From neurodegenerative phenotype to cause and vice versa

Once clinical entities are established, usually based on their phenotypes, clinicians and researchers try to explain these phenotypes from different points of view, such as pathology, biochemistry, physiology, immunology, etc. This inductive way of explanation has been deemed successful and satisfactory, especially when a single causative agent is identified so that a unified reductionist explanation in a deductive way is possible as has been achieved with infectious diseases. Efforts to explain neurodegenerative disease on the basis of human neuropathology, however, have been hampered because of the intrinsic complexities of the human brain and of disease processes, both of them being responsible for disease-specific temporospatial patterns and cytopathologies.

Recent advance in positional cloning stands out by identifying a genetic abnormality as the single determinant tightly linked to the corresponding phenotype. Once the observed linkage between a genetic abnormality and a disease phenotype is established, it is possible to consider this genetic abnormality as a (single) determinant to explain the entire disease process, even before the pathogenetic cascade is completely clarified. This deductive reversal of understanding is revolutionary in explaining neurodegenerative processes. Indeed, this revolution has been very successful and has facilitated our understanding by providing transgenic animals and cellular models carrying genetic abnormalities linked to human neurodegenerative diseases. On the other hand, an increasing number of researchers are now becoming more and more aware not only of the insight provided by animal and cellular models, but also of unexpected discrepancies between models and human neuropathology. Although we have neither reason nor need to decide which one is superior, deduction (transgenic models) or induction (human neuropathology), what we believe instead is that they are complementary. Indeed, intranuclear inclusions found in mice carrying the Huntington’s disease (HD)-mutant gene [2] led to reappraisal of their human counterpart [15], even in other CAG repeat disorders as pointed out by Yamada et al. [21] in this issue of Acta Neuropathologica. Transgenic (tg) mice carrying Alzheimer's disease (AD)-related mutations demonstrated intraneuronal Aβ more clearly than in human AD brains [3]. Moreover, mouse models for ALS (amyotrophic lateral sclerosis) 2 were generated even before the availability of human autopsy reports [9]. Nevertheless, human neuropathology is pivotal in validating these deductively generated models and their positions relative to the original human disease [20].

The following cluster of review articles summarizes pathological findings of major neurodegenerative categories and the relation to their models to highlight similarities and differences. All authors have profound expertise not only in human neuropathology of the disease, but also in its animal models, so that they are able to relate the models to the “reality.” Particular emphasis was placed on unexplained and unexpected aspects of the models and on their discrepancies with human neuropathology.

Sporadic-hereditary parallelism: neuropathology based on lesion quantity

If one separates human neurodegenerative disorders into sporadic and hereditary categories, some sporadic disorders have their hereditary counterpart as with AD, Parkinson's disease (PD), and some tauopathies. Hallmark lesions of these sporadic conditions, such as senile plaques (SPs), neurofibrillary tangles (NFTs) and Lewy bodies (LBs), more or less appear during physiological aging to the extent considered to be not pathological, although the judgment of being pathological or physiological is purely operational based on statistical (quantitative) criteria under the influences of social and scientific backgrounds [19]. Even if the mechanisms to engender these hallmark lesions remain to be clarified and if the mechanisms were completely independent of social, scientific and whatever backgrounds, it is the quantity of hallmark lesions that is helpful for practical judgment to distinguish pathological states from physiological ones based on the “threshold.” From a genetic point of view, an increase in wild-type (wt) gene dosage is associated with these age-related, quantity-based pathological processes as seen with triplication of genes encoding α-synuclein (αS) [18] or amyloid precursor protein (APP) [16] in men. Remarkably, however, tg approaches to increase the expression of wt APP gene or protein have not been very successful in recapitulating Aβ deposition as summarized by Duyckaerts et al. [3] in this issue. Similarly, simple overexpression of wt tau [4] or wt αS [8] genes in neurons did not recapitulate characteristic hallmark lesions.

