (MWU test; comparison of two groups) or using one-way ANOVA followed by Tukey post hoc test or Kruskal-Wallis test followed by Dunn-Holland-Wolfe post hoc test (comparison of > 2 groups) with Igor Pro software (WaveMetrics, United States) and SPSS statistics (IBM, Germany).
Statistical analysis of hearing measurements was performed with SPSS. Click ABR thresholds were analyzed using a oneway ANOVA, DPOAE amplitudes with a Kruskal-Wallis test and frequency-dependent ABR thresholds with a two-way ANOVA; all tests were followed by a Bonferroni post hoc test. ABR growth functions of amplitudes and latencies could not be tested by a two-way ANOVA due to unequal variances. Instead, a regression analysis was performed, and the parameters of the resulting regression lines (slope and y-axis intercept) were tested for differences using Student's t-test or MWU test according to Sachs (1999). In order to assess the phenotype of mice with cochlea-specific ablation of Cacna1d before birth, we analyzed GFP signals and whole-cell Ba 2+ currents through Ca v 1.3 channels in IHCs of conditional knockout (cKO) Pax2::cre;Cacna1d-eGFP flex/flex mice, in short cKO-Ca v 1.3 flex/flex . GFP fluorescence of two distinct intensity levels was present in IHCs of the apical cochlear turn acutely dissected from 3-week-old cKO-Ca v 1.3 flex/flex mice ( Figure 1B) but not in wildtype IHCs ( Figure 1A). Analysis of Ba 2+ currents (I Ba ) using 10 mM Ba 2+ as a charge carrier in response to 8 ms step depolarizations revealed lack of I Ba exclusively in those IHCs with a strong GFP signal ( Figure 1D, green trace). In contrast, I Ba was present in one IHC of cKO-Ca v 1.3 flex/flex mice with weak (blue trace), one IHC without GFP fluorescence (gray trace, Figure 1D) and a wildtype IHC ( Figure 1C). Corresponding individual peak I-V relations show that I Ba was abolished in the IHC with strong GFP fluorescence whereas it was present in the two cKO-Ca v 1.3 flex/flex IHCs with weak or no GFP fluorescence and in the wildtype IHC ( Figure 1E). Averaged peak I Ba from cKO-Ca v 1.3 flex/flex IHCs with strong GFP (-7.9 ± 1.9 pA; n = 4) was significantly reduced compared with IHCs showing weak GFP fluorescence (-101.6 ± 14.2 pA; n = 3; P = 0.0262, MWU test, Figure 1F). We concluded that only those IHCs with strong fluorescence represented true knockout cells with two switched flex alleles whereas IHCs with weak fluorescence represented cells with one switched and one intact flex allele.

Cochlea-Specific
In order to increase the ratio of "true knockout" IHCs without remaining intact flex alleles, cKO-Ca v 1.  (Figure 2E, right). The lack of I Ba was accompanied by a reduction in cell size as evident by a significantly reduced membrane capacitance (P < 0.001, Kruskal-Wallis Test) in cKO-Ca v 1.3 flex/− mice (6.7 ± 0.6 pF; n = 15; P = 0.001, effect of genotype) but not in control Ca v 1.3 flex/− mice (9.5 ± 1.4 pF; n = 10) compared with the wildtype (8.7 ± 1.0 pF; n = 10).
In summary, heterogeneous GFP expression and persistence of I Ba in IHCs with weak GFP fluorescence of cKO-Ca v 1.3 flex/flex mice show that (i) Cre did not faithfully switch both flex alleles at 3 weeks of age and (ii) GFP fluorescence is no reliable marker for deletion of Ca v 1.3 channels in IHCs of cKO-Ca v 1.3 flex/flex mice. In contrast, in IHCs of cKO-Ca v 1.3 flex/− mice containing only one flex allele, GFP fluorescence unequivocally indicated a cellular knockout genotype.

