The Piezo channel is central to the mechano-sensitive channel complex in the mammalian inner ear

Summary The inner ear is the hub where hair cells transduce sound, gravity, and head acceleration stimuli carried by neural codes to the brain. Of all the senses, hearing and balance, which rely on mechanosensation, are the fastest sensory signals transmitted to the central nervous system. The mechanoelectrical transducer (MET) channel in hair cells is the entryway for the sound-balance-brain interface, but the channel’s composition has eluded biologists due to its complexity. Here, we report that the mouse utilizes Piezo1 (Pz1) and Piezo2 (Pz2) isoforms as central components of the MET complex. The Pz channel subunits are expressed in hair-cell stereocilia, are co-localized and co-assembled, and are essential components of the MET complex in vitro and in situ, including integration with the transmembrane channel (Tmc1/2) protein. Mice expressing non-functional Pz1 and Pz2, but not functional Pz1 at the ROSA26 locus under the control of hair-cell promoters, have impaired auditory and vestibular traits that can only be explained if Pz channel multimers are integral to the MET complex. We affirm that Pz protein subunits constitute MET channels and that functional interactions with components of the MET complex yield current properties resembling hair-cell MET currents. Our results demonstrate Pz is a MET channel component central to interacting with MET complex proteins. Results account for the MET channel pore and complex.


Introduction
Transducer ion channels are the sensory modalities' gateway to the nervous system. The activation of transducer channels generates receptor potentials in sensory receptors, which are converted directly into primary sensory neural codes. The mechanoelectrical transducer (MET) channel complex responsible for detecting sound, gravity, and head acceleration in inner ear hair cells (HCs) identity remains an enigma 1,2 . The MET complex is housed in a hair-cell actin-based stereocilium 3 located at the tips 4 of shorter stereocilia connected to the side of longer stereocilia by a tip link 5-7 that gates the MET channels [8][9][10] . In the tip-link's absence, an anomalous HC MET current can be invoked from the cuticular plate membrane 11,12 . Mechanical deflection towards the tallest stereocilia exerts tension in the tip link, transmitting force directly onto the MET channel to increase the open probability (Po), while deflection towards the shortest stereocilia causes slack in the tip link to reduce channel Po 13,14 . Biophysical and pharmacological features of the MET channels include sub-millisecond activation time constants 15,16 , weak cation-selectivity, and large molecule permeability, such as the fluorescent dye FM1-43 [17][18][19] , and aminoglycoside antibiotic 20 .
Accordingly, the MET channel has a sizeable unitary conductance of ~100-pS 21 .
Considerable evidence has been used to identify components of the tip link and candidate protein subunits constituting the MET complex, but the molecular composition remains unclear.
Experiments have shown that the tip link comprises cadherin 23 (Cdh23) at the upper end and protocadherin15 (Pcdh15) at the lower end of the filamentous structure, whereby the two proteins form cis-homodimers that interact in trans via their opposing N-termini 22-24 . The transmembrane inner ear protein (Tmie) 25 and the Ca 2+ /integrin binding family member 2 (Cib2) protein 26,27 may be coupled to the MET channel, specifically to the tip link and the cytoskeleton. The Lipoma HMGIC fusion partner-like 5 protein (Lhfpl5) 28 likely serves as an allosteric modulator of the MET channel. The prevalent promising MET channel candidates have been the transmembrane channel like 1 and 2 (Tmc1) and (Tmc2) 29,30 . Akin to other MET complexes, Tmc1 and Tmc2 are localized at stereocilia tips 31 , and their onset in HC expression coincides with the beginning of the MET current 32 . Multiple mutations of the Tmc1 allele yield hearing loss, which alters the MET channel conductance and Ca 2+ permeability 33 . However, the association between genetic mutations and hearing loss is not unique to the Tmc alleles among candidate MET complex 34,35 .
Tmc has structural similarities to the Ca 2+ -activated Clchannel (TMEM16A) 29 , and Tmc orthologs may constitute a channel 36 37 that is mechanically sensitive in liposomes 38 . Other attempts have failed to demonstrate the Tmcs as ion channels [39][40][41] .
Piezo1 (Pz1) and Piezo2 (Pz2) are mechanically-sensitive ion channels 42, 43 that respond to shear stress and pressure changes in blood vessels and the bladder 44 and are touch-sensitive in Merkel cells 45 . Pz channels are trimeric assemblies in which each modular subunit consists of a bowl-shaped monomer with a central pore, an extracellular cap, and three curved intracellular beams, generating a lipid curvature [46][47][48] . Previous reports did not detect Pz1 in the stereocilia of HC 49 , but Pz2 functions may account for the anomalous reversed polarity current (RPI) 12 -null deletion of Pz2 aborts high-frequency hearing in mice 50 .
Here we show that HCs express Pz1 and Pz2 transcripts and proteins, and the two proteins co-localize at the tips and sides of stereocilia and cuticular plate membranes with and without Tmc1 and Tmc2 co-assembly. Therefore, despite reports regarding Pz1, our results demonstrate a role for both Pz1 and Pz2 in auditory and vestibular HCs. Although sequence homology between Pz1 and Pz2 is ~42% 51,52 , trimeric interface analyses suggest ~84% homology providing the possibility for hetero-trimeric interaction. We generated non-functional Pz1 and Pz2 knockin (ki) mouse models, floxed at the ROSA26 locus. Results demonstrate mutant (ki, MU) Pz1 and Pz2, but not wildtype Pz1 under two different HC-specific Cre-recombinase, Myosin 15 (Myo15 (mc)) and calretinin (Calb2 (cc)), cause hearing and vestibular dysfunction. Coimmunoprecipitation analyses using cochlear tissue and cell lines expressing Tmc-and Pz-fusion proteins revealed the co-assembly of Pz and Tmc and functional coupling, using fluorescence 5 resonance energy transfer (FRET) in situ and in vitro. Pz is central to the MET complex and interacts with multiple proteins, including Pcdh15, Tmc1, Lhfpl5, Tmie, and Cib2, signifying a chief role in the MET complex assembly. We propose the MET complex comprises multimers of the Pz channel subunits serving as the centerpiece, while other regulatory and indispensable binding partners confer a functional MET unit.

