The role of cilia in the development, survival, and regeneration of hair cells

ABSTRACT Mutations impacting cilia genes lead to a class of human diseases known as ciliopathies. This is due to the role of cilia in the development, survival, and regeneration of many cell types. We investigated the extent to which disrupting cilia impacted these processes in lateral line hair cells of zebrafish. We found that mutations in two intraflagellar transport (IFT) genes, ift88 and dync2h1, which lead to the loss of kinocilia, caused increased hair cell apoptosis. IFT gene mutants also have a decreased mitochondrial membrane potential, and blocking the mitochondrial uniporter causes a loss of hair cells in wild-type zebrafish but not mutants, suggesting mitochondria dysfunction may contribute to the apoptosis seen in these mutants. These mutants also showed decreased proliferation during hair cell regeneration but did not show consistent changes in support cell number or proliferation during hair cell development. These results show that the loss of hair cells seen following disruption of cilia through either mutations in anterograde or retrograde IFT genes appears to be due to impacts on hair cell survival but not necessarily development in the zebrafish lateral line.


INTRODUCTION
Primary cilia are microtubule-based organelles protruding from cells that serve as localized signaling environments.Mutations in genes impacting cilia lead to a broad class of diseases, known as ciliopathies, which can impact multiple organ systems (reviewed in Mill et al., 2023;Reiter and Leroux, 2017).One class of genes that has a particularly high impact on cilia is intraflagellar transport (IFT) genes.These genes are important for the transport of proteins in cilia and thus also for the formation and maintenance of cilia (Reviewed in Taschner and Lorentzen, 2016).At a cellular level, one of the defects commonly seen in IFT and a subset of other cilia gene mutants is a decrease in cell number.This is seen in several different cell types with the causes for this decreased cell number varying by cell type.
There is also evidence of increased apoptosis in the developing nervous system in response to mutations in either IFT or transition zone cilia genes (Abrams and Reiter, 2021;Gorivodsky et al., 2009;Wang et al., 2018).In these cases, it is believed the apoptosis is due to disruptions in either sonic hedgehog (Shh) or Akt signaling.Nonneuronal cells also show increased cell death when cilia are disrupted, including intervertebral disc cells, thyroid follicular cells, and thyroid cancer cells (Lee et al., 2021;Li et al., 2020a).In thyroid cancer cells this cell death may be due to mitochondria dysfunction.Knockdown of cilia genes in thyroid cancer cells leads to altered mitochondria morphology, decreased mitochondria respiration and ATP production, and decreased mitochondrial membrane potential (Lee et al., 2018(Lee et al., , 2021)).Additionally, cilia gene knockdown leads to the oligomerization of VDAC1, which can result in cytochrome c release, and VDAC1 inhibition can block the increased apoptosis normally seen in cilia-depleted thyroid cancer cells (Lee et al., 2021).Decreased mitochondria respiration and ATP production are also seen in kidney proximal tubule cells following Ift88 knockdown (Fujii et al., 2021), and neurons are more sensitive to mitochondrial inhibitor-induced apoptosis following Ift88 knockdown (Bae et al., 2019).
In addition to their role in controlling cell number developmentally cilia have also been implicated in the regeneration of numerous cell types.Knockdown of Ift20 decreases proliferation and dedifferentiation of Müller glia cells (Ferraro et al., 2015) and loss of Ift88 in muscle stem cells leads to decreased proliferation and regenerative ability of muscle (Palla et al., 2022).Regeneration in the heart is also reduced when cilia are disrupted, due to a loss of notch signaling (Li et al., 2020b).Cilia have also been shown to be important for normal notch signaling in epidermal keratinocytes, corneal epithelium cells, and hematopoietic stem and progenitor cells (Ezratty et al., 2011;Grisanti et al., 2016;Liu et al., 2019).
Hair cells contain a single primary cilium known as the kinocilium.It has previously been shown that mutations in genes that lead to the loss of the kinocilium result in a reduction of hair cells (Grati et al., 2015;Stawicki et al., 2016Stawicki et al., , 2019;;Tsujikawa and Malicki, 2004).At least part of this decreased hair cell number is likely caused by hair cell death as there is evidence of increased apoptosis in the inner ear of ift88 zebrafish mutants (Blanco-Sánchez et al., 2014;Tsujikawa and Malicki, 2004), along with evidence of the mistrafficking of proteins in these mutants similar to what is seen in photoreceptors when anterior IFT genes are disrupted (Blanco-Sánchez et al., 2014).However, to our knowledge, no one has ever looked at proliferation during hair cell development or regeneration in these mutants, nor has the reason for reduced hair cell number in retrograde IFT mutants that don't show protein trafficking defects (Stawicki et al., 2019) been investigated.Therefore, we wished to more thoroughly investigate how mutations in both anterograde and retrograde IFT mutants impacted hair cell number.We carried out this work in the zebrafish lateral line, which is a system of surface-localized hair cells allowing for easy access to these cells in vivo.The lateral line consists of distinct clusters of cells known as neuromasts that contain centrally located hair cells surrounded by supporting cells.These supporting cells can serve as hair cell precursors during regeneration (reviewed in Thomas et al., 2015).
We found evidence of apoptosis in mutants of both the anterior IFT gene, ift88, and retrograde IFT gene, dync2h1 in lateral line hair cells.However, the number of apoptotic cells observed did not fully account for the hair cell number differences seen in these mutants.We also found that mutations in IFT genes led to a slight reduction in mitochondrial membrane potential and that blocking mitochondrial calcium uptake decreased hair cell number in wild-type but not mutant zebrafish.Hair cell development did not seem as impacted as hair cell survival.We did not see alterations in supporting cell number in either mutant.Likewise, proliferation during hair cell development was not consistently disrupted in both mutants, though we did see a reduction of proliferative cells in ift88 mutants in some situations.In contrast to this, we observed decreases in proliferation during hair cell regeneration in both mutants and a slight decrease in the number of hair cells that were regenerated in ift88 mutants.However, it is not clear whether these defects were due to impacts on regeneration itself, hair cell death during the regenerative process, or the smaller number of initial hair cells in these mutants.Taken together these results suggest that mutations in both anterograde and retrograde IFT genes in zebrafish primarily impact hair cell survival.

