To our knowledge, the possibility that PS1 can be assessed in the CSF has thus far not been considered. Here, we demonstrate that heterodimeric complexes composed of NTF and CTF PS1 are detectable in CSF. To date, the presence of soluble CTF-PS1 has only been demonstrated in culture medium of primary neurons and HEK cells .
The majority of mature brain PS1 protein is present as proteolytic NTF- and CTF-PS1 fragments . The existence of 100–150-kDa complexes containing both endogenous NTF and CTF-PS1 has only previously been described in cultured cells . PS1 exhibiting higher molecular mass bands in SDS-PAGE, has also been described in cellular models over-expressing the protein [24, 25]. These high molecular mass bands of PS1 were detected by alternative anti-NTF and CTF-PS1 antibodies, confirming that heterodimers contain both PS1 fragments. Triton X-100, a common detergent frequently used to extract and solubilise membrane bound proteins, disassembles the 100–150-kDa complexes , and these PS1 heterodimeric species have therefore been mischaracterized in the many previous studies. The sensitivity of these complexes to high temperatures during sample preparation may contribute to its mischaracterization. The presence of the minor 29-kDa NTF and 20-kDa CTF species was only clearly detectable in post-mortem CSF, where artefacts are likely to appear. Similarly, the 50-kDa species was only resolved as a major PS1 band in immunoprecipitates after elution at acid pH.
Ultracentrifugation in sucrose density gradients containing the detergent Brij 97 confirms the existence of stable complexes of 100–150-kDa, and also of large complexes which sediment in regions closer to 200 and 250 kDa. These large complexes are unstable and resolve as 50-kDa components during electrophoresis under denaturing conditions.
γ-Secretase exists on the plasma membrane as an intact complex composed of its subunit components [15, 26], at a stoichiometry of presenilin:PEN-2:nicastrin:APH-1, 1:1:1:1 . However, PS1 can associate intra-molecularly to form higher order complexes , where nicastrin, APH-1 and PEN-2 do not seem to be required for its hetero- and homodimerization . Interestingly, our studies have identified APH-1 and PEN-2, but not nicastrin in the PS1-complexes. γ-Secretase activity depends on the presence of all four components of the complex, including nicastrin [30, 31]. As we were unable to detect γ-secretase activity in human CSF, PS1 complexes in CSF are therefore likely to be the result of nonspecific aggregation of the protein. PS1-CSF complexes appeared as a smear (from ~200 to 1000 kDa) and not as defined complexes on blue native-PAGE. PS2 has been shown to form similar but independent γ-secretase complexes in cell membranes, and PS2 does not co-precipitate with PS1 fragments in membrane extracts [7, 32]. We were able to demonstrate that fragments of PS2 are bound to the same PS1 complexes in CSF. As presenilins are proteins with large numbers of hydrophobic regions, presenilin fragments may be highly unstable in CSF, and spontaneously form complexes. Indeed, like many membrane proteins, PS1 has exhibited a tendency to aggregate under non-native conditions [11, 12]. Thus, in normal conditions, the amount of free CTF and NTF in CSF should be very low as PS1 predominantly exists as complexes. Preliminary analysis indicates that PS1 complexes do not contain some of the abundant CSF proteins known to bind hydrophobic proteins (e.g., albumin). Further characterization of these complexes is yet to be conducted. The possibility that other proteins, such as Aβ, may also be part of the PS1 complex needs to be addressed. Similarly, the source and mechanism of how PS1 appears in CSF is yet to be determined. Active secretion is unlikely, and it is still unclear if passive release from brain cells or neuronal death may be major contributing factors, as recently observed for BACE1 .
Stable 100–150-kDa complexes are more abundant in AD CSF, and this may reflect differences in the hydrophobic and ionic properties of PS1 complexes formed under amyloidogenic conditions. In this context, it is also remarkable that the 29-kDa NTF, detectable only in post-mortem CSF, is less abundant in AD samples, indicating that complexes formed under amyloidogenic conditions are particularly stable. Whether PS1 complexes differ between AD and non-demented subjects requires further research.
It is likely that molecular aberrations in the AD brain are reflected in the CSF. Although it cannot be excluded that the accumulation of Aβ in AD may be partly due to deficient clearance of the peptide, an increase in the generation of Aβ is plausible. Thus, an increase in activity and/or levels of β- and γ-secretase protein components should be expected. These catalytic effectors of β- and γ-secretase are key enzymes in pathological amyloid processing and the subsequent generation of Aβ constitutes a central event in AD progression. However, it is still unclear if γ-secretase activity is altered. Reported levels of PS1 in AD brains have been contradictory. Reports have displayed an increase [34, 35], unchanged levels  or even a decrease [37, 38] in levels, compared to levels in non-demented brains. At the transcriptional level, early reports indicate no differences between PS1 mRNA levels in AD brain compared to controls . However subsequent research suggests that PS1 mRNA levels in human AD brains are significantly higher than in those with no dementia [34, 40]. Interestingly, we have shown that Aβ peptide treatment of cultured cells is able to induce increases in cellular PS1 levels , probably as part of a vicious cycle of Aβ generation.
In our analysis, ventricular post-mortem samples display large differences in the level of PS1 and the stability of its complexes. Lumbar CSF samples only display differences in PS1 complex stability as resolved by sucrose density gradients. Thus, despite high levels of PS1 correlating with an increased proportion of stable complexes in post-mortem CSF, results in lumbar CSF suggest that the early and more significant phenomenon is the change in the dynamics of the assembly of PS1 complexes, a change that appears to be a better marker for discriminating between pathological samples than total PS1 protein levels alone. In any case, the comparison between the two different sets of CSF samples is difficult, as the potential existence of artefacts in the post-mortem samples and the possibility that ventricular and lumbar CSF samples may differently reflect brain protein content. Moreover, our collection of lumbar CSF samples display clear differences for classical AD biomarkers, but the inherent uncertainty in clinical diagnosis should indeed be considered. Although, the more obvious difference between the CSF collections is that post-mortem CSF reflects late stages of the disease, whereas lumbar CSF corresponds to earlier stages of disease in patients. Nonetheless, CSF PS1, combined with other biomarkers, may constitute a potential new marker for AD. Furthermore, CSF PS1 may have value as a marker of disease progression or for monitoring treatment.
As there is still a lack of reliable blood biomarkers for neurodegenerative disorders, it is important to assess if levels in plasma of new potential biomarkers correlate with AD. PS1 is expressed in many peripheral organs as well as the brain [42, 43]. The presence of PS1 complexes in the serum of PS1 cKO mice, in which PS1 is specifically absent in neurons of the forebrain, while mostly absent in CSF, suggests that plasma and CSF PS1 may have distinct cellular origins. Although a small contribution of brain PS1 to plasma levels cannot be discounted, our results indicate that PS1 detection in plasma will not be an efficient marker for brain disorders.