Identification and Preliminary Validation of a Plasma Profile Associated with Cognitive Decline in Dementia and At-Risk Individuals: A Retrospective Cohort Analysis

Abstract

Biomarker discovery is a major need for earlier dementia diagnosis. We evaluated a plasma signature of amyloid, metallo-proteinases (MMPs), and inflammatory markers in a cohort of at-risk individuals and individuals clinically diagnosed with probable Alzheimer’s disease (pAD). Using multiplex arrays, we measured Aβ₄₀, Aβ₄₂, MMP-1, MMP-3, MMP-9, IFN-γ, TNF-α, IL-6, IL-8, and IL-10 in plasma from 107 individuals followed every 6 months for 3 years. Final diagnoses included: pAD (n = 28), mild cognitive impairment (MCI, n = 30), subjective memory impairment (SMI, n = 30), and asymptomatic (NCI, n = 19). Blood was drawn at final follow-up. We used linear and logistic regressions to examine biomarker associations with prior known decline on the Montreal Cognitive Assessment (MoCA) and the Cambridge Cognitive Examination (CAMCOG); as well disease progression by the time of blood-draw. We derived a biomarker composite from the individual markers, and tested its association with a clinical diagnosis of pAD. Lower Aβ₄₀ and Aβ₄₂ and higher IL-8, IL-10, and TNF-α were associated with greater cognitive decline per the MoCA and CAMCOG. MMP-3 was higher in SMI, MCI, and pAD than NCI. Whereas the other investigative molecules did not differ between groups, composite scores—created using MoCA/CAMCOG-based trends in Aβ₄₀, Aβ₄₂, MMP-1, MMP-3, IL-8, IL-10, and TNF-α— were associated with a final diagnosis of pAD (c-statistic 0.732 versus 0.602 for age-sex alone). Thus, plasma amyloid, MMP, and inflammatory biomarkers demonstrated differences in individuals with cognitive deterioration and/or progression to MCI/pAD. Our findings support studying these markers earlier in the continuum of probable AD as well as in specific dementias.

Keywords

Alzheimer’s disease dementia amyloid-β biomarker blood cognitive decline inflammation metallo-proteinases MMP-3 MMP-9 plasma

INTRODUCTION

Early identification of individuals at risk of dementia is crucial to maximize a positive therapeutic outcome. A biomarker signature of incipient dementia could help identify asymptomatic individuals at-risk and permit earlier treatment once disease-modifying therapies become available. For population-wide screening, blood is a preferred source of biomarkers over cerebrospinal fluid (CSF) and various imaging modalities, which are more expensive and/or invasive approaches [1].

Alzheimer’s disease (AD), the most common form of dementia, is a multifactorial brain disorder characterized by amyloid-β (Aβ) and tau accumulation [2 –5], synaptic and neuronal loss [6, 7], neuroinflammation [8 –10], and degeneration of basal forebrain cholinergic neurons [11 –13]. A biomarker signature reflecting these multiple underlying pathologies may be more diagnostically robust than single biomarker measures focusing solely on tau or amyloid. For example, we recently demonstrated that a combined measure of amyloid (Aβ₄₀ and Aβ₄₂) and inflammatory molecules (IL-8 and TNF-α), and their change over time, is a stronger predictor of prospective cognitive deterioration in an AD-asymptomatic Down syndrome population than the individual markers [14]. We also found that the pre-dementia stage in Down syndrome is characterized by significant changes in several plasma analytes, including upregulation of metallo-proteinases (MMP-1, MMP-3, and MMP-9), which participate in inflammation, vascular remodelling, and nerve growth factor (NGF) metabolism [15, 16], as well as by increases in pro-inflammatory cytokines (TNF-α, IL-6 and IL-10) [14].

To evaluate whether this panel of investigative molecules could also have diagnostic utility in the non-Down syndrome context, we studied these markers in the plasma of individuals who received a clinical diagnosis of dementia deriving from probable AD (pAD), mild cognitive impairment (MCI), and subjective memory impairment (SMI) enrolled in the Longitudinal Early Alzheimer’s Disease (LEAD) cohort study. In order to merit further study as plausible biomarkers for AD, we would expect these molecules to demonstrate differences in patients already diagnosed with pAD or MCI, or in patients who already demonstrated to have cognitive deterioration.

Therefore, in this preliminary study, we examined plasma markers of Aβ, metallo-proteinases and inflammatory pathways in participants with established clinical diagnoses of probable AD-dementia, MCI, SMI, and healthy controls as of their final follow-up. We then examined associations between these plasma markers and the participants’ established degree of cognitive deterioration, based on change in cognitive scores or clinical progression at last follow-up. Based on the markers’ associations with cognitive performance, we further derived a pragmatic composite score for validation in further prospective and longitudinal studies, and tested its association with final clinical diagnosis of probable AD.

MATERIALS AND METHODS

Study population

Participants included 107 individuals with and without clinically diagnosed dementia who had been recruited as part of the LEAD study. This was a prospective cohort study in Oxfordshire, UK, consisting of 213 patients followed every 6 months for three years between 2009 and 2015. For this pilot biomarker study, we selected all 28 patients from the LEAD cohort with a final clinical diagnosis of pAD (some of whom had previously progressed through the SMI and MCI categories), and a random sample drawn from the remaining participants and weighted towards representation of symptomatic cases: 30 patients with clinically diagnosed MCI, 30 patients with SMI, and 19 cognitively normal, asymptomatic subjects (NCI: non-cognitively impaired) who did not develop dementia at last follow-up. Participants with other identified causes of cognitive impairment were not included in this study. Study procedures were approved by the South West-Frenchay Research Ethics Committee (National Research Ethics Service Committee, UK). The Institutional Review Board of McGill University, Canada approved the analysis of human plasma samples. Informed consent was provided by all participants.

Clinical assessment and neuropsychological evaluations

All LEAD participants provided demographic information and medical history, including family history of dementia (in first- or second-degree relatives), and underwent a detailed physical and neuropsychological examination, CT/MRI scan to rule out other causes of dementia (e.g., hydrocephalus, vascular dementia) and to identify supportive patterns of atrophy (e.g., hippocampal atrophy), and collection of blood specimens. Participants underwent follow-up assessments up to every six months. The final diagnosis for each patient was determined by a clinical team after the last follow-up assessment and cognitive testing, but prior to any biomarker testing (Fig. 1A).

The cognitive battery included the Montreal Cognitive Assessment (MoCA) [17] and the Cambridge Cognitive Examination (CAMCOG) [18]. Probable Alzheimer’s disease was diagnosed according to the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) criteria [19] and MCI according to the criteria described by Petersen and colleagues [20]. Participants with no symptoms of MCI or probable AD were classified as asymptomatic (NCI), whereas those who reported memory complaints but did not meet MCI criteria were classified as SMI. Further details on the clinical work up are described in the Supplementary Material.

Fig.1

Timeline and progression of participants during follow-up in the LEAD cohort study. A) Experimental timeline of the LEAD cohort, highlighting key time-points for cognitive assessments with the MoCA and CAMCOG tests, as well as time-point for extraction of blood samples, biomarker and data analyses. B) Evolution of participant clinical diagnoses from baseline to final assessments in the cohort. Boxes indicate number of participants and clinical group (asymptomatic, SMI, subjective memory impairment; MCI, mild cognitive impairment; pAD, probable Alzheimer’s disease).

