Behavioral Neuroscience 2003, Vol. 117, No. 3, 478 – 484
Copyright 2003 by the American Psychological Association, Inc. 0735-7044/03/$12.00 DOI: 10.1037/0735-7044.117.3.478
Short-Term -Amyloid Vaccinations Do Not Improve Cognitive Performance in Cognitively Impaired APP⫹PS1 Mice L. Austin, G. W. Arendash, M. N. Gordon, D. M. Diamond, G. DiCarlo, C. Dickey, K. Ugen, and D. Morgan University of South Florida Prior work demonstrated that -amyloid (A) immunotherapy for 8 months prevented cognitive impairment in 16-month-old APP⫹PS1 transgenic mice. In the present study, 4 immunizations administered biweekly to cognitively impaired 16-month-old transgenic mice could not reverse deficits in working memory or reference memory in the radial arm water maze or in visual platform recognition, possibly because of inadequate antibody exposure. Nontransgenic mice showed cognitive savings between the 16and 18-month test periods, but the transgenic groups did not. These results suggest that a longer period of active immunotherapy, or passive immunization, may be required to provide sufficient antibody titers to improve cognition in older transgenic mice. A-based immunotherapy for Alzheimer’s disease will likely be more successful prophylactically than therapeutically.
The process or the product of brain -amyloid (A) peptide deposition appears to be of fundamental importance in the pathogenesis of Alzheimer’s disease (Hardy, 1997; Selkoe, 2001). Consistent with this premise, increased brain processing of amyloid precursor protein (APP) to A and ensuing aggregation of A into amyloid deposits occur prior to the clinical diagnosis of Alzheimer’s disease (Lippa, Nee, Mori, & St. George-Hyslop, 1998). Thus, therapeutics that block brain A aggregation or clear brain A may prevent or lessen the behavioral impairment of this disease. Along this line, active immunization against A through long-term administration of aggregated or soluble A peptides prevents A deposition in several transgenic models of Alzheimer’s disease (Janus et al., 2000; Morgan et al., 2000; Schenk et al., 1999; Sigurdsson, Scholtzova, Mehta, Frangione, & Wisniewski, 2001; Weiner et al., 2000). T-cell activation is apparently not involved in this effect because long-term administration of monoclonal or polyclonal antibodies against A (e.g., passive immunization) can also prevent A deposition (Bard et al., 2000; DeMattos et al., 2001). We have recently explored the behavioral consequences of active immunization in both APPsw and APPsw⫹PS1 transgenic lines (Arendash, Gordon, et al., 2001; Morgan et al., 2000). Im-
munotherapy begun in unimpaired Tg⫹ mice at 7.5 months was found to protect against working memory impairment manifest at 15.5 months while reducing brain A burdens by approximately 25%. This behavioral benefit of A immunization was task specific, being evident against working memory deficits in the radial arm water maze (RAWM) but having no effect on a preexisting balance beam deficit or on Y-maze performance (Arendash, King, et al., 2001). In another study, which involved repeated testing of TgCRND8 transgenic mice between 3– 6 months, Janus et al. (2000) reported that A immunization partially reverses early reference learning impairment and reduces A deposition by 50%. Most recently, Dodart et al. (2002) passively immunized PDAPP transgenic mice through a single injection of the monoclonal antibody m266. In behavioral testing performed 1– 4 days later, m266-immunized mice had a higher level of object recognition and a lower number of holeboard errors. Because no effects on brain A deposition resulted from this single m266 treatment but much higher plasma levels of soluble A were evident, the researchers suggested that the rapid behavioral effects of m266 are due to its facilitation of A efflux from the brain. The above behavioral studies collectively indicate that immunization against A can have beneficial behavioral effects, perhaps through several different or synergistic mechanisms. In our initial study (Morgan et al., 2000), long-term (8-month) active immunization begun in unimpaired 7.5-month Tg⫹ mice provided protection against working memory impairment in the RAWM task at 15.5 months. This finding suggests that a therapeutic effect will result from active immunization begun in humans destined to develop Alzheimer’s disease who are not symptomatic at the beginning of inoculations. The present study addressed the question of whether a relatively short 6-week vaccination period (from 17 months to 18.5 months) would benefit aged APP⫹PS1 transgenic mice confirmed to be impaired in both RAWM working memory and platform recognition (PR) tasks during pretreatment testing at 16 –17 months. In such aged APP⫹PS1 mice, the extent of cognitive impairment in RAWM working memory is highly correlated with A deposition (Aren-
L. Austin and G. W. Arendash, Department of Biology and Memory and Aging Research Laboratory, University of South Florida; M. N. Gordon, G. DiCarlo, C. Dickey, and D. Morgan, Department of Pharmacology, University of South Florida; D. M. Diamond, Department of Psychology, University of South Florida; K. Ugen, Department of Microbiology, University of South Florida. This work was supported by Grant AG18478 to D. Morgan at the University of South Florida Alzheimer’s Research Consortium. We thank Karen Duff and Karen Hsiao for the original founder PS1 and Tg2576 (APP) transgenic lines used in breeding APP⫹PS1 mice for this study. Correspondence concerning this article should be addressed to G. W. Arendash, Memory and Aging Research Laboratory, SCA 110, University of South Florida, Tampa, Florida 33620. E-mail:
[email protected] .usf.edu 478
-AMYLOID VACCINATION IN APP⫹PS1 MICE
dash, Gordon, et al., 2001; Gordon et al., 2001). We report no effect of A vaccinations upon retesting of aged Tg⫹ mice in both RAWM and PR tasks at 18.5–19 months, as well as no effect of immunotherapy in RAWM reference learning done at 20 months. A longer A vaccination period or direct A antibody administration (passive immunization) could provide more behaviorally efficacious effects. Our results suggest that short-term A immunization in symptomatic Alzheimer’s patients is unlikely to reverse their established cognitive impairments in working or reference memory.
