Amyloid precursor protein transgenic mouse models and Alzheimer’s disease: Understanding the paradigms, limitations and contributions

Abstract

Transgenic (Tg) mice overexpressing mutant familial Alzheimer’s disease (AD) amyloid precursor protein (APP) genes have contributed to the understanding of dementia pathology and support the amyloid cascade hypothesis. Although many sophisticated mice APP models exist, none recapitulates AD cellular and behavioral pathology. The morphological resemblance to AD amyloidosis is impressive, but fundamental biophysical and biochemical properties of the APP/Aβ produced in Tg mice differ substantially from those of humans. The greater resilience of Tg mice to substantial Aβ burdens suggests the levels and forms that are deleterious to human neurons are not as noxious in these models. Tg mice have been widely used for testing AD therapeutic agents and demonstrated promising results. Unfortunately, clinical trials resulted in unforeseen adverse events or negative therapeutic outcomes. The disparity between success and failure is in part due to differences in brain environment that separate man and rodent. These observations suggest that the pathogenesis of AD is by far much more intricate than the straightforward accumulation of Aβ.

Amyloid-beta (Aβ deposition combined with neurofibrillary tangles (NFT), are important pathological features of Alzheimer’s disease (AD). Accumulating toxic forms of Aβ disrupt neuronal and glial homeostasis and destroy brain and cerebrovascular tissue architecture. A major step toward the exploration of AD pathophysiology and the evaluation of potential treatments was the creation of transgenic (Tg) mice with gross amyloid deposition due to over-expression of human familial Alzheimer’s disease (FAD) amyloid precursor protein (APP) mutations. The reproduction of these pathological alterations in APP and presenilin (PS) Tg mice added a powerful confirmation to the “amyloid cascade hypothesis”, but although sophisticated mouse models have been created, none faithfully recapitulate the extensive cellular, biochemical and behavioral pathology of AD patients. The preeminence of the amyloid cascade hypothesis as a unique pathogenic factor in sporadic AD (SAD) is being re-examined as a consequence. First, Tg mice reproduction of AD pathology is based on the transfection of mutant APP, PS and tau genes, alone or in combination, that reproduce only some of the pathological changes seen in FAD. Second, no specific mutations in the Aβ peptide explain the massive pathological amyloid deposition in the most common sporadic form of AD. Third, both Aβ and tau accumulation are observed in other neurodegenerative disorders. More important, these features are widely seen in aged, non-demented (ND) individuals. Fourth, it is well accepted that SAD is multifactorial and pleiotropic in its pathological and clinical manifestations and that these interactions are strongly influenced by risk factors such as apolipoprotein E (ApoE) genotype, cardiovascular pathology and disturbances in lipid and glucose metabolism. Finally, immuno-responsive AD patients vaccinated against Aβ42 showed substantial clearance of amyloid plaques from the gray matter, without concomitant dementia abrogation [1;2].

While Aβ plays a central pathologic role, Tg mice affirm that AD entails more than alterations in APP/Aβ overproduction, clearance and deposition. Amyloid-β accumulating in the brains of APP Tg mice is neither physically, chemically or functionally equivalent to that characteristic of human AD [3–8]. The APP Tg mice represent a reductionist approach to AD modeling in which massive overexpression of a single gene profoundly alters mouse physiology and behavior. Ultimately, these unfolding, cascading consequences must be explicable in terms of underlying fundamental APP and Aβ biochemistry. A more complete comparison of Aβ biochemistry between mice and humans might identify the critical but subtle differences that form part of the full AD pathology repertoire, including dementia. Equally important is that interspecies differences in Aβ toxicity might explain why recent therapeutic approaches work so consistently and dramatically in the mouse model but not nearly as effectively in AD patients.

The many differences between FAD and SAD, the inherent biochemical complexity of AD and additional complications imposed by the unique species-specific physiological conditions present in Tg mice, have made unraveling the fundamental causes of this dementia a challenging undertaking. Although many APP Tg mice have now been constructed, the scope of this critical review is focused primarily on those models for which detailed biochemical and morphological assessments, and comparisons of Aβ have been performed in our laboratory. In this synthesis, we combine data from the morphological and chemical analyses of several distinct Tg animals and capitalize on the fact that this body of work, as well as the basic AD Aβ chemistry that serves as a frame of reference, was performed in a single laboratory to assess the similarities of the models to human AD.