Plasma collection

Blood collection was done at the final follow-up visit (see timeline, Fig. 1A). Fasting venous blood was collected in the morning in Becton Dickinson 4 ml K2EDTA vacutainers (BD, #367839) and immediately centrifuged in Eppendorf 5702R tubes (3000 g, 10 min, 4°C) for plasma separation. Plasma samples were divided into 500μl aliquots into Nunc 1 ml polypropylene cryotubes (Thermo Scientific, USA) and stored at – 80°C until shipment and analysis. Samples were shipped on dry ice to McGill University and immediately stored at – 80°C upon arrival. The maximum number of freeze-thaw cycles was three and amyloid peptides were measured first. Samples were used within 6 months after collection.

Analysis of amyloid-β peptides, inflammatory markers, and metallo-proteinases

All multi-analyte kits were from Meso Scale (Meso Scale Diagnostics, MD, USA). The levels of Aβ₄₀ and Aβ₄₂ were measured with a Multi-Spot V-PLEX Aβ Peptide Panel (6E10). Assessment of MMP-1, MMP-3 and MMP-9 levels was done using a Multi-Spot MMP 3-Plex Ultra-Sensitive kit. A custom-built Multi-Spot V-PLEX Pro-inflammatory Panel was used for the quantitative assessment of IFN-γ, TNF-α, IL-6, IL-8, and IL-10. A detailed description of the assays is provided in the Supplementary Material.

Statistical analysis

Analyses were done with STATA 13.1 (StataCorp LLC, TX, USA) and GraphPad Prism v7.00a (GraphPad Software, Inc., La Jolla, CA, USA). Linear regressions and independent-sample t-tests, respectively, were used to assess whether the plasma markers varied by age or sex. Significance of differences between plasma marker levels across clinical groups was tested with the Kruskal-Wallis test and the Dunn’s post-hoc correction.

Age/sex-adjusted linear regressions were used to relate each marker (as a continuous/categorical variable) to the known change in the total MoCA score for each patient between enrolment and the final follow-up assessment when blood was drawn (at 3 years or last visit before death). Changes in the Delayed Recall and Orientation subsections were specifically examined, as these are considered to be most sensitive to assess cognitive deterioration in AD but also in older patients in general [21, 22]. The observed associations were then validated with the change in the CAMCOG score, including its Learning Memory and Orientation subsections. We examined the MoCA first as it is commonly used in clinical practice and is one of the assessments tools recommended by the UK Alzheimer’s Society [23].

Age/sex-adjusted logistic regression was used to examine the association between each marker and the odds of having dropped by ≥5 points on each cognitive test over the course of follow-up, representing a clinically meaningful change [24]. To further validate the observed associations in a more phenotypically homogeneous sample, these analyses were repeated after restricting the sample to those patients who received a final clinical diagnosis of MCI or pAD, thereby reducing the potential confounding factor of non-AD-related pathologies in the SMI sample.

The markers were included in each linear and logistic regression model as continuous variables, and the analyses were repeated coding them as categorical variables (divided into quartiles), in order to elucidate relevant non-linear associations. Correlations were also examined in men and women separately; in this case markers were divided into tertiles. When the markers were used as categorical variables, the Kruskal Wallis test was used to assess the significance of observed differences among the quartiles.

Based on the associations observed between the different markers and the change in cognitive scores in the cohort, a composite was derived for additional analyses. This was done firstly to confirm the associations seen with the individual markers and cognitive decline and secondly to examine new associations with a final clinical diagnosis of pAD, resulting from the combination of plasma markers.

In deriving the composite score, markers that did not show relevant correlations with the MoCA or upon validation testing in the CAMCOG were excluded, as well as those that showed markedly inconsistent correlations, or showed no relevant trends. Using the remainder of the markers, a pragmatic coding system was used to devise a composite score. For each patient, we assigned a score of –1 for each plasma marker concentration in a range that appeared to be relatively “high-risk”, i.e., greater decline in cognitive scores by the time of blood-draw. We assigned a score of +1 for each plasma marker titer in a range that appeared protective or “low-risk” on these analyses, i.e., lower decline in cognitive scores at blood-draw. We added up the score for each retained marker to obtain a composite score. We then examined the association of the composite score – coded both as a continuous variable and dichotomized as a score of +1 or higher, or –1 or lower – with a final clinical diagnosis of pAD, using age/sex-adjusted logistic regression. A Receiver Operating Characteristic (ROC) curve was generated to test the independent sensitivity and specificity of the age/sex-adjusted composite for a final clinical diagnosis of pAD, a ≥5 point decline on the MoCA, or a ≥5 point decline on the CAMCOG. We then compared the goodness-of-fit per the c-statistic of this model with that obtained from a logistic model including only age/sex, and from stepwise logistic regressions with forward/backward elimination that included all the tested markers as well as age/sex in the initial models [25].

RESULTS

Table 1 illustrates the demographics of the study participants. Individuals with a final clinical diagnosis of MCI were older than SMI cases and asymptomatic/NCI participants. There were fewer men in the asymptomatic group. There were no significant differences in education. MCI/pAD individuals were more likely to have a family history of dementia and lower final total scores on the MoCA and CAMCOG tests. The evolution of diagnosis from baseline to final assessment is shown in Fig. 1B. The mean duration of follow-up was 2.96 years (95% CI 2.81–3.12) for the overall cohort, and did not differ among the diagnostic groups (ANOVA p = 0.066). Eight participants died; all of them had completed at least one year of follow-up. One of them had a final diagnosis of SMI, four had MCI, and three had pAD.

Table 1

Characteristics of the study population

	Final clinical diagnosis
	Overall	Asymptomatic	SMI	MCI	pAD	p
	(n = 107)	(n = 19)	(n = 30)	(n = 30)	(n = 28)
Age, Mean (SD)	78.4 (7.2)	74.8 (6.8)	76.6 (7.0)	81.8 (6.7)	78.6 (7.2)	0.01^*
Sex – Male (%)	53 (49.5)	4 (21.1)	17 (56.7)	14 (46.7)	18 (64.3)	0.03^*
University or professional education (%)	39 (36.5)	8 (42.1)	9 (30.0)	11 (36.7)	11 (39.3)	0.83
Cardiovascular disease (%)	10 (9.6)	1 (5.3)	0 (0.0)	5 (17.9)	4 (14.3)	0.09
Family history of dementia (%)	19 (17.8)	2 (10.5)	1 (3.3)	8 (26.7)	8 (28.6)	0.03^*
Median MOCA (IQR) – final	26 (21–29)	28 (27–29)	28 (27–29)	24 (22–26)	15 (8–19)	0.0001^*
Median CAMCOG (IQR) – final	98 (91–102)	102 (100–104)	101 (99–103)	94 (91–96)	77 (65–88)	0.0001^*

SMI, subjective memory impairment; MCI, mild cognitive impairment; pAD, probable Alzheimer’s Disease; MOCA, Montreal Cognitive Assessment; CAMCOG, Cambridge Cognitive Examination; SD, standard deviation; IQR, interquartile range. Kruskal Wallis was used for comparison of numerical data (age, cognitive scores) and chi-squared testing for comparison of categorical data (sex, education, and family history).

