A comparative study of the performance of individuals with fragile X syndrome and Fmr1 knockout mice on Hebb-Williams mazes

Authors


C. S. Kogan, School of Psychology, University of Ottawa, Ottawa, Ontario, Canada K1N 6N5. E-mail: ckogan@uottawa.ca

Abstract

Fragile X syndrome (FXS) is the most prevalent form of heritable mental retardation. It arises from a mutation in the FMR1 gene on the X chromosome that interferes with expression of fragile X mental retardation protein (FMRP) and leads to a wide range of behavioural and cognitive deficits. Previous studies have shown a deficit in basic visual perceptual processing as well as spatial abilities in FXS. How such a deficit may impact spatial navigation remains unknown. The current study extended previous research by evaluating spatial learning and memory using both virtual and physical versions of Hebb-Williams mazes, which allows for testing of humans and animals under comparable conditions. We compared the performance of individuals affected by FXS to typically developing individuals of equivalent mental age as well as the performance of Fmr1 knockout mice to wild-type control mice on the same maze problems. In human participants, performance of the comparison group improved across trials, showing expected significant decreases in both errors and latency. In contrast, the performance of the fragile X group remained at similar levels across trials. Although wild-type control mice made significantly fewer errors than the Fmr1 knockout mice, latencies were not statistically different between the groups. These findings suggest that affected humans and mice show similar spatial learning deficits attributable to the lack of FMRP. The implications of these data are discussed including the notion that Hebb-Williams mazes may represent a useful tool to examine the impact of pharmacological interventions on mitigating or reversing the symptoms associated with FXS.

Fragile X syndrome (FXS) is the most prevalent form of heritable mental retardation (Turner et al. 1996). It arises from a trinucleotide expansion of the fragile X mental retardation 1 gene (FMR1) (Online Mendelian Inheritance in Man® [OMIM] 309550, http://www.ncbi.nlm.nih.gov/omim; Verkerk et al. 1991). In FXS, FMR1 is not expressed in somatic cells (Pieretti et al. 1991), which leads to a wide range of behavioural and cognitive deficits, including deficits in attention (Backes et al. 2002; Baumgardner et al. 1995), visual-spatial cognition (Cornish et al. 1998, 1999; Crowe & Hay 1990), working memory (Jäkäläet al. 1997; Schapiro et al. 1995) and visual perceptual processing (Kogan et al. 2004b). Studies of Fmr1 knockout (KO) mice, a murine model of FXS, have also shown deficits that at times mirror those observed in individuals affected by FXS. Impairments have been described for a range of behavioural tasks including conditioned fear response (Paradee et al. 1999), prepulse inhibition (Chen & Toth 2001), open field activity (Mineur et al. 2002) and social interaction tasks (Mineur et al. 2006; Spencer et al. 2005).

Despite some concordance, results from tests of the murine model of FXS focusing on replicating the spatial processing deficit observed in affected humans have been mixed (D’Hooge et al. 1997; Dobkin et al. 2000; Kooy et al. 1996; Mineur et al. 2002; Paradee et al. 1999; Peier et al. 2000; The Dutch-Belgian Fragile X Consortium 1994). These discrepancies may arise from differences in background strain used (Spencer et al. 2006), number of generations of back crossing (Gu et al. 2002) and possibly the choice of the spatial task employed as well as the equivalency of such tasks to those of human spatial cognition. It therefore remains controversial as to whether the murine model of FXS exhibits a similar spatial deficit as has been documented for affected individuals.

In an effort to resolve the ambiguity in the literature, in the present study we employ Hebb-Williams (H-W) mazes, which are well-established measures of spatial cognition (Hebb & Williams 1946; Rabinovitch & Rosvold 1951) and are reported to be sensitive to detecting alterations of hippocampal-dependent spatial abilities (Winocur & Moscovitch 1990) better than radial or water maze tasks (Pereira et al. 2005). Hebb-Williams mazes have traditionally been used to test spatial learning in animals (Shore et al. 2001). However, more recently, a computerized version of these mazes has been designed to allow researchers to evaluate spatial learning in humans under comparable testing conditions (Shore et al. 2001), a criterion not met in previous studies. The development of this novel experimental paradigm has two distinct advantages. First, this paradigm allows for a comparison across species to directly evaluate the performance of individuals affected by FXS to Fmr1 KO mice. Second, establishing that behavioural assays, such as the H-W maze, are able to detect the behavioural deficits in both human and mice is advantageous, because they can be used to evaluate the pharmacological and behavioural interventions to reverse or mitigate the symptoms of FXS. Finally, findings of comparable deficits among affected individuals and Fmr1 KO mice on a spatial navigation task may provide further insight into the neurobiological basis of the FXS phenotype.