GFP Toxicity in IHCs of cKO-Ca v 1.3 flex/flex Mice
In Ca v 1.3 −/− mice, mild degeneration of IHCs has been reported in the apical cochlear turn starting after P20 and in the basal cochlear turn after P35 (Platzer et al., 2000;Glueckert et al., 2003;Nemzou et al., 2006). Degeneration of IHCs after cochleaspecific deletion of Ca v 1.3 was analyzed in organs of Corti of 4-5 week-old cKO-Ca v 1.3 flex/flex and cKO-Ca v 1.3 flex/− compared with Ca v 1.3 −/− mice, which were double-immunolabeled for GFP and the hair-cell marker calbindin (Figure 3 and Table 1). IHCs of all three genotypes showed mild IHC loss of ≤6.3% in the apical turn (Figures 3A-C and Table 1). In contrast, the majority (81.3%) of basal-turn IHCs of cKO-Ca v 1.3 flex/flex mice was missing ( Figure 3E and Table 1). This pronounced degeneration was not caused by lack of Ca v 1.3 because basalturn IHCs of Ca v 1.3 −/− and cKO-Ca v 1.3 flex/− mice did not show any degeneration (<0.5%; Figures 3D,F and Table 1). We conclude that high expression levels of GFP caused by two functional flex alleles in cKO-Ca v 1.3 flex/flex mice (cf. Figure 1A) resulted in a toxic effect of GFP on basal IHCs. The lack of IHC degeneration in the basal cochlea from cKO-Ca v 1.3 flex/− mice suggests a dose-dependent toxicity of GFP that requires more than one functional GFP allele.
The majority of outer hair cells (OHCs) from the apical (Figures 3A-C) but not basal cochlear turn (Figures 3D-F) were degenerated, as described before for Ca v 1.3 −/− mice (Platzer et al., 2000;Glueckert et al., 2003;Engel et al., 2006). Thus, cochlea-specific deletion of Ca v 1. The rate of true knockout IHCs was assessed by quantification of apical-turn organs of Corti immunolabeled for GFP, Ca v 1.3 and/or BK channels ( Table 2). In cKO-Ca v 1.3 flex/− mice, the knockout rate was 89.2%, which was only slightly higher than the knockout rate of 87.4% in cKO-Ca v 1.3 flex/flex mice at 4-5 weeks of age. Although replacing one flex allele by a knockout (-) allele increased the success rate of Cre in switching one flex allele at 3 weeks of age (cf. Figures 1A,B, 2A), Cre caught up in switching both flex alleles in cKO-Ca v 1.3 flex/flex mice 2 weeks later.
In conclusion, the percentage of knockout IHCs finally was not increased by replacement of one flex allele with a knockout (-) allele to obtain cKO-Ca v 1.3 flex/− mice. However, we found that in cKO-Ca v 1.3 flex/flex mice (i) basal-turn IHCs degenerated, most likely due to dose-dependent GFP toxicity and (ii) GFP expression in IHCs was not unequivocally associated with deletion of Ca v 1.3 channels, thus leading us to further use cKO-Ca v 1.3 flex/− instead of cKO-Ca v 1.3 flex/flex mice.

Similar Phenotypes of IHCs From cKO-Ca v 1.3 flex/− and Systemic
Ca v 1.3 −/− Mice In wildtype mice, up-regulation of BK K + channels around the onset of hearing (P12) and down-regulation of neonatal SK2 K + channels mark the end of terminal maturation and the onset of the mature function of IHCs ( Figure 4A) (Kros et al., 1998;Marcotti et al., 2004). In systemic Ca v 1.3 −/− mice, IHCs maintain an immature-like ion channel composition with persistent expression of SK2 but lack of BK K + channels (Brandt et al., 2003;Engel et al., 2006;Nemzou et al., 2006). The failure of acquiring a mature composition of K + channels may have been caused by lack of Ca v 1.3 currents (i) in the IHC itself or (ii) in brainstem nuclei causing an altered efferent input on the IHC (Hirtz et al., 2011(Hirtz et al., , 2012Satheesh et al., 2012). Precise timing and patterning of Ca 2+ action potentials generated by IHCs during a critical period before the onset of hearing are crucial for their maturation (Johnson et al., 2013a). Altered neuronal activity of the efferent input onto neonatal IHCs therefore might also affect their Ca 2+ action potentials and hence their development.
SK2 immunolabeling was localized at the basolateral pole apart from synaptic ribbons (RIBEYE) of apical turn IHCs from 4 to 5 week-old cKO-Ca v 1.3 flex/− and Ca v 1.3 −/− mice (Figures 4C,D) indicating an immature phenotype. In contrast, no SK2 labeling was found at the basolateral pole of wildtype IHCs ( Figure 4B).