Method Subjects Twenty-six double-transgenic APP⫹PS1 mice and 9 nontransgenic (NT) littermate controls 16 months of age were used for this study. Double-transgenic APP⫹PS1 mice were derived originally from a cross between the mAPP transgenic line Tg2576 and the mPS1 transgenic line 5.1. It should be noted that the Tg2576 mice derive from a C57B6/ SJL ⫻ C57B6 background, whereas the PS1–5.1 line derives from a Swiss Webster/B6D2F1 background, producing progeny with a mixture of these backgrounds. The mice used in this study were mostly from the seventh through ninth generations after the initial cross. Whenever possible, the nontransgenic mice were gender-matched littermates of transgenic mice used in this study. After weaning, the mice were genotyped and group housed until 1 week before testing began; then mice were housed individually. They were maintained on a 12-hr light– dark cycle, with free access to water and rodent chow (Teklad Global 18% protein rodent diet; Harland Teklad, Madison, WI). All behavior testing was done in the light phase of the circadian cycle. Because approximately half of the mice were 1 month older than the others, mice were divided into two subsets for behavioral testing that began at 16 months for each subset. The mice were housed in the same room in which behavior testing was performed.
General Protocol At 16 months, mice underwent 19 days of behavior testing in two cognitive tasks: 15 days in RAWM and 4 days in visible platform (see Figure 1). Two days after testing ended, the transgenic mice were begun on treatment with either A42 peptide or phosphate-buffered saline (PBS) with adjuvant every other week for 6 weeks. One week after the last treatment (i.e., 7 weeks after vaccination began) all mice were retested in both cognitive tasks beginning at 18 months of age. A final vaccination was given 2.5 weeks after completion of these behavior tasks. Most mice (Subset 2) were also posttreatment tested for reference learning in the
479
RAWM for 7 days. Mice that died before completion of behavioral testing or that were considered nonperformers were removed from behavioral analysis. At 20 months (11 days after the final treatment), blood was collected and allowed to clot on ice for more than 1 hr. Serum was collected by centrifugation and was used for determination of A antibody titers, as described previously (Dickey et al., 2001; Morgan et al., 2000).
Vaccination Human A1– 42 peptide was suspended in pyrogen-free Type 1 water at 2.2 mg/ml, then mixed with 10⫻ PBS to yield 1⫻ PBS and incubated overnight at 37 °C. Control Tg⫹ mice were injected with PBS plus adjuvant prepared in the same way. The antigen suspensions of A were mixed 1:1 with Freund’s complete adjuvant and 100 g A, or the control was injected subcutaneously by an experimenter who had no role in the behavioral testing. A boost of the same material (prepared fresh) was made in incomplete Freund’s at 2 weeks and injected biweekly for 6 weeks. To determine the two Tg⫹ treatment groups (Tg⫹/control and Tg⫹/A), we ranked Tg⫹ mice by their pretreatment RAWM performance, and every other mouse went into each group. Mice were vaccinated beginning at 16 months of age. The sample size for each group was as follows: 9 nontransgenics, 9 Tg⫹ vaccinated with A, and 17 Tg⫹ control vaccinations.
