The cerebellum has traditionally been considered to be a brain structure that is primarily involved in the regulation of motor coordination, balance and the motor aspects of speech. This view is changing, however, as recent studies have made it clear that the role of the cerebellum extends to higher cognitive functions including language and executive functioning. The cerebellar hemispheres are substantially enlarged in the human brain and the interaction of their output neurons in the dentate nucleus with higher cortical areas is particularly noteworthy. For a full discussion and list of references the reader is referred to the recent review by Strick et al. [1]. Emphasis on this expanded view of cerebellar function can also be found in the variety of linguistic and behavioral disorders that are found to occur following acquired cerebellar lesions [2]. The common features of these functional deficiencies have been termed "cerebellar cognitive affective syndrome" [3] - a set of symptoms that bears a striking similarity to many of the known symptoms of AD.

Neuronal loss in the cerebellum is part of the normal aging process in human. According to the early study of Hall et al. (1975), the loss of cerebellar Purkinje cells occurs at a rate of 2.5% per decade; the functional consequences of this cerebellar atrophy on neurological functions are unknown [4]. However, the cerebellum is not uniformly affected during the aging process. For example, there is currently no evidence of age-related neuronal loss in the dentate or other cerebellar nuclei [5]. The impact of aging on cerebellum and its multimodal interactions with the rest of the brain led us to question whether the neuropathological changes in this brain region might contribute to the neurological and behavioral symptoms of Alzheimer's disease. The degeneration of Purkinje and granule cells has been reported in both familial and sporadic AD [6, 7]. Studies have shown diffuse amyloid plaques in the cerebellar cortex of AD patients. The majority of these deposits occur in the molecular layer, and while they may extend into the Purkinje cell layer they rarely occur in the granule cell layer [8, 9]. Ubiquitin-immunoreactive dystrophic neurites, increased microglial density and evidence of astrocytosis are also found in the AD cerebellum [6, 10, 11]. In the later stages of AD, a reduction of cerebellar glucose metabolism is observed [12]; and classic cerebellar functions such as balance and gait are also affected [13, 14]. Despite these observations, the cerebellum is still widely regarded as being spared by Alzheimer's disease - often serving as a control tissue or a reference region in imaging studies.

The cerebellum, through its role in the response to pain and emotion, has been linked to behavioral changes such as impulsive behavior. In patients with AD, secondary behavioral manifestations range from hyperactivity to agitation, and impulsive behaviors are common [15]. Motor impulsivity has also been shown in transgenic mice carrying the human APP gene with the Swedish mutation for familial AD (APPsw mice) [16], providing strong evidence of a direct link between mutant APP and this single behavioral AD phenotype both in human and mouse. Impulsive behavior has been traced to abnormalities in the cerebellar deep nuclei, the output nuclei of the cerebellum [17]. While the deep cerebellar nuclei in AD are largely devoid of amyloid deposits and tau inclusions, quantitative MR phase-corrected imaging has revealed increased iron deposition [18]. Iron is often cited for its role in the production of oxidative damage during the progression of AD, raising the significance of this observation. In addition, the loss of cerebellar dentate neurons has been documented in AD with myoclonus [19]. Taken together, these observations suggest that the deep cerebellar nuclei are plausible targets of AD pathology.

Our studies of the biological causes of neuronal death in AD have been guided by our observation of cell cycle reentry in cellular populations that degenerate in human disease [20–22]. Many other laboratories have reported similar findings (references in Herrup and Yang, 2007 [23]). The value of this perspective has been demonstrated in our analyses of mouse models of AD. While neurodegeneration is rarely seen in mouse, in the R1.40 AD mouse model there is a good correspondence between the sites of unscheduled cell cycle events (CCEs) and the areas subject to neurodegeneration in human AD [24–26].

In addition to ectopic cell cycling, AD is also linked to DNA damage. Accumulation of DNA damage in neurons is associated with aging [27] and is exacerbated in neurodegenerative disorders including AD [28, 29]. The correlation of both CCE and DNA damage with the appearance of AD suggests that there may be a link between these two processes. This prompted us to ask whether pathological changes can be found in CDN neurons from AD using early markers of cellular stress. In pursuit of this question, we examined the cerebellar dentate nucleus in 46 AD and control specimens at various disease stages. Our outcome measures included the appearance of DNA damage responses and ectopic cell cycle events revealed by immunohistochemistry and fluorescent in situ hybridization (FISH) analysis of DNA replication. We find that the cerebellar dentate nuclei are indeed involved in the pathogenesis of AD, but that their involvement does not begin until the disease is established. Clinical-pathological correlations reveal that cases with moderate to high levels of cell cycle activity in their CDN are highly likely to show deficits in unorthodox cerebellar functions including speech, language and motor planning. We propose that the symptoms of AD are reasonably ascribed, at least in part, to a loss of cerebellar function.

In this study, we provide evidence that cerebellar dentate neurons are targets of AD pathology. As the neurological and behavioral symptoms and neurobiological aspects of AD are complex, a distributed pathology that includes the cerebellum is not unexpected. We submit that our data argue for the inclusion of the cerebellar dentate nucleus in the list of nervous structures that are affected in AD.

It is intriguing to consider AD symptomatology in light of the non-motor functions of the cerebellum. Decreased verbal fluency is an early finding in AD, and may be related to retrieval slowing. The participation of the cerebellum in speech and language has been recognized for some time (Schmahmann and Sherman, 1998 [3] and references therein). Functional imaging studies show that the lateral cerebellum is engaged when a subject performs a "word generation" task such as determining an appropriate verb to use with a particular noun [35]. Planning and problem solving difficulties may also be related to the degradation of cerebellar function. Whether or not the cerebellum is primarily involved in the actual cognitive dimensions of these processes, decreased cerebellar function places an individual at a disadvantage when it comes to sequencing and planning complex tasks. There is also physiological evidence for anatomical connections between cerebellar and limbic structures [1]. Schmahmann and Sherman designate the totality of these symptoms as the "cerebellar cognitive affective syndrome" [3]. Our findings suggest that the full clinical syndrome of AD includes deficits that are due in part to loss of neuronal function in the deep cerebellar nuclei.

