BACE1 elevation engendered by GGA3 deletion increases β-amyloid pathology in association with APP elevation and decreased CHL1 processing in 5XFAD mice

GGAKO mice have been previously described [10]. The 5XFAD mice co-overexpress familial AD mutant forms of human APP695 with Swedish (K670 N, M671 L), Florida (I716V), and London (V717I) mutation, and presenilin 1 (PS1) with M146 L and L286 V mutation under transcriptional control of the neuronal-specific mouse Thy-1.2 promoter [11]. The 5XFAD mouse line Tg6799 was purchased from the Jackson Laboratories (MMRRC Stock No: 34840-JAX). This line is on the B6/SJL genetic background and develops intraneuronal Aβ42 accumulation starting at 1.5 months of age, just prior to amyloid deposition and gliosis, which begins at 2 months of age. On a congenic C57BL/6 J genetic background (MMRRC stock 34848) it has been the observation of the MMRRC that this phenotype is not as robust as that demonstrated in the B6SJL hybrid background. Hemizygous 5XFAD transgenic mice, on a genetic background of B6/SJL, were crossed to GGA3KO mice, on a genetic background of C57BL/6, to yield GGA3Het;5XFAD male and GGA3Het female mice. Both mice were intercrossed to generate six different genotypes: GGA3WT, GGA3Het, GGA3KO, GGA3WT;5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD. The resulting offspring were genotyped by PCR analysis of genomic DNA from tail biopsies. BACE1−/− (BACE1KO) mice were purchased from the Jackson Laboratories (Stock No: 004714). All animal experiments were performed with the approval of Tufts University Institutional Animal Care and Use Committees.

Mice were anesthetized with isoflurane and transcardially perfused with PBS buffer. One collected hemibrain was used for biochemical analysis and the other for histological analysis. Hippocampus and cortex were rapidly dissected from the hemibrain for biochemical analysis and snap frozen in liquid nitrogen. The snap frozen samples were homogenized in 4 volumes of PBS buffer containing 1X Protease inhibitor cocktail and 1X Phosphatase inhibitor cocktail (Thermo Scientific). Brain homogenates in PBS were separated for immunoblot analysis and ELISA. For immunoblot analysis, 2 times concentrated RIPA (radioimmunoprecipitation assay) buffer (150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, and 50 mM Tris HCl, pH 7.4) containing protease and phosphatase inhibitors were added to brain homogenates in PBS, sonicated briefly, and then incubated for 30 min on ice. The homogenates were centrifuged at 14,000×g for 20 min at 4 °C and then the supernatants were collected for immunoblot analysis. Protein concentrations were determined using BCA assay (Thermo Scientific). RIPA extracted brain lysates (20~ 50 μg) were run on 4–12% Bis-Tris gels (Invitrogen or Bio-Rad) with XT MES or XT MOPS running buffers. 3–8% Tris-acetate gels (Bio-Rad) with XT Tricine running buffer were used to separate and measure the ratio of cleaved N-terminal CHL1 fragment (CHL1_βNTF) to full-length CHL1 (CHL1_FL) in the hippocampus lysates, whereas 4–12% Bis-Tris gels (Bio-Rad) with XT MOPS running buffer were used to assess the CHL1 processing in the cortex homogenates. Proteins were transferred onto polyvinylidene difluoride (PVDF) membrane and blocked in 5% non-fat dried milk in TBST. The membrane was then incubated in the following antibodies: rabbit monoclonal anti-BACE1 (1:1500; D10E5; Cell signaling); rabbit polyclonal anti-GGA3 (1:1500; 4167; Cell signaling); rabbit polyclonal anti-GGA1 (1:2000; H-215; Santacruz); rabbit polyclonal anti-C-terminal APP antibody to detect mouse and human APP (1:5000; A8717; Sigma Aldrich); goat polyclonal anti-N-terminal CHL1 antibody to detect mouse CHL1_βNTF and CHL1_FL (1:1000; AF2147; R&D Systems); rat monoclonal anti-N-terminal CHL1 antibody to detect human CHL1_βNTF and CHL1_FL (1:2000; MAB2126; R&D Systems); mouse monoclonal anti-human APP antibody which recognizes the amino acid sequence 1–16 in Aβ region (1:10,000; 6E10; Biolegend); mouse monoclonal anti-calnexin to detect mouse and human calnexin (1:2000; 610523; BD biosciences); mouse monoclonal anti-GAPDH (1:10,000; MAP374; Millipore); and mouse monoclonal anti-β-tubulin (1:10,000; JDR.3BR; Sigma-Aldrich). Primary antibodies were detected with the species-specific HRP-conjugated secondary antibody, and then visualized by ECL. Chemiluminescent signal was captured on an LAS4000 Fuji Imager. Densitometry analysis was performed using Quantity One software (Bio-Rad).

