Abstract

Introduction

Down syndrome (DS), also known as trisomy 21, is caused by an extra copy of the long arm of human chromosome 21 (Hsa21). While the phenotype of DS is complex and many aspects are variable in penetrance, common to all individuals are both some level of intellectual disability (ID) and the early development of the neuropathology of Alzheimer’s disease (AD).1 The incidence of DS remains at approximately one in 730 live births in the United States and approximately one in 1,000 within Europe.2,3 Given that the average level of cognitive ability is associated with intelligence quotient scores of 40–50, and that a significant percentage is more profoundly impaired,4,5 pharmacotherapies that could lessen or ameliorate the level of impairment would have substantial social consequences.

The major hypothesis that has been and is driving DS research is that the phenotype is a consequence of elevated levels of expression, due to dosage, of trisomic genes.6 Hsa21 genes that are conserved in mouse include ~160 encoding diverse protein functions, a large family (~45 members) encoding apparently redundant keratin associated proteins, and a number of provisionally annotated microRNAs. Hsa21 also encodes several hundred additional genes/gene models of unknown function that lack detectable nucleotide sequence conservation in the mouse genome.7 Many experiments have examined gene expression in DS-related tissues, including fetal brains, amniocytes, neurospheres, and cell lines derived from DS, as well as brain regions and other tissues and cells derived from some partial trisomy mouse models of DS (summarized by Vilardell et al8). Results, most often generated at the RNA level, have largely supported the gene dosage hypothesis; however, not all trisomic genes are elevated in any one experiment, and which genes are elevated and by precisely how much depends on the organism, genetic background, developmental time point and tissue or cell type.

Expression levels of non-Hsa21 genes are also altered in DS.8 This is expected given that among the known Hsa21 protein-coding genes are 20 transcription factors/modulators; ten proteins involved in the processing and/or modification of messenger (m)RNA, transfer (t)RNA, and ribosomal (r) RNA; nine proteins that function directly and indirectly in protein phosphorylation, methylation, and sumoylation; and 16 proteases, protease inhibitors, and proteins that regulate degradation by the ubiquitin pathway.9 Overexpression of genes in each of these classes would be expected to affect levels of expression and/or activity of many non-Hsa21 genes. Because the levels of Hsa21 proteins vary with time and place, the perturbations in the expression and activity of their non-Hsa21 targets also vary with the tissue, developmental time, and model system. Consideration of the number of Hsa21 genes, together with the complexities of their function, regulation, and expression, might suggest that ID in DS is too complex in its genetic basis to be ameliorated by pharmacotherapeutic intervention. However, over the last several years, there have been a number of reports of the successful rescue of learning and memory (L/M) deficits in a mouse model of DS, the Ts65Dn. Drugs and small molecules with diverse targets and mechanisms of action have been tested in a variety of L/M protocols, and for their effects on cellular and electrophysiological features, in mice that range in age from pre- and early postnatal to young and older adults (see Table 1). In this review, we discuss this substantial literature. We consider the challenges of integrating these results, the shortcomings of the Ts65Dn mouse model, and the vexing problem of planning effective human clinical trials based on data generated for the Ts65Dn model.

Modeling DS in mouse and the Ts65Dn

The mouse is the model organism of choice for the preclinical evaluation of drug efficacy. A challenge for modeling of DS is that orthologs of Hsa21 genes are distributed among three mouse chromosomes. Of the ~160 Hsa21 protein-coding genes, ~100 map to the telomeric segment of mouse chromosome 16 (Mmu16), and ~20 and ~40 map to internal segments of Mmu17 and Mmu10, respectively (Figure 1).7 Many partial trisomy models have been created using several different methods, and each is trisomic for a unique subset of Hsa21 genes or their mouse orthologs. These have been reviewed10–12 and will not be discussed here. Instead, this review focuses on a single model, the Ts65Dn, which is the only DS model that has been used in the preclinical evaluation of drugs for L/M.

Open in a separate window

Figure 1

Distribution of Hsa21 genes on mouse chromosomes 16, 17, and 10.

Notes: Giemsa-banded Hsa21 is shown at the left. Dp16 (official name, Dp[16]1Yey), Dp17 (official name, Dp[17]1Yey), and Dp10 (official name, Dp[10]1Yey) are the mouse lines created using chromosomal engineering18,19 carrying, respectively, duplications of the complete segments of mouse chromosomes 16, 17, and 10 that are orthologous to Hsa21. The Ts65Dn is trisomic for the telomere proximal segment of mouse chromosome 16, spanning most, but not all, of the Hsa21 orthologous region (as indicated by the horizontal dashed lines) plus the centromere proximal segment of mouse chromosome 17 (indicated by diagonal shading). Numbers in brackets indicate the number of Hsa21 protein-coding genes conserved in each model. Locations of genes discussed in the text are indicated.