BK channel expression was assessed in wildtype and Ca v 1.3 −/− IHCs co-labeled with calbindin (Figures 5A-C) and in Ca v 1.3 flex/− controls, cKO-Ca v 1.3 flex/flex and cKO-Ca v 1.3 flex/− IHCs co-labeled with GFP (Figures 5D-F). BK channels, which are indicators of a mature IHC phenotype, were present at the neck of wildtype IHCs ( Figure 5A). In Ca v 1.3 −/− mice, BK labeling was absent from apical-turn IHCs (Figure 5B), whereas sparse and faint labeling was found in basal-turn IHCs ( Figure 5C). In Ca v 1.3 flex/− control IHCs, normal BK labeling was found at the neck of IHCs (Figure 5D). In true cKO IHCs i.e., IHCs with strong GFP labeling in cKO-Ca v 1.3 flex/flex mice ( Figure 5E) and with GFP labeling in cKO-Ca v 1.3 flex/− mice (Figure 5F), BK labeling was missing. Unexpectedly, BK immunolabeling in heterozygous IHCs of both cKO genotypes, i.e., IHCs with weak (cKO-Ca v 1.3 flex/flex , Figure 5E) or no GFP labeling (cKO-Ca v 1.3 flex/− , Figure 5F), which appeared   The degree of degeneration was determined for individual ears and cochlear location (apical vs. basal) and is given as number of empty IHC slots (degenerated IHCs) per number of total IHC slots (filled + empty). Numbers per ear are given as mean ± SD. * due to profound degeneration, degenerated IHCs were determined as OHCs of the innermost OHC row minus intact IHCs.

Reduced Ba 2+ Currents and Ca v 1.3 Protein Clusters in Control Ca v 1.3 flex/− IHCs
We noticed that expression of one or two flex alleles without Cre resulted in smaller IHC Ba 2+ currents compared with wildtype IHCs (cf. Figures 1D,E, 2B-E). However, in conditional mouse lines, the function of the target gene should remain unaltered unless it is deleted or manipulated by Cre or other recombinases. For generating the conditional Cacna1d construct, loxP sites were inserted in intronic regions flanking exon 2 of the Cacna1d gene, which should not impair its function (Satheesh et al., 2012).
To  Figure 6A). I Ba normalized to the wildtype (100%) was reduced to 48.4% in Ca v 1.3 flex/flex and 42.1% in Ca v 1.3 flex/− IHCs, respectively ( Figure 6B). In mice with only one wildtype (+) allele, I Ba was slightly but not significantly reduced to -171.6 ± 50.9 pA or 80.9% (Ca v 1.3 +/− ; n = 10) and -149.5 ± 36.5 pA or 70.4% (Ca v 1.3 +/flex ; n = 7; Figures 6A,B), respectively. In contrast, I Ba was reduced to 5.1% in IHCs of cKO-Ca v 1.3 flex/− mice indicating a complete loss of Ca v 1.3 channels leaving a small residual Ca 2+ current that has been described before in the systemic knockout (Platzer et al., 2000;Brandt et al., 2003;Dou et al., 2004). Additionally, Cre expression in the cochlea did not affect I Ba in IHCs of Pax2::cre control mice (-216.4 ± 53.1 pA; n = 10; Figure 6B). Analysis of gating properties by fitting the I-V curves to a secondorder Boltzmann function times Goldman-Hodgkin-Katz driving force yielded a small but significant shift of V h by -2.5 mV in Ca v 1.3 flex/− (-12.6 ± 2.0 mV; n = 10) versus wildtype IHCs (-10.1 ± 2.3 mV; n = 10; P = 0.019, MWU test), whereas the voltage sensitivity of activation determined by the slope factor k was unaffected (Ca v 1.3 flex/− : 11.22 ± 0.97 mV; wildtype: 11.26 ± 0.30 mV; P = 0.762, MWU test). Sketch of a wildtype IHC before (left, neonatal) and after the onset of hearing (right, mature) depicting differences in shape, ion channel composition and innervation. Immature IHCs express SK2 channels (magenta) at the basal pole, which are down-regulated after the onset of hearing, and do not possess BK channels. At the onset of hearing, BK channels are up-regulated and localize mainly to the neck of mature IHCs. Medial olivocochlear (MOC) efferent fibers of the SOC (dark gray) innervate IHCs of neonatal mice. Whereas mature IHCs lack direct efferent innervation, their afferent fibers receive lateral olivocochlear (LOC) efferent fibers (light gray). (B) SK2 labeling (top, magenta) was absent from apical-turn wildtype IHCs at 4-6 weeks of age. (C,D) Dot-like SK2 labeling was present at the basal pole, but not co-localized with synaptic ribbons (RIBEYE, middle, green) of apical-turn IHCs of a Ca v 1.3 −/− (C) and a cKO-Ca v 1.3 flex/− mouse at the age of 4-6 weeks. The merged image is shown with nuclei stained in blue with DAPI. Scale bars: 5 µm.