Spatial navigation learning and memory in general, and performance on the H-W mazes specifically, appears to be dependent on both intact basic visual processing (Tees et al. 1981) as well as hippocampal and parietal cortex functioning (Hunsaker et al. 2008; Rogers & Kesner 2006).

With respect to basic visual functioning that may impact spatial abilities in FXS, Kogan et al. (2004a,b) provided neurobiological and behavioural evidence that individuals with FXS have an early visual processing deficit that impact dorsal stream functioning. In primates, visual information is processed through the dorsal and ventral visual streams (Milner & Goodale 1995; Ungerleider & Mishkin 1982). The posterior parietal cortex is an area that receives primary input from the dorsal visual pathway and is important for spatial processing to guide behaviour (Milner & Goodale 1995) such as egocentric spatial navigational tasks (Andersen & Buneo 2002; Hyvarinen & Poranen 1974; Spiers & Maguire 2007). We hypothesize that impairments in H-W performance indicative of abnormal basic visual functioning would manifest as significantly greater latencies and error rates across all maze problems for affected humans and KO mice for the first trial with persisting deficits observed on subsequent trials.

Spatial navigation is also dependent on intact hippocampal functioning (Ekstrom et al. 2003; Ghaem et al. 1997; Iaria et al. 2003; Morris et al. 1982; O’Keefe & Dostrovsky 1971). Lesion studies in mice have shown that successful performance on paradigms such as the radial arm maze, T maze and water maze rely on intact hippocampal processing (Hock & Bunsey 1998; Mitchell et al. 1982; Morris et al. 1982; Okada & Okaichi 2009). In both typically developing humans and wild-type mice, high levels of FMR1/Fmr1 mRNAs are expressed in the hippocampus (Abitbol et al. 1993; Hinds et al. 1993), suggesting that this is brain area is particularly reliant on and therefore sensitive to changes in fragile X mental retardation protein (FMRP) expression as occur in FXS. Furthermore, structural magnetic resonance imaging (MRI) studies suggest that individuals affected by FXS have enlarged hippocampi (Kates et al. 1997; Reiss et al. 1994). Fmr1 KO mice also exhibit hippocampal abnormalities and are found to have longer dendritic spines in pyramidal cells in subfield CA1 (Grossman et al. 2006), smaller intra-infra pyramidal mossy fibre terminal fields (Mineur et al. 2002), as well as shorter dendrites, fewer dendritic spines and functional synaptic connections (Braun & Segal 2000). Therefore, we hypothesize that abnormal hippocampal functioning alongside intact basic visual functioning in individuals affected by FXS and Fmr1 KO mice will result in normal latencies and errors for the first trial of each new maze problem, and significantly elevated levels of these same dependent variables on subsequent trials.

In the present study, we compared the performance of individuals affected by FXS to typically developing mental age-matched comparison participants on seven H-W mazes of increasing difficulty levels. We also compared the performance of Fmr1 KO mice to wild-type control mice on the same maze problems. We hypothesized that individuals affected by FXS and KO mice as compared with their respective control groups would exhibit poorer performance on mazes deemed more difficult.

Materials and methods

Experiment 1: comparing Fmr1 KO and control mouse performance

Subjects

Twelve male FVB.129P2-Pde6b+Tyrc-ch/AntJ mice (JAX Stock # 004828) and 11 male FVB.129P2-Fmr1tm1Cgr/J mice (JAX Stock # 004624) were obtained from a colony at Jackson Laboratories (Bar Harbor, Maine, USA). Each strain had been backcrossed for 11 generations. Mice were pigmented and did not carry the rd1 mutation, indicating that they do not suffer from blindness because of retinal degeneration. Animals were shipped at 4 weeks of age and were tested when they were approximately 12 weeks old. Eight days prior to behavioural testing, all subjects were individually housed in a climate-controlled vivarium (20–22°C) that was maintained on a 12-h light-dark cycle with lights on from 0700 to 1900. All testing was conducted during the light phase of the cycle. Mice were fed Harlan Global Rodent Chow and tap water. To ensure that mice were motivated during testing, they were maintained at 85–90% of their ad lib body weight. Mice were weighed daily and fed their individually weighed ration of food 30 min after completion of the session. The mice were treated in accordance with the guidelines and principles set by the Canadian Council on Animal Care and tested under the protocol approved by the University of Ottawa Animal Care Committee.

Apparatus

Mice were tested using the H-W maze apparatus as described by Meunier et al. (1986). The maze was constructed using black opaque Plexiglas and was covered with a clear Plexiglas top (Plastics of Ottawa Ltd, Ottawa, Canada). It consisted of a square open field (60 × 60 × 10 cm) with start and goal box compartments (20 × 10 × 10 cm) located at diagonally opposite corners. These compartments were fitted with clear Plexiglas lids that were attached with hinges and could be blocked with removable clear Plexiglas barriers. The goal box was fitted with a ledge (8 × 2.5 cm) with a recessed food cup in the centre (2.5 cm diameter). The floor of the maze was divided into 36-equal squares that were clearly outlined in white. The squares were used as markers for placing the barriers in different maze configurations as described by Rabinovitch and Rosvold (1951) and to define error zones. Removable barriers (10 cm high) were created using black opaque Plexiglas and each were supported by two permanent bases (2.5 × 2.5 cm). Extramaze cues were minimized by conducting the study in an all-black enclosure and by having a dim light as the only source of illumination.

