Here we show that heparin-fractionation of plasma followed by proteomic analyses is reproducible methodology to enrich a subset of plasma proteins that distinguish AD plasma from controls. More than 2800 plasma proteins spanning over 10 orders of magnitude in protein concentration in plasma can be reliably assessed using this methodology. Contextualizing these data by integrating them with previous studies and our extensive human AD brain and CSF data, we show that Hp-enriched AD plasma proteome provides novel insight into the pathobiology of AD by reflecting protein changes in AD brain.
Our primary rational for exploring heparin enrichment of plasma was to evaluate whether this fractionation would enable us to detect and quantify members of an AD brain protein module, referred to as M42, the matrisome (11). Our previous proteomic data had identified M42 as a highly conserved protein module in the AD brain and the CRND8 mouse model of Ab amyloid deposition (11, 19). Notably, selected M42 differential expressed proteins (DEPs) show some of the highest fold increases in the human AD and CRND8 brain proteome, including APP/Ab and APOE, as well as many proteins known to bind heparin (19). The module also includes multiple HSPGs. We have recently shown many members of this module bind amyloid and co-accumulate with Ab in plaques, cerebrovascular amyloid, and/or dystrophic neurites (19). As in the AD brain proteome, we find that many M42 members in Hp-enriched plasma proteome can also distinguish AD from controls. Thus, our Hp-enriched plasma data indicates that M42 proteins are not only intimately linked to AD amyloid pathology, but also appear to be robust AD plasma biomarkers. More unexpectedly, our current data suggests that many protein modules identified initially in the human AD brain proteome are robustly detectable in the AD Hp-enriched plasma proteome. Targeted analyses of AD plasma such as Ab42, Ab40, various pTau species, GFAP and NEFL have provided highly informative blood-based biomarkers that clearly reflect underlying pathological processes in the brain (59). However, broader integration of AD plasma proteome with the human AD brain proteome has not yielded many novel insights into AD pathobiology. As discussed in detail below, the evaluation of our AD Hp-enriched plasma proteome through the lens of our AD brain proteome, suggests that disruptions of multiple brain modules are detectable in the AD plasma proteome.
These findings have many implications relevant to our understanding of AD as a brain only versus a whole-body disorder. Currently it is unclear whether the changes observed in the AD heparin-enriched plasma are simply readouts of protein level changes in the AD brain reflected in the plasma or evidence that AD pathophysiology directly impacts peripheral organ function in a way that drives biomarkers changes which can be observed in plasma. Given the complex relationship between changes in the AD brain and Hp-enriched plasma proteome that is not always conserved in directionality within selected protein modules, previous evidence that plasma biomarkers may reflect changes in the brain proteome, and precedence for robust bidirectional crosstalk between the brain and the periphery, our own bias is that the plasma proteomic changes represent an admixture of central and peripheral processes.
For some brain protein modules, such as M42, we observed consistency between brain, CSF and plasma in terms of direction of change (11, 20, 21). For example, not only are SMOC1 and SPON1 increased in CSF upwards of 30 years in advance of AD (17), these proteins are also increased in the preclinical or asymptomatic stage of disease in brain suggesting each as high value biomarkers if detected in plasma (11). SPON1 was one of the most consistent M42 proteins measured across proteomic platforms as it was increased in AD by aptamer-based and Olink antibody-based measurements. In total, 16 matrisome proteins overlapping with M42 were differentially abundant in AD plasma following heparin enrichment, where SMOC1, SPON1, OLFML3, GPNMB, and HTRA1 are among the most significant ones elevated. Other newly identified members of M42 in plasma include SMOC2 and HGF, which have also been shown to be elevated in AD CSF (60). However, the biological basis for the consistency of changes in these compartments is not yet understood. Notably, the behavior of these proteins is distinct from Ab which does not consistently show an increase in AD CSF or plasma in either absolute level yet does show a decrease in the relative levels of Ab42 to Ab40. Future studies aiming to measure matrisome proteins in plasma from cohorts like Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Dominantly Inherited Alzheimer Network (DIAN) will be essential for assessing their prognostic values in predicting disease outcomes.