The Ts65Dn was the first viable mouse model of DS and has been available for 20 years.13,14 Because of this history and its long reign as the only viable segmental trisomy, the Ts65Dn remains the most popular model. The Ts65Dn carries a freely segregating marker chromosome composed of the telomere proximal region of Mmu16 translocated to the centromere and pericentromeric region of Mmu17. The Mmu16 segment contains a majority (88 of 102) of the Hsa21 orthologous protein-coding genes that map to Mmu16 – ie, ~55% of the (nonkeratin-associated proteins) Hsa21 protein-coding genes conserved in mouse.7 The Ts65Dn thus lacks trisomy for ~45% of Hsa21 orthologs – the remainder of those mapping to Mmu16 and all of those mapping to the Mmu17 and Mmu10 orthologous regions. The Mmu17 region that is trisomic in the Ts65Dn carries 50 protein-coding genes that are not orthologs of Hsa21 genes15,16 (the Hsa21 syntenic region is ~20 Mb distal on Mmu17). This segment is an artifact of the method used in construction, and the genes are not relevant to DS and possibly confound assessments of gene–phenotype correlations and drug responses in the Ts65Dn (see the “Discussion” section).

The Ts65Dn, although not an ideal model, has been extremely valuable for the study of a number of phenotypic features similar to those seen in people with DS, or speculated to contribute to features seen in people with DS (for reviews, see Rueda et al10 and Roubertoux and Carlier,17 as well as the references in Table 1). Particularly well documented are deficits in L/M tasks that require a functional hippocampus, including the Morris water maze (MWM) and several other maze tasks, object recognition tasks, and fear conditioning.10,17 Electrophysiological abnormalities include deficits in gamma-aminobutyric acid (GABA)ergic and glutamatergic transmission in the hippocampus and cerebellum; among them are age- and cell type-specific deficits in hippocampal long-term potentiation (LTP) and long-term depression (LTD). Also well documented are impairments in adult neurogenesis in the dentate gyrus, abnormalities of neuron number and dendritic spine morphologies, and decreased numbers of cerebellar granule cells. Although the Ts65Dn do not show accumulation of beta-amyloid, as in the neuropathology of AD, age-dependent loss of the functional markers choline acetyltransferase (ChAT), nerve growth factor (NGF), and neutotrophic tyrosine kinase receptor type 1 (TRKA) are observed in the basal forebrain cholinergic neurons (BFCN), and the loss of tyrosine hydroxylase (TH) in the locus coeruelus (LC). These supply cholinergic and noradrenergic input, respectively, to the hippocampus (and other brain regions), and their loss is relevant to premature cognitive decline and development of AD-like dementia that is commonly seen in individuals with DS over the age of 60 years (see Table 1).10,17 Each of these features has been explored by multiple research groups, showing these abnormalities to be robust and reproducible (see Table 1).10,17

Most recently, chromosome engineering has been used to create three lines of mice that are separately trisomic for the complete Mmu16 region (the Dp[16]1Yey, or Dp16), the Mmu17 region (the Dp[17]1Yey, or Dp17), and the Mmu10 region (the Dp[10]1Yey, or Dp10).18,19 A full trisomy mouse can be obtained by sequential crossing of the Dp16, Dp17, and Dp10; however, because of viability issues, time, and expense, this has not been practical to pursue.20 Thus, in spite of its limitations, and because of its long and productive history of use, the Ts65Dn remains the most popular choice among DS mouse models.

Over the last few years, a number of experiments have shown that L/M deficits in the Ts65Dn can be rescued pharmacologically (Table 1). To date, more than 20 drugs/small molecules have successfully rescued hippocampal deficits, most of them in adult mice. Based on these, and sometimes additional information, several clinical trials have been carried out and others are currently in progress. The enthusiasm for such trials is strong: drugs that could improve cognitive function – for example, by increasing the intelligence quotient by 10–20 points – would have enormous positive consequences for individuals with DS. It would increase their potential to participate in society more fully and more independently, with consequent positive impacts on their families and on society as a whole.

