UNVEILING POTENTIAL BIOMARKERS IN CEREBROSPINAL FLUID FOR AMYLOID PATHOLOGICAL POSITIVITY IN NON-DEMENTED INDIVIDUALS

 

F. Meng^1,2, X. Zhang^1,2, for the Alzheimer’s Disease Neuroimaging Initiative (ADNI)

 

1. School of Medicine, Nankai University, Tianjin, 300071, China; 2. Department of Neurology, The Second Medical Centre, National Clinical Research Centre for Geriatric Diseases, Chinese PLA General Hospital, Beijing, 100853, China.

Corresponding Author: Prof. Xi Zhang, Department of Neurology, The Second Medical Centre, National Clinical Research Centre for Geriatric Diseases, Chinese PLA General Hospital, Beijing, 100853, China, School of Medicine, Nankai University, Tianjin, 300071, China, E-mail: smrc301@163.com

J Prev Alz Dis 2024;
Published online July 2, 2024, http://dx.doi.org/10.14283/jpad.2024.129

 

---

Abstract

BACKGROUND: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by amyloid-beta (Aβ) plaque accumulation and neurofibrillary tangles. The recent approval of anti-amyloid therapeutic medications highlights the crucial need for early detection of Aβ pathological abnormalities in individuals without dementia to facilitate timely intervention and treatment.
OBJECTIVE: The primary aim of this study was to identify cerebrospinal fluid (CSF) biomarkers strongly associated with Aβ pathological positivity in a non-demented cohort and evaluate their clinical values.
METHODS: A comprehensive analysis was conducted on 51 CSF proteins (excluding Aβ42, pTau, and Tau) obtained from 474 non-demented participants sourced from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. By utilizing the Least Absolute Shrinkage and Selection Operator (LASSO) regression, we identified potential proteins indicative of Aβ pathological positivity and evaluated their performance in tracking longitudinal pathological progression.
RESULTS: Our LASSO analysis unveiled three candidates: apolipoprotein E (APOE), chitinase-3-like protein 1 (CHI3L1), and SPARC-related modular calcium-binding protein 1 (SMOC1). While SMOC1 did not correlate with Aβ42-related cognitive alterations, it displayed better abilities in discriminating both CSF-Aβ positivity and Aβ-positron emission tomography (PET) positivity than the other two candidates. It could precisely predict longitudinal Aβ-PET status conversion. Notably, SMOC1 was the only protein showing associations with longitudinal Aβ-PET trajectory and enhancing the diagnostic accuracy of Aβ42. The assessment of combined Aβ42 and SMOC1 yielded valuable clinical insights.
CONCLUSION: Our findings elucidated SMOC1 as a potential biomarker for detecting Aβ abnormalities. Aβ42 combining SMOC1 offered critical implications in AD pathological diagnosis and management.

Key words: Biomarkers, cerebrospinal fluid, amyloid pathological positivity, non-demented individuals, Alzheimer’s disease.

---

 