Pz1/2 and Tmc1/2 are expressed in inner and outer hair cells (IHCs and OHCs) of cochlear and vestibular organ HCs.
To quantify the relative abundance of candidate MET channel transcripts in HCs, we used single-molecule fluorescent hybridization (smFISH) to determine the expression of Tmc1, Pz1, and Pz2 transcripts in postnatal (P) day 10-12 (P10-P12) cochlear sections. Because the present experiments were restricted to hearing onset, and Tmc2 transcript levels are undetectable after P8 53 , Tmc2 mRNA was not evaluated. The total number of RNA molecules detected per inner and outer hair cell (IHC/OHC) was calculated across independent replicates (Figs. 1a-c). Tmc1 (in green, encoding Tmc1) was richly expressed in IHC, and OHC, followed by Pz2 and Pz1 (in red, encoding Pz2 and Pz1). Negative and positive controls are shown (Supplement Fig. 1 (S1)).
The Pz channel consists of trimeric subunits 46 . Pz1 and Pz2 sequences showed ~42% homology 51,52 . In contrast to the full-length sequence, the identified Pz1 subunit interface revealed 7 75% identical and 84% conserved residues with Pz2 (S2-3), similar to the heteromeric SKchannel subunit interacting interface that occurs in vitro and in vivo 55 , suggesting the possibility for hetero-trimeric interaction between Pz1 and Pz2 (see Method for detailed analyses). Results from proximity ligation assays (PLA) also affirmed at least ~40-nm closeness between Pz1 and Pz2 in HCs (S4). The spatial propinquity of the two channel subunits in HC stereocilia and cuticular plate suggested a functional interaction.
We reasoned that if HC-Pz subunits constitute components of the MET complex, they should localize with Tmc subunits, components of the MET complex 30 . We confirmed this prediction using Tmc2-AcGFP:Pz1-tdT and Tmc1-mCherry:Pz2-AcGFP mice 31 . At P10-P12, AcGFP-Tmc2 and tdT-Pz1 expression was detected in IHCs and OHCs and vestibular HC stereocilia (Figs. 2a-c, e). Co-localized and single-expression of Tmc2 and Pz1 was also detected at the level of the cuticular plate membrane. Fig. 2d shows AcGFP-Pz2 (green) and mCherry-Tmc1 (red) in OHC and IHC stereocilia tips. Serial z-sections revealed co-localization of Tmc1 and Pz2 at stereocilia tips. The Pz1-Tmc2 and Pz2-Tmc1 NND distribution was ~0.4±0.5 µm in IHC stereocilia and ~0.5±0.4 μm in OHC stereocilia (Figs. 2f-h). At the cuticular plate, segregation was apparent between the two proteins (S5). While these results showed Pz was located at the presumed MET-current generation site, we needed to test whether the response properties of the Pz-current resemble the HC MET currents.
Hair cell MET current displays a rectification mechanism, suggesting a reduced pore size at positive voltages 57 . We used small-amine molecules as permeant ions 58 to examine the channel-pore shape by plotting the current-voltage relation of the relative IPz1 at -90 and -120 mV (S7d-e). The channel inner and outer radii were estimated, using reported empirical fitting analyses, by plotting the amine radius vs. the relative current at -120 mV 58,59 . An inner-face pore radius of 7±4 Å and an outer-face radius of 20±5 Å was estimated (S7f-g). The findings align with HC MET channel 53 estimations and the upward concave structure of Pz 48 . Next, the Ca 2+dependent IPz1 decay was evaluated with varying concentrations of patch-pipette Ca 2+ chelators, EGTA and BAPTA. The results reflect the Ca 2+ dependence of HC MET current adaptation 60,61 (S8a-e). At -80 mV, where the driving force for Ca 2+ is greater, with 5-mM pipette EGTA, IPz1 decay was rapid (decay time constant (τdecay) at maximum IPz1 = 24±3 ms (n=15)). In contrast, at 70 mV, the τdecay of IPz1 recorded with patch-pipette-BAPTA, a more efficient Ca 2+ buffer (IPz1 τdecay at -80 mV = 38±4 ms (n=5) and 70 mV = 97±10 ms (n=6)), demonstrated the Ca 2+ -dependence of IPz1 decay. However, the displacement-response relation remained significantly unaltered using different pipette Ca 2+ buffers (S8). The findings affirmed the Ca 2+ dependence of IPz1 decay kinetics, which parallels the HC MET current kinetics.
We co-expressed mouse Pz1 and Pz2 channels and Tmc isoforms as proxies to test the functional significance of co-localization in HC stereocilia (Fig. 2, S4). Expression of mPz1 and mPz2 by themselves yielded distinct current profiles. Pz2 currents (IPz2) showed fast activation and decay kinetics (S9a). The τdecay for maximum IPz2 and IPz1 was <10 and >10 ms, respectively.
The mechanically-activated current decay for mPz1 and mTmc2 transfected N2A cells increased compared to mPz1 alone, and the τdecay decreased ~4.5-fold, indicating the regulation of mPz1-mediated current by Tmc2. Another visible and consistent effect of singly and coexpressed mPz1, Tmc1, and Tmc2 was the robust amplitude of mPz1-Tmc1 current compared to mPz1-Tmc2 or mPz1 current alone. For similar transfection conditions, the total IPz1/T1 density (292±40 pA/pF, n=10) was ~2.4-fold greater relative to IPz1 (122±18 pA/pF, n=10), and IPz1/T2 (113±22 pA/pF, n=10). Co-expression of Tmc1 or Tmc2 with mPz1 shifted displacementresponse relationships rightward of IPz1 alone, with Tmc2 exerting the most pronounced effect (S9). In contrast, the steepness of the displacement-response curves at X1/2 remained relatively unchanged (S9f). Because the present experiments were limited by the relatively slow rise-time of the mechanical stimulator and because of coarse mechanical stimulation, quantitative analyses were not performed of the displacement-response relationship and the time constants of current activation.