Procedure

The protocol consisted of three consecutive phases: habituation, acquisition and testing. Initially, mice were habituated to the maze environment for 20 min/day on 4 consecutive days with barriers and doors to the start and goal box removed. During the last two sessions, the goal box was baited (Rodent Chow, 100 mg) and mice had ad lib access to the food for the duration of the session. Subsequently, mice were trained on six acquisition mazes (Fig. 1a) as described by Rabinovitch and Rosvold (1951). On any given day, mice were tested such that they completed five trials for each of two of the six acquisition mazes. Mice completed all six acquisition mazes in sequence as many times as necessary for them to attain the criterion performance, which was defined as two consecutive sessions completed successfully in less than 30 seconds each. The acquisition phase required an average of 10 days to complete. On each acquisition trials, mice received a small reinforcer (Rodent Chow, 20 mg). Immediately following acquisition, mice were given a selection of the standard test mazes (Rabinovitch & Rosvold 1951) according to the same training protocol used during acquisition sessions, which was conducted over 4 days. None of the acquisition or testing sessions exceeded 180 seconds. Both human participants and mice completed the same Rabinovitch and Rosvold maze configurations and in the same order (i.e. #2, #4, #5, #8, #9, #11 and #12). Latency and number of errors were recorded. Latency was recorded from the moment the barrier at the start box was removed until the animal took its first bite of food. An error was scored each time the animal's two front paws crossed into an error zone (Fig. 1b). Experimenters were blind to the animal genotypes and were never visible to the mice during the runs. The maze was thoroughly cleaned between trials.

Figure 1.

Maze configurations. (a) Testing was conducted using the six practice mazes (A-F) and (b) the seven test mazes depicted. For each maze configuration, the (S) depicted in the bottom right hand corner represents the start box and the (F) in the top left corner represents the goal box. Error zones are depicted by the dotted lines.

Experiment 2: comparing FXS human and comparison performance

Participants

Fifteen male participants with FXS (mean chronological age = 24 years, SD = 4.9, mean verbal mental age = 7.57 years, SD = 1.92) were recruited from patient contact lists at Rush University Medical Center (Chicago, IL, USA). All had a DNA-confirmed diagnosis of FXS and were full mutation carriers except for three participants who had a mosaicism of the full mutation. Seventeen male typically developing comparison participants (mean chronological age = 5 years, SD = 2.9, mean verbal mental age = 8.07 years,SD = 2.03) were individually matched according to the verbal mental age (see Measures, Receptive Language Assessment) of the FXS participants. These individuals were recruited through a collaborator's contact list and newspaper advertising. Informed consent was obtained from caregivers of each participant. Furthermore, assent was obtained from both control children and FXS individuals. All participants were paid $25 per h for their participation in the study and were treated in accordance to the ethical principals established by the Research Ethics Board at the University of Ottawa. Both the ethics committees of the School of Psychology, University of Ottawa and of the Rush University Medical Center approved the study.

Measures

Receptive language assessment. Individual matching of FXS and comparison participants on verbal mental age were accomplished using the Peabody Picture Vocabulary Test Third Edition (PPVT-R, Dunn & Dunn 1981). The PPVT is a measure of receptive vocabulary for standard English and assesses the extent of a participant's receptive verbal ability. The PPVT is a norm-referenced, individually administered test that consists of 175 vocabulary items arranged in increasing difficulty. The participant must select the image considered to best illustrate the meaning of a word presented orally by the examiner from a group of four images. Testing is terminated once a participant makes 8 errors within a block of 12 trials. Raw scores were used to estimate the verbal mental age according to norms provided by the test publisher (Dunn & Dunn 1981). There were no a priori differences between the two group scores (t = 0.468,P = 0.65), indicating a successful matching on receptive verbal ability.

Questionnaires. A brief Medical History Questionnaire was administered to all caregivers of participants to screen for any problems that would exclude them from the study. Exclusion criteria were any significant health or vision difficulties (e.g. colour blindness, amblyopia, astigmatism) that would impact controlling a joystick or viewing the maze stimuli.

Apparatus

Hebb-Williams virtual maze. All participants were tested using a version of the virtual H-W maze designed by Shore et al. (2001). Five mazes were eliminated from the original H-W set for the purpose of this project, because our pilot studies indicated that participants found these mazes too easy. To reduce administration time, only the most challenging mazes were used. All FXS and comparison participants were tested on the remaining subset of mazes, presented in the same order (i.e. #2,#4, #5, #8, #9, #11 and #12).

