Prestin and the cochlear amplifier

Abstract

In non-mammalian, hair cell-bearing sense organs amplification is associated with mechano-electric transducer channels in the stereovilli (commonly called stereocilia). Because mammals possess differentiated outer hair cells (OHC), they also benefit from a novel electromotile process, powered by the motor protein, prestin. Here we consider new work pertaining to this protein and its potential role as the mammalian cochlear amplifier.

Some form of mechanical amplification is a general property of auditory organs (Manley, 2001; Gopfert et al. 2005). In fact, all hair cell-possessing vertebrates express an ancient universal mechanism, based on mechanical feedback by stereovilli (Martin & Hudspeth, 1999; Martin et al. 2000). A second amplifier arose based on OHC somatic electromotility (Brownell et al. 1985; Kachar et al. 1986; Ashmore, 1987). Although the basic phenomenon of electromotility is well known (Holley, 1996), one of the significant remaining issues pertains to the appropriate partitioning of mammalian amplification between its stereovillar and somatic components.

The nature of prestin

When molecular motors are discussed, attention is usually focused on the linear motors myosin, kinesin and dynein. These cytoskeletal proteins generate motion by using nucleoside triphosphate hydrolysis to produce conformational changes in proteins. Other motors, such as helicases, ribosomal motors, chaperonins, etc., also require ATP hydrolysis for energy. Prestin, however, is different. Because it is a direct voltage-to-force converter, it is probably unique in the animal kingdom. Its closest functional relative is the bacterial flagellum, which is driven by a high-speed rotary motor, dependent on proton current across the bacterial cell membrane (Berg, 1975).

Relying upon the knowledge that only OHCs express the putative gene that codes for the motor protein, we identified prestin (Zheng et al. 2000) using suppression subtractive hybridization PCR (Diatchenko et al. 1996). Prestin's amino acid sequence shows that it belongs to an anion-transporter family, solute carrier protein 26 (SLC26A). Prestin shares the family structure, including a highly conserved central core of hydrophobic amino acids (∼400 a.a.) predicted to form 10–12 transmembrane domains, as well as cytoplasmic N- (∼100 a.a.) and C-termini (∼240 a.a.). Although many members of SLC26A are glycoproteins, glycosylation/deglycosylation does not affect function (Karniski et al. 1998; Matsuda et al. 2004). Although prestin is most closely related to SLC26A6 (Dallos & Fakler, 2002; Mount & Romero, 2004), the human and mouse orthologues of A6 have only 78% amino acid identity. In contrast, prestin is a highly conserved protein with 92.7% of amino acids being identical among four different mammalian species: human, mouse, rat and gerbil (He et al. 2006). Such a high degree of conservation is not common among other SLC26A members. Significant changes in prestin's primary sequence occurred after the split between mammalian and avian lines, suggesting that prestin evolved in order to fit special mammalian needs (Franchini & Elgoyhen, 2006).

Aside from its highly conserved primary sequence, mammalian prestin also has several unique features that differ from other members of the family. First, prestin expresses voltage-dependent charge movement and motility. Second, prestin is abundantly expressed in OHCs (> 10⁷ per cell). Third, prestin exists as stable tetramers (Zheng et al. 2006), which probably form the 11 nm particles observed in the OHC's lateral membrane (Forge, 1991; Kalinec et al. 1992). Although there is little evidence suggesting the existence of oligomeric forms in other members, a possible multimeric form of pendrin was reported, even though the majority of pendrin molecules exist as monomers (Porra et al. 2002). Fourth, although the basic function of SLC26A members is to transport anions (Dallos & Fakler, 2002; Mount & Romero, 2004), this is not prestin's principal role. Recent theoretical work suggests, however, that it is best to model prestin as an electrogenic anion exchanger (Muallem & Ashmore, 2006). Unequivocal experimental verification of anion transport is not yet available.

Early work indicated that prestin is voltage dependent (Santos-Sacchi & Dilger, 1988) and, in analogy with voltage-gated ion channels, it was subsequently assumed that the voltage sensor is a distinct charged moiety displaced across the membrane in response to a change in membrane potential. This displacement presumably initiates a conformational change in the motor component of the protein, producing an alteration of surface area. Charge displacement is manifest as a transient current at the onset and cessation of membrane-potential steps. The total charge displaced at a given potential is the product of the total displaceable charge (Q_max) and the probability (p_S) that the displaced charge is in one of its two principal states. This behaviour is expressed as a Boltzmann function. Inasmuch as the charge displaced is voltage dependent, a nonlinear capacitance (NLC) can be associated with it, according to the relationship C = dQ/dV (Ashmore, 1990). Nonlinear capacitance is easier to measure than motility, and it has been shown to provide a surrogate representation (Santos-Sacchi, 1991).