In this review, we discuss results from 29 drugs that have been investigated in the Ts65Dn, both successfully and unsuccessfully (Table 1). Subsets of these drugs have been included in recent reviews.10,11,21 Here, we discuss the complete set, why each was chosen, the details of how each was evaluated and, in particular, the similarities and differences in their outcomes. Also discussed is the current status of clinical trials for cognition in DS and the features of the Ts65Dn that may limit the predictive value of its use in preclinical evaluations.

Ts65Dn and preclinical evaluation of drug efficacy

As with other mouse models of human diseases, there are no standards for the assessment of drug efficacy in the Ts65Dn, so the numbers, types, and stringencies of evaluations vary among studies. For the drugs in Table 1, the rescue of L/M in at least one task has been assessed, and in several cases, effects on adult neurogenesis, LTP/LTD, and age-related neurodegeneration in the BFCN and/or LC have been evaluated. To achieve some consistency in the discussion, the results of only these four categories are included here, although many reports included additional types of assessments (see references in Table 1).

The results in Table 1, even from the same types of experiments, were obtained using protocols that varied in their details. The age of the mice evaluated ranged from embryonic and early postnatal to 16 months. While the majority of experiments used male mice, a few used female mice, some used a mix of male and female mice, but the authors did not separate the sexes in their analyses, and some did not report the sex (see Table 1). Some drugs were administered as acute single doses and others were chronic, with treatments extending from a few days/weeks to several months. Drug delivery methods include stressful subcutaneous and intraperitoneal injections, and benign ingestion through food, water, or milk. When different laboratories tested the same drug, concentrations often differed. Some L/M evaluations used more stressful tests, such as water mazes and electric shocks, which may result in exaggerated poor performance caused, at least in part, by differential reaction to stressors, rather than by solely impaired L/M,22 while others used nonaversive, nonwater mazes, and one used olfaction. Lastly, not all drug treatments provided complete rescue of impairment, and some studies did not provide all statistical analyses needed for such determinations. For example, comparing the performance of drug-treated Ts65Dn mice to that of vehicle-treated Ts65Dn mice merely asks if the drug improved L/M; a comparison to untreated control mice is necessary to determine how well L/M was rescued. Together, such variations in experimental protocols and analyses make comparisons among drug responses, and their extension to predictions for DS, less than straightforward.

The drugs in Table 1 are clearly diverse in their targets and mechanisms. Only three (memantine, epigallocatechin-3-gallate [EGCG], and ethosuximide) have a known/predicted direct link to the overexpression of one or more genes encoded by Hsa21. Most drugs, however, have a rational basis for their choice, frequently from studies that show rescue of a Ts65Dn-relevant phenotypic abnormality in a non-DS mouse model (eg, rescue of neurodegeneration in an AD mouse model) (see references in Table 1). The entries in Table 1 are in chronological order to capture the development of thought and information in the field of treatment of ID in DS.

Comparison of drug effects in the Ts65Dn

Of the 43 reports listed in Table 1, four did not include L/M in the evaluation, and eight of the 39 that did, failed to rescue L/M in the assay(s) used. The 31 reports of the successful rescue of L/M involved 23 different drugs targeting a diverse set of 19 proteins, complexes, or cellular processes. In 14 of the 31 successful reports, rescue was less than complete in that 1) performance improved, but did not reach the level of untreated control mice; or 2) performance was rescued in one test of L/M, but not in another in the same study; or 3) results were inconsistent among tests of the same drug carried out by different laboratories. Incomplete rescue does not diminish the significance of the results; rather, it potentially provides additional opportunities for insight.

L/M responses to drug treatments, of course, are first influenced by the targets and mechanisms of action of the drugs employed. However, responses may be modulated – leading to partial rescue – by details of the experimental design, which may include the use of older mice, suboptimal drug doses, and increased difficulty of the L/M task. Stress could also be a factor, depending on the specific L/M task and the method used for drug delivery. Among the drugs showing only partial rescue in specific L/M tasks, minocycline, memantine, and vitamin E were tested in the WRAM, while melatonin, NAP + SAL, and peptide 6 only partially rescued acquisition in the MWM. Water mazes are generally aversive tasks for mice, and the Ts65Dn mice are more sensitive than are controls to the stresses associated with the MWM.22,119 The stresses of drug delivery by daily injections are not implicated in these cases, because each drug was provided in food, water, or by means of implanted timed release subcutaneous capsules. Partial rescue was observed in one instance of another stressful task, CFC, where the drug, formoterol, was delivered by daily injection.66 On the other hand, there are drugs – among them, PTZ, memantine, and EGCG – that completely rescued performance in the MWM, and performance in CFC (with memantine).47 Thus, effects of stressors are not yet clear.