Experiments were performed on an Asus PC with a 19-in. Acer LCD monitor. Mazes were displayed at a resolution of 640 × 480 in full-screen mode. Participants navigated through the virtual environment at a constant velocity of 12 km/h (forward, backward) and a turn rate of 50 degrees/second (left, right) using a Logitech Attack 3 joystick. Assuming a viewing height of 5 ft 6 in., the projection of the whole maze appeared to participants to measure 20 m2, and the diagonal straight line from start to finish was perceived as being located at a distance of 28.3 m.

Each maze was made up of a 6 × 6 room, with a 1 × 1 alcove at the entrance (start area) and exit (goal area) of the maze. Walls were created using textured rectangles that differed in colour depending on the maze configuration. A different colour was used for each maze configuration to indicate to participants that a new trial within the same maze had started (as opposed to starting a new maze). The start alcove and the floors were textured with black and grey marble effect. Each wall of the goal alcove was white and contained the image of a comic book character to provide motivation and reward for the participants. The roof was textured using beige and brown mottled square tiles (Fig. 2).

Figure 2.

Virtual Hebb-Williams maze. Interior view of maze #12.

Procedure

All participants were individually tested by a research assistant, in a quiet room without their caregivers present. The tasks were administered during a 1-1.5-h session and presented in a standardized order as described under apparatus above. After completing PPVT assessment, participants were trained on two types of practice mazes. An alley maze was presented first and enabled participants to establish how to adaptively maneuver through the virtual environment, while maintaining direct visual contact with the goal area. After meeting this criterion, a T maze was presented in which participants had to choose a virtual navigational pathway in order to practice searching for the goal area of the maze. Criterion was achieved in both acquisition mazes when participants could complete three consecutive maze trials in less than 30 seconds each. At any time if a participant exceeded 120 seconds during a trial, the trial was considered finished and the participant proceeded to the next maze. For both the acquisition and testing mazes, participants received a sticker as reward after each trial, and after completing all three trials of maze they received a small piece of candy to be saved and consumed after the experiment was terminated.

After the acquisition sessions, participants completed three trials of each test maze (Fig. 1b). In between test for each maze, participants were provided with a 2-min break, at which time a children's DVD was played. After completing the fourth maze (#8), all participants were given a 10-min break. Dependent variables were the participants' latency for solving the maze (i.e. time taken from the maze entrance to exit) and number of errors, measured as the number of times a participant crossed a predefined error line (Fig. 1b), suggesting they were heading toward a blind alley rather than the goal box.

Statistical analyses

The number of errors made across maze trials for human and Animal studies was analysed using a mixed-design analysis of variance (anova) with group as between-subjects variable and both maze and trial as within-subjects variables. Latency for completion of maze trials was analysed in the same manner. Prior to the analyses, error and latency variables were transformed to square root scores in order to normalize the distribution of the data. Previous studies suggest that Fmr1 KO mice may exhibit increased activity levels as compared to wild-type mice (e.g. Mineur et al. 2002). Thus, we assessed activity levels by obtaining a count of the number of line crosses per unit of time for trial 1 of maze 12. The latter maze was chosen because it has the least number of partitions, thereby allowing for the clearest observation of locomotion. We restricted our analysis to trial 1 because performance on this trial is independent of learning and memory and reflects exploratory behaviour. All statistical analyses were conducted using PASW Statistics 17.0 (SPSS Inc., Chicago, USA).

Results

Human participants

Latency

A 2 × 7 × 3 anova was conducted with group (FXS, comparison) as independent measures variable and both maze (seven levels) and trial (three levels) as repeated measures variable. The latter showed significant main effects for group (F1,30 = 12.235,P = 0.001), maze (F6,30 = 12.487,P < 0.001) and trial (F2,30 = 14.749,P < 0.001). Furthermore, the interaction between group and trial was significant (F2,60 = 10.933,P < 0.0001) (Fig. 3a). Alpha levels for post hoc analyses of the interaction were adjusted using a Bonferroni correction to control for familywise error rate across trials (α = 0.05/9 = 0.006).Thus, independent samples t-tests showed significant differences between groups for latency for all trials (trial 1 : t30 = 2.78, P< 0.006; trial 2: t30 = −4.512, P = 0.001 and trial 3: t30 = −7.331, P< 0.001). Whereas for trial 1 the mean latency for the FXS group was significantly shorter than for the comparison participants, for subsequent trials, the comparison group completed the mazes significantly faster. We also examined the differences within each group to assess for significant improvements across trials. For the comparison group, post hoc paired samples t-tests showed significant decreases in latency to reach the goal box between trial 1 and trial 2 (t16 = 3.578,P < 0.001), trial 1 and trial 3 (t16 = 4.171,P < 0.001). The difference in latency between groups for trial 2 and trial 3 approached significance (t16 = 2.674,P = 0.017). In contrast, in the FXS group there were no significant differences in latency between any of the trials (trial 1 vs. trial 2: t14 = 0.423,P = 0.679; trial 1 vs. trial 3: t14 = 0.409,P = 0.689 and trial 2 vs. trial 3: t14 = 0.086,P = 0.933).

