Heteromers of amyloid precursor protein in cerebrospinal fluid

Abnormal accumulation of β-amyloid peptide (Aβ) in the brain is characteristic of Alzheimer’s disease (AD). The Aβ peptide is generated by processing a large type I transmembrane spanning glycoprotein, the amyloid precursor protein (APP), through the successive action of proteolytic enzymes called secretases. Sequential processing of APP begins with either the action of α-secretase, or alternatively by β-secretase, followed by γ-secretase cleavage ([1]; see also Figure 1A). When APP molecules are cleaved by α-secretase within the Aβ domain, the generation of the Aβ peptide is precluded. When cleavage is carried out by β and γ-secretase, a 17-43 amino acid Aβ peptide is generated [2]. The Aβ40 peptide is the most abundant species, while the Aβ42 variant is the most amyloidogenic form of the peptide, associated with AD pathogenesis [3]. Cleavage by α- and β-secretase also generates soluble forms of APP, named sAPPα and sAPPβ, respectively, which are present in human CSF [4, 5] and have been postulated as potential new AD biomarkers. However, to date, no consistent change in CSF sAPPα and sAPPβ levels have been identified as associated to the AD condition [6]. Moreover, an unexpected positive correlation has been consistently reported between both forms, indicating a similar shift for sAPPα and sAPPβ levels [7‐10].

Interestingly, it has been demonstrated that full-length APP containing an intact cytoplasmic domain also exists as a soluble form (sAPPf) [11, 12]. The possible contribution of sAPPf when estimating levels of large sAPP fragments in CSF has thus far not been considered, nor has the existence of APP heteromers [13] in CSF. The possibility that sAPP exists in CSF forming oligomers is particularly relevant, since most quantifications of sAPPα and sAPPβ in CSF from Alzheimer’s disease subjects rely on ELISA determinations.

In this study, we investigate the presence of sAPPf in AD CSF, as well as the oligomerization of sAPP species, to determine how this may affect the measurement of sAPPα/sAPPβ levels by currently available ELISA immunoassays.

In this report, we confirm the presence of sAPPf in human CSF and characterize that sAPP species associate in heteromeric complexes. Our studies demonstrate that sAPP heteromers contribute to the estimation of sAPPα and sAPPβ levels by current standard ELISA methods.

Membrane-associated APP is usually detected as a group of 110-130 kDa proteins, and sAPPf in human CSF is expected to have an apparent molecular mass similar to that of cellular APP [11], while sAPPα or sAPPβ fragments are predicted to be ~5-10 kDa smaller than full-length APP [5]. These small size differences make it difficult to distinguish sAPPf from sAPPα or sAPPβ (see Additional file 1: Figure S1). Moreover, there are three APP isoforms expressed in the brain resulting from alternative exon splicing [15]; APP695, the predominant isoform in neurons consisting of 695 amino acids, and APP751 and APP770, both isoforms harboring an amino acid insert homologous to a Kunitz-type trypsin inhibitor (KPI) [16]. Therefore, the expected sAPP banding pattern in CSF is very complex, with three potential variants of each sAPPα and sAPPβ originated from APP695, 751 and 770 isoforms. Our present study adds complexity to this scenario, demonstrating that sAPPf is also present in human CSF.

In our study, no statistical differences were observed in the level of the broad sAPPf band of 110 kDa (attributable by its molecular mass to APP695) between probable AD patients and non-demented controls; however, its levels correlated with Aβ42 levels in the AD group; this may indicate an association with the pathological levels of Aβ42, and requires further research. The 130 kDa band (which should correspond to APP751/770) was found to be increased in AD CSF. Despite the low number of samples analyzed, and the difficulty in quantifying the immunoreactivity of the 130 kDa band, our data suggest that the analysis of KPI-containing sAPPf forms may be of particular interest and needs to be more specifically addressed. Interestingly, the proportion of KPI-containing isoforms appears significantly elevated in the soluble subcellular fraction of AD brains [17].

The origin or functional significance of sAPPf in CSF is currently not known. Early studies failed to demonstrate the presence of sAPP isoforms containing its C-terminal domain in human CSF [4], but later work suggested that sAPPf exists in this fluid [11], in plasma [18] and can be released from platelets [18, 19], chromaffin cells [11, 12] and different cell lines [11, 20]. In chromaffin cells, it has also been proposed that sAPPf could be released from a non-transmembrane APP population [12]. In cell culture, the levels of sAPPf are modulated by cell depolarization in a Ca²⁺-dependent manner, as well as by cholinergic receptor agonists and antagonists [11]. A mechanism for sAPPf to appear in CSF has yet to be determined, but passive release from brain cells or neuronal cell death may be possible, as recently observed for BACE1 [21]. Full-length BACE1 and other membrane resident protein, such as presenilin-1, are present in CSF [22].

Previous studies indicate that cellular APP may form high molecular-weight APP complexes [13, 23]. Co-immunoprecipitation and native gel analysis indicated an interaction among all sAPP forms in CSF, and their assembly into APP heteromeric complexes. This could be due to the self-association properties of APP by hydrophobic and/or ionic interactions resulting in the formation of heteromeric sAPP complexes which contain several forms of sAPP, including sAPPf. Thus, CSF heteromers differ from brain APP membrane-homodimers. Native gel experiments suggest the existence of large sAPP heteromers in CSF. Interestingly, the sAPPf forms released from platelets are more susceptible to sedimentation by ultracentrifugation than sAPP forms lacking the carboxyl terminal [18]; this could indicate a different degree of oligomerization or differences in the conformation of soluble aggregates compared to cellular dimers. Here, we observed differences during ultracentrifugation in sucrose density gradients containing Brij 97 detergent, where APP complexes from the CSF appear to be more unstable than brain APP dimers.