In several cases, a drug rescued L/M in one task but not in another. The effect of task complexity may be evident in the memantine experiments. When memantine was delivered in drinking water (a benign delivery method), learning was completely rescued in NOR, but only partially rescued in the WRAM, a task that places greater demand on working memory than does NOR and other water tasks such as the MWM. These mice were treated with memantine for 6 months, as compared to an acute injection that successfully rescued CFC and NOR.47,49 Therefore, the difficulty of WRAM versus CFC may be a contributing factor in this case. In a different scenario, SAG 1.1 rescued both acquisition and retention on the MWM, but it failed to rescue performance in the Y-maze. Drugs influencing (or not) performance in the spontaneous alternation Y-maze, and the related T-maze, are worth further consideration. Typically, neither is as aversive as the water tasks (if the mice are handled carefully); and in the spontaneous versions most often used with the Ts65Dn, they test only the preference for novelty and the ability to remember prior choices. In addition to SAG 1.1, both lithium and the GABAB receptor antagonist CGP55845 failed to rescue performance in the T-maze, although each rescued performance in other tasks, including NOR, OL and the aversive CFC. Conversely, spontaneous alternation T-maze or Y-maze performances were rescued with fluoxetine, xamoterol, and PTZ, and performance on a related four-arm alternation maze was rescued with physostigmine.58,60,30,64 Taken together, these data are interesting in terms of what they may indicate about drug targets and their distribution within the brain, and regarding what brain regions are involved in each task, as related to cognitive deficits in DS.

Additional limits of the current data are that some other drugs were tested only in a single task, and that the various studies differed in terms of the age of mice used, the dose used, and the details of L/M protocols. While additional, more comprehensive experiments could clarify the contributions of these issues, the data overall suggest that single drugs are unlikely to produce the desired efficacy, and that combinations of drugs should be explored.

Impaired adult neurogenesis has been linked to deficits in L/M in specific tasks. Of interest, ablation of adult neurogenesis in rats by hippocampal irradiation resulted in impaired performance in CFC (but not in a discrimination version of the T-maze).120 Rescue of impaired adult neurogenesis plus L/M in the Ts65Dn mice was investigated with five different drugs with inconsistent results. When neurogenesis was rescued by treating adult mice with lithium or by treating prenatal and perinatal mice with fluoxetine, performance in CFC was also rescued. However, performance on the T-maze was not rescued by lithium, while performance on the Y-maze was rescued by fluoxetine.53,70 (It should be noted that both of these experiments tested for spontaneous alternation; not discrimination as used in Hernández-Rabaza et al120.) Similarly, RO4938581 rescued both neurogenesis and performance on the MWM, but conversely, prenatal choline treatment only partially rescued adult neurogenesis, but it completely rescued performance on the WRAM. Moreover, formoterol failed to rescue neurogenesis, but it partially rescued CFC. While impaired adult neurogenesis is well established in the Ts65Dn mice, the requirement for normal adult neurogenesis for successful learning in CFC and various mazes needs further study.

Two drugs/drug targets were tested by multiple groups with inconsistent results. Of four reports using fluoxetine, one reported severe negative outcomes of seizures and death.57 This study, however, used a much higher dose of fluoxetine than the other three, and indeed, such a high dose has since been shown by a second group to have adverse effects.58 Three drugs targeting cholinesterase activity were tested. Both physostigmine and galantamine successfully rescued performance in the four-arm alternation and an olfactory test, respectively, while donepezil failed to rescue performance on the MWM.60,61

It is also notable that physostigmine was successful in mice at 4 months of age, but not in mice at 10 months or 16 months of age.60 This could reflect an age-dependent level of neurodegeneration and loss of acetylcholine, which is too advanced at 10 months for rescue by a single injection of an anticholinesterase. Conversely, when treatments with minocycline and vitamin E were started at earlier ages – 7 months and 4 months, respectively – and maintained for several months, both prevented neurodegeneration at 10 months. However, when started at 5 months, melatonin only partially rescued neurodegeneration at 10 months, while prenatal choline only partially prevented it at 13–17 months.74,104,105 For L/M amelioration, older age at the start of treatment is not an absolute exclusion – PTZ, memantine, and xamoterol each completely rescued L/M and estrogen, partially, when mice were treated at >9–12 months of age.23,32,47,64