In summary, reduction of I Ba amplitude in IHCs of control Ca v 1.3 flex/flex and Ca v 1.3 flex/− mice, as well as altered gating properties in Ca v 1.3 flex/− control IHCs demonstrate that the unswitched Cacna1d flex allele functionally does not fully replace the wildtype allele.
The functional reduction of Ca v 1.3 channels might be caused by a reduced amount of Ca v 1.3 channel protein in the IHC membrane or by a reduced function of Ca v 1.3 channels in Ca v 1.3 flex/− mice. The abundance of Ca v 1.3 channel protein was assessed by co-immunolabeling for Ca v 1.3 (magenta) and synaptic ribbons (RIBEYE, green, Figure 7). Ca v 1.3 clusters were localized at the synaptic ribbons of wildtype IHCs (Figure 7A,a) and at the majority of ribbons of IHCs from Ca v 1.3 flex/− control mice (Figure 7C,c). In contrast, no specific Ca v 1.3 labeling was found at the synapses of Ca v 1.3 −/− IHCs (Figure 7B,b) and most, but not all IHCs of cKO-Ca v 1.3 flex/− mice (Figure 7D,d ). In part of the IHCs from cKO-Ca v 1.3 flex/− mice, Ca v 1.3 labeling was still present at the synaptic ribbons (Figure 7D,d ) indicating that the flex allele was not switched in these cells. Synaptic ribbons (RIBEYE) of Ca v 1.3-deficient IHCs from Ca v 1.3 −/− and cKO-Ca v 1.3 flex/− mice were agglomerated and localized closer to the nucleus (Figures 7B,D) as described before (Nemzou et al., 2006).
To elucidate the cause of the reduced I Ba amplitude (42% of wildtype, Figures 6A,B) in IHCs of Ca v 1.3 flex/− control mice, a quantitative analysis of the size and number of Ca v 1.3 clusters and synaptic ribbons was performed (Figures 7E-G). Whereas the number of Ca v 1.3 clusters and ribbons was unchanged, the average size of both Ca v 1.3 clusters and synaptic ribbons was significantly reduced to 73 and 89% in Ca v 1.3 flex/− control IHCs compared with wildtype ( Figure 7F). This reduction in size also applied to the total area of Ca v 1.3 clusters to 75% and of ribbons to 83% of the total areas in wildtype, respectively ( Figure 7G). In conclusion, less Ca v 1.3 protein was produced in IHCs of Ca v 1.3 flex/− mice evident by reduced I Ba amplitudes and smaller Ca v 1.3 channel clusters, which was accompanied by smaller ribbons.
Averaged ABR waveforms of Ca v 1.3 flex/− controls had smaller amplitudes than in wildtype mice for click stimuli 40 dB above threshold (Figure 8C). Growth functions of peak-topeak amplitudes showed a significant reduction of all waves in Ca v 1.3 flex/− control (14/7 ears/animals) compared with wildtype (wave I, IV: 14/7 ears/animals; wave II, III: 12/6 ears/animals, Figure 8D), revealed by a regression analysis of the smaller slopes of fits to the amplitudes as a function of level above threshold of wave I to IV (wave I, P < 0.001; wave II, P = 0.015; wave III, P < 0.001; wave IV, P = 0.027; MWU test) and a smaller y-axis intercept value of wave I (MWU test, P = 0.002). Growth functions of latencies, calculated as time between stimulus application and the negative peak of the respective wave, were not significantly altered for all waves (I -IV) in Ca v 1.3 flex/− control mice compared with the wildtype (Figure 8E).
In summary, cochlea-specific deletion of Ca v 1.3 in cKO-Ca v 1.3 flex/− mice resulted in highly elevated ABR thresholds and strongly reduced DPOAEs, reflecting profound hearing loss. Moreover, in Ca v 1.3 flex/− control mice the reduction of mean IHC I Ba amplitude to 42% and of mean Ca v 1.3 cluster size to 73% is accompanied by increases in f-ABR thresholds up to 10 -20 dB and strongly reduced amplitudes of ABR waves I to IV. Notably, click-ABR thresholds and DPOAEs were not affected in Ca v 1.3 flex/− control mice.

