Effect Size of Reference Memory Deficits in the Morris Water Maze in Tg2576 Mice

Abstract

The most widely used mouse model of Alzheimer’s disease is the Tg2576 (APP_SWE) model. While general agreement about their neuropathology prevails, disparate results concerning cognitive changes have been reported. To resolve this controversy, we combined Morris water maze data collected over >10 years to determine the extent of memory impairment. APP_SWE mice exhibited an age-dependent decline in memory, but the effect size was small when compared to non-transgenic littermates. Larger effect sizes were achieved when comparing APP_SWE and Tg5469 (APP_WT) mice.

Alzheimer’s disease (AD) is characterized by an age-related impairment in learning and memory, neuronal loss, gliosis, neuritic changes, amyloid deposition, and abnormal tau phosphorylation and aggregation [1–3]. Animal models of AD should display both the pathological changes observed in AD, as well as changes in memory function that worsen in an age-dependent manner. The latter is particularly important since the primary risk factor for sporadic AD is age.

Over 15 years ago, the Tg2576 mouse (herein referred to as APP_SWE) was developed as an animal model of AD, and has since been used in over 607 articles pertaining to the pathogenesis and treatment of AD. This model over-expresses the 695 amino acid human isoform of the amyloid precursor protein (APP₆₉₅) with the “Swedish” mutation, resulting in the overproduction of the amyloid-beta (Aβ) peptide [4]. The APP_SWE mouse model recapitulates some of the hallmarks of AD, including neuritic changes, inflammation, amyloid deposition and plaques [5–7]. Although most studies have found memory deficits in the APP_SWE mice, discrepancies exist as to when the onset of these deficits first occurs, with reports ranging from 3 months to as late as 15 months, as well as ages in-between [4, 8–11]. Disparate results occur within laboratories as well [4, 8, 12–14]. Some investigators have failed to find deficits altogether [15]. Thus, it has been difficult to discern if an age-related impairment exists in the APP_SWE model.

Several possible reasons for these discrepancies exist, including sensitivity differences in the cognitive tests used. The current paper, however, focuses on two other possibilities: (1) the relatively small effect size seen when comparing APP_SWE mice to transgene negative (Tg Neg) mice using the Morris water maze and (2) the cognitive-enhancing effects of secreted APPα [sAPPα; 16, 17] and the APP intracellular domain [AICD; 18], additional byproducts of APP cleavage.

In the APP_SWE mouse model, sAPPα, AICD, and Aβ are over-expressed following proteolytic processing of APP. sAPPα has been shown to enhance long-term potentiation (LTP), modulate the induction of long-term depression (LTD) [16], and enhance memory performance in a variety of learning-tasks following intracerebroventricular injection [17]. Likewise, AICD facilitates memory and synaptic plasticity [18, 19]. This stands in opposition to Aβ, which impairs memory [4, 14, 20] and synaptic function [9, 21–24]. Consequently, the proper control for APP_SWE mice is the Tg5469 (APP_WT) mouse line. We used previously described methods [20] to show that, with the exception of Aβ, the levels of APP and APP metabolites were similar between the two lines. Briefly, a four-step extraction protocol was used to generate four fractions (extracellular-enriched soluble (EC), intracellular-enriched soluble (IC), membrane-enriched (MB) and insoluble). 1 ug of protein from the EC and MB fractions were loaded onto gels to probe for sAPPα and APP, respectively, using 6E10 (Signet, 1:2500). 150 ug of protein from the MB was used to probe for AICD with an anti-APP C-terminal antibody 0443 (Millipore, 1:5000), and 100 ug of protein from the EC fraction was used to probe for Aβ with 6E10. Blots were stripped and reprobed for α-Tubulin using an anti-α-Tubulin antibody (Sigma, 1:200,000). OptiQuant was used for densitometric analysis, and bands were normalized to α–Tubulin loading controls for each sample. Using this method, we found that the APP_WT mice over-express wild-type human APP at levels equivalent to mutant APP in APP_SWE mice (Fig. 1A & 1E), have equivalently high levels of AICD (Fig. 1B & 1E) and sAPPα (Fig. 1C & 1E) but much lower levels of Aβ (Fig. 1D & 1E).

