Recent clinical findings with memantine should not mean that the idea of neuroprotection in glaucoma is abandoned


Neville N. Osborne
Nuffield Laboratory of Ophthalmology
Oxford University
John Radcliffe Hospital
West Wing, Level 6
Headley Way
Oxford OX3 9DU
Tel: +44 1865 234 797
Fax: +44 1865 234 795


Loss of vision in primary open-angle glaucoma (glaucoma) is caused by retinal ganglion cells dying at a seemingly steady and variable rate in different patients. Present treatments for all glaucoma patients are inadequate and a goal to rectify this is to discover appropriate drugs or chemicals (neuroprotectants) that can be taken orally to slow down retinal ganglion cell death and have negligible side-effects. It was therefore of great disappointment to learn earlier this year that the one clinical trial conducted to test the efficacy of memantine as a neuroprotectant for glaucoma was unsuccessful. In this article, I consider the mechanisms by which retinal ganglion cells may die in glaucoma and suggest that memantine may have benefited patients taking it but to a level that was difficult to detect with present methodologies. Ganglion cells are induced to die by different triggers in glaucoma, suggesting that neuroprotectants with multiple modes of actions are likely to reveal clearer results than was found for memantine. Therefore, the idea of neuroprotection in glaucoma must not be abandoned.

Neuroprotection in glaucoma implies the use of drugs or chemicals to slow down whatever causes loss of vision without influencing intraocular pressure (IOP). Loss of vision in glaucoma is thought to be caused by the death of ganglion cells. In order to be effective, a neuroprotectant must reach the optic nerve head (ONH) and/or ganglion cells and will therefore probably have to be taken orally. Because it will reach other parts of the body, any side-effects of an appropriate neuroprotectant must be reduced to a minimum.

It was a great disappointment when earlier this year the much-awaited finding of an unmasked second phase 3 clinical trial, examining the safety and efficacy of oral memantine as a neuroprotectant for the treatment of glaucoma, was announced. The study showed that the progression of the disease was significantly lower in patients receiving a higher dose of memantine than in patients receiving a low dose of memantine, but there was no clear benefit compared to patients receiving placebo. The results of this second phase 3 trial failed to corroborate the results of the first phase 3 trial. As a consequence, there was no basis for the approval of memantine for use in glaucoma. Needless to say, the financial cost of these clinical trials has been enormous, raising many questions. It is clear that current definitions of clinical trial endpoints for glaucoma are far from ideal; this is one reason why the memantine trial took a long time and, not surprisingly, proved very expensive (estimated cost $80 million over a 7-year period). Perhaps, with better parameters to define more precise endpoints and/or an extension of the trial, the efficacy of memantine might have achieved statistical significance.

Memantine was originally developed to be a superior dopaminergic drug to its derivative amantadine, which was used for the treatment of Parkinson’s disease (Schwab et al. 1969). Laboratory studies eventually showed that memantine has little affinity for dopamine as well as gamma-aminobutyric acid (GABA), serotonin and α-adrenergic receptors (Osborne et al. 1982) but acts as an effective antagonist at glutamate N-methyl-d-aspartate (NMDA) receptors (Chen et al. 1992; Parsons et al. 1993). Because NMDA receptors are implicated in a spectrum of cognitive, motor and behaviour functions, it could be deduced that the inhibition of their normal functions by use of an NMDA antagonist like memantine would result in multiple serious side-effects. This was simply not found to be the case in patients treated with memantine.

Basic science studies eventually provided a reason why an NMDA antagonist such as memantine can be used clinically. Memantine binds at a site within the NMDA ion channel but cannot access this binding site unless glutamate has previously bound to its receptor and induced the channel to open. Memantine is therefore an ‘open channel blocker’ and does not compete with glutamate for its binding site. Moreover, the affinity and kinetics of the channel block of the NMDA receptor by memantine are distinctive. Memantine is a moderate affinity, uncompetitive NMDA receptor antagonist with strong voltage dependency and fast kinetics (Bormann 1989; Chen et al. 1992; Parsons et al. 1993). This argues for the case that when extracellular levels of glutamate are in the normal physiological range, memantine is likely to have no effect at a ‘non-toxic concentration’. However, when the extracellular glutamate levels are elevated excessively to cause pathological damage in situ, then at this concentration memantine becomes effective by having access to a site within the open NMDA channel. This would certainly explain why memantine is both tolerated and has an efficacy for the treatment of dementia (Ditzler 1991; Winblad & Poritis 1999; Jain 2000).

