Cathepsin B Degrades Amyloid-β in Mice Expressing Wild-type Human Amyloid Precursor Protein*

Background: The CysC-CatB axis affects levels of Aβ from hAPP with familial mutations. How it affects Aβ from wild-type hAPP remains unknown. Results: Enhancing CatB reduces and deleting CatB elevates levels of Aβ derived from wild-type hAPP. Conclusion: The CysC-CatB axis regulates Aβ degradation similarly regardless of familial mutations. Significance: Enhancing CatB activity as an Aβ-lowering strategy might be applicable in familial and sporadic AD. Accumulation of amyloid-β (Aβ), believed to be a key trigger of Alzheimer disease (AD), could result from impaired clearance mechanisms. Previously, we showed that the cysteine protease cathepsin B (CatB) degrades Aβ, most likely by C-terminal truncation, in mice expressing human amyloid precursor protein with familial AD-linked mutations (hAPPFAD). In addition, the Aβ-degrading activity of CatB is inhibited by its endogenous inhibitor, cystatin C (CysC). Reducing CysC expression markedly lowers Aβ levels by enhancing CatB-mediated Aβ degradation in hAPPFAD mice. However, because a vast majority of AD patients do not carry familial mutations, we investigated how the CysC-CatB axis affects Aβ levels in mice expressing wild-type hAPP (hAPPWT). Enhancing CatB activity by CysC deletion significantly lowered total Aβ and Aβ42 levels in hAPPWT mice, whereas CatB deletion increased Aβ levels. To determine whether neuron-derived CatB degrades Aβ in vivo, we generated transgenic mice overexpressing CatB under the control of a neuron-specific enolase promoter. Enhancing neuronal CatB activity in hAPPWT mice significantly lowered Aβ42 levels. The processing of hAPPWT was unaffected by increasing or ablating CatB activity. Thus, the CysC-CatB axis affects degradation of Aβ42 derived from hAPP lacking familial mutations. These findings support the notion that enhancing CatB activity could lower Aβ, especially Aβ42, in AD patients with or without familial mutations.

Generation of NSE-catB Transgenic Mice-The cDNA encoding full-length mouse CatB was cloned from the pCMV-SPORT6 vector containing mouse CatB (Addgene) by high fidelity PCR using the following primers (containing a HindIII site): 5Ј-TCT AAG CTT CCA GGA TGT GGT GGT CCT TGA TCC-3Ј and 5Ј-CTC TAA GCT TTT AGA ATC TTC CCC AGT ACT-3Ј. The PCR fragment was inserted into the NSE vector (15) with blunted HindIII linkers to generate the NSE-CatB plasmid. The orientation and sequence of mouse CatB were confirmed by sequencing. NSE-CatB was linearized by SalI digestion and purified by passing over an Elutip column (Schleicher & Schuell). The concentration of the linearized NSE-CatB construct was adjusted to 2 g/ml in injection buffer (5 mM Tris, pH 7.5 and 0.2 mM EDTA) and microinjected into the pronuclei of single-cell embryos harvested from C57BL/6 mice. Transgenic offspring were identified by PCR analysis of tail lysates. Two founder lines of NSE-catB transgenic mice were generated. The line with the higher expression level was crossed with the hAPP-I63 line to generate hAPP WT /NSE-catB mice.
A␤ and Soluble Amyloid Precursor Protein-␤ ELISA-Mice were perfused with 0.9% saline, and their hemi-brains were snap-frozen on dry ice and stored at Ϫ80°C. Hippocampal and cortical A␤ levels were measured by ELISA as described (16). The capture antibodies were 266 (for A␤1-x) and 21F12 (for A␤1-42), and the detection antibody was biotin-conjugated 3D6 (Elan, South San Francisco, CA).
Statistical Analysis-Statistical analyses were conducted with GraphPad Prism 5. Values are expressed as means Ϯ S.E. Differences among multiple means with one variable (catB or cst3 genotype) were evaluated by one-way analysis of variance (ANOVA) and Tukey-Kramer post hoc tests. Differences between two means were evaluated using unpaired Student's t tests. p Ͻ 0.05 was considered significant.