Introduction

Alzheimer’s disease (AD) poses a formidable challenge in the realm of neurodegeneration, characterized by progressive deterioration in cognitive function and memory loss (1, 2). Compelling evidence from clinical investigations suggests that the pathophysiological cascade of AD initiates decades before the onset of overt clinical symptoms (3-6). Notably, amyloid-beta (Aβ) plaques are conceptualized as the initial component of the AT(N) sequence, where Aβ plaques [A] are thought to precede and accelerate tau pathology [T], ultimately leading to neurodegeneration [N] (6, 7).
The recent approval of anti-amyloid therapeutic medicines, such as aducanumab and lecanemab, by the US Food and Drug Administration, underscores the significance of targeting Aβ in the treatment of AD (1, 8). An essential aspect of implementing anti-amyloid therapeutic interventions in clinical practice is the early identification of Aβ pathological positive patients in the initial stages of disease progression when irreversible neurodegeneration is minimal. Therefore, the identification of Aβ pathological positivity in individuals at preclinical or prodromal AD stages presents a crucial window of opportunity for implementing targeted interventions that hold the potential to modify the disease trajectory.
Currently, the two established clinical methods for confirming abnormal Aβ aggregated status are positive Aβ positron emission tomography (PET) scans and positive cerebrospinal fluid (CSF) markers (9-11). It is worth noting that PET scans are expensive and less universally accessible across various healthcare settings compared to CSF markers. Consequently, CSF biomarkers have emerged as promising tools for the early detection of AD pathology. By analyzing specific proteins in the CSF, researchers can glean valuable insights into the molecular alterations occurring in the brain parenchyma(12-14). This study focused on the identification of novel key proteins beyond the traditional AT(N) framework in the CSF that held the potential to serve as biomarkers for indicating amyloid pathology in individuals without dementia. By delving into the CSF proteins and identifying candidate proteins, this research contributed to the ongoing exploration of substantial pathogenesis for individuals in the earliest pathological stages of AD.
We employed the Least Absolute Shrinkage and Selection Operator (LASSO) regression to screen proteins, aiming to identify proteins correlated with the early dysregulation of Aβ in CSF and the subsequent Aβ aggregation detected by PET. Then we conducted a comprehensive evaluation of the candidate proteins by examining their capacity to discriminate and predict Aβ states, their correlations with the longitudinal Aβ pathological progression, and their potential to improve the pathological diagnostic accuracy of Aβ42. Finally, we quantified the clinical efficacy of combining Aβ42 with the final identified proteins in tracking Aβ-PET positivity incident and Aβ aggregation progression. Our ultimate objective was to harness these discoveries to propel the development of personalized medicine approaches and preclinical monitoring strategies to improve outcomes and the quality of life for individuals at risk of AD.

 

Method

Cohort Description

Our study encompassed a cohort of 474 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, comprising 178 cognitively normal (CN) individuals and 296 with mild cognitive impairment (MCI). CSF biomarkers, including Aβ38, Aβ40, Aβ42, pTau, Tau, and 49 other proteins, were measured (refer to Supplementary Table 1). We included a group of proteins whose binding demonstrated significant capabilities of Aβ pathology distinguishing (15) and added TREM2 and PGRN to address the absence of direct microglial cell-related markers. We also included Aβ38 and Aβ40. Despite the limited number of proteins chosen, this strategic protein selection aimed to efficiently utilize resources and ensure sample retention. Besides, several cognitive assessments were administered including the Mini-Mental State Examination (MMSE), the Modified Preclinical Alzheimer Cognitive Composite incorporating the Digit Symbol Substitution Test (mPACCdigit), and the Modified Preclinical Alzheimer Cognitive Composite incorporating the Trail-Making Test Part B (mPACCtrailsB).
Among 474 participants, 385 individuals (151 CN and 234 MCI) had longitudinal whole-brain 18F-florbetapir (FBP) PET summarized data. In our investigation, the selection of candidate variables pertaining to the pathological states of Aβ-PET and CSF-Aβ, as well as the comprehensive evaluation of Aβ pathological diagnostic performance, was anchored in a cohort comprising 474 individuals. Concurrently, the foundational assessment of candidate proteins in predicting the trajectory of longitudinal cognitive decline was also rooted in this cohort. Moreover, when delving into the prognostic capabilities concerning the augmentation of FBP-PET standardized uptake value ratios (SUVRs) and the propensity for Aβ-PET pathological progression among Aβ-PET negative participants, our analysis was predicated upon data derived from a subset of 385 individuals.