Brain PBS extracts were supplemented to reach a final concentration of 5 M guanidine HCl/50 mM Tris HCl and placed on a rotator overnight at 4 °C. Then, samples were diluted 1:20 in Reaction Buffer (Dulbecco’s phosphate buffered saline with 0.03% Tween-20 supplemented with 1X Protease inhibitor cocktail (Thermo Scientific). Samples were then centrifuged at 16,000×g for 20 min at 4 °C, and then the supernatants were collected. Final guanidine HCl concentrations were below 0.1 M. Protein concentrations were determined by BCA assay. The concentration of Aβ42 in Guanidine HCl extracts was measured by human Aβ 42 ELISA kits (KHB3441, Invitrogen) according to the manufacturer’s instruction. Optical signals at 450 nm were read on a Synergy 2 plate reader (Biotek) and sample concentrations were determined by comparison with the respective standard curves.

After perfusion with PBS, collected hemibrain for histological analysis was fixed in 4% paraformaldehyde in PBS for 24 h and cryoprotected in 30% sucrose in PBS containing 0.02% sodium azide. Coronal sections (40 μm) were cut on a sliding microtome. Every twelfth coronal sections were collected in cryoprotectant (30% sucrose, 1% Polyvinyl-pyrrolidone (PVP-40), 0.05 M phosphate buffer, 30% ethylene glycol), and then processed for histological analysis. Coronal sections from 4 and 12 months old male and female GGA3WT:5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD mice were used to assess the effect of GGA3 deletion on amyloid plaques deposition by Thioflavin-S staining. First, free-floating coronal sections were washed three times with PBS. Sections were then treated with 0.1% Sudan Black in 70% ethyl alcohol for 20 min to quench non-specific autofluorescence. Samples were briefly differentiated two times in 70% ethanol, and then washed three times with PBS. Sections were stained with 0.025% Thioflavin-S in 50% ethanol for 8 min, and then differentiated two times with 50% ethanol. After three washes with PBS, sections were mounted on slides, and covered with mounting media. Fluorescent images were acquired on BZ-X700 all-in-one fluorescence microscope (Keyence). Captured images were used to analyze the size, number, and area of plaques in the hippocampus and cortex of male and female GGA3WT;5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD mice with Fiji software (Image J).

Using a BZ-X700 all-in-one fluorescence microscope (Keyence), fluorescent images of 4 and 12 months old male and female GGA3WT;5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD mice were acquired with 10X objective and then stitched using BZ-X analyzer Software. Stitched images (5–8 sections between bregma 0 and − 3.5 per mouse) were quantified using Fiji software (Image J). Captured stitched images were first converted to 8-bit black and white images. Then an intensity threshold (Triangle) for detecting the amyloid plaques area was applied to calculate the number, size, and percentage of total amyloid burden in the hippocampus or cortex.

After weighing the snap frozen mouse hippocampi, samples were homogenized in 10 volumes of PBS buffer containing 1X protease inhibitor and 1X phosphatase inhibitor cocktail with 30 strokes on a Dounce homogenizer. About half of PBS homogenates were ultracentrifuged at 100,000×g for 1 h to obtain PBS fraction (soluble). The remaining pellet was washed with PBS buffer and spun at 100,000×g for 5 min. After removing the supernatant, remaining pellet was lysed in 10 volumes of RIPA buffer, sonicated, rotated for 30 min at 4 °C, and then ultracentrifuged at 100,000×g for 1 h. The supernatant was collected, and then stored in − 80 °C as RIPA fraction (membrane). Remaining half of PBS homogenates were mixed with 2X RIPA supplemented with protease and phosphatase inhibitor cocktail, sonicated and rotated for 30 min at 4 °C. After centrifugation at 14,000 xg for 20 min at 4 °C, the supernatant was collected as RIPA lysates (Total).