Figure 1. Comparison of APP and APP metabolites in 8.5 month old APP_SWE and APP_WT mice.

A–C, Forebrain lysates from 5 APP_WT and 5 APP_SWE mice were analyzed by SDS-PAGE and immunoblot probed with (A) 6E10, a mouse monoclonal antibody, to detect full-length APP (fl-APP) in the membrane-enriched fraction (MB), (B) a rabbit polyclonal anti-APP antibody to recognize AICD in the MB fraction, and (C) 6E10 to detect sAPPα in the extracellular-enriched (EC) fraction. To ensure there was no fl-APP contamination during protein extraction, the same amount of the EC fraction was immunoprecipitated with an anti-APP antibody that recognizes the C-terminal region. No fl-APP was detected from subsequent probing with 6E10 (data not shown). (D) 6E10 was used to detect Aβ species in the EC fraction. All blots were stripped and reprobed with anti-α-Tubulin (bottom rows). (E), APP, AICD, sAPPα, and Aβ were quantified by normalizing the band intensity to that of α-Tubulin to determine the relative intensity between APP_WT and APP_SWE mice. Mean levels in APP_WT mice were defined as 1.0. There were no significant differences in protein levels of α–Tubulin, fl-APP, AICD, or sAPPα (Ps>0.1), but Aβ levels were significantly higher for the APP_SWE mice (P< 0.01).

These are important issues because APP transgenic mice are often used to evaluate potential therapies for AD. Knowing when memory loss first appears as well as the appropriate sample size needed is essential for establishing experimental designs. In addition, an appropriate effect size is needed to ensure that the dynamic range for a particular cognitive task is large enough so that subtle treatment effects can be detected. Comparison of APP_SWE to APP_WT mice should increase the effect size of APP_SWE mice by controlling for the beneficial effects of sAPPα and AICD overexpression, making this a more useful model for therapeutic testing. The current paper tests this possibility.

Here, we combined Morris water maze data collected over more than 10 years in our laboratory in order to determine if an age-dependent impairment in memory exists, and report effect size and sample size needed at various ages. We also report the effect sizes of APP_SWE versus APP_WT mouse models in an effort to distinguish the beneficial effects of sAPPα and AICD expression from the detrimental effects of Aβ.

Spatial reference learning and memory was tested using the Morris water maze in a cohort of 86 Tg5469 (APP_WT), 146 Tg2576 mice (APP_SWE), and 243 Tg negative littermates (Tg Neg) at various time points across their life span (Table 1). All transgenic mice used in this study were generated by breeding transgene positive male APP_WT or APP_SWE mice to female C57Bl6j/SJL F1 mice. The resultant mixed-background mice are F2-like in strain characteristics. To ensure that strain backgrounds in the APP_WT and APP_SWE lines were similar, behavioral scores for Tg Negs littermates from both lines were compared. There were no differences between the Tg Negs for any measure tested (Ps>0.5).

Table 1.

Animals tested in the MWM.

Mice Tested	Included
	Male	Female
APPWT (Tg5469)
4–5 Months	11	13
6–11 Months	14	8
12–18 Months	9	19
20–24 Months	6	6
APPSWE (Tg2576)
4–5 Months	10	15
6–11Months	29	33
12–18 Months	20	16
20–24 Months	15	8
Tg neg
4–5 Months	23	25
6–11Months	41	44
12–18 Months	30	37
20–24 Months	23	20
Total	231	244

Mice were grouped into four age ranges [14] based on previously established changes in soluble Aβ oligomers [20], detergent-insoluble Aβ (Aβ_insol) levels [25], and plaque pathology [4]: (1) very young mice, 4–5 months, after the appearance of Aβ trimers and hexamers but before the appearance of Aβ_insol or plaques; (2) young mice, 6–11 months, after the appearance of Aβ*56, a 56-kDa Aβ oligomer, and during the initial appearance of Aβ_insol and both amyloid plaques and punctate Aβ deposits; (3) middle-aged mice, 12–18 months, during a period in which soluble Aβ oligomers do not change but there is extensive deposition of plaques and Aβ_insol levels are rising rapidly; and (4) old mice, 20–25 months, at a time when soluble Aβ levels are stable, Aβ_insol is leveling off and amyloid loads are comparable to those in patients with AD (see Tables 1, 2). Mice were naïve at each time point tested.