Lateral line hair cells of IFT gene mutants undergo apoptosis
It has previously been shown that hair cells undergo apoptosis in the inner ear of ift88 mutants (Blanco-Sánchez et al., 2014;Tsujikawa and Malicki, 2004).We wished to investigate whether apoptosis also occurred in both anterograde and retrograde IFT gene mutants in the lateral line.To do this we stained wild-type siblings and mutants of both the anterograde IFT gene, ift88, and the retrograde IFT gene, dync2h1, fixed at 4, 5, and 6 days post fertilization (dpf ) with an antibody for both the hair cell-specific protein otoferlin and cleaved caspase-3.Similar to what has previously been found (Stawicki et al., 2016) we saw a significant decrease in hair cell number in both IFT gene mutants, particularly at later ages (Figs 1A,B and 2A,B).Additionally, we saw a small number of cleaved caspase-3 positive hair cells in both IFT gene mutants whereas we never saw any in wild-type sibling controls, suggesting that hair cells in these mutants do undergo apoptosis (Figs 1C and 2C).However, the number of cleaved caspase-3 positive cells we observed in the mutants was considerably lower than the difference in hair cell number between wild-type siblings and mutants.IFT mutants commonly showed three to four fewer hair cells per neuromast, but at most only one or two cleaved caspase-3 positive hair cells total amongst the nine neuromasts we counted for each fish.
Mitochondria dysfunction may play a role in hair cell apoptosis in IFT gene mutants Mutation or knockdown of cilia genes has been shown to alter mitochondria function in multiple cell types (Fujii et al., 2021;Lee et al., 2018) and in some cases, this may be linked to increased cell death (Bae et al., 2019;Lee et al., 2021).Additionally, mutations of IFT genes in hair cells lead to a reduction in rapid FM1-43 uptake suggesting a decrease in hair cell activity (Stawicki et al., 2016), and it has been shown that reduction in hair cell activity can lead to reduced mitochondrial membrane potential (Pickett et al., 2018).
To test if mitochondrial membrane potential was reduced in hair cells of IFT gene mutants we used the ratiometric mitochondrial membrane potential indicator dye, JC-1.JC-1 aggregates in response to mitochondrial membrane polarization, switching from its monomeric green fluorescent form to a red fluorescing form (Reers et al., 1991).Thus, the red/green ratio is influenced by mitochondrial membrane potential with a smaller ratio being indicative of a less polarized mitochondrial membrane potential and vice versa.We found a small but significant decrease in the red/ green JC-1 ratio of both IFT gene mutants at 5 dpf (Fig. 3).
In other hair cell mutants with reduced hair cell number and mitochondrial membrane potential decreases it has been shown that treatment with the mitochondrial calcium uniporter inhibitor RU360 can partially rescue decreases in hair cell number (Santra and Amack, 2021).We wished to test if this was true for IFT gene mutants.To do this fish were treated with RU360 at 500 nM from 4-6 dpf and then fish were fixed and hair cells were labeled for counting.Similar to what had previously been found with extended RU360 treatment (Santra and Amack, 2021), we found a significant decrease in the number of hair cells in wild-type siblings.However, we saw no changes in hair cell number in either IFT gene mutant following RU360 treatment (Fig. 4).Additionally, we noted that hair cell number in wild-type siblings treated with RU360 was similar to hair cell number in IFT gene mutants in both conditions.

Proliferation during hair cell development is not consistently reduced in IFT gene mutants
We next wanted to test whether proliferation was altered during development in cilia gene mutants.To do this we treated fish with EdU from 3-4 dpf to label proliferating cells.We then fixed fish at 5 dpf, labeled hair cells, and counted the number of EdU-positive hair cells in both wild-type siblings and mutants of ift88 and dync2h1 mutants.While the average numbers of EdU-positive hair cells were slightly reduced for each mutant, particularly ift88 mutants, these differences were not statistically significant (Fig. 5A-C).As we know hair cells are dying in IFT gene mutants (Figs 1 and 2) we next wanted to test how delaying the fixation point would impact the number of EdU-positive hair cells seen.Therefore, we again treated fish with EdU from 3-4 dpf to label proliferating cells during this period, but then fixed and stained fish at 6 dpf to label hair cells.Using this paradigm we found that there now were significantly fewer EdU-positive hair cells in ift88 gene mutants whereas there continued to be no significant differences in the number of Edu-positive hair cells between the dync2h1 wild-type siblings and mutants (Fig. 5D).

