Kremen1 regulates mechanosensory hair cell development in the mammalian cochlea and the zebrafish lateral line

Here we present spatio-temporal localization of Kremen1, a transmembrane receptor, in the mammalian cochlea, and investigate its role in the formation of sensory organs in mammal and fish model organisms. We show that Kremen1 is expressed in prosensory cells during cochlear development and in supporting cells of the adult mouse cochlea. Based on this expression pattern, we investigated whether Kremen1 functions to modulate cell fate decisions in the prosensory domain of the developing cochlea. We used gain and loss-of-function experiments to show that Kremen1 is sufficient to bias cells towards supporting cell fate, and is implicated in suppression of hair cell formation. In addition to our findings in the mouse cochlea, we examined the effects of over expression and loss of Kremen1 in the zebrafish lateral line. In agreement with our mouse data, we show that over expression of Kremen1 has a negative effect on the number of mechanosensory cells that form in the zebrafish neuromasts, and that fish lacking Kremen1 protein develop more hair cells per neuromast compared to wild type fish. Collectively, these data support an inhibitory role for Kremen1 in hair cell fate specification.

Scientific RepoRts | 6:31668 | DOI: 10.1038/srep31668 Kremen1 is a single pass transmembrane protein that acts as a receptor to members of the dickkopf (Dkk) family of Wnt antagonists 7 . Kremen1 functions as part of a Wnt inhibitory complex that prevents Lrp5/6 mediated sequestration of Gsk3β 7 , allowing it to target cytoplasmic β -catenin for degradation. On receipt of a Dkk ligand Kremen1 associates with Lrp5/6, removing it from the cell surface via clathrin mediated endocytosis, thus attenuating Wnt signal transduction [7][8][9] .
Based on previous reports that place Kremen1 atop the Wnt cascade acting to inhibit Wnt binding to its cogent receptors, we examined the effects of manipulating receptor composition on Wnt responsive cells of the cochlea. Through gain and loss-of-function experiments, we show that Kremen1 is involved in regulation of cell fate decisions in the mammalian cochlea and the zebrafish lateral line.

Results
Kremen1 is expressed in the developing mouse cochlea. Using reverse transcriptase PCR, we determined that Kremen1 and Dkk family members Dkk2, Dkk3 and Dkkl1 were expressed in the cochlear duct on embryonic (E) day 12.5 and E15.5, coincident with formation of the sensory epithelium (Table 1). We have previously reported that Dkk3 is expressed in the greater epithelial ridge throughout cochlear development and maturation 10 . We observed that Dkk1 was expressed at low levels on E15.5, but was not detected on E12.5 (consistent with previous data we reported in Geng et al. 10 ), and that Dkk4 and Kremen2 were not detected in the developing cochlea on E12.5 or E15.5 (Table 1). We then localized Kremen1 expression using in situ hybridization and immunohistochemistry. On E12.5, when the prosensory region is in the early stages of specification, Kremen1 was expressed in the floor of the duct (Fig. 1A,B). The sensory region of the mammalian cochlea has a developmental gradient such that the developing base of the cochlea is more mature than the developing apex. Sagittal sectioning through the cochlea on E15.5 allowed visualization of the developing sensory epithelium at different developmental stages (Fig. 1C). By E15.5, in the least developmentally advanced region, the apex (Fig. 1C,D), and in the midbase (Fig. 1C,E), expression was restricted to the prosensory region. In the most developmentally advanced region of the cochlea, the base, Kremen1 expression was only localized to the supporting cell region (Fig. 1C,F). On postnatal day 1 (P1), when the developmental process is almost complete, we observed Kremen1 expression only in inner border cells, inner phalangeal cells, pillar cells, and Deiters' cells ( Fig. 1G). We did not detect expression in hair cells. This restricted expression pattern was also observed in adult mice (Fig. 1H). Kremen1 protein was localized to the apical region of supporting cell plasma membrane (Fig. 1I,I' ,I",J,J' ,J"). These results suggest that Kremen1 is continuously expressed in the supporting cells and is implicated in patterning and maintenance of the sensory epithelium.
Over expression of Kremen1 in the mouse sensory epithelium has an inhibitory effect on hair cell formation. Should Kremen1 inhibit Wnt signaling in the cochlea, we hypothesized that introduction of excess Kremen1 would prevent hair cell differentiation. We tested this hypothesis by expressing ectopic Kremen1 in the prosensory epithelium. Plasmid containing the Kremen1 coding sequence (pCIG.Kremen1.IRES.nucGFP) or empty plasmid (pCIG.nucGFP) was electroporated into E13.5 cochlear explants and cultured in vitro for 6 days. Explant cultures were fixed and labeled for the hair cell marker MyoVIIa. Analysis was restricted to the most apical 25% portion of the cochlear duct ( Fig. 2A), targeting prosensory cells prior to hair cell and supporting cell differentiation. GFP positive cells were counted within the sensory epithelium and assigned as hair cells based on expression of MyoVIIa ( Fig. 2C-G). Of control plasmid electroporated cells, in total 49% differentiated as hair cells (Fig. 2C-E) while significantly fewer cells (15%) transfected with Kremen1 plasmid developed as hair cells (Fig. 2C,F,G, Table 2, p = 0.0006).
We also examined the influence of Kremen1 over expression on proliferation. Cochlear explants were harvested and electroporated with Kremen1 plasmid on E13.5 and cultured in vitro for 48 hours before addition of BrdU conditioned media. Explants were maintained for a further 72 hours and then fixed for analysis. GFP positive non-sensory cells of the lesser epithelial ridge within 200 μ m of the sensory epithelium were counted and assigned as having undergone cell division if they contained nuclear BrdU. In the non-sensory region, there was no significant difference in the number of double labeled BrdU and GFP positive cells in the control or Kremen1 expressing cells (Table 3). Kremen1 over expression also did not activate proliferation in the sensory region. We observed no BrdU positive GFP positive cells in the sensory epithelium in either control transfected or pCIG. Kremen1.IRES.nucGFP transfected explants.  [27][28][29][30] . Relative levels of expression are denoted by+ , no expression detected is denoted by − .  We also tested whether loss of Kremen1 protein would allow proliferation in cells of the postmitotic prosensory region. Cochlear explants were harvested and electroporated with negative control or Kremen1RNAi on E12.5 and maintained in BrdU conditioned media for six days. Kremen1RNAi transfected cells were no more likely to exhibit BrdU incorporation than cells expressing the negative control (Table 3, p = 0.36).

