Cognitive Impairment, Neuroimaging, and Alzheimer Neuropathology in Mouse Models of Down Syndrome

Abstract

Down syndrome (DS) is the most common non-lethal genetic condition that affects approximately 1 in 700 births in the United States of America. DS is characterized by complete or segmental chromosome 21 trisomy, which leads to variable intellectual disabilities, progressive memory loss, and accelerated neurodegeneration with age. During the last three decades, people with DS have experienced a doubling of life expectancy due to progress in treatment of medical comorbidities, which has allowed this population to reach the age when they develop early onset Alzheimer’s disease (AD). Individuals with DS develop cognitive and pathological hallmarks of AD in their fourth or fifth decade, and are currently lacking successful prevention or treatment options for dementia. The profound memory deficits associated with DS-related AD (DS-AD) have been associated with degeneration of several neuronal populations, but mechanisms of neurodegeneration are largely unexplored. The most successful animal model for DS is the Ts65Dn mouse, but several new models have also been developed. In the current review, we discuss recent findings and potential treatment options for the management of memory loss and AD neuropathology in DS mouse models. We also review age-related neuropathology, and recent findings from neuroimaging studies. The validation of appropriate DS mouse models that mimic neurodegeneration and memory loss in humans with DS can be valuable in the study of novel preventative and treatment interventions, and may be helpful in pinpointing gene-gene interactions as well as specific gene segments involved in neurodegeneration.

1. INTRODUCTION

Down syndrome (DS) is a developmental genetic condition caused by trisomy of human chromosome 21 (Hsa21). Ninety-nine percent of individuals with DS have an extra copy of Hsa21, and for 90–95% the extra copy is a maternal Hsa21 [1]. DS occurs in 1 in 700 to 1200 live births worldwide, and is estimated to affect approximately 250,000 people in the USA [2]. DS is caused by an extra copy of the long arm of human chromosome 21, which leads to perturbations in expression of genes included in the trisomy segment. Neuropathological and clinical features of Alzheimer’s disease (AD) present in the fourth through sixth decade in individuals with DS [3, 4], and include neurofibrillary tangles (NFTs), amyloid plaques, basal forebrain cholinergic neuron (BFCN) loss, inflammation and dementia resembling AD [5]. People with DS exhibit a higher penetrance of AD pathology in those that survive long enough compared to the disomic population [6, 7]. The co-occurrence of dementia resembling AD in DS (DS-AD) rises steeply with age [8, 9], making DS an excellent natural genetic model for the study of AD-like biological mechanisms and potential biomarkers of both disorders. Moreover, unlike presenilin-linked familial AD, the age of onset of dementia is regulated by the apolipoprotein E (APOE) allele among individuals with DS and can therefore be stratified by the presence of the ε4 allele like typical late onset AD [10]. Despite the molecular and genetic similarities between AD and DS-AD, there is a lack of information on the biological mechanisms underlying the onset of AD pathobiology in people with DS. The development of appropriate mouse models for DS during the last several decades has enabled in depth studies of biological mechanisms, many of which have led to clinical trials of several classes of drugs for DS-AD.

Biological mechanisms involved in DS-AD include extracellular amyloid β protein (Aβ) accumulation, intraneuronal neurofibrillary tangles (NFTs) deposition, BFCN cell loss [11], neuron loss in locus coeruleus (LC) [12], hippocampal abnormalities [13], imbalance of neurotrophic factors [14], alterations in long-term potentiation (LTP) [15], abnormal endosomal signaling [16], presence of neuroinflammation [17], and oxidative stress [18]. Due to its localization on Hsa21, the amyloid precursor protein (APP) gene is triplicated in DS and Aβ deposition is frequently profound in persons with DS. Aβ accumulation in amyloid plaques typically does not begin in the DS brain until after the age of 30 years [19], although diffuse plaques have been observed in individuals with DS in their teens and 20’s [20, 21]. Work from multiple groups including our own has demonstrated increased levels of APP and Aβ, but not in classic amyloid plaque formation, in the Ts65Dn mouse model of DS [22–26]. In addition to the pronounced Aβ deposition in subjects with DS, they also have the presence of intracellular tau-containing NFTs [18, 27]. Previous studies have shown that neuropathology accumulates progressively when individuals with DS are in their 30s and 40s [28]. While a proposed time course for the progression of NFT pathology in the entorhinal cortex is in the mid-30s, hippocampus in the mid-40s, and neocortex in the mid-50s, no consistent appearance of tangle pathology between DS brains has been demonstrated [19, 28–30]. However, there is consensus that NFTs do not appear before the age of 20 in DS [7, 31]. The progressive accumulation of AD pathology suggests that there is a preclinical phase in DS-AD with a delay of at least 10 years between the onset of AD pathology and dementia diagnosis, similar to what is described for AD [32, 33], providing a window for intervention. However, due to the difficulty in obtaining clinically and pathologically well-characterized human DS brain tissue, the need to develop and investigate animal models of this disorder is critical.

The use of mouse models to study DS allows approaches that are not possible in humans. Mouse models are beneficial for: 1) Disease progression studies in a controlled environment, 2) Trials of prevention or intervention strategies, 3) Validation of imaging and surgical procedures, and 4) Studies of gene-gene interactions in the Hsa21 region involved in DS. There are a minimum of 166 protein coding genes in Hsa21, and the regions on Hsa21 are syntenically conserved with three regions located on murine chromosome (Mmu) 10, 16 and 17 [34, 35]. The Ts65Dn mouse, contains a majority of the Hsa21 orthologous protein-coding genes that map to Mmu16, but lacks trisomy for about 45% of Hsa21 ortholog genes [36]. Thus, this model has genetic limitations as a model of age-related AD pathology in DS. Nonetheless, Ts65Dn mice exhibit several deficits seen in individuals with DS, including progressive memory decline [22, 37], adult-onset degeneration of BFCNs and LC-NE neurons [38–40], hippocampal abnormalities [41, 42], and increased APP production [43, 44]. Recent efforts have been made to make a DS mouse model with overexpression of the Hsa21 gene segment in the attempt to more closely mimic human AD pathology such as classical diffuse or neuritic plaques. In the current review we will discuss findings related to aging and AD neuropathology in Ts65Dn mice, as well as comparative studies in other DS mouse models.