Associations of plasma markers with age, sex, and clinical diagnosis

Plasma Aβ₄₀ levels rose significantly with age (coefficient [coef] 1.76 pg/ml/year, 95% CI 0.70–2.83, p = 0.001), as did TNF-α levels (coef 0.03, 0.01–0.04, p = 0.006). None of the other markers varied significantly by age.

The only marker with significant sex differences was MMP-3: levels were significantly higher in men than in women (mean 24,624 pg/ml versus 13,698 pg/ml, p < 0.001). Significant differences in plasma MMP-3 were also evident between clinical groups (Fig. 2, p = 0.02): participants with a final clinical diagnosis of pAD and SMI exhibited higher MMP-3 levels than asymptomatic/NCI individuals (Fig. 2, p = 0.03 and p = 0.046, respectively) with a similar trend in the MCI group (Fig. 2, p = 0.056).

Fig.2

Elevation of plasma MMP-3 in symptomatic participants in the LEAD cohort. MMP-3 concentrations in the overall sample, stratified by final clinical diagnosis showing significant elevations in participants with a final clinical diagnosis of pAD and SMI versus asymptomatic individuals, and a similar trend in the MCI group (p = 0.056). Quantitative analysis of plasma MMP-3 (pg/ml) was performed with a MesoScale MMP Ultrasensitive kit; Kruskal Wallis Test p = 0.020 H = 9.802 and Dunn’s post hoc correction; ^*p < 0.05, n = 17–30/group. Data is displayed in box and whiskers plot with the Tukey method.

Although the plasma concentration of the other markers assessed did not differ significantly between clinical groups (Supplementary Figure 1), participants in the pAD group showed a trend towards lower Aβ₄₀ levels compared to individuals classified as asymptomatic (mean difference pAD vs. NCI: –17.92 pg/ml, 95% CI –38.15 to 2.31 pg/ml, Supplementary Figure 1), and SMI and MCI cases showed a trend towards higher levels of IL-8 compared to NCI participants (mean difference 0.88 pg/ml, 95% CI –0.04 to 1.79; and 0.97 pg/ml, 95% CI –0.04 to 2.00, respectively, Supplementary Figure 1).

Associations between plasma markers, longitudinal cognitive decline, and clinical progression

When the plasma marker concentrations of each participant were plotted against longitudinal changes in MoCA scores over the course of follow-up, a number of correlations (generally non-linear) became apparent (Fig. 3). These associations were often accentuated when the sample population consisted of participants with a final clinical diagnosis of MCI/pAD (Fig. 4). To validate these observations, the analyses were replicated using the CAMCOG scores (Supplementary Figures 2 and 3). The main associations for each plasma marker are described below.

Fig.3

Plasma markers and cognitive decline. Relationship between biomarker plasma concentrations and change in MoCA scores from baseline to last follow-up for the overall sample. The graphs show best-fit fractional polynomial curves with 95% confidence intervals represented in grey. Circles represent individual participants. Analysis revealed associations between the concentration of amyloid peptides, MMP-3 (in women), IL-8, IL-10, and TNF-α with established cognitive decline. Weak correlations were observed between MMP-1 plasma levels and longitudinal cognitive change. No relevant trends were observed for MMP-3 in men or plasma MMP-9, IL-6, and IFN-γ.

Fig.4

Plasma markers and cognitive decline in AD-symptomatic participants. Relationship between biomarker plasma concentrations and change in MoCA scores from baseline to last follow-up for 58 patients who had a final clinical diagnosis of MCI and pAD. The graphs show best-fit fractional polynomial curves with 95% confidence intervals represented by the grey regions. Circles represent individual participants. Analysis revealed associations between the concentration of amyloid peptides, MMP-3 (in women), IL-8, IL-10, and TNF-α with established cognitive decline. Weak correlations were observed between MMP-1 plasma levels and longitudinal cognitive change. No relevant trends were observed for MMP-3 in men or plasma MMP-9, IL-6, and IFN-γ.

Aβ peptides

Participants with Aβ₄₀ levels in the second quartile (176.9–200.5 pg/ml) had suffered greater declines in cognitive scores over time, whereas those with higher concentrations were less likely to have declined (Fig. 4). For example, individuals in the MCI/pAD group with Aβ₄₀ levels in the second quartile had a greater drop in MoCA scores versus the rest (coef: –3.1, –6.2–0.1, p = 0.05), and were more likely to have suffered a decline in MoCA scores by ≥5 points over follow-up (aOR: 5.57, 1.07–28.9, p = 0.04). On the other hand, none of the patients with plasma Aβ₄₀ levels higher than 229.5 pg/ml (top quartile) had such a drop in MoCA scores.

With respect to Aβ₄₂, participants with concentrations in the second quartile (14.1–15.7 pg/ml) had greater drops in scores over time, whilst those with higher concentrations showed less cognitive decline (Figs. 3 and 4 and Supplementary Figures 2 and 3). In the MCI/pAD subgroup, participants with Aβ₄₂ levels in the second quartile had significantly greater drops in total MoCA scores (acoef: –4.9, –8.1 to –1.6, p = 0.004) and Delayed Recall sub-scores over follow-up, versus everyone else (acoef: –1.2, –2.3 to –0.1, p = 0.03). Participants with MCI/pAD who had Aβ₄₂ levels in this range were also more likely to have declined by ≥5 points in their MoCA scores (aOR: 29.3, 2.97–288, p = 0.004). Likewise, on the CAMCOG, MCI/pAD patients with Aβ₄₂ of 14.1–15.7 pg/ml (second quartile) were more likely to have declined by ≥5 points (aOR: 6.04, 1.13–32.4, p = 0.04), and had greater decline in their Orientation scores (acoef: –1.6, –2.9 to –0.4, p = 0.01). Asymptomatic and MCI participants with Aβ₄₂ concentrations in the second quartile were also significantly more likely to have progressed to SMI/MCI/pAD (aOR: 11.6, 2.1–65.0, p = 0.01). Thus, similar to Aβ₄₀, lower Aβ₄₂ titres were associated with greater cognitive decline per the MoCA and CAMCOG tests.

Metallo-proteinases

Lower MMP-3 plasma concentrations were associated with less cognitive decline in women but not in men (Figs. 3 and 4 and Supplementary Figure 3). For example, in the MCI/pAD subgroup, women with MMP-3 concentrations <11,341 pg/ml (first tertile) had significantly less decline in Delayed Recall scores on the MoCA versus those who had higher MMP-3 levels of 11,341–15,221 pg/ml (acoef: 1.6, 0.2–3.0, p = 0.03). On the CAMCOG’s Learning Memory section, women with MMP-3 levels <11,341 pg/ml (first tertile) had significantly smaller drops than those with MMP-3 >11,341 pg/ml (acoef: 1.9, 0.6–3.1, p = 0.004). In other words, the relation between plasma MMP-3 and cognitive decline exhibited a sex-specific effect. Weak correlations were observed between high MMP-1 plasma levels and longitudinal cognitive change (for Delayed Recall, acoef: –1.2, –2.4 to –0.04, p = 0.06). However, on the CAMCOG, MCI/pAD patients with higher MMP-1 levels had greater drops in Orientation scores (acoef: –2.6e-4, –5.1e-4 to –1.7e-5, p = 0.04). No relevant associations were observed between plasma MMP-9 and cognitive decline (Figs. 3 and 4 and Supplementary Figures 2 and 3).