Kremen1 is involved in determining the number of hair cells contained within a zebrafish neuromast.
To examine the influence of Kremen1 on hair cell development in neuromasts of the lateral line in zebrafish, we analyzed a transgenic Kremen1 mutant line (Krm1 nl10 allele) generated on a Tg(cldn:GFP) background 11 . In agreement with previous reports 11 , the mutant fish had fewer neuromasts (Fig. 1A,B). Consistent with our previous work on activation of Wnt signaling in the lateral line 12 , we focused our analysis on neuromasts O2 and P1 (location indicated in Fig. 1A,B). Hair cells (defined by expression of the hair cell marker parvalbumin) within each neuromast were quantified (Fig. 4C-E). We observed significantly more hair cells in the O2 and P1 neuromasts in 5 day post fertilization (dpf) Kremen1 mutants than in age-matched wildtype animals (Kremen1 mutant: O2, 15.42 ± 4.10; P1, 12.66 ± 3.50. Wildtype: O2, 11.10 ± 2.38; P1, 9.41 ± 2.87), (O2, p = 0.0002; P1, p = 0.002, two tailed t-test). These data concur with previously observed reports of Kremen1 inhibition at position O2 13 .

Discussion
We report that in the embryonic mouse cochlea, Kremen1 was expressed first in the prosensory domain and that Kremen1 expression was restricted to supporting cells as development proceeded. Over expression of Kremen1 had a significant inhibitory effect on the differentiation of hair cells, while reduction in the levels of Kremen1 made a prosensory cell more likely to take a hair cell fate. These data are consistent with an antagonistic effect on canonical Wnt signaling, as Wnt signaling is required for hair cell formation 3,4 . Previous studies have revealed that β -catenin is necessary for hair cell differentiation and that stabilization of β -catenin in all prosensory cells leads to supernumerary hair cells and prosensory cell proliferation 3,4 . Given that current literature reports that Kremen1 is an inhibitory component of the Wnt signaling network, our results suggest that canonical Wnt signaling takes place in cells fated to become hair cells, and that Kremen1 prevents receipt of this signal in cells fated to become supporting cells. This effect was limited to cells that had not yet been committed to a hair cell or supporting cell fate; as the cochlea matured the ability of Kremen1 knock down to induce hair cell fate was reduced. We speculate that on E15.5, the hair cells and supporting cells had been specified, and knock down of Kremen1 had no significant effect. Furthermore, by reducing expression of Kremen1 on E12 prior to the onset of hair cell differentiation, we found that while transfected cells were more likely to differentiate as hair cells, they did not undergo cell division. The lack of a proliferative response when Kremen1 was downregulated indicated that should Kremen1 normally inhibit canonical Wnt signaling in the ear, endogenous levels of Wnt signaling were not sufficient to return post mitotic cells to a proliferative state in the absence of Kremen1. It is more likely that if endogenous Wnt signaling in the sensory region of the cochlea is modulated by Kremen1, it would control cell fate decisions, rather than cell cycle check points in the prosensory cells E13.5 onwards. This does not preclude a role for canonical Wnt signaling induced proliferation prior to E12.5 or in the periotic mesenchyme. Kremen1 reportedly acts by disrupting the composition of Wnt receptors and co-receptors on the cell membrane 7 . Supporting cells maintain expression of members of the Wnt receptor complex throughout development, but do not proliferate or change fate unless exogenous factors are applied or β -catenin signaling is activated intracellularly 3,4 . Wnt signal transduction is repressed, likely either by the presence of secreted Wnt antagonists, or the absence of key members of the Wnt receptor complex. We demonstrated that Dkk1, Dkk2, Dkk3 and Dkkl1 are expressed in the cochlear duct during development of the cochlea ( Table 1), suggesting that Dkk family of secreted Wnt antagonists are putative ligands for Kremen1. The presence of three out of five members on E12.5 and four out of five on E15.5 indicates that there is likely functional redundancy between these ligands. Membrane bound components of Wnt signaling, including Frizzled receptors, Lrp co-receptors and Wnt potentiating Lgr receptors, are expressed in supporting cells during