Construct Validity

Construct validity refers to the similarity of the genetic context used to develop the mouse model to that found in HD patients (Fleming et al., 2005). The genetic context should be viewed at multiple levels. At the DNA level, one should consider whether the types of mutations introduced (i.e., CAG repeat length) resemble those in the patients. In this regard, the targeted insertion of expanded CAG repeat and/or mhtt exon 1 into the endogenous murine huntingtin disease homolog (Hdh) locus in the mouse genome (i.e., knockin models) is the most genetically precise model of HD (Gusella and Macdonald, 2006). In the transgenic models of HD (with random insertion of transgenes into the mouse genome), one also needs to consider whether the genomic DNA context of the mutant transgene is similar to that in the patients. At the RNA level, one should consider whether the regulation of mhtt transcription is similar to the endogenous human htt, whether it is driven by an exogenous promoter, and whether endogenous mhtt mRNA processing such as splicing is preserved (i.e., models with an intact htt genomic locus). Finally, at the protein level, one needs to consider how similar the expressed mhtt protein in a given mouse model is to that in patients. Any divergence from the full-length human mhtt protein sequence may also reduce the construct validity of a model. For example, mouse models only expressing mhtt N-terminal fragments may have reduced construct validity at the protein level. Moreover, because the murine Hdh differs from its human counterpart in the polyproline region (polyP, a known protein–protein interaction domain of htt) and in another 273 amino acids outside the polyQ and polyP regions, these amino acid differences may subtly modify the disease processes mediated by the mutant Hdh compared with those by human mhtt.

Face Validity

Face validity refers to whether the overall phenotypes of a disease mouse model, both behavioral deficits and neuropathology, faithfully recapitulate those in HD patients. As described in more detail below, the phenotypes of the current preclinical HD mouse model can be studied at multiple levels to fully uncover its disease-relevant phenotypes. Commonly used behavioral phenotyping tools in HD mice include longitudinal behavioral observations (i.e., SHIRPA) (Rogers et al., 1997); motor behaviors such as rotarod performance (the ability to run a rotating wheel) and open field activity (automated measurement of spontaneous activity in an open field); cognitive tests that include a variety of learning tasks that may depend on the hippocampus or the corticostriatal circuit, such as swimming T mazes (Holmes et al., 2002) and instrumental conditioning (Balleine, 2005); and psychiatrically related behaviors such as anxiety and depressive behaviors and prepulse inhibition (PPI) (Arguello and Gogos, 2006; Holmes et al., 2002). Commonly used neuropathological studies in HD mice include stereological measurement of striatal and cortical volume and stereological counting of the striatal neurons (Menalled et al., 2002), counting striatal degenerating dark neurons as measured by toluidine blue staining of semithin brain sections (Gray et al., 2008; Hodgson et al., 1999; Turmaine et al., 2000), gliosis (Gu et al., 2005), and microgliosis (Simmons et al., 2007). In addition, weight loss and premature death are also used as secondary readouts in several preclinical mouse models. Each mouse model of HD, before entering the preclinical studies, has been extensively characterized to determine the extent of its phenotypic similarities to HD. Those models with relatively early-onset, progressive, and robust disease-like phenotypes (Table 7.1) are now being pursued for preclinical studies in HD.

Predictive Validity

An ultimate test for the validity of an HD mouse model is its predictive validity, which refers to whether key mechanistic findings and therapeutic efficacy in a given HD mouse model can predict positive clinical outcomes in human patients and vice versa. Currently, there is no effective disease-modifying therapy available for HD that can be used to validate the existing HD mouse models. Conversely, because the mouse genetic models of HD are widely used to identify pathogenic mechanisms and to test preclinical candidates, future clinical studies in HD patients using therapeutic leads originating from the mouse studies may be crucial to reveal the predictive validity of these models.

The concepts of construct validity, face validity, and predictive validity will be crucial to guide us in appreciating the strengths but also recognizing the weaknesses of the existing preclinical mouse models of HD. They may also be helpful in choosing the appropriate mouse models for preclinical studies. In the following section, we will describe the three major types of HD mouse models that are currently used or being prepared for use in HD preclinical studies. They include those expressing a small toxic N-terminal fragment of mhtt, those with precise genetic modification of the endogenous murine Hdh, and those human genomic locus transgenic mice expressing full-length human mhtt under its endogenous regulatory elements (Table 7.1). We will describe for each model its genetic construct and key phenotypic outcomes that may be used in preclinical studies.