In our previous consensus analysis of the human brain, we did not incorporate the APOE-specific isoforms (APOE2 and APOE4) into the human protein database (11). Since these isoforms differ due to substitutions of cysteine residues with arginine residues, they can be easily distinguished in the human proteome after trypsin cleavage which releases novel peptides, which can then be accurately mapped and quantified using mass spectrometry (54). Notably, by integrating genomics and proteomics data from the same individuals, we previously demonstrated that the individuals carrying an APOE ε4 allele exhibited higher M42 levels in brain, and this regulation was not solely driven through the levels of the APOE protein itself (11). In this study, the inclusion of the APOE ε4 specific protein isoform rather than APOE in M42 serves to further strengthen the genetic association between APOE ε4 genotype and M42 levels. This also indicates that the APOE ε4 isoform may have a stronger predisposition for interaction with heparin and HBPs within M42 in the brain than other APOE isoforms (55). In the future, the implementation of integrated genomics and proteomics pipelines will be needed for assessing whether M42 proteins in plasma are under genetic regulation by APOE4. This will provide valuable insights into whether this relationship is consistent across both the central nervous system and the peripheral system.
Heparin and HS accelerate the formation of Aβ fibrils (26–28) and loss of this heparin-APOE binding interaction has been suggested as a possible mechanism for the protection of the APOE Christchurch loss-of-function mutation recently described in a PSEN1 ADAD mutation carrier (55). However, more recently, a rare RELN variant has been proposed to delay the age of onset of siblings with ADAD who do not carry the Christchurch APOE variant (61). Like APOE, Reelin (RELN) is also an HBP, but in contrast to the APOE Christchurch variant, the RELN variant is associated with heightened interactions with heparin and a consequent reduction in tau phosphorylation via the Dab1 signaling pathway (61). It is worth noting that in our study, we observed increased levels of RELN in the Hp-enriched fraction compared to input and FT (Supplemental Fig. 1). Hence, there appears to be an intricate relationship between heparin interactions, APOE, and AD risk. The exact mechanisms by which HBPs, including APOE and members of M42, influence amyloid deposition and potential clearance still requires further investigation.
Beyond their interactions with APOE and M42 members, HSPGs alone have been implicated in AD progression and pathogenesis (62). Specifically, HSPGs are believed to play a crucial role in mediating the internalization and propagation of specific proteopathic seeds of tau (63). Namely, the HSPG glypican-4 (GPC4) has been identified as a contributor to APOE4-induced tau hyperphosphorylation (64). It is also noteworthy that 6-O sulfation on HSPGs is presumed to regulate the cell-to-cell propagation of tau (65). Interestingly, glypican-5 (GPC5), which is structurally homologous to glypican-4, is a core member of M42 (Fig. 6A). While it is yet to be established whether glypican-5 plays a role in the regulation of tau internalization, the co-expression between APOE4 and GPC5 in the brain suggests the possibility of such involvement. Evidence supporting the role of HS in the etiology of AD is also emerging in genome-wide association studies (GWAS). For example, GWAS meta-analysis identified the heparan sulfate-glucosamine 3-sulfotransferase 1 gene (HS3ST1) as a risk locus associated with late onset AD (66). Furthermore, a recent study reported a seven-fold increase in total brain HS in AD compared to controls and other tauopathies (67). These findings collectively suggest that dysfunction in HS and HBPs in brain, CSF and plasma may play a central role in the etiology and the clinicopathological presentation of AD.
We also identified a significant number of novel Hp-enriched plasma proteins that exhibited increased levels in AD. Among these proteins were the proteoglycan biglycan (BGN), which is typically found in the extracellular matrix of blood vessels (68), and Endothelial Cell–Specific Molecule 1 (ESM1), also known as endocan. Both proteins play roles in regulating endothelial cell function and angiogenesis, and they have been implicated in processes related to inflammation and vascular disease (68, 69). Furthermore, we observed an elevation in Colony-Stimulating Factor 1 (CSF1) in the plasma of individuals with AD. CSF1 primarily functions in the regulation of the immune system (70). Elevated levels of CSF1 in plasma have been associated with various diseases and conditions, including inflammation, cancer, and certain autoimmune disorders (71, 72). Taken together, these findings suggest that widespread systemic inflammation may also manifest in the plasma of individuals with AD. However, whether this phenomenon is specific to AD or extends to other subtypes of dementia remains a subject that requires further investigation.