2. SUMMARY OF MOUSE MODELS OF DOWN SYNDROME

One of the first models of DS to be developed was the complete trisomy of the murine chromosome 16 (Ts16), which proved to be lethal [45]. Bambrick et al. [46] demonstrated that hippocampal neurons from the Ts16 mouse die at an accelerated rate via caspase-mediated apoptosis and that caspase inhibitors prevent abnormal neuronal loss in Ts16 mice. Because complete Mmu16 trisomy is lethal, this model can only be used for in vitro studies. Other DS mouse models have also been developed to investigate the role of different gene segments of Hsa21 on AD pathology and memory impairment associated with DS-AD. A summary of these different mouse models is provided in Fig. 1. The Ts65Dn mouse model for DS, developed in the early 90’s by Muriel Davisson [47], is the most widely used mouse model for DS, and represents about 120 orthologs of Hsa21 protein encoding genes through a segmental trisomy for Mmu16 (Fig. 1). Mrpl39 is the proximal endpoint for the segmental trisomy in Ts65Dn [48]. This mouse model is also trisomic from a segment of about 10 Mb of the Mmu17, which contains 60 protein encoding genes, none of which are homologous to Hsa21 genes. Most of these have orthologues on Hsa6 [35, 48]. Although the aneuploidy in Ts65Dn is not lethal, there is some evidence that these mice have a shorter lifespan than diploid mice [49]. Another mouse model for DS, Ts1Cje, has shorter Mmu16 trisomy than the Ts65Dn mouse, contains 62 orthologs of Hsa21 genes, and excludes the gene segment containing APP and superoxide dismutase 1 (SOD1, Fig. 1). These mice display certain neurodegenerative features of the processes identified in Ts65Dn mice, including oxidative stress [50], tau hyper-phosphorylation, mitochondrial dysfunction [51], and show some learning and memory deficits [52]. The Ms1Ts65 mice are trisomic for the region of difference between Ts1Cje and Ts65Dn, contains 56 orthologs of Hsa21 genes within genetic segment from Mrp139 to Sod1.

Fig. 1.

Summary of available DS mouse models to date. Hsa = human chromosome, Mmu = murine chromosome, gray lines delineates included gene segments per mouse, and light gray dotted box represents the so called Down syndrome Critical Region (DSCR).

The closely related Ts65Dn, Ts1Cje, and Ms1Ts65 models have been examined by Sago and collaborators in order to assess differences in behavioral phenotypes from their aneuploid segments. Spontaneous locomotor testing showed that Ts65Dn mice were hyperactive, Ts1Cje mice hypoactive, and Ms1Ts65 mice exhibited normal spontaneous activity [53]. The performance in Morris water maze of Ts1Cje mice was similar to that of Ts65Dn mice, but reverse probe tests demonstrated a more severe deficit in Ts65Dn compared to Ts1Cje mice, suggesting consolidation problems unique to the Ts65Dn mice. This is consistent with observations by Lockrow et al., showing that Ts65Dn mice exhibit significant overnight consolidation problems in a water radial arm maze [54]. The memory deficits in Ms1Ts65 mice were much less severe than in Ts65Dn mice. Sago and collaborators [53] concluded that triplication of SOD1 to Mx1 plays a significant role in memory loss observed in Ts65Dn, and that an imbalance of the APP-SOD1 gene segment also contributes to the poor performance. These findings further indicate that genes in the Ms1Ts65 trisomic region interact with others in the Ts1Cje region to produce hyperactivity in Ts65Dn mice. Histomorphological studies of these related DS models have revealed key hallmark changes that may underlie the behavioral phenotypes observed. APP overexpression, seen in Ts65Dn but not in Ts1Cje mice, is necessary to produce cholinergic cell loss [55] and DS-AD neuronal endosomal abnormalities seen at the earliest stage of disease [25, 56, 57] and during infancy long before disease hallmarks are present [57]. Moreover, APP overexpression is similarly linked to cholinergic degeneration associated with NGF transport deficits observed with aging in Ts65Dn mice. These findings suggest that over-expression of APP plays a significant role in AD neurodegeneration and dementia in DS.

Another relatively novel model for DS is the Ts2Cje mouse model Ts(Rb(12.1716))2Cje (Ts2) [58]. Ts2 mice have a similar aneuploid segment to Ts65Dn mice except that a chromosomal rearrangement of the Ts65Dn genome caused a translocation to Mmu12 forming a Robertsonian chromosome [59]. Studies performed in the Nixon and Falangola laboratories indicate that the pattern of brain network abnormalities is similar between the Ts2 and the Ts65Dn mice (see Neuroimaging section below) and further analyses by Jiang et al., (submitted report) show that the sequence of characteristic age-related endosome abnormalities and neurodegenerative changes of basal forebrain cholinergic neurons are similar, confirming the validity of these models for neurobiological studies of DS abnormalities. Finally, Yu and collaborators have developed another novel mouse model for DS that carries complete aneuploidy (spanning the entire Hsa21 syntenic region) [60]. After generating separate aneuploid mouse models: Mmu10− Dp(10)1Yey/+ (Ts1Yey), Mmu17− Dp(16)1Yey/+ (Ts2Yey), and Mmu16− Dp(17)1Yey/+ (Ts3Yey), the triple aneuploid mice (Ts1Yey; Ts2Yey;Ts3Yey) were established by standard breeding [60]. The investigators found that the genotype of Ts3Yey resulted in abnormal hippocampal LTP, but the genotype of Ts2Yey resulted in impaired performance in learning and memory tasks, as well as abnormal hippocampal LTP, suggesting that genes syntenic to Hsa21 on Mmu16 may represent a crucial element for memory loss in DS. Trisomy-based differences in brain morphology were similar in Ts3Yey compared to Ts65Dn mice [60]. These results validate the genetic basis for behavioral and morphological DS mouse model phenotypes, and carry promise for developing more appropriate and complete mouse models for DS in the near future. Moreover, this collection of mouse models allows for assessment of the importance of the Mmu17 and Mmu10 homologous regions in generation of the phenotypes of DS. For example, Zhang et al. [61] have recently shown that the Mmu17 region is a significant determinant of DS related cognitive deficits.

While many mouse models of DS involve aneuploidy of Mmu segments, a notable DS model was created from the trans-species insertion of Hsa21. The Tc1 mouse model for DS carries most of human chromosome 21 in addition to the normal complement of mouse chromosomes, and is trisomic for approximately 212 Hsa21 protein-encoding genes ([36]; Fig. 1). Gardiner and collaborators have shown that Tc1 mice show many of the phenotypes characteristic of DS, including abnormalities in learning and memory and synaptic plasticity [36]. In comparison to Ts65Dn mice, the Tc1 exhibit elevated S100B calcium-binding protein, AMPK, and the mTORC1 proteins RAPTOR and downstream kinase P70S6 [62], which are important regulators of cellular metabolism and aging. Therefore, these mice may represent a novel and yet fairly unexplored model for DS, aging and AD. Due to a genetic rearrangement involving the APP gene, expression of the full length APP transcript cannot be detected in the Tc1 brain, and no human APP protein is found in the Tc1 brain [63]. Therefore, even in the absence of a functional third copy of the APP gene, Tc1 mice still demonstrate a DS phenotype and associated molecular signaling changes that may underlie neuropathology. These data establish the importance of aneuploidy of non-APP genes in DS neuropathology. Although the Tc1 DS mouse model may be an excellent model to examine aging, to our knowledge no aging or intervention studies have been carried out to date. Tc1 mice are mosaic for and have some Hsa21 rearrangements attributed to gamma irradiation of the Hsa21 during production of the Tc1 mice, which complicates their use for some studies but they can, nonetheless, be useful for comparison studies of Hsa/Mmu traits [63]. Interestingly, the production of the Ts65Dn mouse model also involved irradiation of the testes of the male mice used to make these mice [47]. The extent of damage due to Cesium¹³⁷ irradiation of the extra chromosome in Ts65Dn mice has not been tested.