Supporting cell number is not altered in IFT gene mutants
In addition to playing a role in proliferation during development cilia are also important for the differentiation of some cell types (Liu et al., 2019;Song et al., 2007).In the zebrafish lateral line, a subset of cells in neuromasts are specified as hair cells with the remaining cells becoming support cells (reviewed in Thomas et al., 2015).Therefore, defects in hair cell specification may result in alterations in supporting cell number.To test for changes in support cell number we stained 6 dpf zebrafish with a Sox2 antibody to label supporting cells (Hernández et al., 2007).Doing this we failed to see any significant differences between the number of Sox2-positive cells in wild-type siblings versus mutant zebrafish in either ift88 or dync2h1 mutants (Fig. 6).

Hair cell regeneration is slightly reduced in IFT gene mutants
Regenerative ability has also been shown to be reduced in some cell types following the loss of cilia (Palla et al., 2022) in some cases due to disruption in notch signaling (Li et al., 2020b).However, in hair cells, disruption of notch signaling usually leads to increased regeneration (Daudet et al., 2009;Ma et al., 2008;Mizutari et al., 2013;Warchol et al., 2017).Therefore, we investigated what hair cell regeneration would look like in cilia gene mutants.As IFT gene mutants are strongly resistant to short-term neomycin-induced hair cell death (Stawicki et al., 2016), we used a long-term gentamicin treatment, which they are less resistant to (Stawicki et al., 2019), to trigger hair cell regeneration.Both wild-type siblings and IFT mutant fish were treated with 200 µM gentamicin for 24 h at 4 dpf after which ½ the fish were fixed to confirm hair cell death and the other ½ were left in plain EM for 48 h to allow for hair cell regeneration.Another group of fish was kept in plain EM for the first 24 h and again ½ were fixed and ½ were moved to fresh EM for an additional 48 h.Both wild-type siblings and IFT mutants experienced significant hair cell death in response to gentamicin treatment (Fig. 7A-D).There was also significant hair cell regeneration in both wild-type siblings and IFT gene mutants in the 48-h period after gentamicin treatment.In the case of both ift88 and dync2h1 mutants, the number of hair cells in the regeneration group was significantly reduced compared to wild-type siblings (Fig. 7A-D).However, when subtracting the average number of hair cells/neuromast immediately following gentamicin treatment from the average number of hair cells/neuromast 48 h later to get a value for the number of cells regenerated we found that only ift88 mutants showed a significant difference from their wild-type siblings (Fig. 7E, F).Interestingly, despite the slight reduction in hair cells following regeneration, the difference in hair cell number between IFT gene mutants and their wild-type siblings was greater in the groups that did not undergo regeneration than those that did not.Relatedly, while wild-type siblings showed comparable, if not slightly reduced, hair cell numbers in their 48-h fix groups that had undergone regeneration groups as compared to the 48-h fix group that had not undergone regeneration, in both IFT gene mutants there were significantly more hair cells in the 48-h fix group that had undergone regeneration as compared to the one that had not (Fig. 7C,D).
Given the reduction in hair cell number following regeneration in IFT gene mutants we next wanted to see if proliferation during hair cell regeneration was disrupted in these mutants.To do this fish (C) Quantification of the JC-1 Red/Green ratio for both ift88 and dync2h1 mutants and wild-type siblings at 5 dpf.There was a significant difference in both mutants by unpaired Student's t-test, P=0.0004 for ift88, shown as ***, and P=0.0105 for dync2h1, shown as *.n=19 wild-type siblings and 18 mutants for ift88 and n=21 wild-type siblings and 18 mutants for dync2h1 mutants.n numbers are not equal for all groups as some images were not usable due to the movement of fish during imaging.
were again treated with 200 µM gentamicin for 24 h at 4 dpf, however, this time EdU was included during the gentamicin treatment.Co-treatment of EdU with gentamicin was performed because we found that treating with EdU after gentamicin did not result in the expected increase in EdU-positive hair cells in wildtype siblings treated with gentamicin as compared to the EM group (data not shown).Fish were then washed out of gentamicin and EdU and given 48 h to recover in plain EM before they were fixed and labeled with hair cell markers.Both ift88 and dync2h1 mutants saw a significant decrease in the number of EdU-positive hair cells as compared to wild-type siblings in the groups that received gentamicin treatment to trigger hair cell regeneration (Fig. 8).We also saw a significant decrease in the number of EdU-positive hair cells in the ift88 mutant EM control group as compared to the ift88 wild-type sibling EM controls (Fig. 8), which is similar to what we previously saw in ift88 mutants when fish were fixed 48 h after EdU treatment (Fig. 5C).

