Background
Alzheimer’s disease (AD) is characterized by two types of pathological lesions: neurofibrillary tangles, which are composed of hyperphosphorylated forms of the protein Tau, and amyloid plaques, which consist of aggregations of the peptide β-amyloid (Aβ). Cerebral accumulation of Aβ, especially the 42-amino acid isoform Aβ42, is thought to play a critical early role in the development of AD [
1,
2], so there is great interest in decreasing Aβ levels in the brain to prevent or treat AD. Aβ is generated by the sequential cleavage of Amyloid Precursor Protein (APP) by β-secretase (BACE1) and γ-secretase. APP transgenic mouse models of AD that are devoid of BACE1 expression via BACE1 gene knockout (BACE1
−/− mice) fail to generate Aβ and lack the amyloid plaques and cognitive impairments found in APP transgenic mice that express both BACE1 alleles [
3‐
7]. These results validated BACE1 as the major β-secretase enzyme in the brain and suggested that inhibition of BACE1 could be of therapeutic benefit for AD (reviewed in [
8]). Indeed, BACE1 inhibitor drugs are currently being tested in clinical trials in AD and mild cognitively impaired individuals (reviewed in [
9]). However, the long-term safety and efficacy of BACE1 inhibitor drugs are unknown at present.
Initially, BACE1 null mice were reported to be healthy with no obvious phenotype [
5,
10], but more extensive analysis revealed that they have subtle neurological abnormalities [
8,
9]. Subsequent studies have found that BACE1 null mice have higher offspring mortality, decreased myelination, impaired memory, hyperactivity, axon mis-guidance, schizophrenia-like phenotypes, and increased seizure activity, but these phenotypes are largely absent from BACE1
+/− mice [
6,
7,
11‐
18]. The BACE1 substrates responsible for some of these phenotypes are known, such as the role of neuregulin in myelination [
13,
17] and CHL1 in axon guidance [
18], but others remain unexplained. Proteomic screens of BACE1
−/− compared to BACE1
+/+ primary neurons [
19,
20] have revealed even more potential BACE1 substrates that are not yet validated
in vivo but could have a role in these phenotypes, and others yet to be described.
Since complete loss of BACE1 activity has detrimental effects in BACE1
−/− mice it seems likely that almost complete inhibition of BACE1 for treatment or prevention of Alzheimer’s disease could have mechanism based side-effects in humans. The 50% BACE1 reduction observed in in BACE1
+/− mice, on the other hand, seems to have no ill effects. If 50% inhibition of BACE1 is able to decrease Aβ production enough to delay disease onset or slow disease progression, this could represent a therapeutic strategy to avoid side effects of almost total BACE1 inhibition. The BACE1
+/− heterozygous null mouse is a useful model for 50% BACE1 inhibition, and several publications have described BACE1
+/− mice on various backgrounds of APP transgenic mouse models, with most observing some reduction in Aβ levels, but the degree of Aβ lowering varies from model to model [
5,
14,
21‐
26]. It is also unclear whether 50% reduction in BACE1 leads to a long-lasting decrease in cerebral Aβ. It has been reported in the PDAPP mouse model that BACE1
+/− genotype led to a small reduction in Aβ at 3 months of age, but dramatic Aβ decreases at 13 and 18 months [
24]. On the other hand, in transgenic mice co-expressing APP Swedish (swe) and presenilin 1 exon 9 deletion (PS1Δ9) familial AD (FAD) mutations, BACE1
+/− genotype led to decreased cerebral Aβ and plaques at 12 months, but not at 20 months of age [
14].
This work extends the study of 50% BACE1 inhibition as a therapeutic approach, demonstrating that 50% BACE1 reduction in 5XFAD transgenic mice, which display aggressive, early onset amyloid pathology [
27], decreases Aβ42, plaques, and BACE1-cleaved APP fragments (C99 and sAPPβ) at 4, 6 and 9 months of age, but unexpectedly only in females, which have higher levels of Aβ42 and amyloid plaques than males. Other work reported a reduction in Aβ, amyloid deposition, and amelioration of cognitive deficits in 5XFAD/BACE1
+/− mice, but did not differentiate between the sexes [
21‐
23]. We attribute the elevated Aβ42 and amyloid deposition in female 5XFAD to higher levels of APP transgene expression due to an estrogen response element (ERE) found in the Thy-1 promoter of the transgene. The 5XFAD mouse model has become quite widely used in the Alzheimer’s field, and this study highlights the importance of using cohorts of the same gender, or containing equal numbers of each sex. If experimental and control groups are not gender balanced, effects on cerebral Aβ and amyloid pathology may be observed that are not due to experimental manipulation, but to higher Aβ levels in female mice.