Evidence for the functional coupling of Pz and Tmc using Förster resonance energy transfer (FRET)
Motivated by Pz and Tmc co-localization in HC stereocilia and IPz property modulations by Tmc1/2 co-transfection (IPz/T1/2), we tested additional evidence for functional coupling between the two proteins. In OHC stereocilia, we examined the co-expression of Pz2-GFP and Tmc1-mCherry in mice that were crossbred from heterozygotes for each construct, whereas in N2A cells in vitro we examine co-transfected Pz1 tagged with mClover3 (mPiezo1-mClover3) and Tmc1 tagged with mRuby3 (mTmc1-mRuby3). Results examined after FRET-fluorophore activation are shown in S10. FRET efficiencies were quantified using acceptor photobleaching (S10). Results from FRET experiments in vitro showed FRET efficiency of ~10%. Experiments conducted using OHCs stereocilia expressing Pz2-GFP and Tmc1-mCherry showed similar FRET efficiency (S10), suggesting that the two proteins are located spatially and are functionally coupled.

IPz1/T1 exhibited similar pharmacological properties to MET current.
MET current block by aminoglycosides is a pharmacologic trademark 20, 56 , consequently, IPz or current ensuing from a Pz-and Tmc association should be affected to resemble dihydrostreptomycin (DHS) effects on HC MET currents. Thus, we tested for IPz1 sensitivity towards DHS. DHS application on N2A cells expressing mPz1 alone inhibited the IPz1 in a dosedependent manner. Results show an IC50 of 270±135 µM (n = 7), with ~65% of the maximum IPz1 was blocked by 200 µM neomycin (n = 2, S11 a-c). Typically, HC MET current is at least ~20-fold more sensitive towards DHS than IPz1 DHS-sensitivity in N2A cells (S11). Since previous results showed that Tmc and its mutations reduced HC MET current sensitivity to DHS 62, 63 , we tested IPz1/T1 block by DHS. Pz1-Tmc1 co-transfection yielded current IC50 of ~70±6 nM (n=7) in the presence of DHS (S11). The ~1000-fold increased sensitivity of IPz1/T1 towards DHS provided additional evidence for the Pz-Tmc functional association.

mPz1 in lipid bilayer exhibited similar single-channel conductance as MET current
HC MET single-channel conductance is ~100 pS 64,65 . We subjected the Pz1-transfected N2A-cell membrane to pressure at varying voltages, and the results showed a Pz1 unitary conductance of ~36 pS (S12), consistent with previous reports 42 but at odds with HC MET channel conductance. However, IPz is modulated by regulatory and interacting proteins, and the forceresponse properties depend on membrane lipids 66 . Moreover, recent studies suggest that HC Tmc may exhibit lipid scramblase activity 67 . Using purified mPz1 from brain lipid extract in lipid bilayer with defined lipid composition (PE: PS: PC 5:3:2), we demonstrate that the unitary mPz1 current was ~10 pA at -100 mV with single-channel conductance of ~100 pS, in keeping with the reported HC MET single-channel conductance. DHS, GsMTx4, and ruthenium blocked the mPz1channel activity in lipid bilayers, consistent with HC MET's current pharmacology (S13).