DISCUSSION
It has previously been shown that mutations that lead to the loss of kinocilia in hair cells lead to a reduction in hair cell number in both the inner ear and lateral line of zebrafish (Grati et al., 2015;Stawicki et al., 2016Stawicki et al., , 2019;;Tsujikawa and Malicki, 2004).In this work, we wished to further investigate the mechanism behind this hair cell number reduction in addition to seeing if there were similar hair cell number changes during hair cell regeneration.Past work has suggested that decreases in hair cell number in the inner ear in mutants of the anterograde IFT gene, ift88, are caused by hair cell apoptosis potentially due to the mistrafficking of proteins and subsequent ER stress (Blanco-Sánchez et al., 2014;Tsujikawa and Malicki, 2004).However, it is not obvious that proteins are mistrafficked in hair cells in mutants of retrograde IFT genes like dync2h1 (Stawicki et al., 2019).This is similar to what has previously been shown in photoreceptors where anterograde IFT gene mutants show mistrafficking of opsins and subsequent cell death (Doerre and Malicki, 2002;Hudak et al., 2010;Lopes et al., 2010;Pazour et al., 2002;Tsujikawa and Malicki, 2004), however, in retrograde IFT gene mutants there is either no evidence of opsin mistrafficking or cell death (Krock et al., 2009), or the defects seen are less severe than those seen in anterograde IFT gene mutants (Tsujikawa and Malicki, 2004).There is evidence that some IFT genes can function outside of cilia (Cong et al., 2014;Delaval et al., 2011;Finetti et al., 2009Finetti et al., , 2014)), so some of the differences seen in phenotypes of different IFT gene mutants may be due to roles of these genes outside of cilia rather than their function in generating and maintaining cilia.
In contrast to what was previously seen in photoreceptors, we found evidence of hair cells undergoing apoptosis in comparable numbers in the lateral line of both ift88 and dync2h1 gene mutants (Figs 1 & 2).This suggests cilia formation and maintenance generally is important for hair cell survival rather than ift88 playing a unique gene-specific role.It also shows that IFT genes are important for the survival of both inner ear and lateral line hair cells of zebrafish.However, the number of apoptotic cells we saw did not fully account for the hair cell number differences seen in these mutants.We usually saw between three to four fewer hair cells/neuromast in IFT gene mutants at 5 and 6 dpf, whereas we would only see one or two hair cells undergoing apoptosis in a small subset of lateral line neuromasts from 4-6 dpf (Figs 1 and 2).There could be a few reasons for this discrepancy.As we are only fixing at distinct time points we are certainly not seeing all the hair cells that will undergo apoptosis.Additionally, it is possible that some hair cells are dying by mechanisms other than apoptosis, such as necrosis, and thus are not staining positive for cleaved caspase-3.It is also possible that cilia are impacting hair cell numbers by affecting something other than hair cell survival.Though, as discussed below, we did not see obvious impacts on hair cell development in these mutants.
We also found evidence that disruption in mitochondria may play a role in the hair cell death seen in IFT gene mutants.This is consistent with data in other cell types showing disrupted mitochondria activity and/or morphology in cilia gene mutants (Fujii et al., 2021;Lee et al., 2018Lee et al., , 2021) ) and that these mitochondria disruptions are linked to cell death (Bae et al., 2019;Lee et al., 2021).We saw a small but significant decrease in mitochondrial membrane potential in both IFT gene mutants (Fig. 3).We also saw that treatment with RU360, a mitochondrial uniporter inhibitor, led to a decrease in hair cell number in wild-type siblings but not IFT gene mutants (Fig. 4).While it is true that IFT gene mutants are resistant to other drugs, such as aminoglycosides, due to decreased entry (Stawicki et al., 2016) we do not believe this is the reason for the lack of impact of RU360 in these mutants.Aminoglycosides rely on mechanotransduction activity to enter hair cells (Gale et al., 2001;Marcotti et al., 2005) whereas RU360 is a cell-permeable drug (Matlib et al., 1998) and should be able to enter hair cells normally even if there are mechanotransduction defects.The mitochondrial uniporter functions to allow the mitochondria to uptake calcium from the cytoplasm (Bernardi, 1999;Gunter and Fig. 4. Inhibition of the mitochondrial uniporter decreases hair cell number in wild-type siblings, but not IFT gene mutants.Quantification of hair cells/neuromast in (A) ift88 and (B) dync2h1 wild-type siblings and mutants with or without RU360 treatment.In the case of both mutants genotype, drug treatment, and the interaction between the two variables was a significant source of variation by two-way ANOVA (P<0.0001).****=P<0.0001by S ̌ıdaḱ's multiple comparisons test comparing the two groups at the edges of the lines under the asterisks.n=10 for all groups with the exception of the ift88 WT sibling EM group where n=9 due to the loss of fish during the staining process.Gunter, 1994;Kirichok et al., 2004) and plays key roles in cell death (Demaurex and Distelhorst, 2003;Scorrano et al., 2003).Our results are consistent with a model where mitochondrial uniporter function is impaired in IFT gene mutants and this leads to cell death, though they do not prove this model definitively.Future work is warranted to more thoroughly investigate how mitochondria are impacted in the hair cells of cilia gene mutants and what role this plays if any in hair cell death.
It is worth noting that decreases in hair cell activity can lead to reduced mitochondrial membrane potential (Pickett et al., 2018)   Biology Open cells of IFT gene mutants show reduced uptake of the dye FM1-43 (Stawicki et al., 2016;Stawicki et al., 2019).As this rapid FM1-43 uptake is known to be dependent on hair cell activity (Gale et al., 2001;Meyers et al., 2003;Seiler and Nicolson, 1999) this suggests hair cell activity is reduced in these mutants.Experiments looking at calcium imaging in response to water jet stimulation of lateral line hair cells also suggest hair cell activity is reduced in ift88 gene mutants (Kindt et al., 2012).The reductions in mitochondrial membrane potential we observe in IFT gene mutants are not as dramatic as those observed in cdh23 mutants (Pickett et al., 2018), which may be because cdh23 mutants show a complete loss of hair cell activity (Nicolson et al., 1998;Seiler and Nicolson, 1999) whereas hair cell activity is only reduced but not eliminated in IFT gene mutants (Kindt et al., 2012;Stawicki et al., 2016Stawicki et al., , 2019)).Thus it is not clear if the mitochondria defects we see in IFT gene mutants are due to direct impacts of the cilia on the mitochondria or indirect impacts due to changes in hair cell activity.Additionally, as the mitochondria is critically involved in the apoptosis process (reviewed in Wang and Youle, 2009) it is possible the mitochondrial changes we observed are a consequence of the apoptosis rather than the cause.Defects in proliferation during development are another way by which IFT gene mutants could impact hair cell number.This is commonly seen in neuronal development (Amador-Arjona et al., 2011;Breunig et al., 2008;Chizhikov et al., 2007;Han et al., 2008;Lepanto et al., 2016;Pruski et al., 2019;Spassky et al., 2008;Tong et al., 2014).Our results suggest that dync2h1, and thus cilia in general, do not play a role in hair cell proliferation during development, however, it is not clear if ift88 is playing a role.To investigate whether decreased proliferation during hair cell development is responsible for the reduction in hair cell number seen in IFT gene mutants we treated fish with EdU from 3-4 dpf and then hair cells were labeled at 5 dpf.These time points were chosen as decreases in hair cell number in dync2h1 mutants are first observed at 5 dpf (Stawicki et al., 2016) (Fig. 2).When doing this we failed to see any significant differences in the number of EdU-positive hair cells in either the ift88 or dync2h1 mutants, though the ift88 