We hypothesize that the lower level of expression of the APP transgene in 5XFAD males is the cause of the decreased cerebral Aβ42 and amyloid, and leads to a situation where BACE1 is in excess of APP, even when reduced by 50%, so that no Aβ42 lowering occurs in 5XFAD/BACE1+/− mice. In contrast, because of higher transgenic APP levels in 5XFAD females, BACE1 is not in excess over APP, thus resulting in substantial Aβ42 lowering with 50% BACE1 reduction. These results suggest that 50% BACE1 inhibition would be an effective therapeutic approach to decreasing cerebral Aβ42 levels only under conditions where BACE1 is not in excess over APP. In the case of human AD patients, APP is not overexpressed, and BACE1 is increased during the course of disease, suggesting BACE1 is likely to be present in excess of APP, limiting the therapeutic efficacy of reducing BACE1 activity by 50%. Finally, we also report an accurate, simple, and inexpensive dot blot assay to measure cerebral Aβ42 levels as an alternative to Aβ42 ELISA.
Discussion
Here, we report that 50% BACE1 reduction lowers cerebral Aβ42 levels by ~30-40% in female, but not male, 5XFAD/BACE1
+/− mice. Similarly, 50% BACE1 reduction lowers sAPPβ and C99 levels in female 5XFAD/BACE1
+/− mice, paralleling the Aβ42 decrease and suggesting reduced BACE1 cleavage of APP, although other mechanisms such as increased Aβ42 clearance remain formally possible. We also show that female 5XFAD mice have higher cerebral Aβ42 levels than males, which is most pronounced in the BACE1
+/+ genetic background and correlates with higher steady state transgenic APP levels in female compared to male 5XFAD mice. Importantly, an ERE exists in the neuron-specific 5XFAD transgene murine Thy-1 promoter sequence, which potentially could explain the increased transgenic APP level and, in turn, the higher cerebral Aβ42 accumulation in female 5XFAD mice [
27]. Finally, we report a robust and accurate Aβ42 dot blot assay for measuring relative Aβ42 levels in the brain as a cost-effective alternative to Aβ42 ELISA. Together, these results reveal that the effectiveness of 50% BACE1 inhibition in reducing cerebral Aβ42 is affected by relative levels of BACE1 and APP, which has important implications for use of BACE1 inhibition in preventing or treating AD.
We have previously reported that our dot blot assay can measure Aβ42 in the brains of cognitively normal, aged humans, with results comparable to those generated by ELISA [
29]. By ELISA, Aβ42 varied from 1.5-30 ng/mg total protein in homogenates of normal human brains, which is in the linear range of our dot blot method. Here we demonstrate that the dynamic range of the Aβ42 dot blot assay is at least 500 fold, and it is sensitive and linear in the range required for measuring Aβ42 in the 5XFAD mouse model as well (Figure
3G). This dot blot assay can be easily adapted to measure absolute Aβ42 levels by including known quantities of synthetic Aβ42 on the dot blot for generating a standard curve. The range of the standards for a commonly used commercial Aβ42 sandwich ELISA (e.g., Wako, Inc.) is 4-400 pg/ml, while the standards we used for the dot blot assay range from 7.7-3900 ng/ml. While the ELISA has the ability to detect very low concentrations of Aβ42, as with plasma or cerebral spinal fluid samples, this is not necessarily advantageous for samples with high Aβ42 concentrations, such as brain homogenates of APP transgenic mice or aged humans, because they have to be diluted up to 5000-fold, leaving room for pipetting error, while the dot blot assay requires little if any dilution, and is thus potentially less error-prone. Additionally, the linear range of the ELISA readout is only around 100-fold, while our dot blot is 500 fold, allowing the comparison of diverse samples. In short, we suggest the dot blot technique for measuring Aβ42 levels in brain homogenates of mouse models of AD and in human subjects, though it is likely to be less sensitive for measuring low concentrations of Aβ42 in samples such as cerebral spinal fluid or plasma.
The findings reported here are congruous with those of other studies showing reduced Aβ and amyloid pathology in APP transgenic mice with one inactive allele of BACE1 [
5,
14,
21‐
26]. The degree of Aβ lowering with 50% BACE1 reduction varies depending on the study, which might be related to differences in the APP transgenic mouse models employed, such as APP mutation, overexpression level, and strain background, among others. The level of Aβ lowering in female 5XFAD/BACE1
+/− mice in our study was most similar to that observed in APPSwe/PS1E9/BACE1
+/− mice, in which Aβ42 levels are decreased by 40% and 27% at 3 and 12 months, respectively [
14].