Figure 3.

Human data. (a) Human latency. Mean human latency measured in seconds for each of the three maze trials. Performance of comparison and FXS individuals on each of the three trials collapsed across the seven mazes. Error bars represent the standard error of the mean. (b) Human error. Mean human error for each of the three maze trials. Performance of comparison and FXS individuals on each of the three trials collapsed across the seven mazes. Error bars represent the standard error of the mean.

Error

A 2 × 7 × 3 anova was conducted with group (FXS, comparison) as independent measures variable and both maze (seven levels) and trial (three levels) as repeated measures variables. The latter showed main effects for group (F1,30 = 7.398,P = 0.011), maze (F6,30 = 23.376,P < 0.001) and trial (F2,30 = 37.036,P < 0.001). Furthermore, the interactions between group and trial (F2,56 = 10.435,P < 0.001) (Fig. 3b) as well as maze and trial (F12,336 = 2.138,P = 0.014) were significant. Alpha levels for post hoc analyses for the interaction were adjusted using a Bonferroni correction to control for familywise error rate across trials (α = 0.05/9 = 0.006). Thus, independent samples t-tests showed significant differences between groups for errors only for trials 2 and 3 (trial 1: t28 = 1.257,P = 0.210; trial 2: t28 = −4.018,P < 0.001 and trial 3: t28 = −4.556,P < 0.001). We also examined the differences within each group to assess for significant reductions in errors across trials. For the comparison group, post hoc paired samples t-tests showed significant decreases in the number of errors made during completion of trial 1 and trial 2 (t16 = 6.059,P < 0.0001), trial 1 and trial 3 (t16 = 5.591,P < 0.0001), but no significant difference in errors between trial 2 and trial 3 (t16 = 0.741,P = 0.470). In contrast, for the FXS group there were no significant differences in errors between any of the trials (trial 1 vs. trial 2: t12 = 1.656,P = 0.124; trial 1 vs. trial 3: t12 = 1.492,P = 0.162 and trial 2 vs. trial 3: t12 = 0.526,P = 0.609).

Mice

Latency

A 2 × 7 × 5 anova was conducted with group (Fmr1 KO, control) as the independent measures variable and both maze (seven levels) and trial (five levels) as repeated measures variables. The latter showed significant main effects for maze (F6,126 = 3.677,P = 0.002) and trial (F4,84 = 26.490,P < 0.0001) but not for group (F1,21 = 0.107,P = 0.746), indicating that the latency to complete the mazes did not differ significantly between the KO and control group. However, a significant three-way interaction between maze, group and trial (F = 1.633,P = 0.03) was found (Fig. 4). Post hoc independent samples t-tests showed only one large significant pair-wise difference attributable to longer latencies for the KO mice on trial 1 of maze 11 (t21 = −2.09,P = 0.049). However, when the α level was adjusted using a Bonferroni correction to control for familywise error rate (α = 0.05/35 = 0.0014), this difference was no longer significant.

Figure 4.

Mouse latency. Performance of wild-type control and Fmr1 KO mice for each of the seven mazes on each of the five respective trials. Error bars represent the standard error of the mean.

Error

A 2 × 7 × 5 anova was conducted with group (Fmr1 KO, control) as the independent measures variable and both maze (seven levels) and trial (five levels) as repeated measures variables. The latter showed significant main effects for group (F1,21 = 11.088,P = 0.003), maze (F6,126 = 8.189,P < 0.001) and trial (F4,84 = 16.620,P < 0.0001). These results indicate that on average Fmr1 KO mice made more errors than control mice. Although no significant interaction between group and trial was found (F4,84 = 1.458,P = 0.222), a significant interaction between maze and trial was found (F24,504 = 2.066,P = 0.002). Finally, a significant three-way interaction was found for maze, trial and group (F24,1008 = 2.140,P = 0.001) (Fig. 5). Post hoc independent samples t-tests showed only one significant pair-wise difference attributable to significantly greater errors committed by the KO mice on trial 1 of maze 11 (t21 = −2.33,P = 0.03). However, when the α level was adjusted using a Bonferroni correction to control for familywise error rate (α = 0.05/35 = 0.0014), this difference was no longer significant.

Figure 5.

Mouse error. Performance of wild-type control and Fmr1 KO mice for each of the seven mazes on each of the five respective trials. Error bars represent the standard error of the mean.