CatB Ablation Elevates A␤ Levels without Affecting hAPP
Processing in Transgenic Mice overexpressing hAPP WT -CatB degrades A␤ in hAPP FAD -J20 mice without affecting the processing of hAPP FAD (7). To directly determine the effects of CatB deletion on hAPP WT , we crossed catB Ϫ/Ϫ mice with hAPP WT -I63 mice, which express hAPP at a level similar to hAPP FAD -J20 mice but produce much less A␤. The levels of total A␤ (A␤1-x) in the hippocampuses and cortices of hAPP WT mice were slightly increased by deleting one catB allele and significantly increased by complete deletion of catB (Fig. 1, A and C). The levels of A␤1-42 were also significantly increased in the cortices of hAPP WT /catB Ϫ/Ϫ mice and were slightly increased in the hippocampuses, although not statistically significantly in the latter (Fig. 1, B and D). Nevertheless, these results suggest that endogenous CatB lowers the level of A␤ derived from hAPP WT. Deleting CatB did not affect the levels of s-APP␤ in the cortex, as determined by ELISA (Fig. 1E), indicating that CatB may not function as a ␤-secretase for hAPP WT . In addition, Western blot analyses with a C terminus-specific antibody revealed that the levels of full-length hAPP (FL-hAPP) and ␣and ␤-CTFs were similar in hAPP WT mice with or without CatB (Fig. 1, F and G), further supporting that neither the ␤nor ␣-cleavage of hAPP WT is affected by CatB deletion.
Neuronal Overexpression of CatB Reduces the Levels of A␤ Derived from hAPP WT -CysC reduction enhances CatB activity in the brain, where CatB is expressed in neurons and glia. To determine whether neuron-derived CatB degrades A␤ in vivo, we established a transgenic mouse line that overexpresses mouse CatB under the control of a neuronal enolase promoter (NSE-catB) (Fig. 3A). CatB activity in the cortex was significantly higher in NSE-catB mice than in non-transgenic controls (Fig. 3B), indicating that overexpressed CatB is functionally active in vivo. Moreover, CatB staining was enhanced in hippocampal neurons expressing NeuN (Fig. 3C) but not in glial fibrillary acidic protein-expressing astroglia (Fig. 3D) or Iba1-positive microglia (Fig. 3E). These results suggest that CatB is increased predominantly in the neurons of NSE-catB mice.
We next crossed NSE-catB mice with hAPP WT mice and determined the effects of elevated neuronal CatB on A␤ and hAPP processing. Enhancing neuronal CatB in hAPP WT /NSE-catB mice significantly lowered the levels of A␤1-42 in both the hippocampus and cortex (Fig. 4, B and E). The levels of total A␤ were modestly lower in hAPP WT /NSE-catB mice compared with hAPP WT mice (Fig. 4, A and D). Increased CatB activity in neurons also significantly reduced the relative abundance of A␤42 in both the hippocampus and cortex (Fig. 4, C and F). These results suggest that neuronal CatB reduces A␤1-42 preferentially, likely by truncating at the C terminus as described previously (7). Neuronal overexpression of CatB did not affect the levels of FL-hAPP, ␣-CTF, or ␤-CTF in the hip- pocampus (Fig. 4G) or in the cortex (Fig. 4H), confirming that neuronal CatB does not affect hAPP WT processing.

DISCUSSION
This study shows that CysC-CatB affects A␤ levels in hAPP WT mice in a similar fashion as in hAPP FAD mice as described in our previous studies (7,8). CatB removal elevated the levels of A␤ in hAPP WT mice, as in hAPP FAD mice (7). In mice expressing hAPP WT , neuron-derived CatB effectively reduced the levels of A␤1-42, the most pathogenic of the A␤ peptides in the brain (19). Removing CysC, the endogenous inhibitor of CatB, in hAPP WT mice lowered total A␤ and A␤1-42 levels. However, hAPP WT processing was unaffected by removal of CatB or CysC. Enhancing neuronal CatB also had no effect on hAPP WT processing. Thus, targeting the CysC-CatB axis to enhance A␤ degradation might be applicable not only in AD patients with FAD mutations but also in sporadic AD cases.
