MSR1 and NEP Are Correlated with Alzheimer’s Disease Amyloid Pathology and Apolipoprotein Alterations

INTRODUCTION

In addition to extracellular amyloid-β (Aβ) plaques [1], neuronal and synaptic loss [2, 3], intra-neuronal neurofibrillary tangles [4], and cortical thinning in the temporal, orbitofrontal, and parietal regions [5], Alzheimer’s disease (AD) is characterized by chronic inflammation [6] as well as an altered cholesterol metabolism [7, 8]. According to multiple studies, scavenger receptors are directly implicated in the neuropathological features that characterize AD. In addition to mediating the uptake and the degradation of acetylated and oxidized low-density lipoproteins (LDLs), which both lead to massive cholesterol deposition in the neuropil [9], these pattern recognition receptors are highly expressed by amyloid deposit-associated microglia [10]. Moreover, the scavenger receptors have been implicated in the binding and uptake of oligomeric and fibrillar Aβ by microglia [11 –13]. In opposition to the familial form of AD, which is caused by the overproduction of Aβ₄₂ [14], the sporadic form of AD, which is more common, is mostly characterized by a reduced Aβ₄₂ clearance [15], a process that is mediated in part by the microglial uptake of this peptide.

Macrophage scavenger receptor 1 (MSR1) is one of the main receptors implicated in Aβ uptake by immune cells. In humans, MSR1 is predominantly found on macrophages [16] and monocytes in the periphery [11], and microglia in the CNS [13]. A diverse array of ligands is recognized by MSR1: Aβ [17], heat shock proteins [18], modified (i.e., acetylated or oxidized) LDL [9], as well as surface molecules from hepatitis C virus [19] and bacteria [20]. The majority of endogenous MSR1 ligands are connected to age-related degenerative diseases; oxidized lipoproteins being the driving force behind atherosclerosis, advanced-glycosylated end products resulting from diabetic glucose overload, and Aβ fibrils representing the central component of amyloid plaques in AD [21]. This receptor has also been shown to modulate the TLR4-mediated inflammation, which has been well documented in AD [17 , 22–28].

In recent years, multiple human studies have identified a correlation between MSR1 and AD. Indeed, a study using the Lothian Birth Cohort 1936 have shown that a significant proportion (36%) of the negative correlation between plasma MSR1 levels and cognitive ability was mediated by the total brain volume [29]. Moreover, a GWAS performed on nearly 4,000 European AD cases identified a variant in the 3’-UTR region that displays a suggestively significant correlation with the rate of cognitive decline [30]. In autopsied cases, elevated microglial MSR1 expression was correlated with a higher risk of having a diagnosis of AD and was correlated, in AD subjects, with reduced pre-mortem cognitive performances as well as higher densities of neuritic/diffuse plaques and neurofibrillary tangles [31]. Single-cell RNA sequencing also revealed that MSR1, along with APOE, are highly expressed in microglia from AD subjects (Single-cell atlas of the Entorhinal Cortex in Human Alzheimer’s Disease; ddnetbio.com) [32].

MSR1 has been implicated in the pathological processes of multiple mouse models of amyloidosis, tauopathy, neuroinflammation, and cerebral ischemia. RgTg4510 tau transgenic mice, a model of tauopathy [33], showed increased microglial Msr1 mRNA levels in contrast to wild-type mice [34]. Eight and fourteen-month-old APP/PS1 mice showed elevated microglial Msr1 mRNA levels compared to age-matched wild-type littermates [35]. Furthermore, APP/PS1 mice deficient for Msr1 (Msr1 –/–) displayed increased mortality rates over time, enhanced Aβ deposition, and a decreased gene expression for neprilysin (NEP) and insulin [17], two key Aβ-degrading enzymes [36, 37]. Moreover, glatiramer acetate, which is known to reduce spatial learning or memory deficits and to decrease the number of hippocampal Aβ plaques in APP/PS1 mice [38], increased MSR1 immunoreactivity one hour and 24 h after treatment on cultured monocyte-derived macrophages extracted from WT mice [39].

NEP is an integral membrane zinc metallopeptidase that is produced by mononuclear phagocytes [35], whose active site is present in the extracellular space [40]. NEP-deficient mice showed a significantly reduced but not completely abolished Aβ degradation, suggesting that, although being the most potent Aβ-degrading enzyme [36, 41], NEP degrades Aβ₄₂ along with other enzymes [42]. Intronic variants in the NEP gene were also correlated with sporadic AD [43]. A recent meta-analysis showed that AD subjects displayed lower levels of NEP gene expression in the frontal cortex [44]. Cerebrospinal fluid (CSF) NEP levels were also reduced in mild cognitive impairment (MCI) and AD subjects and were correlated, in these patients, with lower Mini-Mental State Examination scores as well as higher CSF total tau (t-tau) levels [45].