Rodent Aβ analogs

Transgenic mice over-expressing human FAD APP mutations develop vascular and parenchymal amyloid deposits morphologically resembling those observed in demented humans. However, the extent of this resemblance is limited. Detailed microscopic examination of Congo red and thioflavin-S-stained Tg mouse plaques reveals that although β-pleated sheet structure is present, other aspects of tertiary and quaternary structure differ between AD mouse models. The compact plaques produced in mice generally exhibit a less dense and a more amorphous morphology than those observed in AD patients [9;10]. In general, while the amyloid plaques of AD patients resist disruption by all but the harshest denaturing agents, the analogs produced in Tg animals, such as the Tg2576 [11], APP23 [12], PDAPP [13], TgCRND8 [14] and to a lesser degree in the Tg-SwDI [15], are completely soluble in detergent-containing buffers. In part this is due to Tg mice Aβ lacking the extensive post-translational modifications such as Asp isomerizations, Asp and Ser racemizations, Met oxidations, pyroglutamyl cyclizations, intermolecular cross-links and extensive N-terminal degradation that are characteristic and abundant in SAD senile plaque deposits [3;4;7;11–13;16–19]. Amino-terminal truncations due to extensive proteolytic degradation that produce more hydrophobic Aβ species are prevalent in humans but virtually absent in APP Tg mice [9;11–13]. In addition, approximately 35–40% of total human amyloid plaque mass is an assortment of ancillary glycoproteins and glycolipids that apparently increases their insolubility and resistance to proteolysis [20;21]. These various incongruities may be explained by differences in lifespan and physiology that allow longer periods of amyloid accumulation while facilitating the deposition of a diverse array of different amyloid-associated proteins in humans. Recent transcriptome analyses revealed that the gene expression pattern of aging in mice extensively differs from that of humans [22].

All APP and PS Tg mouse models engineered thus far express human FAD mutations that result in aggressive, early-onset dementia. These rare mutations have yielded important insights into the amyloid cascade, but when inferring fundamental mechanisms from these studies, it is important to bear in mind that FAD biochemical manifestations differ from those observed in SAD [23]. In addition, Tg mice themselves only partially mimic the amyloidosis of FAD. Transgenic PDAPP mice over-expressing the FAD APP V717F mutation are widely employed as an experimental model for SAD and in Aβ immunotherapeutic trials. In humans, this mutation results in characteristic early-onset dementia with numerous flocculent amyloid plaques, relatively few compact core plaques and an enormous production of NFT [24]. The biochemical properties of PDAPP Tg mice amyloid deposits were compared to those of a FAD case with the corresponding APP mutation [13;24]. The Tg mice exhibited prominent flocculent amyloid deposits and comparatively few core plaques with substantial thioflavin-S-positive β-pleated sheet structure. However, while the amyloid deposits produced in the human resisted disruption in buffers containing SDS and EDTA, the analogous structures formed in Tg mice disaggregated in such solutions [13]. These studies confirm that although PDAPP Tg mice recreate select aspects of the Aβ pathology found in humans carrying the corresponding APP mutation, it is an abbreviated recapitulation and does not include other important facets of AD, such as induction of NFT.

The temporal deposition and type of Aβ also varies among the different species of Tg mice. For instance, in the APP23 at 2 months of age, the levels of the soluble Aβ in brain tissue are very high and remain as such during the lifespan of the rodent [25], while in the Tg2576, soluble Aβ levels are very low up to 12–14 months of age and elevate exponentially from this age onward [26]. This an important feature since the APP23 rapidly accumulates large quantities of vascular amyloid. The Tg2576, on the other hand, deposits a moderate amount of vascular Aβ at later stages suggesting a direct relationship between the soluble Aβ fraction and the degree of cerebrovascular amyloidosis [25;26]. Chemical analysis of the APP Tg2576 and the APP23 Tg mice, expressing the Swedish double mutations flanking the N-terminal region of Aβ, indicates that these models contain abundant quantities of Aβ monomers (Aβ:1–37, 1–38, 1–39, 1–40 and 1–42) in their amyloid deposits and no detectable Aβ dimers and oligomers [11;12;25]. In contrast, using the same analytical tools, other Tg mice such as PDAPP [13], Tg-SwDI [15] and the 3xTg (Roher, unpublished observations) generate large quantities of monomeric, dimeric and oligomeric Aβ peptides. These observations suggest that internal Aβ mutations or mutations at the APP C-terminal flanking region of these peptides are necessary for the stable aggregations of Aβ in Tg models. In SAD, on the other hand, no mutations are needed for the stable association of Aβ into oligomeric forms, suggesting that post-translational modifications play an important role in Aβ aggregation and hinder degradation of these peptides [3;5–7;16;21;27–29]. In addition, SAD patients accumulate large quantities of Aβ 17–42 in diffuse deposits [4], which are not commonly observed in Tg mouse models.