In this study we also leveraged the consensus brain protein co-expression network to explore the relationship between the Hp-enriched plasma and brain proteomes. Within the network modules, certain plasma proteins consistently exhibited increases or decreases in AD that mirrored changes in the brain, while others displayed divergent alterations as previously described (30). Among those consistently increased, beyond M42 members, were proteins mapping to M26 in the brain, associated with the ‘Acute phase response’, which suggests that proteins related to complement activation potentially associated with immune function are enriched in AD plasma. Notably, proteins of interest in M26 included SERPINA3, which was recently identified through a large-scale analysis of the plasma proteome using Mendelian randomization as potentially causal in AD pathogenesis (73). Additionally, we observed proteins in plasma that map to neuronal modules related to synaptic biology in the brain which displayed consistent decreases in AD. Whether this change in plasma reflects synapse loss in the brain will require further investigation. Nevertheless, it is intriguing that a signature of synaptic loss typically associated with cognitive decline in brain and CSF appears in the Hp-enriched plasma in AD. There was also some discordance in the direction of change between AD plasma and brain proteomes. For instance, proteins specifically in M7 ‘MAPK signaling’ and M25 ‘Sugar metabolism’, exhibited increased levels in the brain, yet decreased levels in the Hp-enriched plasma. This trend contrasts with the direction of change observed in CSF (30), where glycolytic signature is increased in AD even in the preclinical phase (14, 30, 74). The precise mechanisms underlying this discordance remain unclear. However, it is plausible that blood-brain barrier dysfunction might contribute to these differences (75), where proteins increased in plasma are decreased in CSF, and vice versa.
Although our goal was to capture HBPs in plasma, a consequence of the heparin affinity enrichment was the clearance of highly abundant proteins such as albumin. This reduction in albumin from the Hp-enriched fraction resulted in the comprehensive coverage of nearly 3,000 plasma proteins following high-pH off-line fractionation and TMT-MS. Thus, in contrast to immune-depletion methods (76), antibody-free affinity enrichment-based approaches utilizing nanoparticle, cationic/anionic or hydrophobic/hydrophilic-based strategies appear to substantially enhance plasma proteome coverage through MS-based technologies (77). Collectively, this progress marks a step toward overcoming one of the major limitations of plasma MS-abased proteomics, which is the vast dynamic range of protein abundances. Furthermore, the utilization of more advanced mass spectrometers, such as the Orbitrap Astral, which quantifies five times more peptides per unit time than state-of-the-art Orbitrap mass spectrometers (78), is expected to significantly enhance the depth of proteome coverage in plasma when employing these affinity enrichment strategies. This is of particular significance due to the complementary coverage of the Hp-enriched plasma proteome by TMT-MS in contrast to the SomaScan and Olink platforms, which will further enhance the depth of the plasma proteome when measurements are integrated across platforms (30).
While this study provided a comprehensive proteomic analysis of Hp-enriched plasma from human subjects, several limitations should be acknowledged. Notably, the study participants predominantly consisted of non-Latino white individuals. Recent reports have highlighted disparities in AD prevalence, with Black and Hispanic populations showing a higher likelihood of developing AD compared to older white Americans (79–81). Additionally, it has been observed that cognitively impaired African American individuals have lower levels of CSF tTau and pTau compared to Caucasians (38). An important ongoing initiative of the Accelerating Medicines partnership for AD (AMP-AD) (82) is the inclusion of African American and Hispanic individuals in plasma biomarker studies. Research efforts employing heparin enrichment techniques should aim to encompass a more diverse participant population to better capture the complexities of AD across different racial and ethnic groups. Future studies that investigate the interplay between age, sex, and race within the Hp-enriched plasma proteome will yield valuable insights. Moreover, our plasma proteomics study exclusively focused on control and symptomatic individuals who were AD biomarker positive based on their tau/amyloid ratio in the CSF. Future plasma proteomic studies aimed at exploring the pre-symptomatic stages of AD before cognitive impairment manifests will be needed to identify Hp-enriched plasma proteins that undergo early changes in the disease course. Nevertheless, this study offers a global view into the Hp-enriched plasma proteome, reinforcing a hypothesis that increased matrisome proteins are shared between the brain and blood in AD.