Experimental design and novelty of the study
In the present studies, we aimed to elucidate the interaction between activated glia and neurons in the earliest stages of the AD pathology. Several reports in the field have documented the effect of anti-inflammatory treatments with NSAIDs in Tg models of AD-like pathology [40–42]. However, these studies were designed to assess the effect of treatment on plaque pathology, and the animals were sacrificed after plaque onset. Our approach differed from previous reports in that we wanted to clarify the role of inflammation in the early, pre-plaque stages of the pathology, which likely mimic the earliest, pre-clinical stages in AD.
This is a matter of high clinical relevance, as the available data in humans suggest a differential role of inflammation in early versus late stages of the disease. The contribution of the inflammatory process in disease onset has been highlighted by epidemiological, retrospective studies indicating a lower incidence of AD in populations receiving long-term treatment with NSAIDs [43–46]. In contrast, prospective trials applying NSAIDs to patients clinically diagnosed with AD have failed to reverse or slow down the disease, often worsening it [12–14]. Taken together, the clinical evidence is consistent with the concept that inflammation would contribute to and accelerate the AD neuropathology in its early pre-clinical stages, while it would be neutral or even beneficial in later, clinical stages. It is important to note that this hypothesis has been supported by the recent extended results from an AD anti-inflammatory preventive trial (ADAPT). The study had to be halted after only two years for safety concerns ; at that point in time, no beneficial effects were observed with the anti-inflammatory treatment . However, the follow-up results indicated that the naproxen treatment outcome critically depended on the stage of the pathology at the moment of enrolment in the trial [49, 50]. In this prospective study, in fact, an increased risk of AD onset was detected in patients who displayed some cognitive impairment (but no dementia) when they entered into the study. Such a cohort would likely represent patients closer to the disease onset. On the other hand, asymptomatic individuals treated with NSAIDs had a reduced AD incidence . The concept of a differential role of inflammation at early versus late stages of the disease has been suggested for other neurodegenerative conditions, such as amyotrophic lateral sclerosis , and it might very well be related to the differential subsets of monocytic cells involved . Therefore, the temporal window in which anti-inflammatory treatment can be beneficial and the mechanism involved in such a beneficial effect need to be further studied. We propose that pre-plaque Tg mice can be a valuable model for this type of investigation.
Early microglial activation and its inhibition with minocycline
In agreement with our previous observations , we gathered biochemical (Figure 1A-C) and morphological (Figure 1D-H) evidence indicating the presence of microglial activation in the cerebral cortex and hippocampus at pre-plaque stages in the transgenic mouse model McGill-Thy1-APP.
To investigate the pathological participation of such pro-inflammatory process in the AD-like amyloid pathology, we chose to administer the tetracyclic derivative minocycline. In addition to its antimicrobial activities, this drug easily crosses the blood-brain barrier and has been shown to be beneficial in several CNS neuropathological conditions and in neurodegeneration . Minocycline appears to exert its action through a plethora of mechanisms, including inhibition of key inflammatory enzymes (such as iNOS, matrix metalloprotease 9 and 5-lipoxygenase), blocking caspase-dependent and independent apoptosis and demonstrating anti-oxidant effects (for a review, see ).
Previous reports have studied the effect of minocycline on the full-blown amyloid pathology in APP Tg mice [28, 54–56]. Overall, the drug appeared to reduce neuroinflammation and the behavioral deficits observed in Tg mice. However, in all the previous reports, the animals were sacrificed after plaque onset and the effect on pre-plaque pathology was not documented. Our study therefore represents the first report of the effect of minocycline on early, pre-plaque stages of AD-like amyloid pathology in Tg mice.
As expected, minocycline was indeed effective in reducing inflammation, as COX-2, iNOS and IL-1β levels were all found to be down-regulated in Tg animals treated with minocycline compared to placebo (Figure 1A-C). After minocycline treatment, the microglial soma size of hippocampal cells appeared to be significantly reduced (Figure 1E and 1G), along with an increase in the complexity of microglial arborization (as illustrated in the representative micrograph in Figure 1E). Both biochemical and morphological findings suggested decreased pro-inflammatory activity. Given the well-known complexity of the microglial phenotype , it is possible that minocycline exerted its effect by switching the microglial cells to a more neuroprotective, M2-like phenotype. Further studies will be required to pinpoint the features of microglia in response to minocycline treatment, using alternative markers such as CD45 or arginase-1 [58–60]. Interestingly, the anti-inflammatory effect of minocycline was specific to the hippocampus and the cortex, a region burdened with intracellular Aβ-oligomers . In fact, microglial cells of the thalamus, an area largely devoid of Aβ material at this early age, did not appear to be affected by the treatment (Figure 1F and 1H). This result indicates that minocycline specifically interfered with a pathological inflammatory process dependent on intracellular Aβ accumulation.
Minocycline effects on the amyloid pathology
Having assessed the ability of minocycline to inhibit the pre-plaque inflammatory process, we set ourselves to study the consequences of such anti-inflammatory treatment on the intracellular, pre-plaque phase of the amyloid pathology.