Behavioral Assessment RAWM: Working memory. The six-arm water maze was made by placing six dividers into a 100-cm circular pool to create six arms (30.5 cm long ⫻ 19 cm wide) that emanated from the center. The swim arm walls extended above the surface of the water by about 5 cm. Spatial cues were present on the walls surrounding the pool and ceiling. A submerged platform (9 cm in diameter) was placed near the end of one arm. The 15 days of pre- and posttreatment testing consisted of four acquisition trials and, after a 30-min delay, a retention (delayed-recall) trial. Each day, the platform’s location was changed to a different one of the six arms. A semirandom sequence of the remaining swim arms was used for the order of start arms for Trials 1–5 (T1–T5). The number of errors during the last acquisition trial (T4) was used as an index of daily learning. The memory retention trial (T5) was begun from the one remaining start arm for that day. For each trial, an error was determined by the mouse swimming into an arm that did not contain the escape platform, after which the mouse was returned (in the water) to that trial’s start arm to make another arm choice. A trial ran until the mouse found the platform or until 1 min of searching had elapsed. Mice had to learn the procedural strategy that the platform is in the same arm for each trial per day but shifts across days. For each trial, the mouse was placed at the mouth of the swim arm facing the center, and the number of errors made before finding the escape
Figure 1. Time line of the general protocol for this study. RAWM ⫽ radial arm water maze; Vax ⫽ vaccination; M ⫽ months.
480
AUSTIN ET AL.
platform was recorded. If a mouse swam in the central area without making a choice approximately every 20 s, it was returned to the start arm, and an error was recorded. After finding the platform, mice were allowed to stay there for 30 s before going on to the next trial. If the mouse did not locate the platform after 1 min, it was guided to the escape platform and given 30 s before proceeding. After the four acquisition trials, mice were returned to their home cages and dried by heat lamps during the 30-min delay period. Nonperforming animals (i.e., those that consistently refused to make arm choices, circled tightly, or floated) were removed from behavioral analysis. The experimenter was naive to the transgenicity and treatment of the mouse. Platform recognition (water maze). A 9-cm platform was placed in the 100-cm circular pool used for RAWM testing, with the six dividers removed. The platform was elevated 0.8 cm above the surface of the water with an attached 10 ⫻ 40-cm black and white foam structure (cue). Preand posttreatment testing began the day after RAWM testing finished and extended over a 4-day period. Animals were placed in the pool at the same wall location for all 4 daily trials, but the platform was changed to a different one of four quadrants for each trial. Mice were given a maximum of 1 min to find and ascend the platform; latency was recorded. For statistical analysis, we averaged latency for each of the day’s four trials. Those mice that did not find the platform within the given time were placed on the platform. A 30-s rest period was allowed between each trial. Nonperformers in this task were removed from statistical analysis. RAWM: Reference learning. For reference learning, the RAWM set up was the same as previously described, but the submerged platform stayed in the same arm location across 6 days of testing, and only four consecutive trials were performed each day. A semirandom sequence was generated for the start arms of T1–T4 each day. For each of the four trials, the mouse was placed at the mouth of that trial’s start arm and given up to 1 min to locate the platform. If the mouse made an error by swimming into an arm without the platform, it was returned to the start arm and charged an error (as in RAWM working memory). The number of errors was recorded for each trial. Mice were given 30 s between each trial to rest on the platform. If the mouse was not able to locate the platform after 1 min, it was guided to the platform and given a 30-s stay. Following an initial day of shaping, we calculated the average number of errors over four trials per day for each animal. Then the averages of all 6 days were combined for statistical analysis. Only the second subset of mice was tested posttreatment in this task (5 Tg⫺, 6 Tg⫹/control, and 4 Tg⫹/A mice, respectively).
Results RAWM: Working Memory Pretreatment testing. In pretreatment RAWM testing, there were no differences between Tg⫹ and Tg⫺ mice on T4 errors through all five blocks (3.8 errors vs. 3.2 errors), indicating that both groups had similar overall RAWM acquisition (see Figure 2). This was particularly apparent for the last block of testing (Block 5), wherein both groups were able to significantly reduce their errors between T1 and T4. By sharp contrast, analysis of the memory retention trial (T5) through all blocks of pretreatment testing revealed significantly poorer memory performance of Tg⫹ mice compared with Tg⫺ mice (4.3 errors vs. 2.8 errors), F(1, 19) ⫽ 8.68, p ⬍ .01. Compared with Tg⫺ mice, Tg⫹ mice made significantly more T5 errors during both Block 4 and Block 5 of testing ( p ⬍ .01, and p ⬍ .05, respectively; see Figure 2). Posttreatment testing. For all trials evaluated (T1, T4, and T5) over all blocks of posttreatment testing, RAWM performance of Tg⫹ mice given A vaccinations was not improved in comparison with Tg⫹ mice given adjuvant injections (see Figure 3). Analysis of T4 errors over all five blocks did, however, reveal a significant groups effect, F(2, 18) ⫽ 4.82, p ⬍ .025. Post hoc analysis indicated that both A vaccinated and adjuvant-treated Tg⫹ mice exhibited significantly more T4 errors overall than did Tg⫺ mice (Tg⫹/A ⫽ 3.4 errors, Tg⫹/control ⫽ 3.3 errors, Tg⫺ ⫽ 1.8 errors; p ⬍ .02 for both Tg⫹ groups vs. Tg⫺). Analysis of T4 in individual blocks revealed that one or both Tg⫹ groups was usually impaired (see Figure 3). Thus both Tg⫹ groups had impaired postvaccination RAWM acquisition. For T5, however, there was neither a significant group effect, F(2, 18) ⫽ 1.95, p ⫽ .17, nor a Group ⫻ Block interaction, F(8, 72) ⫽ 0.56, p ⫽ .81, over all blocks of testing (Tg⫹/A ⫽ 3.4 errors, Tg⫹/PBS ⫽ 3.1, Tg⫺ ⫽ 2.2). On the last block of posttreatment testing (Block 5), both T4 and T5 performance of Tg⫹/A mice was no different from that of Tg⫹/control mice (see Figure 3), although both T4
Statistical Analysis For RAWM, we averaged the data into five 3-day blocks. Pretreatment comparison of NT versus Tg⫹ groups, as well as posttreatment analysis of the three treatment groups (NT, Tg⫹/control, Tg⫹/A) was done by repeated measure analyses of variance (ANOVAs) across blocks, followed by Fisher’s least significant difference post hoc comparisons. Also, repeated measure ANOVAs and post hoc analysis were used to determine group differences in T4 and T5 across all blocks for both pre- and posttreatment RAWM testing. Savings for RAWM performance was analyzed by repeated measure ANOVAs comparing each group’s pre- versus posttreatment performance over all blocks. Also to evaluate savings, we used ANOVAs to compare each group’s Block 5 pretreatment performance to its Block 1 posttreatment performance, as well as to compare Block 1 pretreatment versus Block 1 posttreatment performance. For PR, we performed repeated measure ANOVAs with post hocs between groups for preand posttreatment performance across all 4 days. Then, using an ANOVA, we compared all groups on the last day of pre- and posttreatment testing. Pretreatment versus posttreatment behavior for each group was compared using repeated measure ANOVAs. Lastly, in reference RAWM, posttreatment data from the three groups (NT, Tg⫹/control, Tg⫹/A) were averaged over all 6 days of testing and compared using an ANOVA.
Figure 2. Radial arm water maze pretreatment acquisition (Trial [T]1– T4) and memory retention (T5) in transgenic (Tg⫹) mice and nontransgenic (NT) control mice over five 3-day blocks (B1–B5). Acquisition involved four daily trials and was measured by the number of errors to find a submerged platform in the goal arm (which was randomly changed each day). During memory retention (T5), Tg⫹ mice exhibited a working memory impairment over all five blocks, as well as for the last two blocks compared with NT controls. Data are means (⫾ SEM). * ⫽ significantly different from NT ( p ⬍ .05).
-AMYLOID VACCINATION IN APP⫹PS1 MICE
Figure 3. Radial arm water maze posttreatment acquisition (Trial [T]1– T4) and memory retention (T5) in transgenic control (Tg⫹/Con) treatment, transgenic -amyloid (Tg⫹/A) vaccinated, and nontransgenic (NT) mice over five 3-day blocks (B1–B5). NT mice showed learning between T1 and T4, whereas both groups of Tg⫹ mice were impaired overall in T4 learning, as well as at individual blocks, compared with NT mice. Data are means (⫾ SEM). * ⫽ significant difference between Tg⫹/A and NT ( p ⬍ .05); † ⫽ significant difference between Tg⫹/Con and NT ( p ⬍ .05).
and T5 errors were significantly higher in Tg⫹/A mice in Block 5 compared with Tg⫺ mice ( p ⬍ .05 for both trials). Pre- versus posttreatment testing. To further determine any pre- versus posttreatment change in RAWM performance (or cognitive savings) following the 7-week treatment period, we compared each group’s pretreatment RAWM performance with their posttreatment performance. Over all five blocks of RAWM testing, there were no pre- versus posttreatment differences in T1, T4, or T5 for the Tg⫹/A and Tg⫹/control groups. These results indicate an overall stabilization of RAWM performance in Tg⫹ mice but at a level of appreciable T4 and T5 error numbers. Although NT mice also exhibited a stabilization in overall RAWM performance for T1 and T5, the stabilization in T5 was at a low number of errors (i.e., 1–2 errors by the last block of pretreatment and posttreatment testing). Moreover, overall T4 (end learning) performance of NT mice was significantly better in posttreatment testing (3.2 pretreatment vs. 1.8 posttreatment errors; p ⬍ .05). As further indicative of improved RAWM learning by NT mice following the treatment period, they had (a) significantly fewer T4 errors in Block 1 posttreatment versus pretreatment (2.1 vs. 3.9 errors; p ⬍ .2) and (b) an identical number of T4 errors (2.1) in pretreatment Block 5 versus posttreatment Block 1 testing. By contrast, pretreatment Block 5 versus posttreatment Block 1 analysis of T4 errors for Tg⫹ groups revealed significantly poorer performance of Tg⫹/A mice (2.2 vs. 3.9 errors; p ⬍ .01) and nearly significant poorer performance of Tg⫹/control mice (2.2 vs. 3.6 errors; p ⫽ .08). Nonetheless, T5 (memory) performance of both Tg⫹ groups was similar in pretreatment Block 5 versus posttreatment Block 1 testing. Thus, NT mice clearly exhibited cognitive savings (i.e., improved or low-error stabilized performance) between RAWM test periods separated by 7 weeks, whereas Tg⫹ mice showed little or no evidence of cognitive savings between these RAWM test periods.
