Improvement of neuropathology and transcriptional deficits in CAG 140 knock-in mice supports a beneficial effect of dietary curcumin in Huntington's disease

A primary goal of the experiment was to determine whether curcumin, which has anti-amlyoid properties, could affect protein aggregates of huntingtin in the striatum of CAG140 KI mice. High concentrations of curcumin increases aggregate size in vitro [41] and indeed, using PC12 cells that inducibly express EGFP-tagged exon 1 of mutant htt [42], we found that 10 or 20 μM curcumin, applied at the same time as protein induction, increased aggregate size markedly at 48 or 72 h post mutant-protein induction (Figure 3; 48 h, Kruskal-Wallis ANOVA p < 0.01, *p < 0.05 compared to control treated cells; 72 h, Kruskal-Wallis ANOVA p < 0.01, *p < 0.05 compared to control treated cells). In contrast to 20 μM curcumin, 5 nM curcumin caused a slight reduction in aggregate size at 48 h (Figure 3; 48 h, Kruskal-Wallis ANOVA as above, *p < 0.05 compared to control treated cells) but the effect was very small (-8% reduction in size of all aggregates) and was only observed with the lowest concentration tested, which is well below the concentration of curcumin measured in brain in our study (Table 1).

Despite the lack of clear effects of low concentrations of curcumin in vitro on aggregate sizes, we observed a clear decrease in aggregate numbers in the striatum of KI mice after administration of curcumin in vivo. Only tissue from KI mice was examined because huntingtin-stained nuclei and aggregates are not detected in the brain of WT mice [12]. An histological approach to analyzing aggregates was chosen to allow for the separate analysis of several types of aggregates; in addition, we performed a regional analysis by examining two levels of striatum, and the level of striatum was included as a factor in the ANOVA analyses because it had a significant effect on the distribution of neuropil aggregates (F(1,20) = 9.74, p < 0.01), diffusely stained nuclei (F(1,20) = 25.37, p < 0.0001) and microaggregates (F(1,20) = 29.76, p < 0.0001), independent of treatment. Importantly, power calculations show that a group size of 2-3 mice are sufficient to detect a 30% change in the number of stained nuclei, or number of nuclei containing microaggregates (α = 0.05, 80% power), thus, our group sizes were well powered to detect treatment effects.

At 4.5 m of age, curcumin-fed KI mice had less diffusely stained nuclei, the earliest huntingtin-related pathology observed in mutant mice, than control-fed mice (effect of treatment F(1,20) = 5.02, p < 0.05; Figure 4A, B, C). Curcumin-treated KI mice also showed fewer microaggregates, defined as numerous, small, nuclear puncta, similar to those previously observed in other models of HD [44‐46] (effect of treatment F(1,20) = 12.73, p < 0.01; Figure 4A, B, C). Nuclear inclusions were reduced by curcumin but this effect did not reach significance (effect of treatment F(1,20) = 4.3, p = 0.052). By far the largest reduction observed in curcumin-treated mice (-17% in lateral sections, -45% in medial sections) was on neuropil (cytoplasmic) aggregates in the striatum (effect of treatment, F(1,20) = 5.22, p < 0.05, Figure 4A, B, C). These data indicate that curcumin improved a key pathological consequence of the HD-causing mutation in vivo.

Having established a profound effect on htt aggregation, we went on to assess whether transcription was also improved. Together with the formation of protein aggregates, transcriptional dysregulation is a major effect of mutant huntingtin [48]. We have previously detected profound changes in striatal transcripts for enkephalin but not substance P at 4 months of age in a similar line of KI mice with 94 CAG repeats (CAG94; [28]). We have also detected a significant decrease of D1, D2 dopamine and CB1 receptors, enkephalin and DARPP-32, but not substance P mRNA at 4 months of age in CAG140 KI mice [49]. Power calculations showed that group sizes of 2-7 (depending on transcript) are required to detect a 50% change in the number transcripts in the KI mice (α = 0.05, 80% power), thus our group sizes were well powered to detect treatment effects (n = 7 for each KI group). The housekeeping gene used was HPRT, as used previously [50‐52]. In a separate group of mice (n = 23), we found that HPRT showed excellent correlation to results obtained using other housekeeping genes, with r² = 0.862 and slope of 1.08 (Figure 5, p < 0.0001 [53]).

