Transient pharmacologic lowering of Aβ production prior to deposition results in sustained reduction of amyloid plaque pathology

Background Alzheimer’s disease (AD) is the leading cause of dementia among the elderly. Disease modifying therapies targeting Aβ that are in development have been proposed to be more effective if treatment was initiated prior to significant accumulation of Aβ in the brain, but optimal timing of treatment initiation has not been clearly established in the clinic. We compared the efficacy of transient pharmacologic reduction of brain Aβ with a γ-secretase inhibitor (GSI ) for 1–3 months (M) treatment windows in APP Tg2576 mice and subsequent aging of the mice to either 15M or 18M. Results These data show that reducing Aβ production in a 2-3M windows both initiated and discontinued before detectable Aβ deposition has the most significant impact on Aβ loads up to 11M after treatment discontinuation. In contrast, initiation of treatment for 3M windows from 7-10M or 12-15M shows progressively decreasing efficacy. Conclusions These data have major implications for clinical testing of therapeutics aimed at lowering Aβ production, indicating that; i) these therapies may have little efficacy unless tested as prophylactics or in the earliest preclinical stage of AD where there is no or minimal Aβ accumulation and ii) lowering Aβ production transiently during a critical pre-deposition window potentially provides long-lasting efficacy after discontinuation of the treatment.


Background
Alzheimer's disease (AD) is the leading cause of dementia among the elderly. Numerous studies in humans have demonstrated the sequential development of various pathological features that characterize AD and the relationship of these pathologies to diagnosis of dementia [1]. These studies support the amyloid cascade hypothesis which posits that accumulation of Aβ aggregates triggers a series of downstream pathologies that results in clinical AD [2]. AD has recently been proposed to have three pre-clinical stages, characterized in sequence by the development of Aβ pathology (stage 1), the presence of plaque pathology, neurofibrillary pathology and initial evidence for neurodegeneration (stage 2), and finally by the presence of stage 2, pathologies plus evidence for subtle cognitive decline (stage 3). This progression likely occurs over a 10-20 year timeline [3].
The amyloid cascade hypothesis predicts that attenuating or preventing Aβ aggregation and accumulation will attenuate or prevent the subsequent development of AD. Strong experimental support for the amyloid hypothesis has supported the rationale for many therapeutic approaches targeting Aβ. Major investments have been made to develop inhibitors or modulators of Aβ production [4]. γ-secretase inhibitors (GSIs), β-secretase inhibitors (BSIs) and γ-secretase modulators (GSMs) have been or are currently being tested in human AD trials. Both the GSM tarenflurbil (R-flurbiporfren), and the GSI semagacestat, failed to show efficacy in phase III human trials [4,5]. Several other GSMs, GSIs, and BSIs are currently in early stage trials. Based on the clinical failures of tarenflurbil and semagacestat as well as other anti-Aβ therapies not targeting Aβ production but its aggregation or clearance, we and others have begun to question whether targeting of Aβ in symptomatic AD may be a futile endeavor [6].
Given the framework provided by the amyloid cascade hypothesis, disappointing results from completed trials of anti-Aβ therapies in patients with AD, and several but limited preclinical studies supporting the concept that when targeting Aβ "earlier is better" [7][8][9], we explored the relationship between long-term efficacy with respect to reducing amyloid loads and timing and duration of Aβ lowering. In designing these studies, we were struck by the parallels between Aβ accumulation in the mouse brain and in vitro studies of Aβ aggregation. In vitro, Aβ aggregates in a nucleation dependent polymerization reaction, which exhibits three phases: a lag phase in which nucleation occurs but not fibril formation, an exponential growth phase in which fibril formation and growth occur rapidly, and a plateau phase where fibril growth and formation slows [10]. Except for being more extended in terms of time, studies in APP Tg2576 transgenic mice show that Aβ aggregation and accumulation follows a similar course in the brain [11]. There is an initial lag phase where no Aβ aggregation and accumulation occurs, followed by an exponential accumulation phase, and finally when amyloid loads become AD-like, the growth slows or plateaus (see Figure 1A). Given these parallels, we tested the efficacy of transient lowering of Aβ production in 1-3M treatment "windows" corresponding to these various phases of deposition using the GSI LY-411,575 (LY) in APP Tg2576 mice. These data show that GSI treatment during the pre-deposition window (4-7M) before the early exponential phase of deposition (7-10M) or late exponential phase (12-15M) has by far the most efficacy in terms of reduction of plaque loads at 15M of age.

