A novel Aβ isoform pattern in CSF reflects γ-secretase inhibition in Alzheimer disease

Accumulation of amyloid β (Aβ) peptides in senile plaques in the cerebral cortex is an early event in the pathogenesis of Alzheimer disease (AD) [1]. The longest isoform of Aβ, consisting of 42 amino acids (Aβ1-42), is produced from amyloid precursor protein (APP) by sequential cleavage by β- and γ-secretase in the amyloidogenic APP-processing pathway (Figure 1A) [2]. β-Secretase activity originates from an integral membrane aspartyl protease encoded by the β-site APP-cleaving enzyme 1 gene (BACE1) [3‐6] whereas γ-secretase is an intramembrane-cleaving complex composed of at least four essential subunits with the presenilin (PS1 or PS2) proteins at its enzymatic core [7‐9]. γ-Secretase, which is one of the top targets for developing AD therapeutics with disease-modifying effects, cleaves the transmembrane region of APP to produce Aβ of variable length [10‐12], with Aβ1-40 being the most abundant isoform, whereas Aβ1-42 is most prone to aggregation [13, 14].

The main focus of using disease-modifying drugs is to inhibit brain Aβ production and aggregation and to increase Aβ clearance from the brain. Several drug candidates, including β-secretase and γ-secretase inhibitors or Aβ immunotherapy, are being evaluated in different phases in clinical trials. However, in a slowly progressive disorder like AD, evaluation of the clinical effect of a disease-modifying drug by using rating scales requires large patient numbers and extended treatment periods. This type of drug cannot be expected to have any early effect on symptoms, but instead may show a less-pronounced decline in cognitive function over years. Thus, a great need exists for biomarkers to identify and monitor the biochemical effect of disease-modifying drugs in AD clinical trials. Such biomarker data providing evidence directly in patients with AD that the drug has the predicted mechanism of action may be valuable for making a go/no-go decision for large and expensive Phase II or III clinical trials.

In animal studies, a decrease in soluble Aβ levels and Aβ accumulation has been reported in response to γ-secretase inhibition [15‐18]. Recently, in a human clinical trial, it was shown, by using enzyme-linked immunosorbent assay (ELISA), that the γ-secretase inhibitor LY450139 reduces the concentration of Aβ1-40 and Aβ1-42 in human plasma in a dose-dependent manner, whereas a significant decrease in human cerebrospinal fluid (CSF) Aβ levels was not detected [19]. However, by using a stable isotope-labeling kinetic (SILK) method, which can be used to measure central nervous system (CNS) production and clearance rates of total Aβ in the CSF of humans [20], it was shown that LY450139 significantly decreased Aβ production without affecting Aβ clearance in the CNS [21].

We recently identified a set of 18 N- and C-terminally truncated Aβ peptides in human CSF by using immunoprecipitation-mass spectrometry (IP-MS) and showed that CSF Aβ1-16 may be a novel positive biomarker for AD that was elevated in AD patients [22]. This novel biomarker for AD is generated through a previously unrecognized metabolic pathway by concerted β- and α-secretase cleavage of APP (Figure 1B) [23]. Further, the relative level of Aβ1-16, along with Aβ1-14 and Aβ1-15, has previously been shown to be elevated in cell media and in CSF from mice in response to γ-secretase inhibitor treatment [23, 24], suggesting that these shorter Aβ isoforms may be sensitive biomarkers for γ-secretase inhibitor treatment. Here, we tested the hypothesis that treatment of AD patients with the γ-secretase inhibitor LY450139 can be monitored by increased CSF levels of Aβ1-14, Aβ1-15, and Aβ1-16.

Representative CSF Aβ isoform mass spectra from an AD patient before treatment and after 14 weeks on the high dosage of the γ-secretase inhibitor LY450139 are shown in Figure 3a. The treatment greatly increased the mass spectrometric signal for Aβ1-14, Aβ1-15, and Aβ1-16, whereas the mass spectrometric signal corresponding to Aβ1-34 decreased as a response to treatment (Kruskal-Wallis test, P = 0.009, 0.01, 0.05, and 0.000041, respectively). The other Aβ isoforms reproducibly detected in all spectra (Aβ1-17, Aβ1-18, Aβ1-19, Aβ1-33, Aβ1-37, Aβ1-38, Aβ1-40, and Aβ1-42) did not differ significantly in intensity before and after treatment. The levels of the Aβ isoforms Aβ1-13 and Aβ1-20 were low, were not detected in all samples, and were therefore omitted from further analysis. The mass-to-charge (m/z) ratio of Aβ1-18 [M + H]⁺ is almost identical to the m/z of the second charge state of Aβ1-40 [M + 2H]²⁺ and therefore difficult to quantify; it was excluded from further analysis.

In the 100-mg group, after drug administration at week 14, the CSF levels of Aβ1-14, Aβ1-15, and Aβ1-16 were significantly increased by a mean of 57%, 21%, and 30%, respectively, whereas Aβ1-34 decreased 39% (see Figure 3b and Table 2) compared with study baseline. The same result was seen in the 140-mg group, Aβ1-14, Aβ1-15 and Aβ1-16 were increased by means of 74%, 35%, and 67%, respectively, whereas Aβ1-34 decreased 37% (Table 2). No statistically significant changes were seen for any other Aβ isoforms detected. α-sAPP showed a tendency to increased levels (13%) in the 140-mg group, whereas β-sAPP was unaffected by all the treatments (Figure 3c).