To our knowledge, age-related memory deficits and AD pathology that occur in Ts65Dn mice have not been examined in other DS mouse models. The newly developed DS mouse models may more closely mimic neuropathology occurring in humans with DS-AD, and most are made available as a shared resource to researchers. Attrition studies have not been conducted officially on any of the DS mouse models to date, although we have found that Ts65Dn mice can age quite successfully, until at least 20 months of age, with similar age-related attrition as seen in normosomic littermates (Granholm, personal communication). A summary of studies focused on cognitive impairment in DS mouse models, along with the age of the cohorts studied, is shown in Section 3 (Table 1). In addition to the mouse models listed, there are numerous mouse models transgenic for one or more Hsa21 genes and in which one or more Hsa21 genes have been inactivated by targeted mutagenesis (knockout mice). These mice can be used either separately or in conjunction with the models described above. For example, APP knockout mice have been used to reduce the APP gene in Ts65Dn mice to two copies, thus allowing for the dissection of the effects of trisomy of a single gene. Until recently, this technology was somewhat limited by the availability of appropriate knockout mice. However, with the advent of CRISPR/Cas9 technology, the generation of appropriate mice for these applications is likely to improve markedly [64].

Table 1.

Behavioral alterations in DS mouse models.

Mouse model:	Behavior:	Direction of change	Age	References
Tc1	-Sleep disturbance	Fragmented sleep pattern	adult	[191]
Ts65Dn	-Nest building task	Severely impaired	young	[192]
Ms2Yah	-social novelty interaction -contextual fear-conditioning	Decreased increased associative memory	young	[193]
Ts[Rb(12.1716)]2Cje; Ts2Cje (Ts2)	-nest building behavior -spontaneous alternation	Impaired Impaired	Young adult	[88]
Ts65Dn/DnJ	-Object recognition and consolidation	Impaired	Adult	[194]
Ts65Dn	-Spatial working and reference memory	Impaired (age-dependent)	6–18 month old	[67, 68, 195]
Ts1Cje	-Morris Water maze -Spontaneous locomotion	Less severely affected Hypoactive	Adult	[53]
Ts65Dn	Elevated plus maze	Increased activity	Adult	[196]
Ms1Ts65 (trisomic for APP to SOD1)	-Morris water maze -Spontaneous locomotion	Less severe deficits Normal activity	Adult	[53]
Ts65Dn	-Impulsivity -motor activity	Unaltered	Adult/ aged	[75]
Ts65Dn	-conditioned taste aversion	Impaired	12 months	[158]
Ts65Dn	-Locomotor activity	increased	2–12 months	[137, 197]
Ts65Dn	-WRAM	Severely impaired	13–17 months	[174, 198]
Ts65Dn	Sustained attention task	Impaired (error-induced stereotypy)	15–17 month	[65]
Dp(10,16,17)1Yey	Morris water maze	Impaired	Young	[61]

3. COGNITIVE IMPAIRMENT

Several research groups have demonstrated that Ts65Dn mice exhibit a progressive loss of learning and memory function with age, which starts when they are approximately 6 months old and worsens with age [37, 65, 66]. The Ts65Dn mouse has undergone extensive behavioral testing, and has been shown to exhibit severe spatial and recognition memory deficits [67–69]. Impaired performance has been observed in several different memory tasks including the 8-arm water radial arm maze [70], novel object recognition tasks [52, 71], and the Morris water maze [72–74], see also Table 1. In a meta-analysis of Ts65Dn performance in the WRAM maze at 4, 8, and 10 months of age (Fig. 2A), we found significant genotype effects in the last four trials (F=4.708; p=0.0026), strongly suggesting deficiencies in the learning process in Ts65Dn mice compared to Normosomic (NS) littermates. In post hoc analyses using Bonferroni correction, we also found that 10 month old Ts65Dn mice exhibited the worst performance of both genotypes at any age tested (p=0.0030), suggesting a progressive reduction in performance in this task with age. Stasko and collaborators [73] performed memory and learning studies of Ts65Dn mice, and noted that the behavioral phenotype of Ts65Dn mice was milder than previously described in the literature. These studies were performed on young adult mice, and did not include aged cohorts, demonstrating the importance of aging in DS mouse investigations. Ts65Dn performance on several different behavioral tasks gets worse from 6 to 12 months of age and 4-month old mice are not severely deficient, at least not on spatial memory tasks [23]. Ts65Dn mice exhibit hyperactivity already at a young age [75], and a meta-analysis of Ts65Dn spontaneous activity demonstrated that hyperactivity also increases as the mice age (Fig. 2B).

Fig. 2.

(A) = Analysis of WRAM performance in Ts65Dn mice (gray symbols) and age-matched NS mice (black symbols) from several different experiments across age. Note that the Ts65Dn mice exhibit no statistical difference in average errors the first 4 trials, indicating that they are capable of performing the task, but have significantly more average errors during the last 4 trials, indicating significant reduction in learning. (B) Analysis of total spontaneous activity (in cm) in Ts65Dn mice (gray) and age-matched normosomic mice (NS, black) across several different experiments. Note that there is significant hyperactivity at all ages examined (6–13 months of age). (C) Analysis of locus coeruleus (LC) phenotypic loss from 4 to 13 months of age. TS = Ts65Dn; NS = normosomic. Comparative study across several different cohorts of mice. There is a progressive loss of LC neurons with age, most notable at 13 months of age.

As can be seen in Table 1, few studies [76] have been conducted on the aging process of other DS mouse models. However, it will be exciting to evaluate the aging process and the development of AD pathology and age-related memory loss in these newly developed DS mouse models.

4. NEUROIMAGING OF MOUSE MODELS OF DOWN SYNDROME

Despite the fact that mouse models of DS have been characterized cognitively and histomorphologically, no neuroimaging biomarker evaluation of developmental or neurodegenerative abnormalities has been performed. Only a few in vivo imaging studies have been published in humans [76–84] or mouse models [85–89] of DS. In humans, positron emission tomography (PET) measurements of β-amyloid burden [77, 79, 81], regional PET measurements of the cerebral metabolic rate for glucose [76] and structural and functional connectivity differences by anatomical and functional magnetic resonance imaging (MRI) studies [78, 80, 82, 84] demonstrate the safety, efficacy, and acceptability of these imaging techniques to study DS-AD progression. Using diffusion tensor imaging (DTI), Powell et al. [83] recently reported that changes in white matter integrity, particularly in the frontal tracts, were associated with poorer cognition in adults with DS. This suggested that late myelinating fiber bundles in the frontal cortex are vulnerable to aging in DS. Hence, neuroimaging is an ideal method for analyzing the anatomical and functional phenotypes of DS mouse models, which may exhibit unknown cellular defects that develop in three-dimensions (3D) over time.

