Amyloid β modification is a complex process that occurs both enzymatically and non-enzymatically. Many proteins are already shown to interact with Aβ leading to alteration of its structure or repair of pathogenic modifications. However, for many modified Aβ species, purified from AD brain tissue, the source of origin remains unknown (Kummer and Heneka, 2014).

Prior to discussing the role of modified Aβ in sAD, we will first focus on the life cycle of the Aβ molecule from its formation to its degradation or aggregation; taking into consideration, all modifications along the way. Beta-amyloid is produced via proteolytic cleavage of APP protein by beta-secretase (BACE) and gamma-secretase (Huang and Mucke, 2012); this is termed the amyloidogenic pathway. The non-amyloidogenic pathway is mediated by alpha-secretase (ADAM10). It is important to mention that cleavage by gamma-secretase is imprecise and results in production of an Aβ peptide ranging from 37 to 43 amino acids in length; notably the 42 residue species is considered to be the most pathogenic (Haass and Selkoe, 2007). Production of Aβ may occur at three different sites: on the plasma membrane, in the ER/Golgi or in endocytic vesicles. This decision helps determine its fate and defines the set of possible modifications that can be made to the peptide, as different Aβ-modifying enzymes are assigned to specific cellular compartments (Hartmann et al., 1997; Thinakaran and Koo, 2008). In Figure 1 below, we present a putative scheme for the amyloid peptide modification process inside and outside of the cell and both in solution and as aggregates (Figure 1). Some of the modifications present are enzymatic, some are triggered by low-molecular compounds such as peroxynitrite or 4-hydroxynonenal (HNE), and two of them are spontaneous, namely racemization and isomerization. Formation of N-truncated amyloid has not been well studied, however it is possible that it originates from proteolytic cleavage by aminopeptidase A (ENPEP), meprin or BACE or alternatively via non-enzymatic hydrolysis of peptide bonds (Kummer and Heneka, 2014). A huge body of evidence supports the pathogenic role of individual Aβ modifications; however, no research has been done to investigate the orchestrated action of different modifications on a single molecule of amyloid peptide. Additive or synergistic effects of such modifications may potentially increase the pathogenic properties of Aβ peptide far above the level of the widely studied intact Aβ. These modifications can promote accumulation of amyloid and plaque formation as they hamper its clearance and increase aggregation (Kumar, 2011; Kozin S. A. et al., 2013). Another blind spot in the studies of Aβ PTM is its connection with tau pathology. Tau hyperphosphorylation (HP) is presumably induced by Aβ, leading to systemic brain pathology (Oddo et al., 2006) and this transition might be caused by modified Aβ peptides. However, the association of HP-tau and Aβ modifications was only studied and observed for pyroglutamylated amyloid peptide (Mandler et al., 2014). We propose that studies of the relationship between Aβ PTM and tau pathology may contribute substantially to the understanding of AD development.

The principal fact that drew our attention to amyloid PTMs as a presumable cause of sAD, is that the PTM process is disrupted with aging. It was shown both directly in studies where the accumulation of modified proteins was measured (Levine and Stadtman, 2001), and indirectly as we know that proteostasis itself is disturbed in the aged body (Dubnikov and Cohen, 2015; Labbadia and Morimoto, 2015). It is known that reactive oxygen species (ROS) production and neutralization is destabilized in the aged body due to elevation of NOX activity (Dasuri et al., 2013) and an increase in mitochondrial respiratory chain leakage that is accompanied by an accumulation of mutations in mitochondrial DNA (Bratic and Larsson, 2013). HNE is a product of lipid peroxidation and its production increases with aging as a side effect of chronic oxidative stress (Castro et al., 2016). It has also been shown that nitric oxide synthase is upregulated in AD; however, it is not clear whether it is a normal part of the aging process or a pathological event (Domek-Łopacińska and Strosznajder, 2010). Meanwhile, isomerized and deaminated proteins have a tendency to accumulate naturally in an aging organism, and in carboxyl methyltransferase-deficient mice damaged proteins have also been shown to accumulate in the brain (Kim et al., 1997; Clarke, 2003). The phosphorylation process is regulated by balancing kinase and phosphatase activity and is also disrupted with aging (Magnoni et al., 1991; Rajagopal et al., 2016; Thomas and Haberman, 2016). Citrullination is another modification that is known to increase in aged body (Osaki and Hiramatsu, 2016). The important point is that modifications can create pathogenic networks with a positive feedback. It was shown that ROS-induce dityrosine crosslinking of Aβ results in formation of stable and poorly degradable oligomers (Al-Hilaly et al., 2013). Aβ increases ROS production (Butterfield and Swomley, 2012), which then promotes further inhibition of the amyloid clearance system and may result in a positive feedback-driven cascade of accumulation; such a cascade has already been shown for the Aβ and HNE interaction (Ellis et al., 2010). Taken together, all of these findings make it probable that disturbance of Aβ modification processes with age leads to the rise of different pathogenic processes including AD.