Several factors may account for the marked qualitative differences existing between Tg mouse and human AD senile plaques. First, the underlying Aβ accumulation mechanisms are substantially different. While the factor(s) leading to Aβ accumulation in SAD have not been elucidated, in most instances, Tg mice employ brute force transcription overexpression of mutant APP genes. Murine APP processing patterns are clearly different from those of AD [13–15]. In addition, the Aβ peptides produced in Tg animals predictably exhibit proteolytic cleavages that are idiosyncratic for the particular FAD mutation rather than replicating the activities characteristic of SAD patients. The TgCRND8 and Tg-SwDI mice demonstrated an alteration in the order of APP processing by secretase enzymes that resulted in a novel C-terminal (CT) fragment generation and the accumulation of longer than Aβ42 peptides [14;15]. Although the CT99/83 fragments of APP degradation are less abundant in the TgCRND8 and Tg-SwDI mice than in humans, multiple additional Aβ-related peptides were also present. This finding suggests that the initial point of the precursor molecule hydrolysis lies upstream of the sites observed in normal and AD individuals, with subsequent sequential secretase cleavages occurring at the γ- and ε-positions or elsewhere due to uncharacterized murine endopeptidases [15]. Transgenic PDAPP mice also exhibit an altered CT fragment pattern on Western blots [13] with extended length peptides terminating beyond position 42 as well as N-terminally truncated molecules [13], analogous to those observed in the FAD V717F mutation [24]. In contrast, Tg2576 and APP23 Tg mice accumulate severalfold greater levels of APP than those observed in AD patients, but interestingly, are very efficient in generating Aβ40 and Aβ42 in contrast to Tg-SwDI [11;12;15;25;26]. These observations reveal substantial strain-specific heterogeneity in proteolytic processing in APP Tg mice. Another explanation for the differing responses of mice and humans could be that, despite the fact that the rodent genome is 14% smaller than that of humans, it has substantially more protease genes than humans [30;31]. The vast overproduction of APP in most Tg animals undoubtedly perturbs and likely overwhelms the entire range of processing enzymes, with implications for cell signaling and homeostasis. Although much of the excess APP is processed to yield Aβ peptides, the extraordinary pace of production and its scale in most Tg mice may preclude the processing and modification routinely observed in SAD. Along this vein, in the double PS1 and PS2 knock-out mice, enzymatic degradation of membrane-bound substrates is impaired, resulting in the accumulation of membrane-tethered peptide stubs. To compensate, in these mice, there is an increased expression of the calpain-cathepsin enzymes [32].

A growing body of evidence suggests that tertiary and quaternary structures are critical determinants of Aβ pathogenicity [33], with soluble oligomers of mis-folded Aβ peptides perhaps representing the most toxic species [34–36]. Transgenic mice accumulate amyloid deposits that stain avidly with thioflavin-S, revealing the presence of extensive β-pleated sheet structure. However, with the exception of the triple APP mutation-expressing Tg-SwDI mice [15], Aβ oligomers are not produced at levels comparable to those found in AD patients. If Aβ dimers are a key toxic species, then explaining why most APP Tg mice tolerate substantial Aβ burdens is easy; they simply lack the most pathologic species. The Tg-SwDI mice expose a logical conundrum for the simple Aβ dimer abundance model, because they produce and tolerate large amounts of what should be a quite toxic species. One possibility is that the Aβ42 dimers produced in Tg-SwDI mice are firmly bound to the vasculature or within the plaque cores that occur in close proximity to the vessels, suggesting that the toxic species are sequestered in such a way that they are rendered less toxic. Another possibility is that the dimers produced in Tg mice are different in structure and pathologic potential from those produced in humans. Binding of the Aβ-specific “Pittsburgh” compound is less avid in Tg mice, a difference hypothesized to arise as a consequence of the comparative lack of pyroglutamyl modification in these animals [37]. However, all Tg animals are not equivalent. Examination of synthetic Aβ42 peptide carrying the Dutch/Iowa mutation reveals a propensity to produce dimeric structures resistant to IDE degradation. This form was likewise abundant in the Tg-SwDI mice [15]. Finally, it is possible that Aβ dimers are simply not as neurotoxic in mice as they seem to be in humans for unknown physiological reasons.