Thus far, only a few in vivo neuroimaging studies in mouse models of DS have been published using MRI [85–89]. Chen et al. [85] reported decreased transverse relaxation time (T2) in the cortical and hippocampal cholinergic circuitry in middle-aged (> 12 months old) Ts65Dn mice, demonstrating the utility of quantitative in vivo MRI for identifying AD-relevant cholinergic changes in animal models of DS. Structural MRI revealed ventricular enlargement that may be associated with impaired neurogenesis in the brains of Ts1Cje (Ts1) and Ts2Cje (Ts2) mouse models [86]. In vivo magnetic resonance spectroscopy (MRS), demonstrated reduced glutamate in the hippocampus of Ts2 mice, which was accompanied by reduced mRNA and protein levels of N-methyl-D-aspartate receptors (NMDA-R1) [88]. Recently, in a study using high-field MRS, investigators reported lower levels of glutamine and higher levels of myo-inositol in the hippocampus of Ts65Dn mice, and elevated myo-inositol has also been found via MRS in humans with DS [87]. Lastly, our group recently published the first diffusion MRI (dMRI) results in Ts65Dn mice, using diffusional kurtosis imaging (DKI) [89]. Diffusion MRI is uniquely positioned for exploring brain tissue microstructure, function and connectivity, and to provide evidence of brain plasticity in vivo, allowing longitudinal assessment of ongoing processes. DKI is a specific dMRI technique that quantifies the non-Gaussian behavior of water diffusion [90–92]. Aside from yielding all of the diffusion indices conventionally obtained with diffusion tensor imaging (DTI), DKI also gives metrics of non-Gaussian diffusional, such as mean (MK), axial (K_//) and radial (K_⏊) kurtoses, which more comprehensively characterize brain cytoarchitecture. Our own animal investigations, as well as studies from other groups, have shown that DK metrics are sensitive to changes in brain microstructural complexity that may be associated with brain development [93], Aβ deposition [94], stroke [95, 96], and myelin abnormalities [97].

In our diffusion MRI study in the Ts65Dn mice, DKI metrics allowed detection of early brain tissue abnormalities, with distinct patterns in age-related microstructural trajectories in frontal cortex, striatum and hippocampus of the Ts65Dn mouse model of DS [89]. We showed that Ts65Dn mice have higher DKI metrics in the frontal cortex compared with normosomic (NS) littermates at 2 months and do not change significantly through 8 months of age. We also showed that, in the dorsal hippocampus, Ts65Dn mice have lower DK metrics compared with NS and do not change significantly with age, except for the axial kurtosis (K_//). Although interpretation of changes in diffusion MRI metrics is complex, DKI metrics tend to increase with diffusional heterogeneity and are altered by water exchange or by diffusion barriers. Previous studies suggest that increased kurtosis values in the frontal cortex of 2-month old Ts65Dn mice reflect increased heterogeneity due to abnormal cortical lamination and altered structural dendrite and spine abnormalities in this region [98, 99]. Also, lower DK values in the dorsal hippocampus of Ts65Dn mice may be related to defects in axonal spread and enlargements of dendritic spines [98, 99] causing diffusion dead-space micro-domains, as well as, abnormal myelination of the hippocampal formation, as described in humans [100, 101]. Lower radial kurtosis (K_⏊) values in the striatum may also be related to abnormalities in the myelination process [101].

Since the Ts2 mouse model shows a similar degree of neurogenesis impairment, plasticity alterations and behavioral impairments as described in Ts65Dn mice [59], we investigated if DKI metrics would also detect brain tissue abnormalities. Therefore, we studied 10-month old male Ts2 mice (Ts2; n = 5) and control NS littermates of Ts2 mice (NS; n = 4). The in vivo MRI experiments were performed on a 7T Agilent MR system at Nathan Kline Institute (NKI) using a respiratory gated 6-shot SE-EPI sequence for DKI acquisition (see Table 2 for imaging details). All of the dMRI metrics were estimated from the diffusion and diffusional kurtosis tensors [90], and we studied the same brain regions of interest (ROIs) as in our previous study on the Ts65Dn mice (e.g., frontal cortex (FC), striatum (ST) and hippocampus (Dorsal - HD; Ventral - HV) [89]. Similar to the Ts65Dn results (Table 2), DT metrics showed no group differences in any of the brain ROIs assessed in the Ts2 mice. In contrast, DK metrics showed significant differences compared with NS mice, with decrease in all DK metrics in the ST, HD and HV, which supports the known morphological and plasticity alteration similarities between these two DS mouse models [59]. The changes observed in the hippocampus in both DS mouse models at later ages (8–10 months of age), when cognitive defects are present [52, 69, 74, 102], may be associated with neurodegenerative changes related to neuroinflammation, increased oxidative stress, BFCN degeneration and age-related increased APP transcription and Aβ levels, as previously described [5, 103].

Table 2.

Diffusion metrics estimates (mean ± standard error of the mean (SEM)) in each brain region of interest for the 8-month old Ts65Dn and 10-month old Ts2 mice. The sequence parameters were: TR/TE=4000/29 ms, δ/Δ=5/17 ms, slice thick-ness=1 mm, 14 slices with 0.1 mm gap, data matrix=96×96 zero filled to 128 × 128, image resolution=208×208 μm², 1 average, 30 gradient directions and two b-values for each gradient direction (1 and 2 ms/μm²). A non-diffusion weighted image (b = 0 ms/μm²) was also acquired. DKI post-processing was performed using in-house Diffusional Kurtosis Estimator (DKE) (http://nitrc.org/projects/dke) [199]. Parametric maps of the conventional diffusion tensor (DT) metrics of mean (MD), axial (D_‖) and radial (D_⏊) diffusivities, as well as the additional DK metrics of mean kurtosis (MK), axial kurtosis (K_|), and radial kurtosis (K_⏊) were subsequently computed. One-way analysis of variance (ANOVA) was performed to compare the mean of diffusion metrics between the two groups, with significant p-values in bold; ROIs: frontal cortex (FC), striatum (ST) and hippocampus (dorsal – HD; ventral - HV).

			MD μm2/ms	D// μm2/ms	D⏊ μm2/ms	MK	K//	K⏊
Ts65Dn	FC	NS	0.79 ± 0.06	0.92 ± 0.09	0.72 ± 0.05	0.75 ± 0.05	0.72 ± 0.04	0.75 ± 0.06
		TS	0.78 ± 0.02	0.92 ± 0.03	0.71 ±0.02	0.68 ± 0.04	0.64 ± 0.05	0.66 ± 0.06
	p-values		0.81	0.94	0.64	0.04	0.03	0.06
	ST	NS	0.75 ± 0.02	0.91 ± 0.04	0.67 ± 0.02	0.86 ± 0.06	0.86 ± 0.05	0.97 ± 0.07
		TS	0.74 ± 0.02	0.88 ± 0.03	0.67 ± 0.01	0.80 ± 0.05	0.78 ± 0.06	0.86 ± 0.09
	p-values		0.48	0.25	0.92	0.15	0.03	0.04
	HD	NS	0.79 ± 0.01	0.87 ± 0.01	0.74 ± 0.01	0.82 ± 0.01	0.83 ± 0.01	0.85 ± 0.02
		TS	0.78 ± 0.01	0.87 ± 0.01	0.73 ± 0.00	0.69 ± 0.02	0.71 ± 0.02	0.70 ± 0.03
	p-values		0.35	0.84	0.24	0.001	0.001	0.002
	HV	NS	0.79 ± 0.01	0.88 ± 0.02	0.75 ± 0.01	0.80 ± 0.02	0.77 ± 0.03	0.82 ± 0.03
		TS	0.81 ± 0.02	0.91 ± 0.02	0.76 ± 0.02	0.71 ± 0.04	0.71 ± 0.04	0.71 ± 0.04
	p-values		0.4	0.22	0.52	0.06	0.26	0.07
Ts2	FC	NS	0.85 ± 0.01	1.06 ± 0.02	0.80 ± 0.05	0.54 ± 0.03	0.71 ± 0.05	0.46 ± 0.02
		Ts2	0.87 ± 0.01	1.07 ± 0.02	0.76 ± 0.05	0.48 ± 0.02	0.62 ± 0.04	0.40 ± 0.02
	p-values		0.31	0.57	0.07	0.11	0.22	0.08
	ST	NS	0.82 ± 0.01	1.08 ± 0.03	0.81 ± 0.08	0.76 ± 0.02	0.91 ± 0.02	0.72 ± 0.04
		Ts2	0.84 ± 0.01	1.07 ± 0.02	0.72 ± 0.07	0.65 ± 0.02	0.84 ± 0.02	0.59 ± 0.04
	p-values		0.37	0.66	0.48	0.002	0.04	0.05
	HD	NS	0.89 ± 0.02	1.05 ± 0.04	0.82 ± 0.01	0.65 ± 0.02	0.83 ± 0.03	0.56 ± 0.03
		Ts2	0.91 ± 0.01	1.09 ± 0.04	0.83 ± 0.01	0.56 ± 0.03	0.72 ± 0.03	0.45 ± 0.03
	p-values		0.25	0.55	0.67	0.01	0.02	0.02
	HV	NS	0.86 ± 0.02	1.05 ± 0.04	0.78 ± 0.02	0.63 ± 0.02	0.82 ± 0.02	0.50 ± 0.02
		Ts2	0.90 ± 0.02	1.09 ± 0.04	0.81 ± 0.02	0.52 ± 0.02	0.72 ± 0.02	0.40 ± 0.02
	p-values		0.11	0.4	0.21	0.01	0.01	0.02