We confirmed striatal transcriptional dysregulation in the control-fed KI mice compared to WTs (Table 3, effect of genotype: Enkephalin F(1,26) = 78.1, p < 0.0001; D1 F(1,26) = 27.5, p < 0.0001; D2 F(1,26) = 29.6, p < 0.0001; DARPP-32 F(1,26) = 51.7, p < 0.0001; CB1 F(1,26) = 21.8, p < 0.0001; and post hoc tests using Bonferroni's multiple t-tests, corrected for 4 comparisons).

Curcumin treatment attenuated several deficits in KI mice, with D1 and CB1 mRNA no longer being different to control-fed WT mice, and the levels of DARPP-32 and D1 receptor mRNA being significantly increased compared to control-fed KIs (Table 3). No transcripts were significantly affected in WT mice (Table 3).

In addition to curcumin's effect on pathological and molecular changes induced by mutant huntingtin we also examined the effects of curcumin on the early behavioral deficits in KI mice. Analyses were carried out at 1 month (open field), 1.5 month (climbing) and 4 months of age (pole task and rotarod). Several behavioral deficits are typically present in these tests at these ages in the KI mice, indicative of the extensive neural dysfunction present at these early ages [11, 12, 49].

For mice treated with curcumin or control diets from conception, there was no effect of gender on any of the tasks examined (effect of sex: for open field first 5 mins F(1,85) = 3.6, ns; climbing: F(1,80) = 3.1, ns; pole task: F(1,77) = 0.003, ns; rotarod 10 rpm: F(1,81) = 2.8, ns, 20 rpm, F(1,81) = 0.21, 30 rpm F(1,81) = 0.92, ns; no interaction between genotype, treatment and sex: for open field: first 5 mins, F(1,85) = 2.1, ns); climbing: (F(1,80) = 1.4, ns); pole task: (F(1,77) = 2.6, ns) or rotarod (10 rpm: F(1,81) = 0.03, ns; 20 rpm: F(1,81) = 0.38, ns; 30 rpm, F(1,81) = 0.6, ns). Therefore, data from males and females were grouped for analysis.

Curcumin treatment rescued the reduced rearing in KI mice observed in the first five min in the open field (genotype × treatment F(1,89) = 4.36, p < 0.05, Figure 6). As previously shown [11], climbing was decreased at 1.5 months in KI mice (Table 4, effect of genotype F(1,84) = 4.6, p < 0.05; no overall interaction between genotype and treatment F(1,84) = 3.1, ns). Post hoc analysis showed that compared to control diet curcumin abrogated the difference between KI and WT curcumin-fed mice suggesting a beneficial effect of treatment. However, curcumin impaired climbing in WT mice (see Table 4). This suggests that the effect of curcumin on HD pathogenesis may be stronger than, and opposite to, the "off-target" effects in WT mice.

As previously shown [11], 4 month old control-fed KIs were impaired on the pole task (Table 4, effect of genotype F(1,81) = 10.1, p < 0.01). Curcumin-treated KIs were no longer significantly different from WT controls, indicating a small beneficial effect (Table 4). In agreement with other studies of rotarod performance of knock-in mice [55], we have previously shown that impairments on the rotarod in CAG140 KI mice are very subtle, with no impairments during accelerating protocols and small impairments during fixed speed protocols [11]. In the present study, control-treated KIs showed no defects in rotorod peformance, and actually performed slightly better than WTs on training days 3 and 4. However, curcumin impaired performance in both WTs and KIs throughout training. The curcumin-treated mice appeared to learn the task during training, since all mice improved over successive days, having started out with similar performances (see Figure 7, day 1). However, over time, both control-fed WTs and control-fed KIs showed improved performance above that of their genotype matched curcumin-fed counterparts (Figure 7, training: days 1 through 4, accelerating paradigm, smooth axle, effect of treatment F(1,86) = 9.2, p < 0.01, no interaction of genotype and treatment F(1,86) = 1, ns). The effect of treatment was also seen after the training phase, on day 5 with the smooth axle fixed speed paradigm (effect of treatment F(1,85) = 12.6, p < 0.001; for example at 20 rpm: WT control 139 ± 39 s, WT curcumin 37 ± 7, KI control, 182 ± 64, KI curcumin 34 ± 9, WT control versus WT curcumin p < 0.01, KI control versus KI curcumin p < 0.01, Fishers LSD post hoc, no interaction of genotype and treatment F(1,85) = 0.5, ns), and on day 6 with the grooved axle (effect of treatment F(1,85) = 16.3, p < 0.001; for example 30 rpm WT control 274 ± 47 s, WT curcumin 81 ± 26, KI control, 248 ± 68, KI curcumin 91 ± 40, WT control versus WT curcumin p < 0.01, KI control versus KI curcumin p < 0.01, Fishers LSD post hoc; no interaction of genotype and treatment F(1,85) = 0.6, ns; 1 WT control mouse died before days 5 and 6).

