Click here for the PDF version

Contents

Mammals show little or no capacity to replace hair cells [1]. Progressive sensory loss in the auditory and vestibular systems leads to hearing and balance deficits that compromise the quality of life for millions of people and cost our health services many millions of pounds. Non-mammals do not suffer the same problem. Birds have sophisticated ears with the ability to process language and music yet they have not lost their capacity to recover from extensive loss of hair cells [2]. They continuously replace their vestibular hair cells and can replace auditory hair cells following damage. A potential explanation for lack of replacement in the mammalian auditory system is that the organ of Corti has an extremely specialised cellular architecture and a unique tuning mechanism based upon the motility of outer hair cells [3]. This is correlated with a much higher frequency range than that found in non-mammals. However, this explanation is not entirely convincing because hair cell replacement occurs only at minimal levels even in mammalian vestibular epithelia [4, 5].

Why should we be optimistic about regenerative therapies for the replacement of mammalian hair cells in vivo? The low level of regeneration in vestibular epithelia provides crucial evidence that at least some hair cells can be replaced. In other systems, for example the spinal cord, the regeneration of neurons has long been considered impossible. However, there is now strong evidence that regeneration of spinal neurons can lead to functional recovery. The problem is complex. It is necessary not only to find ways of stimulating regrowth of nerve fibres but also to remove inhibitory molecules on surrounding cells and barriers caused by scar tissue [6]. Functional activity in associated cells is also an important component of successful recovery. In thinking about the inner ear we should consider the need to replace hair cells by stimulating cell division and differentiation, but we should also be aware of possible negative effects due to scar tissue and inhibitory molecules in the immediate cellular environment. A further consideration is that the recent explosion in attention to stem cell research has revealed the presence of multipotent cells in many tissues, including the central nervous system [7]. We should search for such cells in the adult cochlea [8].

Therapeutic approaches to regeneration are far from the stage of clinical trials, and distinct from approaches designed to prevent sensory loss. Numerous ototoxic drugs are employed throughout the world, including aminoglycosides such as gentamycin and anti-cancer drugs such as cisplatin. Animal experiments have provided compelling evidence that antioxidants, anti-apoptotic drugs, specific neurotransmitter antagonists and growth factors can prevent hair cell loss following exposure to excessive noise or to ototoxic drugs [9]. Some clinical trials have already taken place, and progress should be significant over the next few years. However, the prevention of hair cell damage and death is based upon quite different mechanisms to those required to stimulate regeneration, which involves cell division and differentiation.

The development and behaviour of cells depends upon their tissue environment. They receive signals from their neighbours, from the connective tissues around them and from soluble factors secreted either by themselves or other cells. When a hair cell dies there are a number of potential outcomes. The first is that the neighbouring supporting cells fill the space and seal the cell layer, forming a scar. This is essential to maintain separation between the perilymph and endolymph and thus the endocochlear potential. It is what happens in our own ears. We currently know little about the biology of this response in terms of the nature of the remaining cells and the potentially inhibitory environment that they create. Are they the same as supporting cells and can they be stimulated to divide or redifferentiate? The second outcome is that one of the supporting cells detects the loss of the hair cell and changes into one. This 'phenotypic conversion' might occur in some animals. The third outcome is that one of the supporting cells divides, leaving one daughter to differentiate as a hair cell and the other to replace the supporting cell. This is the principle method adopted by the chick [2] but it also occurs at low levels in mammalian vestibular epithelia [4, 5]. It reflects the facts that hair cells and supporting cells share a common progenitor during development and that supporting cells, unlike hair cells, are not irreversibly differentiated. Thus, the aim of many studies in regeneration is to stimulate proliferation of supporting cells and differentiation of hair cells (see animation).

Much of the research in regeneration revolves around the identification of master, regulatory genes that govern fundamental cellular processes. For example, the differentiation of pancreatic beta cells is governed by the transcription factor Pax4 [10]. This gene regulates expression of a functionally coherent set of genes that endow the cell with the ability to secrete insulin in a glucose-dependent manner. In therapeutic terms, it may thus be possible to treat insulin deficiency by inserting Pax4 into existing pancreatic cells. However, it would be much more convenient to develop a drug that can activate Pax4 in vivo. Activin, a member of the TGF-beta superfamily, upregulates Pax4 and stimulates insulin production in vitro [11]. This example illustrates the principle that knowledge of the key regulatory genes and associated signalling pathways can provide valuable insights into potential therapeutic approaches. Application of this principle does of course depend upon cell competence. Not all cells possess the genetic background to respond to activin or Pax4 in the same way.

Several molecules that regulate proliferation and differentiation in hair cells have already been identified. The cyclin-dependent kinase inhibitor p27^kip1 regulates the proliferation of sensory epithelial cells during development, and is the first known marker for the patch of epithelial cells that are destined to become sensory epithelium [12]. The cells that express it stop dividing and then proceed to differentiate as supporting cells and hair cells. Mice that lack p27^kip1 develop more hair cells and supporting cells than normal [13, 14]. Interestingly, normal developing hair cells downregulate p27^kip1 whilst supporting cells maintain it. There is no evidence that hair cells ever re-enter the cell cycle so other cell cycle inhibitors must control suppression of cell proliferation. The fact that supporting cells maintain p27^kip1 suggests that they might be induced to re-enter the cell cycle if its function can be blocked. A more detailed analysis of the factors that regulate supporting cell proliferation has been undertaken in cultures of rat vestibular epithelia stimulated with human recombinant glial growth factor 2 (rhGGF2) [15]. This preparation has helped to reveal the molecular cascade of messengers and enzymes that trigger a mitogenic response in mammalian supporting cells.