Taken together, neuroimaging studies demonstrate that structural MRI, MRS and dMRI can detect changes in the brain of DS mouse models. DKI provides a sensitive indicator of complex neurite architecture abnormalities seen in both Ts65Dn and Ts2 mouse models of DS, which supports the viability of using DKI-based metrics as potential neuroimaging biomarkers of disease processes and for monitoring future neuroprotective interventions in DS. These studies may also assist in the interpretation of results from future translational studies related to human brain microstructure abnormalities in DS.

5. NEURONAL CELL LOSS

BFCNs provide the major cholinergic innervation to frontal cortical regions and the hippocampus, and thereby influence the processing of information necessary for attention and cognition in both animal models and in humans [104–106]. The phenotypic loss of cholinergic neurons is a hallmark of AD, and correlates strongly with memory loss [107]. Individuals with DS exhibit progressive degeneration of BFCNs [11, 108–110]. BFCN neuropathology is mirrored in the Ts65Dn mouse model of DS-AD, with an age-related degeneration, commencing around 6–8 months of age [38–40, 44, 55, 111] with significant atrophy of BFCN cell bodies shown at 6 months of age, and significant phenotypic loss between 8 and 10 months of age (Table 3; Fig. 3A and B). Further, a decrease in immunohistochemistry for both choline acetyltransferase (ChAT) and the high-affinity nerve growth factor (NGF) receptor TrkA has been demonstrated in BFCN neurons at this age. These findings are consistent with the increased deficiency on hippocampal-dependent spatial memory tasks, such as the WRAM, during this age span [39].

Table 3.

Literature summary of BFCN cell loss in Ts65Dn mice compared with age-matched NS mice.

Sex	Age	Marker	Region	Results	References
M	4 mos	ChAT, TrkA	MS, VDB	– cell size	[39]
M	4 mos, 8 mos		MS, VDB	– region area
M	6–10 mos	ChAT	MS, VDB	↓ cell density
M	6 mos, 8 mos, 10 mos	TrkA	MS	↓ cell size
M	8mos, 10 mos	ChAT	MS	↓ cell size
M	6 mos	p75NTR	MS	– cell count	[38]
M	12 mos, 18 mos	p75NTR	MS	↓ cell count ↓ cell size
ns	6 mos	p75NTR	MS	– cell count ↓ cell size	[40]
ns	20 mos	p75NTR	MS	↓ cell count ↓ cell size
M	8–10 mos	TrkA	MS	↓ cell count ↓ cell size	[70]
M	10 mos	ChAT	MS	↓ cell count	[174]
M	10 mos	ChAT	VDB	– cell count
F	18 mos	ChAT	MS	↓ cell count
F	18 mos	ChAT	VDB	– cell count
M,F	10 mos	p75NTR	MS	– cell count	[111]
M,F	19 mos	ChAT, p75NTR	MS	↓ cell count
M,F	19 mos	p75NTR	VDB	– cell count
F	11–14 mos	ChAT	MS	– cell count – cell size	[114]
ns	12 mos	ChAT	MS/VDB/ HDB*	↓ cell count – cell size	[44]
M	19–22 mos	p75NTR, TrkA	MS	↓ cell count ↓ cell size	[55]

NS, normosomic mice born to Ts65Dn dams; ChAT, choline acetyltransferase; F, female; HDB, horizontal limb of diagonal band; M, male; MS, medial septum; ns, not specified; p75^NTR, protein 75 kilodaltons neurotrophic receptor; sc, subcutaneous; TrkA, tyrosine kinase receptor A; Tx, treatment; VDB, vertical limb of diagonal band

ns = sex of mice studied not specified

NS = normosomic

^*Regions combined

Fig 3.

Cell loss in basal forebrain (TrkA staining), hippocampus (Calbindin), and locus coeruleus (tyrosine hydroxylase, TH) in middleaged Ts65Dn mice compared to age-matched normosomic mice (NS). In all these three neuronal populations, there is a progressive loss of phenotype with age, starting at 6 months of age. For BFCNs, this phenotypic loss is significant at 8 months of age, and for the LC and hippocampal neurons at 10–12 months of age. As can be seen in these representative images, phenotypic loss is accompanied by neurodegenerative shrinkage of cell bodies. Scale bar in (A) represents 150 microns, in (D) 60 microns, and in (F) 50 microns.

Evaluation of the age-dependent changes in the activity of ChAT and the acetylcholine-degrading enzyme acetylcholinesterase (AChE) revealed that in 10-month old Ts65Dn mice, ChAT activity was significantly higher compared to NS littermates in the hippocampus and prefrontal cortex, but not in other neocortical regions [111]. At 19 months of age, no differences in ChAT activity were found between the two genotypes, despite significant BFCN degeneration at this age [44]. Thus, alterations of ChAT activity in these forebrain areas seem to recapitulate changes described in patients with mild cognitive impairment (MCI) or mild AD, who exhibit an increase in ChAT activity in cortical regions, presumably as an adaptative reaction to BFCN degeneration [107]. Contestabile et al. [111] reported that neurochemical markers associated with disease progression, including inflammatory markers and APP cleavage products, were altered in some brain regions at the oldest age examined (19 months), suggesting continued progression of AD neuropathology with age in Ts65Dn mice. A summary of cholinergic deficits that occur in Ts65Dn mice and other DS models is provided in Table 3.