Table 2.

Effect and sample sizes of APP_SWE and APP_WT B6/SJL mice tested in the Morris water maze.

Age	Probe	Males and Females Combined APPSWE vs. Tg Neg				Males and Females Combined APPSWE vs. APPWT				Males and Females Combined APPWT vs. Tg Neg
		Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference2	Pooled Standard Deviation	Effect Size	N Needed per Group
4–5 Months	Probe 1	−3.57	12.79	0.28	203	−2.27	13.16	0.17	528	−1.30	11.11	0.12	1146
	Probe 2	−2.00	12.87	0.16	650	0.22	12.89	0.02	53885	−2.22	11.41	0.20	416
	Probe 3	−3.81	14.11	0.27	216	−4.51	13.70	0.33	146	0.70	14.06	0.05	6330
	Mean Probe	−3.13	10.01	0.31	162	−2.19	10.33	0.21	350	−0.94	8.71	0.11	1347
6–11 Months	Probe 1	2.68	11.16	0.24	273	9.03	11.95	0.76	29	−6.35	11.14	0.57	50
	Probe 2	1.49	13.70	0.11	1326	8.49	15.02	0.57	51	−7.00	12.49	0.56	51
	Probe 3	4.51	13.77	0.33	148	8.60	15.96	0.54	55	−4.09	12.78	0.32	155
	Mean Probe	2.90	9.32	0.31	163	8.71	10.18	0.86	23	−5.81	8.19	0.71	33
12–18 Months	Probe 1	6.18	13.70	0.45	79	6.78	12.36	0.55	54	−0.60	14.66	0.04	9368
	Probe 2	7.99	13.26	0.60	45	8.38	13.93	0.60	45	−0.39	13.46	0.03	18695
	Probe 3	5.70	13.86	0.41	94	7.31	14.07	0.52	60	−1.61	15.10	0.11	1380
	Mean Probe	5.40	10.28	0.53	58	7.49	9.84	0.76	29	−2.09	11.18	0.19	450
20–24 Months	Probe 1	−0.08	11.32	0.01	314300	7.53	12.79	0.59	47	−7.61	11.27	0.68	36
	Probe 2	4.21	11.98	0.35	128	9.68	13.84	0.70	34	−5.47	11.09	0.49	66
	Probe 3	14.03	11.84	1.19	13	19.11	11.74	1.63	8	−5.08	13.03	0.39	105
	Mean Probe	6.05	8.85	0.68	35	12.11	9.24	1.31	11	−6.06	8.91	0.68	35

¹A negative value indicates a higher mean for APP _SWE (Tg2576)

²A negative value indicates a higher mean for APP _WT (Tg5469)

We previously described in detail the Morris water maze procedure used here [14]. Briefly, at each age tested, mice received visible platform training for 3 days, eight trials per day, followed by hidden platform training for 9 days, four trials per day. Three probe trials of 60 s duration were performed at the beginning of the 4th, 7th, and 10th day of hidden platform training. The mean platform score (MPS) was used to assess retention of spatial reference memory and was calculated for each mouse by averaging time spent in the quadrant area for the three probes conducted. Although similar trends were observed using the platform crossing index (PCI), percent time is reported here because it is the most popular dependent measure reported in the Morris water maze literature [26], including our own search of the APP_SWE literature.

All trials were monitored using a computerized tracking system (Noldus EthoVision 3.0; Noldus Information Technology, Wageningen, The Netherlands), and performance measures were extracted using Wintrack (Wolfer, et al. 2001). Statistical analysis consisted of t-tests, ANOVA and repeated-measures ANOVA (RMANOVA). Post hoc comparisons were performed using Bonferroni with p values of <0.05 considered significant.

Effect size and sample size needed was calculated for each probe and MPS using the software package Systat (2004). Cohen’s d for t-tests was used to calculate effect sizes and was chosen for two reasons. First, this calculation is one of the most popular, allowing for easy comparison to other published studies. Second, Cohen’s [27] classification of effect sizes into categories (.20 - small, .50 - medium, and .80 - large) makes evaluation of this experiment’s effect-size results easy to compare to known benchmarks.