We are not the first to suggest cerebellar involvement in AD [6–11, 36]. However, to our knowledge, there have been no reports that directly link the cerebellar deep nuclei to AD. Our findings also add a new type of evidence to the argument. We have shown that in AD a portion of the cerebellar dentate neurons manifest higher levels of markers associated with DNA damage and cell cycle events. If the appearance of the S-phase cyclin, cyclin A, is used as a marker there is a significant cell cycle involvement of the neurons of the deep cerebellar nuclei in AD cases compared to controls. The levels of cyclin A are significantly correlated with disease state. Paradoxically, however the appearance of elevated PCNA levels (a second S-phase marker of the cell cycle) are not disease correlated. The discrepancy may be explained by the fact that PCNA is involved in both cell cycle progression and DNA damage repair [30].

FISH evidence for high levels of hyperploidy (extra chromosomal material) in both AD and control samples raises several important issues. The absence of correlation between the appearance of cell cycle proteins and evidence of DNA replication might be due to the fact that expression of specific cell cycle proteins can be turned on and off. Polyploidy, by contrast, is irreversible. If the polyploid cells do not die, or if they die slowly, they would remain a substantial fraction of the population, even if regulation of cell cycle protein levels returned to normal. Perhaps more significant is the fact that the high levels of hyperploidy we observed in the non-demented control cases is at odds with our earlier studies in hippocampus and basal nuclei [21]. This discrepancy can be partly explained by the increased sensitivity of the method used in the current study compared to that used by us previously. Other labs have addressed the same problem [37–40]; the common thread in all reports is that while the absolute numbers of hyperploid neurons in normal individuals may be uncertain, this number is significantly increased in affected regions of the Alzheimer's disease brain. We also take note of recent lineage tracing studies showing that the deep cerebellar neurons are derivatives of the rhombic lip rather than the ventricular zone of the fourth ventricle [41]. As the rhombic lip is the source of one of the few known tumors of the CNS that is neuronal in origin (medulloblastoma) it is possible that cell cycle regulation in this lineage is unique among the neurons of the central nervous system.

The expression of DNA damage-related proteins in the deep cerebellar neurons is a robust and consistent marker of disease in the cerebellar dentate nucleus. The appearance of DNA damage during the course of late-onset neurodegenerative disease has been attributed in part to the fact that neurons exhibit high mitochondrial respiration, which is known to lead to the production of reactive-oxygen-species. Over time this oxidative stress results in the accumulated damage of mitochondrial and nuclear DNA [29]. In the current report we show that the chromosome binding protein, 53BP1, and the activated DNA damage response kinase, P-ATM, are both significantly elevated in AD cases with Braak stage of V or above. The occurrence of DNA damage in AD hippocampus has been reported previously using the modified histone, γ-H2AX as marker [42]. In this study, however, the response was restricted to astrocytes and discussed as representing a part of the reactive astrocytosis. In the current study, it is the neurons within the cerebellar dentate nucleus that show the robust DNA damage response. This discordance will be pursued in a future study, especially as we have shown that the Chk2 checkpoint kinase protein, a downstream substrate of ATM, is activated in the astrocyte population of the cerebellar dentate nuclei (in keeping with the findings of Myung et al. [42]).

The cases we were able to study of individuals who died at early stages of their dementia (CDR 0.5), while few in number, are particularly informative as to the onset of deep nuclear abnormalities in the Alzheimer's disease process. Of the 5 individuals we examined, only two had evidence of either DNA damage or cell cycle protein expression in CDN. This can be contrasted with the hippocampus data from the same individuals in this study (Table 3) and our earlier work in the entorhinal cortex and hippocampus [22]. In every case, hippocampal neuronal cell cycle events were found to be as prevalent in individuals with mild cognitive impairment (MCI) as they were in cases that had progressed to full AD. The implication is that the involvement of the cerebellar dentate nucleus occurs later in the progression of the disease. The fact that cell cycle and DNA damage events are found in two different cases of early stage AD also suggests that these two events are temporally sequential or mutually exclusive. Indeed, accumulating evidence has suggested a weak response to DNA damage is associated with the initiation of a quick death, and this correlation may underlie aspects of the selective neuronal vulnerability observed in aging and neurodegenerative disease. For example, the number of Purkinje cells is significantly reduced during aging, but there is no parallel increase in DNA damage. By contrast, granule cell loss is insignificant during the aging process, but there is a significant increase in DNA damage markers in these cells [43, 44]. In this regard it is worth considering that P-ATM and 53BP1 immunoreactive neurons in cerebellar dentate nucleus might persist, while cyclin A positive neurons might be destined to die. This prediction is consistent with the clinical-pathological data that link high cyclin A expression to difficulty in speech/language and motor planning.

The results presented here reveal that the CDN neurons are under cellular stress in AD and suggest that at least some of the non-motor symptoms found in patients with AD may be partly cerebellar in origin. Our findings also emphasize the value of using direct markers of neuronal distress - cell cycle events and DNA damage - as pathological markers in AD. They augment the classical histopathological picture achieved by staining for amyloid plaques and tau inclusions by providing an early neuronal vulnerability marker. As with the appearance of hyperphosphorylated tau, increased neuronal cell cycle events are not disease specific [23]. However, they unify the disease phenotype in human AD and its transgenic mouse models, and provide a perspective on the human disease that adds depth and insight.