R6/2 Mice

Genetic Construct and Expression

The R6/2 model is the first and one of the most influential HD mouse models for preclinical studies to date (Mangiarini et al., 1996) (Table 7.1). It was created on the (C57BL/6J × CBA) F1 mixed genetic background. R6/2 mice carry about 1 kilobase (kb) of the human huntingtin promoter, which drives the expression of mhtt exon 1 with ~150 CAG repeats. The transgene also contains ~262 base pair (bp) of human htt intron 1 sequence after the exon 1. R6/2 mice express the mhtt-exon 1 transgene at about 75% of the endogenous Hdh level (Mangiarini et al., 1996). One issue with R6/2 is its genetic heterogeneity as a result of the CAG repeat instability in the germ line, which can range from low 100s to >300 repeats. Such dramatic repeat alterations may significantly alter R6/2 mouse phenotypes. Therefore, in a preclinical study, the repeat length for the R6/2 mice should be closely monitored to avoid genetic drift of the repeat lengths during the study.

Behavioral Phenotypes

The earliest motor deficits in R6/2 mice were detected at 4.5 weeks and include decreased running wheel activities in mice individually housed, climbing a wired cylinder, and hypoactivity in the open field test (Hickey et al., 2005). At 7–8 weeks, grip strength and rotarod deficits can be readily detected, and these deficits increase until the animal’s premature death around 16 weeks of age (Carter et al., 1999; Lione et al., 1999; Murphy et al., 2000). In addition, R6/2 mice also exhibit a variety of other motor phenotypes, including changes in circadian rhythms (Morton et al., 2005), clasping (a dystonic posture when the mice were suspended by the tail), tremor, involuntary jerky movements, and seizures (Li et al., 2005; Mangiarini et al., 1996).

Before the onset of overt motor deficits (at 8 weeks of age), R6/2 mice already exhibit a variety of cognitive deficits reminiscent of HD patients. For example, HD patients exhibit perseverance of learned tasks indicating frontal-striatal inhibition deficits (Aron et al, 2003). R6/2 mice also demonstrate difficulty in reversing prelearned tasks in the two-choice swim tank, T maze (Lione et al., 1999), and an automated video-based reversal task (Morton et al., 2006). R6/2 mice also exhibit deficits in spatial learning in the Morris water-maze task (Murphy et al., 2000) and in contextual fear conditioning (Bolivar et al., 2003). In the psychiatrically related deficits, R6/2 mice exhibit impairment in PPI, a sensorimotor gating abnormality also seen in the HD patients (Swerdlow et al., 1995) and in patients with schizophrenia (Swerdlow et al., 1994).

Neuropathology

R6/2 mice demonstrate robust brain atrophy (about 20% brain weight loss at 12 weeks). Striatal atrophy in these mice can be readily measured using unbiased stereology (Li et al., 2005). R6/2 mice exhibited only moderate dark neuron degeneration in the late stage (Mangiarini et al., 1996; Turmaine et al., 2000), but the striatal neuronal atrophy can be readily quantified as a preclinical outcome. Another frequently used pathological endpoint in R6/2 mice is the presence of mhtt-containing nuclear inclusions (NIs) and neuropil aggregates (NAs), some of which can be detected with antibodies against ubiquitin or against the N-terminal region of huntingtin, for example, EM48 antibody (Davies et al., 1997; Gutekunst et al., 1999; Li et al., 1999). The aggregates in this model are much more widespread than those observed in patients using the same EM48 antibody (DiFiglia et al., 1997; Gutekunst et al., 1999). Moreover, because the significance of mhtt aggregates to the pathogenesis of HD neuronal dysfunction and degeneration remains unclear (Bates, 2003), and there is accumulating evidence suggesting that the large nuclear inclusions may be neuroprotective (Arrasate et al., 2004), mhtt aggregates should not be used as the sole pathological outcome in a preclinical study.