The purpose of the present study is to examine the relationship between MSR1, NEP, and neuropathological traits of AD in both the pre-symptomatic and symptomatic phases of the disease. Therefore, we systematically examined the relationships between plasma/CSF MSR1 and NEP levels versus disease progression and CSF Aβ_1-42, ApoE, and ApoJ concentrations in a living cohort of asymptomatic individuals with a parental history of AD (the Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease [PREVENT-AD]) [46] as well as in MCI/AD subjects from the Canadian Consortium on Neurodegeneration in Aging (CCNA) [47]. Data from autopsied brain samples from the Religious Orders Study and Memory and Aging Project (ROSMAP) [48] were used to examine cortical MSR1 gene expression in relation to APOE and CLU mRNA levels as well as CERAD scores [49].

METHODS

Statistical analysis

All analyses were performed in SPSS. In analyses combining PREVENT-AD and CCNA samples, Pearson correlations adjusted for age and sex (for Figs. 2B, 3, 4B-D, 5C-D, and 6C-D) were used. Analyses performed only on PREVENT-AD subjects (Figs. 1, 2A, 4A-C, 5A-B, and 6A-B, as well as Supplementary Figure 1) used Pearson correlations adjusted for age, sex, and APOE ɛ4 status. In ROSMAP, because mRNA levels were not distributed normally, analyzes were performed using Spearman correlations adjusted for age groups (65–79/80–89/90+), sex, postmortem interval, and APOE ɛ4 status (Figs. 7 and 8). In ROSMAP, MSR1, CLU and APOE mRNA levels that were 3 interquartile ranges above the median were removed.

OPEN IN VIEWER

Fig. 1

Correlation between plasma and CSF MSR1 A) and NEP B) levels in asymptomatic subjects from PREVENT-AD. Each dot represents an individual subject. CSF, cerebrospinal fluid; MSR1, macrophage scavenger receptor 1; NEP, neprilysin; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease.

OPEN IN VIEWER

Fig. 2

Correlation between plasma (A) or CSF (B) NEP and MSR1 levels in asymptomatic individuals from PREVENT-AD and MCI/AD subjects from CCNA. Each dot represents an individual subject. AD, Alzheimer’s disease; CCNA, Canadian Consortium on Neurodegeneration in Aging; CSF, cerebrospinal fluid; MCI, mild cognitive impairment; MSR1, macrophage scavenger receptor 1; NEP, neprilysin; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease.

OPEN IN VIEWER

Fig. 3

CSF MSR1 A) and NEP B) levels in PREVENT-AD and CCNA according to diagnosis. Bars represent mean NPX values±SEM. AD, Alzheimer’s disease; CCNA, Canadian Consortium on Neurodegeneration in Aging; CSF, cerebrospinal fluid; MCI, mild cognitive impairment; MSR1, macrophage scavenger receptor 1; NEP, neprilysin; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease; SEM, standard error of the mean.

OPEN IN VIEWER

Fig. 4

Relationship between plasma (A, C) or CSF (B, D) MSR1 (A-B) or NEP (C-D) and CSF Aβ_1-42 levels in asymptomatic individuals from PREVENT-AD and MCI/AD subjects from CCNA. Each dot represents an individual subject. Aβ_1-42, amyloid-beta 1-42; AD, Alzheimer’s disease; CSF, cerebrospinal fluid; CCNA, Canadian Consortium on Neurodegeneration in Aging; MCI, mild cognitive impairment; MSR1, macrophage scavenger receptor 1; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease.

OPEN IN VIEWER

Fig. 5

Correlation between plasma (A-B) or CSF (C-D) MSR1 levels and CSF ApoE (A,C) or ApoJ (B, D) levels in asymptomatic subjects from PREVENT-AD. Each dot represents an individual subject. ApoE, apolipoprotein E; ApoJ, apolipoprotein J; CSF, cerebrospinal fluid; MSR1, macrophage scavenger receptor 1; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease.

OPEN IN VIEWER

Fig. 6

Correlation between plasma (A-B) or CSF (C-D) NEP levels and CSF ApoE (A,C) or ApoJ (B, D) levels in asymptomatic subjects from PREVENT-AD. Each dot represents an individual subject. ApoE, apolipoprotein E; ApoJ, apolipoprotein J; CSF, cerebrospinal fluid; NEP, neprilysin; NPX, normalized protein expression; PREVENT-AD, Presymptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease.