The age for pharmacological treatment is an important consideration not only for AD in DS, but also for ID in DS. For two drugs, fluoxetine and SAG 1.1, treatment solely in the early postnatal time periods resulted in successful L/M at young adult ages of 6 weeks and 16 weeks, respectively.53,77 Indeed, for rescue on the MWM (but not on the spontaneous alternation Y-maze), SAG 1.1 treatment involved only a single P0 injection.77 Three drugs were tested by the treatment of only pregnant dams: fluoxetine and NAP/SAL were injected during gestation;59,98 and choline was provided in food to dams during gestation and through weaning.105 In another treatment, vitamin E was provided in food both to dams during gestation and to the offspring through to the age of L/M testing.94 In each case, the prenatal drug exposure resulted in at least the partial rescue of L/M in adult offspring. As promising as such results are, they should be interpreted with caution. Adverse effects of fluoxetine on the developing fetus are a concern,121,122 and as Reeves et al have discussed,76,77 activation of the sonic hedgehog pathway is seen in many malignancies and thus cannot be assumed to be completely benign either. While choline is an essential nutrient, its biochemistry is linked to folate and one-carbon metabolism; as will be discussed, several genes in this pathway map to the Mmu10 syntenic region and therefore their effects in trisomy cannot be assayed in the Ts65Dn mice. Lastly, clinical trials with vitamin E failed to prevent cognitive decline (see the following sections).

In spite of the partial rescues, it is of interest that so many drugs of such diverse targets can all be successful in rescuing complex tasks of L/M. Just how diverse the targets and mechanisms of action are, however, remains to be fully elucidated. For example, memantine, in addition to antagonizing NMDA receptor activation, also acts as a dopamine receptor agonist123 and indirectly increases dopamine levels.124 Memantine and fluoxetine act synergistically in models of depression to rescue behavior in the forced swim test.125 NMDA receptor activation modulates GABA receptor activity through phosphorylation of the γ subunit and fluoxetine alters GABA transmission in the hippocampus.126,127

Current status of clinical trials for ID in DS

Several clinical trials for cognitive deficits in DS have been completed recently and/or are in progress. Such trials are challenging. Outcome measures are complex and not standardized, and improvements in a specific task(s) are not straightforward enough to translate to improvements in important quality-of-life factors such as confidence, independence, or job satisfaction. These challenges are not unique to ID in DS. They are being addressed in other causes of ID (eg, Fragile X), and proposed solutions may be applicable to DS.128

Some very small trials (fewer than ten participants) have tested the cholinesterase inhibitors, donepezil, and rivastigmine, based on the observations discussed earlier regarding abnormal levels of acetylcholine in the DS brain.129–131 While some positive outcomes were reported for individual participants, overall, these treatments were considered ineffective, and the trials were hampered by low power.

The efficacy of folate supplementation was tested in a larger trial where infants with DS, aged 3–30 months, received leucovorin (folinic acid) or placebo daily for 12 months.132 The choice of folate was based on the fact that seven Hsa21 genes are involved in folate metabolism, and evidence for their elevated expression and resulting perturbed concentrations of folate pathway components have been observed in DS. The primary outcome measure was a significant increase in global development age. While only a small effect was seen with folate, this effect was significantly increased in children who were also receiving thyroxine for hypothyroidism. Only two of the seven genes predicted to affect folate metabolism are trisomic in the Ts65Dn mice, making it an inadequate model for preclinical evaluations of folate-related supplements. The Hsa21 connection to the thyroid hormone may involve NRIP1, which is also not trisomic in the Ts65Dn mice.133 A larger trial with folate plus thyroxine is in progress (see http://clinicaltrials.gov).

Other trials are based, at least in part, on positive preclinical evaluations in Ts65Dn mice. A 2-year trial evaluated the effects of daily antioxidant intake for 53 individuals with DS and dementia.134 No difference in cognitive function and decline was seen between the treated and placebo groups.134 Also to evaluate treatment in older individuals with DS (>40 years old), the first large randomized, double-blind trial135 evaluated memantine, which is already in use for moderate to severe AD in the typical population, for the prevention of the development or worsening of dementia. Eighty-eight participants received memantine and 85 received placebo for 52 weeks. There were no differences in outcomes between the groups; both groups declined in terms of cognitive level and function.135 It cannot be excluded that the neurodegeneration was too advanced in this age group for antioxidants and memantine to provide a benefit.