Survival and Weight Loss

Two commonly used preclinical outcomes in R6/2 mice are weight loss and premature death. Because these phenotypes can be readily modified by environmental enrichment (Hockly et al., 2002), which may confound any therapeutic study, it is now recommended that standardized living conditions with a moderate level of enriched environment should be implemented for this model (Hockly et al., 2003). Because the underlying etiology for weight loss and death in R6/2 mice remains unclear, these phenotypes are often only used as surrogate markers to support findings from the primary behavioral and neuropathological outcomes (Bates and Hockly, 2003; Hockly et al., 2003).

Other Phenotypes

R6/2 mice also have a set of molecular readouts, such as gene expression changes, that could be readily used in preclinical studies. A recent microarray analysis comparing gene expression changes in HD mice and HD patients revealed that many of the gene expression changes seen in the R6/2 striatum significantly overlap with those occurring in patients (Kuhn et al., 2007). This result supports the idea of using gene expression signatures as a biomarker for preclinical efficacy in R6/2 mice. Other promising biomarkers in R6/2 mice include using NMR spectroscopy to detect a robust (53%) nonlinear decrease in in vivo N-acetyl aspartate (NAA) levels beginning at 6 weeks of age, a brain energetic deficit also seen in patients (Jenkins et al., 2000). Finally, 8-hydroxy-2-deoxyguanosine (8-OHDG), a biomarker for oxidized DNA damage, which is elevated in HD patient serum (Hersch et al., 2006), is also elevated in R6/2 and R6/1 mice (a model with similar construct to R6/2) (Bogdanov et al., 2001; Kovtun et al., 2007). These latter biomarkers (NAA and 8-OHDG), once validated in HD patients, may be used as relatively noninvasive and unbiased endpoints for preclinical and clinical studies in HD (Hersch et al., 2006).

N171-82Q Mice

Genetic Construct

Another commonly used N-terminal fragment mouse model of HD in preclinical studies is the N171-82Q mouse model. N171-82Q mice express an N-terminal fragment of human htt with 82 polyglutamine repeats and the first 171 amino acids of htt in all the neurons of the brain driven by a mouse prion protein promoter (Schilling et al., 1999). The N171-82Q line was created on the C3H/HeJ × C57BL/6J background. This model expresses mhtt fragment at about 20% of levels of the endogenous murine Hdh protein (Schilling et al., 1999).

Behavioral Phenotypes

N171-82Q mice exhibit progressive motor deficits starting at 12 weeks of age, including tremors, uncoordination, hypoactivity, abnormal gait, and clasping (Schilling et al., 1999). These symptoms are progressive until premature death at approximately

5–6 months of age. The most used quantifiable motor phenotype in this model is rotarod impairment, beginning at approximately 11 weeks of age (Andreassen et al., 2001; Schilling et al., 2004b). The behavioral deficits in N171-82Q mice appear to be

more variable compared with R6/2 mice; thus, to attain adequate statistical power, a typical preclinical study with N171-82Q mice would require a large number of mice (i.e., minimum of 20) than a study with R6/2 (Hersch and Ferrante, 2004).

Neuropathology

N171-82Q mice have several pathological features mimicking HD. First, they exhibit brain weight loss at 120 days of age (Andreassen et al., 2001; Hersch and Ferrante, 2004), striatal atrophy (as measured by stereological measurement of the ventricular volume), and striatal neuronal atrophy (Gardian et al., 2005). Moreover, these mice exhibit other neuropathological features more similar to HD patients, including more mhtt aggregates in the cortex than in the striatum, robust reactive gliosis at 4–5 months of age, and apoptotic neuronal degeneration can be detected in the cortex and striatum at 4.5 months of age (Yu et al., 2003).

Survival and Weight Loss

Similar to R6/2 mice, N171-82Q mice also exhibit progressive weight loss (beginning at approximately 90 days of age) and shortened lifespan (average survival of about

130 days). These phenotypes are frequently used as surrogate markers of therapeutic efficacy in the preclinical studies. Because environmental enrichment can also significantly modify these phenotypes (Schilling et al., 2004a), standardized housing conditions are critical in preclinical trials using this model.

Other Phenotypes

N171-82Q mice exhibit hyperglycemia (i.e., increased baseline fasting glucose level and impaired glucose tolerance test) at about 80 days of age, which could be used as an outcome measure (Andreassen et al., 2001). The utility of the test is relatively limited because of its unclear etiology and high variability.