It is reasonable to assume that the voltage-sensing charged group is different for prestin than in other SLC26A proteins, which produce no motility. With this in mind, Oliver et al. (2001) altered each charged, non-conserved amino acid in the putative membrane-interacting region of prestin, individually or in groups. Surprisingly, no combination of mutations eliminated NLC or altered its gain. These results led to the suggestion that the voltage sensor may not be an intrinsic component of the protein, but an extrinsic ion. Using inside-out and outside-out membrane patches, Oliver et al. demonstrated that intracellular Cl⁻ functions as the extrinsic voltage sensor. Subsequent investigations (Fakler & Oliver, 2003; Rybalchenko & Santos-Sacchi, 2003; Santos-Sacchi et al. 2006) showed that as intracellular Cl⁻ concentration decreases, the amount of charge transferred also decreases and voltage sensitivity shifts in the depolarizing direction. The direction of shift implies that the net charge moved across the membrane is positive. Thus, two alternatives exist to the idea that Cl⁻ is the voltage sensor. It is possible that monovalent anions need to attach to a binding site and their combination, with net positivity, is translocated across the membrane. Alternatively, chloride binding could enable an allosteric change, thereby allowing a positive gating charge to be moved (Rybalchenko & Santos-Sacchi, 2003). The two alternatives are demonstrated in the cartoons of Fig. 1.

Figure 1. Two models of prestin gating by voltage, in which the presence of intracellular chloride is an essential factor in both, but the gating mechanisms are different.

The long (or extended) state of the molecule corresponds to hyperpolarization of the cell; the short (or compact) state to depolarization. The no-chloride case is arbitrarily modelled as long. In A, Cl⁻ is assumed to associate with a positive binding site and the combination is translocated across the membrane. In B, chloride binding enables a positive gating particle to unlock and be translocated. The cartoons depict incomplete transporters. If full anion transport is demonstrated, a pore region needs to be incorporated.

It was also shown (He & Dallos, 1999, 2000) that voltage-driven motility is accompanied by a voltage-dependent change in axial stiffness. Voltage change can modulate cell stiffness over a range of about 10-fold, with a potential to influence micromechanics. He et al. (2003) also demonstrated that interference with prestin's function, via substitution of pentane sulphonate for chloride, reduces cell stiffness to ∼1/3 of its normal value. Because stiffness changes occur on a cycle-by-cycle basis without significant attenuation as frequency increases, it is conceivable that stiffness change may be the primary voltage-dependent process in OHCs. If so, electromotility would be a simple consequence of changing the stiffness of a preloaded ‘spring’.

Finally, prestin, like other transducers, demonstrates reciprocity (Iwasa, 1993), which can be modelled by assessing the molecule's piezoelectric properties. When viewed in this context, the efficiency of conversion from mechanical force to electrical charge is ∼20 fC nN⁻¹. This number is four orders of magnitude greater than that obtained for the best man-made material (Dong et al. 2002).

Prestin mutations

In order to understand the mechanism underlying electromotility, various mutations, including point mutants, chimeric mutants (hybrid proteins made from portions of prestin and other SLC26A members) and truncation mutants have been investigated (He et al. 2006). Almost all mutants fall into two groups. For group one, NLC is maintained with or without some shift in V_1/2, the membrane potential at which half of the molecules are in the contracted state. Group-two mutants lose NLC, in most cases due to their inability to insert into the plasma membrane (PM). Like other members of the SLC26A family, prestin contains a sulphate transporter and antisigma-factor antagonist (STAS) domain, the latter located in the C-terminus. Mutations occurring in this region result in improper targeting and loss of partial or complete function (Taylor et al. 2002; Karniski, 2004). Prestin is no exception (Zheng et al. 2005).

Mutations in other regions of prestin can also cause improper PM targeting. In a heterologous system, prestin mutants are often misfolded or aggregated, resulting in their retention and accumulation in the endoplasmic reticulum and other cytoplasmic membranes. In fact, deletion of more than 21 a.a. at the N terminus, or more than 32 a.a. at the C terminus, results in loss of function, probably due to improper targeting (Navaratnam et al. 2005; Zheng et al. 2005). Other family members demonstrate similar problems, such as Pendred syndrome (Rotman-Pikielny et al. 2002; Taylor et al. 2002). In fact, there is a strong negative correlation between the severity of the human disease phenotype and the level of SLC26A protein present in the PM (Scott et al. 2000; Karniski, 2004).