PR For both pre- and posttreatment PR testing, Tg⫺ mice had low escape latencies, even on the first day of testing (see Figure 4).
481
They were thus able to quickly change from the spatial working memory strategy of RAWM to the search–recognition strategy of PR. Compared with these Tg⫺ controls, both Tg⫹/A and Tg⫹/ control mice had much more difficulty in shifting to the PR strategy for both pre- and posttreatment PR testing. Significant group effects were present for both pretreatment, F(2, 19) ⫽ 5.10, p ⬍ .02, and posttreatment, F(2, 19) ⫽ 6.27, p ⬍ .01, testing. Post hoc analysis revealed that both Tg⫹ groups were not different from each other and both groups had significantly higher overall escape latencies compared with Tg⫺ mice ( p ⬍ .05 or higher level of significance). Significant Group ⫻ Block interactions were also present for both pre- and posttreatment testing, F(6, 57) ⫽ 3.30, p ⬍ .01; F(6, 57) ⫽ 3.16, p ⬍ .01, indicative of group differences in the rate of learning. Although both Tg⫹/A and Tg⫹/PBS groups showed overall impairment in both pre- and posttreatment PR performance, both of these groups were able to reduce their escape latencies to the level of Tg⫺ mice by the last day of testing (Day 4) for both pre- and posttreatment testing (see Figure 4). Comparing each group’s pretreatment performance to their respective posttreatment performance revealed no significant pre- versus posttreatment differences for any group.
RAWM Reference Learning In posttreatment RAWM reference learning (see Figure 5), a significant groups effect was present, F(2, 12) ⫽ 6.66, p ⬍ .02. Post hoc analysis showed that both Tg⫹/A and Tg⫹/control
Figure 4. Water maze platform recognition for transgenic control (Tg⫹/ Con) treatment, transgenic -amyloid (Tg⫹/A) vaccinated, and nontransgenic (NT) mice. Pretreatment (A) and posttreatment (B) acquisition was measured as latency (mean [⫾ SEM] of four trials per day) to find a distinctly marked platform over 4 test days. Both Tg⫹/control and Tg⫹/A vaccinated mice were equally impaired overall compared with NT controls in ability to locate the platform during both pretreatment ( p ⬍ .01 and p ⬍ .02, respectively) and posttreatment ( p ⬍ .005 and p ⬍ .05, respectively) testing. By the final day of testing however, there were no differences between the groups in pretreatment or posttreatment testing.
AUSTIN ET AL.
482
Figure 5. Radial arm water maze reference learning for transgenic control (Tg⫹/Con), treatment, transgenic -amyloid (Tg⫹/A) vaccinated, and nontransgenic (NT) control mice. Acquisition was measured as number of errors (mean [⫾ SEM] of four trials per day) to find a submerged platform positioned in the same arm over 6 test days. NT showed excellent reference learning, but the performance of both Tg⫹ groups was significantly worse than that of NT mice. * ⫽ significantly different from NT ( p ⬍ .05 or higher level of significance).
groups performed similarly, with both Tg⫹ groups being impaired in comparison with Tg⫺ mice over all three blocks of testing ( p ⬍ .005, and p ⬍ .05, respectively).
A Antibody Titers All Tg⫹ mice vaccinated with A developed circulating antibodies capable of recognizing A in enzyme-linked immunosorbent assays. The average half-maximal antibody titers (or IC50) was calculated to be 2,700⫾1,200. In contrast, no mouse injected with adjuvant possessed measurable levels of anti-A antibodies.
