Successful long-term treatment for Alzheimer's disease (AD) has remained as elusive as the disease's cause. Lifestyle treatments for AD, such as exercise, are being researched in both animal and human models 12345. Exercise has been found to be a successful preventative measure in delaying disease onset. Exercise interventions in patients who already exhibit AD symptoms have had mixed results, but are able to show improvement to varying degrees 2345.

The pathology in AD is well-known. Aβ and tau deposits are the chief pathological hallmarks. Aβ has been implicated in impairing learning and memory 678910. Aβ deposition stimulates a local immune response by the microglia, which become macrophagic 111213. The macrophagic phenotype in AD is characterized by the presence of CD11b and immunohistochemistry revealing attempted phagocytosis of Aβ 11141516. This response is not sufficient to clear Aβ deposits and instead contributes to the chronic progression of AD 171819. Chronically activated microglia are characterized by their macrophagic morphology, CD11b expression, and release of neurotoxic cytokines IL-1β and TNF-α 2021. This response of microglia seems to contribute to AD progression, rather than clearing Aβ pathology 222324.

However, the immune responses possible in the brain are more complex than was once thought. Aside from the initial innate response, characterized by macrophagic microglia and cytokine production, adaptive responses in which antigen is presented to T cells infiltrating from the periphery has been increasingly observed in various neurodegenerative disease states and in Aβ vaccination studies 1425262728. In addition, exciting new evidence exists of an alternative immune response, characterized by perivascular cells that participate in both innate immunity, via phagocytosis and cytokine production, as well as adaptive immunity via antigen presentation and co-stimulation at the blood brain interface 293031. Evidence indicates that the AD brain is capable of clearing Aβ if the microglial response is manipulated 32. For instance, injected bone marrow cells travel to the brain and differentiate into microglia. These newly differentiated microglia acquire a dendritic, rather than macrophagic phenotype and are associated with antigen presentation and a decrease in Aβ 32. Further, microglia stimulated with glatiramer acetate or IFN-γ change phenotype and become antigen presenting cells 333435. This antigen presenting phenotype stimulates the adaptive immune response, characterized by increased major histocompatibility complex II (MHCII) expression and CD11c positive cells 3436. Butovsky et al.(2007) and Ziv et al (2006) have also shown this shift to the adaptive immune response coincides with decreases in cytokines associated with the innate immune response, namely IL-1β and TNF-α 3437. The identity of the antigen presenting cells was initially unclear in these studies. However, it has recently been suggested that the clearance of Aβ is dependent on invasion of the brain by peripheral monocytes, which differentiate into antigen presenting microglia 3538. A stimulus to the immune response in late AD may confer benefit through clearance of soluble species of Aβ, as evidenced in several studies using lipopolysaccharide or Aβ immunization 394041. The role of microglia in AD is clearly complex and can be either detrimental or beneficial, depending on the nature of the microglial response.

While physical exercise can delay the onset of cognitive decline in AD, the mechanisms remain largely unknown 4243444546. Adlard et al. recently demonstrated that exercise reduced Aβ deposition in AD transgenic mice 1. We suggest that exercise may be also altering the immune response in AD, which in turn affects Aβ levels. We hypothesize that the protective effects of exercise in AD may be due in part to an immune shift from innate to antigen presenting, similar to the shift observed by Butovsky et al (2005–2007) 3335. Specifically, we hypothesize that this shift is characterized by decreases in markers of the innate response, such as TNF-α and IL-1β, as well as increases of adaptive and alternative activation response stimulating cytokine IFN-γ. We further suggest that exercise induces a microglial phenotype shift from a macrophagic phenotype to an antigen presenting phenotype, expressing CD11c and MHC II. We also expect to see increased CD40, which communicates from the antigen presenting cell to CD40 ligand on T cells. It is possible this shift in immune response could decrease Aβ species in the Tg2576 mouse, as seen in the Butovsky studies 33343538. In contrast to Butovsky, our study is unique in that we examined whether physical exercise can induce this immune shift without pharmacological intervention.

In the current study, we used the Tg2576 AD mouse model developed by Karen Hsiao 47. Tg2576 mice exhibit plaque pathology at 10 months, but show cognitive deficits as early as 6 months 848. For the present study, we were interested in using aged mice with advanced pathology and memory deficits. The older mouse with advanced pathology serves as a model for later stages of AD. We have previously demonstrated that three weeks of voluntary running improved water maze performance in aged (17–19 month) Tg2576 mice 49. In this investigation, we examine a shift in the immune response to Aβ as a potential mechanism for the behavioral improvements.

No significant differences in running distances existed between genotypes. (Table 1)

IL-1β and TNF-α protein were significantly greater in aged sedentary Tg2576 mice (TG_SED) than in sedentary wildtype (WT_SED) (IL-1β: p = 0.006; TNF-α: p = 0.04) (Figure 1). Three weeks of exercise decreased the levels of both IL-1β and TNF-α to a point where they no longer significantly differed from the WT_SED. The IL-1β in exercised Tg2576 mice (TG_RUN) was significantly lower than in the TG_SED (p = 0.01)(Figure 1). When we compare current cytokine markers to our previously published behavioral data, the decreased levels of the pro-inflammatory cytokine IL-1β observed in transgenic animals correlate with decreased mean latency to find the escape platform on day 2 of the radial arm version of the Morris water maze (correlation = 0.31, p = 0.001)49. We observed CD11b+ microglia in TG_SED and TG_RUN animals (Figure 2). Levels of microglial activation, as measured by Iba-1 western blot, did not change with exercise (Figure 2).