Pro-inflammatory markers

Participants with IL-8 concentrations in the second quartile (2.81–3.41 pg/ml) had the least decline in MoCA Orientation scores (acoef: 0.4, 0.03–0.8, p = 0.03). For IL-10, individuals with higher plasma concentrations showed significantly greater drops in MoCA scores (acoef: –4.8, –9.5 to –0.1, p = 0.045). For TNF-α, individuals with concentrations in the second quartile (1.55–2.00 pg/ml) showed less decline in Delayed Recall scores than the rest (overall sample acoef: 0.7, 0.1–1.4, p = 0.04; MCI/pAD acoef: 1.2, 0.2–2.2, p = 0.02). Asymptomatic/MCI participants with TNF-α levels in the lower ranges (1.55–2.00 pg/ml) were also less likely to have progressed to MCI/pAD (aOR: 0.03, 0.001–0.61, p = 0.02). No relevant associations were observed between plasma IL-6, IFN-γ and change in cognitive scores (Figs. 3 and 4 and Supplementary Figures 2 and 3). In short, higher levels of the inflammatory markers IL-8, IL-10 and TNF-α correlated with established cognitive decline.

Derivation and testing of the composite biomarker score

Based on the most consistent associations from the retrospective analyses of MoCA and CAMCOG scores with the plasma markers, a biomarker composite was devised by assigning possible scores ranging from –6 to +7, using quartile-based definitions for cut-offs, except for MMP-3 where we used sex-specific tertiles to accommodate the observed sex differences (Table 2). We excluded MMP-9, IFN-γ, and IL-6 from the composite owing to the absence of any meaningful associations with cognition. The composite included the following markers: Aβ₄₀, Aβ₄₂, MMP-1, MMP-3, IL-8, IL-10, and TNF-α. Based on the associations observed with the individual markers, we expected that the composite would correlate with cognitive decline, and we further tested its association with a final clinical diagnosis of pAD.

Table 2

Proposed composite biomarker score

Biomarker	Proposed cutoff (in pg/ml)	Points
Aβ₄₀	25th–50th percentile (176.9–200.5)	−1
	75th–100th percentile (>229.5)	+1
Aβ₄₂	25th–50th percentile (14.1–15.7)	−1
	50th–75th percentile (15.7–18.4)	+1
MMP-1	25th–50th percentile (2083–2985)	+1
	>50th percentile (>2,985)	−1
MMP-3	Women: 1st Tertile (<11,341)	+1
	3rd Tertile (>15,221)	−1
IL-8	25th–50th percentile (2.81–3.41)	+1
	75th–100th percentile (>4.45)	−1
IL-10	25th–50th percentile (0.19–0.25)	+1
	75th–100th percentile (>0.31)	−1
TNF-α	25th–50th percentile (1.55–2.00)	+1
	Total can range from –6 to +7

Protective or low-risk scores (+1) were assigned based on lower decline in cognitive scores over follow-up. A score of –1 was assigned for biomarkers that appeared to be relatively “high-risk” on these analyses i.e. greater decline over follow-up or higher odds of having progressed to MCI/pAD by the time of blood-draw. The scores were added for each retained biomarker to obtain a composite biomarker score.

As expected from this derivation, more favorable (higher) composite scores were associated with less decline on cognitive tests by the time of blood-draw (Fig. 5). Participants with higher biomarker composite-scores showed less decline over follow-up on the MoCA (acoef: 0.5, 0.1–1.0, p = 0.01; MCI/pAD acoef: 1.1, 0.5–1.7, p = 0.001), and were significantly less likely to have dropped by ≥5 points (aOR: 0.59, 0.40–0.88, p = 0.009). MCI/pAD patients with higher biomarker composite-scores also had less decline in Delayed Recall (acoef: 0.3, 0.1–0.5, p = 0.009). Similarly on the CAMCOG, participants with higher composite-scores showed less decline (acoef: 1.3, 0.4–2.1, p = 0.003), and were less likely to drop by ≥5 points (aOR: 0.60, 0.44–0.83, p = 0.002); in fact, none of those participants with composite-scores ≥+2 had shown such a decline. Favorable composite scores also correlated with less decline on the Orientation section (acoef: 0.2, 0.04–0.4, p = 0.02).

Fig.5

Biomarker composite and cognition. Relationship between the proposed biomarker composite (comprising Aβ₄₀, Aβ₄₂, MMP-1, MMP-3, IL-8, IL-10, and TNF-α) and (A) the change in MoCA scores from baseline to last follow-up for all 107 participants, (B) the change in MoCA scores for the MCI/pAD subset, (C) the change in CAMCOG scores for all participants, and (D) the change in CAMCOG scores for the MCI/pAD subset. Negative biomarker scores reflect less favorable composites (i.e., associated to higher decline on cognitive tests by the time of blood-draw) while positive biomarker scores reflect more favorable composites (associated to lower = cognitive decline). Grey regions represent 95% confidence intervals for the best-fit curves. Circles represent individual participants.

Participants with a final clinical diagnosis of pAD had more negative composite scores (pAD: median –1, IQR –2.5 to 1; rest of participants: median 0, IQR –1 to 1, Wilcoxon rank-sum p = 0.01). Individuals with higher composite scores carried lower odds of a final clinical diagnosis of pAD by the last follow-up (aOR: 0.66, 0.51–0.87, p = 0.003).

The age/sex-adjusted logistic model that included the plasma composite-score had a c-statistic (or area under ROC curve, AUC) of 0.732 (95% CI 0.614–0.849) for the association with a final clinical pAD diagnosis, representing an improvement over a model that included only age/sex (AUC = 0.602, 0.477–0.747) and comparable to the final age/sex-adjusted model selected by stepwise logistic regression consisting of IL-10 and MMP-1 (AUC = 0.689, 0.565–0.813; Supplementary Table 1, ROC curves in Supplementary Figure 4). The age/sex-adjusted logistic model with the biomarker composite score for identifying decline on the MOCA by ≥5 points had an AUC of 0.751 (95% CI 0.544–0.958) and was also an improvement over age/sex alone (AUC = 0.635, 0.452–0.818) and the age/sex-adjusted model consisting of IL-10 selected by stepwise regression (AUC = 0.685, 0.472–0.899). The biomarker composite-score also identified those participants whose CAMCOG declined by ≥5 points (AUC = 0.844, 95% CI 0.751–0.936) better than age/sex alone (AUC = 0.728, 0.609–0.846) and comparable to the best-fit age/sex-adjusted model from stepwise regression consisting of IL-10 and MMP-3 (AUC = 0.787, 0.685–0.889).

DISCUSSION

In this preliminary study, we measured a panel of plasma markers at final clinical follow-up, and examined their correlation with the participants’ established clinical diagnoses and cognitive trajectories at that time-point. First, we showed that specific profiles of individual markers of the amyloid, metallo-protease, and inflammatory pathways were associated with established decline in cognitive scores per the MoCA and CAMCOG tests. Second, we derived a pragmatic composite-score based on these associations, which was more strongly associated with final dementia status than age/sex alone. This panel extends beyond the core AD analytes to include inflammatory cytokines as well as metallo-proteinases, reflecting other aspects of the complex pathology underlying diagnoses of pAD.