The nature and composition of sAPPf oligomers in CSF, as well the possibility that other proteins could also be part of the sAPP complexes, have yet to be deciphered. Based on our co-immunoprecipitation and ELISA analysis, we assume that all the possible sAPP heteromers are present in CSF.

Previous reports on the levels of sAPP in CSF draw a complex scenario. A strong positive correlation between sAPPα and sAPPβ levels has been consistently found in healthy elderly individuals and AD subjects in our and other studies [6, 7, 9, 10, 24]. Since the production of sAPPβ should be inversely proportional to that of sAPPα, this is an unpredicted finding. The changes in APP gene expression levels in the brain of AD subjects are modest [25, 26], and do not to explain the concurrent positive correlation in sAPPα and sAPPβ levels. It has been suggested that there may be a common factor that regulates the metabolism of APP by influencing the activity of both α- and β-secretases or by a mutual influence on both secretases [6]. Cross-reactivity between sAPPα and sAPPβ assays has also been suggested as a contributing factor for their unexplained correlation [27]; however, development of assays based on pan-specific antibodies further confirms this positive correlation. Instead, we cannot totally exclude the possibility that cross-reactivity also contribute in our determination of sAPP heteromers; although our results indicate that such a strong correlation may be explained by the presence of heteromers containing sAPPβ and sAPPα. These sAPPβ/sAPPα heteromers would contribute proportionally to the determinations of both sAPPα and sAPPβ by ELISAs, leading to this positive correlation.

In the same line, studies of sAPPα and sAPPβ levels in CSF likely would reflect potential changes in the balance between non-amyloidogenic and amyloidogenic pathways, but reports are also contradictory. Studies of sAPPα levels in CSF from AD subjects range from an increase [8, 27‐29], to no significant change [30‐33], or a slight decrease [4, 34‐37], compared to levels in NDC subjects. Similarly, some reports indicate that sAPPβ levels increase in AD compared to NDC cases [27], but others show no change or even a decrease [7, 32, 35, 37]. Simultaneous determination in CSF of sAPPα levels with respect to sAPPβ could potentially reflect APP brain metabolism, but comparative studies have again resulted controversial. Given that the α-secretase pathway is the predominant APP processing pathway, higher levels of sAPPα would be expected compared to sAPPβ levels in CSF from non-disease individuals. Some studies have reported higher levels of CSF sAPPα than sAPPβ in non-disease controls [11, 32], although most published reports do not support these results [7, 10, 24, 27]. Our results demonstrate that the presence of sAPPf, detectable by some ELISAs developed for sAPPα, should be considered a contributing factor for these contradictory findings. Thus, comparison of sAPPβ and sAPPα levels by IBL ELISA kits (using specific anti-APP antibodies which recognize the C-terminal of sAPPα or sAPPβ) showed higher amounts of sAPPβ in CSF than that of sAPPα. Nonetheless, comparison of sAPPβ with those sAPP levels obtained with the Novex Life kit does not confirm this appreciation. The Novex Life kit, like other kits based on the 6E10 antibody [11, 32], is based on an antibody recognizing the N-terminal part of the Aβ peptide for detection, thus binding sAPPf and sAPPα. The contribution of sAPPf to the estimation of sAPPα levels will explain discrepancies between the Novex Life and IBL kits. Altogether, our data suggest that sAPPβ is more abundant in CSF than sAPPα, and that sAPPf is the least represented. However, since all those ELISA kits do not discriminate between homodimers, heterodimers and large heteromers, an analysis of sAPPα and sAPPβ under conditions that favor the measurement of sAPP isoforms in their monomeric state is warranted to estimate their correct potential as AD biomarkers.

Finally, we have speculated about the abundance of sAPP dimers by our ELISA assays, taking into account that from these data we cannot discriminate sAPP dimers from large sAPP oligomers. Based on the Novex Life ELISA analysis after IBL-β immunoprecipitation, we presumed that the most abundant oligomers include sAPPβ, probably being sAPPα/sAPPβ, since the oligomers which have not been pulled-down, sAPPf/sAPPα, sAPPα/sAPPα and/or sAPPf/sAPPf, constitute altogether ~10% of the non-immunoprecipitated samples. Immunoprecipitation with the Sigma-Ct antibody followed by ELISA analysis for sAPPα and sAPPβ (IBL) revealed in both cases that sAPPf/sAPPα or sAPPf/sAPPβ heteromers represent ~30% of oligomers that include sAPPα and sAPPβ. We speculate that the sAPPα/sAPPα, sAPPf/sAPPf, but also sAPPβ/sAPPβ homodimers, would constitute the least abundant population. In any case, specific characterization of each sAPP oligomer in CSF, and potential differences in the AD condition, will require further research.

The use of CSF biomarkers for AD in clinical practice is challenging due to the high variability in their levels. Such variability has been partially attributed to different pre-analytical and analytical procedures between laboratories, highlighting the need to establish standardized operating procedures [38, 39]. CSF-sAPPα and sAPPβ have been proposed as promising CSF biomarkers for AD and other neuropathological conditions [40‐45], but have failed to meet expectations with their often contradictory findings. We describe that CSF-sAPP forms, including soluble full-length APP, assemble into oligomers and that heteromers contribute to the estimation of sAPP levels by ELISAs, and thus, potentially account for the underestimation of real sAPP levels. Our finding could explain the controversy about CSF-sAPPα and sAPPβ levels, and may change how these biomarkers are assessed. These findings indicate the need to design new assays for CSF-sAPP, with new conditions to selectively analyze sAPP monomers, which could unmask pathological differences, and to develop new approaches for sAPP heteromers, which could reveal new biomarkers.