Activity levels

An independent samples t-test showed no significant differences in activity levels between the Fmr1 KO and control group (t = 0.661,P = 0.516).

Maze difficulty

The seven H-W mazes presented to participants and mice were selected because of reported differences in their level of difficulty (Shore et al. 2001). As described earlier, whereas a significant main effect of maze was found for human and mouse latency as well as mouse error, a significant interaction between maze and trial was found for human error. Thus, contrary to our hypothesis, neither the FXS group nor the Fmr1 KO group performed any worse than the respective comparison groups for mazes deemed more difficult. The main effect of maze suggests there were differences in latencies and number of errors dependent on the particular maze completed (Fig. 6a,b). Post hoc independent samples t-tests analyses were conducted (Table 1), with α adjusted using a Bonferroni correction (α = 0.05/21 = 0.002). For human participants, the pattern of latency results suggests a trend in terms of difficulty, with mazes 5, 9 and 12 tending to have longer latencies than mazes 2, 4 and 11 (Fig. 6a). Mice took significantly longer to complete maze 9. However, the pair-wise differences in latency for the remaining mazes were non-significant. In contrast, the pattern of error results for human participants indicates that participants made fewer errors on mazes 2, 4, 8 and 12 than mazes 9 and 11, with maze 5 resulting in significantly more errors than for all other mazes. This pattern is generally consistent with the mouse error data, with the exception of maze 2, for which mice made significantly more errors than on mazes 4 and 12.

Figure 6.

Maze difficulty. (a) Latencies for human and mouse performance across the seven mazes collapsed across groups. (b) Number of errors for human and mouse performance across the seven mazes collapsed across groups. Note: for ease of comparison, mazes are presented according to the human data in ascending order of difficulty.

Table 1.  Maze difficulty
HumanMaze #122845911
Maze #
  1. P-values for pair-wise post hoc analyses of latency and error data from mice and humans (α = 0.002, significant results are presented in bold). Whereas latency data is presented in the lower triangular matrices, error data is presented in the upper triangular matrices.

12  0.0310.0030.1220.0020.0020.002
2 < 0.002 0.1990.7930.0020.0020.018
8 0.0510.001 0.2530.0020.0010.141
4 0.0020.2340.057 0.0020.0020.021
5 0.7490.0020.1030.001 0.0020.002
9 0.9310.0020.0310.0020.674 0.341
11 0.0010.4150.0580.8200.0030.002 
MiceMaze #122845911
Maze #
12  0.0020.0020.3040.0020.0020.020
2 0.014 0.5210.0020.0180.2000.955
8 0.0490.595 0.0020.0040.0600648
4 0.0140.6930.399 0.0020.0020.006
5 0.0460.6860.9150.467 0.2350.096
9 0.0860.0020.0020.0020.002 0.421
11 0.0050.4770.2420.7880.2980.002 

Discussion

The present study examined the spatial navigation abilities of individuals affected by FXS and a mouse model of FXS, Fmr1 KO mice, on a similar task. The results show significant differences in performance for both groups as compared with mental age-matched comparison individuals and wild-type mice, respectively. In contrast to the FXS group, performance of the comparison group improved as indicated by significantly fewer errors across trials. A similar pattern of results was observed when latency across trials was analysed. Although significant differences between groups were found for all trials, on trial 1 the FXS group completed mazes faster than their peers while committing a similar number of errors. That participants affected by FXS were able to successfully complete the mazes suggests that basic visual functioning necessary for solving a novel spatial task is intact in these individuals. However, with subsequent trials, the FXS group appeared either unable to learn or recall the maze solution and continued to use a trial and error strategy to find the goal box.

Contrary to our hypothesis, neither the FXS group nor the Fmr1 KO group performed any worse than the respective comparison groups for more difficult mazes. However, there were differences in latencies and number of errors dependent on the particular maze that either humans or mice completed (Fig. 6a,b). Consistent with the results from Shore et al. (2001), the pattern of maze performance across species was similar, with the exception of the number of errors committed by mice on maze 2 (Fig. 6b). The latter may be explained by the aversion mice have to open spaces evident in the configuration of maze 2. Finally, human participants tended to take longer to complete the mazes, which may be related to the differences in the presentation of the H-W mazes, i.e. real vs. virtual.

Our findings are consistent with previous research suggesting impairment of spatial processing in FXS. Individuals affected by FXS have reliably been shown to experience difficulties on visual-motor tasks including those requiring participants to manipulate objects in space, an ability similar to the one used to navigate successfully through virtual H-W mazes (Cornish et al. 1998, 1999; Crowe & Hay 1990; Freund & Reiss 1991; Mazzacco et al. 2006). Kogan et al. (2004b) provided evidence for underlying impairments in neural functioning within the thalamic magnocellular (M) pathway and dorsal visual stream integral to the visual control of action (Milner & Goodale 1995). Therefore, we hypothesized that individuals affected by FXS would perform worse on the virtual H-W mazes because spatial navigation relies in part on intact vision (Tees et al. 1981). The findings of the current study argue against this hypothesis. In fact, the FXS group was able to complete the first trial of the mazes faster than comparison participants making similar number of errors (Fig. 3).