Primates and mice diverged about 85 million years ago [38] and consequently exhibit vast differences over a wide range of fundamental attributes, including life span, and an age-dependent repression of broad-spectrum neuronal genes, a feature of humans and Rhesus macaques that is not replicated in mice [22], suggesting that the assumption of evolutionarily conserved biochemical equivalence between human and mouse aging is erroneous. Marked differences observed in the response to Aβ injected into aged rodent and primate brains have led to the hypothesis that while the brain becomes more sensitive to Aβ toxicity with aging, a “species barrier” reduces the deleterious effects in rodents [39].

Transgenic Mice and Tau Pathology

Despite exhibiting a massive and often rapid accumulation of Aβ peptides, none of the APP Tg mice examined replicated the full spectrum of AD pathology. Notably, these animals failed to produce dementia-correlated NFT structures. This suggests that subtle aspects of the amyloid cascade disease process are not duplicated in these mice and that the presence of Aβ is necessary, but not sufficient to induce the tau-involved pathological changes typical of SAD and some FAD patients [24]. It may be important to recognize that these limitations mean that with Tg mice we are studying the direct effects of Aβ overproduction and deposition, not AD. Transgenic mouse models may suffer proportionately less deleterious effects than AD patients and therefore respond to amyloid disruption/removal therapies far better in terms of cognitive function recovery than even minimally demented humans who are contending with a broader scope of pathology. Eliminating amyloid deposits by Aβ immunization may represent a cure in Tg mice, but notwithstanding the apparent total disruption of senile plaques in immunized patients with SAD, successful clearance of these structures could neither reverse nor halt the progression of dementia [1].

Triple Tg (3xTg) mice, carrying the APP Swedish mutations K670N/M671L, the presenilin-1 M146V mutation and the tau P301L mutation in the C57BL6 mice background, were developed as an AD model by Oddo et al. [40]. This widely employed Tg model is of paramount importance because it develops intracellular Aβ and extracellular plaque deposits as well as tau pathology similar to that observed in AD. In addition, these mice demonstrate altered long-term potentiation (LTP), synaptic dysfunction and age-related learning deficits and have potential for testing therapeutic interventions against AD [41;42].

However, as is the case for APP Tg mice, it may not be correct to view the 3xTg mice as duplicators of the amyloid cascade as it unfolds in humans [43]. The key APP/Aβ and tau genes are overexpressed simultaneously by transcriptional forcing methods, while tau pathology emerges subsequent to Aβ production in AD. Nevertheless, this model is useful to study the dynamic interactions between human Aβ and tau proteins in mice and further biochemical characterization is needed.

Vascular amyloidosis in Tg mice and AD

Another important difference between AD and Tg mice relates to the distribution and morphology of the Aβ40 and Aβ42 peptides associated with the vasculature. In humans, the deposition of cortical Aβ associated to the capillary network affects the form of globular cores, resembling pussy-willows that are mainly composed of Aβ42. Amyloid associated to the arteriolar and small cortical arteries is in the form of sheets and contains a mixture of Aβ40 and Aβ42. Larger leptomeningeal vessels, on the other hand, mostly contain Aβ40 organized in large concentric sheets replacing the smooth muscle cells [27]. This clear pattern of amyloid deposition is unique to humans as a well as its composition and chemical structure [44]. With the exception of the APP23 [12] and the Tg-SwDI [15] Tg mice, which show some cores of amyloid encroached upon cortical and thalamic microvasculature, respectively, other Tg mice do not entirely replicate the vascular amyloid composition and pattern of AD. The unique distribution of vascular Aβ in AD is related to the drainage of the cerebral interstitial fluid, which occurs along the periarterial spaces that extend from capillaries to the major leptomeningeal cerebral arteries and drain into the lymphatic vessels of the head, eventually reaching the systemic venous circulation [45;46]. It is along the interstitial fluid drainage pathways that Aβ accumulates. The size of this system in humans is by far larger and much more complex than that of rodents. Obstruction of the periarterial spaces by amyloid deposition in the cortical regions results in the accumulation of interstitial fluid and dilation of the periarterial spaces in the white matter and edema [47], with dire pathological consequences for both gray and white matter. Again, APP Tg mice do not recapitulate these important changes in brain hemodynamic physiology and chemistry [48] that are important components of AD dementing process.