Hdh^Q111 Mice

Genetic Construct

The Hdh^Q111 model is a KI mouse model of HD generated on a CD1 × 129Sv background. This model targeted a chimeric murine Hdh/human mhtt exon 1 into the endogenous Hdh locus, and the human mhtt portion includes 111 CAG repeat and human polyP region (Wheeler et al., 1999).

Behavioral Phenotypes

The behavioral phenotypes of Hdh^Q111 are very mild. Gait abnormalities are detected at 24 months of age, and no rotarod, clasping, or open field abnormalities were detected in heterozygous or homozygous Hdh^Q111 mice up to 17 months of age (Wheeler et al., 2002).

Neuropathology

Hdh^Q111 homozygous mice display selective accumulation of nuclear mhtt at 2.5 months of age, as stained by EM48 antibody (Wheeler et al., 2000). This phenotype progresses as EM48-positive puncta are detected at 5 months of age. Nuclear inclusion formation is first detected at 10 months of age (Wheeler et al., 2000, 2002). At 24 months, these mice also demonstrate reactive gliosis in the striatum and a small number (about 3.5%) of striatal dark neurons as stained by toluidine blue (Wheeler et al., 2002). Brain atrophy has not been reported in this model.

Other Phenotypes

Hdh^Q111 mice exhibit a variety of molecular phenotypes, including but not limited to increased expression of a ribosomal signaling protein Rrs1, somatic Hdh CAG- repeat instability in the striatum, and low mitochondrial ATP levels (Fossale et al., 2002; Lloret et al., 2006; Seong et al., 2005; Wheeler et al., 2003).

CAG140 Mice

Genetic Construct

A second preclinical Hdh-KI mouse model has targeted replacement of the endogenous murine Hdh exon 1 with a chimeric mouse and human exon 1 with 140 CAG repeats (Menalled et al., 2003). They were generated on a 129Sv × C57BL/6J background.

Behavioral Phenotypes

CAG140 mice exhibit early hyperactivity as measured by an increase in rearing at 1 month of age, followed by hypoactivity at 4 months of age (Menalled et al., 2003). This pattern of locomotor activity changes is similar but with an earlier onset compared with another Hdh-KI model, CAG94 (Menalled et al., 2002). CAG140 mice exhibit gait abnormality at 12 months of age, with a decrease in stride length (Menalled et al., 2003).

Neuropathology

CAG140 mice exhibit selective striatal and cortical nuclear staining of mhtt microaggregates by EM48-positive antibody (Menalled et al., 2003). Nuclear inclusions are present in the striatum at 4 months of age and do not appear in the cortex until 6 months of age. At this age, nuclear staining and nuclear microaggregates are also seen in the cerebellum, an area relatively spared in HD. Neuropil microaggregates are also present in the striatum, globus pallidus, and piriform cortex at 2 months of age and increase with age. To date, no overt cell loss or brain atrophy has been reported for this model. However, CAG94, which is a model very similar to CAG140, exhibits significant striatal atrophy (15%) but no neuronal loss at 18–26 months of age as measured by unbiased stereology (Menalled et al., 2002). Therefore, it is very likely that CAG140 may exhibit significant striatal atrophy phenotype that could facilitate its utility in a preclinical study.

Hdh^(CAG)150 Mice

Genetic Construct

The third preclinical Hdh-KI model is Hdh^(CAG)150, which replaced the short CAG repeat in the murine Hdh exon 1 with a repeat of 150 CAG (Lin et al., 2001). These mice, created in a 129/Ola × C57BL/6J background, were analyzed either in this mixed genetic background (Heng et al., 2007) or in a (CBA × C57BL/6) F1 background generated from C57BL/6 and CBA congenic lines (i.e., backcrossed for >12 generations) (Woodman et al., 2007). The latter breeding strategy ensures the homogeneity of the genetic background in the mutant and control mice (Silva, 1997).

Behavioral Phenotypes

Homozygous Hdh^(CAG)150 mice were extensively studied using a battery of behavioral paradigms that revealed several slowly progressive motor abnormalities (Heng et al., 2007; Woodman et al., 2007). In the mixed genetic background, a mild exploratory deficit is present in Hdh^CAG(150) mice at 70 weeks of age, with more robust deficits seen at 100 weeks (Heng et al., 2007). In the rotarod test, homozygous mice display a significant deficit on the rotarod at 40 weeks (Lin et al., 2001) and 100 weeks of age (Heng et al., 2007). At the latter time point, these mice also exhibit gait abnormalities, difficulty in traversing balance beams, and clasping.