To calculate effect size, we computed the standardized mean difference (SMD) as the difference between the APP_SWE mice and either the Tg Neg or the APP_WT mice divided by the pooled standard deviation. In addition, we compared the Tg Neg mice to the APP_WT mice.

$$d=M_1-M_2/{\sigma }_{\mathit{pooled}}$$ (a)

$${\sigma }_{\mathit{pooled}}=\surd [((n_1-1){s_1}^2+(n_2-1){s_2}^2)/(n_1+n_2)]$$ (b)

Key to symbols:

• d = Cohen’s d effect size

• M₁ = mean (average of Tg Neg or APP_WT)

• M₂ = mean (average of APP_SWE)

• s = standard deviation

• n = number of subjects

Effect sizes were computed as Cohen’s d where a positive effect size represents better performance for the (1) Tg Neg mice when compared to either the APP_WT or APP_SWE mice or (2) APP_WT when compared to the APP_SWE mice (Tables 2 & 3). In order to estimate an appropriate sample size, the SMD was determined, the probability of a Type I error (α) and power were set at 0.05 and 0.80, respectively, and the alternative was specified as not equal.

Table 3.

Gender effects on effect and sample size of APP_SWE and APP_WT B6/SJL mice tested in the Morris water maze.

Age	Probe	Males Only APPSWE vs. Tg Neg				Males Only APPSWE vs. APPWT				Males Only APPWT vs. Tg Neg
		Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference2	Pooled Standard Deviation	Effect Size	N Needed per Group
4–5 Months	Probe 1	−3.89	10.48	0.37	115	−4.84	12.74	0.38	110	0.95	9.99	0.10	1735
	Probe 2	−6.15	11.94	0.52	61	0.97	9.44	0.10	1486	−7.12	11.61	0.61	43
	Probe 3	−6.61	12.35	0.54	56	−10.32	12.41	0.83	24	3.70	13.41	0.28	207
	Mean Probe	−5.55	8.71	0.64	40	−4.73	8.33	0.57	50	−0.82	9.17	0.09	1961
6–11 Months	Probe 1	2.37	11.66	0.20	381	11.79	13.51	0.87	22	−9.42	10.89	0.87	22
	Probe 2	0.25	13.87	0.02	48315	6.58	15.32	0.43	86	−6.33	12.25	0.52	60
	Probe 3	2.20	13.79	0.16	616	5.63	17.44	0.32	152	−3.44	12.81	0.27	219
	Mean Probe	1.61	9.74	0.17	574	8.00	11.53	0.69	34	−6.39	7.69	0.83	24
12–18 Months	Probe 1	11.43	12.15	0.94	19	13.30	13.51	0.98	18	−1.86	14.85	0.13	1000
	Probe 2	11.33	13.04	0.87	22	12.12	15.82	0.77	28	−0.79	13.12	0.06	4326
	Probe 3	8.09	13.30	0.61	44	6.36	14.23	0.45	80	1.73	15.03	0.12	1184
	Mean Probe	10.28	9.81	1.05	16	10.59	10.67	0.99	17	−0.31	12.36	0.03	24951
20–24 Months	Probe 1	−1.33	10.29	0.13	939	18.71	11.65	1.61	8	−20.04	9.30	2.16	5
	Probe 2	9.07	10.79	0.84	24	21.23	13.18	1.54	8	−12.16	9.79	1.24	12
	Probe 3	14.41	11.75	1.23	12	18.04	10.95	1.65	7	−3.63	12.00	0.30	173
	Mean Probe	7.38	8.04	0.92	20	19.33	8.11	2.39	4	−11.94	8.19	1.46	9