The Tg-SwDI mice express very low levels APP and exhibit a unique pattern of amyloid deposition and composition. In addition to diffuse cortical amyloid deposits, thioflavin-S-positive, senile, core-like structures were observed in close association with the thalamic vasculature, but absent in the cerebral cortex [15]. Unlike the situation with the parenchymal tissue plaque analogs of other APP Tg mice, these vascular-associated core plaques were resistant to physical disruption and were more reminiscent of the recalcitrant vascular amyloid deposits observed in other Tg mouse strains. The Tg-SwDI mice also produced large quantities of dimeric, insulin degrading enzyme (IDE)-resistant Aβ, a feature unique to this animal [15]. This finding is significant because the thalamic vascular-associated amyloid pathology produced in the Tg-SwDI strain is far more extensive in total scale and scope than in other APP Tg mice [15]. Similar to APP23 and Tg2576 Tg mice, which produce high levels of Aβ, Tg-SwDI animals tolerate the presence of enormously elevated levels of Aβ.

Alzheimer’s disease and vascular pathology often exhibit an intricate, perhaps co-dependent relationship [47;49]. Whatever role vascular disease and inflammation play in AD promotion and progression, one thing seems certain: APP Tg mice develop a variable degree of amyloid deposits in the brain vasculature. Unlike the case for the senile plaque analogs in Tg mice, the vascular deposits are more resistant to immuno-disruption and thus mimic the situation present in AD [2;50]. Although parenchymal deposits decreased subsequent to anti-Aβ immunization, vascular deposits increased [2;51]. Once the vasculature of Tg mice becomes amyloid burdened, the dynamic flow of Aβ out of the brain tissue may be impaired [25] and this situation may not be completely reversible by vaccination [52;53]. Moreover, there have been reports of microhemorrhages in vaccinated Tg mice [54;55] and signs of vasogenic edema in Aβ-immunized humans (Alzforum, August 11, 2008; http://www.alzforum.org/new/detail.asp?id=1894).

Recent follow-up studies of AN-1792-vaccinated patients suggest that clearance of vascular amyloid deposits may occur in a subset of patients or over long-term treatment [1]. However, it is impossible to judge the degree of vascular amyloid deposition or damage, if any, that existed in any of these patients prior to immunization. Studies of Tg mice as well as humans [2;51;54], have suggested an increased vascular pathology associated with vaccination-induced disruption of parenchymal amyloid plaques. Some experiments have demonstrated that in Tg mice immunotherapy reduced plaque burden despite the presence of substantial vascular pathology [56]. However, ELISA quantifications of total Aβ were not performed although the vascular accumulation of Aβ strongly suggests that disrupted plaque components exit the brain slowly or not at all. The results of the first human immunization trial revealed that even patients with mild AD may have already sustained sufficient vascular damage to prevent disrupted Aβ from egressing the brain [2]. That may mean anti-Aβ immunotherapy will be most effectively employed as a preventive rather than a therapeutic method. This increases the sense of urgency to vaccinate or treat patients before cognitive problems are evident. Concerns over potential deleterious effects on the brain vasculature as a consequence of anti-Aβ immunotherapy have prompted investigators to assess novel antibodies with differing binding specificities and chemistry in Tg mice [54;57]. These studies suggest that specific antibody forms coupled with dose modulation may produce beneficial effects with minimal collateral vascular damage.

Although mutant APP genes are overexpressed within the brain, it is important to remember that in humans, Aβ peptides are both produced and found outside the CNS. Several tissues in the periphery express APP isoforms and harbor substantial levels of Aβ peptides. While any beneficial function(s) for Aβ is uncertain, by the same token these molecules could contribute to as yet undefined processes in the vasculature and other peripheral sites [58]. The implications for AD immunotherapy are obvious, but at the same time remain frustratingly obscure.

Aβ therapies in Tg mice and humans

Immunization of Tg mice against Aβ42 resulted in the apparent reduction of plaque formation and the regression of existing plaque deposits [52], although total Aβ peptide levels remained elevated along with a redistribution of Aβ in the brain [59–61]. Despite the sharp dichotomy between the physicochemical properties of SAD patient and Tg mouse senile plaque amyloid, Aβ vaccination of humans with mild to moderate SAD also resulted in an apparent reduction of plaque deposits [1;2;51;62–64]. This equivalency in the outcome of plaque disruption suggests that APP Tg mice are faithful models of the initial phases of the amyloid cascade in humans. However, studies undertaken in vivo on a real time basis have revealed that amyloid plaques rapidly emerge and disappear in Tg mice [65]. The fact that murine amyloid deposits are substantially more labile than their human counterparts suggests that plaques could be far more dynamic over the entire lifespan of Tg mice as well. These observations bolster the idea that the most effective amyloid disruption therapies must commence with the first clinical evidence of dementia or earlier.