Discussion Our results indicate that aged, cognitively impaired APP⫹PS1 transgenic mice do not benefit behaviorally from short-term immunization with human A. Aged Tg⫹ mice given A immunotherapy over a 2–3 month period were impaired to the same extent as Tg⫹ controls in three different tasks: RAWM working memory, RAWM reference learning, and PR. These findings are in contrast with our earlier study (Morgan et al., 2000), wherein long-term A immunization of younger, unimpaired Tg⫹ mice for 8 months protected them from RAWM working memory impairment in old age. A longer period of active immunization, or perhaps passive immunization, could enhance the possibility of cognitive improvement in aged, impaired Alzheimer’s disease transgenic mice. The APP⫹PS1 mice used in this study have some amyloidbearing plaques by 6 months of age and exhibit a progressive increase in brain A deposition; marked 20 –50% A burdens are evident by 16 months (Gordon et al., 2001). We have previously found that at 6 months old these Tg⫹ mice are behaviorally normal in a variety of tasks, including RAWM working memory (Arendash, King, et al., 2001). By 16 months, however, they are impaired in both Morris water maze reference learning and RAWM working memory, with the later impairment being closely correlated with the extent of brain A deposition (Arendash, King, et al., 2001; Gordon et al., 2001; Morgan et al., 2000). In the present study, pretreatment RAWM testing of similarly aged 16-
month Tg⫹ mice revealed a clear impairment in T5 working memory, which confirms our earlier reports (Arendash, King, et al., 2001; Morgan et al., 2000). Following 2 months of treatment, A-immunized Tg⫹ mice remained as impaired as Tg⫹ controls in RAWM working memory, with both Tg⫹ groups making significantly more T4 errors overall. In posttreatment T5 working memory, however, both Tg⫹ groups were statistically no different overall from Tg⫺ mice. This lack of overall T5 impairment by Tg⫹ mice (posttreatment) reflects their greater T5 error reduction between pre- and posttreatment testing compared with that exhibited by Tg⫺ mice (4.3–3.2 errors vs. 2.8 –2.2 errors, respectively). An additional determinant of RAWM performance is the extent of cognitive saving between pre- and posttreatment RAWM testing. Tg⫺ mice showed clear evidence of cognitive saving (i.e., improved or low-error stabilized performance) between these RAWM test periods separated by 2 months. For example, their overall T4 performance was significantly better in posttreatment testing and their T5 errors at the end of both test periods was stabilized at a low number (1–2 errors). In sharp contrast, neither Tg⫹/A nor Tg⫹ controls exhibited cognitive savings between RAWM test sessions. Their performance was largely stabilized but at a relatively high number of errors. Thus, aged Tg⫹ mice could not improve their poor RAWM performance across several test periods, and short-term A immunotherapy did not provide aged Tg⫹ mice with cognitive savings. Although not reported as such, we have previously observed cognitive saving for Tg⫺ mice in RAWM testing done at 11.5 and again at 15.5 months (Morgan et al., 2000). Between these two test periods, Tg⫹ control mice became substantially impaired, whereas Tg⫹ mice given A immunizations since 8 months of age were protected from impairment. Thus, A immunotherapy has the ability to provide a degree of cognitive savings to Tg⫹ mice when begun before the onset of cognitive impairment. Similar to the RAWM data, results from the PR task revealed impaired performance of aged Tg⫹ mice and no beneficial effect of short-term A immunotherapy. As a task requiring recognition of a variably placed visible platform to escape from the water, the PR task has an appreciable cognitive component (de Bruin, Sa`chez-Santed, Heinsbroek, Donker, & Postmes, 1994; Lindner, Plone, Schallert, & Emerich, 1997). Indeed, we have found performance in this task to correlate closely with performance in other cognitive tasks (Arendash & King, 2002; King & Arendash, 2002). In both pre- and posttreatment testing, Tg⫺ mice were quickly able to change from the spatial working memory strategy of RAWM to the recognition strategy of PR. By contrast, both Tg⫹/A and Tg⫹ control mice showed equal impairment in shifting strategies, as evidenced by significantly higher escape latencies compared with Tg⫺ mice in both pre- and posttreatment PR testing. Nonetheless, all groups had similar low escape latencies by the last day of testing, suggestive of no group differences in visual acuity, motivation, or attention. Nonetheless, it is possible that noncognitive aspects of this task (unlikely to be modified by A immunotherapy) contributed to the longer period needed for Tg⫹ mice to switch from a spatial to a recognition strategy. In RAWM reference learning evaluated at approximately 20 months of age and after 3 months of treatment, aged Tg⫺ mice were nearly flawless in making very few errors overall. However, aged Tg⫹ mice exhibited impaired RAWM reference learning in committing substantially more errors than Tg⫺ mice. These results
-AMYLOID VACCINATION IN APP⫹PS1 MICE
are consistent with those of our prior study (Arendash, Gordon, et al., 2001), wherein we found 16-month (but not 6-month) APP⫹PS1 mice to be impaired in similar Morris water maze reference learning. In the present study, an extended 3 month period of A immunization did not benefit aged Tg⫹ mice in RAWM reference learning because they were equally impaired to Tg⫹ controls. Their old age at behavioral testing (20 months), profound A pathology at that age, and late-starting immunotherapy are all probable factors in not observing improved performance in this reference learning task. In that context, Janus et al. (2000) repeatedly tested TgCRND8 transgenic mice for Morris maze reference learning between 3– 6 months of age and found that A immunization begun at 6 weeks of age provided behavioral protection from early impairment in this task. Thus, A immunization can apparently protect against spatial reference learning impairment when started early in adulthood but not (as the present study found) when started at an age when memory impairments