In our study, as opposed to previous reports of minocycline in AD models, the treatment was started and finished when the animals were devoid of plaques. Therefore, instead of plaque number, we focused our investigation on the cerebral levels of APP, APP-related products and soluble Aβ following minocycline treatment. Soluble levels of Aβ are particularly important indicators of the disease state, as they were shown to correlate with the degree of dementia in AD patients . On the other hand, it is well established that the amyloid plaque burden does not correlate with the severity of the disease [62, 63].
In this regard we first noticed that, at this early time point, inhibition of inflammation was associated with the down-regulation of APP (Figure 2). While the cellular mechanisms for such an effect remain to be determined, there is evidence in the literature that inflammatory mediators can modulate APP synthesis. IL-1β, for instance, was shown to induce APP synthesis in neurons [64, 65]. The decrease in flAPP could therefore be a consequence of reduced IL-1β levels.
Besides the down-regulation of flAPP, the most significant effect we observed was the reduction of a 12-kDa band recognized with the monoclonal antibody 6E10. This band co-migrates with trimers of synthetic Aβ and is considered by some authors as oligomeric-Aβ [66, 67]. However, the epitope recognized by 6E10 is shared by the C-terminus fragments of the amyloidogenic pathway (β-CTF) and 6E10 is often used to detect β-CTF from cell lysates and homogenates . The fact that CTFs migrate around 12 kDa, as do Aβ trimers, complicated the interpretation of the band. To clarify the nature of this material we sought to specifically quantify CTF fragments and human Aβ in the same samples. This analysis also allowed us to determine the relative abundance of the two species in each brain.
Western blots using a specific antibody directed against the C-terminus of full-length APP (pab27576) revealed a reduction in the CTF content, which did not reach significance (Figure 2D and 2E). Even though the band recognized by pab27576 perfectly overlapped with the band seen with 6E10, the results did not fully match our analysis with 6E10. It is possible that the discrepancy is due to different specificity of the antibodies. Alternatively, some CTF material from the non-amyloidogenic pathway (αCTF which can be detected by pab27576 but not by 6E10) might have affected our quantification. We also considered the possibility that the 12-kDa band was constituted mostly of Aβ species (trimers) and so performed a highly sensitive ELISA assay for human Aβ40 and Aβ42. No plaques were detected in the animals and, from our previous study in this model, the Aβ-ir species are either monomeric or oligomeric at this stage . Therefore, the Aβ material measured via ELISA can be considered as soluble in nature. Our analysis of such soluble Aβ material did reveal some degree of reduction after treatment with minocycline, but the high variability resulted in no statistical significance (Figure 2F). In summary, while the three analyses (western blot with 6E10, pab27576 and ELISA) all showed a reduction of APP-related products, the pattern observed with the 6E10 antibody was not fully reproduced by either that of CTF or Aβ alone. A likely explanation for this is that the 12-kDa band recognized by 6E10 represents a mixture of β-CTF and Aβ-oligomers. In this view, it is possible that minocycline treatment resulted in the reduction of both species, which together reached significance. Nevertheless, further investigations are needed to clarify this point.
The simultaneous presence of Aβ and APP-related products in early, pre-plaque stages of the disease in Tg models of AD is a highly controversial issue [69–73]. In particular, their relative abundance and their specific contribution to the neuropathology have not been clarified. We therefore took advantage of the data set presented here to explore the relative abundance of Aβ and CTF, and the effect of an anti-inflammatory treatment on their ratio. We compared the absolute levels of CTF (extrapolated from a recombinant CTF calibration curve, using a semi-quantitative method) and Aβ (measured via ELISA) from each sample. In Tg Placebo animals we observed (on average) 6.76 pg of Aβ40 per each ng of CTF, and 24 pg of Aβ42 per each ng of CTF. In terms of molar ratio, it appears that McGill-Thy1-APP mice harbor about 84 molecules of CTF per each molecule of Aβ40 and about 24 molecules of CTF per each molecule of Aβ42. It is therefore very likely that, while the species co-exist, β-CTF fragments represent the vast majority of the material seen with 6E10, as suggested by McAlpine et al. .
Beta-site APP cleaving enzyme 1-deregulation in young, pre-plaque mice and its correction with minocycline
To further elucidate the effect of minocycline on APP processing, we studied the levels and activity of BACE-1, the most important β-site APP cleaving enzyme in the brain .
Our analysis revealed that BACE-1 levels and activity were up-regulated in the McGill-Thy1-APP Tg model, in agreement with reports from sporadic AD [76–79] and Tg models [24, 80, 81]. As described in young V717V Tg mice , BACE-1 levels and activity were up-regulated in McGill-Thy1-APP mice prior to plaque deposition. Therefore, deregulation of APP processing might be an early event in the progression of the AD-like amyloid pathology. Minocycline treatment restored BACE-1 activity to control levels, and corrected BACE-1 protein content in young, pre-plaque Tg mice (Figure 3). These results would agree with the western blotting of 6E10 (Figure 2) in indicating reduced β-cleavage of APP upon anti-inflammatory treatment.