Frozen human hippocampus samples from subjects with Down syndrome (DS) or Unaffected Controls (Ctrls) were obtained from NIH NeuroBioBank (Table 1). Down syndrome samples were obtained from University of Maryland Brain and Tissue bank and unaffected control samples were obtained from University of Miami Brain Endowment Bank. All subjects are described in Table 1.

For serial extraction of human samples, we used a modified sequential extraction method [12]. After weighing the frozen human hippocampus tissue, samples were homogenized in 10 volumes of ice-cold TBS buffer (50 mM Tris-HCL, 150 mM NaCl, pH 7.4) containing 1X protease inhibitor and 1X phosphatase inhibitor cocktail with 30 strokes on a Dounce homogenizer. About half of TBS homogenates was used for serial fractionation, the other half of TBS homogenates was used for RIPA extraction. About half of TBS homogenates was ultracentrifuged at 175,000×g for 30 min. The supernatant was collected, and then stored in − 80 °C as TBS fraction (soluble fraction). The remaining pellet was washed with TBS buffer and spun at 175,000×g for 5 min. After removing supernatant, the remaining pellet was resuspended in 10 volumes of TBS-TX buffer (1% Triton X-100 in TBS buffer containing 1X protease inhibitor and 1X phosphatase inhibitor cocktail), sonicated, rotated for 30 min at 4 °C, and then ultracentrifuged at 175,000×g for 30 min. The supernatant was collected, and then stored in − 80 °C as TBS-TX fraction (membrane fraction). The remaining half of TBS homogenates was mixed with 2X RIPA supplemented with protease and phosphatase inhibitor cocktail, sonicated and rotated for 30 min at 4 °C. After centrifugation at 14,000 xg for 20 min at 4 °C, the supernatant was collected as RIPA lysates (Total).

Total RNA was extracted from hippocampi of non-5XFAD and 5XFAD mice using RNeasy Mini kit (Quiagen) according to manufacturer’s instructions. Total RNA was analyzed using the Agilent 2100 Bioanalyzer to measure purity and integrity of RNA (Aligent Technologies). 1 μg of total RNA from each sample was subjected to reverse transcription reaction to synthetize cDNA using QuantiTect Reverse Transcription kit (Quiagen) according to manufacturer’s instruction. RT-qPCR was performed on an MX3000p qPCR System (Aligent Genomics) using the SYBR Green PCR Master Mix kit (Life Technologies). Primers were used to measure relative Chl1 gene expression levels (Primerbank ID: 110347544c3) normalized to the housekeeping gene Gapdh (Primerbank ID: 6679937a1). Relative expression levels were determined according to the ΔΔCt method when the expression level of the CHL1 mRNA is given by 2^-ΔΔCT where ΔΔCT = ΔCT CHL1 mRNA – ΔCT reference mRNA (Gapdh) in the same sample.

Data are expressed as mean ± standard error of the mean (SEM), represented as error bars. One-way or two-way analysis of variance (ANOVA) with Fisher’s LSD post-hoc test was performed to evaluate statistical difference among the groups. Unpaired t-test with Welch’s correction was used for statistical comparison between means of amyloid burden in GGA3WT;5XFAD and GGA3KO;5XFAD mice. P < 0.05 was considered statistically significant for all experiments.

It has been previously reported that levels of BACE1 increase with age in 5XFAD mice starting at the 6 months of age [13]. Given that GGA3 deletion increases cerebral levels of BACE1 in mice [10], we hypothesized that GGA3 deletion would lead to BACE1 elevation in 5XFAD mice at an earlier age (e.g. 4 months of age). Thus, we first analyzed the levels of BACE1 in the hippocampus and cortex of 4 months old of GGA3WT, GGA3Het, GGA3KO, GGA3WT;5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD mice (Fig. 1a and b). We found that BACE1 levels were increased in the hippocampus of GGA3WT;5XFAD mice compared to GGA3WT mice. As we have previously reported, BACE1 levels were increased both in the hippocampus and cortex of GGA3KO mice compared to GGA3WT mice (Fig. 1c and d). However, levels of BACE1 were similar in GGA3KO and GGA3KO;5XFAD mice (Fig. 1c and d). Interestingly, GGA3 haploinsufficiency potentiates BACE1 elevation both in the hippocampus and cortex of 5XFAD mice. Levels of BACE1 were increased both in the hippocampus and the cortex of GGA3Het;5XFAD compared to GGA3WT;5XFAD mice, and GGA3Het;5XFAD compared to GGA3Het mice (Fig. 1c and d). Two-way ANOVA showed no significant difference in BACE1 levels between sex in hippocampus (F(1, 60) = 2.12, p = 0.15) and cortex samples (F(1, 61) = 0.043, p = 0.83), thus the data from both sexes were combined. Taken together, these data indicate that while 5XFAD mutations did not further increase BACE1 levels when GGA3 is deleted, GGA3 haploinsufficiency potentiated BACE1 elevation in 4 months old 5XFAD mice.