Alternatively, hippocampal functioning, also required for spatial navigation, may explain our findings (Ekstrom et al. 2003; Ghaem et al. 1997; Iaria et al. 2003; Morris et al. 1982; O’Keefe & Dostrovsky 1971). In typically developing individuals, high levels of FMR1 mRNAs are expressed in the hippocampus (Abitbol et al. 1993; Hinds et al. 1993), suggesting that this brain region is reliant on and therefore sensitive to changes in FMRP expression. Indeed, functional impairments in hippocampal circuitry have been shown in Fmr1 KO mice, such as enhanced long-term depression (LTD), a form of synaptic plasticity (Huber et al. 2000, 2002; Zhang et al. 2009). Therefore, it is possible that similar impairments are present in affected individuals. In support of this notion, structural MRI studies suggest that affected individuals have enlarged hippocampi (Kates et al. 1997; Reiss et al. 1994). Behaviourally, individuals affected by FXS exhibit impaired performance on spatial tasks (Cornish et al. 1998, 1999; Crowe and Hay 1990; Freund & Reiss 1991; Mazzacco et al. 2006). We speculate that impaired performance of affected individuals on the virtual H-W mazes is attributable to hippocampal deficits, which have been shown to affect spatial learning (Logue et al. 1997; Morris et al. 1982).

Interestingly, Fmr1 KO mice exhibit impairments similar to those observed in the FXS group for the same H-W mazes. Overall, wild-type control mice made significantly fewer errors than the Fmr1 KO group. However, the latency to complete the mazes for both groups was not significantly different. These findings suggest that for the Fmr1 KO but not the wild-type mice there is a speed-accuracy trade-off in the navigational strategies employed. Pollard and Lysons (1969) suggest that learning and memory may be best accounted for by measurements of error, whereas measurements based on time may better reflect non-problem solving behaviours, such as exploration and/or motivational factors. Similar to the human data, increased errors observed in Fmr1 KO mouse performance is more accurately a reflection of a learning or memory deficit, rather than non-specific factors or visual perceptual ability per se.

An alternative interpretation for the greater number of errors committed by the Fmr1 KO mice relates to reported increased exploratory behaviour and hyperactivity (The Dutch-Belgian Fragile X Consortium 1994; Zupan & Toth 2008), less freezing behaviour (Paradee et al. 1999), more open field entries (Yan et al. 2004), as well as decreased anxiety (Yan et al. 2004) in this transgenic group. Therefore, increases in both hyperactivity and reductions in neophobia characteristic of Fmr1 KO mice may result in increased entries in to error zones, which in turn resulted in poorer learning or encoding of inefficient solutions. To test this hypothesis, a post hoc analysis of activity levels between the murine groups was conducted for the first trial of the first maze tested. Consistent with previous reports, our results indicate that there are no significant differences in activity levels between Fmr1 KO and control mice (Mineur et al. 2002; Peier 2000; Spencer 2005; The Dutch-Belgian Fragile X Consortium 1994; Zupan & Toth 2008). However, the limited number of data points included in the analysis precludes definitive evidence ruling out activity level as a confound. Analyses of the paths taken by the respective groups to reach the goal box is part of an ongoing study to establish whether increased errors committed by the KO mice is a cause or consequence of a learning deficit.

The deficits observed in the Fmr1 KO mice may also relate to hippocampal dysfunction. Similar to typically developing humans, wild-type mice express high levels of Fmr1 mRNAs in the hippocampus (Hinds et al. 1993). Cultured hippocampal neurons from Fmr1 KO mouse pups harbour abnormal dendritic spines (Braun & Segal 2000; Grossman et al. 2006) and fewer functional synaptic connections (Braun & Segal 2000). Smaller intra-infra pyramidal mossy fibre terminal fields have also been observed in the hippocampi of Fmr1 KO mice (Mineur et al. 2002). Furthermore, these mice show behavioural deficits on tasks thought to be dependent on hippocampal function, such as the radial maze (Mineur et al. 2002), reversal trials of the Morris water maze (D’Hooge et al. 1997; Kooy et al. 1996; The Dutch-Belgian Fragile X Consortium 1994;Van Dam et al. 2000) and performance on the cross-shaped maze (Dobkin et al. 2000). Therefore, we speculate that impaired performance of Fmr1 KO mice on the H-W mazes is attributable to abnormal processing in the hippocampus. Future studies should examine whether these deficits are attributable to impairment in encoding, storage, retrieval or some combination thereof. It is also possible that the greater number of errors committed by the KO mice relates to higher rates of perseveration for incorrect solutions to a given maze problem. An ongoing study of the paths taken to solve the mazes will address this issue.