Age	Probe	Females Only APPSWE vs. Tg Neg				Females Only APPSWE vs. APPWT				Females Only APPWT vs. Tg Neg
		Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference1	Pooled Standard Deviation	Effect Size	N Needed per Group	Mean Difference2	Pooled Standard Deviation	Effect Size	N Needed per Group
4–5 Months	Probe 1	−3.54	14.38	0.25	260	−0.29	13.38	0.02	33412	−3.25	11.95	0.27	213
	Probe 2	1.41	13.27	0.11	1389	−0.70	14.79	0.05	7004	2.11	10.75	0.20	408
	Probe 3	−2.22	15.16	0.15	732	−0.30	14.19	0.02	35117	−1.92	14.50	0.13	895
	Mean Probe	−1.45	10.86	0.13	880	−0.43	11.48	0.04	11185	−1.02	8.28	0.12	1034
6–11 Months	Probe 1	2.96	10.69	0.28	206	3.98	9.62	0.41	93	−1.03	11.11	0.09	1825
	Probe 2	2.64	13.50	0.20	411	11.71	14.59	0.80	26	−9.07	12.57	0.72	32
	Probe 3	6.53	13.57	0.48	69	11.47	13.90	0.83	25	−4.94	12.73	0.39	106
	Mean Probe	4.04	8.88	0.46	77	9.05	8.45	1.07	15	−5.01	8.69	0.58	49
12–18 Months	Probe 1	2.58	13.32	0.19	419	4.85	10.06	0.48	69	−2.27	12.32	0.18	463
	Probe 2	4.99	13.18	0.38	111	6.10	11.97	0.51	62	−1.11	13.35	0.08	2268
	Probe 3	9.35	13.62	0.69	35	9.99	13.50	0.74	30	−0.64	14.68	0.04	8256
	Mean Probe	−4.36	10.17	0.43	87	6.98	8.58	0.81	25	−11.34	9.58	1.18	13
20–24 Months	Probe 1	0.15	11.80	0.01	97141	−5.10	10.92	0.47	73	5.26	10.69	0.49	66
	Probe 2	−3.56	12.66	0.28	200	−4.50	11.32	0.40	101	0.94	10.64	0.09	2009
	Probe 3	13.79	11.93	1.16	13	20.43	12.80	1.60	8	−6.64	14.05	0.47	72
	Mean Probe	3.46	9.68	0.36	124	3.61	8.93	0.40	97	−0.15	8.83	0.02	54394

¹A negative value indicates a higher mean for APP _SWE (Tg2576)

²A negative value indicates a higher mean for APP _WT (Tg5469)

To justify individual comparisons of transgene at each age range, we first examined the effect of age, gender, and transgene on probe scores to determine if there were any main effects or interactions among these variables. This analysis revealed a main effect of (1) age (P=0.004), performance declined as the mice aged, (2) transgene (P<0.0001), APP_WT mice performed better than Tg Negs and APP_SWE mice (ps<0.01) while Tg Negs were superior to APP_SWE mice (p<0.01), (3) gender (P=0.0057), males performed better than females, and (4) an interaction between transgene and age (P<0.0001). To clarify the interaction between age and transgene, we examined the effects of transgene at each age range.

When compared with APP_WT mice, spatial reference memory was significantly impaired in APP_SWE mice at all ages tested after 4–5 months of age (Fig. 2). At 6–11 and 12–18 months of age, a medium to large effect size was observed for every measure when comparing APP_SWE to APP_WT mice, regardless of gender (Table 2 and 3). However, at 20–24 months of age the effect size was gender dependent; APP_WT males performed better than APP_SWE males at each probe, whereas APP_WT females were more variable (Table 3). In contrast, comparison between Tg Neg and APP_SWE mice revealed much smaller effect sizes, regardless of gender (Table 3), leading to a lack of statistical difference at 6–11 months of age (Fig. 2). In addition, these small effect sizes mandated the need for much larger sample sizes compared to the practical numbers needed for comparisons between APP_SWE and APP_WT mice, particularly when comparing males (Table 2 & 3), making this latter comparison the more prudent of the two.

Figure 2. Assessment of memory function in APP_SWE, APP_WT, and Tg Neg mice using the Morris water maze.