Conditional Tissue-Specific Knockout Mice -Benefits and Pitfalls
An unwanted side effect of transgenic animals is unexpected germline expression of Cre recombinase resulting in embryonal recombination of loxP sites that might even occur in Crenegative offspring carrying a flex or lox allele (Song and Palmiter, 2018). This can be monitored (i) by adapting the genotyping protocol to recognize excised or switched lox or flex alleles and (ii) in the flex switch system as eGFP expression in cells of Cre-negative flex control mice with one or two flex alleles.
Without the GFP reporter function, we might not have detected the incomplete recombination of flex alleles in IHCs. Partial recombination of floxed alleles, resulting in a mixture of cells with recombination of both, one or even no allele, is a frequent problem in conditional knockout mice (Saam and Gordon, 1999;Schulz et al., 2007;Weis et al., 2010). Lack of knowledge about the amount of successful cellular deletion events may lead to wrong conclusions caused by residual functions contributed by non-knockout cells. In this study, about 10% of the IHCs in cKO-Ca v 1.3 flex/− mice carried an unswitched flex allele resulting in residual hearing function compared with complete deafness of Ca v 1.3 −/− mice (Platzer et al., 2000;Dou et al., 2004).
Our attempt to increase the success rate of Cre in switching the flex alleles by replacing one flex by a constitutive knockout ("−") allele resulted in a higher ratio of true knockout IHCs in cKO-Ca v 1.3 flex/− compared with cKO-Ca v 1.3 flex/flex mice (cf. Figures 1, 2) at 3 weeks of age. However, this difference was no longer present 2 weeks later (Table 2), indicating that Cre managed to switch most flex alleles by this time point. An alternative approach to increase the recombination rate of loxP sites would be to increase expression of Cre recombinase (Schnütgen et al., 2003;Schulz et al., 2007) using Pax2::cre/cre instead of Pax2::cre/+ mice. However, high Cre expression levels on the other hand might increase the risk of possible side effects. High levels of Cre expression in α-myosin heavy chain-Cre mice have for example been demonstrated to be cardiotoxic causing altered cardiac function, DNA damage and inflammation (Bhandary and Robbins, 2015;Pugach et al., 2015). A dose dependence of Cre toxicity has been confirmed in cell culture titration experiments (Loonstra et al., 2001;Baba et al., 2005).
In summary, due to ambiguous eGFP expression in cKO-Ca v 1.3 flex/flex mice when only one flex allele was switched resulting in heterozygous IHCs that still produced Ca v 1.3, we decided to further use cKO-Ca v 1.3 flex/− mice, where eGFP expression was a reliable marker of Ca v 1.3 ablation. Since incomplete recombination of both flex (or lox) alleles is likely to be a general problem in conditional mice, a combination with a systemic knockout allele (flex/-or lox/-) should be used if possible.

Toxicity of Excessive eGFP
Degeneration of IHCs in the basal cochlear turn as early as P25 in cKO-Ca v 1.3 flex/flex but not Ca v 1.3 −/− mice suggests toxicity of excessive eGFP. Furthermore, direct fluorescence of eGFP can be seen in non-fixed IHCs (this study) but not in the auditory brainstem (Bartels, 2009), further indicating particularly high expression of Ca v 1.3 in wildtype and eGFP in cKO-Ca v 1.3 flex/flex IHCs, respectively. GFP toxicity has been demonstrated in cell lines where its expression induced apoptosis (Liu et al., 1999) or inhibited polyubiquitination (Baens et al., 2006). In mice, neuronal expression of yellow fluorescent protein induced multiple dose-dependent stress responses (Comley et al., 2011). A possible cause for these damaging effects is that the pre-mature, colorless form of eGFP, which is present in variable proportions in GFP-expressing cells, produces the free radical O2 •− and hydrogen peroxide (H 2 O 2 ) under consumption of NAD(P)H (Ganini et al., 2017). Such a GFP-induced oxidative stress may explain why only IHCs of cKO-Ca v 1.3 flex/flex mice but not of cKO-Ca v 1.3 flex/− degenerated because of the higher dose of eGFP produced by two flex alleles.