mutants did show slightly fewer EdU-positive hair cells/neuromast, with the average being 4.4 for wild-type siblings and 3.5 for mutants (Fig. 5A-C).However, during the regeneration experiments when fish were treated with EdU from 4-5 dpf and then hair cells labeled at 7 dpf, we did see a significant difference in the number of EdU-positive hair cells/ neuromast in the EM control group for the ift88 mutants which had not undergone regeneration, with the average being 4 Edu-positive hair cells for wild-type siblings and 2 for mutants (Fig. 8).One possibility for this disparity is that the role of IFT genes in proliferation during hair cell development varies by age, however, unlike dync2h1 mutants, ift88 mutants are already showing significant decreases in hair cell number at 4 dpf (Fig. 1).If proliferation was playing a role in that difference, one would expect it to be present early in development.Alternatively, as the regeneration experiment labeled hair cells 2 days rather than 1 day after EdU labeling, and we know there is hair cell death in these mutants, there may be more death of new hair cells in those experiments thus lowering the EdU-positive hair cell counts in Fig. 5. Proliferation during hair cell development is not significantly decreased in both IFT gene mutants.Representative images of neuromasts from (A) ift88 and (B) dync2h1 wild-type siblings (top) and mutants (bottom) at 5 dpf.Cells that proliferated from 3-4 dpf are EdUpositive as shown in magenta.Nuclei are labeled in blue with DAPI and hair cells in green with the otoferlin antibody.Scale bar: 10 µm.(C) Quantification of the number of EdU-positive hair cells/neuromast for each mutant when fixed at 5 dpf.There were no significant differences when comparing wildtype siblings to mutants by unpaired t-test for either mutant (P=0.096for ift88 and P=0.481 for dync2h1).For ift88 Welch's correction was used due to the unequal variances between wild-type siblings and mutants.n=10 for ift88 mutants and dync2h1 wild-type siblings and n=9 for ift88 wild-type siblings and dync2h1 mutants.(D) Quantification of the number of EdUpositive hair cells/neuromast for each mutant when fixed at 6 dpf.ift88 mutants showed a significant reduction in the number of EdU-positive hair cells as compared to their wild-type siblings (P=0.0411) by unpaired t-test.There were still no significant differences when comparing dync2h1 wildtype siblings to mutants (P=0.1055).n=9 for both ift88 groups and n=10 for both dync2h1 groups.n numbers were not equal in all groups due to the loss of fish in the staining process or issues with images.Fig. 6.Support cell number is not altered in IFT gene mutants.Representative images of neuromasts from (A) ift88 and (B) dync2h1 wildtype siblings (left) and mutants (right) at 6 dpf stained with a Sox2 antibody to label support cells.Scale bars: 10 µm.(C) Quantification of the number of Sox2 positive support cells/neuromast for each mutant.There were no significant differences when comparing wild-type siblings to mutants by unpaired t-test for either mutant (P=0.6982for ift88 and P=0.5632 for dync2h1).n=9 for both ift88 groups and n=10 for both dync2h1 groups.n numbers were not equal in all groups due to issues with images.
mutants.Upon repeating the developmental EdU experiments and fixing 2 days after EdU labeling rather than 1 day (thus fish were labeled with EdU from 3-4 dpf and fixed at 6 dpf) we did see a significant reduction in the number of EdU-positive hair cells in ift88 mutants (Fig. 5D) suggesting this is what is happening.
Another way by which hair cell number could be changed in IFT gene mutants is defects in hair cell differentiation.To investigate this, we looked at supporting cell number reasoning that if hair cells were not differentiating, we may see an increase in cells kept at a supporting cell fate.However, we did not observe any differences in supporting cell number between wild-type siblings and IFT gene mutants for either mutant (Fig. 6).Thus, this was another situation where we failed to see impacts on neuromast or hair cell development in IFT gene mutants.Though as hair cells are derived from symmetrically dividing precursors a difference of 3-4 hair cells could be accounted for by only 1-2 support cells, it is possible that we simply were not able to detect that small of a difference.We also, from our data, cannot rule out defects in hair cell differentiation that do not change supporting cell numbers.However, the fact that there were not dramatic decreases in support cell number does support the theory that there are not widespread proliferation defects during hair cell development in IFT gene mutants, as these would presumably decrease supporting cell number as well as hair cell number.
The idea that IFT genes are playing a greater role in hair cell survival rather than hair cell development is also supported by the fact that earlier in development IFT gene mutants have normal hair cell numbers.It has previously been shown that at 2 dpf ift88 mutants have a normal number of hair cells in the lateral line (Kindt et al., 2012) and at 3 dpf they have a normal number of hair cells in the inner ear (Tsujikawa and Malicki, 2004).Likewise, dync2h1 mutants have normal lateral line hair cell numbers at 3--4 dpf (Stawicki et al., 2016) (Fig. 2B).However, one potential caveat to this data is that RNA for cilia genes is known to be maternally loaded into zebrafish eggs (Bisgrove et al., 2005;Sun et al., 2004).Therefore, as the mutants we used were from heterozygous parents it is possible that they expressed these genes early in development from this maternal RNA.Maternal zygotic mutants would need to be used to definitively rule out a role for these genes in hair cell development.
While we did not see consistent evidence for defects in proliferation during hair cell development across the two different IFT mutants, both mutants did show significant decreases in proliferation and subsequent hair cell number following regeneration (Figs 7 and 8).However, as it is known that neuromasts generally regenerate back to their starting size (Ma et al., 2008), it is not clear how much of this is due to impairment in the regenerative process versus the smaller starting size of neuromasts in IFT mutants.Also as we know hair cells are dying in IFT gene mutants some of the reduction we see in the regenerative group for these mutants may be due to hair cell death rather than reduced regeneration.Indeed, mutants that had undergone hair cell regeneration actually had more hair cells, and thus closer hair cell numbers to their wild-type siblings, than those that had not undergone regeneration (Fig. 7C,D).This would fit with a model where hair cells are dying over time creating the difference in hair cell number between wild-type siblings and mutants, as the regeneration response would have led to the birth of new hair cells that would not yet have had time to undergo apoptosis.
Overall, our results suggest that the decrease in hair cell number in dync2h1 mutants seen in normal situations is in large part due to hair cell death.We do not see evidence of defects in proliferation during development in these mutants.Additionally, significant hair cell number differences are not seen until 5 dpf in these mutants (Fig. 2) (Stawicki et al., 2016) which would fit more with a model of hair cell death than impaired initial development.The more dramatic differences in hair cell number in fish that have not undergone regeneration than those that did also fit this model as discussed above.In contrast to this, ift88 may be playing other roles in controlling hair cell number due to the earlier observed decrease in hair cell number (Fig. 1) and potential decreases in proliferation in situations where regeneration is not occurring (Figs 4 and 8).It also appears IFT genes are not required for hair cell regeneration as it can still occur in mutants, but the number of hair cells following regeneration and proliferation of cells during regeneration is reduced in these mutants (Figs 7 and 8).This could be due to an impairment in the regenerative process itself or reductions in initial hair cell number and hair cell death.