Our observation that 50% BACE1 reduction in the 5XFAD transgenic line decreases Aβ42 levels in females only was unexpected. Previous studies of BACE1
+/− mice on various APP transgenic backgrounds reported no differences in Aβ levels between males and females [
14,
26] or did not analyze the genders separately [
21‐
24]. The reason for the lack of Aβ42 lowering in male 5XFAD/BACE1
+/− mice remains enigmatic, although it might involve the lower level of expression of the human APP transgene in male compared to female 5XFAD mice. We hypothesize that even with 50% BACE1 reduction in male 5XFAD/BACE1
+/− mice, BACE1 levels are still sufficiently in excess of those of APP in the intracellular compartments where β-secretase processing occurs such that the same amount of APP can be cleaved as in male 5XFAD/BACE1
+/+ mice. In contrast, in female 5XFAD/BACE1
+/− mice, BACE1 levels might not be in excess of those of APP because of the higher female expression of the APP transgene. This could render BACE1 rate-limiting, so that at the 50% level BACE1 cleaves less APP within β-secretase intracellular compartments, leading to a reduction in Aβ42, C99, and sAPPβ in 5XFAD/BACE1
+/− compared to 5XFAD/BACE1
+/+ females. This notion is supported by studies showing that PDAPP/BACE1
+/− mice with very high (~10 fold) APP transgene expression [
34] have 90% reduction in Aβ at 13 months and 50% reduction at 18 months [
24], while decreases in endogenous Aβ are small in non-Tg BACE1
+/− mice lacking an overexpressed APP transgene [
25,
26]. However, contrary to the above argument, we also observed that in both males and females, full-length transgenic APP levels are significantly increased in 5XFAD/BACE1
+/− compared to 5XFAD/BACE1
+/+ mice at 9 months of age, and trend higher at the younger ages, regardless of gender, indicating reduced β-secretase cleavage of APP in 5XFAD/BACE1
+/− of both sexes. Thus, there may be additional factors other than transgenic APP level causing the difference in Aβ lowering between male and female 5XFAD/BACE1
+/− mice.
These data, along with our current findings suggest that the degree to which BACE1 activity must be inhibited to significantly reduce cerebral Aβ42 levels will depend on whether BACE1 is in excess of APP and by how much. Depending of the study, in wild type mice, where APP is not overexpressed, BACE1 heterozygosity has either no effect on Aβ40 or Aβ42 [
25] or reduces Aβ40 by ~10% [
26]. In AD patients, BACE1 is elevated in dystrophic neurites surrounding plaques, as is APP [
28,
30], so it difficult to determine whether BACE1 is in excess of APP or not. BACE1 elevation in AD brains is detectable by immunoblot and by enzymatic activities [
28,
30,
35,
36] but APP elevation by immunoblot has not been reported. The elevation of BACE1 protein levels and enzymatic activity in AD brains suggests that BACE1 could be in excess of APP, so we predict that BACE1 inhibition greater than 50% will be required to significantly reduce cerebral Aβ42 levels in AD. While BACE1 inhibitors in clinical trials are able to lower cerebral spinal fluid Aβ by ~50-90 % indicating highly effective BACE1 inhibition (reviewed in [
9]), concerns about mechanism-based side effects related to BACE1 over-inhibition exist [
8]. BACE1
−/− mice have multiple complex but subtle neurological abnormalities, though it is not yet clear whether these stem from the absence of BACE1 during crucial developmental periods, or during adult life.
It should be noted that the presence of the APP Swedish mutation in the 5XFAD mouse line could also affect the degree to which 50% BACE1 reduction lowers Aβ42 generation, although probably not dramatically. This mutation makes APP a better substrate for BACE1 [
37] so a larger proportion of APP molecules become cleaved to produce Aβ. Rabe et al. [
26] showed that heterozygous BACE1 gene deletion reduces Aβ levels by 16% in mice expressing a wild-type APP transgene, which more likely reflects the situation in sporadic AD, and by 20% in mice expressing an equivalent level of transgenic APP with the Swedish mutation. Although 50% BACE1 reduction lowered Aβ levels in Swedish mutation APP mice greater than in wild-type APP mice by a modest amount, these results suggest the Swedish mutation does not dramatically alter the effect of partial BACE1 reduction on Aβ generation, at least in the mouse strains studied [
26].
Our study furthers the characterization of the 5XFAD mouse model, which is now widely used by the AD research community, and highlights the variation between males and females in this line. Differences between males and females in Aβ levels and plaque deposition have been reported in other mouse models of AD such as the Tg2576, 3xTg-AD, and APPswe/PS1de9 lines [
38‐
40], and may be related to estrogen levels [
41]. These results mirror observations in the human population that AD risk is higher for females than males, even after correcting for increased longevity [
42], but it is still not clear why women are more susceptible. Our observation that female 5XFAD mice have higher steady-state levels of transgenic APP compared to males is likely responsible for increased female 5XFAD Aβ42 level and amyloid deposition and appears to be related to the presence of an ERE in the 5′ upstream regulatory region of the murine Thy-1 transgene promoter. Levels of endogenous APP did not differ between male and female non-Tg mice, suggesting that female humans do not have higher cerebral APP expression. Our results caution that the 5XFAD mouse and other AD transgenic models that employ the Thy-1 promoter are not appropriate models for the gender disparity observed in AD. Thus, a mechanism other than increased steady-state APP level is likely responsible for the higher incidence of AD in women. Additionally, these results illustrate the critical importance of designing studies in which the control and experimental groups contain equal numbers of males and females, and that data from the two genders should be analyzed separately to determine differential effects of the experimental condition, and in order to avoid false gender-specific differences. If gender effects are observed in this model, or others using the Thy-1 promoter, it is important to determine whether the cause may be related to different levels of transgene expression.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
KRS participated in study design, performed immunoblotting, analyzed data, and drafted manuscript. WAE participated in study design and performed immunostains. SLC participated in study design and performed ELISA. RV conceived of the study, participated in its design and coordination and designed research, and drafted manuscript. All authors read and approved the final manuscript.