Knockdown of Pz by CRISPR/Cas9-mediated genome engineering
Data accrued motivated us to test Pz functions in vivo in the inner ear, using non-functional Pz1 and Pz2 gene knockin and HC-specific Cre-recombinase strategies. While other gene manipulation strategies, such as mutated-gene knockin at the promoter sites could have been adopted, we opted against such a design to mitigate potential embryonic lethality because of the channels' essential roles in other tissues. Additionally, the amassed data suggested that Pz1 and Pz2 are expressed in HCs and the two may form heterotrimers (S2-3).
The non-functional forms of Pz1 (Pz1 MU ) and Pz2 (Pz2 MU ) were flanked by loxP sites and inserted into intron 1 of the ROSA26 locus in the two ki mice. Additionally, a control wildtype (WT) Pz1 WT was generated. Both the Pz1-mutant and Pz2-mutant ki (denoted as Pz1 MU , and Pz2 MU ) coding sequences contain four amino acids in the C-terminal domain (CTD), where the sequence "MFEE" was replaced with amino acids "AAAA" (Fig 3a-b). The sequences correspond to amino acids 2493-96 for Pz1 and 2767-2770 for Pz2. The selection of EE:AA residue substitutions stems from earlier reports 68 demonstrating that while mutant Pz proteins with substitutions of the AA residues can be targeted to the plasma membrane, they are non-conductive. Similar results were confirmed for the MFEE:AAAA mutants transfected in N2A cells (S14). We performed analyses in silico to predict the top five most likely off-target mutations that could occur with the CRISPR approach and ensured the mice did not harbor them. Mice were also backcrossed unto C57 mouse lines for ten generations to mitigate potential off-target effects. Genotyping strategies are described ( Fig. 3c-d; see Methods). We denote the mutant ki as Pz1 MU and Pz2 MU . A control mouse with wildtype (WT) Pz1 at the ROSA 26 locus was also generated (Pz1 WT ).
Calretinin is encoded by the Calb2 gene, subserving as a major Ca 2+ buffer 69,70 in HCs, whereas myosin 15 is encoded by the Myo15 gene, showing early postnatal expression in HCs 71,72 . Calb2 is also expressed in auditory neuron subtypes 73 . We used the two Cre lines as complementary strategies to regulate Pz knockin mouse lines. Procedures, generation, genotyping, and validation of Myo15-(mc) and Calb2-Cre (cc) mouse lines are described in the Methods section and Supplement 15 (S15). We first confirmed HC-specific Cre expression by crossing mc and cc with the Ai9-tdT (Ai9(RLC-tdT; JAX strain #00709) and Calb2-GFP mouse lines (S15). Additionally, we inactivated Pz1 and Pz2 in HCs individually, using Pz1 MU and Pz2 MU mice 74,45 crossed with mc and cc mouse lines to assess the potential differential roles of the two proteins. We will refer to the resulting lines after crosses as mc-Pz1 MU and cc-Pz1 MU , mc-Pz2 MU and cc-Pz2 MU . The controls used were cc, mc or Pz1 WT , where the wildtype Pz1 channel was knockin at the ROSA 26 locus. The auditory and vestibular traits in the control mice were similar and thus used interchangeably.
Pz conditional knockout mouse models Pz1-knockout (Pz1 ko ) cross with the Cre-lines were assessed as mc-Pz1 ko , mc-Pz2 ko , cc-Pz1 ko , and cc-Pz2 ko . The crosses of heterozygous floxed-lines with and without the Cre allele yielded offspring with a 1:2:1 ratio for the wild-type (WT), heterozygous, and homozygous floxed allele, respectively, suggesting no embryonic lethality. Pz1 MU and Pz2 MU mice appeared normal without obvious behavioral defects (rotarod test, data not shown) and body weight issues (S16a). Real-time RT-PCR analysis of wild-type and Pz1/2 MU cochlear samples at P10 demonstrated normal Pz transcript levels (data not shown). We performed immunostaining of whole-mount cochlear tissue using the Pz1 antibody and detected correctly targeted Pz1 in HC stereocilia in the Pz1 MU mouse samples as expected (S16b).