We have previously shown that that levels of GGA3 are decreased and inversely correlated with BACE1 levels in post-mortem AD brains [8]. Our original findings have been confirmed by two independent studies conducted in AD brain samples of Australian and European origin [14] (US Patent Application 20120276076, Annaert, Wim; et al. November 1, 2012). More recently, we have reported that not only GGA3 but also its homologue GGA1 is a substrate for caspase-3. Accordingly, we found that not only GGA3 but also GGA1 was depleted in postmortem AD brains with elevated BACE1 levels [10]. Thus, we measured GGA3 and GGA1 levels in the hippocampus and cortex of mice from the six different genotypes at 4 months of age (Fig. 2a). We found that GGA3 levels were decreased in both hippocampus and cortex of GGA3WT;5XFAD mice compared to GGA3WT littermates (Fig. 2b and c). In contrast, levels of GGA1 were similar in all genotypes (Fig. 2d and e). Given that Aβ pathology is still at early stage in 5XFAD mice at 4 months of age, these data suggest that the GGA1 depletion observed in AD brains may be a late event in the progression of the disease. No difference was found between sexes for GGA3 and GGA1 levels in the hippocampus (Two-way ANOVA, F(1, 41) = 0.24, p = 0.63 and F(1, 59) = 2.98, p = 0.90, respectively) and cortex (F(1, 41) = 1.69, p = 0.20 and F(1, 61) = 1.38, p = 0.24, respectively), thus, data from male and female mice were combined. These data collectively suggest that AD pathology leads to depletion of GGA3 not only in human brains but also in a mouse model of AD. Thus, 5XFAD mice represent a good model to study the impact of GGA3 depletion on AD pathology and in particular BACE1 accumulation.

Given that BACE1 levels were increased in the GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice, we next determined whether GGA3 deletion leads to an increase in levels of Aβ42 and amyloid plaques. Guanidine hydrochloride (GuHCl) protein extracts of hippocampus and cortex of 4 months old GGA3WT;5XFAD, GGA3Het;5XFAD and GGA3KO;5XFAD mice were analyzed by ELISA (Invitrogen) to measure Aβ42 levels. Since female 5XFAD mice have higher Aβ42 levels than male 5XFAD mice [11, 15], we analyzed Aβ42 levels in males and females separately. In spite of the increase in BACE1 levels in GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice (Fig. 1), Aβ42 levels were similar in all genotypes in both male and female mice (Fig. 3a-b and Additional file 1). We furthered our experiment by measuring the amyloid burden by Thioflavin-S staining in the hippocampus and cortex of GGA3WT;5XFAD, GGA3Het;5XFAD and GGA3KO;5XFAD mice. We found no difference among different genotypes at 4 months of age in both male and female mice (Fig. 3c-f). Accordingly, we found that levels of BACE1-derived APP-CTFs (phosphoC99, C99, and phosphoC89) were similar among the different genotypes (Additional file 2).

In addition to APP, more than 40 BACE1 substrates have been identified by quantitative proteomics analysis [16‐18] and some of them have been validated in vitro and in vivo (for review see [19, 20]). The neural cell adhesion molecule L1-like protein (CHL1), which is required for axonal targeting, has been shown to be a BACE1 substrate [21, 22]. Given that elevated BACE1 did not increase APP processing in GGA3KO;5XFAD mice at 4 months of age, we next investigated the effect of GGA3 deletion and consequent BACE1 elevation on the processing of endogenous CHL1.