Since the anti-inflammatory treatment with minocycline was able to correct BACE-1 up-regulation in pre-plaque Tg mice, it is very likely that the early deregulation of BACE-1 prior to plaque deposition is related to the pro-inflammatory process. This view is in line with the body of evidence indicating that neuroinflammation has a pivotal role in regulating BACE-1. In fact, several studies have indicated that BACE-1 behaves as a stress-response protein and its levels are increased by cytokines , oxidative stress , astrocytic activation , ischemia , hypoxia  and energy inhibition . On the other hand, we did not detect any effect of minocycline treatment on the levels of Aβ-degrading enzymes such as insulin-degrading enzyme and neprilysin (data not shown).
Minocycline mechanism of action and effect on NFkB
In an attempt to clarify the mechanism of action of minocycline, we measured the levels of NFkB, a key transcription factor which is known to regulate the expression of several inflammatory markers as well as BACE-1 and APP [38, 88, 89]. Increased NFkB expression is associated with neuroinflammatory conditions and it has been reported in AD [90–93]. Elevated NFkB activity was detected in Tg models of AD . Interestingly, since the promoter of the IkB gene contains several NFkB binding sites , IkB expression is elevated in response to NFkB activation following cerebral ischemia  and lipopolysaccharide injections . Increased levels of IkB have also been reported in AD .
Consistent with the results from AD samples, we found up-regulated levels of NFkB and its inhibitor, IkB, in Tg animals compared to Non Tg Placebo animals. Given the relatively high variability of the data, these changes did not reach statistical significance; however, they support the notion of a pro-inflammatory state in these brains. Accordingly, NFkB levels were reduced following minocycline treatment, while IkB was further up-regulated. A reduction of NFkB with concomitant up-regulation of IkB has been reported for other anti-inflammatory agents such as ibuprofen  and for glucocorticoids . The increase in IkB levels is thought to further potentiate the anti-inflammatory effect, as any NFkB molecule synthesized by the cell will associate with the inhibitor and be prevented from entering the nucleus. These results, even though they did not reach significance, suggested that minocycline treatment in Tg animals resulted in an overall decreased activity of NFkB.
Furthermore, the levels of NFkB strongly correlated with BACE-1 levels in each sample. These results are consistent with the concept that BACE-1 levels and activity are tightly linked to NFkB levels in vivo. Even though this type of correlative analysis cannot prove causality, several indications exist that NFkB regulates BACE in vivo. Paris et al. have recently shown that the NFkB inhibitor celastrol is capable of inhibiting BACE and reducing amyloidogenic pathway in a mouse Tg model of AD . Similarly, a reduction in BACE-1 and Aβ levels was found in Tg mice acutely treated with the NSAID ibuprofen . This drug is endowed with multiple COX- independent mechanisms of action, including inhibition of NFkB signaling , peroxisome proliferator activated receptor-gamma activation  and gamma-secretase modulation . It is very likely that, like ibuprofen, minocycline exerts its beneficial effects via multiple mechanisms of action.
Based on our results, one could speculate that the inflammation-induced hyperactivity of NFkB is responsible for the increased transcription of BACE-1 in Tg animals. This might represent a possible mechanism for the glia-to-neuron or neuron-to-glia communication in early AD, whereby the activation state of microglia can instruct the processing of APP in neurons. Alternatively, the reduction of inflammatory markers and the reduction in BACE-1 (levels and activity) following minocycline treatment might be parallel, unrelated events sharing the same up-stream events (that is, inhibition of NFkB in glia and neurons).
Minocycline adverse effects
It is important to note that the intraperitoneal application of 50 mg/Kg/day of minocycline resulted in some toxicity: one out of eight mice (12.5%) in the Non Tg group and two out of seven mice (28%) in the Tg group died, while the remaining mice showed signs of liver toxicity and peritoneal irritation. These adverse effects precluded the completion of behavioral testing for learning and memory, such as the Morris water maze task. Liver toxicity  and peritoneal inflammation  are known side effects of intraperitoneal administration of minocycline which are seldom referred to in numerous experimental published studies. It has been established that the peripheral inflammatory process can have an impact on the microglial status in the CNS [104, 105]. Indeed, the occurrence of some glial activation following minocycline treatment is supported by the rise in NFkB levels and BACE-1 activity in the Non Tg Mino group. However, these alterations were not accompanied by classical pro-inflammatory activity, as COX-2, iNOS and IL-1β were not found to be significantly different from Non Tg Placebo. As the intraperitoneal application of the drug was not inert, we cannot rule out the possibility that peripheral toxicity could have had some role in the CNS effects observed in the Tg-treated mice.
While the dose and administration route of the drug need to be optimized to avoid adverse peripheral effects, our overall results indicate that the inhibition of neuroinflammation with minocycline can be beneficial in early pre-plaque stages of AD-like amyloid pathology.