Measurement of CSF Protein Levels

The quantification of Aβ38 and Aβ40 in the CSF was performed using a two-dimensional ultraperformance liquid chromatography-tandem mass spectrometry technique. CSF pTau, Tau, and Aβ42 were measured with the Elecsys platform. Measurements of CSF TREM2 and PGRN were conducted through an MSD platform-based assay. The remaining proteins were quantified using a standard flow Agilent 1290 Infinity II liquid chromatography system. Detailed methodological information is accessible at LONI (ida.loni.usc.edu).
For certain proteins, multiple peptides were assayed, and we calculated the correlation coefficients among peptides corresponding to the same protein, yielding coefficients ranging from 0.72 to 0.99 (P < 0.001). Subsequently, we derived the CSF protein levels by averaging the contents of their respective peptides.

APOE Genotyping

The APOE genotyping for the ADNI cohort was conducted using DNA samples obtained and processed by Cogenics (Beckman-Coulter, Pasadena, California).

FBP-PET SUVRs Data and Aβ Pathological Positivity Identification

The whole-brain summarized FBP-PET SUVRs data were downloaded from the LONI website (ida.loni.usc.edu). Aβ-PET positivity was defined based on thresholds for FBP-PET SUVRs (summarized SUVRs based on the whole cerebellum reference region ≥ 1.11 for cross-sectional and summarized SUVRs based on the composite reference region ≥ 0.78 for longitudinal data) as described on the LONI website. CSF-Aβ positivity was defined as Aβ42/pTau < 39.20 in Aβ-PET negative participants (16). CSF-Aβ positivity was recognized as indicative of the pre-pathological stage leading to amyloid plaque aggregation (Aβ-PET positivity).

Statistical Analysis

Demographic Characteristics

To compare demographic characteristics across different Aβ-PET groups, we employed Chi-square tests for categorical variables such as gender and APOE ε4 carrying status. For continuous variables, the Mann-Whitney U test and t-test were utilized for non-normally and normally distributed data, respectively. Baseline CSF protein levels and cognitive condition across different Aβ-PET statuses were compared using general linear models, with adjustments of age, gender, APOE ε4 carrying status, and years of education. The CSF protein levels were preprocessed through log transformation.

Candidate Hub Variable Identification

For candidate hub variables screening, we initially randomly sampled 70% of the total participants for the LASSO regression (excluding Aβ42, pTau, and Tau). The protein levels were preprocessed through log transformation and standardization (z-score normalization). This selection was based on the lambda.1se criterion derived from the ‘glmnet’ package in R. This process was iterated 1000 times, with variables chosen in over 500 iterations deemed most pertinent to the Aβ-PET pathological state (17, 18). The same procedure was conducted in Aβ-PET negative participants to pinpoint hub variables associated with CSF-Aβ positivity. Subsequently, we overlapped these two protein subsets to identify proteins of significance in both Aβ-PET positive and pre-positive phases.