The APP Tg mice models have been widely used for the testing of several AD potential therapeutic agents [52;66–70]. Overall, in the Tg mice, the various curative approaches demonstrated promising results such as improved learning, significant reduction in amyloid plaques, a decrease in soluble and insoluble Aβ burden in brain and a reduction of Aβ in plasma. Unfortunately, clinical trials to date have resulted in unforeseen adverse events or failed to produce clinically beneficial outcomes. The disparity between therapeutic success and failure must in part be due to differences in the complexities of the brain biological environment, metabolism and lifespan that separate humans from rodents. Furthermore, these observations suggest that the pathogenesis of SAD is far more complex and intricate than the straightforward accumulation of Aβ peptides.

If the senile plaques of human AD are intrinsically more neurotoxic than their Tg mouse analogs, it is possible that disrupting such structures by immune or other methods could produce an array of serious consequential changes in patients. In addition, such adverse events, if they occur in Tg mice, could be too subtle to recognize easily. Senile plaques, despite their apparent toxicity to surrounding cells might represent the optimal means to sequester toxic monomeric/oligomeric Aβ [6;71]. Unless all Aβ released from plaques is promptly degraded or removed, deposit disruption may be a treatment that ultimately turns out to be worse than the original disease. Postmortem examination of AN-1792-vaccinated patients, both with and without meningoencephalitis, revealed an unexpected and potentially ominous finding [2]. Although senile plaques were disrupted, there was no proportionate decrease in total Aβ levels within the brain tissue. These findings reveal that disrupted plaque Aβ was unable to exit the brain and suggest that some of this material ended up in the white matter. The toxic and inflammatory properties of Aβ suggest that a permanent elevation of these peptides will ultimately be deleterious to white matter function with profound implications for cognitive function.

Conclusions

Deposition of Aβ may be the penultimate link in a much longer chain of pathological events. Processing of APP is both dynamic and variable, and subtle changes may underlie synaptic functional declines and neuronal death [72;73]. The precise temporal sequence of concerted protease processing determines whether or not the toxic Aβ species are produced at all [72]. In addition to Aβ peptides, APP proteolysis yields a variety of larger products that affect LTP [74] and transcription-mediated signaling processes, including glycogen synthase kinase-3β expression which promotes tau hyperphosphorylation, all of which potentially impact cognitive functions [75;76]. Studies of FAD mutations have revealed that Aβ is of central importance and that even subtle alterations in the relative mix of APP proteolytic peptides yield profound effects on neuropathology, cognition and dementia development that can be manifested even in the absence of a cardinal defining AD lesion, the senile plaque [24]. Recent investigations of several AN-1792-vaccinated patients, who survived for extended periods, revealed that all ultimately died with dementia despite an apparent therapy-induced absence of amyloid deposits [1]. The factors that both influence or are influenced by APP processing remain undefined, but may include fundamental energy deficiency [71;72;77–79] arising as a consequence of increasing vascular deficiencies [47;49;80], mitochondrial failure [81–84] or other stressing mechanisms and co-morbidities associated with aging decay that, in principle, could exert a broad range of synergistic impacts beyond amyloid deposition.

Despite of the anatomical and physiological differences between humans and rodents, Tg mouse models have and will continue to represent promising venues to investigate the amyloid cascade and the efficacy of anti-Aβ immunization and other therapeutic interventions. It is important to recall the underlying complexity and long-lasting effects of AD as both Tg mice and AD therapies are evaluated. Depending on the precise chronicle of the disease process, remediation of late-stage pathology may reverse only a fraction of the damage to the brain. Extrapolating Tg mouse results to humans may be even more complicated because the instigating events of complex aging processes, and hence, the exact pathway to disease, is not necessarily identical in each individual. This implies that just as we should not demand the perfect and comprehensive AD model, we may wish to re-think how we assess therapeutic interventions such as immunization in rodents. It may be better to evaluate the relative success or failure of amyloid disruptive treatments by focusing on that specific target and outcome rather than anticipating and demanding a global AD cure. The pathology underlying AD is complex and its management or ultimate treatment may be equally complicated.