To determine whether the deleterious effects of curcumin in WT mice on climbing and rotarod performance were related to exposure to curcumin during early development, we conducted a separate trial of curcumin in normal C57Bl/6 J WT mice. Treatment began at 2.7 months and continued until 8 months of age, to match the total length (gestation and postnatal) of treatment used in the CAG140 trial (approximately 23 weeks).

As in the life-long curcumin trial, both male and female curcumin-fed WT mice were impaired in rotarod performance by endpoint (Figure 8A, 8 months, treatment × training session F(2,70) = 4.1, p < 0.03). We did not observe curcumin-induced impairments in climbing until 8 months of age. This deficit was only in males, however, climbing activity was reduced in all females at this age, possibly obscuring any treatment effects (Figure 8B, gender × treatment interaction at 8 months, F(1,35) = 6.2, p < 0.02). The reduction in rotarod performance was not due to a general effect of curcumin on movement, since there was no effect of curcumin on distance moved in the open field (Figure 8C, effect of day F(3,105) = 36.7, p < 0.0001, treatment × day interaction F(3,105) = 0.93, ns). Grip strength was actually increased by curcumin treatment and then normalized by the endpoint (Figure 8D). As expected, gender impacted muscle strength, thus, the analysis was conducted separately in males and females (Figure 8D, treatment × gender × time, F(2,70) = 4.3, p < 0.02). These data demonstrate that muscle strength was not impaired in curcumin-treated mice, and as such does not explain the impaired rotarod performance. Latency to immobility was unchanged between treatment groups in the forced swim task indicating that behavioral despair, also, was not the cause of the impaired rotarod performance (data were analyzed separately in males and females since gender impacted levels of immobility, latency to immobility: treatment × gender F(1,35) = 0.26, ns; data not shown).

Taken together, these data in adult WT mice demonstrated that fine motor coordination, muscle strength, depressive behavior and activity levels were not the cause of impaired rotarod or climbing performance induced by curcumin treatment. Nevertheless these deleterious effects are not unexpected, since CoQ10 is an anti-oxidant and component of the electron transport chain, also impairs rotarod performance in WT mice [56, 57].

PC12 cells inducibly expressing EGFP-tagged exon 1 of mutant htt were a kind gift from Dr Erik Schweitzer (UCLA [42]). Cells were cultured in DMEM containing 5% horse serum, 5% calf serum and 1% L-glutamine (Fisher Scientific, Pittsburgh, PA), 1% penicillin/streptomycin and 1% geneticin (Invitrogen, Grand Island, NY) in a humidified atmosphere containing 9.5% CO₂ at 37°C. Cells were cultured in collagen-coated T75s and plated for experiments onto poly-D-lysine-coated 96-well plates (20,000 per well, no cells were plated in the outermost wells; plates and flasks from BDBiosciences, San Jose CA). On the following day after plating, the cells were induced with 0.1 μM Tebufenozide, a kind gift from Dr Erik Schweitzer (UCLA), or treated with ethanol (vehicle) and also treated with curcumin (Sigma Aldrich, St. Louis, MO; 5 nM, 50 nM, 500 nM, 5 μM, 10 μM or 20 μM) or DMSO (Fisher Scientific; vehicle for curcumin) using dilutions of 1:1000 to prevent toxicity from the DMSO or ethanol. Four wells were used for each treatment (8 for control-treated induced cells), the positions of which were pseudorandomized between independent experiments (n = 4). Cells were cultured for 48 or 72 h and then the medium was removed, the cells were washed in warmed PBS and then fixed in cold 4% PFA for 30 mins. The cells were washed again and covered with fluorescent mounting medium and stored in the dark for analysis. For analysis of aggregates, photomicrographs, at 10× magnification, were taken and were analyzed using ImageJ. One photo, centered over the well, was taken per well. The mean size of all aggregates per field of view (FOV, i.e., 1 FOV per well) was calculated by ImageJ. This mean size was then expressed as a proportion of the mean aggregate size of all control-treated induced wells. These proportions were then used for the final quantification and statistical comparisons, such that each treatment group contained n = 4 replicates, each from an independent experiment.