Side Effects of Gene-Targeted Alleles Without Gene Deletion by Cre Recombinase
In conditional models, the modifications of the target gene should not affect its function unless recombined by Cre. In the study of Satheesh et al. (2012), who first described the Ca v 1.3-flex model, Ca 2+ currents were not analyzed. Normal ABR thresholds of Ca v 1.3 flex/flex control mice led the authors to the conclusion that unswitched flex alleles did not affect Ca v 1.3 channel function. Since Ca v 1.3 channels mediate the majority of Ca 2+ current in IHCs, the present study provided the unique opportunity to analyze potential side effects of the flex construct in detail by measuring Ca 2+ channel currents and quantitatively analyzing Ca v 1.3 protein clusters. We found reduced I Ba amplitudes and a lower amount of Ca v 1.3 protein in IHCs of cre-negative Ca v 1.3 flex/flex and Ca v 1.3 flex/− control mice demonstrating that the unswitched flex allele did not fully replace the wildtype function. As has been shown before, a considerable reduction or increase in peak Ca 2+ current amplitudes of IHCs has only minor effects on click ABR thresholds (Scharinger et al., 2015;Fell et al., 2016), which can be misleading when used as the only method to assess the function of IHCs. Here, frequencydependent ABR thresholds were increased by 10-20 dB at most frequencies in Ca v 1.3 flex/− control mice upon reduction of I Ba to 42% of the wildtype value, which is in accordance to threshold increases of 5 -20 dB in null mutants of the auxiliary α 2 δ2 Ca 2+ channel subunit causing reductions of I Ba to 60-70% (Fell et al., 2016). The most prominent consequence of I Ba reduction with respect to hearing function of Ca v 1.3 flex/− control mice are the reduced growth functions of peak-to-peak amplitudes of the ABR waves, especially wave I, indicating strongly reduced IHC output at all levels above threshold. It should be kept in mind that both click and frequency-specific ABR thresholds are determined by only one afferent fiber type, the low threshold, high spontaneous rate fibers, whereas growth functions of ABR amplitudes cover the activity of all (high, medium and low spontaneous rate) afferent fiber types (Kiang et al., 1965;Liberman, 1978Liberman, , 1982Petitpré et al., 2018;Shrestha et al., 2018;Sun et al., 2018).
The question arises as to why I Ba was reduced in IHCs of Cre-negative mice containing at least one flex allele (Figure 6). For generating the conditional Ca v 1.3 mouse, the Cacna1d-eGFP flex construct was placed outside and a few hundred base pairs up-and downstream of the coding regions of exon 2 in the Cacna1d gene to avoid unintended manipulation of regulatory elements flanking the exon (Bartels, 2009). This insertion might have disrupted unknown regulatory elements and thus reduced the expression level of the channel. Moreover, I Ba gating properties were altered in Ca v 1.3 flex/− control mice, suggesting that insertion of the flex construct might have affected splicing of Cacna1d mRNA (Bock et al., 2011;Scharinger et al., 2015).

The IHC and Auditory Phenotype Following Systemic Versus Cochlea-Specific Deletion of Ca v 1.3 Channels
The ablation of Ca v 1.3 channels before birth in cKO-Ca v 1.3 flex/− mice caused an IHC phenotype similar to that of Ca v 1.3 −/− mice, including persistent expression of SK2, lack of BK expression in apical-turn IHCs, and a reduced cell size (Brandt et al., 2003;Glueckert et al., 2003;Nemzou et al., 2006).