The auditory and vestibular traits of Pz MU and Pz ko mice
We analyzed auditory brainstem responses (ABR) to broadband click-and pure tones of 4, 8, 16, and 32 kHz stimuli at various sound-pressure levels (SPL). ABRs were measured at 4and 8-weeks of age. Representative ABR traces show reduced characteristic waveform peaks and increased latencies using click sounds at 60-and 80-dB SPL (Fig. 4a). From 4 to 8 weeks, mc-Pz1 MU mice exhibited an ~20 to 35 dB threshold increase, while mc-Pz2 MU mice showed an ~40 dB threshold shift compared to age-matched control mc and mc-Pz1 WT mice for broadband click (Fig. 4b) and across all tone pip stimuli (S17). Results indicated hearing dysfunction by progressively elevated ABR thresholds, with severe hearing loss greater than 70-80 dB SPL thresholds for mc-Pz1 MU and mc-Pz2 MU and profound hearing loss by 12 weeks (threshold at or above 90 dB for click sound) and concomitant threshold elevations across all tone stimuli.
However, the elevated thresholds were more significant at 32 kHz than at 4-16 kHz in 4-week-old mice, and a progressive increase in ABR thresholds was observed at 8 weeks in the mu relative to control mice (S17). Almost complete hearing loss (threshold ≥ 90 dB) was recorded for the double mutant mc-Pz1-2 MU mice by 8 weeks of age (Fig. 4b). ABR thresholds for the cc-Pz1 MU and cc-Pz2 MU were similar to mc-Pz1 MU and mc-Pz2 MU mice (Fig. 4, S17). In contrast, null deletion of individual Pz1 or Pz2 alone produced modest ABR threshold elevations, using Myo15 and Calb2 Cre lines and click sound as documented (Fig. 4). Attempts to generate knockouts of both isoforms, thus far, have failed. We suspect embryonic lethality despite using multiple Cre lines.
Increased ABR thresholds may suffice to capture IHC malfunction, whereas distortion product otoacoustic emissions (DPOAEs) are acoustic measurements of OHC activity. DPOAE thresholds for the mc-Pz1 MU and mc-Pz2 MU mice were elevated relative to age-matched control mice (Fig. 4d). Similar ABR and DPOAE results were obtained in the cc-Pz1 MU and cc-Pz2 MU mouse lines. Results suggest compromised integrity of electromechanical transduction and a decline in the competence of IHC and OHC functions.
Pz channels' functional roles in vivo in vestibular end-organ HCs were examined by measuring compound action potentials, the vestibular sensory evoked potential (VsEP) 75 . VsEP originates from HCs, the vestibular nerve, and its central relay activity in response to linear acceleration pulses. Figure 4e shows VsEP waveforms for cc control and cc-Pz1/2 MU and mc-Pz1-2 MU double mice. The summary data show that VsEP thresholds in single Pz1 or Pz2 mutant knockins are significantly elevated relative to those for the age-matched controls (Fig. 4f).
Compared to the double mutant mice (Pz1-2 MU ), relative to controls, the VsEP threshold differences were even more remarkable. However, results from a null deletion of individual Pz1 and Pz2 show no significant effects (Fig. 4f).

Mechanotransduction is attenuated in Pz MU mice.
We monitored the extent-of-altered HC transducer currents in the Pz MU mice to explain the observed hearing phenotype. The rapid uptake of the lipophilic dye FM1-43 occurs through the MET channels 17,19 . At rest, HC MET channels' Po is ~0.1, so we reasoned that a nonconducting Pz subunit would further reduce the resting Po and inhibit HC FM1-43 dye uptake. We examined FM1-43 dye loading in mc-Pz1 MU and mc-Pz2 MU HCs from the apical-to-middle cochlear region, representing characteristic frequencies (CFs) of 4 to 10 kHz from P12 mouse cochleae and compared them to aged-matched mc controls. Local perfusion of FM1-43 dye resulted in intense labeling first at the hair bundle level 1 (L1), followed by dye membrane-partitioning and diffusion across the cuticular plate (L2) and basolateral (L3) aspects of HCs. On average, chronological images from the three levels show intense labeling after 3-10 sec of 10 µM FM1-43 exposure in the mc control and faint labeling in mc-Pz1 MU HCs (S18). Z-stack images were taken at the cuticular plate level in 5-sec intervals post-dye exposure (S18). The time constants (τ) of dye loading at L2 were: 23±3 sec (n=5) for controls; 61±6 sec (n=4) for Pz1-ki m-c ; and 58±8 sec (n=4) for Pz2-ki m-c . For clarity, only L2 data are plotted (S18). We conclude that the expression of a nonfunctional Pz subunit suffices to reduce but not altogether abolish FM1-43 dye loading in IHCs and OHCs.
To examine whether the reduced FM1-43 loading in mc-Pz MU mouse HCs matches with transducer currents and to obtain direct evidence for Pz's role in HC MET currents, we recorded IHC and OHC MET currents in the whole-cell configuration from P10-P12 cochleae ( Fig. 5a-

b).
HCs were held at -80 mV. IHC and OHC MET current in control mice showed varied activation and adaptation kinetics, as shown previously 49,76,77 (Fig. 5a-b). The size of the maximum MET current was significantly smaller in mc-Pz1 MU and almost negligible in the mc-Pz1-2 MU compared to control HCs (mc) (Fig. 5c). The current-displacement plots showed a notable difference in MET currents from mc compared to mc-Pz1 MU and mc-Pz2 MU mice. While the normalized currentdisplacement curve for control HCs was well-fitted with a two-state Boltzmann function, the relationships derived from the mc-Pz1 MU and mc-Pz2 MU IHCs were best matched with a threestate Boltzmann function (Fig. 5d). Similar results we obtained for OHCs MET currents (Fig. 5c, f). We examined the RPI following the post-BAPTA application to disrupt the tip links. Figure 5g shows the MET current generated using sinusoidal mechanical stimuli. In Figs. 5h-I, we illustrate the remaining current after a 5-mM BAPTA solution bath application. The expanded trace, in blue, shows the features of the RPI. At P12, the maximum RPI (in pA) from IHCs was significantly smaller in mc-Pz1 MU compared to mc control mice (Fig. 5i).