Further, female mice had smaller BFCNs in all cholinergic subregions examined compared with genotype-matched males. These findings demonstrate that differences between the genders in BFCNs of young-adult Ts65Dn and NS mice are region- and genotype-specific. The gender differences in BFCN degeneration also have clinical relevance. Studies have shown that women with DS are likely to have early menopause, and that the age of menopause onset correlates significantly with onset of dementia [113]. Our previous findings demonstrated that estrogen treatment prevented age-related BFCN loss and spatial memory loss in female Ts65Dn mice [114], providing evidence that adequate hormone replacement therapy may be a relevant investigative path for older adult women with DS.

Other neuronal populations also degenerate in human and mouse models of DS. Calbindin-D28K (Cal) neurons are known to degenerate in the human brain with age, with greater loss associated with AD [112]. Cal is a neuronal calcium binding protein, which may act as a buffer for neuronal calcium. Cal-positive pyramidal cells do not accumulate NFT in AD, which suggests that loss of Cal expression in specific hippocampal neuronal populations may be associated with reduced resistance to injury and loss [116]. Previous work in our laboratory demonstrated an age-related and significant loss of Cal immune-reactive neurons in the hippocampal CA1 region in Ts65Dn mice (Fig. 3C and D) [22]. These neurons are regulated by neurotrophin expression and play an important role in AD neuropathology observed in the hippocampus [116, 117]. Neurotrophin regulation provides a possible biological mechanism for hippocampal neuronal loss in aged Ts65Dn mice, since these mice exhibit age-related loss of mature NGF and BDNF levels in the hippocampus [14, 43, 44, 68] (also discussed in the Cuello et al. chapter in the current issue [118]).

Finally, a great deal of interest has been generated in the loss of LC-NE neurons and noradrenergic (NE) cortical innervation in both DS-AD and AD [119–123]. NE neurotransmission exerts effects on neurons, glia, and blood vessels throughout the neuro-axis [124, 125]. LC-NE lesions using the selective NE neurotoxin, N-(2-chloroethyl)-N-ethyl-2-bromobenzylamine (DSP-4), give rise to aggravated amyloid accumulation, oxidative stress, and memory loss in transgenic AD models [126, 127], and our previous work has shown that DSP-4 lesions in Ts65Dn mice give rise to accelerated memory loss, inflammation, and neurodegeneration [128]. LC-NE neurons degenerate with age in Ts65Dn mice, to the same extent as BFCN neurons (Fig. 2C, Fig. 3E and F) [128]. Effects of LC-NE degeneration on brain function are mediated directly, via neurotransmission, and indirectly, via aggravation of Aβ accumulation, inflammation, and oxidative stress pathways [129, 130]. NE circuitry exerts a direct influence on BFCNs via α₁-adrenergic receptor activity [131], and pharmacological stimulation of NE receptors leads to improved cognitive performance in rodent models and in humans [132–135]. Salehi et al. [136] demonstrated successful recovery from memory loss in Ts65Dn mice using the NE precursor Droxidopa (L-threo-dihydroxyphenylserine). In support of this study, a meta-analysis of several different groups of Ts65Dn mice confirmed the reported LC degeneration with age in Ts65Dn mice (Fig. 2C, 3E and F), see also [128]. In further studies aimed at rescuing LC function, Fortress and Hamlett, et al. used a Gq-coupled designer receptor exclusively activated by a designer drug (DREADD-hM3Dq) to selectively stimulate aged LC-NE neurons in Ts65Dn mice [137]. The DREADD receptor was administered via AAV into the LC under a synthetic promoter to dopamine-beta-hydroxylase (DBH), PRSx8, to selectively stimulate LC neurons by exogenous administration of the inert DREADD ligand, clozapine-N-oxide (CNO). DREADD stimulation of LC-NE neurons enhanced performance in a novel object recognition task and reduced hyperactivity in Ts65Dn mice, without significant behavioral effects in normosomic age-matched controls. We further confirmed that the NE transmitter system was responsible for the enhanced memory function, by administering the NE pro-drug L-threo-dihydroxyphenylserine (L-DOPS) in separate groups of Ts65Dn and NS mice, which produced similar behavioral results [137].

A recent in vivo study demonstrated that administration of physiological levels of NE prevented Aβ (Aβ_1–42 and Aβ_25–35) toxicity in human NT2N (hNT) neurons and rat primary hippocampal neurons as well as inhibiting Aβ_25–35-mediated increases in intracellular reactive oxygen species, mitochondrial membrane depolarization, and caspase activation in hNT neurons [138]. Taken together, these data suggest that the neuroprotective effects of LC-NE afferent projections result from stimulating neurotrophic NGF and BDNF autocrine or paracrine loops via β-adrenoceptor activation of the cAMP response element binding protein pathway. Thus, NE stimulation may prevent memory loss in Ts65Dn mice and may hold promise for treatment in individuals with DS and dementia as the LC-NE transmitter system is intricately involved in attention, executive function, and several other domains of learning and memory [139, 140]. LC-enhancing medications, such as Atomoxetine or L-DOPS [141] have been utilized for individuals with Attention Deficit Disorder (ADHD), and could be considered in future clinical treatment paradigms for DS-AD patients, since attention deficits and executive dysfunction are common also in DS [142].

6. AMYLOID AND TAU PATHOLOGY

Individuals with DS develop Aβ depositions which are characteristic of early-onset AD in mid-life, presumably due to an extra copy of the chromosome 21- located APP gene [7, 16, 31, 143, 144]. The APP gene is triplicated in the Ts65Dn mouse, and this gene triplication is required for early endosomal abnormalities [25], diminished NGF transport, BFCN loss and cerebellar neuron loss in the Ts65Dn mouse [55]. These pathologies occur despite the absence of amyloid plaques in the Ts65Dn mouse [22, 44], though elevations in both APP, ßCTF and Aβ occur with aging in the hippocampus of this model. APP accumulates beyond the gene dosage in the striatum at 6–8 months of age and in the hippocampus at 13–16 months of age in Ts65Dn mice [43]. Similar elevation of soluble peptide levels (sAPPa and sAPPβ) were also seen at 12 months, but not at 4 months of age [24]. Recent evidence implicates ßCTF, which is elevated in AD and DS brain as well as DS models [145, 146] as the APP metabolite primarily responsible for the development of the endosome defects leading to cholinergic neuronal deficits [55]. In support of the ßCTF pathological role, endosomal transport and cholinergic deficits in fibroblasts from individuals with DS and neuron models require elevated levels of the ßCTF, and not Aß [145, 146]. Most recently, the adaptor protein APPL1 (adaptor protein containing pleckstrin homology domain, phosphotyrosine binding (PTB) domain, and leucine zipper motif) was identified as the critical molecule that directly binds to endosomal ßCTF and activates rab5, the effector of endosome dysfunction leading to defective endosome transport and hallmark cholinergic degenerative changes [146, 147]. Although most investigators have not found extracellular Aβ aggregates in mouse models of DS, Lomoio et al. described small deposits of Aβ in the cerebellum of middle-aged T65Dn mice [26]. These investigators also found evidence for tau protein hyper-phosphorylation in cerebellar neurons, suggesting that Ts65Dn mice are affected by a cerebellar tauopathy. Degeneration of cholinergic and cerebellar neurons in the Ts65Dn mouse has been shown to result in part from the additional gene load of APP [55], clearly linking the APP hypothesis with neurodegeneration that occurs in DS-AD (see Head et al. in the current issue [21]).