In the (CBA × C57BL/6) F1 background (Woodman et al., 2007), homozygous Hdh^CAG(150) mice also have onset of significant and progressive rotarod deficits at 18 months of age. A grip-strength deficit was observed at 6 months of age, but it is nonprogressive until 10 months of age. No locomotor deficits were detected in the mutant mice up to 18 months of age, suggesting this phenotype may depend heavily on the genetic background of the mutant mice.

Neuropathology

Mutant huntingtin aggregates can be detected in the striatum and hippocampus in the homozygous Hdh^CAG(150) mice at 6 months of age and become widespread in the brain at 10 months of age (Tallaksen-Greene et al., 2005). Similar to what was observed in the patients, the recent study reveal that at 100 weeks, Hdh^CAG(150) homozygous mice exhibit reduced striatal D1 and D2 dopamine receptor ligand binding and reduced dopamine transporter (Dat) ligand binding (Heng et al., 2007). Using stereological counting of NeuN(+) striatal neurons, the same group found that the Hdh^CAG(150) mice exhibit 40% reduction in striatal volume and 43% reduction in striatal NeuN(+) neurons. This latter stereological result is very promising to use striatal atrophy and/or neuronal loss as preclinical outcomes in this model. However, because these results were obtained from a relatively small sample size, it would be important for the study to be replicated using a much larger sample size, and preferably also in recombinant inbred F1 background.

Survival and Body Weight

Hdh^CAG(150) mice do not have a shortened lifespan. However, in both mixed and F1 genetic backgrounds, the mutant mice exhibit significant and progressive weight loss at 70 weeks of age (Heng et al., 2007) and at 52 weeks in the pure F1 background (Woodman et al., 2007).

Molecular Phenotypes

Brain gene expression profiling in the Hdh^CAG(150) F1 mice revealed parallel changes in 22-month-old Hdh^CAG(150) mice and 12-week-old R6/2 mice (Woodman et al., 2007). Decreases in the expression of chaperones (i.e., heat shock protein 70, heat shock cognate 70) can be confirmed by Western blots in both models. Such parallels in the molecular phenotypes between the two different types of HD mouse models support the further exploration of gene expression changes as biomarkers in preclinical studies.

YAC128 Mice

Genetic Construct

A series of YAC transgenic models of HD expressing full-length human mhtt with 18, 46, 72, and 128 CAG repeats (i.e., YAC18, YAC46, YAC72, and YAC128) were generated in the inbred FvB background (Hodgson et al., 1999; Slow et al., 2003; Van Raamsdonk et al., 2007). In these models, the YAC transgenes carry the entire 170-kb genomic locus of human htt gene plus 25 kb of the 5′ flanking sequence and about 120 kb of 3′ flanking sequence. YAC128 mice are the latest of the series and exhibit by far the most robust phenotypes among all the YAC models; therefore, YAC128 is used as a preclinical model in HD (Slow et al., 2003). At the protein level, the YAC128 line expresses human mhtt at about 75% of the level of the endogenous murine Hdh (Slow et al., 2003).

Behavioral Deficits

YAC128 mice exhibit hyperactivity at 2 months of age and hypoactivity at 8–12 months of age. They also exhibit rotarod deficit initially at 4 months of age but become more prominent at 6 months of age (Graham et al., 2006; Van Raamsdonk et al., 2005c). YAC128 mice demonstrate a variety of cognitive deficits, including motor learning deficits on rotarod beginning at 2 months, a reversal learning deficit in a swimming T maze beginning at 2 months, PPI deficits at 12 months, openfield habituation test at 9 months, and a linear swimming test at 8 months (Van Raamsdonk et al., 2005e).

