Recent breakthroughs, especially in-depth and large-scale omics studies of brain tissue, have dramatically extended our discovery of molecular pathogenesis in AD [
18,
49]. New dysregulated genes/proteins/pathways have been increasingly identified and linked to AD pathogenesis, suggesting of multifactorial pathologies in AD. However, in-depth proteome discovery in the proximal body fluids (e.g. CSF and serum/plasma) is still rare, largely due to technical challenges to address the complexity of CSF and serum/plasma proteomes. The relationship between diverse brain pathologies and protein alterations in body fluids is not fully explored at the proteome level. Furthermore, the variable results obtained from studies evaluating proteins involving in amyloid and tau pathology as disease biomarkers underlined the importance of novel biofluid biomarkers [
68]. To meet the challenges to analyze CSF and serum/plasma proteomes in AD, we have recently developed an in-depth biofluid profiling platform [
25] that combines un-depleted biofluid sample processing, multiplexed TMT labeling, extensive two-dimensional LC fractionation and high-resolution tandem mass spectrometry. The platform enables the quantification of 5941 and 4826 proteins in CSF and serum, respectively, providing the most in-depth biofluid proteome landscape so far for the AD community.
In addition to the issue of proteome coverage, reproducibility is often another concern in many of previously published AD biomarker studies. Even some novel CSF protein biomarker candidates have been proposed, many of them, however, are not successfully repeated across different laboratories, distinct proteomic platform, and/or independent cohorts, raising a substantial bottleneck for selecting reliable candidates for large-scale validation. To address this issue, we systematically integrated our ultra-deep CSF proteome with two other discovery-driven deep CSF proteomic studies in AD, resulting in 6 biomarker candidates that were repeatedly emerged in at least two independent studies including SMOC1, C1QTNF5, OLFML3, SPON1, SLIT2 and GPNMB. Remarkably, all of them were reported to be highly linked to AD pathogenesis in brain tissue [
18]. SMOC1 has been shown to accumulate in plaque structures of AD in brain. The expression levels of SMOC1, OLFML3, SLIT2, and GPNMB were highly correlated with the Aβ level in AD brain, and these findings were also recapitulated in the 5xFAD mouse model [
18]. Consistently, SMOC1 and GPNMB were reported to be CSF biomarker candidates for AD in other recent biomarker studies [
62,
69]. These proteins represent the most promising CSF biomarker candidates of AD for future large-scale studies.
Mitochondrial function and energy metabolism are known to be severely compromised processes repeatedly reported in AD [
18,
70]. Emerging lines of evidence suggest the growing importance of mitochondria damage and energy defects in AD pathogenesis [
70], and mitochondrial deficit is proposed as a major hallmark of AD pathogenesis besides amyloid and tau pathologies [
70]. Studies in a
C. elegans model expressing pan-neuronal human Aβ show that metabolic stress is a primary pathogenic event [
71] and impaired mitochondrial calcium efflux contributes to disease progression [
72]. Enhanced mitochondrial proteostasis may reduce amyloid-β proteotoxicity [
73] and NAD+ supplementation normalizes key Alzheimer’s features and DNA damage responses in an AD mouse model [
74]. Mitochondria have gradually been recognized as a major novel therapeutic target in AD [
70]. In this study, we identified consistent and evident mitochondrial protein decreases in AD CSF and serum samples, which have rarely reported until the availability of the deep CSF/serum profiling (our reference datasets) [
25,
33]. This is understandable because, as our analysis suggested, high proteome coverage is a prerequisite to detect mitochondrial changes due to their low abundance, explaining why they are missing in numerous previous biofluid studies of shallow proteome coverage. Although the causative factors of mitochondrial dysfunction in AD are not fully understood, we believe that the mitochondrial changes in the cortex, CSF and serum are highly associated in AD based on several lines of evidence. Mitochondrial changes can co-occur with amyloid deposition early in the brain of asymptomatic cases with amyloid pathology, as well as mild cognitive impairment subjects [
18,
33,
75]. Amyloid peptides have been reported to directly aggregate in mitochondrial compartment [
75,
76]. Recently, vascular deposits of Aβ peptides (amyloid angiopathy) are increasingly recognized as a common pathology in AD cases, supporting that Aβ peptides circulate within the interstitial fluid, including CSF, and blood vessels through perivascular (e.g. lymphatic) drainage pathways during the crosstalk between the brain and the vascular system [
77]. As Aβ can form vascular deposition, it is likely that Aβ could lead to mitochondrial damage in the vascular system. This local Aβ-induced mitochondrial damage may partially address an important question – where is the origin of the identified mitochondrial proteins? Interestingly, emerging data suggest that mitochondria can be released into extracellular space, and transferred between cells [
78], although mitochondria have traditionally been known as the intracellular powerhouse. For instance, neurons can transfer damaged mitochondria to astrocytes for disposal and recycling, and astrocytes can also release mitochondria to neurons under stress [
79]. Astrocytic mitochondria may also be released to the CSF as a biomarker for evaluating brain integrity, with low CSF mitochondrial quantity and activity indicating brain damage [
80]. Nevertheless, the origin of the mitochondria proteins is worth future investigation. We acknowledge that our results only indicate a correlation between mitochondria changes in proximal body fluids and brain lesions in AD. Further studies are clearly required to understand the mechanism behinds the associated mitochondrial protein changes in the cortex, CSF and serum in AD.
Here we demonstrate that a mitochondrial signature is the most significant and consistent changes detected across human brain cortex, CSF and serum in AD, and it has been recapitulated in the 5XFAD mice as well. It has been mentioned that these mitochondrial changes can only be confidently detected in an ultra-deep proteomic setting. This exciting finding provides a strong rationale not only for the development of disease diagnostic biomarkers but also the implementation of novel prognostic biomarkers for therapeutic strategies targeting mitochondria in AD.