We first confirmed that CHL1 is a BACE1 substrate by analyzing its processing in the hippocampus and cortex homogenates from BACE1WT, BACEHet, and BACE1KO mice. In agreement with a previous study [21], in the adult hippocampus 3 bands were detected: CHL1_FL (185 kDa), CHL1_βNTF (~ 165 kDa) and a band of ~ 175 kDa that most likely is an alternative splicing isoform or glycosylated CHL1. Western blot analysis using anti-N-terminal CHL1 antibody (AF2147) revealed an increase in full-length CHL1 (CHL1_FL) and significantly reduced N-terminal CHL1 fragment (CHL1_βNTF) in both hippocampus and cortex of BACE1KO mice compared to BACE+/+ mice. A slight increase in CHL1_FL and decrease in CHL1_βNTF were observed in both hippocampus and cortex of BACE+/− mice compared with BACE1WT mice (Fig. 4a). Next, the ratio of CHL1_βNTF to CHL1_FL was measured by immunoblot analysis in the hippocampus and cortex lysates from mice of the six different genotypes at 4 months of age (Fig. 4b). Since there was no sex difference in BACE1 levels among all genotypes, data from male and female mice were combined. BACE1-mediated CHL1 cleavage was increased by 29.4% (p = 0.003) in the hippocampus of GGA3KO mice compared to GGA3WT mice. Similarly, the ratio of CHL1_βNTF to CHL1_FL was increased by 37.6% (p = 0.02) in hippocampus of GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice (Fig. 4c and d). In contrast, BACE1-mediated processing of CHL1 was similar among all genotype in the cortex. A possible explanation for the increased CHL1 processing observed in the hippocampus and not in the cortex of GGA3KO mice is that CHL1 is predominantly expressed in the subiculum, mossy fiber bundles, inner molecular layer of dentate gyrus, and stratum oriens [23]. Similarly, BACE1 is highly expressed in the mossy fiber bundles [21, 24]. Accordingly, axonal guidance defects are observed in the hippocampal mossy fibers of both CHL1−/− and BACE1−/− mice [21, 23].

Collectively, these data indicate that the BACE1 elevation engendered by GGA3 deletion was sufficient to increase CHL1 but not APP processing in both 5XFAD and non-5XFAD mice, most likely because BACE1 preferentially cleaves CHL1 over APP.

Levels of BACE1 increase with age in 5XFAD mice starting at the age of 6 months [13]. Thus, we investigated whether GGA3 haploinsufficiency or deletion affects the age-dependent BACE1 elevation in the 5XFAD mice. Levels of BACE1 were quantified in the hippocampus homogenates from GGA3WT;5XFAD, GGA3Het;5XFAD, and GGA3KO;5XFAD mice at 2, 4, 7 and 12 months of age (Fig. 5a and b). BACE1 elevation was observed in the hippocampus homogenates of 12 months old GGA3WT;5XFAD mice compared to 2 and 4 months old GGA3WT;5XFAD mice. Also, BACE1 levels were increased in the hippocampus of 4 months old GGA3Het;5XFAD mice compared to 2 months old GGA3Het;5XFAD mice. However, the levels of BACE1 were similar in the hippocampus of GGA3KO;5XFAD mice at all ages (Fig. 5a and b). As expected, BACE1 levels were significantly increased in the hippocampus of GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice or to GGA3Het;5XFAD of same age (Fig. 5c). Since Two-way ANOVA showed no significant difference in BACE1 levels between males and females at 2, 4, 7 and 12 months of age in hippocampus (F(1, 24) = 2.6, p = 0.12; F(1, 36) = 0.11, p = 0.74; F(1, 21) = 0.13, p = 0.73; F(1, 42) = 0.30, p = 0.59, respectively) samples, the data from both sexes were combined. We did not observe a significant increase in BACE1 levels between 2 and 12 months old GGA3KO;5XFAD mice, thus these data indicate that GGA3 deletion prevents age-dependent elevation of BACE1 in 5XFAD mice.