Neuropathology

The earliest pathological marker in YAC128 mice is the selective nuclear localization of EM48-positive mhtt in the striatum at 1–2 months of age, which intensifies at 3 months of age (Van Raamsdonk et al., 2005a). EM48-positive mhtt nuclear accumulation is present in the cortex and hippocampus at 3 months of age, and EM48-positive nuclear inclusions appear in the striatum by 18 months of age (Van Raamsdonk et al., 2005a). The YAC model is the first HD model to demonstrate selective atrophy in the striatum and cortex but not in the cerebellum, a pattern reminiscent of that in HD (Van Raamsdonk et al., 2005a). Stereological studies reveal striatal volume loss first detected at 9 months of age, as well as a significant decrease in striatal volume (10%), cortical volume (8.6%), and globus pallidus (10%) at 12 months of age but no change in the hippocampus or cerebellar volumes (Van Raamsdonk et al., 2005a). The striatal volume loss is associated with a loss of NeuN positive neurons (18%) (Van Raamsdonk et al., 2005a).

Survival and Body Weight

The lifespan of YAC128 mice is slightly decreased. YAC128 mice gain weight between 2 and 6 months of age (about 27%), and the weight gain occurs in all organs except the brain and testis, where mhtt is particularly toxic and causes weight loss (Van Raamsdonk et al., 2006). The weight gain phenotype appears to be caused by the overexpression of human htt because YAC18 mice overexpressing wild-type htt also exhibit similar weight gain phenotype. Because HD patients exhibit weight loss rather than weight gain, the weight gain phenotype in the YAC128 model is not a feature of the disease and should not be used as a preclinical endpoint in this model.

BACHD Mice

Genetic Construct

BACHD is a novel transgenic mouse model of HD generated and maintained in the FvB inbred background. BACHD mice express full-length human mhtt from its own regulatory elements on a 240-kb BAC, which contains the intact 170-kb human htt locus plus about 20 kb of 5′ flanking genomic sequence and 50 kb of 3′ sequence (Gray et al., 2008). The BAC was engineered to include an mhtt exon 1 containing a mixed CAA/CAG repeat encoding an intact polyglutamine stretch (Kazantsev et al., 1999; Yang et al., 1997). A recent more precise sizing of the BACHD CAA/CAG repeat by Laragen (Los Angeles, CA), using both direct sequencing and GeneMapper methods, shows that the BACHD CAA/CAG repeat length is 97. Unlike Hdh-KI and R6/2 mice, the CAA/CAG repeat length in BACHD mice appears stable in the germline over many generations. Another feature of the BACHD transgene design is the inclusion of two LoxP sites flanking mhtt-exon 1, which permits the selective removal of mhtt expression in any cell types expressing the Cre recombinase (Branda and Dymecki, 2004; M. Gray and X. W. Yang, unpublished data) Therefore, this model is particularly useful to study the cell autonomous toxicity and pathological cell–cell interactions elicited by mhtt (Gu et al., 2005). BACHD mice have five copies of the transgene integrated and express fl-mhtt protein at about 1.5- to 2-fold of the endogenous Hdh level. The expression of mhtt is in an endogenous pattern and functional because the BACHD transgene can rescue the embryonic lethality of the murine Hdh knockout mice (Gray et al., 2008; Zeitlin et al., 1995).

Behavioral Phenotypes

BACHD mice exhibit mild but significant rotarod deficits at 2 months, but repeated testing revealed that the rotarod deficits in these mice are progressive and become very pronounced at 6 and 12 months of age. At 6 months, BACHD mice also exhibit hypoactivity in the open field test. In the cognitive domain, a preliminary study reveals BACHD mice exhibit deficits in an instrumental reward learning paradigm, in which these mice are unable to efficiently learn the relationships between stimulus, action, and the rewarding outcomes (B. Balleine, unpublished data). The BACHD mice also exhibit significant impairment in acquiring the swimming T-maze task at 4–6 months of age (L. Menalled and D. Howland, personal communication). Finally, BACHD mice also exhibit several phenotypes in the psychiatric-like behavioral domain (Holmes et al., 2002) starting at about 6 months of age, including enhanced anxiety in the light-dark box and elevated plus maze, enhanced depressive-like behavior in forced swim test, and altered PPI (Menalled et al., 2009; M. Gray and X. W. Yang, unpublished data).