development and at neonatal stages 5 . Frizzleds 1, 2 and 4 are expressed in both hair cells and supporting cells 5 ; expression of Frizzled receptors is necessary in hair cells after differentiation in order to transduce non-canonical Wnt signaling to orientate stereociliary bundles 14 . Lgr receptors function to allow secreted Rspondin molecules to potentiate Wnt signaling by retaining members of the Frizzled/Lrp5/6 receptor complex at the cell surface 15 . While Lgr receptors are expressed in supporting cells, supporting cells respond modestly to stimulation with Rspondin1 16 ; this could be because Kremen1 blocks Rspondin function by removing Lrp5/6. The ability of a given cell to respond to Wnt signaling is dependent on the combination of receptors that it expresses. We place Kremen1 on the cell surface of prosensory cells and then  supporting cells, and we show that Kremen1 expression biases cells away from a hair cell fate; together this indicates that Kremen1 acts to disrupt the transduction of hair cell inducing signaling events.
In zebrafish, Kremen1 over expression also had a negative effect on hair cell formation. Over expression of Kremen1 led to a reduction in the number of hair cells per neuromast, while loss of Kremen1 protein resulted in an increase in the number of hair cells per neuromast, but a decrease in the number of neuromasts at the posterior end of the trunk lateral line. It has previously been proposed that Kremen1 inhibits proliferation of prosensory cells located in deposited neuromasts since morpholino-mediated knock down of Kremen1 resulted in an increase in neuromasts size, consistent with our results in the krm1 nl10 mutant 13 . However, loss of Kremen1 inhibited proliferation of neuromast precursor cells in the migrating lateral line primordium by increasing levels of free Dkk1b and Dkk2 11 , resulting in the formation of fewer neuromasts 11 . It is therefore possible that in the krm1 nl10 fish, the anterior neuromasts contain additional hair cells due to increased deposition of precursors in these neuromasts, rather than to increased proliferation post-deposition. Over expression of Kremen1 limited the number of hair cells that form in the O2 and P1 neuromasts but did not increase the number of neuromasts that were deposited. This raises the question of whether hair cell precursors are segregated between hair cell and supporting cell fate by Wnt signaling during development, whether they require Wnt signaling in order to proliferate after being deposited, or whether the distribution of hair cells between neuromasts is controlled by Wnt signaling.
Kremen1 signaling likely acts at multiple points within lateral line development, with lateral line patterning in the krm1 nl10 fish resembling Wnt loss-of-function, rather than Wnt gain-of-function 11 . It is unlikely that the function of Kremen1 in sensory cell development recapitulates its role in migration of the posterior lateral line primordium. If Kremen1-mediated capture of Dkk ligands allowed Wnt signaling to upregulate proliferation within the deposited neuromasts, we would expect to see fewer hair cells and supporting cells per sensory organ in fish lacking Kremen1. Our experiments, and those by Wada et al. 13 , demonstrate that the opposite is true. Loss of Kremen1 results in an increase in the number of hair cells per neuromast, while over expression of Kremen1 resulted in fewer hair cells per neuromast. Since neuromasts are deposited and mature progressively, it is possible that Kremen1 has two distinct roles. Dkks are secreted ligands that can act at a distance from their source to inhibit Wnt signaling independently of Kremen1. Dkk expression is expanded in the migrating primordium of krm1 nl10 fish, but expression in deposited neuromasts is unknown, nor is it known whether Dkk and Kremen1 act cooperatively in the later stages of neuromast development 11 . The role of endogenous Wnt signaling in neuromast hair cell formation is still unclear as manipulation of Dkk2 13 and Kremen1 appear to affect hair cell number, whereas manipulation of downstream Wnt components such as Lef1 have no effect 17,18 . Given that Kremen1 appears necessary for both propagation and inhibition of proliferation depending on the stage of development, further study is required to clarify temporal functions of Wnt signaling during development of hair cell epithelia.