Our results differ from a previous study that showed that deleting CatB reduces the levels of s-APP␤, A␤1-40, and A␤1-42 in hAPP WT mice (9). In contrast, we found that total A␤ (A␤1-x) and A␤1-42 levels were reduced, whereas the levels of s-APP␤ were unaffected. Although the exact reason underlying this discrepancy remains unclear, one key difference between the two studies is the A␤ ELISA used to measure total A␤. We used a well established ELISA that allowed us to detect all A␤ species starting at position ϩ1 and ending at or after position ϩ23, which would include A␤1-40, A␤1-38, and A␤1-42 among others (16), whereas previous ELISA probing total A␤ levels measured A␤1-40 specifically (9). Because CatB could convert A␤42 to A␤40 and A␤38, it is conceivable that CatB deficiency could lead to a reduction in A␤40 and/or A␤38 because A␤42 could not be cleaved to generate A␤40 and/or A␤38. In support of this notion, a recent study showed that enhancing CatB activity pharmacologically significantly reduced A␤x-42 but increased levels of A␤1-38 (20).
To determine the effects of the CysC-CatB axis on A␤ derived from hAPP WT , we used three approaches to modulate CatB levels. Besides deleting CatB, we enhanced CatB activity either by deleting CysC or by overexpressing CatB in neurons. In both cases, enhancing CatB activity significantly lowered A␤1-42 levels. Moreover, the enhanced activity reduced the relative abundance of A␤1-42, supporting the notion that enhanced CatB activity reduces A␤1-42 preferentially, most likely by C-terminal truncation. Consistent with these findings, systemic injections of the CatB-enhancing drug benzyloxycarbonyl-Phe-Ala diazomethyl ketone significantly reduced A␤ deposition and A␤x-42 levels in two independent AD models, even at advanced ages (10 -12 and 20 -22 months) (20). In contrast, deleting CatB in hAPP WT mice led to a modest increase in A␤ levels. One possible explanation is that the physiological concentration of endogenous CysC is much greater than that of CatB (21). Because the A␤-degrading activity of CatB is normally inhibited strongly by CysC, the effects of removing CatB would be modest. Indeed, CatB ablation had more robust effects on A␤42 levels when CysC levels were reduced (8).
In contrast to the effects on A␤ levels, modulating CatB or CysC did not affect the levels of s-APP␤, ␣-CTF, ␤-CTF, or FL-hAPP, suggesting that CatB is not involved in ␤or ␣-cleavage of hAPP with or without FAD mutations. However, in another study, deleting CatB in hAPP WT mice reduced the levels of A␤ and ␤-CTF (9). The reason for this discrepancy is unclear. However, it is important to note that CatB activities were altered with three independent genetic approaches in hAPP WT mice in the present study, none of which led to changes in ␤or ␣-CTF. It is thus very unlikely that CatB acts as a major ␤-secretase on hAPP WT . Indeed, ablation of BACE1 completely abolished processing of the ␤-secretase site and A␤ generation in PDAPP mice, which overexpress hAPP V717F with no mutations at the ␤-secretase site (22), further supporting the notion that BACE1, not CatB, is the major ␤-secretase of hAPP with or without FAD mutations.
Accumulating evidence supports the notion that disruption of substrate proteolysis within lysosomal pathways is a mechanism underlying early pathological changes in AD (23). The importance of lysosomal dysfunction in AD pathogenesis is further supported by the discovery that mutant presenilin-1 linked with early-onset AD disrupts autophagy directly by impeding lysosomal proteolysis (24). Our studies of the CysC-CatB proteolytic axis in hAPP mice provided one of the first examples of lysosomal pathways in A␤ degradation (7,8). In agreement with our findings, deleting cystatin B, another endogenous cysteine protease inhibitor, rescued autophagic-lysosomal pathology, reduced the accumulation of intraneuronal A␤ and plaque load, and prevented behavioral deficits in hAPP mice (25). Indeed, deleting cystatin B accelerated protein turnover and enhanced the activities of lysosomal enzymes, including CatB, suggesting that the enhanced activities of CatB or other lysosomal enzymes could be responsible for the beneficial effects of cystatin B deletion.