Until the onset of hearing, SK2 channels mediate efferent inhibition of IHCs via α9α10 Ca 2+ -permeable nicotinic acetylcholine receptors (nAChRs) (Oliver et al., 2000;Elgoyhen et al., 2001), The origin of these efferent fibers lies in cholinergic neurons in the SOC. Shortly after birth, neurons of the auditory brainstem are spontaneously active and undergo a developmental program including synaptic pruning and establishment of tonotopy (Blankenship and Feller, 2010;Clause et al., 2014). The spiking pattern of SOC neurons is modulated by ascending information from the cochlea, where IHCs produce spontaneous Ca 2+ action potentials, which are synchronized by Ca 2+ waves in the transient Kölliker's organ (Tritsch and Bergles, 2010;Johnson et al., 2011Johnson et al., , 2017Sendin et al., 2014;Eckrich et al., 2018;Mammano and Bortolozzi, 2018). In turn, efferent inhibition from the SOC closes a feedback loop by shaping the spontaneous activity of IHCs (Guinan, 2006;Frank and Goodrich, 2018). In the systemic Ca v 1.3 knockout mouse both SOC neurons and IHCs lack Ca v 1.3 currents (Platzer et al., 2000;Hirtz et al., 2011), the latter of which as a consequence cannot produce action potentials (Brandt et al., 2003). In the SOC of Ca v 1.3 −/− mice, depolarization-induced spiking of lateral superior olive (LSO) neurons was changed from a single to a multiple firing pattern due to a reduction in K v 1.2 channels (Hirtz et al., 2011). This was most likely caused by the specific lack of Ca v 1.3 channels in brainstem neurons despite intact cochlear expression as confirmed in brainstemspecific Ca v 1.3 knockout mice (Satheesh et al., 2012). In the present study with SOC neurons expressing Ca v 1.3 channels, the phenotype of Ca v 1.3-deficient IHCs from cKO-Ca v 1.3 flex/− mice (lack of BK channels, persistence of SK2 channels, smaller cell size) was very similar to that of systemic Ca v 1.3 −/− mice. Therefore a potentially altered feedback signaling by Ca v 1.3deficient SOC neurons onto IHCs cannot be causative for the IHC phenotype of Ca v 1.3 −/− mice. Nevertheless, the spiking pattern of SOC neurons and, thus, efferent signaling back to immature IHCs might still be altered due to the loss of afferent activation by IHCs. In α9and α10-nAChR knockout mice, maturation of IHC K + channels was normal despite the complete lack of cholinergic efferent input from SOC neurons (Gomez-Casati et al., 2009;Johnson et al., 2013b). In summary, maturation of the IHC's K + channel composition is mainly controlled by intrinsic Ca 2+ signaling within the IHC and does not depend on Ca v 1.3 expression in the SOC exerting efferent feedback.

BK Channel Expression in Basal-Turn
IHCs of Ca v 1.3 −/− and Cre-Negative IHCs of cKO-Ca v 1.3 flex/− Mice We found residual BK labeling in IHCs of the basal but not the apical cochlear turn of Ca v 1.3 −/− mice. So far it is unknown why BK protein is missing in IHCs of Ca v 1.3 −/− mice along most of the cochlear length (Brandt et al., 2003) despite expression of the respective Kcnma1 mRNA (Nemzou et al., 2006).
The faint and dot-like BK labeling in GFP-negative IHCs of cKO-Ca v 1.3 flex/− mice clearly differed from the large BK patches found in IHCs of wildtype and control Ca v 1.3 flex/− mice ( Figure 5F). Assuming that Cre is not active in these IHCs their phenotype should be the same as that of Ca v 1.3 flex/− controls ( Figure 5D) but this was not the case. We can thus exclude that the reduced BK expression was caused by the incomplete wildtype function of the unswitched flex allele. But what are the differences between IHCs of Ca v 1.3 flex/− controls and GFPnegative IHCs of cKO-Ca v 1.3 flex/− mice? Differences intrinsic to the IHCs are: (i) Presence of the Pax2::cre allele at an unknown location in the genome, which might interfere with modulatory sequences affecting BK expression; (ii) Cre might be expressed in these IHCs without switching the flex allele, but it could still interfere with BK expression. Alternatively, a factor extrinsic to the IHC might be causing the reduced BK expression. GFPnegative IHCs of cKO-Ca v 1.3 flex/− mice are surrounded by Ca v 1.3-deficient, electrically silent IHCs, whereas neonatal IHCs produce Ca 2+ action potentials in wildtype and presumably Ca v 1.3 flex/− mice (Brandt et al., 2003). This activity causes periodic efflux of K + ions from the IHCs, which depolarizes neighboring phalangeal cells and IHCs, thereby amplifying and synchronizing Ca 2+ AP activity (Wang et al., 2015;Eckrich et al., 2018). In summary, impaired expression of BK channels in solitary GFP-negative IHCs surrounded by true Ca v 1.3 knockout IHCs of cKO-Ca v 1.3 flex/− mice may result from a lack of mutual activation and synchronization of Ca 2+ AP activity among IHCs during the critical developmental period. It would be interesting to analyze whether the Ca 2+ action potential activity in GFP-negative IHCs of cKO-Ca v 1.3 flex/− is altered compared to Ca v 1.3 flex/− controls.