Breeding mice (5 males, 10 females per group) were fed normal chow (NIH-31 number 7013, Harlan Teklad, Indianapolis, IN) or the same chow containing 555 ppm curcumin (Sigma Aldrich, St. Louis, MO) for at least 1 week prior to breeding and throughout breeding. This regimen was chosen based on previously published preclinical trials in mouse models of Alzheimer's disease showing drug effects in brain, and measurable curcumin in brain parenchyma following oral administration [20, 25, 26]. Breeding pairs were checked daily for litters, and the number of pups in each litter was noted at birth and at weaning. In some cases, the number of pups increased from the number at birth and this is due to incomplete parturition at the time that the litter was first counted. Two rounds of breeding were used to generate mice for the preclinical trial, and equal numbers of litters were generated in both treatment groups (n = 18 per treatment group). Groups were matched for gender, and group sizes were n = 30 - 32 for WTs, and n = 16-17 for KIs. HET mice (n = 15-19) were used only to monitor body weight and curcumin brain levels, and were not used for any behavioral testing. Progeny were genotyped and weaned by 21 days and curcumin or control chow was continuously fed to the mice until 4.5 months of age. Body weight was recorded weekly. Tail samples were taken at the end of the experiment in order to confirm genotyping and to measure CAG repeat length. The mean CAG repeat length alleles, measured by Laragen Inc., Culver City, CA, using an ABI 3730 sequencer and Genemapper software, were 117 ± 1 (n = 50) in control diet-treated, and 113 ± 1.4 (n = 46) (mean ± sem) in curcumin-treated KI mice (frequency distributions not significantly different, Chi square independence: ns).

To control for effects of developmental exposure to curcumin, an additional trial was conducted in 39 adult WT C57Bl/6 J (B6) mice purchased from Jackson laboratories (Bar Harbor, ME). Following 1 week habituation to housing, the mice were divided into weight- and gender-matched control-fed groups (N = 10 males, N = 10 females) and curcumin-fed groups (N = 10 males, N = 9 females). Mice were fed normal or curcumin chow, as above, from 2.7 months of age until 8 months. Prior to start of trial, mice were tested for climbing, grip strength and body weight, and groups were balanced using these data.

Food and water were available ad lib, and stocks of chow were stored at 4°C. Mice were housed in a temperature and humidity controlled room, on a reverse light-dark cycle (11 am lights off, 11 pm lights on). For food utilization for the trial in adult mice, chow was weighed twice weekly and the amount used per mouse per day calculated.

Measurements of curcumin in brain tissue were carried out as described previously [20]. Briefly, fresh frozen (WT, n = 2 control, n = 3 curcumin; mice were from the behavior trial) or PBS-perfused and subsequently fresh frozen (HET, n = 4 control, n = 5 curcumin; HETs were treated in parallel to mice in the behavioral trial) hemispheres of brain from mice treated with control- or curcumin-containing food were weighed, then finely powdered and homogenized in 10 vol of 1 M ammonium acetate (pH 4.6). Mice were euthanized at the end of the trial thus following approximately 5.5 m of treatment, which includes gestation, weaning and up to approximately 4.5 m of age. Homogenates were extracted in 95% ethyl acetate/5% methanol and dried under a continuous flow of N₂. Dried extracts were redissolved in acetonitrile/methanol/water/acetic acid (41/23/36/1, all by volume), and injected onto a reverse phase HPLC column (Supelco Ascentis Express C18, 150 × 2.1 mm) equilibrated in 10 mM ammonium acetate and eluted (100 μl/min) with an increasing concentration of acetonitrile/isopropanol. Samples were detected at 262 nm, using tetramethoxycurcumin as an internal standard. The effluent from the column was passed directly to an Ionspray™ ion source connected to a triple quadrupole mass spectrometer (PerkinElmer Life Sciences Sciex API III+). The retention times of curcumin and internal standard were 28.24 and 30.27 min, respectively.