Neuropathology

At 6 months of age, BACHD mouse brains are indistinguishable from the wildtype controls and have comparable cortical and striatal volumes. At 12 months of age, BACHD brains are visibly atrophic with 20% reduction in the forebrain weight but normal cerebellar weight compared with the wild-type mice (Gray et al., 2008). Stereological measurement reveals a 28% reduction in the striatal volume and 32% reduction in the cortical volume compared with the wild types. Furthermore, using toluidine blue staining of striatal semithin sections, we found the 12-month-old BACHD mice contain about 15% dark degenerating neurons in the lateral striatum compared with only 0.3% in the control littermates. One distinguishing feature of BACHD mice from the other full-length huntingtin mouse models is the lack of early nuclear localization of EM48-positive mhtt in the striatum and cortex. Instead, at 12 months of age, EM48 staining reveals large and predominantly neuropil mhtt aggregates in the deep cortical layers and very few small aggregates in the striatum. This distribution pattern of mhtt aggregates is reminiscent of that in adult-onset HD (DiFiglia et al., 1997; Gutekunst et al., 1999), albeit the abundance of the aggregates in BACHD mice appears to be less than that in the patients.

Survival and Body Weight

BACHD mice have a normal lifespan compared with their wild-type littermates. Similar to YAC128 mice, BACHD mice also exhibit significant weight gain between 2 and 6 months and maintain the weight differential without further weight gain until 12 months. As discussed above, this phenotype may be related to the overexpression of human huntingtin. Importantly, rotarod performance and body weight at 6 months demonstrate that body weight in BACHD mice is not correlated with their poor rotarod performance; furthermore, a subset of BACHD mice within the normal weight range still exhibit robust rotarod deficits (Gray et al., 2008; Menalled et al., 2009). These results suggest that the neuronal dysfunction and degeneration in this model are related to the toxicity of mhtt in the brain and are unrelated to the body weight changes.

Power Analyses to Determine the Sample Size for the Study

A very important consideration in the design of preclinical trials is to perform power analyses for a given preclinical outcome to determine the number of HD mice needed for a given set of outcomes in a preclinical trial (Bates and Hockly, 2003). Similar to the human clinical trials, insufficiently powered studies can be costly and may not provide the necessary information to assess the efficacy of a compound. Performing power analyses before the mouse trial will help one to determine how many animals will be needed in a study to have a predetermined reasonable chance (i.e., 80% or 90%) of detecting a significant improvement at a predetermined magnitude of improvement (i.e., 30% improvement). The basic formula for sample size estimation is the following (adapted from Hockly et al., 2003):

In this formula, n represents the number of mice in each treatment group; μ1 and μ2 are the means of the two groups; and SD1 and SD2 are the standard deviations in the two groups. Furthermore, y is 1.96 for P = 0.05 (two-sided normal distribution); and z is 1.28 for power of 90% and 0.84 for power of 80%.

Because of cost concerns, power analyses are crucial for minimizing cost by selecting outcomes and/or HD mouse models that may require a relatively smaller number of mice per treatment group to detect a significant therapeutic benefit. For example, the use of R6/2 mice normally requires only about 10 mice per group for most of the phenotypic outcomes, whereas the use of N171-82Q mice may require 20 mice per group (Hersch and Ferrante, 2004). For the full-length models, most of the potential preclinical outcomes in these models should be analyzed by power analyses to determine whether they are suitable for preclinical studies. In the case of the YAC128 model, striatal atrophy is a particularly good readout because only eight animals per group are needed to provide 80% power to detect a significant (P < 0.05) treatment benefit that is 33% or greater (Slow et al., 2003). In BACHD mice, both our analyses (Gray et al., 2008) and an independent analysis at PsychoGenics (Tarrytown, NY) (L. Menalled and D. Howland, personal communications) reveal that the rotarod deficit at 6 months is a particularly robust phenotype outcome: only 22 BACHD mice are needed per group to provide 80% power to detect significant (P < 0.05) improvement of 33% or better. Thus, in both fragment and full-length models of HD, power analyses should be done for all the major outcomes so that one can begin the study with sufficient power to detect a therapeutic benefit.