The clinical use of memantine where neuronal death is caused by excessive activation of NMDA receptors is based on a sound footing. Therefore, it would seem reasonable to suggest that if ganglion cell death in glaucoma is caused by excessive activation of their NMDA receptors then memantine would be worth exploring as a potential neuroprotectant. Undoubtedly, the reports by Dreyer et al. (1996) and Brooks et al. (1997) that the vitreous humour level of glutamate was significantly elevated in patients, dogs and monkeys with glaucoma provided a catalyst for such a trial to be initiated. Unfortunately, one of the authors associated with all these studies was later discredited and attempts by others to reproduce some of these findings have not been successful (Levkovitch-Verbin et al. 2002; Wamsley et al. 2005). As a result, reliance on these data as a basis for undertaking a clinical trial is hard to justify. The relevance of any elevation of glutamate in the vitreous humour in relation to ganglion cell death is also unclear (Salt & Cordeiro 2006). Also, because NMDA receptors are located to both amacrine and ganglion cells in the retina (Massey 1990; Fletcher et al. 2000; Zhou & Dacheux 2004) it is not easy to explain why only ganglion cells should die in glaucoma if this is caused by the excessive activation of their NMDA receptors. In retinal ischaemia, extracellular glutamate is undoubtedly raised and is the major cause of death of both amacrine and ganglion cells (Osborne et al. 2004). Glaucoma, in contrast, is a disease where ganglion cells are specifically affected: other neurone types may eventually become affected in advanced glaucoma because of transynaptic degeneration caused by the death of ganglion cells (Osborne et al. 1999). In addition, for NMDA receptor activation to occur both glutamate and glycine are required as co-agonists; their combined effects can be modulated by a variety of extracellular substances including magnesium, zinc and nitric oxide (Chen & Lipton 2006; Lipton 2007). Thus a rise solely in the extracellular glutamate level surrounding an NMDA receptor may not result in these receptors being affected.

It is universally accepted that the primary site of injury in glaucoma occurs in the ONH and is caused by various interdependent factors (Flammer et al. 2002; Quigley 2005; Pache & Flammer 2006; Boland & Quigley 2007; Flammer & Mozaffarieh 2007). The insults to the ONH are thought to result ultimately in mechanical and/or ischaemic damage to the different components that make up the region, including glial cells (microglia and astrocytes), ganglion cell axons and the laminae cribrosa (Hernandez & Pena 1997; Osborne et al. 2001; Boland & Quigley 2007). Because glaucoma patients are only treated following diagnosis, any design of successful therapies must focus on slowing down whatever causes the ganglion cells to die and/or reducing the impact of the insults to the ONH. In some patients an insult to the ONH is caused by elevated IOP; the progression of the diseases can be blunted by the use of IOP-lowering drugs (Quigley 2005). However, in many patients IOP-lowering drugs are ineffective, making it necessary to explore the possible use of neuroprotectants to slow down ganglion cell dysfunction/death. It is therefore imperative that the idea of neuroprotection is not abandoned because of the apparent negative findings of the memantine trial, but rather that we learn from it. What needs to be asked is whether memantine was the right choice of drug and whether the methodologies used were not sensitive enough to detect a small but significant positive effect. Is it possible that memantine produced a benefit equivalent to the lowering of IOP by more than 5 mmHg compared to the best available IOP-lowering drug available, yet this was not detected?

It is important to stress that ganglion cell death in glaucoma is thought to occur by apoptosis (Kerrigan et al. 1997; Quigley 2005) and that this process takes place within a very short period. It could therefore be argued that once glaucoma is initiated ganglion cells become weaker because of cumulative damage and die at different rates by apoptosis. This would be consistent with the ‘cumulative hypothesis’ (Clarke et al. 2000, 2001), which predicts that apoptotic ganglion cell death would increase over time; however, this appears not to be the case in glaucoma (Kerrigan et al. 1997). Indeed, what we do know about apoptotic ganglion cell death in glaucoma (Kerrigan et al. 1997), combined with the fact that the disease does not appear to progress faster as aging occurs, suggests that cell death is exponential – as is found to occur in neuronal degenerations generally (Clarke et al. 2000, 2001). An alternative explanation is therefore to predict that after injury to the ONH all ganglion cells exist in a reduced energetic state, meaning that they are still functionally normal but are less resistant to any secondary insult. Secondary insults to such ganglion cells are proposed to occur primarily because of a gradual rise in the variety of chemicals released from activated glial cells (Osborne et al. 2001, 2006) (Fig. 1). Importantly, it is predicted that a gradual increase of specific chemicals in the extracellular space affects ganglion cells differentially depending on their receptor profiles and other individual characteristics. In this way, ganglion cell apoptosis might be triggered at a fairly constant rate over many years. This is also consistent with the ‘one-hit model’ of cell death in various other neuronal degenerations (Clarke et al. 2000, 2001). Thus, a ganglion cell having 10 NMDA receptors would logically be less likely to be triggered to die by apoptosis following a defined slight increase in extracellular glutamate than neurones that have a thousand NMDA receptors. It is known that ganglion cell death is variable following an intravitreal injection of NMDA, supporting the contention that ganglion cells probably have different densities of NMDA receptors.