There is substantial evidence that peptide monomers and oligomers of Aβ, derived from β-secretase activity, are toxic in the brain, disrupting cell signaling, causing synapse degeneration, and in some cases leading to neuronal death, especially in vulnerable neuronal populations such as the BFCNs [148] although evidence implicating Aß toxicity as the primary factor leading to cholinergic dysfunction in DS are less clear. APP-FL levels are elevated according to gene dose in Ts65Dn mice, while sAPP accumulates with age [24, 43]. The homeostasis between the a - and β-secretase cleavage pathways may be modified by oxidative stress, since anti-oxidant treatments can reduce Aβ formation [149]. We have shown that vitamin E treatment in Ts65Dn mice prevents AD neuropathology (including neuroinflammation, BFCN cell loss, and gliosis) and memory loss, and also augments frontal cortex carboxyl terminal fragment (CTF) peptide cleavage [70]. Specifically, our findings suggested that vitamin E gave rise to increased CTFa peptide relative to APP-FL and CTFβ in Ts65Dn mice with aging [70], suggesting that the anti-oxidant treatment could, at least partially, rectify aberrant APP processing in DS.

The over-expression of APP may have consequences not only for age-related AD neuropathology, but also for brain development in DS. APP affects neural precursor cell proliferation, cell fate, and maturation of neurons, as well as gliogenesis [98, 150]. There are both intracellular and secreted domains of APP that appear to play a role for brain development [151]. This study showed that normalization of APP expression was able to restore neurogenesis and reduce astrocytosis, as well as increase neurite length of trisomic neuronal precursor cells, strongly suggesting that normal expression levels of APP are needed for development of both neuronal and glial elements of the brain. In a recent study by this group [151], the investigators found that deregulation of GSK3β activity was due to higher levels of the intracellular fragment of the trisomic gene APP that directly bound to GSK3β, and by restoring GSK3β phosphorylation either with lithium or the 5-HT receptor agonist fluoxetine, they could rescue neurogenesis and behavior in Ts65Dn mice, and restore GSK3β phosphorylation [151], therefore providing potential novel treatment paradigms for DS-AD. Fluoxetine is also shown to normalize GABA neurotransmission as well as reverse memory loss in adult Ts65Dn mice [152], providing further incentive towards exploring this serotonin agonist clinically in DS-AD, even though some studies indicate adverse effects of fluoxetine in this model [153].

Hyperphosphorylation of tau (p-Tau) is another prominent neuropathological hallmark of AD and DS-AD [5, 27, 154–156]. Tau abnormalities have been observed in the Ts1Cje [51] and the Ts65Dn mouse models of DS [26, 58, 157, 158]. Studies indicate that dual-specificity tyrosine-phosphorylation regulated kinase 1A (Dyrk1A), which is overexpressed in DS and in Ts65Dn mice, may play a role for AD pathology development related to p-Tau. Qian et al. [157] have shown that Dyrk1A regulates tau expression in a dose-dependent manner, and increased tau levels were found in the brains of Ts65Dn mice that overexpress Dyrk1A by stabilizing its mRNA. Further, tau aggregations, which are elevated in Ts65Dn mice, were reduced following hippocampal implantation of neural precursor cells [158]. It is possible that the neural precursor cells increase endogenous growth factors or anti-inflammatory factors to exert these neuroprotective effects.

7. COMPARISONS BETWEEN DS AND AD MOUSE MODELS

The major mouse models employed for the study of AD are humanized, transgenic mice with human gene insertions, which increase the level of human Aβ production, most commonly through transgenes that include mutations in APP or presenilin, or combinations of the two [159]. Many of these transgenic models develop similar pathology: a progressive accumulation of Aβ deposits that begin in the hippocampus and entorhinal cortex. These deposits are surrounded by astrocytic gliosis and dystrophic neurites, and are highly similar to those seen in early AD in humans [160]. Along with plaque deposition, AD mouse models also exhibit elevated oxidative stress and neuroinflammation, but most AD models fail to show significant neuronal loss in any brain region, contrary to DS mouse models [161]. Transgenic AD mouse models expressing high levels of Aβ have memory loss and neuroinflammation/oxidative stress processes that are typical for the early phase of AD (for review, see Bilkei-Gorzo [162]). Neuropharmacological agents that have some clinical beneficial effects in humans with AD are often neuroprotective in AD mouse models, including the Tg2576, APP23, APP/PS1 and 3x-Tg AD models [162]. Of the most used AD models, the triple transgenic La Ferla model appears to have the highest validity in terms of development of AD pathology, neuronal degeneration, and memory loss [162]. The exogenous promoter for APP in these mice may not respond appropriately to inflammation since it is not murine APP, and therefore may not accurately model the “inflammatory hypothesis” in AD, which posits that inflammatory cells drive the production of Aβ through cytokine production. On the other hand, inflammation in the AD mouse models does appear to be driven by Aβ accumulation [163, 164], and anti-inflammatory treatment can reduce amyloid accumulation in AD mouse models [165]. The relationship between inflammation and amyloid accumulation appears to be different in DS compared to AD mouse models. As discussed above, inflammation, as well as memory deficits and oxidative stress are present in the hippocampus of Ts65Dn mice at 4 months of age, a period prior to significant increases in APP or Aβ levels [22, 43], and in the notable absence of Aβ deposits, which rarely develop in trisomy models for DS.

8. NEUROINFLAMMATORY AND OXIDATIVE STRESS PATHWAYS

As in AD, individuals with DS-AD consistently exhibit chronic neuroinflammation, with increases in microglial and astrocytic activation coupled with IL-1β and TNF-α cytokine release [17, 166, 167]. Microglial activation has been demonstrated in several brain regions in AD, including entorhinal cortex, hippocampus and surrounding cortex, and basal forebrain. BFCNs are highly sensitive to inflammation and oxidative stress [131, 168, 169], but specific biological mechanisms for their selective loss in AD and in DS-AD have not been revealed, although TNF-α induced cortical inflammation affecting cholinergic terminals can give rise to retrograde degeneration of BFCNs [170]. Since Ts65Dn mice exhibit significant degeneration of both BFCNs and LC-NE neurons, which both exert a massive innervation into the limbic system, it is not surprising that age-related astrocytosis and microgliosis are accelerated in the hippocampus of this mouse model of DS. Depletion of NE terminals in murine models of AD using the selective NE toxin DSP-4 results in increased inflammatory cytokine production, activated microglial morphology, and accelerated Aβ deposition [125, 126, 169, 171], and an NE lesion has similar results in Ts65Dn mice [128]. The tetracycline derivative minocycline has shown promise as a neuroprotective treatment for neurodegenerative disorders, based on its anti-inflammatory entities [172, 173]. Lockrow and collaborators demonstrated that minocycline treatment inhibited microglial activation, prevented progressive BFCN decline, and markedly improved performance of middle-aged Ts65Dn mice on a working and reference memory task [174], strongly suggesting that anti-inflammatory treatment strategies may succeed in DS-AD to at least partially prevent AD neuropathology and memory loss. Based on these studies, it is difficult to determine whether BFCN and LC-NE degeneration activates inflammatory pathways, or if the cytokine production by astrocytes and microglia, in turn, cause the neuronal degeneration in DS-AD and AD. Most likely, both processes have interactive and escalating effects on each other, leading, in the end, to memory loss and AD pathology, providing hope that anti-inflammatory therapies may be useful for prevention in DS-AD.