OHC electromotility in Pz MU mutant and control mice was relatively intact
DPOAE measurements, which represent the coarse assessment of cochlea-sound amplification, suggest mild OHC dysfunction in the Pz1 MU and Pz2 MU mouse models. We determined the roles of the Pz channels in OHCs by assessing OHC electromotility. Electromotility describes OHC functions expressed visibly as length changes in the OHC cell-body, mediated by a voltage-dependent gating-charge movement and measured as nonlinear capacitance (NLC) changes 78,79 . OHC NLC at P18-P21 in control and Pz1 MU and Pz2 MU mice showed substantial differences between the control and the mutant mice (S21). However, the results alone may not account for the DPOAE threshold increase 80 .

Hair cell bundle structure in Pz1 MU and Pz2 MU mice
Mechanoelectrical transduction appears to be a requisite for HC maturation 81 , and with dwindling current magnitude in Pz MU HCs, we expected HC morphological alterations and degeneration. We counterstained HC-stereocilia actin with phalloidin-TRITC in P12 and 4-week (P28)-old cochleae. We observed no gross changes in hair bundle structure and the characteristic three rows of OHCs and one row of IHC in the mc-Pz MU P12 cochlea (data not shown), but by 4week, loss of OHCs was apparent in the cochlea (Fig. 6a). We used scanning electron microscopy (SEM) for high-resolution analyses to evaluate HC and bundle morphology. Cochlear HCs of mc-Pz MU mice at P21 had normal morphology, including intact hair bundles and stereocilia linkages in both IHCs and OHCs resembling controls (Figs. 6b-d), but the loss of a few OHCs was apparent at the basal cochlea at P21 (Fig. 6d). In older mc-Pz MU mice >P42, IHC enlargement, and multiple OHC loss was evident (S20). Profound OHC degeneration at the cochlear basal and middle turns was observed at P56 and older, and other hair bundle abnormalities were noticeable, such as fused OHC stereocilia and basal-turn IHC degeneration (S20).

Co-immunoprecipitation of Pz and MET complex proteins
We reasoned that if the Pz protein is an integral and central component of the MET protein complex, it should co-immunoprecipitate as a complex in cochlear tissue. Cochlear tissue was harvested from Tmc2-AcGFP:Pz1-tdT and Tmc1-mCherry:Pz2-GFP mice, whereas negative controls were from non-transgenic mice cochlear tissues. Results show Pz1 coimmunoprecipitated with Tmc2, anti-GFP, and as demonstrated in immunoblot using anti-tdT antibodies. Similarly, to determine whether Pz2 forms multiprotein complexes with Tmc1, we performed immunoprecipitation using anti-GFP and immunoblotting using anti-mCherry antibodies. We show that Pz1 and Tmc2, and Pz2 and Tmc1 interact in a complex (Fig. 7a).
Because of limited cochlear tissue and unreliable Pz and Tmc protein antibodies, we were compelled to use the HEK 293 expression system and tagged MET apparatus proteins. Results from HEK 293 cells transiently transfected with either mCherry-Tmc1 or mCherry-Tmc2 confirmed the mCherry-tagged Tmcs pulled-down and anti-mPz1 labelled Pz1, given that HEK 293 cells endogenously express Pz1 (Fig. 7b) 82 . A complement experiment using cells transfected with mCherry-Tmc1/2 and Flag-mPz1 showed positive pull-down with anti-mCherry or anti-Flag, revealing the interaction between mTmc1/2 and mPz1. Next, similar experiments were performed using tagged proteins, including His-Tmie, Myc-Lhfp15, HA-Pcdh15, and V5-Cib2 co-expressed individually with Flag-mPz1. Co-immunoprecipitation investigations revealed that Flag-mPz1 successfully immunoprecipitated with His-Tmie, Myc-Lhfp15, HA-Pcdh15, and V5-Cib2, suggesting the proteins form complexes (Fig. 7c-f). The original gel blots are presented in