A smaller randomized, placebo-controlled trial tested memantine in 20 younger adults with DS (18–32 years old).136 After 16 weeks of treatment, there were no differences in the primary outcome measures assessing episodic and spatial memory between the memantine group and the placebo controls; however, there was a marginally significant outcome in one secondary memory measurement.136 Results of a small trial with the green tea component, EGCG, were recently reported.84 Fifteen participants aged 14–29 years were treated with EGCG for 3 months. Compared with the placebo group, those treated with EGCG showed a significant improvement in a test of visual memory.

Clinical trials currently recruiting participants include a Phase II trial with another GABAA α5 antagonist, RG1662 (Hoffman-La Roche Ltd., Basel, Switzerland), that started in May 2014 (http://clinicaltrials.gov). This is an international trial with centers in the US, France, Italy, and Spain already recruiting, and additional centers are planned in several other countries. The expected end date is August 2015. Treatment will run for 26 weeks. Also in the recruiting stage (in Australia) is a trial using PTZ (BTD-001; Balance Therapeutics, Inc., San Bruno, CA, USA). Participants in both of these trials will be in the age range of 13–30 years. Moreover, Elan Corporation Ltd (Dublin, Ireland) has started recruiting for a trial using scyllo-inositol as a method for inhibiting the aggregation of Aβ and, therefore, for preventing cognitive decline.

Validity of the Ts65Dn in preclinical drug evaluations

Clinical trials are expensive in terms of time and resources, as well as the goodwill of the individuals they are designed to help, and their families. So far, clinical trials for ID and AD in DS have not shown the positive outcomes predicted from evaluations in Ts65Dn mice. The validity of a mouse model for DS (and, indeed, any disease system) needs careful consideration if the rather poor results obtained to date in the clinical trials overall are to be improved.137 For almost 20 years, the Ts65Dn has been the primary mouse model for DS, and its study has vastly increased our understanding and predictions of the cellular, molecular, physiological, and behavioral deficits in DS. However, as a model for testing drug efficacy, the Ts65Dn needs to be judged critically and dispassionately. There are two features of the Ts65Dn that need to be considered: 1) the Ts65Dn is trisomic for orthologs of only 88 of 161 Hsa21 protein-coding genes (ie, it is not trisomic for 73 protein-coding genes [45%] that are trisomic in the large majority of people with DS); and 2) it is trisomic for 50 protein-coding genes that are not orthologs of Hsa21 genes and are not trisomic in individuals with DS. This gene content raises several questions: 1) would the phenotypic features be modified if the 73 missing orthologs were trisomic?; 2) do the 50 non-Hsa21 orthologs make any contribution to the phenotypic features of the Ts65Dn?; and 3) would drug responses be altered in either case? The answers to these questions are not known at this time, but a brief consideration of some gene functions is useful. Even with our limited knowledge regarding the functions and interactions of Hsa21 genes (a complete list of Hsa21 protein-coding genes and their distribution among mouse chromosomes and the Ts65Dn can be found in Ahmed et al16), many that are not trisomic in the Ts65Dn can be predicted to contribute to some aspect of ID or AD in DS; Table 2 lists 13 examples.

It has been repeatedly observed in cell lines and in tissues from both DS and DS mouse models that gene dosage does not result in exactly 50% increases in trisomic gene expression.8 Rather, there is considerable interindividual variation in the level of expression perturbation, with some genes showing <50% and others showing >>50% elevation. There are also gene, tissue, and development timing specificities in the observed levels of aberrations. Interindividual variation in the levels of gene expression is a natural phenomenon and is expected, at least in part, from allelic variation. Such variation is likely to contribute to the observed variation in phenotypic severity seen in DS, including in the level of cognitive impairment. It has also been proposed that the expression and phenotypic variation in DS could be due to epigenetic perturbation – ie, trisomy-induced aberrations in patterns of DNA methylation and histone modification.138 Indeed, altered patterns in DNA methylation in DS tissues have been reported.139 A number of Hsa21 genes could be relevant to epigenetics – among them, DNMT3L, which interacts with DNA methyl transferases, as well as the transcriptional regulators, HMGN1 and NRIP1, which bind histones and regulate chromatin configuration. Ten Hsa21 proteins, based on known functions, and Hsa21-encoded microRNAs, were proposed to potentially influence epigenetic phenomena.138 The same authors further proposed that these genes be used as a first group of targets for drugs. They did not discuss the challenges inherent in the design and simultaneous use of multiple drugs each directed at a different target, nor the potential advantages in considering targets that are downstream of the Hsa21 individual perturbations – eg, DNA methylase activities or histone modifications. It is noteworthy that several of the genes they propose as targets (eg, DNMT3L and NRIP1) are not trisomic in the Ts65Dn.