Another hallmark now recognized in the progression of AD is oxidative stress. Brain regions affected in AD show elevated oxidative stress markers [175, 176], and neurofibrillary tangles and amyloid plaques, both hallmarks of AD, colocalize with markers of oxidative stress in AD mouse models [161]. As in AD, individuals with DS-AD have elevated levels of DNA damage and lipid peroxidation [18, 177], and a pro-oxidant state is present early in the life, possibly due to the fact that superoxide dismutase (SOD) is located on the trisomic segment of Hsa21, leading to a dysregulation of the reactive oxygen species (ROS) scavenger pathway [178, 179]. Enhanced oxidative stress has also been observed in DS mouse models [50, 180]. DS neurons show increased sensitivity to oxidative stress in vitro [177, 181], and previous work suggests that vitamin E may directly protect neuronal function and improve survival. Studies in trisomy of Mmu16 show that defects in mitochondrial complex I contributed to elevations in reactive oxygen species (ROS) [180]. The potent antioxidant vitamin E was sufficient to protect BFCNs and LC neurons via reduced levels of ROS and neuroinflammation in aged Ts65Dn mice [70]. Similarly, previous studies assessing various antioxidants, including vitamin E, melatonin, and resveratrol have also shown similar benefit to mice transgenic for APP [161, 182]. Vitamin E delivery reduces isoprostane levels in both young and aged Tg2576 mice, yet only successfully reduces amyloid pathology in young mice [182]. These findings indicate that vitamin E appears capable of prevention, but perhaps not rescue from AD-like pathology. In a recent 2-year randomized, double-blind, placebo-controlled trial, investigators explored whether Vitamin E supplementation was safe and tolerable for individuals with DS over the age of 40 [183]. Vitamin E was well tolerated but ineffective as a treatment for dementia in individuals with DS-AD, although Vitamin E supplementation has also been tested in children and teenagers with DS, showing that antioxidant intervention with vitamins E and C attenuated the systemic oxidative damage present in DS children and teenagers [178], therefore holding promise as a neuroprotective therapy if initiated at an earlier age.

9. NEUROPROTECTION TRIALS IN MICE

Prior to the recent approval of memantine, a glutamate NMDA receptor antagonist, the primary treatment option for AD and DS-AD was acetylcholinesterase inhibitors (AChEIs), such as the second generation drugs donepizil, galantamine, and rivastigmine [184]. Each of these drugs has shown minor yet significant improvements to cognition and global status in clinical studies of mild to moderate AD patients. In particular, four small clinical trials using Donepezil, which included individuals with DS-AD suggest beneficial effects, including decreased confusion and cognitive improvement, though these trials were short (often under 6 months) and all involved relatively few patients [185, 186]. Despite studies in animal models that indicate that AChE inhibitors may act to modulate APP processing away from the amyloidogenic pathway [187, 188], more prolonged treatment paradigms using AChE inhibitors in AD patients have not lead to sustainable improvement or marked reductions in progression of the AD disease process. The consensus from meta-analyses of these trials is that AChEI are not capable of either neuroprotective or disease modifying effects [189]. Despite the short list of drugs available for DS-AD and AD patients in the clinic, a multitude of drug therapies have been tested and found relatively successful in DS and AD mouse models. A recent review by Costa et al. [190] summarizes both preclinical and clinical drug trials for the adult DS population. Table 4 shows a brief summary of drug trials in DS mouse models. The anti-oxidant and anti-inflammatory treatments that have been initiated in DS mouse models show promise, along with transmitter-modulating drugs such as the GABA receptor inhibitors and the NE-enhancing drugs (see Table 4), suggesting that the DS mouse models can be used to validate novel pathways towards successful drug therapy in DS-AD.

Table 4.

Summary of pharmaceutical trials in DS mouse models.

Drug	Mechanism	Mouse model	Results	References	Clinical trials Yes/No
Estrogen	Anti-oxidant/protective	Ts65Dn	Protection/rescue	[22, 114]	No
Donepezil	AChE inhibitor	Ts65Dn	No beneficial effects	[200]	FDA approved
Memantine	NMDA reverse agonist	Ts65Dn	Symptomatic	[54, 201, 202]	Ongoing
Vitamin E	Anti-oxidant	Ts65Dn	Preventive	[70]	Yes [178, 183, 203]
Fluoxetine	Serotonin reuptake inhibitor	Ts65Dn	Increased neurogenesis Normalize GABA release	[152, 153, 204]	No formal trial
GABA α5-selective inverse agonists	GABA inhibition	Ts65Dn	Rescue learning and memory	115	Ongoing
GABA(B) receptor antagonist CGP55845	GABA inhibition	T65Dn	Restored spatial memory and novel object recognition	[197]
pentylenetetrazole (PTZ)	GABA(A) antagonist	Ts65Dn	rescued TS performance in the Morris water maze	[200]
L-DOPS	NE precursor	Ts65Dn	Rescue AD pathology and memory	[136, 137]	No
epigallocatechin-gallate (EGCG)	DYRK1A inhibitor	Ts65Dn	Rescues cognitive deficits	[205]	Pilot studies

10. CONCLUSION

This review demonstrates the need for comparative aging studies in several newly developed mouse models of DS, to determine which gene segments on Mmu10, 17 and 16 are involved in the neuropathology and memory loss observed in DS with aging. Development of more appropriate models will allow better pre-clinical studies of drug interventions as well as validation of studies performed in existing models in the past. This review also demonstrates the need for identifying in vivo neuroimaging biomarkers for the different neuropathological processes in DS especially as a function of aging, and mouse studies are crucial for validating neuroprotective interventions. A wealth of previous findings suggest that inflammation, oxidative stress, amyloid and tau toxicity, as well as an imbalance in neurotrophic factor maturation, transport, and cleavage are involved in the pathogenesis and dementia observed in DS-AD.

LIST OF ABBREVIATIONS

AD

Alzheimer’s disease

APP

Amyloid precursor protein

ChAT

Choline acetyltransferase

D_//

Axial diffusivity

D_⏊

Radial diffusivity

DKI

Diffusional kurtosis imaging

dMRI

Diffusion MRI

DS

Down syndrome

DTI

Diffusion tensor imaging

F

Female

HDB

Horizontal limb of diagonal band

Hsa

Human chromosome

K_//

Axial kurtosis

K_⏊

Radial kurtosis

L-DOPS

L-threo-dihydroxyphenylserine

M

Male

MD

Mean diffusivity

MK

Mean kurtosis

Mmu

Murine chromosome

MRI

Magnetic resonance imaging

MRS

Magnetic resonance spectroscopy

MS

Medial septum

Mx1

Myxovirus (Influenza) resistance 1

NS

Normosomic mice born to Ts65Dn dams

p75^NTR

Protein 75 kilodaltons neurotrophic receptor

sc

Subcutaneous

SOD1

Superoxide dismutase 1

TrkA

Tyrosine kinase receptor A

TS

Trisomic

Tx

Treatment

VDB

Vertical limb of diagonal band

WRAM

Water radial arm maze

Biography

Ann-C.E. Granholm