are already manifest. Serologic analysis after the completion of all behavioral testing at 20 months of age indicated that five A immunizations given to Tg⫹ during the preceding 3 months modestly elevated A antibody titers (IC50 ⫽ 2,700⫾1,200). Serum from control Tg⫹ mice (and Tg⫺ mice) had no anti-A activity, even at final serum dilutions of 1:128, thus indicating Tg⫹ mice do not spontaneously generate antibody reactivity to their own endogenous A. The modest elevation in antibody titers in aged A-immunized Tg⫹ mice is substantially below the high titers (IC50 ⫽ 27,000⫾5,000) we found in 16-month Tg⫹ mice that had been given A immunizations for the preceding 8 months (Morgan et al., 2000). The presently reported lower antibody levels in aged Tg⫹ mice probably reflect not only a shorter period of A immunotherapy but also treatment initiation in old age (16.5 months), when an immune response is slower to develop. Though not within the scope of this behavioral study, it is noteworthy that the aged Tg⫹ mice given 3 months of A immunotherapy exhibited enhanced microglial activation, as indicated by CD45 expression (Wilcock et al., 2001). Along this line, one possible mechanism for the behavioral benefits previously reported with A immunization (Janus et al., 2000; Morgan et al., 2000) involves enhanced microglial clearance of deposited A through A antibody stimulation of Fc receptormediated phagocytosis (Bard et al., 2000). Other potential mechanisms to explain behavioral benefit through A antibody production include direct antibody prevention of brain A aggregation (Solomon, Koppel, Frankel, & Hana, 1997; Solomon, Koppel, Hana, & Katzav, 1996) and enhanced A clearance from the brain through plasma sequestration of A (DeMattos et al., 2001). In the present study, there are several possible reasons to explain the inability of A immunotherapy to improve the cognitive performance of aged, cognitively impaired Tg⫹ mice. First, only modest A antibody titers were present at the end of posttreatment behavioral testing, therefore the integrated antibody exposure of these mice was likely to be relatively low. Second, because brain A levels and/or deposition in aged APP⫹PS1 were very high even months prior to posttreatment behavioral testing at 18.5–20 months (Gordon et al., 2001; Morgan et al., 2000; Wilcock et al., 2001), it is possible that neural circuitry was already irreversibly impaired. Because a longer period of A immunotherapy induces a much higher A antibody response (Wilcock et al., 2001), aged Tg⫹ mice may have benefited behaviorally from a longer period
483
of immunization than the 3 months provided in this study. Alternatively, direct administration of A antibodies (passive immunization) may improve the poor cognitive performance of Tg⫹ mice because of the advantage of immediate increases in plasma A antibody titers. In this context, a single injection of the monoclonal antibody m266 to 11-month-old PDAPP transgenic mice has been reported to result in improved object recognition and a decreased number of holeboard errors (Dodart et al., 2002). This immediate behavioral benefit of A antibody treatment did not involve decreases in brain A deposition and may involve a generalized mechanism, such as increased cerebral blood flow. The inability of active immunization in the present study to provide cognitive benefit contrasts with active–passive immunization studies (Dodart et al., 2002; Janus et al., 2000; Morgan et al., 2000) that reported cognitive benefit. This may reflect differences in (a) behavioral tasks used (e.g., different cognitive domains tested), (b) transgenic lines used, (c) type of immunotherapy, and (d) initiation point and length of treatment. In summary, the present study indicates that A active immunization for several months does not improve the cognitive performance of aged, cognitively impaired APP⫹PS1 mice. Although there are several possible explanations (e.g., submaximal antibody levels), irreversible damage to neuronal circuitry by A must be presently considered a prime factor working against immunotherapy-induced behavioral benefit in aged Tg⫹ mice. In that context, our results suggest that immunotherapy may be most beneficial to Alzheimer’s individuals very early in the disease and may require prolonged exposure to benefit symptomatic Alzheimer’s patients.
References Arendash, G. W., Gordon, M., Diamond, D., Austin, L., Hatcher, J., Jantzen, P., et al. (2001). Behavioral assessment of Alzheimer’s transgenic mice following long-term A-beta vaccination: Task specificity and correlations between A-beta deposition and spatial memory. DNA and Cell Biology, 20, 737–744. Arendash, G. W., King, D., Gordon, M., Morgan, D., Hatcher, J., Hope, C., & Diamond, D. (2001). Progressive, age-related behavioral impairments in transgenic mice carrying both mutant amyloid precursor protein and presenilin-1 transgenes. Brain Research, 891, 42–53. Arendash, G. W., & King, D. L. (2002). Intra- and intertask relationships in a behavioral test battery given to Tg2576 transgenic mice and controls. Physiology and Behavior, 75, 643– 652. Bard, F., Cannon, C., Barbour, R., Burke, R., Games, D., Grajeda, H., et al. (2000). Peripherally administered antibodies against amyloid beta peptide enter the central nervous system and reduce pathology in a mouse model of Alzheimer’s disease. Nature Medicine, 6, 916 –919. de Bruin, J. P., Sa`nchez-Santed, F., Heinsbroek, R. P., Donker, A., & Postmes, P. A. (1994). Behavioural analysis of rats with damage to the medial prefrontal cortex using the Morris water maze: Evidence for behavioural flexibility, but not for impaired spatial navigation. Brain Research, 652, 323–333. DeMattos, R., Bales, K., Cummins, D., Cosme, J., Paul, S., & Holtzman, D. (2001.) Peripheral anti-A antibody alters CNS and plasma A clearance and decreases brain A burden in a mouse model of Alzheimer’s disease. Proceedings of the National Academy of Sciences, USA, 98, 8850 – 8855. Dickey, C. A., Morgan, D. G., Kudchodkar, S., Weiner, D., Bai, Y., Cao, C., Gordon, M. N., & Ugen, K. E. (2001). Duration and specificity of humoral immune response in mice vaccinated with the Alzheimer’s
484
AUSTIN ET AL.