As higher plasma levels of Aβ₄₀ and Aβ₄₂, and lower levels of the inflammatory markers TNF-α, IL-8, and IL-10 appeared “protective” in our analysis, correlating with better cognitive performance and a lower likelihood of having progressed to MCI/pAD, our results support the role of both the amyloid cascade and inflammatory processes in the progression of AD pathology. Inflammation in AD is a complex and yet incompletely understood phenomenon [8], traditionally ascribed to a late “protective reaction” to amyloid plaque deposition and advanced pathology [26 –28], but now increasingly thought to precede the deposition of amyloid plaques with a detrimental effect on synaptic and cognitive processes [29 –31]. These concepts, supported by strong experimental and clinical evidence, have been recently reviewed [8] and a call for action from an international expert panel to re-discuss these views has been put forward [32]. Recently, an analysis of the Atherosclerosis Risk in Communities cohort found that increases in a different set of inflammatory markers—fibrinogen, albumin, white blood cells, von Willebrand factor, and factor VIII— during midlife were associated with smaller brain volumes as well as reduced episodic memory 24 years later [33]. Our findings complement this work by exploring additional inflammatory markers associated with cognitive decline and progression to dementia.

Despite the central role that amyloid pathology is ascribed in the pathogenesis of AD, single amyloid species measured in plasma have not previously been shown to associate with AD risk, although a small negative correlation between the Aβ_40/42 ratio and AD risk has been observed [34]. However, here we found that higher plasma levels of either Aβ₄₀ or Aβ₄₂ were associated with a lower risk of having progressed to MCI/pAD. This apparent discrepancy may reflect a methodological difference; the multi-spot MSD assays used here have better sensitivity and a broader dynamic range than conventional ELISA assays, so fewer samples would fall below detection limit. In addition, matrix effects, which are an issue with complex biological samples such as plasma, are greatly reduced with this technology [35]. Reductions in plasma Aβ while brain amyloidosis increases likely reflect a failure of brain-to-vasculature clearance mechanisms [36], paralleling the Aβ “sink” that is believed to explain AD-associated reductions in CSF Aβ levels [37]. In contrast, higher levels of Aβ in plasma correlate to increased risk of AD and cognitive decline in people with Down syndrome [14 , 39]. This may reflect different causes of increased brain amyloidosis in the two conditions, with a failure of Aβ clearance potentially playing a more dominant role in sporadic AD [36] versus enhanced Aβ production predominating in Down syndrome due to chromosome 21 and APP gene triplication [40].

Our study also provides further insights into the role of metallo-proteinases in dementia as reflected in peripheral blood. Our correlation analysis revealed that lower titers of MMP-1 were associated with better cognitive scores and less decline, while MMP-3 exhibited a sexually dimorphic effect with respect to cognition. Matrix metallo-proteinases have been implicated in various aspects of AD pathology, including amyloid metabolism [41, 42], vascular integrity [43], synaptic remodeling [44], inflammation [45], and NGF degradation [16]. Our findings are in agreement with other fluid biomarker studies showing MMP-3 elevations in plasma or CSF of individuals with MCI, AD, and Down syndrome [14, 46]. It intrigued us that we did not find a consistent relationship between plasma MMP-9 and cognition, nor between MMP-9 and disease progression, given its role in NGF degradation [16]. We have previously reported significant correlations between MMP-9 zymogenic activity measured in postmortem MCI brains and cognitive scores at time of death [47]. It is possible that analysis of activity rather than levels is a better predictor of cognition, or alternatively, that plasma MMP-9, which mostly derives from neutrophil production [48], inadequately reflects MMP-9 content in the brain.

Whereas we did not see differences in MMP-9 titers among the different groups, we found a significant and consistent elevation in MMP-3 at the non-demented stages as well as in pAD patients. The finding of MMP-3 elevation in the non-demented groups is interesting, as it suggests a potential for this metallo-protease to be a helpful biomarker in pre-symptomatic stages of AD, an issue warranting future investigations. Indeed, in a population of young adults with Down syndrome, who develop AD pathology throughout their lifetime, we also found higher MMP-3 levels in plasma in the absence of AD-like symptoms [14]. In the brain, MMP-3 derived from neurons under metabolic stress is reported to participate in microglia activation and in the induction of neuroinflammation [49, 50]. Interestingly, MMP-3 can cleave and activate MMP-9 [51], suggesting that MMP-3 elevations could indirectly exacerbate NGF degradation in AD brains.

Another interesting finding of our study is the sex difference in MMP-3, with higher plasma levels associating with greater cognitive decline in women and correlating with increased pAD risk but not in men. A strong sex relationship with plasma MMP-3 was also reported in a Down syndrome population [14]. Although these sex-specific differences require further validation and elucidation, our results agree with emerging studies highlighting important sex-effects in AD epidemiology and phenotypic variability [52 –54]. Notably, a polymorphism in the MMP-3 gene was shown to interact with the APOE ɛ4 allele and to predispose carriers to AD [55]. It is worth investigating whether this polymorphism varies by sex or whether MMP-3 interacts with other sexually dimorphic genetic risk factors that differentially modulate the risk of AD in men and women. In sum, the above observations invite further investigations regarding the fate of metallo-proteases in the progression of AD neuropathology and neurodegenerative conditions, in general.

Several biomarkers are currently used to identify individuals with incipient AD or to confirm diagnoses of AD. Of these, the most common include CSF Aβ₄₂, CSF phosphorylated tau or total tau, and tracer retention on amyloid (typically Florbetapir F-18 or Pittsburg Compound B) PET, following the IWG-2 criteria [56]. Other putative markers, including both total and region-specific tau PET tracers [57], FDG-PET [58], and structural imaging measures [59] have demonstrated utility for the identification of preclinical AD, for assessing risk of conversion from MCI to AD, and for confirming clinical diagnoses of AD. While the sensitivity and specificity of our investigative biomarker composite is lesser than standard PET and CSF modalities, this is unsurprising as our composite was associated with cognitive decline and probabilistic diagnoses of AD by neuropsychological testing, which entail greater heterogeneity than is found in a confirmed diagnosis of AD by CSF or amyloid imaging.

Furthermore, the advantages of plasma markers reside primarily in their accessibility and cost-effectiveness; PET scanning is too expensive to use for population-level screening for preclinical AD and requires equipment unavailable to remote or developing populations, while lumbar puncture is invasive and requires an expert clinician. In this regard, we propose that our composite could serve as part of a larger composite plasma test for AD or could be used to identify aged individuals at risk for dementia before more costly and invasive testing is pursued.

A number of potential limitations of the present study merit discussion. First, we measured the plasma markers of interest at the end of the follow-up period and related them retrospectively to the patients’ clinical trajectories, and did not assess for changes in these markers over time. We chose to do this as a first step in assessing whether further study of this plasma panel would be of value, since the absence of any relevant differences at the final follow-up— by which time clinical diagnoses and cognitive deterioration would be most apparent— could potentially help us eliminate markers that were least likely to be useful. However, this means that we cannot infer causal relationships based on the observed associations, nor can we assume that plasma marker levels were similar at baseline. Considering the associations found between the plasma panel, cognitive decline and disease progression, these findings justify further study of these investigative biomarkers earlier in the disease course and longitudinally.