Prestin as amplifier

It is known that the low-pass cutoff frequency of the OHC membrane filter is at most ∼1 kHz (Housley & Ashmore, 1992; Preyer et al. 1996). Thus, the cell's receptor potential, assumed to control motility, is attenuated above the cutoff frequency and is progressively less effective. In contrast, the motor itself is fast, i.e. OHC motility can be induced at greater than 70 kHz (Frank et al. 1999). In order to envision how motility occurs in vivo, ingenious schemes have been proposed to overcome the filter problem. Among these are various models utilizing prestin's piezoelectric properties to compensate for the voltage attenuation (Mountain & Hubbard, 1994; Dong et al. 2002; Ospeck et al. 2003; Spector et al. 2003), interaction of prestin with localized chloride conductances (Rybalchenko & Santos-Sacchi, 2003) and reliance on extracellular voltage gradients (Dallos & Evans, 1995). This latter conjecture has been tested experimentally (Fridberger et al. 2004) and found to merit consideration. Finally, the recognition that cochlear amplification can be construed as a negative feedback process (Mountain & Hubbard, 1994) was used to demonstrate that while individual OHCs may have low-pass filter properties, the resulting system bandwidth is much wider when their collective gain is high (Lu et al. 2006). Although it is not known which of the above possibilities, or their combinations, might be employed in the living cochlea, there clearly are several plausible means whereby the bandwidth limitation can be overcome.

Prestin knockout mice

Measurements in prestin knockout (KO) mice demonstrate that prestin is required for normal auditory function. In the absence of prestin, OHCs do not exhibit electromotility (Liberman et al. 2002). In vivo, brain-stem evoked-potential thresholds increase by ∼50 dB (Liberman et al. 2004) as do compound action potential thresholds (Cheatham et al. 2004). Frequency selectivity is also absent (Cheatham et al. 2004). In order to associate these physiological deficits with a change in OHC motor function, it must be established that forward transduction is normal in KO mice. Homozygotes have wildtype-like nonlinear responses including harmonic and intermodulation distortion (Liberman et al. 2004; Cheatham et al. 2004, 2006), CM pseudotransducer functions (peak ± ac (alternating current) response versus peak ± sound pressure), both summating potential polarities, as well as normal uptake of the dye AM1-43 via transducer channels (Cheatham et al. 2004). Data from OHCs isolated from KO mice also demonstrate large asymmetrical transducer currents similar to those in wildtype controls (Jia et al. 2006). Thus mechano-electrical transduction appears to be normal in mice lacking prestin and therefore changes in sensitivity and frequency selectivity cannot be attributed to changes in forward transduction. Because prestin KO mice exhibit a progressive apoptosis of both inner and outer hair cells (Liberman et al. 2002; Wu et al. 2004), older adult mice are probably better models for deaf adult humans with prestin mutations (Liu et al. 2003). Data also suggest that heterozygous prestin mice have normal sensitivity and frequency selectivity, as well as near normal levels of prestin protein and OHC somatic electromotility (Liberman et al. 2004; Cheatham et al. 2005). Because prestin mRNA is ∼50% of wildtype (Liberman et al. 2002), prestin protein expression appears to be autoregulated (Cheatham et al. 2005).

Somatic and stereovillar amplification

The stereovillar motor can do work against an external load (Benser et al. 1996), may have the necessary speed (Kennedy et al. 2005) and can produce some nonlinear amplification (Chan & Hudspeth, 2005). However, there is no evidence that it dominates mammalian amplification. Because non-mammalian sense organs lack OHCs and prestin, it is assumed that the stereovillar amplifier is their principal means of boosting hearing sensitivity, while the newer somatic motility evolved to fulfil some specific mammalian requirement, possibly associated with their extended frequency range of hearing. If stereovillar amplification were the primary mechanism throughout this range, the evolutionary advantage of adding somatic motility becomes obscure. Hence, we assume that both mechanisms are present in mammals. Although de novo somatic motility probably dominates, spontaneous otoacoustic emissions may be more intimately tied to stereovillar mechanisms. In this context, we note reports showing spontaneous oscillations of stereovillar bundles (Crawford & Fettiplace, 1985; Rusch & Thurm, 1990; Martin & Hudspeth, 1999; Camalet et al. 2000). In contrast, no spontaneous oscillation has ever been seen in OHC length or stiffness.

Conclusions

Remaining problems include an experimental examination of the molecule's putative transport function, its means of voltage sensing, its companion proteins, the nature of its conformational change and its possible interactions with surrounding lipid (Raphael et al. 2000; Zhang et al. 2001). Many of these questions may be answered when a structural description becomes available. The most important outstanding issue, however, is the apportioning of mammalian amplification between stereovillar and somatic motility (Chan & Hudspeth, 2005; Jia et al. 2006; Kennedy et al. 2006).