Discussion
The sound and balance gateways to the brain are through MET-channel activation in HC stereocilia. The identity of HC MET channel-complex composition has been a puzzle, and multiple attempts have revealed essential elements of the protein complex, but the pore-forming protein remains debatable 13 . We demonstrate that IHCs and OHCs express Pz1 and Pz2 transcripts. Notwithstanding the caveats associated with expression systems, the expression of mTmc1 or 2 alone did not bear a mechanically activated current, but Pz1/2 did. Co-transfection of Pz1/2 with Tmc1/2 produced a mechanically-gated current with increased density and varied decay kinetics Evidence for HC Pz2 transcripts and protein expression was previously shown 49, 50 , but detection was missed for Pz1 49 . A more sensitive in situ hybridization strategy, smFISH 84 , allowed for Pz1 and Pz2 RNA detection and quantification in HCs, and high-resolution microscopy enhanced by protein-fluorophore fusion expression revealed Pz and Tmc proteins localized at the stereocilia tips and sides in propinquity (Figs 1-2). Given the close localization of Pz1/2 in HC stereocilia, predicted structural and functional interaction between the two subunits in HC and N2A cells, and the trimeric structure of the Pz subunits, we sought to attenuate Pz channel  (Fig. 4).
Tmc is an integral element of the MET complex in mammalian HCs 29 and may serve as ion channels in invertebrates 36 . Cell lines expressing Tmc provide evidence for Pz1 interacting with MET complex proteins, including Tmie, Lhfp15, Pcdh15, and Cib2, indicating that the Pz subunits are central to the MET complex in both cochlear tissue and a reconstituted system. Tmc may be an essential allosteric regulator for Pz-channel functions in mammalian HC. Among the prevailing evidence that Tmc serves as the pore-forming subunits for MET channels is a set of elegant experiments in which specific Tmc1 residues were mutated into cysteine, and cysteinemodification reagents were able to mediate MET currents 29 . In a scenario where the Pz-channel complex was intact except for the mutated Tmc1, it is conceivable to use cysteine modification to restore the MET complex function. Suppose Tmc is an allosteric anchor for force transmission through lipid interaction with Pz and is required for Pz engagement in the HC setting. In that case, results from the previous report can be envisioned. An improbable setup for two parallel ion permeation pathways, one for the Tmcs and the other for the Pzs, can be envisioned and tested in the future. However, the simplest explanation for the current findings is that the Pz channel is the central pore-protein, and a vital allosteric regulatory protein, such as Tmc, sculpts the MET current properties. From the findings, Pz-Tmc current sensitivity to aminoglycosides also offers persuasive evidence for the importance of Tmc in the function of Pz in HCs. The findings provide   Fig. 1 (S1). c, The mean number of RNA molecules detected per HC was calculated as described 85 (Supplement Fig. 1, S1). Data are plotted to show individual replicates (animals) and mean ± SEM. There were significant differences at the p<0.05 level for        The protein backbone is shown in a ribbon with basic and acidic amino acids in the sphere.
Coloring for Pz1 is as follows: ribbon = yellow, conserved basic = light blue, conserved acidic = red/orange, non-conserved basic or acidic = same color as protein ribbon. Pz2 coloring is as follows: ribbon = purple/pink, conserved basic = dark blue, conserved acidic = dark red, nonconserved basic or acidic = same color as protein ribbon.

Supplement Figure 3 (S3). Alignments of Pz1 and Pz2 interacting interfaces.
Sequence alignment for Pz1 trimer interface. Alignment includes human Pz1 (Piezo1_H), mouse Pz1 (Piezo1_M), human Pz2 (Piezo2_H), mouse Pz2 (Piezo2_M). The sequence coloring is based on amino acid properties according to the "Taylor" scheme found in Jalview.  Densitometric data of mean pixel intensity were measured in arbitrary grayscale units (a.u) as described 19 . The number of animals tested is indicated for controls and genotypes. Frames were taken at the three rows of OHCs at the apical one-third of the cochlea, and a similar loading pattern is observed at the middle third of the cochlea. The change in fluorescence was fitted with an exponential function, and the time constants (τ, in secs) of FM1-43 dye loading in control, mc-Pz1 MU , and mc-Pz2 MU apical cochlear OHCs at L2, 23+3 (n = 5), 61+6 (n = 4) and 58+8 (n = 4). Figure 19 (S19). Cellular degeneration of Pz MU mice. a, Whole-mount cochlea of a ~6-week-old Calb2-Cre (cc) control mouse showing the low-frequency segment at the ~6-kHz cochlear region. Myo7a, the hair cell marker, is stained (white), and the stereocilia marker is stained (green) for actin. Scale bar = 10 µm. b, The high-frequency segment at the ~32-kHz cochlear region. ABR thresholds for the mouse were (dB); 4-kHz=40, 8-kHz=20 dB, 16-kHz=15, 32-kHz=60, and click=35. c-d, A 6-week-old cc-Pz2 MU cochlea shows 6 and 32 kHz segments.

Supplement
Note enlarged IHCs (red arrows) at the low-frequency ~6-kHz segment and lost OHCs (* in red) at the high-frequency ~32-kHz segment. Recorded ABR thresholds for the mouse were (dB); 4- Cm is the total capacitance of the cell, Cln is the linear capacitance, Cv is the nonlinear capacitance, V is the membrane potential, Vh is the voltage at half-maximal nonlinear charge transfer, e is the electron charge, k is Boltzmann's constant, T is the absolute temperature, z is the valence, and Qmax is maximum nonlinear charge transfer. For the plots shown, values for myo15-cre (mc; controls), (Cln (pF), z, Qsp (fC/pF) and Vh (mV) were (9, -0.9, 120, -47); mc-Pz1 MU , (7, - Fig. 4c).
Calb2-cre PCR assay was performed with a pair of primers for wild-type, wt: (525 bp: F:5"-GTTGATAGGAAGGTCCATTCGGT-3" and R:5"-CAGAAGCCTAAATCATACAGCGAAG-3"), and homozygote, homo: 271 bp: F:5"-ACTTAAACCCACTCTCACCTCTTT-3" and R:5"-TACGGTCAGTAAATTGGACACCTT-3")) (Supplement Fig. 7 VsEPs were recorded to normal and inverted stimulus polarities (1024 points, 10-µs per point, 128 responses per averaged waveform). Three response parameters were quantified: threshold, peak latencies, and peak-to-peak amplitudes. The threshold is measured in dB re 1.0 g/ms and is defined as the stimulus amplitude midway between that which produced a discernible VsEP and that which failed to produce a response. The threshold measures the sensitivity of the gravity receptor end organs, utricle, and saccule. Thresholds, latencies, and amplitudes are averaged for each group and analyzed using variance (one-way ANOVA). A. Patapoutian.