As we observed the age-dependent BACE1 elevation in 5XFAD mice, we went on to investigate if levels of GGA3 and GGA1, two proteins that negatively correlated with elevated BACE1 in AD brains, reduce with age, and could be an underlying mechanism of BACE1 accumulation in aging brains. Hippocampal levels of GGA3 in GGA3WT;5XFAD and GGA3Het;5XFAD mice were measured at four different ages (2, 4, 7 and 12 months) using immunoblot analysis (Fig. 6a). We found that GGA3 levels were significantly decreased in the hippocampus of 4, 7 and 12 months old GGA3WT;5XFAD mice compared to 2 months old GGA3WT;5XFAD mice (Fig. 6b). Moreover, GGA3 levels were reduced in the hippocampus, in 12 months old GGA3Het;5XFAD mice compared to 2 and 4 months old GGA3Het;5XFAD mice (Fig. 6b). In contrast, levels of GGA1 were decreased only in 12 months old GGA3WT;5XFAD mice compared to 2 months old GGA3WT;5XFAD mice, suggesting that GGA1 depletion is associated with more advanced pathology (Fig. 6c). Since two-way ANOVA showed no significant difference in GGA3 levels between males and females at 2, 4, 7 and 12 months of age in hippocampus (F(1, 17) = 0.51, p = 0.49, F(1, 25) = 0.005, p = 0.94; F(1, 14) = 0.24, p = 0.64, F(1, 28) = 0.21, p = 0.65 respectively) samples, the data from both sexes were combined. These results indicate that the amount of GGA3 in the hippocampus decreases while BACE1 increases with age in 5XFAD mice, similar to what is observed in human AD brains, and suggest that GGA3 depletion is a leading candidate mechanism underlying BACE1 accumulation in AD.

To further assess the effect of GGA3 deletion on AD pathology in aged brain, we next analyzed levels of Aβ42 by ELISA in 12 months old GGA3WT;5XFAD, GGA3Het;5XFAD and GGA3KO;5XFAD mice. We found that levels of Aβ42 were increased in the hippocampus of both male and female GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice (Fig. 7a and Additional file 3). In contrast, levels of Aβ42 in the cortex of all genotypes were similar in both sexes (Fig. 7b and Additional file 3).

We furthered our experiment by measuring the amyloid burden by Thioflavin-S staining in the hippocampus and cortex of GGA3WT;5XFAD and GGA3KO;5XFAD mice. We found that amyloid burden was increased in the hippocampus but not the cortex of 12 months old GGA3KO;5XFAD females compared to GGA3WT;5XFAD females (Fig. 7c-f and Additional file 3). Accumulation of Thioflavin-S positive amyloid plaques were more frequently observed in dentate gyrus (DG), stratum lucidum (slu), and CA1 stratum oriens (CA1so) of GGA3KO;5XFAD females compared to GGA3WT;5XFAD females at 12 months of age (Fig. 7c-d). A trend of increased amyloid burden was observed in the hippocampus of GGA3KO;5XFAD males, but it did not reach a statistical difference when compared to GGA3WT;5XFAD males (Fig. 7e). A further analysis revealed that the increase in amyloid burden in the hippocampus of 12 months old female GGA3KO;5XFAD mice compared to GGA3WT;5XFAD littermates was due to an increase in the plaque size rather than the number of plaques (Fig. 7g and h).

Given that levels of BACE1 do not increase in GGAKO;5XFAD mice with age (Fig. 5), we eliminate a mechanism based on the effects of the enzymatic function that generates Aβ. Next, we wonder if differential expression of the BACE1 substrate, namely APP, could be the underlying reason for our findings in Aβ42 levels and amyloid burden. The fact that sexual difference in Thioflavin-S staining was observed, consistent with sex-difference in APP expression in 5XFAD mice that has previously described [15], makes investigating APP a logical next step. Therefore, we next analyzed transgenic human APP levels using anti-human APP antibody (6E10).