Candidate Hub Variable Evaluation

To explore the associations between baseline candidate hub protein levels and the trajectory of cognitive decline alongside FBP-PET SUVRs increase, we employed the following two linear mixed-effect (LME) models, each incorporating random intercepts and slopes. The candidate proteins and Aβ42 levels were preprocessed through log transformation and z-score normalization. All other continuous variables were also z-scored. The formulas for calculating the LME models were as follows:
Assessing the predictive capacity of candidate hub proteins on longitudinal FBP-PET SUVRs increase and cognitive decline:
Dependent variable (Cognitive scores, SUVRs) (Time-varying) ~ β0 + β1 * (Baseline hub protein levels) * (Time) + βx * (Age, Gender, APOE ε4 carrying status, Years of education) * (Time) + βy * (Main terms of each variable)
Assessing the predictive capacity of candidate hub proteins on longitudinal Aβ42-related cognitive alterations:
Dependent variable (Cognitive scores) (Time-varying) ~ β0 + β1 * (Baseline hub protein levels) * (Baseline Aβ42 level) * (Time) + β2 * (Baseline hub protein levels) * (Time) + β3 * (Baseline Aβ42 level) * (Time) + βx * (Age, Gender, APOE ε4 carrying status, Years of education) * (Time) + βy * (Main terms of each variable)
To ascertain the clinical implications of candidate proteins, we conducted receiver operating characteristic (ROC) analyses to calculate the area under the curve (AUC) values of Aβ pathological positivity incidents (both PET and CSF). Additionally, we employed logistic regression analyses to calculate odds ratios (ORs) and 95% confidence interval (CI) of each hub protein. Moreover, we performed Cox hazard regression analysis to derive hazard ratios (HRs) and 95% CI for predicting the risk of Aβ-PET pathological progression in Aβ-PET negative participants. Then Aβ-PET negative individuals were stratified into high-level and low-level groups based on the median level of the protein, and their risks of Aβ-PET positivity conversion were compared using log-rank tests and illustrated using Kaplan-Meier plots. For logistic regression and Cox hazard regression analyses, the protein levels were preprocessed through log transformation and z-score normalization.
We then evaluated whether the incorporation of the individual candidate protein could enhance the diagnostic ability of Aβ42 in Aβ-PET pathological discrimination, we partitioned 70% of the total sample as the training set and the remaining 30% as the testing set. Within the training set, logistic regression was used to determine the coefficients of Aβ42 and the other protein, which had been preprocessed through log transformation and z-score normalization, to create a composite protein profile. The discriminative abilities of the combined protein profile and Aβ42 were compared directly using DeLong’s test to identify statistically significant differences in AUC values (P < 0.05). Moreover, decision curve analysis (DCA) was employed to assess the standardized net benefit of the combined protein profile by weighing the true-positive and false-positive predictions across various threshold probabilities for the pathological outcome.
All statistical analyses were performed using R version 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria). Notably, for exploratory investigations, we refrained from conducting multiple comparison corrections.

 

Results

Cohort Characteristics

Our study encompassed 474 non-demented individuals, as detailed in Table 1. The Aβ-PET positive group displayed a greater prevalence of MCI individuals and an increased frequency of APOE ε4 carriers compared to the negative group (P < 0.001). Additionally, individuals in the Aβ-PET positive group were characterized by an older age profile than those in the negative group (P < 0.001). As anticipated, the cognitive status of Aβ-PET positive individuals was notably poorer than that of negative individuals (P < 0.001). Differences were observed in 26 of the 54 proteins between Aβ-PET positive and negative participants. Detailed summaries of protein levels across different groups can be found in Supplementary Table 1.

For continuous variables, the values were presented as mean (standard deviation, SD). Demographic comparisons were conducted using Chi-square tests for categorical variables, the Mann-Whitney U test for the continuous variable with a non-normal distribution (years of education), and t-test for the continuous variable with a normal distribution (age). Disparities in cognitive scores were assessed using linear models with age, gender, APOE ε4 allele status, and years of education as covariates. MMSE: Mini-Mental State Examination, mPACCdigit: Modified Preclinical Alzheimer Cognitive Composite incorporating the Digit Symbol Substitution Test, mPACCtrailsB: Modified Preclinical Alzheimer Cognitive Composite incorporating the Trail-Making Test Part B.

 

Candidate Hub Proteins Identification

Our analysis revealed 20 biomarkers that showed significant associations with Aβ-PET positivity, along with 3 biomarkers exhibiting strong correlations with CSF-Aβ positivity. Notably, proteins apolipoprotein E (APOE), chitinase-3-like protein 1 (CHI3L1), and SPARC-related modular calcium-binding protein 1 (SMOC1) were the intersecting proteins correlated with both Aβ-PET positivity and CSF-Aβ positivity (Figure 1). Upon further investigation of the levels of the aforementioned proteins across various Aβ pathological statuses, we observed that CHI3L1 and SMOC1 displayed distinctions between the CSF-Aβ negative and Aβ-PET negative group (CSF-PET-) and the CSF-Aβ positive and Aβ-PET negative group (CSF+PET-), as well as between the CSF-PET- and CSF-Aβ positive and Aβ-PET positive group (CSF+PET+). Following statistical adjustments, APOE exhibited a significant difference only between the CSF-PET- and CSF+PET+ groups. However, it did show a tendency towards differentiation between the CSF-PET- and CSF+PET- groups (unadjusted P = 0.03, adjusted P = 0.10) (Supplementary Table 2).