In contrast, IFN-γ was significantly lower in aged TG_SED than in WT_SED (p = 0.02) and showed a trend to increase in TG_RUN (p = 0.06) such that it returned to levels similar to the WT (Figure 3). Similarly, protein levels of chemokine MIP1α that tended to be lower in the TG_SED than the WT_SED (p = 0.07), were significantly increased in TG_RUNanimals (p = 0.05) to levels similar to those observed in the WT (Figure 3). The MHC II blots confirm that the IFN-γ increase observed is associated with increased antigen presentation in the TG_RUN animals, who expressed significantly more MHC II than TG_SED (p = 0.04) (Figure 4). In the TG_RUN animals, we observed abundant antigen presenting cells, indicated by CD11c (Figure 5B). In TG_RUN, we observed CD11c+ blood vessels and in individual cells that appeared linearly arranged, as if inside of blood vessels (Figure 5B,C). Vessels labeled strikingly with CD68 (red) in proximity to, but not within the same cells as CD11c (green), suggesting alternatively activated macrophages may be present (Figure 5D–F). Vessels were double labeled for Iba-1 (green) and mannose receptor (red), a marker for perivascular macrophages (Figure 5G–I). No such cells or vessels were observed using the same markers in TG_SED animals.

The presence of MHCII and CD11c labeled cells indicate antigen presentation. Increased MIP-1α levels suggest T cells and monocytes are being recruited to the brain (Man, 2007). CD40, expressed by the antigen presenting cell communicates with its ligand on T cells. In sedentary TG animals, a trend of reduced expression of CD40 is present compared to WT_SED (p = 0.10), but there is a significant elevation in CD40 from TG_SED to TG_RUN (p = 0.008) (Figure 4). Additionally, alternative activation of macrophages could explain the shift in immune markers observed 29. Indeed, if we consider all of the cytokine and chemokine changes collectively, along with the conflicting reports of T cell presence and function in the brain 27, alternative activation of macrophages may be a more likely candidate for repair in an AD state 52.

Finally, we examined whether exercise leads to reduced Aβ in aged Tg2576 mice. We quantified the levels of various Aβ species in cortex by ELISA, and in hippocampus and serum using multiplex technology as well as dot blot. Multiplex revealed no statistically significant differences in Aβ aggregates in hippocampus, though direction of change was to decrease in run (Figure 6A). There was no detectable aggregated Aβ in the serum samples and immunohistochemical staining for plaques appeared similar between TG_SED and TG_RUN (data not shown). ELISA of Aβ₄₂ in soluble and insoluble fractions from cortex did not statistically differ between TG_SED and TG_RUN though we should note that there was a large degree of variation between the RUN animals (Figure 6B–C) Total soluble levels of Aβ₄₀ did show a significant decrease (p = 0.01) in theTG_RUN animals compared to the TG_SED (Figure 6B). Soluble Aβ is composed of both fibrillar and pre-fibrillar oligomeric species51. We used a conformation specific fibrillar antibody to explore changes with exercise in the Tg2576 39.

The dot blots of hippocampal lysates are able to selectively probe different soluble Aβ species in a soluble fraction, spun at low speeds. First, we analyzed total Aβ using 6E10 on dot blots. We did not detect any significant differences between SED and RUN (Figure 7A). The OC antibody detects soluble fibrillar Aβ that is conformationally and immunologically distinct 51. Levels of soluble Aβ fibrils were 40% lower in TG_RUN compared to TG_SED as assessed by OC antibody (Figure 7B). This decrease was significant (p = 0.01).

We have demonstrated that physical exercise can shift the immune response in the brain of an AD mouse model. AD is characterized by an innate immune response characterized by microglial activation. Though likely an intended protective mechanism to phagocytose Aβ, these glia are not capable of combating chronic AD pathology 53. In addition, the cytokines released by these microglia, IL-1β and TNF-α, become cytotoxic when chronically expressed, contributing to the overall neurodegeneration 20212435545556. We confirmed this immune state in our sedentary aged TG animals. Our findings confirm previous findings that high IL-1β levels are correlated with poor cognitive performance 57 and indicate that exercise induces reductions in TNF-α and IL-1β in the hippocampus of Tg2576 mouse at advanced age.