Animals
All experiments were carried out using fish from either the dync2h1 w46 or ift88 tz288 mutant lines (Doerre and Malicki, 2002;Stawicki et al., 2016).Mutant alleles were maintained in heterozygous animals in the *AB wildtype background and these heterozygotes were incrossed to get larvae for experiments.Wild-type siblings and mutants were separated based on body morphology phenotypes (Ryan et al., 2013;Tsujikawa and Malicki, 2004).Fish larvae were raised in embryo media (EM) consisting of 1 mM MgSO 4 , 150 μM KH 2 PO 4 , 42 μM Na 2 HPO 4 , 1 mM CaCl 2 , 500 μM KCl, 15 mM NaCl, and 714 μM NaHCO 3.They were housed in an incubator maintained at 28.5°C with a 14/10 h light/dark cycle.The Lafayette College Institution Animal Care and Use Committee approved all experiments.

Cell counts
Hair cell and cleaved caspase-3 positive cells were counted using an Accu-Scope EXC-350 microscope under the 40X objective.For most experiments the posterior lateral line neuromasts, P1-P9 were counted (Alexandre and Ghysen, 1999).For the RU360 experiments, the following anterior lateral line neuromasts were counted OP1, M2, IO4, O2, MI2 and MI1 (Raible and Kruse, 2000).In both cases for hair cell counts the total number of hair cells counted for each fish was divided by the number of neuromasts counted to get an average hair cell/neuromast number for each fish.For the cleaved caspase counts data is shown as the total number of cells counted for each fish due to the small number.Representative images were imaged on a Zeiss LSM800 confocal microscope using the Zen Blue software using a 40× water immersion objective.Z-stacks were taken consisting of 15 (caspase) Fig. 7. Hair cell number following regeneration is slightly reduced in IFT gene mutants.Representative images of neuromasts with hair cells labeled using the otoferlin antibody in (A) ift88 and (B) dync2h1 wild-type siblings (left) and mutants (right), pre (top) and post (bottom) regeneration.Scale bars: 10 µm.Quantification of hair cells/neuromast for (C) ift88 and (D) dync2h1 wild-type siblings and mutants during the regeneration experiment.EM groups were fish that were kept in EM the first 24 h of the experiment whereas Gent groups were fish that were put in 200 µM gentamicin for the first 24 h of the experiment.0 h groups were fixed immediately following those first 24 h, whereas 48 h groups were fixed following a 48 h recovery period in EM.A three-way ANOVA was carried out for both experiments which found all variables (time, drug, genotype) to be significant sources of variation as well as all two-way interactions between the three variables (P<0.001).The three-way interaction between variables was found to be significant for the dync2h1 experiment (P=0.003),but not the ift88 experiment (P=0.796).****=P<0.0001and ***=P<0.001by S ̌ıdaḱ's multiple comparisons test comparing the two groups at the edges of the lines under the asterisks.The 48-h gentamicin data shown in C and D is also shown in panels E for ift88 and F for dync2h1 following subtraction of the average 0-h gentamicin data for each group.There was a significant difference between ift88 wild-type siblings and mutants ****=P<0.0001,but not for dync2h1 (P=0.1639) by unpaired t-test following this normalization.n=10 for all groups.A two-way ANOVA was carried out for both experiments.For ift88 both drug treatment and genotype were significant sources of variation (P<0.0001)whereas the interaction between the two variables was not (P=0.7884).For dync2h1 both of the individual variables were also significant sources of variation (P<0.0001) as was the interaction between the two variables (P=0.0016).**=P<0.01 and ****=P<0.0001by Tukey's multiple comparisons test comparing the two groups at the edges of the lines under the asterisks.n=10 for all groups.
or 12 (regeneration hair cells) slices separated by 1 µm and then maximum projection images were made in Fiji.