disease-associated -amyloid 1– 42 peptide. DNA and Cell Biology, 20, 723–729. Dodart, J.-C., Bales, K. R., Gannon, K. S., Greene, S. J., DeMattos, R. B., Mathis, C., et al. (2002). Immunization reverses memory deficits without reducing brain A burden in Alzheimer’s disease model. Nature Neuroscience, 5, 452– 457. Gordon, M., King, D., Diamond, D., Jantzen, P., Boyett, K., Hope, C., et al. (2001). Correlation between cognitive deficits and A deposits in transgenic APP⫹PS1 mice. Neurobiology of Aging, 22, 377–385. Hardy, J. (1997). Amyloid, the presenilins and Alzheimer’s disease. Trends in Neurosciences, 20, 154 –159. Janus, C., Pearson, J., McLaurin, J., Mathews, P., Jiang, Y., Schmidt, S., et al. (2000, December 21–28). A peptide immunization reduces behavioral impairment and plaques in a model of Alzheimer’s disease. Nature, 408, 979 –982. King, D. L., & Arendash, G. W. (2002). Behavioral characterization of the Tg2576 transgenic model of Alzheimer’s disease through 19 months. Physiology and Behavior, 75, 627– 642. Lindner, M., Plone, M., Schallert, T., & Emerich, D. (1997). Blind rats are not profoundly impaired in the reference memory Morris water maze and cannot be clearly discriminated from rats with cognitive deficits in the cued platform task. Cognitive Brain Research, 5, 329 –333. Lippa, C., Nee, L., Mori, H., & St. George-Hyslop, P. (1998). Abeta-42 deposition precedes other changes in PS-1 Alzheimer’s disease. Lancet, 352, 1117–1118. Morgan, D., Diamond, D., Gottschall, P., Ugen, K., Dickey, C., Hardy, J., et al. (2000, December 21–28). A peptide vaccination prevents memory loss in an animal model of Alzheimer’s disease. Nature, 408, 982–985.
Schenk, D. B., Seubert, P., Lieberburg, I., & Wallace, J. (2000). -peptide immunization: A possible new treatment for Alzheimer disease. Archives of Neurology, 57, 934 –936. Selkoe, D. (2001) Alzheimer’s disease: Genes, proteins and therapy. Physiological Reviews, 81, 741–766. Sigurdsson, E. M., Scholtzova, H., Mehta, P. D., Frangione, B., & Wisniewski, T. (2001). Immunization with a nontoxic/nonfibrillar amyloid-beta homologous peptide reduces Alzheimer’s diseaseassociated pathology in transgenic mice. American Journal of Pathology, 159, 439 – 447. Solomon, B., Koppel, R., Frankel, D., & Hana, E. (1997). Disaggregation of Alzheimer beta-amyloid by site-directed mAb. Proceedings of the National Academy of Sciences, USA, 94, 4109 – 4112. Solomon, B., Koppel, R., Hana, E., & Katzav, T. (1996). Monoclonal antibodies inhibit in vitro fibrillar aggregation of the Alzheimer beta amyloid peptide. Proceedings of the National Academy of Sciences, USA, 93, 452– 455. Weiner, H., Lemere, C., Maron, R., Spooner, E., Grenfell, T., Mori, C., et al. (2000). Nasal administration of amyloid beta peptide decreases cerebral amyloid burden in a mouse model of Alzheimer’s disease. Annals of Neurology, 48, 567–579. Wilcock, D. M., Gordon, M. N., Ugen, K. E., Gottschall, P. E., DiCarlo, G., Dickey, C., et al. (2001). Number of Abeta inoculations in APP⫹PS1 transgenic mice influences antibody titers, microglial activation, and congophilic plaque levels. DNA and Cell Biology, 20, 731–736.
Received July 8, 2002 Revision received December 9, 2002 Accepted December 18, 2002 䡲
