Alzheimer’s disease (AD), the leading cause of dementia in the elderly, is a progressive neurodegenerative disorder identified clinically by a characteristic decline in cognitive function. Symptoms of AD dementia usually begin with impairments in episodic memory whereby patients have difficulty forming new memories of events. As the disease progresses, deficits in semantic (facts and general knowledge) and remote memory (e.g., childhood memories) become apparent, as do deficits in visuo-spatial ability, and attentional and executive function (1– 4). Patients can also display other noncognitive behavioral symptoms, such as anxiety, aggression, depression, agitation, hallucinations, circadian rhythm disturbances, wandering, and personality disturbances (5–10). These cognitive and behavioral changes are accompanied by characteristic neuropathological changes – both familial and sporadic forms of AD display the same hallmarks including accumulation of extracellular amyloid plaques and intracellular neurofibrillary tangles (NFTs), synaptic and neuronal loss, and inflammation (characterized by astrocytic and microglial activation).

An important early discovery in understanding the pathogenesis of AD was that the primary component of the amyloid deposits found in AD brain was an approximately 40-residue long peptide, now known as amyloid-β peptide (Aβ) (11–15). Subsequent studies established that Aβ is the product of proteolytic processing of amyloid precursor protein (APP) following sequential cleavage by beta- and gamma-secretases (16–18). This processing of APP to create Aβ is heterogeneous, with gamma-secretase cleavage generating Aβ peptides ranging in length from 38–43 amino acids, although under normal conditions of APP processing, the two major forms of Aβ that have been observed are Aβ40 and Aβ42. Numerous AD-linked mutations in APP and presenilins 1 and 2 (19–22) alter APP metabolism to increase Aβ production and deposition, resulting in accumulation of Aβ42, the Aβ species essential for the formation of parenchymal and vascular amyloid deposits (23), and proposed to initiate the cascade leading to AD (21).

The Aβ hypothesis of AD posits that regardless of whether the disease is familial or sporadic, cerebral accumulation of Aβ to form amyloid plaques is the primary driver of AD pathogenesis (21, 24), and additional disease processes (NFT formation and inflammation) result from the imbalance between Aβ production and clearance. In light of recent studies, the hypothesis has been modified – it is now thought that synaptic failure can be attributed to the presence of soluble Aβ oligomers (25–27), so recent modifications of the amyloid hypothesis suggest that rather than the amyloid plaques being solely and primarily responsible for AD dementia, soluble and diffusible products of APP processing also play a critical role in disease pathogenesis (28–31).

Importantly, the identification of the major proteins like Aβ involved in AD pathogenesis has allowed creation of rodent models based on elevation of Aβ levels in brain. Initially, this was achieved by direct intracerebral or intraventricular injection of Aβ into rodent brain. Elucidation of gene mutations involved in familial AD allowed development of AD transgenic mice engineered to overexpress these gene mutations. More recently, the use of viral vector gene transfer technology has allowed the development of “somatic transgenic” models, whereby genes putatively involved in AD pathogenesis can be selectively overexpressed in specific brain regions involved in AD. A complete ­animal model of AD would need to replicate all the cognitive, behavioral, and neuropathological features of the disease, and account for the contribution of risk factors such as aging. Not one model recapitulates all these features of AD – at best, current models are partial only, allowing us to model specific components of AD pathology. The scope of this chapter is to summarize and provide the reader with methodologies for setting up nontransgenic Aβ rodent models of AD based on infusion of Aβ – either direct infusion of Aβ into rodent brain or use of viral vectors to overexpress APP or selected Aβ species.

The Aβ-related cognitive deficits and some neuropathological features of AD can be induced by intracerebral or intraventricular infusion of Aβ peptides into the rodent brain. Infusion models were originally developed under the assumption that brain Aβ deposition is required to induce behavioral impairments, however, studies demonstrating the disruptive action of soluble ­species of Aβ on long-term potentiation (LTP) and behavior (27, 32–34) suggest that acute Aβ infusion is sufficient to induce some ­learning and memory impairments. Overall, infusion of Aβ peptides can lead to both spatial and nonspatial learning and memory deficits (35–39) and behavioral alterations such as decreased activity and exploration (35). However, the severity of the deficits observed varies markedly depending on species of Aβ used, and the time interval between peptide administration and behavioral testing. Another layer of complexity to be considered when interpreting the data is that cognitive deficits can be measured at different stages of the learning and memory process – acquisition, consolidation, or retention. Therefore, Aβ peptides may impact on different aspects of learning and memory depending on the time interval between peptide administration and testing.