To conduct CSF protein selection, 70% of the total dataset was randomly sampled for LASSO regression, with this process repeating 1000 times. Variables appearing over 500 times were considered robust. For exploratory investigations, we refrained from conducting multiple comparison corrections. This procedure was applied to both Aβ-PET (Orange) and CSF-Aβ statuses (Purple). The outer boundary of the red circle delineated proteins selected more than 500 times.

 

Candidate Hub Proteins Assessment in Predicting Aβ Pathological Status and Forecasting Aβ Pathological Transformation

SMOC1 showed the strongest association with Aβ pathology, displaying AUCs [95%CI] of 0.83 [0.72-0.95] for CSF-Aβ positivity and 0.74 [0.70-0.79] for Aβ-PET positivity. Additionally, APOE demonstrated AUCs [95%CI] of 0.69 [0.54-0.83] for CSF-Aβ positivity and 0.71 [0.66-0.75] for Aβ-PET positivity, while CHI3L1 exhibited AUCs [95%CI] of 0.73 [0.60-0.87] for CSF-Aβ positivity and 0.65 [0.60-0.70] for Aβ-PET positivity (refer to Figures 2A and 2B). Individuals with low levels of these hub proteins displayed a diminishing likelihood of Aβ-PET and CSF-Aβ positivity (as shown in Figures 2C and 2D). ORs [95% CI] for the candidate hub proteins were as follows: 1.61 [1-2.54] for APOE, 3.78 [2.1-7.32] for SMOC1, and 2.66 [1.52-4.87] for CHI3L1 concerning CSF-Aβ positivity, and 2.11 [1.78-2.52] for APOE, 2.5 [2.03-3.12] for SMOC1, and 1.76 [1.44-2.16] for CHI3L1 regarding Aβ-PET positivity.

(A and B) ROC analysis of candidate hub proteins in predicting CSF-Aβ and Aβ-PET positivity incidents. (C and D) Forest plots of ORs and 95% CI of candidate hub proteins in predicting CSF-Aβ and Aβ-PET positivity incidents. For exploratory investigations, we refrained from conducting multiple comparison corrections.

 

We further employed Cox hazard regression analysis to explore whether the hub proteins could precisely predict the longitudinal Aβ-PET pathological conversion among non-demented Aβ-PET negative individuals. The findings indicated that SMOC1 and APOE were significantly linked to the risk of progression to Aβ-PET positivity. Specifically, elevated baseline levels of each protein were indicative of an increased risk of pathological conversion, with HRs [95% CI] of 1.36 [1.03-1.78] for APOE and 1.53 [1.12-2.09] for SMOC1. However, CHI3L1 did not correlate with Aβ-PET positivity transforming risk with HR [95%CI] as 1.15 [0.84-1.57].

Candidate Hub Proteins Assessment in Predicting Longitudinal Cognitive Decline and SUVRs Increase

CHI3L1 and SMOC1 showed significant associations with longitudinal cognitive decline in non-demented individuals who were Aβ-PET positive. Specifically, CHI3L1 demonstrated correlations with longitudinal cognitive decline as indicated by β [S.E.] values of -0.24 [0.08] for MMSE (P = 0.004), -0.11 [0.05] for mPACCdigit (P = 0.039), and -0.14 [0.05] for mPACCtrailsB (P = 0.01). Similarly, SMOC1 exhibited associations with longitudinal cognitive decline with β [S.E.] values of -0.16 [0.06] for MMSE (P = 0.01), -0.10 [0.04] for mPACCdigit (P = 0.01), and -0.11 [0.04] for mPACCtrailsB (P = 0.01) (Supplementary Table 3). Although APOE did not directly correlate with longitudinal cognitive decline, it was the only candidate protein that displayed an association with Aβ42-related cognitive alterations in Aβ-PET positive non-demented individuals. This association was observed through the examination of the interactive relationship between baseline CSF APOE, CSF Aβ42 levels, and follow-up time in longitudinal cognitive scores, resulting in β [S.E.] values of 0.20 [0.08] for MMSE (P = 0.02), 0.14 [0.05] for mPACCdigit (P = 0.006), and 0.15 [0.05] for mPACCtrailsB (P = 0.007) (Supplementary Table 4). Furthermore, SMOC1 emerged as the only CSF protein exhibiting an association with the longitudinal increase of Aβ-PET SUVRs with β [S.E.] value of 0.02 [0.01] (P = 0.01) within Aβ-PET negative cohort (Supplementary Table 3).