JC-1 labeling and analysis
JC-1 was used as previously described (Pickett et al., 2018).Fish were incubated with 1.5 µM JC-1 (ThermoFisher, T3168) diluted in EM for 30 min, washed 3× with plain EM and then left in the third EM wash for 90 min.After 90 min fish were anesthetized with MS-222 and imaged on a Zeiss LSM800 using a 40× water immersion objective.For imaging both the green and red signals were obtained using a 488 nm laser.The green signal looked at 410-546 nm emissions whereas the red signal looked at 585-700 nm emissions.One neuromast was imaged for each fish.A z-stack was taken consisting of five slices separated by 1 µm.Image analysis was carried out in Fiji.Maximum projections were made and then the average red and green fluorescence was measured.From these numbers a red/green ratio was generated.For each mutant data from two separate experiments were combined.Thus, for each individual experiment the data was normalized to the average red/green ratio of the wild-type siblings for that experiment.

Drug treatment
For all drug treatments fish were loaded into netwell inserts in six-well plates containing EM.The netwell inserts containing the fish were then moved into new six-well plates containing the drugs being used for treatment diluted in EM.Drugs used were either 500 nM RU360 (MilliporeSigma, 57440) or 200 µM Gentamicin (MilliporeSigma, G1272).Following drug treatment fish were washed three times with plain EM, again by moving netwell inserts to new 6 well plates containing plain EM.

EdU labeling and analysis
EdU labeling was carried out similar to what has previously been described (Thomas and Raible, 2019).Fish were incubated in EdU (ThermoFisher, A10044 or C10340) at 500 µM for 24 h.After fixation EdU-positive cells were labeled via Click-iT reaction using the Click-iT TM EdU cell proliferation kit for imaging (ThermoFisher, C10340) following the protocol that came with the kit.Hair cells and nuclei were then labeled as mentioned above in immunostaining.To count the number of EdU-positive hair cells, fish were imaged on a Zeiss LSM800 microscope.A z-stack was taken consisting of 15 (developmental) or 12 (regeneration) slices separated by 1 µm.Image analysis was carried out in Fiji.Composite images were generated combining the EdU, Otoferlin and DAPI labels and then the number of EdU-positive hair cells were manually counted in Fiji using the cell counter tool.For each fish the P1, P2, and P3 neuromasts were counted and then an average EdU-positive hair cell/neuromast number was generated.Counts were carried out using the full z-stacks, however, the representative images shown are maximum projections which were also made in Fiji.