Methods and Materials
Animals. Mouse tissue was harvested from CD-1 mice of either sex (Charles River) maintained and euthanized in accordance with Sunnybrook Research Institute and Canadian Council on Animal Care guidelines for the care and use of laboratory animals. All mouse work was approved by Sunnybrook Research Institute Animal Care and Use Committee.
Zebrafish were maintained and euthanized in agreement with animal welfare regulations set out by the NIH. All zebrafish work was approved by the Washington State University Institutional Animal Care and Use Committee.

PCR.
Positive control tissue (Forelimb, E13.5; skin, E15.5; eye, E13.5) was collected and 10 cochlear ducts each (E12.5 and E15.5) were dissected out of the temporal bone and RNA harvested using Qiagen RNeasy purification kit. RNA was harvested for each tissue type on three independent occasions, and each RT PCR experiment was carried out three independent times. Reverse transcription was performed using Accupower RT pre-mix (Bioneer). PCR was performed for 35 cycles. Primers were: Kremen1 F-GTGCTTCACAGCCAACGG R-ACGTAGCACCAAGGGCTC; Kremen2 F-ACGACTAGGCATCTATGAAG, R-AAGGCACGGAGTAGGTTC; D k k 1 F -T C T G C TA G G A G C C A G T G C C , R-G AT G G T G AT C T T T C T G TAT C C 2 0 ; D k k 2 F-TG C C AC AG TC C C C AC C A AG G ATC , R-C C TG ATG G AG C AC TG G T T TG C AG 20 ; D k k 4 F-G C C C TG G T TC TG G AC T T TA AC A AC , R-C TG AC AC C TC C TG C G A AC TC TAC 21 ; Dk k l 1 F-ACTGAGGGTCTTGCTGCTGCT, R-GGAAGTTCCTAGGAAGGTCTC 22 . In Situ hybridization. In situ hybridization was performed as previously described 10 . Immunofluorescence. Immunofluorescence was performed as previously described 23,24 . Phalloidin Kremen1 gain and loss-of-function. Explant cultures and electroporations were performed as previously described 3,25 ; BrdU was used as previously described 3 . Explant cultures were maintained in vitro for six days before fixation and analysis.
Zebrafish over expression and loss-of function. Kremen1 capped RNA was generated using the mMessage Machine SP6 kit (Ambion) and purified with Ambion's MegaClear kit. mRNA was microinjected into Brn3c:mGFP zebrafish embryos at the 1-2 cell stage. Controls were injected with 1-2% phenol red (also included in mRNA to visualize the injection site). Kremen1 mutant zebrafish were obtained from the Nechiporuk lab at Oregon Health & Science University 11 . Image analysis and statistics. Bright field and epifluorescence images were captured using an Axioscope microscope fitted with an Axiocam (Zeiss). Fluorescence images were captured using a Leica SP5 or SP8 confocal microscope (Leica). Each experiment was carried out on a minimum of three independent occasions. Within each replicate, control and experimental cochlear explant electroporations were performed on litter-mate cochleae with developmental stage determined using the Theiler staging criteria 26 . All cell counts were performed using image J and statistics were performed using Graph Pad Prism statistical software (Graph Pad). In order to quantify effects on cell fate in the cochlea, we counted GFP positive cells throughout the apical quarter of the explant in the region marked by MyoVIIa ( Fig. 2A). We then used orthoganol views, and scanned down through each Z-plane to ascertain whether the GFP signal was localized to the nucleus of MyoVIIa positive cells, since MyoVIIa does not stain hair cell nuclei and the GFP was bright in the nucleus. Student's T-tests were performed on mouse and zebrafish gain and loss-of-function experiments.