Candidate Hub Proteins Assessment in Enhancing the Diagnostic Performance of Aβ42 in Pathological Status

Since all three CSF proteins exhibited the potential to differentiate between Aβ-PET positivity and negativity, we aimed to investigate whether combining Aβ42 with each single hub protein could improve its discriminative capability. The dataset was divided into a training set (70%) for the composite protein profile construction by integrating the proteins into a linear regression equation with careful consideration given to the assignment of variables to their corresponding parameters based on the logistic regression and the rest as the testing set for composite protein assessment. Among the three proteins, only the combination of Aβ42 with SMOC1 demonstrated superior diagnostic accuracy in both training and testing sets compared to Aβ42 alone. In the training set, the combination yielded an AUC of 0.91 [0.88-0.94], surpassing the AUC of Aβ42 alone at 0.86 [0.82-0.91] with a statistically significant difference (DeLong test P < 0.001). This trend was consistent in the testing set, with the combination achieving an AUC of 0.91 [0.86-0.96] compared to Aβ42’s AUC of 0.86 [0.80-0.92] (DeLong test P = 0.008) (refer to Figures 3A and B). Furthermore, DCA curves supported the use of the composite Aβ42 and SMOC1 profile for risk stratification, indicating a higher clinical standardized net benefit across various threshold probabilities compared to using Aβ42 alone, exhibiting stronger potential clinical utility, as illustrated in Figures 3C and D. Although the addition of CHI3L1 to Aβ42 in the training set improved the Aβ42’s diagnostic performance of Aβ-PET pathological status (AUC = 0.91 [0.87-0.94], DeLong test P = 0.001), this enhancement was not sustained in the testing set (AUC = 0.89 [0.83-0.94], DeLong test P = 0.08).
Considering the binding of SMOC1 and Aβ42 can enhance the diagnostic performance of Aβ42, we used this as a starting point to explore the addition of the hub variables CHI3L1 and APOE in an attempt to further optimize the diagnostic model. However, the results indicated that neither the inclusion of CHI3L1 nor APOE alone nor their combination led to further optimization of the diagnostic model (Supplementary Figure 1).

(A and B) ROC analysis of Aβ42 alone and the combination of Aβ42 with APOE, CHI3L1, or SMOC1 in predicting Aβ-PET positivity incident. The discriminative abilities of the combined protein profile and Aβ42 were compared directly using DeLong’s test to identify statistically significant differences in AUC values (P < 0.05) (C and D) DCA curves compared the standardized net benefit between Aβ42 alone and the combination of Aβ42 with APOE, CHI3L1, or SMOC1 in predicting Aβ-PET positivity incident.