Support cell number quantification
Fish were fixed at 6 dpf and hair cells and support cells were labeled using the Otoferlin and Sox2 antibody respectively as mentioned above in immunostaining.To count the number of support cells, fish were imaged on a Zeiss LSM800 microscope.A z-stack was taken consisting of slices separated by 1 µm through the entirety of each neuromast.Image analysis was carried out in Fiji.Composite images were generated combining the Sox2 and otoferlin labels and then the number of Sox2 positive cells were manually counted in Fiji using the cell counter tool.For each fish the P1, P2, and P3 neuromasts were counted and then an average Sox2 positive support cell/neuromast number was generated.Counts were carried out using the full z-stacks, however, the representative images shown are maximum projections which were also made in Fiji.

Statistics and graphing of data
Most statistics were calculated in GraphPad Prism 6 with the exception of the three-way ANOVA for the hair cell regeneration data which was calculated in SPSS version 29.The specific statistical tests used for each experiment are mentioned in the figure legends.The graphs were all generated in GraphPad Prism 6. Graphs show individual data points along with the averages and standard deviations as superimposed lines.

Fig. 1 .
Fig. 1. ift88 mutants show decreased hair cell number and hair cells undergoing apoptosis.(A) Representative images of neuromasts from ift88 wild-type siblings (top) and ift88 mutants (bottom) at 6 dpf.Nuclei are labeled in blue with DAPI, hair cells in green with the Otoferlin antibody and cells undergoing apoptosis in magenta with the cleaved caspase-3 antibody.Scale bar: 10 µm.(B) Quantification of hair cells/neuromasts in ift88 wild-type siblings and mutants at 4, 5 and 6 dpf.Genotype and age were significant sources of variation by two-way ANOVA (P<0.0001), but the interaction between the two variables was not (P=0.2377).****=P<0.001by S ̌ıdaḱ's multiple comparisons test comparing the wild-type siblings and mutants at the different ages.n=10 for each group.(C) Quantification of the total number of cleaved caspase-3 positive hair cells in the nine neuromasts of the posterior lateral line in ift88 wild-type siblings and mutants at 4, 5 and 6 dpf.Genotype was a significant source of variation by two-way ANOVA (P=0.001) whereas age (P=0.1089) and the interaction between the two variables (P=0.0189) were not.n=10 for each group.

Fig. 2 .
Fig. 2. dync2h1 mutants show decreased hair cell number and hair cells undergoing apoptosis.(A) Representative images of neuromasts from dync2h1 wild-type siblings (top) and dync2h1 mutants (bottom) at 6 dpf.Nuclei are labeled in blue with DAPI, hair cells in green with the Otoferlin antibody and cells undergoing apoptosis in magenta with the cleaved caspase-3 antibody.Scale bar=10 µm.(B) Quantification of hair cells/neuromasts in dync2h1 wild-type siblings and mutants at 4, 5 and 6 dpf.Genotype (P<0.0001),age (P=0.0003), and the interaction between the two variables (P<0.0001) were significant sources of variation by 2-way ANOVA.****=P<0.001by S ̌ıdaḱ's multiple comparisons test comparing the wild-type siblings and mutants at the different ages.n=10 for each group.(C) Quantification of the total number of cleaved caspase-3 positive hair cells in the 9 neuromasts of the posterior lateral line in dync2h1 wild-type siblings and mutants at 4, 5 and 6 dpf.Genotype was a significant source of variation by 2-way ANOVA (P=0.0169)whereas age (P=0.1918) and the interaction between the two variables (P=0.1918) were not.n=10 for each group.

Fig. 3 .
Fig. 3. Mitochondrial membrane potential is slightly reduced in IFT gene mutants.(A) Representative images of neuromasts from (A) ift88 or (B) dync2h1 wild-type siblings (top) and mutants (bottom) labeled with JC-1.Scale bar: 10 µm.(C)Quantification of the JC-1 Red/Green ratio for both ift88 and dync2h1 mutants and wild-type siblings at 5 dpf.There was a significant difference in both mutants by unpaired Student's t-test, P=0.0004 for ift88, shown as ***, and P=0.0105 for dync2h1, shown as *.n=19 wild-type siblings and 18 mutants for ift88 and n=21 wild-type siblings and 18 mutants for dync2h1 mutants.n numbers are not equal for all groups as some images were not usable due to the movement of fish during imaging.

Fig. 8 .
Fig. 8. Proliferation during hair cell regeneration is reduced in IFT gene mutants.Representative images of neuromasts from (A) ift88 and (B) dync2h1 wild-type siblings (top) and mutants (bottom) at 7 dpf following gentamicin and EdU treatment from 4-5 dpf.Cells that proliferated during the 4-5 dpf time period are EdU-positive as shown in magenta.Nuclei are labeled in blue with DAPI and hair cells in green with the otoferlin antibody.Scale bar: 10 µm.The graphs show the quantification of EdU-positive hair cells/neuromast for (C) ift88 and (D) dync2h1 wild-type siblings and mutants 48 h after either EM or gentamicin treatment.A two-way ANOVA was carried out for both experiments.For ift88 both drug treatment and genotype were significant sources of variation (P<0.0001)whereas the interaction between the two variables was not (P=0.7884).For dync2h1 both of the individual variables were also significant sources of variation (P<0.0001) as was the interaction between the two variables (P=0.0016).**=P<0.01 and ****=P<0.0001by Tukey's multiple comparisons test comparing the two groups at the edges of the lines under the asterisks.n=10 for all groups.