The cyclic nucleotide, cyclic guanosine monophosphate (cGMP), functions as an intracellular second messenger that has established roles in synaptic plasticity.140 Several phosphodiesterases regulate cGMP levels through degradation. Among them, the Hsa21-encoded protein, PDE9A, has the highest affinity for cGMP and is highly expressed in the hippocampus.140 Relevance to DS stems from the recent demonstrations that inhibition of PDE9A successfully rescued learning impairments in AD mouse models and prevented Aβ-induced neurotoxicity.141 The overexpression of PDE9A in DS could be an exacerbating influence both for cognitive deficits in DS, as well as for AD in DS. The PDE9A gene maps to the Mmu17 orthologous region of Hsa21, and therefore its effects cannot be assayed in the Ts65Dn.

Endostatin is an antiangiogenic fragment derived from the C-terminal of the Hsa21-encoded protein collagen 18A1. Endostatin is found with Aβ in plaques, and with cerebral vasculature in the AD brain and apoptotic neurons.142 It is induced following stroke and traumatic brain injury.143 It is not clear from these observations if endostatin is exacerbating neurodegeneration and damage, or if it is possibly serving a compensatory role. More recently, Al Ahmad et al144 studied the functions of endostatin in vitro in PC12 cells and demonstrated that endostatin inhibits NGF-induced differentiation, migration, and neurite outgrowth. Notably, these effects were dependent upon endostatin concentration, although the concentrations used could not be directly extended to predictions in the DS brain. In drosophila, endostatin functions in the regulation of neurotransmitter release, and it is required for maintenance of presynaptic homeostasis.145 The possible contribution of elevated COL18A1 and endostatin to phenotypic features in DS has not been studied. Of note, COL18A1 maps to the Mmu10 orthologous region.

Three trefoil factor genes, TFF1–TFF3, are clustered within distal Hsa21q and map to the Mmu17 orthologous region. TFF3 has been studied as a neuropeptide. In control mice, acute injection of TFF3 prior to training in NOR was shown to enhance acquisition and retention.146 While it has not been examined with respect to DS, it is notable that in the three DS mouse models trisomic for TFF3, the Tc1 (which carries a partially deleted Hsa21 encodinĝ120 Hsa21 protein-coding genes) shows a more mild phenotype than Ts65Dn, while the Dp17 shows enhanced LTP, and the Ts1Yah (which is trisomic for a subset of nine genes from the Mmu17 region) shows enhanced learning on the MWM.16,147,148 Therefore, unlike predictions for most other Hsa21 genes, overexpres-sion of TFF3 in trisomy may have positive or compensatory contributions to the DS phenotype.

Hsa21 components of the folate pathway have already been discussed in relation to a successful preliminary trial; one (cystathionine beta-synthetase) maps to the Mmu17 region and four to the Mmu10 region. The NRIP1 gene may play a role in the same clinical trial through the regulation of thyroid hormone expression.

While triplication of APP is promoted as the cause of AD in DS, it is important to note that there is considerable evidence to suggest that at least four Mmu10 genes – CSTB, SUMO3, S100B, and COL18A1/endostatin – may modulate this process. Each has been shown to affect the processing of APP and/or neuroinflammation, and/or to accumulate in AD plaques,149–152 which is discussed further.153 Furthermore, S100B is sensitive to the redox environment, showing either neuroprotective or neurotoxic properties, and it causes L/M deficits and neurodegeneration when overexpressed. Elevated levels of the Mmu10 gene, TRPM2, in DS might directly affect the efficacy of melatonin signaling and indirectly affect memantine efficacy through the impact on NMDA receptor activity.154,155

The majority of these non-Ts65Dn Hsa21 proteins are encoded by the Mmu10 region, and therefore it is notable that Dp10 mice have shown no impairment in CFC, NOR, MWM, or LTP at 2–4 months of age.19 These mice do, however, have very abnormal protein profiles at 8 months of age.153 These facts suggest that the abnormal protein profiles may include compensatory responses and protective effects, and/or that these mice develop age-related L/M impairments coincident with protein abnormalities. These possibilities need to be investigated. That genes mapping to the Mmu17 region also need to be considered is emphasized by a recent report showing that L/M deficits and LTP abnormalities present in Dp16:Dp17 mice that are trisomic for both the Mmu17 and the Mmu16 regions are not rescued by the same genetic manipulations that rescue them in the Dp16.156 This is an especially salient observation; a similar phenotype can have differing underlying molecular genetic causes and, therefore, may also have differing responses to pharmacotherapies.

