ARF: Outline for us what you think are the steps lead to AD pathogenesis? What role do beta amyloid and tau play in AD? Are they related or independent events?

Mesulam: Amyloid plaques (AP) and neurofibrillary tangles (NFT) are the two diagnostic markers of Alzheimer's disease (AD). The neuropsychological features of AD are closely correlated with the distribution of the NFT and therefore favor a disease process revolving around neurofibrillary degeneration. The genetics, however, favor a disease process revolving around the AP, principally because mutations in the amyloid precursor protein (APP) are sufficient to cause AD. The inability to reconcile these two aspects of AD has prevented the formulation of a unified theory of pathogenesis.
     My hypothesis of AD pathogenesis shows that all genetic causes and risk factors of AD (amyloid mutations, presenilin mutations, apoE4, estrogen deficiency, multiple trauma, aging) can increase the physiological burden of neuroplasticity. It is hypothesized that the resultant intensification of the plasticity burden leads to an initially adaptive upregulation of tau phosphorylation and APP turnover, to the subsequent formation of NFT and AP as independent consequences of excessive plasticity-related cellular activity, and to the eventual loss of neurons, dendrites and synapses as the ultimate expression of plasticity failure. The two pathological markers of AD are therefore independent manifestations of a more fundamental process through which the many different genotypes of AD cause an identical clinical and neuropathological phenotype.

ARF: How do you account for the anatomical pattern and cellular specificity in the progression of AD?

Marsel Mesulam: All AD-promoting factors are likely to perturb processes that normally facilitate neuroplasticity. The hypothesis is based on the assumption that the resultant barriers to neuroplasticity occur at downstream dendritic and synaptic sites and that they trigger a reactive (or compensatory) upstream intensification of plasticity-related perikaryal activity. In other words, AD-promoting factors create a setting where neurons must work harder to meet neuroplasticity needs at their axonal and dendritic terminals. Over many years, such compensatory processes would lead to chronically high and eventually unsustainable levels of plasticity-related cellular activity.

ARF: Is there evidence for enhanced plasticity-related cellular activity in experimental models or AD brains?