Recent investigations into bone marrow derived cells that differentiate into microglia in the brain, in concert with studies where microglial phenotype is experimentally altered to an antigen presenting phenotype collectively suggest that if the adaptive immune response is stimulated, Aβ pathology in AD can be reduced and cognition improved 5859343538. The antigen presenting response is characterized by CD11c positive cells, MHC II, and CD40. IFN-γ has been shown to be a stimulus for this microglial phenotype and has also been found to be capable of inducing Aβ clearance 37. We demonstrate that the aged TG_SED mouse express lower levels of IFN-γ than the WT and that exercise results in increased IFN-γ and increased markers of the antigen presenting response. In addition to antigen presenting microglia from the local environment, it is also possible that exercise increases recruitment of peripheral monocytes to the brain, where they too can become antigen presenting cells. MIP-1α recruits central and/or peripheral macrophages to the area of plaque deposition in AD and MIP-1α was found to increase in TG_RUN6061. However, the possibility of peripheral recruitment is in debate, as recent data indicates that much of the evidence in favor of monocyte recruitment to the brain only occurs in the presence of damage to the blood-brain barrier due to the whole-body irradiation used in the original experiments 3859626364. The integrity of the blood brain barrier is compromised in AD and in the Tg2576 mouse model, however, so we cannot rule out either a peripheral or central source for these microglia 656667. Though the origin or the microglia will be a topic for continued investigation, we do demonstrate that the microglia present in the hippocampus after exercise in the aged TG mice exhibit characteristics of the adaptive immune response and antigen presentation.

Another interesting possibility presented by the vascular localization of macrophagic markers observed in TG_RUN animals is the alternative activation of perivascular cells as macrophages and antigen presenting cells. Though microglia are the initial immune responders in the CNS, perivascular cells may be more efficient at both phagocytosis of antigen and presentation of that antigen than adult microglia 2568. In particular, our data revealing extensive mannose receptor labeling in TG_RUN animals strongly indicates perivascular macrophages are involved in the immune response, as this receptor is not expressed by microglia 31.

We suggest that the antigen being presented is Aβ or an Aβ fragment. Fibrillar Aβ is a highly cytotoxic species that has been shown to adhere to microglia 6970. If microglia are stimulated to take on an antigen presenting phenotype, fibrillar Aβ can be cleared from the brain 53. In the current study, soluble Aβ₄₀ and soluble fibrillar Aβ significantly decreased after three weeks of running coincident with the increase in adaptive immune markers and the improvement in behavior 49. We also observed CD11c adjacent to mannose receptor and CD68 positive cells within the vasculature of TG_RUN. This presence of perivascular macrophages in conjunction with markers of antigen presentation might be indicative of clearance of Aβ into the periphery, supporting the "peripheral sink" hypothesis, in which Aβ is proposed to efflux from brain into plasma after immunization therapy 717273. Indeed, one proposed function of the mannose receptor which we observed in TG_RUN is antigen transport 31. Vasilevko et al. (2007) recently showed that direct immunotherapy with anti-Aβ antibodies resulted in decreased diffuse Aβ deposits in brain, but increased Aβ₄₀ in plasma 40.

In this study, we present evidence that exercise decreases the chronic pro-inflammatory response well known to associate with AD. The decrease in at least one of the cytokines, IL-1β, correlates with improved ability to solve a water maze task. Unlike ibuprofen and other anti-inflammatory drugs used in AD treatment, which decrease pro-inflammatory markers, the current study provides evidence that exercise not only decreases pro-inflammatory markers like IL-1β and TNF-α, but also increases adaptive inflammatory markers IFN-γ, CD40, MHC II, CD11c, and MIP-1α. We observed multiple macrophage markers (CD68, mannose receptor) in and around the vasculature of TG_RUN animals. Further, we have shown that exercise decreases soluble Aβ₄₀ and soluble fibrillar Aβ. We support the hypothesis put forward by Butovsky et al (2006) that the innate immune response in AD becomes dysfunctional after chronic activation, but that an initiation of the adaptive or alternative immune response may reduce pathology 35. The decrease in fibrillar Aβ and improvement in behavior observed in our investigations of the Tg2576 mouse at 17–19 months of age suggests that physical exercise can trigger this immune shift in late stages of AD, leading to cognitive improvement 49.

Our data suggests that exercise intervention may pave the way for successful Aβ clearance even in late stages of AD. We suggest that exercise not only decreases neurotoxic cytokines, but also increases Aβ clearance through its induction of an adaptive or alternative immune response. In summary, our investigation opens a potential avenue for behavioral interventions that would be complementary to long term NSAIDs or anti-inflammatory drugs.

WT: Wildtype; TG: Transgenic; SED: Sedentary; AD: Alzheimer's disease; IL: Interleukin; TNF: Tumor necrosis factor; MHC: Major histocompatibility complex; IFN: Interferon; MIP: Macrophage inflammatory protein; CNS: Central nervous system; NSAIDS: Non-steroidal anti-inflammatory drugs.

CGG is a consultant for Kinexis, Inc. The remaining authors declare that they have no competing interests.

KEN designed and implemented the investigation and was responsible for data analysis, interpretation, and preparation of the manuscript. WWP performed the for Aβ dot blots. AIP aided in study design and preparation of tissue for histology. DHC provided Aβ ELISA data and contributed to the direction of the investigation and data interpretation. CGG created the unique antibodies used to assess Aβ in the dot blot analysis. CWC contributed to the writing of the manuscript and the data interpretation.