A subset of curcumin- and control-fed CAG140 KI mice (n = 6 per group) were anesthetized and perfused with 4% paraformaldehyde and 0.5% glutaraldehyde, their brains removed, post-fixed for 6-8 h in 4% paraformaldehyde, cryoprotected in 30% sucrose and frozen for use. Sagittal cryosections at the level of 1.32 mm and 2.28 mm lateral of the midline [47] were used for analysis. Tissue cryosections (35 μm) were stained with polyclonal EM48 (EM48; X-J Li, Emory University) as described in [12]. Briefly, sections were washed in 0.01 M PBS and then endogenous peroxidases were inactivated by incubating in 1% H₂O₂ and 0.5% Triton X-100 in PBS, for 20 min. Non-specific binding sites were then blocked by incubating sections for 30 min at room temperature in PBS containing 3% bovine serum albumin (BSA) and 2% normal goat serum (NGS). The primary antibody, EM48, was diluted (1:300) in PBS containing 3% BSA, 2% NGS, 0.08% sodium azide, and 0.2% Triton X-100 and sections were incubated overnight at room temperature. The following day the sections were washed in PBS and then incubated in biotinylated goat anti-rabbit antibody (1:200; Vector ABC Elite; Vector, Burlingame, CA) for 2 h at room temperature, washed and then reacted with avidin-biotin complex (Vector ABC Elite) in PBS containing 0.2% Triton X-100 for 2 h. Immunoreactivity was visualized by incubation in 0.03% 3-3-diaminobenzidine tetrahydrochloride (Sigma, St. Louis, MO) and 0.0006% H₂O₂ in 0.05 M Tris buffer, pH 7.6. After rinses in Tris buffer, the sections were dehydrated, defatted, and mounted with Eukitt (Calibrated Instruments, Hawthorne, NY). Control sections, processed in parallel, were incubated in the absence of the primary or secondary antibodies. No staining was noted in control sections (data not shown). Huntingtin-stained nuclei and aggregates were analyzed with Stereo Investigator 5.00 software (Microbrightfield, Colchester, VT). Briefly, the contours of the striatum were drawn at 5× magnification. The software then laid down a sampling grid of 200 × 200 μm, on which counting frames of 20 × 20 μm were placed. Counting frames were located on the top left corner of each sampling grid, thus allowing for unbiased sampling, and these counting frames were used for quantification of each type of aggregate per section. Quantification was done at 100× magnification, using a 1.4 NA lens and 1.4 NA oil condenser, with a DVC real-time digital camera. The mean number of 1) stained nuclei, 2) nuclei containing microaggregates, 3) nuclei containing inclusions and 4) neuropil aggregates, per counting frame, was calculated per striatal section, per mouse. These data were then used to generate group means. Microaggregates [12, 74] were defined as numerous, small, nuclear puncta, similar to those previously observed in other models of HD [44‐46].

A subset of curcumin- and vehicle-treated CAG140 WT and KI mice (control-treated: WT n = 8, KI n = 7; curcumin-treated: WT n = 8, KI n = 7) were quickly decapitated and their brains frozen in powdered dry ice. Total RNA was purified from one striata of fresh frozen tissue using QiaGen RNeasy mini kit (protocol as described by manufacturer, QIAGEN Sciences, Maryland, USA). During the RNA extraction procedure DNAse 1 treatment was performed to remove contaminating genomic DNA (RNase-Free DNase Set, QIAGEN, Hilden, Germany). The Invitrogen ThermoScript RT-PCR System (Invitrogen, Carlsbad, CA, USA) was used for cDNA synthesis with oligo dT primers. The cDNA was then analyzed by quantitative real time PCR using a Roche LightCycler 480 (UCLA Genotyping and Sequencing Core). PCRs were performed using LightCycler FastStart DNA Master plus SYBR Green 1 kit (Roche Diagnostics, Mannheim, Germany). Each assay included: 1) a standard curve of five serial dilution points of control cDNA (mouse EST clone of appropriate fragment of the gene of interest (Invitrogen, CA), 2) sample cDNA, 3) no template control. All samples were run in triplicates. The PCR cycling parameters were: 95°C for 5 min (1 cyc); 95°C for 10 sec, 65°C for 10 sec, 72°C for 10 sec (40 cyc). A dissociation protocol was established at the end of each run to verify the presence of a single product. The relative expression of genes of interest was calculated from Ct values using the Pfaffl method [54]. PCR efficiencies of each primer pair were calculated from standard curve analysis and incorporated into relative quantification calculations. The endogenous control was HPRT, hypoxanthine phosphoribosyltransferase, which was reported to be unchanged in mouse and human microarray studies [50, 75] and has been used in other previous RT-PCR studies examining transcript changes in HD [50‐52]. Designed primers yielded a product of about 200 bp for each gene (Table 5).