We found that levels of transgenic APP were higher in 12 months old mice compared to 2 months old mice with the same genotype and sex (Fig. 8a and c). These findings were confirmed in another APP transgenic line (APPswe/PS1∆E9) (Additional file 4). Moreover, levels of transgenic human APP were higher in 12 months old females compared to males within the same genotype of 5XFAD mice (Fig. 8b). Interestingly, levels of APP were decreased in GGA3KO;5XFAD males compared to GGA3WT;5XFAD males. GGA3KO;5XFAD females also showed a similar trend toward decreased APP levels compared to GGA3WT;5XFAD females. We speculate that increased BACE1 levels via GGA3 deletion might result in increased APP processing, and decreased APP levels were consequently observed in GGA3KO;5XFAD mice at 12 months of age. We also analyzed APP-CTFs and found that levels of BACE1-derived APP-CTFs (phosphoC99, C99, and phosphoC89) were similar among the different genotypes (Additional file 5). Given that BACE1 elevation engendered by GGA3 deletion is ~ 30% compared to GGA3WT;5XFAD mice, we were not surprised to find no changes in levels of phospho-C99, C99, and phospho-C89 fragments in 12 months old GGAWT;5XFAD, GGA3Het;5XFAD and GGA3KO;5XFAD mice. Similarly Sadleir et al. [15] found that Aβ42 levels and amyloid burden decreased in BACE1Het;5XFAD female mice. While levels of APP were increased no significant changes in APP-C99 and sAPPβ were observed at 6 and 9 months of age in BACE1Het;5XFAD female mice. Furthermore, McConlogue et al. [25] reported a decrease in Aβ42 levels but no change in sAPPβ in 13 month-old BACE1Het;PDAPP mice. All together, these data indicate that increased Aβ42 levels and amyloid burden due to GGA3 deletion in 5XFAD mice, an effect that is more profound in females, is associated with elevated levels of transgenic APP.

In 4 months old 5XFAD mice, we found that elevated BACE1 due to GGA3 deletion increased BACE1-mediated cleavage of CHL1 (ratio of CHL1_βNTF to CHL1_FL) without affecting the generation of amyloid beta. In contrast, Aβ42 levels were increased in 12 months old GGA3KO;5XFAD mice compared to GGA3WT;5XFAD mice, most likely owing to elevated transgenic human APP levels. These findings raised the question of whether the elevation of APP and its increased processing could affect the cleavage of CHL1 at 12 months of age.

First we measured the levels of CHL1_FL and CHL1_βNTF in the hippocampus homogenates of mice from the six genotypes. Male and female mice were analyzed separately. CHL1_FL levels were slightly decreased in both male and female GGA3KO mice compared to wild type mice, consistent with an increased BACE1-mediated processing. In contrast, CHL1_FL levels were dramatically increased in 5XFAD mice compared to GGA3 genotype-matched non-5XFAD mice, similar to the increased CHL1_FL levels observed in BACE1KO mice, and such increase was more pronounced in females than males (Figs. 4a and 9a-c). While levels of CHL1_βNTF were similar across genotypes in non-5XFAD mice, they were increased in both male and female 5XFAD mice compared to non-5XFAD mice (Fig. 9d). In spite of such elevation, the CHL1_βNTF/CHL1_FL ratio was significantly decreased in 5XFAD mice compared to genotype-matched non-5XFAD mice and such decrease was greater in females than males (Fig. 9f-g). We also found that the ratio of CHL1_βNTF to CHL1_FL was increased in 12 months old GGA3KO mice compared to GGA3WT mice (Fig. 9a and f), similar to the observation in 4 months old mice and consistent with increased levels of BACE1 (Fig. 4).

In order to further confirm our data and better characterize CHL1 processing, we carried out serial fractionation of the hippocampus from BACE1WT, BACEHet and BACEKO mice to separate soluble and membrane fractions. Western blot analysis of CHL1 clearly detected two membrane bound CHL1 fragments in the membrane fraction (Additional file 6), corresponding to CHL1_FL (~ 185 kDa) and the ~ 175 kDa band (arrowhead) also detected in the total extract (Figs. 4b and 9a). CHL1_FL levels were increased in BACE1KO in the membrane fraction and total extract as shown in Fig. 4a. CHL1_βNTF was detected in the soluble fraction and corresponds to the ~ 165 kDa band detected in the total extract (Additional file 6). As previously described [26], a band of slightly lower molecular weight (indicated by an asterisk) was detected in the soluble fraction from BACE1KO mice. Such soluble fragment is most likely derived by a compensatory increased cleavage of CHL1 by ADAM8 or ADAM10 [27, 28].

Next, we prepared hippocampal fractions from 12 months old GGA3WT, GGAKO, BACE1Het, BACE1KO, 5XFAD, and BACE1Het;5XFAD mice. Samples from BACE1KO were included as negative control for BACE1-mediated processing of CHL1. BACE1Het and BACE1Het;5XFAD mice were analyzed to assess the effect of 50% reduction of BACE1 on levels of CHL1_FL and CHL1_βNTF.