Aβ40 and Aβ42 are the two most common lengths of Aβ found in both normal human brain and in the cortical and vascular deposits of AD patients. Hence, many studies have been conducted using synthetic forms of these peptides infused into the ventricles or hippocampus of both rats and mice. Comprising over 90% of the total Aβ load, Aβ40 has been the focus of many in vivo studies, with the results overall supporting a role for Aβ40 in development of cognitive deficits, although the effects observed are dependent on the Aβ40 administration regime. For example, Aβ40 administered immediately after acquisition in an active avoidance testing paradigm resulted in consolidation deficits but no effect on short-term spatial memory using the radial arm maze (40) or operant conditioning (41). Multiple, repeated injections, on the other hand, produced performance decrements several weeks later (41). Chronic, repeated dosing also produced deficits in the Morris water maze and passive avoidance acquisition and retention (38, 42–47). Aβ42, the species essential for plaque deposition, is the second most abundant form of Aβ and is generally considered the most toxic form. Following both acute and chronic infusion into rodent brain, this too has been observed to have deleterious effects on both short- and long-term memory and consolidation processes, depending on the Aβ administration regime. For example, deficits in acquisition and retention of spatial memory, passive avoidance, and spontaneous alternation tasks have been observed following infusion of Aβ42 (37, 48–52).

Other groups have demonstrated that intraventricular injection of other Aβ species, including Aβ25-35 and Aβ1-28, into rats and mice impairs memory (35, 36, 50, 53–55). Although species like Aβ25–35 are easy to synthesize, readily form toxic aggregates, and can induce learning and memory deficits in rodents, it is unclear whether these species play a significant role in human disease (15).

In general, acute injections of Aβ cause deleterious effects on learning and memory consolidation rather than short-term ­memory or retrieval processes. However, effects on both short-term and long-term memory acquisition and retention have been observed when behavioral testing is carried out several days after Aβ administration or following long-term chronic Aβ administration.

In addition to measurable deleterious effects on learning and memory tasks in rodents, exogenous administration of Aβ species can lead to neuropathological changes reminiscent of human AD, although the full complexity of the human pathology is not reproduced and these pathologies are not widespread as in the human condition. Accumulation of Aβ deposits in brain parenchyma have been reported – these deposits are localized to the injection site and can be induced by infusion of aggregated Aβ species (52, 56, 57), by injection of the soluble form of putative fibrillogenic Aβ (58–62) or following infusion of Aβ deposits into rodent brain (34, 63, 64). These deposits can be associated with inflammation whereby activated astrocytes and microglial cells surround and infiltrate the deposit (56, 61, 65–67). Local cell loss, proximal to the injection site only (68–70), has also been observed. Other changes, such as local production of oxidative stress mediators (66) and cytokines (71), alterations in N-methyl d-aspartate (72), and acetylcholine receptors (52) and neuropeptides (39, 73) have also been observed.

Overall, within the published literature there is a wide variation in the reported behavioral and neuropathological effects of Aβ infusion. These inconsistencies in cognitive deficits and neuropathology that have been observed across Aβ infusion models may be due in part to variations in methodologies used by different groups. It is increasingly clear that not only the species of peptide infused (Aβ40, Aβ42, or Aβ25–35), but the aggregation state also has a major impact on its toxicity and actions in vivo. Variations in effects of Aβ are also dependent on the type of Aβ preparation (synthetic vs. recombinant peptide; fresh vs. aged Aβ), the duration of the infusion (acute vs. chronic dosing), the site of infusion (e.g., ventricle, hippocampus, entorhinal cortex), the time interval between Aβ administration and behavioral testing, peptide concentration, and even the solvent used to dilute peptides. These methodological differences need to be considered when interpreting data from Aβ infusion studies and prior to embarking on any in vivo study. The impact of these points on the model are elaborated later.

Although by no means a complete model of AD, and despite variation in results between labs, the intracerebral Aβ infusion model can contribute to AD research on at least three levels. First, data from this model provides support for the hypothesis of the pivotal pathogenic role of Aβ and allows us to gain insight into the mechanisms and effects of Aβ toxicity in vivo. Second, given the plethora of data from infusion models describing the effects of various Aβ species on different parameters of memory (acquisition, consolidation, and retention), this model can be used to study the effect of Aβ on cognition and can be used to test new therapies that potentially alleviate the cognitive symptoms of AD. Third, since infusion of peptide has been shown to induce inflammation and microglial activation, this model could serve not only to provide insights into the role of inflammation in AD, but could be used to test the protective effects of pharmacological modulation of microglial signaling (86).