Neuropathology based on lesion quality

In contrast to age-related, quantity-based pathologies seen in AD, PD and some tauopathies with both sporadic and hereditary categories, CAG repeat disorders including HD are genetically defined disorders without sporadic counterparts. Indeed, hallmark lesions (intranuclear inclusions or diffuse nuclear staining) of CAG repeat disorders including HD never appear in normal individuals even at advanced age. The qualitative nature of these hallmarks is shared with Bunina bodies and ubiquitin-positive inclusions in ALS, Pick bodies and glial cytoplasmic inclusions of multiple system atrophy [19]. As summarized by Kato [9] in this issue, lower motor neuron degeneration with Lewy body-like hyaline inclusions (LBHI) is induced by superoxide dismutase 1 (SOD1) mutations in human and tg mice. However, this process is not identical to that of sporadic ALS often with Bunina bodies but not with SOD1-positive LBHI. Indeed, genetic abnormalities linked to lower or upper motor system pathologies are highly variable and heterogeneous [10]. Tables provided by Frank et al. [4] and Kahle [8] in this issue demonstrate enhanced pathogenicity of mutations in tau or αS genes to form more aggregated or even fibrillary lesions. If these mutations represent some qualitative pathogenicity, is it then possible to assume that familial AD mutations provide some qualitative differences distinct from sporadic AD? Certainly, some of these mutations have been shown to recapitulate Aβ deposition in mouse brains, which has not been achieved by over-expression of wt APP, and this qualitative change is more or less related to an increase in specific Aβ species leading to accelerated Aβ deposition, as seen in human brains with FAD mutations, but not in their sporadic counterpart [3]. Tg approaches for these principally hereditary disorders by introducing pathological mutations are then always based on this qualitative abnormality. It is largely the pathogenic nature of the mutations that confers the qualitative change specific to these disorders, and combinations of mutations further enhance these pathogenicities. It remains to be clarified, however, whether model animals carrying pathological mutations will provide opportunities to identify mutation-independent factors responsible for pathological processes relevant to the more frequent sporadic counterparts. Duyckaerts et al. [3] discussed this point by showing that identical lesions do not necessarily reflect identical mechanisms.

Aβ and tau

If the human brain is equipped with the machinery that facilitates AD pathologies, it is plausible that the mouse brain is sharing this machinery to some extent. Indeed, both Aβ [3] and tau [4] depositions are inducible in the mouse brain, but parallel development of Aβ and tau deposition as seen in human AD brains has been achieved only by parallel manipulation of the relevant genes [13]. Both Duyckaerts et al. [3] and Frank et al. [4] have summarized a series of tg mouse data, which failed to confirm the so-called “Aβ-cascade hypothesis” that Aβ deposition leads to subsequent formation of NFTs. The missing link between Aβ and tau in mice, however, does not necessarily contradict their possible link in human brain. Indeed, interactions between tau and Aβ [5, 11, 14] are being identified even in the mouse brain.

Since the answer (explanation) is dependent on the question (experimental strategy) and vice versa [1], tg mice provide opportunities to answer specific questions on some aspects of the disease. Validity of the models is then dependent on the way that they are evaluated, as pointed out by Vonsattel [20], or how the models are used to answer the questions. Indeed, Aβ deposition recapitulated by introducing FAD mutations to mice has proved to be useful to test the therapeutic significance of some agents or maneuvers [3]. Among them, successful reduction of Aβ deposits after Aβ vaccination [17] with or without behavioral improvement in mice harboring FAD mutations was undoubtedly a promising forerunner of its therapeutic application to sporadic AD patients [3]. Unexpected complications of meningoencephalitis and expected reduction of Aβ deposits, but not of tau deposits in these patients after Aβ vaccination [12] again demonstrated some disparities between human and mouse brains and between Aβ and tau.

Gene down-regulation and reversibility

It is exciting that downregulation of mutant huntingtin in mice leads to disappearance of degenerative lesions [22], because this in vivo observation possibly suggests the reversible nature of the degenerative process, paving the way for therapeutic trials to be realistic. It is also possible that some intrinsic processes that may counteract the pathological insults are at work [20]. However, spontaneous downregulation of mutant ataxin-7 after early development did not prevent progression of retinopathy in transgenic mice [6]. Similarly, interrupted expression of mutant tau halted neuronal degeneration, while development of NFTs was progressive, suggesting that some of the toxic effect may be irreversible. Probably, different pathological aspects of a disease are not necessarily linked to each other even if both are induced by a single genetic abnormality [14]. Persistence of NFTs in AD patients after clearance of Aβ deposits by Aβ vaccination may be interpreted in this context [12].

From hypothesis to reality

The essentially inductive nature of human neuropathology is related to its descriptive strategy, which usually raises more questions than providing mechanistic explanations. In contrast, the tg approach is rather a kind of strategy expected to provide explanations for the hypotheses obtained elsewhere. As reviewed in this cluster, however, these models not only have provided explanations, but also have raised new questions as well. Judgment of success or limitation of models is highly dependent on the question addressed and the measures to validate the models. Of course, human neuropathology also suffers from inherent limitations. Awareness of these aspects, their similarities and differences will then be necessary for improvement of experimental paradigms and sound interpretations of the findings. More importantly, this awareness may provide further opportunities to raise more relevant and meaningful questions, which are not to be solved by either human neuropathology or tg models alone. All contributing authors will be more than happy if this cluster is of some help for researchers interested in neurodegenerative disorders to understand the current status and to go forward so that the sufferings of patients with these so far intractable disorders will be substantially attenuated in the near future.