OPEN IN VIEWER

Fig. 7

Correlation between brain MSR1 mRNA levels and CERAD scores in ROSMAP. Bars represent mean MSR1 FPKM values±SEM. AD, Alzheimer’s disease; CERAD, Consortium to Establish a Registry for Alzheimer’s Disease; FPKM, Fragments Per Kilobase of transcript per Million mapped reads; mRNA, messenger ribonucleic acid; MSR1, macrophage scavenger receptor 1; ROSMAP, Religious Orders Study and Memory and Aging Project; SEM, standard error of the mean.

OPEN IN VIEWER

Fig. 8

Relationship between brain expression of MSR1 and APOE/CLU in ROSMAP. Each dot represents an individual subject. AD, Alzheimer’s disease; APOE, apolipoprotein E; CTL, control, CLU, clusterin (APOJ); FPKM, Fragments Per Kilobase of transcript per Million mapped reads; MCI, mild cognitive impairment; mRNA, messenger ribonucleic acid; MSR1, macrophage scavenger receptor 1; ROSMAP, Religious Orders Study and Memory and Aging Project.

RESULTS

DISCUSSION

Plasma and CSF MSR1 levels are strongly correlated in asymptomatic subjects from PREVENT-AD (Fig. 1). This observation suggests that a large proportion of MSR1 originates from peripheral macrophages [16] and monocytes [11] that infiltrated the brain. Since CSF/plasma ratios of albumin and IgG, which are indicators of blood-brain barrier permeability [56], and CSF MSR1 levels are not correlated in this asymptomatic cohort (Supplementary Figure 1A, B), this infiltration likely occurs via active transport in the brain and not through a leakage in the blood-brain barrier [57]. Once in the brain, cell-surface MSR1, like for proteins such as TREM2, is released from the microglial plasma membrane upon the action of specific sheddases such as ADAM10 or ADAM17 [58] and can therefore be detected in the CSF. On the other hand, since 1) CSF NEP levels do not correlate with CSF/plasma ratios of albumin and IgG (Supplementary Figure 1C, D) and 2) there is nearly no NEP gene expression in the CNS (data not shown), this suggests that most of the NEP that is detected in the CSF originates from the active infiltration of NEP-expressing peripheral monocytes/macrophages in the CNS. However, the lack of correlation between plasma and CSF NEP levels also suggests that some signals unique to the brain environment might influence CNS NEP protein levels.

As expected from Frenkel’s original observations in APP mice [17], CSF levels of MSR1 and NEP are highly correlated in pre-symptomatic subjects, sharing almost half of their variance (Fig. 2B). Plasma MSR1 and NEP levels are not correlated (Fig. 2A), which is consistent with a CNS-specific process. CSF MSR1 and NEP levels are also inversely correlated with clinical disease progression, with lower levels found in MCI and AD subjects (Fig. 3). Results for NEP are consistent with the literature [45], while those for MSR1 are consistent with our hypothesis that CSF levels of the latter protein reflect the proportion that is released from the surrounding environment of amyloid plaques. Therefore, a reduction in the CSF concentrations of MSR1 could be viewed as an informative marker of higher levels of this protein in amyloid plaque-associated brain tissues.

Relationships between levels of MSR1 and CSF Aβ_1-42 are also quite different depending on the origin of MSR1. In PREVENT-AD, while CSF MSR1 concentrations are positively correlated with CSF levels of Aβ_1-42, plasma MSR1 levels are not (Fig. 4A, B), consistent again with a CNS-specific process with little or no contribution from the periphery. As the disease progresses toward a higher density of neocortical amyloid plaques (i.e., a definite diagnosis of AD according to the CERAD classification), higher MSR1 gene expression is observed in cortical areas (Fig. 7), presumably as a compensatory response to control for the increased amyloid deposition in brain tissues [11]. Although there have been some discrepancies in the ability of peripheral NEP to reduce brain Aβ [59, 60], our results tend to show that plasma NEP levels have no impact on CSF Aβ_1-42 levels (Fig. 4C). In the CSF, NEP levels correlate positively with Aβ_1-42 levels (Fig. 4D), which implies that lower CSF NEP levels are correlated with higher brain amyloid deposition. This observation is quite consistent with a recent study performed by Grimmer et al. on AD patients [61].