Results and discussion
In vitro and in vivo dosing studies with Ly-411,575 Ly-411,575 is potent orally bioavailable γ-secretase inhibitor [12]. We synthesized large quantities for these studies and then performed both in vitro and in vivo studies to validate Ly-411,575 potency. The IC50 in cell culture for LY-411,575 was between 1-3 nm (see Additional file 1: Figure S1A). Six hours after a single IP dose, Aβ40 levels are reduced by~75% in the brain of non-depositing Tg2576 mice (Additional file 1: Figure  S1B). Oral administration (10 mg/kg/day) administered as a single dose reduces brain Aβ40 levels 24 hours after the final dose (7 days) by~65% (Additional file 1: Figure S1C) and plasma Aβ40 by~95% (Additional file 1: Figure S1D). Plasma Aβ levels are reduced more quickly than brain levels following oral dosing with maximal inhibition seen after 3 days (Additional file 1: Figure S1D). For long term in vivo mouse studies, we tested administration of Ly-411,575 using various dosing strategies and time points (Table 1). Oral administration (formulated in rodent chow) continually for 2 weeks with either -1 mg/kg/day or 2.5 5 mg/kg/day reduced brain Aβ40 levels by~50% and plasma Aβ40 by~80%, without overt toxicities ( Table 1). However, continuous oral dosing (>2 weeks) using the higher dose of 5 mg/kg/day or 10 mg/kg/day resulted in overt toxicities in mice (Table 1). Therefore, in all long term studies presented in this manuscript, we have used the 2.5 mg/kg/day dose formulated in rodent chow.

Transient Aβ reduction dramatically reduces the subsequent accumulation of amyloid plaques
For these experiments, we transiently dosed Tg2576 mice with LY from 4-7M, 7-10M or 12-15M and aged treated mice to 15M (see Figure 1B for experimental design). To confirm efficacy of LY in all groups, we measured plasma Aβ levels immediately after the last day of treatment in sentinel mice and showed a similar reduction in plasma Aβ levels in all cohorts ( Table 2). Following sacrifice at 15M, levels of Aβ in the brain were examined. The 4-7M LY treatment significantly reduced Aβ deposition; plaque burden in the frontal cortex and hippocampus was decreased by~68% ( Figure 1C) and FA-solubilized brain Aβ levels were reduced by~60% ( Figure 1D). The 7-10M LY treatment non-significantly decreased Aβ plaque burden by~19% compared to controls ( Figure 1C), whereas FA-solubilized Aβ levels were reduced by~34% ( Figure 1D). The 12-15M GSI treatment group had no significant effect on either plaque burden or FA-solubilized Aβ levels ( Figure 1C, D). We further evaluated effect of the 4-7M LY treatment on cored amyloid plaques and cerebral amyloid angiopathy (CAA). Compared to untreated control mice, there were significant reductions both in cored plaques in the frontal cortex and hippocampus (~47% reduction) and CAA in the leptomeninges (~41% reduction, Figure 2A, B). To further determine whether the magnitude of the effect diminished with further aging, we performed an additional experiment, where we aged 4-7M LY treated mice to 18M, and again we observed significantly reduced Aβ plaque burden and FA-solubilized Aβ levels ( Figure 2C), demonstrating that the suppression of deposition was maintained even up 11 months after treatment was halted.
Because Aβ deposition in mice follows an exponential course, we sought to determine how the observed reductions in Aβ accumulation correlate with Aβ accumulation to untreated mice. We compared Aβ levels in the brains of untreated Tg2576 mice from various ages ranging from 10-18 months and plotted the natural logarithm transformed value of these against Aβ levels from the 4-7M LY treated mice, aged to 15M and 18M ( Figure 2D). To minimize variance, we assayed all of the samples in a single set of ELISA studies. This analysis showed that the 4 -7M GSI treatment time frame "shifts" amyloid depositing by approximately 3 months, essentially delaying Aβ accumulation by a time roughly equivalent to the treatment window.
We next attempted to determine whether this 4-7M narrow therapeutic "window" can be further refined. To do this, we transiently dosed Tg2576 mice with LY for 1, 2 or 3M intervals between the 4-7M age time frame and again aged the mice to 15M. Aβ levels were again significantly reduced in this second 4-7M LY cohort (~59% reduction) at 15M of age compared to controls ( Figure 2E). Similarly, i) the 4-6M LY cohort showed a significant reduction in Aβ levels (~51% reduction) and; ii) the 4-5M LY group showed a modest but nonsignificant reduction in Aβ levels (~24% reduction) ( Figure 2E). However, in the remaining 5-6M, 6-7M or 5-7M LY cohorts tested, none showed any significant reductions in Aβ levels at 15M compared to controls ( Figure 2E). Since we show that the 4-7M LY "shifts" amyloid deposition equivalently to the duration of treatment, we wondered whether we will see a similar shift in amyloid deposition in the shorter treatment intervals examined that showed efficacy, e.g., in the 4-5M and 4-6M cohorts. Again, we compared Aβ levels in the brains of untreated Tg2576 mice from various ages ranging from 10-15 months against Aβ levels from the 4-5M, 4-6M, and 4-7M LY treated cohorts aged to 15M ( Figure 2F). This analysis seems to show that the 4-5M GSI treatment "shifts" amyloid depositing by approximately 1 month, however, both the 4-6M and 4-7M treatment intervals essentially show a similar shift in Aβ accumulation, namely by roughly 3 months ( Figure 2F). These data would suggest, at least within the 4-7M time frame that we have studied, that the 4-6M age interval may play a critical role in the early seeding phases of amyloid deposition. Additional studies, using either earlier treatment times (e.g. starting at 3M of age) or longer treatment intervals may be necessary to fully appreciate the usefulness of this early "therapeutic window" to achieve the best possible outcome on subsequent amyloid deposition.