All analysis was carried out blinded to genotype and treatment. Mice were habituated to the testing rooms for 15-20 mins prior to all testing. Body weight was monitored in both trials and the amount of chow utilized was also quantified in the adult trial. Open field, pole task, and rotarod testing all took place in the dark phase, while climbing activity was recorded during the light phase [11, 12]. For the life-time curcumin trial, CAG140 KI and WT mice were tested at 1 m of age in the open field, 1.5 months in the climbing test, and at 4 months in the pole test and the rotarod [11, 12]. This schedule was chosen to avoid excessive repeated testing that could influence the progression of the disease. For the trial in adult WT mice, mice were tested 4.5 and 8 months of age in the climbing test and the rotarod, and at 8 months of age in the open field (for 15 mins daily over a period of 4 d). In addition, WT mice in the adult trial were also tested for grip strength, a measurement of forelimb muscle strength as described previously [76] at 3.5, 4.5 and 8 months of age. Finally, in this trial, WT mice were tested using the Porsolt swim task at 8 months [77]. They were also tested using the tail suspension task [78] however, several mice "escaped" the tail suspension task by climbing up their tail ([79] n = 2 male control, n = 5 female control, n = 1 male curcumin, n = 3 female curcumin), thus forced swim test data only were used.

For rearing activity mice were placed individually in the center of the open field (Truscan, Coulbourn Instruments, Allentown, PA) and video recorded for later analysis. Mice were tested in 2nd through 4th hours of the dark cycle, using a red light (25 W) for illumination. For climbing, mice were placed in a wire cylinder (3 ¾″ height × 4″ diameter) for 5 mins and their behavior recorded for later video analysis, as described previously [11]. For the pole task mice were placed head facing upwards on a vertical pole and trained to turn around and descend to the bottom of the pole, as described previously [11]. For rotarod performance mice were trained to walk on an accelerating rotating axle (Ugo Basile, Varese, Italy), as previously described with slight modifications [11]. Briefly, mice were given 3 trials/day over 4 days (5-40 rpm over 10 mins, with approximately 30 mins between successive trials) and the latency to fall was measured. For training the axle of the rotarod was covered with smooth rubber (smooth axle). On the 5th day mice received 1 trial at each of 10, 20 and 30 rpm (smooth axle) and on the final day mice received 1 trial at each of 20, 24, 30 and 36 rpm using a grooved axle. Testing was carried out approximately half way through the dark phase, under a red light (25 W). Any mice that clung to the axle for 3 successive rotations were removed and the time of removal recorded and used as the latency. The proportion of mice that cling is very small, thus we did not analyze these animals separately [11]. For the Porsolt swim task (adult trial only), mice were placed in 12 cm of water (temperature: 26°C) in a large Plexiglas beaker for 6 mins and their behavior was videotaped for analysis. Latency to become immobile and duration of immobility were quantified.

GBstat (V8.0) and SAS (V 9.1) were used for statistical analyses. Comparisons of in vitro aggregate size proportions were completed using Kruskal-Wallis ANOVAs followed by Mann Whitney U tests. For in vivo data, outliers were detected using Grubbs test and removed from analyses. Measures of huntingtin pathology (immunostained nuclei and aggregates) were analyzed with completely randomized ANOVAs followed by Fisher's LSD for post hoc analysis. For qRT-PCR results, data were analyzed with completely randomized ANOVAs followed by Bonferroni t tests, corrected for 4 comparisons (control-fed WT V control-fed KI; control-fed WT V curcumin-fed WT; control-fed WT V curcumin-fed KI; control-fed KI V curcumin-fed KI). Correlations of mRNA data were carried out in GraphPad prism V4. Comparisons of husbandry data were made with Student's t tests or repeated measures ANOVA followed by Fisher's LSD post hoc tests. Body weights were analyzed using mixed generalized linear model ANOVAs in SAS using Bonferroni's adjusted Student's t-tests for post hoc analysis. Behavioral data and food intake were analyzed using one, two or three factor ANOVAs followed by Fisher's LSD for post hoc analysis. Significance was defined as α = 0.05 for all analyses.