Second, we only measured the molecules of interest in a subset of patients from the whole LEAD cohort, and our selection did not involve controlling for age, sex, or risk factors like APOE genotype, so there is a risk of selection bias. However, we controlled for age/sex in all our key analyses.

Third, the proposed quartile/tertile-based cut-offs for each marker in our composite-score are primarily illustrative at this point; the distribution of concentrations may be considerably different in the general population.

Fourth, at the time this cohort was recruited we did not have access to amyloid imaging or CSF testing to diagnose pAD or MCI, therefore we cannot exclude the possibility that some of these clinically diagnosed subjects could be amyloid negative. However, the “gold standard” in our study was the final research diagnosis based on the integration of clinical history, examination, cognitive tests, and CT/MRI findings, which reflects how dementia has been diagnosed at the time of study conception, prior to routine analysis of CSF tau and amyloid PET imaging.

Fifth, our analysis did not include patients known to have non-AD-related dementia, so we cannot comment on the specificity of our findings for AD versus other causes of dementia, such as vascular cognitive impairment, frontotemporal dementia, or others. However, this opens new avenues for investigations of the proposed plasma panel in other neurodegenerative conditions.

Sixth, our initial analyses in SMI will have incorporated the well-described heterogeneity that still exists within this classification [60, 61], and any changes observed within this group cannot be assumed to be AD-specific. In later correlational analyses we included only those SMI individuals that later progressed to MCI or pAD, for which the proposition that SMI reflects early AD is much more plausible. Therefore, more rigorous prospective cohort-based studies in independent populations will be required to compare the diagnostic accuracy of different cut-offs of the proposed composite-score— as well as different cut-offs for each marker included in the score—against amyloid imaging, CSF testing, and other blood-based signatures, both for identifying patients at risk of cognitive decline and for distinguishing AD from non-AD pathologies.

Despite these limitations, the key strengths of our study are its nesting within a well-characterized prospective cohort, with rigorous age/sex-adjusted longitudinal analyses for each marker for both final MoCA scores and change over time, as well as validation of these trends in the CAMCOG test, and in analyses of clinical progression. This is an important improvement over prior studies of composite blood-based signatures, which have often not demonstrated added predictive power beyond demographic metrics for the individual components [62].

We further validated our findings by demonstrating the added value of using a composite score derived from the most promising markers; this also demonstrated a “dose dependence” of these markers— the more protective (+) factors present, the lower the odds of a pAD diagnosis by blood-draw, and the more detrimental (–) factors present, the stronger the association with a pAD diagnosis by blood-draw. Importantly, the relative simplicity of the scoring system that we have proposed can facilitate easier use in clinical practice or trial-based settings, should these preliminary findings be validated in prospective studies. This potential additive effect of different markers may explain why we did not observe significant differences in individual concentrations among diagnostic groups (except for MMP-3), despite the significant correlations seen for cognitive decline, but, importantly, observed differences in composite scores between those with and without a final pAD diagnosis.

Conclusions

Plasma Aβ, metallo-proteinases, and inflammatory cytokines were associated with established cognitive deterioration by final follow-up in this pilot study. A composite profile including Aβ₄₀, Aβ₄₂, MMP-1, MMP-3, IL-8, IL-10, and TNF-α better correlated with clinically diagnosed pAD than individual markers or age/sex alone. These results need to be replicated in prospective longitudinal studies and ideally, combined with other promising markers to potentially create optimized panels for screening. As therapeutic strategies for AD progress, it will be crucial to develop inexpensive and widely adoptable screening methods to identify patients at a pre-clinical stage and monitor participants over time, first in cohorts for secondary prevention trials of AD, and then for screening in the general clinic population.

Footnotes

ACKNOWLEDGMENTS

The authors express gratitude to participants in the LEAD study for their time and commitment. The LEAD team dedicates this paper to Rose M. Wharton, MSc, statistician and beloved team member at the Centre for Prevention of Stroke and Dementia, University of Oxford, who unfortunately passed away recently. The authors thank Dr. Mark Hancock from McGill University for technical assistance with the Sector Instrument and data extraction, Louise Silver and Aimee Gornall for assisting with data management/extraction, and Donald Warden for assistance with plasma sample shipment between Oxford and McGill University.

ACC acknowledges financial support from the Canadian Institutes of Health Research and the Alzheimer Society of Canada. He holds the McGill University Charles E. Frosst/Merck Chair in Pharmacology and is a member of the Canadian Consortium of Neurodegeneration in Aging. GKW was supported by the Oxford NIHR Biomedical Research Centre, UK. AG was funded by the Rhodes Trust and has received a grant from the Murray Speight foundation. RP was the recipient of a Student Fellowship from the McGill Integrated Program in Neuroscience and a CIHR Doctoral Award. LFA was supported by a Doctoral Merit Scholarship from the Fonds de recherche du Quebec-Nature et Technologies and a Doctoral Award granted by the National Council for Science and Technology (CONACyT) of Mexico. The funding bodies had no role in the design of the study or in the collection, analysis, and interpretation of data or in writing the manuscript.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Hampel

, Lista

, Khachaturian

(2012) Development of biomarkers to chart all Alzheimer’s disease stages: The royal road to cutting the therapeutic Gordian Knot. Alzheimers Dement 8, 312–336.

Glenner

, Wong

(1984) Alzheimer’s disease: Initial report of the purification and characterization of a novel cerebrovascular amyloid protein. Biochem Biophys Res Commun 120, 885–890.

Grundke-Iqbal

, Iqbal

, Quinlan

, Tung

, Zaidi

, Wisniewski

(1986) Microtubule-associated protein tau. A component of Alzheimer paired helical filaments. J Biol Chem 261, 6084–6089.

Selkoe

, Hardy

(2016) The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO Mol Med 8, 595–608.

Goedert

, Wischik

, Crowther

, Walker

, Klug

(1988) Cloning and sequencing of the cDNA encoding a core protein of the paired helical filament of Alzheimer disease: Identification as the microtubule-associated protein tau. Proc Natl Acad Sci U S A 85, 4051–4055.

DeKosky

, Scheff

(1990) Synapse loss in frontal cortex biopsies in Alzheimer’s disease: Correlation with cognitive severity. Ann Neurol 27, 457–464.

Terry

, Masliah

, Salmon

, Butters

, DeTeresa

, Hill

, Hansen

, Katzman

(1991) Physical basis of cognitive alterations in Alzheimer’s disease: Synapse loss is the major correlate of cognitive impairment. Ann Neurol 30, 572–580.

Cuello

(2017) Early and late CNS inflammation in Alzheimer’s disease: Two extremes of a continuum? Trends Pharmacol Sci 38, 956–966.