Marsel Mesulam: In vivo and in vitro experiments show that high levels of neuroplasticity tend to be associated with the increased expression and phosphorylation of tau (Brion, 1994; Black, 1996; Busciglio, 1987; Lovestone, 1997; Trojanowski, 1993; Viereck, 1989). For example, the olfactory bulb of the adult rat, a region that continues to show very active neuroplasticity, expresses particularly high levels of the more extensively phosphorylated, fetal forms of tau (Viereck, 1989; Lovestone, 1997). Furthermore, transfected PC12 cells overexpressing tau extend neurites more rapidly, and neurite extension in response to NGF is associated with a 10-20-fold induction of tau (Drubin, 1985; Esmaeli-Azad, 1994).
     In the course of the processes that lead to AD, a chronically high demand for plasticity-related activity could thus upregulate the expression of tau, favor its phosphorylation, and potentially promote the polymerization of tau into NFT. The NFT produced by such a sequence of events would initially appear within limbic-paralimbic neurons because these neurons have the highest baseline levels of plasticity and would thus have the highest exposure to compensatory upregulations of plasticity-related cellular activity.
     The resultant cytoskeletal dysfunction in these limbic-paralimbic neurons would eventually lead to a degeneration of their dendrites and a loss of their synapses at axonal projection targets. The adjacent limbic and paralimbic neurons (many of which share the same connectivity patterns) would then face at least two additional plasticity demands: 1) more reactive synaptogenesis at their projection targets in order to replace the synapses originally provided by the degenerated axons of adjacent NFT-containing neurons, and 2) more local dendritic remodeling to receive the synapses which can no longer be accommodated by the degenerated dendrites of adjacent NFT-containing neurons. Because of the downstream barriers to plasticity, these attempts at reactive remodeling would be relatively ineffective and would also induce an excessive upstream intensification of plasticity-related neuronal activity, eventually leading to the formation of NFT in these additional neurons. This sequence of events would promote the "horizontal" spread of NFT within the tightly interconnected components of limbic-paralimbic cortices.
     The loss of dendrites and synapses belonging to limbic-paralimbic neurons would eventually increase the plasticity burden of the association cortices with which they are reciprocally interconnected. These association areas would need to accelerate dendritic remodeling to cope with the loss of inputs from limbic-paralimbic neurons and would also need to remodel axonal endings to cope with the loss of synaptic sites at their limbic targets. This would cause a "vertical" expansion of the disease during which the neurofibrillary degeneration (and eventually cellular death) would spread centrifugally from limbic-paralimbic areas to association neocortex.
     In support of this scenario, tangle-bearing hippocampal neurons show more extensive dendritic trees than immediately adjacent tangle-free neurons, suggesting that NFT formation may be accompanied or preceded by increased plasticity (Gertz, 1990). Furthermore, tau mRNA is increased in the hippocampus, but not in the visual cortex or cerebellum, of patients with AD (Barton, 1990). The initial stages of AD are also associated with increased tau in the CSF (Galasko, 1997), suggesting that an upregulation of this protein occurs at a time when the NFT are undergoing a steep increase in density and distribution.
     Animal experiments show that injury- and denervation-induced neuroplasticity can also lead to an upregulation of APP (Banati, 1993; Chauvet, 1997; Wallace, 1993; Beeson, 1994). Fimbria-fornix lesions in adult rats, for example, elicit a marked accumulation of APP immunoreactivity in the denervated areas within the hippocampus (Beeson, 1994). Such reactive APP accumulation may be quite selective. In the case of the fimbria-fornix lesions, it seems to occur predominantly in the CA1 region but not in the dentate gyrus. In the course of events leading to AD, an initial upregulation of APP would be expected to occur at sites of maximal plasticity burden, namely in limbic-paralimbic areas and their axonal projection targets. This APP would then be processed into the sAPP and Aß moieties, giving rise to complex combinations of neurotrophic and neurotoxic effects. The released Aß would first have a soluble form and would diffuse within the extracellular fluid in the form of 10-100 kDa monomers and oligomers (Kuo, 1996). Upon exceeding local concentration thresholds, this Aß would be expected to form fibrils and condense into initially inert diffuse plaques which would eventually mature into neurotoxic structures. The formation of plaques by local condensation after the diffusion of the amyloid from sites of upregulation, the fact that these sites can overlap with NFT-prone limbic-paralimbic neurons or their widespread projection targets, the apparent regional selectivity of plasticity-induced APP accumulation, and the initial inertness of the deposited amyloid may explain why plaques do not mirror the distribution of the NFT, and why they do not necessarily display a spatial and temporal distribution that fits the clinical features of the dementia.
     The premature development of NFT and Aß deposits in the brains of ex-boxers provides further circumstantial support for the contention that a heightened state of neuroplasticity (in this case, injury-induced) can trigger the neuropathological changes of AD (Tokuda, 1991; Geddes, 1996). This is perhaps why head injury and stroke have both been implicated as risk factors for AD (Salib, 1997; Snowdon, 1997). However, such relationships are probably relevant only when the injury is widespread, chronic, and when it occurs on a background of additional factors which erect downstream barriers to neuroplasticity. Otherwise, all neuronal diseases would eventually lead to AD pathology.

ARF: Can you describe the significance of cholinergic neuron pathology in AD brains, specifically as it relates to your neuroplasticity model?