The Aβ infusion model also offers some advantages over the use of transgenics. Current transgenic models of AD utilize overexpression of mutant human APP and PS1 to increase Aβ production and recapitulate AD cognitive deficits and pathologies (87–90). However, overexpression of APP results not only in increased production of both Aβ40 and Aβ42, but in elevated levels of other APP fragments, which can have neuroprotective (91,92), neurotoxic (93), or signaling functions (94) and influence learning and memory (95–98). Use of direct Aβ infusion allows the dissemination of the contribution of each Aβ peptide to the dementia and pathology associated with AD, allowing researchers to directly test the effect of different Aβ species on the development of symptoms. The infusion model in particular gives the researcher the ability to administer defined amounts of Aβ of known sequence and length or to introduce controlled co-factors related to plaque development. An additional advantage is the ability to accelerate the experimental timeline rather than waiting for transgenic rodents to age (34, 99). It can take 6–12 months to observe deposition of plaques in transgenic mice (89, 100, 101). Following Aβ infusion, plaque deposition, although local only, can be observed in a timeline of weeks (58, 64).

There are, however, limitations to the use of the Aβ infusion model. By necessity, and in contrast to the use of transgenics, the infusion approach involves inevitable injury to the brain associated with the invasive procedure (injection into brain parenchyma or implantation of a cannula) required to deliver Aβ into the brain. Additionally, the vehicle in which Aβ is dissolved (quite often a nonphysiological solvent) can affect Aβ neurotoxicity. Both these factors could contribute to the inflammatory processes induced by the infusion procedure. To a certain extent, however, these limitations can be overcome by adjusting the infusion rate, volume of injection, and time delay between infusion and examination of the animal to minimize these confounding effects (34). It also needs to be ensured that controls, such as infusion of vehicle only or a scrambled peptide sequence, are run concurrently. Additionally, use of the Aβ infusion model does not take into account the contribution of factors other than Aβ known to be critical in AD pathogenesis, such as neurofibrillary tangles and perturbations in vasculature, and largely bypasses the effect of aging on AD progression.

Viral vector-mediated gene delivery has opened up new possibilities for the development of animal models of neurodegenerative disorders. Localized in vivo gene transfer of putative pathogenic genes has emerged as a viable alternative strategy to transgenic technology for the generation of genetic models of neurological disease. One vector system particularly useful for this application is adeno-associated virus (AAV), recombinant vectors derived from nonpathogenic, replication-deficient members of the Parvovirus family. These vectors are particularly efficient at transducing neurons and following infusion of AAV into the brain stable, long-lasting neuronal transgene expression is observed, with no apparent toxicity (102–105). AAV vectors are well established as versatile tools, allowing upregulation or knockdown of gene expression in specific brain regions, and can be used not only as vehicles for neurological gene therapy (106–109) but for in vivo functional genomics studies (110–112) and to create animal models of neurodegenerative disease.

The use of viral vectors to overexpress genes implicated in disease pathogenesis has already been successfully implemented to generate nontransgenic rat models of both Parkinson’s (113, 114) and Huntington’s disease (115–117), with successful transfer of the method to a primate Parkinson’s model (118). Similar strategies have been undertaken to produce rodent models of AD; however, these models have yet to be fully characterized for development of behavioral deficits and associated neuropathological features. The first reported viral vector-mediated models of AD were based on overexpression of a mutant form of APP. In one study, lentiviral-mediated overexpression of APP751_SWE in mouse hippocampus led to an increase in Aβ40 expression at 10 days and 4 weeks postinfusion (119). However, these short-term time points were the only ones examined, and Aβ40 expression levels determined via immunohistochemistry were the only parameter measured. Notably, most of the Aβ-specific signals observed with immunohistochemistry appeared to be intracellular, suggested by the authors to be a processed, presecretory form of Aβ located in the endoplasmic reticulum. The effect of lentiviral-mediated APP overexpression on hippocampal-dependent behavioral tasks was not undertaken.

In a similar study, an AAV2 vector was used to overexpress mutant APP (APP695_SWE) bilaterally in rat hippocampus (120) for prolonged periods of time (up to 15 months). Detectable expression of both APP and Aβ42 was seen at 6 and 12 months postinfusion. Significant behavioral deficits (in both acquisition and retention of the Morris water maze task) were observed at 12 months post-AAV. Notably at this time point, there was no evidence of plaque formation, neuronal loss, or Fluorojade labeling (to indicate presence of dying neurons). There was no measurable decrease in synaptophysin immunostaining, indicating no detectable loss of synaptic arborization following long-term APP overexpression. All these observations confirm data from AD transgenic mice that development of cognitive deficits is not necessarily accompanied by overt neuropathological changes. Additionally, gliosis, indicative of inflammation, was not observed. Notably, in models utilizing mutant APP overexpression, production of multiple fragments of APP processing would occur – any cognitive deficits or pathology observed cannot be attributed to a specific Aβ fragment.