Heneka

, Carson

, El Khoury

, Landreth

, Brosseron

, Feinstein

, Jacobs

, Wyss-Coray

, Vitorica

, Ransohoff

, Herrup

, Frautschy

, Finsen

, Brown

, Verkhratsky

, Yamanaka

, Koistinaho

, Latz

, Halle

, Petzold

, Town

, Morgan

, Shinohara

, Perry

, Holmes

, Bazan

, Brooks

, Hunot

, Joseph

, Deigendesch

, Garaschuk

, Boddeke

, Dinarello

, Breitner

, Cole

, Golenbock

, Kummer

(2015) Neuroinflammation in Alzheimer’s disease. Lancet Neurol 14, 388–405.

10.

McGeer

, Rogers

, McGeer

(2016) Inflammation, antiinflammatory agents, and Alzheimer’s disease: The last 22 years. J Alzheimers Dis 54, 853–857.

11.

Bowen

, Smith

, White

, Davison

(1976) Neurotransmitter-related enzymes and indices of hypoxia in senile dementia and other abiotrophies. Brain 99, 459–496.

12.

Davies

, Maloney

(1976) Selective loss of central cholinergic neurons in Alzheimer’s disease. Lancet 2, 1403.

13.

Pearson

, Sofroniew

, Cuello

, Powell

, Eckenstein

, Esiri

, Wilcock

(1983) Persistence of cholinergic neurons in the basal nucleus in a brain with senile dementia of the Alzheimer’s type demonstrated by immunohistochemical staining for choline acetyltransferase. Brain Res 289, 375–379.

14.

Iulita

, Ower

, Barone

, Pentz

, Gubert

, Romano

, Cantarella

, Elia

, Buono

, Recupero

, Romano

, Castellano

, Bosco

, Di Nuovo

, Drago

, Caraci

, Cuello

(2016) An inflammatory and trophic disconnect biomarker profile revealed in Down syndrome plasma: Relation to cognitive decline and longitudinal evaluation. Alzheimers Dement 12, 1132–1148.

15.

Rosenberg

(2009) Matrix metalloproteinases and their multiple roles in neurodegenerative diseases. Lancet Neurol 8, 205–216.

16.

Bruno

, Cuello

(2006) Activity-dependent release of precursor nerve growth factor, conversion to mature nerve growth factor, and its degradation by a protease cascade. Proc Natl Acad Sci U S A 103, 6735–6740.

17.

Nasreddine

, Phillips

, Bedirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

18.

Roth

, Tym

, Mountjoy

, Huppert

, Hendrie

, Verma

, Goddard

(1986) CAMDEX. A standardised instrument for the diagnosis of mental disorder in the elderly with special reference to the early detection of dementia. Br J Psychiatry 149, 698–709.

19.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

20.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

21.

Sano

, Raman

, Emond

, Thomas

, Petersen

, Schneider

, Aisen

(2011) Adding delayed recall to the Alzheimer Disease Assessment Scale is useful in studies of mild cognitive impairment but not Alzheimer disease. Alzheimer Dis Assoc Disord 25, 122–127.

22.

O’Keeffe

, Mukhtar

, O’Keeffe

(2011) Orientation to time as a guide to the presence and severity of cognitive impairment in older hospital patients. J Neurol Neurosurg Psychiatry 82, 500–504.

23.

Ballard

, Burns

, Corbett

, Livingston

, Rasmussen

(2013) Helping you assess cognition: A practical toolkit for clinicians. Alzheimer’s Society, https://www.alzheimers.org.uk/downloads/download/1045/helping_you_to_assess_cognition_a_practical_toolkit_for_clinicians

24.

Andrew

, Rockwood

(2008) A five-point change in Modified Mini-Mental State Examination was clinically meaningful in community-dwelling elderly people. J Clin Epidemiol 61, 827–831.

25.

DeLong

, DeLong

, Clarke-Pearson

(1988) Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 44, 837–845.

26.

Eikelenboom

, Rozemuller

, van Muiswinkel

(1998) Inflammation and Alzheimer’s disease: Relationships between pathogenic mechanisms and clinical expression. Exp Neurol 154, 89–98.

27.

McGeer

, Itagaki

, Tago

, McGeer

(1987) Reactive microglia in patients with senile dementia of the Alzheimer type are positive for the histocompatibility glycoprotein HLA-DR. Neurosci Lett 79, 195–200.

28.

Rogers

, Cooper

, Webster

, Schultz

, McGeer

, Styren

, Civin

, Brachova

, Bradt

, Ward

et al., (1992) Complement activation by beta-amyloid in Alzheimer disease. Proc Natl Acad Sci U S A 89, 10016–10020.

29.

Ferretti

, Bruno

, Ducatenzeiler

, Klein

, Cuello

(2012) Intracellular Abeta-oligomers and early inflammation in a model of Alzheimer’s disease. Neurobiol Aging 33, 1329–1342.

30.

Hanzel

, Pichet-Binette

, Pimentel

, Iulita

, Allard

, Ducatenzeiler

, Do Carmo

, Cuello

(2014) Neuronal driven pre-plaque inflammation in a transgenic rat model of Alzheimer’s disease. Neurobiol Aging 35, 2249–2262.

31.

Hong

, Beja-Glasser

, Nfonoyim

, Frouin

, Li

, Ramakrishnan

, Merry

, Shi

, Rosenthal

, Barres

, Lemere

, Selkoe

, Stevens

(2016) Complement and microglia mediate early synapse loss in Alzheimer mouse models. Science 352, 712–716.

32.

Rogers

(2018) Principles for central nervous system inflammation research: A call for a consortium approach. Alzheimers Dement 14, 1553–1559.

33.

Walker

, Hoogeveen

, Folsom

, Ballantyne

, Knopman

, Windham

, Jack

CR Jr

, Gottesman

(2017) Midlife systemic inflammatory markers are associated with late-life brain volume: The ARIC study. Neurology 89, 2262–2270.

34.

Koyama

, Okereke

, Yang

, Blacker

, Selkoe

, Grodstein

(2012) Plasma amyloid-beta as a predictor of dementia and cognitive decline: A systematic review and meta-analysis. Arch Neurol 69, 824–831.

35.

Rosenberg-Hasson

, Hansmann

, Liedtke

, Herschmann

, Maecker

(2014) Effects of serum and plasma matrices on multiplex immunoassays. Immunol Res 58, 224–233.

36.

Mawuenyega

, Sigurdson

, Ovod

, Munsell

, Kasten

, Morris

, Yarasheski

, Bateman

(2010) Decreased clearance of CNS beta-amyloid in Alzheimer’s disease. Science 330, 1774.

37.

Fagan

, Mintun

, Mach

, Lee

, Dence

, Shah

, LaRossa

, Spinner

, Klunk

, Mathis

, DeKosky

, Morris

, Holtzman

(2006) Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans. Ann Neurol 59, 512–519.

38.

Coppus

, Schuur

, Vergeer

, Janssens

, Oostra

, Verbeek

, van Duijn

(2012) Plasma beta amyloid and the risk of Alzheimer’s disease in Down syndrome. Neurobiol Aging 33, 1988–1994.

39.

Schupf

, Patel

, Silverman

, Zigman

, Zhong

, Tycko

, Mehta

, Mayeux

(2001) Elevated plasma amyloid beta-peptide 1-42 and onset of dementia in adults with Down syndrome. Neurosci Lett 301, 199–203.

40.

Head

, Lott

(2004) Down syndrome and beta-amyloid deposition. Curr Opin Neurol 17, 95–100.