 

Clinical Applications of the Combination of Aβ42 and SMOC1 in Predicting Pathological Progression

Given the combination of Aβ42 and SMOC1 offered better discriminative power than Aβ42 alone, we conducted an in-depth analysis of the relationship between the combined protein (Combined protein = -0.84 + 1.10 * SMOC1 – 1.85 * Aβ42) and the progression of pathology over time, utilizing longitudinal changes in whole-brain amyloid deposition as a marker. The coefficients for SMOC1 and Aβ42 in the linear combined protein formula were determined through logistic regression analysis on the training set. Our finding indicated the potential for precise risk stratification of non-demented populations by leveraging the median levels of combined proteins.
Firstly, the combined protein could accurately predict the longitudinal FBP-PET SUVRs increase within Aβ-PET negative cohort, where higher baseline levels of the combined protein were associated with more significant amyloid pathological progression over time (β [S.E.] = 0.06 [0.008], P < 0.001). Stratifying Aβ-PET negative individuals into high and low-level groups based on the median level of the combined protein profile showed that those with higher levels exhibited a more pronounced increase in amyloid deposition (β [S.E.] = 0.06 [0.01], P < 0.001) (Figure 4A).
Secondly, Cox hazard regression demonstrated a significant association between both continuous and categorical representations of the combined protein and the risk of progression to amyloid pathological positivity. Specifically, elevated baseline levels of the combined protein were suggestive of an increased likelihood of amyloid deposition onset (HR [95% CI] = 1.68 [1.41-1.99]). This association persisted after stratifying Aβ-PET negative participants based on the median value of the combined protein, with the high-level group showing a substantially elevated Aβ-PET positivity-conversion risk (HR [95% CI] = 3.75 [2.00-7.04]) as depicted in Figure 4B.
Besides, the combined protein demonstrated robust predictive performance over time, with AUC values of 0.74, 0.75, 0.81, and 0.85 for 3-year, 5-year, 7-year, and 9-year Aβ-PET positivity incidents, respectively, underscoring its potential utility as a prognostic marker (Figure 4C).

(A) Differences in longitudinal FBP-PET SUVRs increase between groups stratified by the median levels of the Aβ42 and SMOC1 combination. (B) Kaplan-Meier plot illustrating the difference of Aβ-PET positivity-conversion risk between groups stratifying based on the median levels of the Aβ42 and SMOC1 combination. (C) AUCs of ROC curves to assess the prognostic accuracy of Aβ42 combined SMOC1 for Aβ-PET positivity conversion at 3-year, 5-year, 7-year, and 9-year intervals.

 