Marsel Mesulam: A severe depletion of cortical cholinergic innervation is one of the most consistent features in the neuropathology of AD (Geula, 1996). The cholinergic innervation of the cerebral cortex arises from the nucleus basalis of Meynert, a limbic structure that maintains an unusually high level of plasticity into late adulthood (Arendt, 1995). Neurons of the nucleus basalis are consequently among the very first cells of the brain to display an accumulation initially of hyperphosphorylated tau and then of NFT (Mesulam, 1996). Numerous experiments have shown that cholinergic neurotransmission plays an essential role in supporting reactive and experience-induced synaptic reorganization in the cerebral cortex (Baskerville, 1997; Kilgard, 1998; Zhu, 1998). Furthermore, cortical cholinergic innervation also promotes the alpha-secretase pathway and therefore the release of the neurotrophic sAPP moieties (Nitsch, 1992). The early loss of cholinergic innervation in AD could thus contribute to the acceleration of the pathological process by further jeopardizing the potential for neuroplasticity in the cerebral cortex, both directly and through changes in APP metabolism .

ARF: What's the relative importance of genetics versus environmental factors, such as increasing age?

Marsel Mesulam: The biological capacity for plasticity decreases with age, explaining why age is the single most important and universal risk factor for AD. According to this formulation, the AD of old age may not be a disease at all but the inevitable manifestation of a failure to keep up with the increasingly more burdensome work of plasticity. Other factors such as trisomy 21, the e4 allele of apoE, estrogen deficiency, head trauma and the AD-related mutations of APP, PS1 and PS2 accelerate the time course of the events leading to AD by increasing the burden of neuroplasticity. The fact that all of these factors operate through a common downstream mechanism helps to explain how the numerous genotypes of AD cause an identical clinical and neuropathological disease phenotype. Genetic mutations do not really cause AD, they simply accelerate the temporal course of events that lead to plasticity failure and therefore lower the age at which the pathological process begins to gather momentum. The advanced cognitive and mnemonic activities of the human brain impose a very high plasticity burden. The combination of this property with a long life span may endow the human brain with its unique susceptibility to AD.

ARF: What therapeutic targets do you see as the most promising?

Marsel Mesulam: The first wave of AD therapy has aimed to reverse the depletion of cortical cholinergic neurotransmission. Subsequent strategies may aim to inhibit tau polymerization and Aß formation. For reasons that have been described above, such interventions may not be entirely successful unless the underlying plasticity failure is also addressed. The proposed role of plasticity suggests that important insights related to the pathophysiology and prevention of AD may come from the fields of developmental biology. One of the most important goals will be to understand the processes that influence plasticity in the adult human brain and to determine whether their vulnerability to aging and to the other AD-causing factors can be modified.

ARF: You have mentioned the term "plasticity" as the unifying feature in AD pathogenesis. Would you summarize what is meant by the term and how it can cause AD?

Marsel Mesulam: "Neuroplasticity" is a generic term referring to processes of vital importance for the structural upkeep of the brain and for the functional adaptation of the organism to the environment. My hypothesis proposes that the remote initiators of AD may be traced to factors that erect physiological barriers to the downstream manifestations of neuroplasticity, and that the NFT and AP represent independent by-products of initially compensatory but eventually excessive and maladaptive plasticity-related cellular activity. The exceedingly complex events associated with neuroplasticity are influenced by multiple genetic and environmental factors. Consequently, a large number of variables can determine the ease or difficulty with which the neuroplasticity demands are met and also the time at which perturbations of neuroplasticity reach the critical threshold and duration needed to trigger the events that lead to AD.

ARF: How have you seen the field of AD change over the past 30 years?

Marsel Mesulam: From a torpid and esoteric field at the fringes of neuroscience, to one which is vibrant with excitement and which holds the key to fundamental questions related to aging and cognition.

ARF: Finally, what advice do you have for young investigators or graduate students in the field of AD research?

Marsel Mesulam: The best is yet to come.

Additional Reading
Mesulam M-M. 1999. Neuroplasticity failure in Alzheimer's disease: Bridging the gap between plaques and tangles. Neuron, 24:521-529. Abstract.

Marsel Mesulam, M.D., is director of The Cognitive Neurology and Alzheimer's Disease Center at Northwestern University Medical School, Chicago, IL.