Transient GSI does not permanently alter APP levels or processing
To determine if the 4-7M LY treatment had longlasting effects on Aβ production or APP levels and processing, we conducted a number of studies. We measured plasma Aβ levels at termination of GSI treatment and then determined how long before steady state levels of Aβ were normalized. Plasma Aβ levels were significantly lower immediately after treatment, but returned to control levels between with 1-2 weeks after treatment was halted ( Figure 3A). We also measured levels of full-length APP and APP C-terminal fragments (CTFs) in the brain at the end of the 4-7M treatment and at 15M. Immediately following discontinuation of treatment, APP levels were unchanged but APP CTFs were increased ( Figure 3B,D). However, at 15 months, there were no significant differences in full length APP or CTFs ( Figure 3C,D). Finally, as GSI treatment has been shown to affect peripheral lymphocytes [13], we performed FACS analysis and quantified both B and T cell numbers in the spleen after the 4-7M LY transient treatment and at 15M. B and T cell numbers were not affected ( Figure 4A). However, when  we analyzed Th1 and Th2 cytokine profiles, splenic CD4+ T cells from the 4-7M LY treated cohort showed a bias towards Th1 polarization with increased IFN-γ secretion (a canonical Th1 cytokine) and concurrent decreases in IL-4 levels (a signature Th2 cytokine), which persisted even after the treatment was halted for 8 months (Figure 4B, C,). The significance of this longlasting effect on peripheral CD4 + T cell immune responses is not clear, but certainly warrants further investigation.

Conclusion
Our current data show that transiently lowering Aβ production in Tg2576 mice during the pre-deposition "seeding" phase has a major impact on subsequent Aβ accumulation with the effect persisting for at least 11M. Later treatment windows showed decreasing efficacy of treatment. These current data have major implications for trial design targeting Aβ production in humans. Indeed, they would suggest that efficacy in terms of plaque reduction is likely to be maximal during seeding phase, and rapidly decline if treatment is initiated during the exponential phase of deposition. A recent report [14] using passive immunotherapy provided more limited evidence that a similar effect may be observed with anti-Aβ passive immunotherapy initiated during the early deposition phase persisting for 3 months after discontinuation of treatment. Given the plethora of anti-Aβ therapies currently in development, it will be essential to determine for each modality whether efficacy is both maintained after discontinuation and similarly influenced by timing of the treatment. The near equivalence of the delay in subsequent deposition to the length of the treatment window during a pre-deposition phase suggests that the GSI treatment from 4-7M alters or reverses seeding of Aβ. Indeed, there are no plaques forming during this period of time. Future studies will be necessary to determine the exact mechanisms by which lowering Aβ in this window alters subsequent deposition. As even in vitro detection of Aβ nucleation relies on indirect assays that have limited sensitivities, these studies will likely require a better understanding of nucleation and better assays to detect it. Developing such assays will be critical as direct translation of these studies would be facilitated by defining the equivalent window in humans. Regardless of mechanism, these preclinical window therapy studies have major implications regarding the potential of prophylactic therapy targeting Aβ production in humans. Not only do these data suggest that prophylactic therapy in the appropriate window will be more efficacious then initiating therapy in individuals with any preexisting Aβ pathology, but they suggest that such therapy, even if discontinued might delay subsequent onset AD by a time roughly equivalent to the time of treatment.

Tg2576 mice
APP Tg2576 mice were generated, maintained and genotyped as described previously [7]. All animal husbandry procedures performed were approved by the Mayo Clinic Institutional Animal Care and Use Committee in accordance with National Institutes of Health guidelines. All animals were housed three to five to a cage and maintained on ad libitum food and water with a 12 h light/dark cycle.

LY-411, 575
LY-411, 575 (LY) was synthesized at the Mayo Clinic Chemical Core as previously described [15]. LY at a  Tg2575 mice (n = 7-15/group/experiment) were transiently treated with LY chow at 1, 2 or 3 months interval (between 4-7 months of age), from 7-10 months of age and from 12-15 months of age. Untreated control mice from each age group were as aged and used for analysis. Groups of mice (n =3) were sacrificed immediately after the end of each treatment time-point to access efficacy of LY inhibition. After treatment was halted, mice were then aged till 15 months or 18 months of age and then sacrificed for Aβ analysis. Chow consumption, general health and body weight were monitored on a weekly basis in all treatment groups.

Immunohistological analysis
Sagittal sections of paraffin embedded sections were used for analysis. Immunohistochemical staining was done using pan-Aβ antibody (mAb 33.1.1; Mayo), mAb 13.1.1 (human Aβx-40 specific, binds cored plaques and CAA; Mayo) as previously described [17]. Immunohistochemically stained sections were captured using the Scanscope XT image scanner (Aperio, Vista, CA, USA) and analyzed using the ImageScope program. Aβ plaque burden was then calculated using the Positive Pixel Count program (Aperio). At least 5 sections/brain 30 μm apart, were used and averaged by a blinded observer to calculate plaque burden.