CatB is expressed in various cell types in the brain, including neurons and microglia. Previously, we showed that A␤ and A␤42 levels in primary neuronal cultures were reduced by enhancing neuronal CatB and elevated by inhibiting CatB (7). By enhancing neuronal CatB in hAPP WT /NSE-catB mice, we showed unequivocally that neuron-derived CatB reduces A␤42 in vivo. The modest A␤-lowering effects of neuronal CatB suggest that CatB in non-neuronal cells could also be involved in degrading A␤. For example, CatB is highly expressed in cultured microglia, and its expression levels are elevated by treatment with A␤42 in culture (26,27). Moreover, in cultured microglia, inhibition of CatB prevented the degradation of oligomeric A␤, whereas inhibitors of neprilysin, matrix metalloproteinases, or insulin-degrading enzymes had no effect (28). In hAPP FAD mouse brain, we previously observed strong CatB immunoreactivity in microglia surrounding the plaques (7). However, in hAPP WT /NSE-catB mice, we did not detect clear CatB immunoreactivity in microglia. One possible explanation is that CatB expression is very low in quiescent microglia in mice with low levels of A␤42. Indeed, CatB expression is FIGURE 4. Neuron-derived CatB overexpression reduces levels of A␤ but does not affect hAPP processing in hAPP WT mice. A and B, ELISA measurements of hippocampal A␤1-x (A) and A␤1-42 (B) in 4-month-old hAPP WT and hAPP WT /NSE-catB mice. C, the A␤1-42/A␤1-x ratio was significantly lower in hAPP WT /NSE-catB mice than in hAPP WT mice (n ϭ 8 -12 mice/genotype). **, p Ͻ 0.01 (unpaired Student's t test). D and E, ELISA measurements of cortical A␤1-x (D) and A␤1-42 (E) in 4-month-old hAPP WT and hAPP WT /NSE-catB mice. F, the A␤1-42/A␤1-x ratio was significantly lower in hAPP WT /NSE-catB mice than in hAPP WT mice (n ϭ 6 -8 mice/genotype). *, p Ͻ 0.05 (unpaired Student's t test). G and H, quantification of hippocampal (G) and cortical (H) levels of FL-hAPP (relative to those of GAPDH) and ␣and ␤-CTFs (relative to those of FL-hAPP) in hAPP WT mice by Western blotting (n ϭ 8 mice/genotype). Values are means Ϯ S.E. directly up-regulated by A␤42 in cultured microglia (26,27), which might help explain the readily detectable CatB in microglia surrounding the plaques but not those away from the plaques (7). Further study is needed to determine the importance of microglial CatB in A␤ degradation and clearance in vivo.
Our results support an important role for neuronal CatB in A␤ degradation. However, where the CysC-CatB axis acts to modulate A␤ degradation at the subcellular level remains unknown. In most pyramidal neurons, mature and proenzyme forms of CatB are present in early endosomes (11,29). Consistent with this finding, CatB in endosomes catalyzes C-terminal truncations of epidermal growth factor and insulin-like growth factor and inhibits their signaling (30,31). Although CysC is generated mainly in astroglia, CysC immunoreactivity is found in neurons with a punctate distribution, which co-localize with CatB in AD brains (32). Thus, CysC could be taken up by neurons or microglia into the endocytic pathway and inhibit the A␤-degrading activity of CatB. Regardless of the specific sites at which the CysC-CatB axis regulates A␤ degradation, our findings support enhancing proteolytic activity of CatB as a potential new A␤-lowering strategy in sporadic AD.