After the patch of sensory epithelial cells has been defined, individual hair cells are specified. We do not know the precise mechanism but it involves signalling mediated by 'notch' receptors in the cell membrane [16]. These receptors are activated by membrane-bound ligands on neighbouring cells and involve a feedback system ensuring that hair cells do not differentiate next to each other. Downstream of these events, two transcription factors known to be essential for hair cell differentiation are Math1 [17] and Brn3c [18]. No hair cells develop or survive in the ears of mice that lack either of the genes for these factors. Math1 is the mouse atonal homolog 1 of the Drosophila gene atonal, which is essential for hair cell differentiation in fly mechanosensory sensilla. If it is transfected into cells of the greater epithelial ridge (GER) in mouse organs of Corti, then it can convert these cells into hair cells [19]. This exciting result raises the possibility of replacing lost hair cells by transfecting Math1 into the sensory epithelia. However, there are a number of problems. Not all cells in the cochlea are competent to respond to Math1 and the effect has only been observed in neonatal tissue, suggesting that adult cells from the GER are unresponsive. Even if hair cells in the greater epithelial ridge are functional, they are not placed to receive appropriate stimulation from the basilar membrane. There are also major challenges in the development of appropriate techniques in therapeutic gene delivery. Nevertheless, it is worth looking for extrinsic signals that can activate Math1 and to search for other regulatory genes that are required during normal development. Structural and functional recovery has recently been reported in vestibular epithelia of adult guinea pigs following lesions generated with gentamycin and treatment with a cocktail of growth factors with retinoic acid [20]. It is not yet clear if a real regenerative response explains this result and the cocktail is not an attractive therapeutic solution. However, elucidation of the mechanisms by which it works may be extremely important for future drug design.

To search for more information on genes that regulate differentiation of specific cell types in the organ of Corti, we have established conditionally immortal cell lines from the sensory epithelia of developing mouse otocysts [21]. Cells were derived from transgenic mice carrying a temperature-sensitive variant of the immortalising 'T' antigen from the SV40 virus under the control of a promoter inducible with g-interferon. Cells cultured at 33 °C with g-interferon continue to proliferate without further differentiation, but when transferred to 39 °C without g-interferon, they stop dividing and continue to differentiate for a limited period. By taking cells at defined times and locations during the development of the otocyst, it is possible to represent critical events in specific cell types. The cell line UB/OC-1 expresses numerous hair cell genes under differentiating conditions and represents the early stages of hair cell differentiation. It was cultured for 14 days at 39 °C and messenger RNA was prepared daily throughout this period. The RNA was then screened against Affymetrix, oligonucleotide arrays representing up to 11,000 mouse genes. The approach allows us to identify genes involved in hair cell differentiation, and to correlate temporal expression profiles of genes that have a high probability of being functionally related [22]. We can apply the same approach to cells in which key genes, such as p27^kip1, have been functionally inhibited, and start to form an understanding of the appropriate gene networks and signalling pathways.

Conditionally immortal cell lines are not just convenient research tools since they have been used successfully in transplantation studies in the brain and retina [9]. Although it is possible that some types of stem cell, notably embryonic stem cells, might potentially be able to differentiate into any cell type, it may be more useful to use cells derived from as close to the target tissue as possible. This issue is reflected in the fact that specific regulatory genes achieve different results in different tissues. For example, Math1 regulates differentiation of cerebellar granule cells as well as hair cells. Conditionally immortal cell lines are particularly flexible because they can be characterised and cultured in unlimited quantities without the need to derive them repeatedly from embryonic tissue. Some preliminary transplantation experiments with stem cells have been attempted in the central auditory pathways [23] and the inner ear [24]. Replacement of hair cells by this method will not be easy because appropriate cells must be targeted to specific regions of a tightly sealed epithelial layer. However, in the context of cochlear implants the presence of replica hair cells might provide important, secreted factors that help to attract and maintain a nerve supply to the electrode. It may also be possible to replace spiral ganglion neurons directly for the same purpose. At this stage, very few animal experiments have been conducted and the prospect of real progress lies a long way into the future.

In research terms, there are several reasons to be optimistic about the prospects for hair cell regeneration. New discoveries regarding the regenerative potential of the central nervous system and other organs are teaching us to be more open-minded about tissue regeneration. In clinical terms, hair cell regeneration is a long way away. However, within the next 5 years we will certainly see significant advances in protective measures against sensory loss, and will be gathering information about the potential for regeneration at a much faster rate. Thus, it would be worth considering a serious programme of interaction between trainee clinicians and basic researchers to ensure that we harness this potential as quickly as possible.