A, Compared with APP_WT and Tg Negs, the time spent swimming in the target quadrant during probe trials did not differ in APP_SWE mice at 4–5 months of age. Mean platform score (MPS) represents the average time spent in the quadrant area for the three probes conducted. RMANOVA data are as follows: transgene (MPS): P= 0.44; probe versus transgene: P=0.86. B, At 6–11 months of age, APP_WT mice spent significantly more time in the target quadrant than both APP_SWE and Tg Neg mice. RMANOVA data are as follows: transgene (MPS): P=0.001; probe versus transgene: P=0.83. C, APP_SWE mice were impaired compared to both APP_WT and Tg Neg mice. RMANOVA data are as follows: transgene (MPS): P=0.003; probe versus transgene: P=0.97. D, Although APP_WT and Tg Neg mice improved with repeated probe testing, APP_SWE mice exhibited stable performance (probe trial RMANOVA: P=0.46), resulting in a probe by transgene interaction. RMANOVA data are as follows: transgene (MPS): P=0.001; probe versus transgene: P=0.004. # APPSWE vs. APPWT p<0.05, APPSWE vs. Tg Neg p<0.05, * APPWT vs. Tg Neg p<0.05.

To determine whether an age-dependent decline in retention of spatial reference memory was present for any of our three groups, the 4 time points were compared for each group separately. We then compared performance at 6–11, 12–18, and 20–24 months of age to that at 4–5 months of age, a time at which the groups did not differ. For the APP_SWE mice, the first indication of impaired spatial reference memory was observed at 6–11 months (Fig. 3). As the mice aged, retention of spatial memory became more dramatically impaired, whereas for both the APP_WT and Tg Neg mice, there were no age-related declines in performance (Fig. 3).

Figure 3. APP_SWE mice develop age-dependent memory deficits not seen in APP_WT or Tg Neg mice.

Mean platform score (MPS) represents the average time spent in the quadrant area for the three probes conducted.. APP_SWE mice exhibited an age-dependent decline in MPS performance (P=0.0002) that was not observed in APP_WT (P=0.23) or Tg Neg mice (P=0.07). 4–5M vs. older ages *p < 0.05, **p < 0.001, ***p < 0.0001.

These results suggest that APP_SWE mice do in fact have an age-dependent decline in memory but that the effect size is quite small from 6–11 months when compared to Tg Neg mice. This small effect size is most likely due to the opposing effects in APP_SWE mice of sAPPα and AICD, which enhance cognition [16–18], and the accumulation of Aβ oligomers, which disrupt cognition. Support for this comes from comparisons between APP_SWE and APP_WT mice, which over-express sAPPα and AICD, but not Aβ, to the same extent as APP_SWE mice. This comparison results in a much greater effect size, particularly at 6–11 months of age. Likewise, APP_WT mice outperform Tg Neg mice at 6–11 and 20–24 months of age, similar to previous reports [18]. Thus, one potential reason for disparate results in the APP_SWE literature is the use of Tg Neg mice. When using a small number of animals with a relatively small effect size, which is often the case when experimenters compare APP_SWE to Tg Neg mice, there is greater potential for erratic results due to variations in random sampling. By using the APP_WT mice with their greater effect size, variability between experiments should be reduced. We should note that although we compared the Tg Negs from the APP_WT and APP_SWE lines to ensure that background strains were similar between the two lines, it is possible that parental strains of the two lines have diverged over a 10–15 year period and may differ from those in other laboratories. The emergence of sublines of APP_SWE during the 15 years since the first founder was created could explain the acquisition deficits seen in mice purchased from Taconic (for example, [28–30]).

The use of APP_WT mice, instead of Tg Neg mice, for comparison with APP_SWE mice results in a larger difference at 6–11 months of age. Because this is the first time point at which an age-dependent decline in performance is seen (Fig. 3), the study of cognitive enhancers at this age would most likely inform therapeutic prevention in AD. Therefore, it is imperative that differences between the APP_SWE mice and the control group be as large as possible to detect subtle enhancements in performance. The difference between the Tg Neg and APP_SWE mice at this time-point is not statistically significant when a large cohort of animals are compared. Therefore, subtle changes in performance are likely to be undetected.

It should be noted that the APP_SWE mouse is only a partial model of AD that lacks both neurofibrillary tangles and neuronal loss. This mouse may best be thought of as a latent AD model. If this is the case, it is not surprising that deficits are subtle and that variability is greater in this model than in those exhibiting substantial deficits at a very young age. While these characteristics make the model difficult to use, they also make it one of the best models for research on the prevention of AD. The subtle and slow progression of cognitive deficits allows for therapeutic intervention before extensive pathology is present. A key factor in the use of this model lies in understanding the small effect size when compared to Tg Neg mice and utilizing the APP_WT mouse line to increase this effect size.