Rigorous Standardization of Preclinical Mouse Trials

A key to reducing the variability in a mouse trial is the rigorous standardization of the testing condition (see HDF Cardiff Workshop Report; also Bates and Hockly, 2003; Hockly et al., 2003). There are several areas to which the investigator should pay particular attention. First, the genetics of a given HD mouse model has to be carefully controlled. If possible, an inbred mouse background should be used for a study. We do not recommend that all the mouse trials for all the different types of HD models should be done in the same genetic background, although it is well known that certain behavioral readouts may be more robust in one inbred background than another (Crawley et al., 1997). We do recommend, however, within the same type of model, a consistent and relatively pure genetic background (i.e., inbred or recombinant F1) should be used, and preferably a genetic background in which a given model exhibits the most robust phenotypes that will be used as preclinical outcomes (i.e., YAC128 model in FvB background) (Van Raamsdonk et al., 2007). For the fragment models such as R6/2, which can only be maintained in a mixed F1 background, one should consistently split transgenic littermates between the treatment group and control group, and use litters from different breeding pairs to avoid potential breeder effects (Hockly et al., 2003). Because CAG repeat size can vary in certain mouse models, such as R6/2 and Hdh-KI mice, the mice used in the study should be routinely genotyped to rule out any mice with dramatic expansion or contraction of the repeats, which may inf luence the phenotypes. Second, mouse housing and diet conditions should also be standardized. There are several reports that demonstrate environmental enrichment and dietary inf luences, which may significantly improve outcomes in HD mice (Hockly et al., 2002; Schilling et al., 2004a; Spires et al., 2004). For R6/2 mice, a moderately enriched housing condition is recommended (Hockly et al., 2003). For other HD mouse models, such as the full-length models, the housing conditions should be maintained constant in different trials and comparable across different laboratories. Finally, the outcome testing conditions and statistical analyses should also be the same for a given model in different trials and different laboratories. For example, the equipment and methods for behavioral testing and neuropathological studies (i.e., stereological counting and dark neuron quantization) may be different from laboratory to laboratory. Therefore, it is critical that those laboratories using the same preclinical mouse model reach a consensus on a set of standardized protocols for various outcome studies and make such standardized protocols accessible to others in the community. For the phenotypic studies, the investigators conducting the tests should be blind to the genotypes. Finally, statistical analyses should be performed in the same way for a given outcome in a given model, so one can readily compare the preclinical results of different compounds and their ability to be replicated by different laboratories.

Improving the Outcome Measures in Preclinical Mouse Studies

In the majority of preclinical trials with HD mice, the primary phenotypic outcomes are rotarod for motor behaviors and stereological measurement of the striatal atrophy. Secondary outcomes, such as weight loss and lethality, are only possible for the fragment models, and their relevance to HD remains unclear. However, in HD patients, the clinical features are far more complex and consist of motor, psychiatric, and cognitive deficits. Because we are currently unclear whether this clinical triad is caused by neurodegeneration or neuronal dysfunction in HD, and whether different brain regions may mediate these distinct deficits, the exclusive use of motor and striatal pathology outcomes may be insufficient, particularly in evaluating candidates that may selectively improve cognitiveand/or psychiatric-related phenotypes (which may be highly relevant to HD). For example, a recent preclinical study with R6/2 mice showed that the imposition of sleep with alprazolam selectively improves the cognitive deficits in this model but not motor deficits (Pallier et al., 2007). Currently, there is a concerted effort to develop more sensitive potentially automatable cognitive behavioral outcomes for the preclinical models of HD (Lione et al., 1999; Morton et al., 2006; Van Raamsdonk, et al., 2005e). In our view, the use of robust cognitive measures, especially those affecting the same corticostriatal neural circuit as in patients, should be evaluated carefully and should be included when possible as primary outcome measures in HD preclinical studies.

Finally, another concerted effort in the HD therapeutic field is the attempt to discover biomarkers that may reflect the disease process in patients and may be used in future clinical studies. Some of these potential biomarkers, such as gene expression alterations and/or specific molecular changes in the serum such as increase in 8-OHDG (a marker for oxidized damage to DNA), have already been elucidated in patients (Dalrymple et al., 2007; Hersch et al., 2006; Runne et al., 2007). Importantly, in a recent clinical study with creatine, 8-OHDG was shown to dramatically normalize with the treatment, thus indicating it may be a useful peripheral biomarker that can predict therapeutic benefit (Hersch et al., 2006). If similar biomarkers also are altered in preclinical mouse models, they could be incorporated in the mouse trial to provide the opportunity to noninvasively monitor preclinical outcomes in mice and clinical outcomes in patients.