In vitro and ex vivo fluorescence resonance energy transfer (FRET) experiments
Mouse Pz1 tagged with mClover3 (mPz1-mClover3) and mTmc1 tagged with mRuby3

FM1-43 dye loading
A stock solution, 10 mM, was prepared in PBS. A new working solution was prepared.
Cochlear samples were incubated in 10 µM of FM1-43 for 5 s, washed with PBS, and observed under the Leica SP8 confocal microscope as described 19 .

Mechanical Stimulation.
Mechanical stimulation was achieved using a fire-polished glass pipette (tip diameter 1μm) at 60° to the recorded cell. Downward movement of the probe toward the cell was driven by a Clampex-controlled piezo-electric crystal micro stage (E660 LVPZT Controller/Amplifier; Physik Instruments). The probe was typically positioned close to the cell body without visible membrane deformation. Mechanical displacement was applied with a fire-polished blunt pipette driven by a piezo-electric device. To assess the mechanical sensitivity of a cell, a series of mechanical steps in ~140 nm increments were applied every 10 to 20 s, which allowed for full recovery of mechanosensitive currents. Inward MA currents were recorded at a holding potential of -80 mV.
For I-V relationship recordings, voltage steps were applied 750 ms before the mechanical stimulation from a holding potential of -60 mV.
For mechanical stimulation of hair bundles, we employed stepwise and sinewave stimulation of OHCs and IHCs using a piezo-driven fluid-jet hair bundle mechanical stimulator on P10-P12 cochleae and recorded IHC and OHC MET currents in the whole-cell configuration. We represented the bundle displacement in the form of applied piezo-driven voltage. Bundle displacement was not calibrated for each cell because of variations in stimulating probe positions relative to the stimulated hair bundle.

Estimating the pore size
We followed the approach used for the MET channel in hair cells to estimate the Piezo1 pore size from the inner face and amines radius. Plots were fit empirically to where a is the radius of the amines, r is the radius of the channel, and A is a scaling factor. Internal solutions were made with monovalent amines of different sizes, with size estimated as previously described from CPK models 87,88 The external face pore size data were fit to Ix/INa= (A (1-a/r) 2 /a where Ix/INa is the current ratio, a is the radius of the amine compound, r is the radius of the channel, and A is a scaling factor 58 . The external solution used the same approach to substitute Na + with amines of different sizes. Amines were purchased from Sigma Aldrich.

Single-channel recordings and analysis
Single channel gating properties were recorded using a Planar Lipid Bilayer Workstation Channel data was acquired with pClamp 10 software (Molecular Devices, Sunnyvale, CA). Data analysis and plots were made with OriginPro 2018b and Corel Draw 2020.

Sequence and structural evidence for Pz1 and Pz2 heterodimers:
To assess whether Pz1 and Pz2 could form heterodimers, we explored the concept of conservation of salt bridges as the matching of positive and negative charges is widely used as a tool to design drug and protein ligands. Using the Pz1 structure (PDB: 6BPZ), we selected one chain (see S2) (a) of the Pz1 structure and expanded our selection to include any amino acid on the remaining two chains (b, c) within 5 Å. We then kept this selection on the two chains (b, c) and expanded this selection to any amino acid on the original chain (A) by 5 Å to back-select amino acids on chain (A) to obtain all inter-chain interacting amino acids in the Pz1 trimer. We then selected the basic and acidic amino acids and mapped these onto the Pz1 mouse structure (S2-3). We then calculated a local sequence alignment between mouse-Pz1, human-Pz1, mouse-Pz2, and human-Pz2. The local alignment included the trimer interacting amino acids and surrounding amino acids to align the interacting amino acids properly. We determined the conservation of the interacting amino acids, which reports a 74.59% conservation and an 83.6% amino acid chemical property conservation (S3). The sequence on Pz2 that matched the Pz1 basic and acid salt bridges was also mapped onto the mouse-Pz2 structure (PDB 6KG7). Due to this very high conservation, the chains could be interchanged. We verify that the conserved ionic charges match in 3D coordinates and the location of the non-conserved charges. We replaced Chain B in Pz1 with Chain B from the Pz2 structure using the matchmaker (UCSF Chimera) to make a 2:1 Pz1:Piezo2 structure. We repeated the procedure to replace chain C of Pz1 with chain C of Pz2 to build a 1:2 Pz1:Pz2 model. Visual inspection of these models shows the pore domain contains the most conserved charged residues while the cap domain contains the most non-89 conserved charged residues (S2). In addition, the 3D positions of the acidic and basic residues have high conservation, suggesting that heterotrimer structures are structurally plausible.

Statistical analyses
Where appropriate, pooled data are presented as means ± SD. Significant differences between groups were tested using t-test and ANOVA, where applicable. The null hypothesis was rejected when the two-tailed p-value < 0.05 is indicated with *, < 0.01 with **, < 0.001 with ***.
The number of mice and neurons is reported as n.