In the early 1990s, MSR1 was identified as one of the receptors that mediate the uptake of both acetyl-LDL and oxidized LDL by foam cells (i.e., cholesterol-loaded macrophages) [62 –64]. Ligand detection for MSR1 using purified lipoprotein complexes identified ApoAI, ApoAII, and ApoE as selective ligands that compete with known MSR1 ligands, such as acetylated or oxidized LDL, for their binding to the receptor [65]. Furthermore, MSR1-mediated adhesion of apolipoproteins to macrophages/microglia may represent a mechanism by which MSR1 contributes to the trapping of these cells at sites of pathological apolipoprotein accumulation or deposition, such as amyloid plaques and tangles [65].

These findings prompted us to investigate the possible interactions between MSR1 and the two major HDL-associated apolipoproteins found in the CSF, which are also widely believed to be implicated in AD [66]: ApoE and ApoJ. MSR1 levels are significantly correlated with CSF concentrations of ApoE and ApoJ, which is again consistent with a CNS-specific process (Fig. 5). In the injured or diseased brain, ApoE and ApoJ are known to work in tandem to coordinate the lipid transport and mobilization associated with synaptic remodeling [67 –69]. The involvement of MSR1 in the early phase of synaptic damage is certainly consistent with its known function during apoptosis after neuronal injury [70]. In cognitively unaffected individuals from ROSMAP, brain MSR1 mRNA levels correlate with the cortical expression of CLU (i.e., the gene encoding ApoJ), but not APOE (Fig. 8). In the CSF of cognitively unaffected individuals, NEP concentrations are however correlated with both ApoE and ApoJ levels (Fig. 6C, D). It is conceivable that it is the APOE genotype, which significantly modulates ApoE protein levels independently from its gene expression in the CNS, that serves as the driving force behind the correlation between CSF MSR1 and ApoE concentrations, as seen in Fig. 5C [66 , 71–73] These observations are consistent with evidence that suggests that lipidated ApoE might be implicated in the cleavage of Aβ by neprilysin within microglia [74]. Plasma NEP levels are inversely correlated with CSF ApoE levels (Fig. 6A), which might be explained by the implication of CNS ApoE in the reallocation of peripheral NEP-expressing mononuclear phagocytes to the CNS. Plasma NEP levels are however not correlated with CSF ApoJ levels (Fig. 6B).

Although we believe it might unravel a portion of the mechanisms by which MSR1 and NEP collaborate to clear amyloid plaques, our study contains a few limitations. First, CCNA’s COMPASS-ND cohort lacked a few important data, such as APOE ɛ4 genotypes and CSF/plasma ratios of albumin and IgG, which would have allowed us to perform comparisons on the whole AD spectrum. Although CSF collection and measurements were performed in the same way between the PREVENT-AD and the COMPASS-ND studies, important differences, such as those pertaining to the eligibility criteria, might have also introduced biases when comparing data from these cohorts. Furthermore, CSF samples were centrifuged, which would remove all the microglia cell surface content from our analyses. Therefore, we can only assume that the soluble portion of the proteins that we measured constitutes a proxy of the microglial-bound content. Another potential confounding issue in this article is the fact that the discrepancies observed between PREVENT-AD/CCNA and ROSMAP can be explained by factors other than differences related to mRNA versus protein levels, such as age, sample sizes, and soluble versus membrane-bound protein regulations. Finally, while we have shown multiple significant correlations throughout this article, these do not necessarily imply direct causality, and future research is needed to better understand the mechanisms in place.

Overall, these results suggest that, in addition to be highly correlated to one another, CSF MSR1 and NEP levels are both correlated with the disease status, as well as CSF Aβ_1-42, ApoE, and ApoJ concentrations. Additionally, both CSF MSR1 and NEP levels are not correlated with CSF/plasma ratios of albumin and IgG. However, while plasma NEP concentrations are correlated with CSF ApoE levels, plasma MSR1 concentrations are not. Moreover, although CSF MSR1 levels are correlated with plasma MSR1 concentrations, CSF NEP levels are not correlated with plasma NEP concentrations. Furthermore, while brain MSR1 mRNA levels are correlated with the CERAD scores and CLU mRNA levels, parenchymal NEP expression is nearly inexistant. The multiple discrepancies described in this paragraph could be partly explained by the differential gene/protein expression of NEP and MSR1 by peripheral monocytes/macrophages before and after their infiltration in the CNS. Additional research is needed to investigate the possible use of MSR1 and NEP as potential plasma or CSF biomarkers of the response to Alzheimer’s disease pathology.