Since we propose an aging-related disturbance in the Aβ modification process as a cause of AD, one may ask whether every aging individual is destined to suffer from Aβ accumulation, AD, and a resulting steady cognitive decline with age. This does not happen and AD obviously requires additional triggers besides senescence-related pathogenic modification of Aβ peptides. These triggers are widely discussed and a plethora of work has been conducted to identify AD risk factors. The risk factors identified include: smoking, sleep deprivation, brain trauma, diabetes, bacterial infections, viral infections, and gut microbiota alteration (Itzhaki et al., 1997; Miklossy, 2008; Kang et al., 2009; Naseer et al., 2014; Reitz and Mayeux, 2014). The most studied trigger of AD is genetic background. According to estimations, based on human pedigree analysis, up to 80% of all Alzheimer's cases are hereditary (Bergem, 1994). The first genes that were identified as genes in which mutations lead to fAD were: APP, BACE, and gamma-secretase genes, called PSEN1 and PSEN2. They are responsible for most cases of familial AD and can dramatically increase overall Aβ production or shift the production ratio in favor of Aβ 1-42 (Bertram et al., 2010). Increased Aβ burden results in early amyloid accumulation and this in turn leads to an early-onset development of brain pathology (Huang and Mucke, 2012). However, fAD only accounts for about 1% of registered AD cases and late-onset AD (LOAD) or sAD is more or less beyond prediction (Campion et al., 1999; Bertram et al., 2010). In sAD, ApoE gene variants were associated with an increased risk of disease and could account for up to 20% of LOAD cases (Ertekin-Taner, 2010). However, to date the efforts to identify other sAD-modifying genes with a comparable magnitude of influence have been unsuccessful and we suggest that future work should focus on the genetics of Aβ-modifying enzymes. Currently an association between genetic variants of Aβ-modifying enzymes and AD is only shown for NOS2 (Akomolafe et al., 2006) and QPCT (Saykin et al., 2010), however for the latter the association does not have genome-wide significance. It is very possible that the lack of such associations is due to the nature of the tool that was used in prior studies to identify disease-modifying genes. The primary tool for such investigations is genome-wide association studies (GWAS), which have brought many gene-disease associations to our attention over the years (Singleton and Hardy, 2016). However, GWAS usually lacks full-genome coverage and fails to detect statistically significant associations with small effects (Naj et al., 2017). GWAS findings are dependent on a chosen cohort and many candidate genes are often thrown away. GWAS associations do not permit an inference of causation (Naj et al., 2017), so the role of individual genetic studies based on additional data is not diminished. Most enzymes featured in Figure 1 are already associated with other genetic pathologies, including neurological diseases; ENPEP is associated with Koch Hypertension (Kato et al., 2011); QPCT is associated with schizophrenia and frontotemporal dementia (Zhang et al., 2016); and PCMT1 with premature ovarian failure (Pyun et al., 2009). Variants of these may likewise be important for the development of AD. Such guidance may facilitate further genetic studies taking into consideration the potential synergy between the impairment of different modifications.

Aβ modifications and genetic alterations is a vast, yet poorly studied field with the potential to contribute substantially to the understanding of AD pathogenesis. The modification process results in the formation of pathogenic Aβ species, the level of which may increase with age and due to hereditary factors. To summarize, we hypothesize that modification of Aβ is a major contributor to sAD and targeting of the modified peptides or modification enzymes could serve as a novel therapeutic mechanism or provide a new means of diagnosis.

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

The study was funded by the Russian Science Foundation (grant #14-24-00100).