In contrast, studies undertaken in our laboratory have used AAV-mediated gene transfer to create a rat model of AD based on overexpression and secretion of specific Aβ peptides (121). AAV1 vectors encoding BRI-Aβ cDNAs, fusions between human Aβ peptides, and the BRI protein involved in amyloid deposition in British and Danish familial dementia (122, 123) were used to achieve high-level hippocampal expression and secretion of the specific encoded Aβ peptide, either Aβ40 or Aβ42, in the absence of APP overexpression (23, 124). Notably, use of these BRI-Aβ fusions results in enhanced Aβ secretion, distinguishing this approach from overexpression of Aβ minigenes, a strategy that generates high levels of intracellular Aβ but minimal secreted Aβ (125). Animals were infused bilaterally in the hippocampus with the AAV1 vectors and tested for development of behavioral deficits 3 months after AAV infusion. At this time point, a range of behavioral impairments were found, with learning and memory deficits in the Morris water maze, passive avoidance, and novel object recognition tasks. At the conclusion of behavioral testing, brain tissue was removed and analyzed for Aβ levels and evidence of extracellular Aβ deposition. Only AAV-BRI-Aβ42 animals developed extracellular Aβ deposits, located within the hippocampus. These diffuse deposits were immunopositive for Aβ, but did not stain with thioflavin S or Congo Red and were not associated with astrogliosis or proliferation of microglia, indicating that they are “noncored” diffuse structures similar to the diffuse Aβ deposits observed in humans that are primarily composed of Aβ42 and not associated with significant reactive pathology. Expression for 9 months enhanced Aβ42 accumulation but still did not result in formation of cored plaques. This data showed that viral delivery of BRI-Aβ42 can foster considerable Aβ accumulation in a relatively short timeframe, and confirms both transgenic Aβ Drosophila (126) and transgenic BRI-Aβ mouse studies (23) where visible Aβ deposits were obtained only with Aβ42, but not Aβ40, overexpression. Overall, the results demonstrated that while virally mediated overexpression of Aβ42 alone is sufficient to initiate plaque deposition, both Aβ40 and Aβ42 levels contribute to the development of cognitive deficits. The lack of correlation between the presence of Aβ plaques and the severity of cognitive dysfunction observed confirmed observations from AD transgenic mice that measurable behavioral deficits are not dependent on the presence of Aβ plaques (89, 100, 101, 127) and may correlate better with other Aβ assemblies such as oligomers (128–131).

Viral vector-mediated gene transfer of APP or Aβ is a tool that can be used to model two important aspects of AD pathology in vivo. First, the model can be used to study the effect of Aβ on various parameters of memory formation and can be used to test new therapies that potentially alleviate the cognitive symptoms of AD. Second, this model could be used to study the effects of ­particular Aβ species on deposition of amyloid and testing of therapeutics aimed at clearing Aβ. An advantage over the Aβ infusion model is that depending on the Aβ construct encoded, the AAV model can be used to examine the effects of either secreted Aβ (i.e., with use of the BRI fusions) or the effect of intracellular accumulation of Aβ. An additional advantage over Aβ infusion is that the effect of chronic Aβ exposure can be examined without the need for a long-term cannula implanted into the brain, a procedure with inherent technical problems such as maintaining cannula patency and sterility over an extended time period. A single infusion of AAV allowed us to examine the effect of Aβ overexpression for up to 9 months.

The use of viral-vector-mediated gene transfer to create genetic models of neurodegenerative disease also offers some advantages over traditional transgenic technology. First and foremost, use of viral-vector technology affords the researcher temporal and spatial control over the onset of transgene expression. This means that the consequence of transgene expression specifically in (aged) adult brain can be examined, without onset during embryonic development as is the case with many transgenics. The experimenter can also choose where in the brain the transgene is expressed – for example, in AD either the hippocampus and/or cortex could be targeted. Potentially shortened timelines are another advantage of using this technology – it is not necessary to wait for animals to age as the onset of pathology is faster than observed in AD transgenic mice. A single researcher skilled in stereotaxic surgery can easily generate 10–12 vector-injected animals per day. Finally, viral vector-mediated delivery is not limited by species – transgenic techniques are so far largely limited to mice whereas use of viral vector technology means it has been possible to generate rat and even primate models of neurodegenerative ­disease (118).

Arguably, the main disadvantage of using this method is the difficulty in reproducibility between labs – this is largely due to variability between viral vector stocks. Also, inevitable damage is caused by infusion of the AAV vector into brain parenchyma, similar to what is observed in Aβ infusion models; however, this is not an issue with the use of transgenics. Although viral vector technology can be used to model cognitive deficits reminiscent of AD, it must be borne in mind that the encoded Aβ is present continuously throughout testing (i.e., during acquisition, consolidation, and retention testing) meaning that this model cannot be used to distinguish between effects of different Aβ species on these different parameters of memory formation.