41.

Hernandez-Guillamon

, Mawhirt

, Blais

, Montaner

, Neubert

, Rostagno

, Ghiso

(2015) Sequential amyloid-beta degradation by the matrix metalloproteases MMP-2 and MMP-9. J Biol Chem 290, 15078–15091.

42.

Lammich

, Kojro

, Postina

, Gilbert

, Pfeiffer

, Jasionowski

, Haass

, Fahrenholz

(1999) Constitutive and regulated alpha-secretase cleavage of Alzheimer’s amyloid precursor protein by a disintegrin metalloprotease. Proc Natl Acad Sci U S A 96, 3922–3927.

43.

Lee

, Yin

, Hsin

, Chen

, Fryer

, Holtzman

, Hsu

, Xu

(2005) Matrix metalloproteinase-9 in cerebral-amyloid-angiopathy-related hemorrhage. J Neurol Sci 229-230, 249–254.

44.

Restituito

, Khatri

, Ninan

, Mathews

, Liu

, Weinberg

, Ziff

(2011) Synaptic autoregulation by metalloproteases and gamma-secretase. J Neurosci 31, 12083–12093.

45.

, Wang

, Cuzner

(2002) Extracellular proteolysis in brain injury and inflammation: Role for plasminogen activators and matrix metalloproteinases. J Neurosci Res 69, 1–9.

46.

Hanzel

, Iulita

, Eyjolfsdottir

, Hjorth

, Schultzberg

, Eriksdotter

, Cuello

(2014) Analysis of matrix metallo-proteases and the plasminogen system in mild cognitive impairment and Alzheimer’s disease cerebrospinal fluid. J Alzheimers Dis 40, 667–678.

47.

Bruno

, Mufson

, Wuu

, Cuello

(2009) Increased matrix metalloproteinase 9 activity in mild cognitive impairment. J Neuropathol Exp Neurol 68, 1309–1318.

48.

Ardi

, Kupriyanova

, Deryugina

, Quigley

(2007) Human neutrophils uniquely release TIMP-free MMP-9 to provide a potent catalytic stimulator of angiogenesis. Proc Natl Acad Sci U S A 104, 20262–20267.

49.

Yoshiyama

, Asahina

, Hattori

(2000) Selective distribution of matrix metalloproteinase-3 (MMP-3) in Alzheimer’s disease brain. Acta Neuropathol 99, 91–95.

50.

Kim

, Choi

, Block

, Lorenzl

, Yang

, Kim

, Sugama

, Cho

, Hwang

, Browne

, Kim

, Hong

, Beal

, Joh

(2007) A pivotal role of matrix metalloproteinase-3 activity in dopaminergic neuronal degeneration via microglial activation. FASEB J 21, 179–187.

51.

Ogata

, Enghild

, Nagase

(1992) Matrix metalloproteinase 3 (stromelysin) activates the precursor for the human matrix metalloproteinase 9. J Biol Chem 267, 3581–3584.

52.

Barnes

, Wilson

, Bienias

, Schneider

, Evans

, Bennett

(2005) Sex differences in the clinical manifestations of Alzheimer disease pathology. Arch Gen Psychiatry 62, 685–691.

53.

Gamberger

, Lavrac

, Srivatsa

, Tanzi

, Doraiswamy

(2017) Identification of clusters of rapid and slow decliners among subjects at risk for Alzheimer’s disease. Sci Rep 7, 6763.

54.

Neu

, Pa

, Kukull

, Beekly

, Kuzma

, Gangadharan

, Wang

, Romero

, Arneric

, Redolfi

, Orlandi

, Frisoni

, Au

, Devine

, Auerbach

, Espinosa

, Boada

, Ruiz

, Johnson

, Koscik

, Wang

, Hsu

, Chen

, Toga

(2017) Apolipoprotein E genotype and sex risk factors for Alzheimer disease: A meta-analysis. JAMA Neurol 74, 1178–1189.

55.

Saarela

, Lehtimaki

, Rinne

, Hervonen

, Jylha

, Roytta

, Ahonen

, Mattila

(2004) Interaction between matrix metalloproteinase 3 and the epsilon4 allele of apolipoprotein E increases the risk of Alzheimer’s disease in Finns. Neurosci Lett 367, 336–339.

56.

Dubois

, Hampel

, Feldman

, Scheltens

, Aisen

, Andrieu

, Bakardjian

, Benali

, Bertram

, Blennow

, Broich

, Cavedo

, Crutch

, Dartigues

, Duyckaerts

, Epelbaum

, Frisoni

, Gauthier

, Genthon

, Gouw

, Habert

, Holtzman

, Kivipelto

, Lista

, Molinuevo

, O’Bryant

, Rabinovici

, Rowe

, Salloway

, Schneider

, Sperling

, Teichmann

, Carrillo

, Cummings

, Jack

Jr , Proceedings of the Meeting of the International Working Group the American Alzheimer’s Association on The Preclinical State of AD July Washington DC, USA (2016) Preclinical Alzheimer’s disease: Definition, natural history, and diagnostic criteria. Alzheimers Dement 12, 292–323.

57.

Maass

, Landau

, Baker

, Horng

, Lockhart

, La Joie

, Rabinovici

, Jagust

, Alzheimer’s Disease Neuroimaging Initiative (2017) Comparison of multiple tau-PET measures as biomarkers in aging and Alzheimer’s disease. Neuroimage 157, 448–463.

58.

Mosconi

, Tsui

, Herholz

, Pupi

, Drzezga

, Lucignani

, Reiman

, Holthoff

, Kalbe

, Sorbi

, Diehl-Schmid

, Perneczky

, Clerici

, Caselli

, Beuthien-Baumann

, Kurz

, Minoshima

, de Leon

(2008) Multicenter standardized 18F-FDG PET diagnosis of mild cognitive impairment, Alzheimer’s disease, and other dementias. J Nucl Med 49, 390–398.

59.

Eskildsen

, Coupe

, Fonov

, Pruessner

, Collins

, Alzheimer’s Disease Neuroimaging Initiative (2015) Structural imaging biomarkers of Alzheimer’s disease: Predicting disease progression. Neurobiol Aging 36 (Suppl 1), S23–31.

60.

Jessen

, Wolfsgruber

, Wiese

, Bickel

, Mosch

, Kaduszkiewicz

, Pentzek

, Riedel-Heller

, Luck

, Fuchs

, Weyerer

, Werle

, van den Bussche

, Scherer

, Maier

, Wagner

, German Study on Aging C, Dementia in Primary Care P (2014) AD dementia risk in late MCI, in early MCI, and in subjective memory impairment. Alzheimers Dement 10, 76–83.

61.

Balash

, Mordechovich

, Shabtai

, Giladi

, Gurevich

, Korczyn

(2013) Subjective memory complaints in elders: Depression, anxiety, or cognitive decline? Acta Neurol Scand 127, 344–350.

62.

Burnham

, Rowe

, Baker

, Bush

, Doecke

, Faux

, Laws

, Martins

, Maruff

, Macaulay

, Rainey-Smith

, Savage

, Ames

, Masters

, Wilson

, Villemagne

(2016) Predicting Alzheimer disease from a blood-based biomarker profile: A 54-month follow-up. Neurology 87, 1093–1101.