Interestingly, although deficits in spatial memory in Fmr1 KO have been shown for many tasks, findings have often been inconsistent and difficult to replicate, with most studies reporting either mild or no differences in spatial abilities (Paradee et al. 1999; Peier et al. 2000; Yan et al. 2004). Many possible explanations for these inconsistencies have been suggested, such as differences in background strain (Spencer et al. 2006), the presence of a retinal degeneracy mutation in older FVB strains (Errijgers et al. 2007) and number of generations of back crossing (Gu et al. 2002). For example, Paradee et al. (1999) showed that the C57BL/6 and FVB-129 background strains produced different results on the Morris water maze task, which is consistent with the innate spatial abilities of each strain. Furthermore, Dobkin et al. (2000) showed reduced learning on the Morris water maze in the FVB-129 strain but not in C57BL/6 strain, consistent with the excellent spatial ability of C57BL/6 mice. In the present study, an FVB background was used. With its more modest spatial abilities, it may be a better murine model to investigate the visual phenotype of FXS.

Interpretation of our results may be limited by the choice of comparison groups included. For the human participants, mental age matching was chosen as optimal to meet our stated objectives (see Burack et al. 2004 for a discussion of this process). We sought to control for general cognitive ability, which allows conclusions to be drawn regarding specific cognitive deficits, in this case of spatial abilities, among affected individuals. This strategy resulted in a significant chronological age difference between the groups. Thus, the present findings may be attributable to differences in the stage of development of the comparison group (mean chronological age = 5 years) rather than a specific spatial cognition deficit among FXS-affected participants (mean chronological age = 24 years). Chronologically age-matched individuals were not included because significantly stronger cognitive abilities of this group would have resulted in ceiling level performance on the mazes precluding meaningful interpretation (e.g. Kogan et al. 2004b; Kogan et al. 2009). In contrast, Fmr1 KO mice were compared with chronologically matched controls. This choice could have increased the likelihood of observing a significant difference between the two groups tested. Studies of individuals affected by FXS primarily compare their performance to individuals who are typically developing but score similarly on a given measure of intellectual functioning. Therefore, the comparison participants are almost always chronologically younger than the affected individuals in order to avoid the confound of general intelligence. As there is no established methodology for determining mental age of either KO or wild-type mice, matching on this variable was not possible. A survey of the literature on Fmr1 KO mice shows that studies either fail to inform the reader about the age of the subjects or use chronological age matching.

The present study included three participants with FMR1 mosaicisms. That is, these individuals express FMR1 in some of their somatic cells. Inclusion of these individuals could have minimized the likelihood of observing significant differences, because mosaicisms tend to create less severe impairments across most cognitive measures (Cohen et al. 1996; McConkie-Rosell et al. 1993; Merenstein et al. 1996). Despite their inclusion, a significant difference in the latency of the FXS individuals was observed, as compared with mental age-matched peers. Furthermore, review of data obtained from individuals with mosaicisms showed that their performance was similar to that of the participants with the full mutation.

A clear strength of the present study is its novelty. Specifically, there have been few studies comparing human and rodent performance on the same task (Shore et al. 2001) and to our knowledge, none that have directly compared individuals affected by FXS to a murine model on the same task. This is important because using a translational approach increases the validity of the murine model as well as strengthens the notion that our results reflect similar underlying spatial deficits that are targeted and expressed in both species. An additional strength of this study was the establishment of the validity of the H-W mazes as a behavioural assay to be able to detect the behavioural deficits in both human and mice. We propose that H-W mazes could be used as a tool to evaluate the therapeutic strategies both during early development in animals and later in clinical trials. Pharmacological compounds have been used to treat cognitive and behavioural deficits of FXS (Berry-Kravis et al. 2004, 2006), but there have been some challenges in establishing reliable outcome measures. Many cognitive measures used previously have been too difficult for the majority of individuals with FXS to complete and further have produced unacceptable levels of variability (Berry-Kravis et al. 2006). Additional studies investigating the reliability of the H-W paradigm in its ability to detect spatial deficits as well the stability of performance measures over time would be beneficial.

Acknowledgments

This work was supported by a grant from the National Fragile X Foundation and the National Sciences and Engineering Research Council of Canada to C.K. The authors would like to thank Dr Joe MacInnis for generously providing us with the maze program, Dr Isabelle Boutet for her expert assistance in the behavioural data acquisition and Ms Katherine Bendell for her assistance in testing participants. We also thank Dr Dwayne Schindler for his assistance with the statistical analyses and Dr Matthew Holahan for providing helpful input on a revision. Finally, we express our gratitude to the families and participants who gave generously of their time.

Ancillary