Our data revealed a hitherto unknown process of structural remodeling of broken tip links that is necessary for regeneration of MET in auditory hair cells (). After stereocilia link disruption, CDH23 largely disappears while PCDH15 migrates from the side link area around the circumference of the stereocilium and becomes available for forming the upper end of a nascent tip link. At the same time, PCDH15 molecules at the lower end of a disrupted tip link also migrate circumferentially but remain at the stereocilia tip. Then, transient tip link filaments are assembled using PCDH15 molecules at both sides of the filament (, red). These transient tip links mediate MET current with normal amplitude but an abnormal extent of adaptation. Finally, new or perhaps refolded/repaired CDH23 molecules (, blue) replace PCDH15 at the upper end of the tip link, completing the regeneration of mature tip links and the concomitant recovery of MET adaptation. This final step seems to be essential for long-term maintenance of MET because CDH23 is apparently required for MET in mammalian hair cells 
. The high affinity interaction between CDH23 and PCDH15 
may favor the replacement of PCDH15 at the upper end of the tip link with newly synthesized or repaired CDH23, whose appearance in regenerating links is delayed relative to PCDH15.
Formation of a transient PCDH15–PCDH15 tip link and its subsequent replacement with PCDH15–CDH23 link.
Because HL5614 antibodies recognize an extracellular epitope that seems to be common to numerous PCDH15 alternative splice isoforms 
and because these isoforms may have redundant function 
, we cannot distinguish whether nascent regenerating tip links have at their ends identical or different PCDH15 isoforms. It is yet uncertain whether nonclassical cadherins like PCDH15 or CDH23 could form stable extracellular links through homomeric interaction. Evidence for CDH23–CDH23 homomeric interaction have been reported 
, but the same group later suggested that this interaction, along with homomeric PCDH15–PCDH15 interaction, is likely to have a low affinity 
, which is consistent with our data that PCDH15–PCDH15 tip links are only temporary intermediates in the tip link regeneration process. A recent report also provided no direct evidence for homomeric interaction between the first EC domains of either CDH23 or PCDH15 
. However, heteromeric interaction of different PCDH15 isoforms is still possible. Alternative splicing of the Pcdh15
gene can generate several protein products that differ at the N-terminus (Figure S7
) and provides a dazzling variety of potential heteromeric interactions between different PCDH15 isoforms, which are known to be targeted specifically to different stereocilia compartments 
. In addition, a secreted isoform of PCDH15 lacking the transmembrane domain 
may also be recruited to help span the distance between adjacent stereocilia.
It is yet unclear exactly how molecular remodeling of tip links affects the extent of adaptation of MET responses. It is possible that the decreased length of PCDH15–PCDH15 tip links changes the angle of inclination of tip link and therefore affects the extent of adaptation (Figure S8A
–B). However, such changes are predicted to affect the slope of the MET current–displacement curve, which was not observed in our experiments (). The extent of adaptation may also decrease if the elastic harmonin-based complex connecting the upper end of the tip link to the cytoskeleton is replaced to a complex with nonlinear saturating stiffness, for example with PCDH15 that is bound to the cytoskeleton with molecules that “slip down” along the actin core when tip link tension reaches a certain level (Figure S8C
). Indeed, recent studies identified harmonin and THMS as interacting partners of CDH23 and PCDH15, correspondingly 
. These very different interacting partners may contribute to different mechanical properties of macromolecular complexes connecting CDH23 and PCDH15 to the actin cores of stereocilia. Another possibility is that CDH23 at the upper end of the tip link may affect homomeric interaction of parallel PCDH15 filaments at the lower end of the link, thereby affecting the adaptation machinery, or recruit additional adaptation-related components of tip link, which can be seen as a third “foot” at the lower end of the link 
. In any case, our data open new perspectives to study the molecular basis of MET machinery by characterizing redistribution of the proteins normally associated with PCDH15 and CDH23 during tip link regeneration.
Delayed acquisition of fully developed adaptation is a common finding of studies investigating formation of the mechanotransduction apparatus either after tip link disruption 
or during development 
. However, there are different interpretations of these data. According to one model, the fully functional mechanotransduction machinery is assembled somewhere around the base of the hair bundle and transported by myosin-based motors to the tips of stereocilia, where it undergoes final “refinement” 
. Another model postulates that the mechanotransduction apparatus is assembled at its final location at the tips of the stereocilia from components that arrive there in a stepwise progression 
. We did not find evidence for upward movement of stereocilia links or PCDH15 molecules along the stereocilium (), which is consistent with direct assembly of the transduction apparatus at the tips of stereocilia. On the other hand, replacement of PCDH15 with CDH23 at the final step of molecular remodeling of the tip links may be considered as a “refinement” of the transduction apparatus. We should note, however, one caveat of the direct comparison of our data with the previously published reports. In this study, we investigated regeneration of mechanotransduction in young postnatal mouse IHCs, capitalizing on the unique sensitivity of all stereocilia links in these cells to BAPTA treatment. Previous reports described regeneration or development of mechanotransduction in different hair cell types—e.g., hair cells of chick basilar papilla 
, mouse vestibular hair cells 
, and cochlear outer hair cells of rat 
and mouse 
. The diversity of preparations suggests caution in comparing results. For example, an unusually fast recovery of MET current–displacement relationship () may result from specific mechanical properties of the IHC bundles or from better control of vertical positioning of the stimulating probe in our experiments ().
Our data reconcile two previous molecular models of tip link formation. One model proposed that the tip link has a PCDH15–CD3 isoform at the lower end and a PCHD15–CD1 isoform at the upper end of the link 
. A more recent model suggests that the tip link is built with PCDH15 and CDH23 at the lower and upper ends of the link, correspondingly 
. We show that the molecular composition of the tip links is not static. During tip link regeneration, the hair cell temporarily utilizes PCDH15 molecules at both ends of the link to recover MET responses until mature PCDH15–CDH23 links are established.
It is generally believed that regeneration of tip links and mechanotransduction in vitro
recapitulates the assembly of the transduction apparatus in vivo
(also see and ) 
. Therefore, the two-step molecular remodeling process that we have discovered may underlie formation of the tip links in the developing hair bundle, their turnover in the mature bundle, and their regeneration after intense acoustical stimulation in order to safeguard hearing.