Western blot analysis revealed that levels of CHL1_FL were decreased in the membrane fraction while CHL1_βNTF levels were increased in the soluble fraction of GGAKO compared to GGA3WT mice (Additional file 7A–B). As a result the ratio CHL1_βNTF/CHL1_FL was increased, consistent with an increased BACE1-mediated processing of CHL1 in GGAKO mice as reported in Figs. 4c and 9f. In contrast, levels of CHL1_FL were significantly increased in the membrane fraction from BACE1Het compared to wild type mice (indicated as GGA3+/+) and in BACE1Het;5XFAD compared to BACE1Het and 5XFAD mice. However, levels of soluble CHL1_βNTF were not significantly reduced in the soluble fraction of BACE1Het and BACE1Het;5XFAD compared to control (GGA3WT and 5XFAD mice, respectively). Furthermore, the ratio CHL1_βNTF/CHL1_FL was reduced in BACE1Het and BACE1Het;5XFAD but not reach a significant difference when compared to control. These findings indicate that a 50% decrease of BACE1-mediated processing leads to an accumulation of CHL1_FL in absence of a reduction of CHL1_βNTF. Similarly, levels of CHL1_FL were significantly increased in the membrane fraction from 5XFAD mice compared to wild type mice (indicated as GGA3+/+) as reported in Fig. 9a and b. Levels of CHL1_βNTF were increased in the soluble fraction as well as in the total extract of 5XFAD mice compared to wild type (Additional file 7A–B and Fig. 9d). As observed in BACE1Het mice, the ratio CHL1_βNTF/CHL1_FL was reduced in the 5XFAD but did not reach a significant difference when compared to control most likely due to the fact that soluble fragments are detected even in absence of BACE1 in the soluble fraction (see BACE1KO fractionation). Another possible explanation is that while in the total extract CHL1_βNTF/CHL1_FL ratio was performed within the same sample, in the soluble and membrane fractions fragments are normalized against loading controls specific for each fraction.

To rule out the possibility that the increase in CHL1_FL observed in 5XFAD mice was due to increased expression we assessed the mRNA levels of Chl1 in hippocampi from both non-5XFAD and 5XFAD mice by RT-qPCR. The levels of Chl1 mRNA were similar in the hippocampus of 5XFAD mice compared with non-5XFAD mice (Additional file 8A). Furthermore, Chl1 expression was similar in 5XFAD males and females in spite of the increased levels of CHL1_FL observed in 5XFAD females compared to 5XFAD males (Additional file 8B and Fig. 9c). Altogether, these results indicate that CHL1 expression is not affected by transgenic APP expression in 5XFAD mice. An alternative explanation is that both reduced BACE1-cleavage and protein stabilization associated with aging result in increased levels of CHL1_FL and CHL1_βNTF.

These data suggest that the elevated transgenic APP might compete with CHL1 for BACE1 processing, resulting in the accumulation of full-length CHL1 in 12 months old 5XFAD mice. Such speculation was further confirmed with the observation that levels of CHL1_FL were significantly higher in female 5XFAD mice compared to males with matching genotypes (Fig. 9b). Since female 5XFAD mice have higher levels of transgenic APP than males with corresponding genotype (Fig. 8e), the competition of APP over CHL1 on BACE1 is furthered in female 5XFAD mice, resulting in a sexually differential stabilization of CHL1_FL.

In order to further support our finding that increased levels of APP are associated with decreased BACE1-mediated processing of CHL1, we analyzed hippocampal tissue from subjects affected by Down Syndrome (DS) carrying an extra copy of chromosome 21. The APP gene lies on chromosome 21 and as a consequence, levels of APP are increased in the brain of DS patients compared to controls. Soluble and membrane fractions were prepared from snap frozen hippocampus samples. Western blot analysis revealed a decrease in CHL1_βNTF/CHL1_FL ratio in brains from DS patients compared to controls owing to a significant decrease in CHL1_βNTF (Fig. 10a and b). Levels of CHL1_FL were increased but not reached a significant difference in DS vs. control samples. However, a direct correlation was observed between APP and CHL1_FL both in DS and control samples (Fig. 10c). Altogether these results indicate that increased APP levels reduce CHL1 processing not only in the brain of 5XFAD mice but also in human DS brains.