Figure 1.

 Hypothetical cascade of events to account for differential ganglion cell death as occurs in glaucoma (32,39). It is proposed that the initial insults in glaucoma take place in the optic nerve head (ONH), resulting in mechanical and/or ischaemic damage. Insults to ganglion cell axons result in the cells existing at a reduced energetic state. In this state, ganglion cells still function normally but are more vulnerable than before. An insult to the ONH will also result in glial cells (microglia and astrocytes) being activated, causing them to release different chemicals into the extracellular space. Initially, these chemicals will be removed by the retinal Müller cells but this process will become inefficient with time. A gradual increase of these chemicals in the extracellular space will occur over time. Some of these chemicals will be potentially toxic (e.g. glutamate) or protective [e.g. transforming growth factor-β (TGF-β)] to neurones containing the appropriate receptors. Ganglion cells are likely to be particularly affected in the retina because they exist at a reduced energetic state. As a result, ganglion cells will be triggered to die by apoptosis at varying times depending on their receptor profile and density for different receptor types. Another insult to ganglion cells at a reduced energetic state is proposed to be light, and this might vary for different neurones depending on their number of mitochondria and position in the globe (Osborne et al. 2006, 2008).

The hypothesis as depicted in Fig. 1 provides an explanation of how ganglion cells may die by apoptosis at different times in glaucoma. Moreover, if laboratory studies conducted with memantine are taken into account, it is possible to understand why the results from the first phase 3 memantine clinical trial suggested optimism and why they were not eventually substantiated in the second trial. Laboratory studies have shown that the NMDA receptor ion channel in cell membranes maintained at a healthy resting potential is largely blocked by magnesium ions but if the tissue is energetically compromised (e.g. by injury or anoxia) then the membrane potential will depolarize and this will then relieve the magnesium ion block of the NMDA receptors. As a result, even a slight elevation of extracellular glutamate will have a much greater effect on activation of the NMDA receptor (Mayer & Westbrook 1987; Meldrum 2000). It could therefore be argued that once glaucoma is initiated in the ONH, ganglion cell NMDA receptors become potentially dysfunctional because of injury to their axons and any elevation of extracellular glutamate. Activated astrocytes, as seen in glaucoma (Hernandez & Pena 1997), are known to release glutamate (Bezzi et al. 2004) and other substances such as tumour necrosis factor-α (TNF-α) (Tezel & Wax 2000) and transforming growth factor-β (TGF-β) (Mahesh et al. 2006). In contrast to energetically compromised ganglion cells, NMDA receptors associated with amacrine cells in the glaucomatous eye will not be affected by small increases in extracellular glutamate.

Chemicals that might rise steadily in the extracellular space as glaucoma progresses to cause ganglion cell apoptosis are shown in Fig. 1. The apoptotic trigger(s) for any single ganglion cell will not be constant but will depend on, for example, the ganglion cell’s receptor profile and other factors such as their number of mitochondria and axonal length and position within the globe (Osborne et al. 2006). This would mean that apoptosis can be triggered in different ganglion cells for different reasons and at time intervals that can be many years apart. For example, some ganglion cells containing different amounts of NMDA receptors may additionally possess TGF-β, TNF-α and endothelin receptors. It is also possible that some ganglion cells lack NMDA receptors, although animal studies suggest this to be unlikely (Fletcher et al. 2000). In this scenario, the use of an NMDA receptor antagonist like memantine may only be very effective to slow down apoptosis in cells containing a defined density of NMDA receptors. Thus, a neuroprotectant that has a single mode of action like memantine will have a limited positive effect at slowing down ganglion cell death and might be restricted to cells containing a defined density of NMDA receptors. This significant positive influence to the patient might have been difficult to demonstrate clinically and could account for the results of the memantine trial.

Significant neuroprotection in glaucoma might only be demonstrated clearly in clinical trials if the process of ganglion cell death could be prolonged in a greater number of neurones than was possible with memantine. This might be achieved by a substance like epigallicatechin gallate (ECGC), which has multiple actions (Mandel et al. 2005) that theoretically will blunt a number of potential secondary insults, and also possesses the appropriate profile of a neuroprotectant for glaucoma (i.e. it can be taken orally to reach the retina and produces negligible side-effects). It is worth noting that one secondary insult to ganglion cells in glaucoma might be light (Osborne et al. 2006, 2008) where the antioxidant characteristics of a substance like ECGC will be effective. Using a combination of drugs (e.g. memantine plus EGCG) or even a cocktail of antagonists/agonists acting directly on not only NMDA receptors is another possible way forward. Present treatments for glaucoma remain inadequate: instead of abandoning the idea of neuroprotection in glaucoma, we should use the results from the memantine trials positively to stimulate further research.


The author thanks Professor Thomas Salt for his valuable comments.