The 50 non-Hsa21 orthologs trisomic in the Ts65Dn have not been considered by DS researchers for their potential influences on phenotypic features. These genes are, however, overexpressed in the Ts65Dn,15 and thus there is no logical motivation for assuming that they make no contribution. Among them are sorting nexin 9, which complexes with the Hsa21 DS cell adhesion molecule (DSCAM) involved in axon guidance, and that causes abnormal synaptic vesicle recycling when overexpressed.157,158 Several dynein light chain genes are also present; their overexpression perturbs early endosome signaling,159 similar to that seen in the Ts65Dn, which is notably only partially rescued with APP gene copy number reduction.160 Other Mmu17 genes trisomic in the Ts65Dn are paralogs of the Hsa21 genes and include the guanine nucleotide exchange factor TIAM2, the phosphoinositol phosphatase SYNJ2, and the cGMP degrading PDE10A. If their functions and expression patterns even partially overlap with the Hsa21 genes, they may be increasing perturbations seen in the Ts65Dn over what occurs in trisomy Hsa21 alone.

Lastly, Troca-Martin et al161 have proposed the mechanistic target of rapamycin (MTOR) pathway inhibitor, rapamycin, as a potential treatment for cognitive deficits in DS. This is based on the well-established role of the MTOR pathway in L/M, and on the observation of elevated phosphorylation of MTOR and AKT in the Ts1Cje mouse model of DS. The Ts1Cje is trisomic for a subset of ~70 protein-coding genes trisomic in the Ts65Dn. Rapamycin is US FDA approved and is being investigated in clinical trials for other IDs, including Fragile X, tuberous sclerosis, and Rett syndrome (http://clinicaltrials.gov). Similar aberrations in AKT and MTOR were not seen in the Ts65Dn, or in the Tc1 DS mouse model;16,162 however, several abnormalities in levels of MTOR pathway components are present in the Dp10 and Dp17 models153 (source: Gardiner). In addition, when the Ts65Dn are exposed to CFC, the dynamic responses in the MTOR pathway seen in control littermates are either absent or abnormal in the Ts65Dn (submitted to PLoSOne; currently under review). However, significant perturbations in the mitogen-activated protein kinase (MAPK) pathway, that is also robustly associated with L/M, have also been reported in the Ts65Dn and Tc1,16,162 and are present in the Dp10 and Dp17 (source: Gardiner et al). Thus, as with other proposed drug treatments, multiple options must be considered.

Conclusion

The reproducible rescue of L/M deficits in the Ts65Dn, with a large and diverse panel of drugs, is a valid reason for optimism that effective human clinical trials for cognitive deficits in DS will be possible. That these preclinical evaluations were successful in a broad age range of adult (as well as in pre and perinatal) Ts65Dn mice is of major significance. There are an estimated 300,000 people with DS in the US, thus forming a very significant target treatment population. While prevention of cognitive deficits is an important goal, prenatal treatments come with their own more stringent and challenging requirements for safety demonstrations. The details of the Ts65Dn experiments used in preclinical evaluations are important; different drugs rescued L/M in different tasks, or they produced only partial rescue, with and without attendant rescue of neurogenesis. This suggests that combinations of drugs may be both necessary and advantageous. This approach is common in cancer therapy and viral infections, and recent methods for the rapid identification of optimally effective combinations of drugs could be easily applied to in vitro DS systems.163,164 In spite of reasonable cause for optimism, however, the genetic limitations of the Ts65Dn are clear and may have consequences for outcomes of clinical trials. If phenotypic features and drug responses are affected in DS by Hsa21 genes that are not trisomic in the Ts65Dn, as well as by the non-Hsa21 genes that are trisomic, then preclinical evaluations may be less reliable and their effective translation to human clinical trials less successful than outcomes in the Ts65Dn suggest. Solutions to this challenge are not clear. The use of the full trisomy mouse that can be created by crossing the three duplication lines20 is very expensive, especially when testing of multiple drugs and multiple L/M tasks is the objective. Exploration of drug responses in human, neurally relevant, DS-derived cell lines could complement studies with additional mouse models of DS. Elucidation of the molecular basis of drug responses has received little attention to date, but it could be used to identify features common to successful drugs, leading to potentially more efficacious drug targets.

Acknowledgments

Footnotes

References