Intense multidisciplinary research efforts during the last decades have provided detailed knowledge on the molecular pathogenesis of AD, which has been translated into novel promising therapies with putative disease-modifying effects. Several promising drug candidates, such as Aβ immunotherapy and secretase inhibitors, are now being tested in clinical trials. However, because the predicted clinical effect of this type of disease-modifying drugs is a less-pronounced slope in the rate of cognitive deterioration, without any early symptomatologic effect, very large clinical trials with extended treatment periods will be needed to identify a beneficial clinical effect by using rating scales. Thus, biomarker evidence from smaller short-term clinical trials that the drug has the predicted biochemical mode of action directly in patients with AD would be valuable for making a go/no-go decision for expensive Phase III clinical trials. Thus, a great need exists for biomarkers to identify and monitor the biochemical effect of disease-modifying drugs in AD clinical trials.

The main focus with disease-modifying drugs is to inhibit brain Aβ production and aggregation and to increase Aβ clearance from the brain. γ-Secretase inhibitors have previously been shown to reduce Aβ1-40 and Aβ1-42 production and secretion in cells and to reduce soluble Aβ and amyloid plaque burden in mice [10, 15‐18, 23]. These results have made the γ-secretase complex one of the top targets for developing AD therapeutics. Here we show that the novel Aβ isoforms Aβ1-14, Aβ1-15, and Aβ1-16, together with Aβ1-34, may serve as sensitive biomarkers for γ-secretase inhibition by LY450139 in the CNS of AD patients.

In a previous study using ELISA measurements of CSF Aβ1-40 and Aβ1-42, the expected reduction of the peptides in response to LY450139 treatment was not found [19]. It was suggested that this lack of changes might be the result of a rapid transport of Aβ from CSF into plasma or that longer treatment duration may be required to identify changes [19]. In another study using the SILK method to examine whether an effect on Aβ production could be identified with LY450139 treatment, it was shown that Aβ production in the CNS decreased while Aβ clearance remained stable [21]. The lack of effect on CSF Aβ1-40 and Aβ1-42, despite the reduced Aβ production, may be because the different techniques are measuring different targets. SILK analysis of Aβ turnover requires that all Aβ isoforms are digested with trypsin before analysis, and a cleavage product consisting of Aβ17-28 is then measured by using MS [20]. Thus, total Aβ (that is, the mean of all of the very different Aβ isoforms that contain the Aβ17-28 sequence) is measured. This means that all longer isoforms detected in the present study (Aβ1-33, Aβ1-34, Aβ1-37, Aβ1-38, Aβ1-40, and Aβ1-42) will contribute to the mass spectrometric signal. The results presented herein suggest that the reduction in Aβ1-34, the generation of which is γ-secretase dependent [10, 11, 27], may contribute to the overall reduction of Aβ detected in CSF by using the SILK method.

Previous experimental studies on certain cultured cells expressing wild-type human APP have shown that γ-secretase inhibitor can increase in α- and β-secretase cleavage products along with the expected increase in APP C-terminal fragments (C99 APP-CTF) [28]. Further, recent data have shown that the levels of Aβ1-14, Aβ1-15, and Aβ1-16 are elevated in cell media and CSF from transgenic mice treated with γ-secretase inhibitors and that these shorter isoforms are derived from concerted cleavages of APP by β- and α-secretase, thus reflecting a third metabolic pathway for APP [23, 24]. Data in this study suggest that these shorter Aβ isoforms may be sensitive novel biomarkers for γ-secretase inhibitor treatment, even at doses that do no affect the CSF levels of Aβ1-40 or Aβ1-42. The increased levels of the isoforms detected after γ-secretase inhibitor treatment might be explained by an increased amount of substrate (that is, β-CTF or C99) (Figure 1), for α-secretase, after APP is cleaved by β-secretase.

Here we verify these findings directly in living AD patients (that is, a major increase of the shorter isoforms and a slight increase of α-sAPP, in CSF from AD patients treated with a γ-secretase inhibitor). The unchanged CSF levels of Aβ1-40 and Aβ1-42 are in agreement with previous studies [19] and suggest that these markers are less sensitive to detect γ-secretase inhibition as compared with Aβ1-14, Aβ1-15, and Aβ1-16, possibly because they may be influenced by other factors, such as brain amyloid load and/or Aβ oligomerization or that they are present at higher levels in CSF and that higher doses are needed to detect effects on these biomarkers in AD patients. Unchanged β-sAPP levels and only slightly elevated α-sAPP levels suggest that γ-secretase inhibition does not result in any major change in the release of these APP fragments to the CSF. It is at this stage unclear whether these fragments are degraded further or if they simply are present in such high concentrations in the CSF that a very high γ-secretase inhibitor dose is needed to detect an effect on CSF α-sAPP and β-sAPP levels.

This study suggests that Aβ1-14, Aβ1-15, and Aβ1-16 are positive and very sensitive biomarkers for γ-secretase inhibition and can be used to detect biochemical effects on APP processing in AD patients treated with LY450139. Long-term clinical trials are needed to reveal whether these biomarkers predict a beneficial clinical treatment effect.