Discussion

Some studies have already reported a strong association between SMOC1 and AD (15, 19, 20). Furthermore, other findings highlighted the direct link between SMOC1 and Aβ pathology (21). However, there were fewer studies systematically investigating the specific diagnostic potential of SMOC1 in both early and late amyloid pathological stages and its performance as an auxiliary diagnostic marker for Aβ42. In our investigation, among 51 CSF proteins, we revealed the good discriminative ability of individual CSF SMOC1 in detecting both the early signs of Aβ dyshomeostasis (CSF-Aβ positivity) and late-stage Aβ aggregation measured by PET (Aβ-PET positivity), with respective AUCs of 0.83 and 0.74. Furthermore, SMOC1 emerged as the only CSF protein exhibiting an association with the longitudinal increase of Aβ-PET SUVRs. Additionally, individual assessment of CSF SMOC1 effectively predicted the conversion to Aβ-PET positivity from Aβ-PET negativity in non-demented individuals. Moreover, the addition of SMOC1 amplified the diagnostic accuracy of Aβ42 in amyloid pathological status from AUCs of 0.86 to 0.91 in both training and testing sets, while the other two candidate hub proteins did not achieve this. DCA curves further supported using Aβ42 combined SMOC1 to make decisions for anti-amyloid treatment could earn more clinical benefits in both training and testing sets. Furthermore, upon stratifying all the non-demented participants into CN and MCI cohorts, the positive effect of SMOC1 on improving the diagnostic accuracy of Aβ42 persisted (see Supplementary Figure 2). These findings collectively emphasized the potential of SMOC1 as a valuable biomarker for amyloid pathology. Additionally, the observations of the absent CSF SMOC1 level alterations in patients with non-AD neurodegenerative diseases in other studies accentuated the specificity of SMOC1 as a potential biomarker for AD (19, 22). However, despite these observations, the precise role of SMOC1 in amyloid pathology remains enigmatic. Building on the established roles of SMOC1 in glucose homeostasis and angiogenesis from previous studies (23, 24), aberrant SMOC1 levels might reflect the pathological status of cerebral neurovascular, which will further impede timely Aβ plaque clearance (25-27). The combination of Aβ42 and SMOC1 might address both the increase in Aβ42 aggregation and the decrease in Aβ plaque clearance so that it exhibited better clinical performance compared to Aβ42 alone.
Although we found potential discriminative power of APOE and CHI3L1 between CSF-Aβ negative/positive groups and Aβ-PET negative/positive groups, our subsequent analyses did not reveal the correlation between APOE, CHI3L1, and the longitudinal increase in Aβ-PET SUVRs. We also did not observe CHI3L1’s ability to predict the risk of Aβ-PET positivity conversion. For APOE, previous studies have demonstrated the protective functions of APOE2 and APOE3 protein isoforms, as well as the detrimental impact of APOE4 protein isoform, concerning Aβ clearance, aggregation, and amyloid plaque formation (28). In our study, the lack of correlation between APOE protein and longitudinal SUVRs changes may be attributed to the possibility that the imbalance between protective and harmful APOE isoforms may have a greater impact on Aβ pathology progression than the overall level of APOE.
Regarding CHI3L1, although some studies supported our findings that there were differences in CHI3L1 levels between CSF-Aβ negative/positive groups and Aβ-PET negative/positive groups (29, 30), clinical research suggested that CHI3L1 was more associated with tau pathology rather than Aβ pathology (31, 32). Our study also found that CSF CHI3L1 could not predict the risk of Aβ-PET positivity conversion or the progression of longitudinal Aβ-PET SUVRs. On the other hand, another astrocyte marker, GFAP, was also found to exhibit differences between the aforementioned groups in other studies (33, 34) but was closely linked to Aβ pathology rather than tau pathology (31, 33). GFAP is a structural hallmark of astrocytes, while CHI3L1 is only expressed in certain subtypes of astrocytes (32). We speculated that this may be due to the activation of astrocytes following the transition of Aβ pathological status in the brain (early Aβ dyshomeostasis or late amyloid plaque deposition), where the activated subtypes were involved in not only the Aβ pathology-related subtype but also downstream tau pathology-related subtype. CHI3L1 may serve as an important marker for tau pathology-related astrocyte subtype, acting as a crucial link between Aβ pathology and tau pathology.
Given the inherent limitations in the sensitivity and specificity of conventional clinical assessments, the biomarker-based approaches emerge as crucial for identifying abnormal amyloid pathology. Furthermore, an essential component for the effective clinical implementation of anti-amyloid therapy is the prompt identification of individuals with Aβ-pathological positivity before or during the onset of limited irreversible neurodegeneration. Notably, given the heterogeneity of AD (11, 35, 36), there exist AD patients whose primary pathological presentation does not center on amyloid deposition. These individuals should not be hastily recommended for anti-amyloid therapy, emphasizing the importance of accurately diagnosing Aβ pathologic status during the early phases of disease advancement.
The identification of SMOC1 as a potential biomarker shows promise for the development of a reliable diagnostic tool for early detection of amyloid pathology. The synergistic effect of CSF SMOC1 and Aβ42 demonstrated higher accuracy and predictive power than Aβ42 alone. With the aging population on the rise and the expected increase in approved AD treatments requiring initial confirmation of underlying AD pathology through diagnostic testing (2, 37, 38), the identification and clinical assessment of such a diagnostic adjunct factor as SMOC1 will be more helpful in the future.
The present study, however, is subject to several limitations. One such limitation pertains to the restricted number of proteins selected for analysis which is due to a strategic choice made to ensure sample retention within the constraints of available resources. However, it is still confined by a relatively small sample size and the absence of an external validation dataset, potentially introducing sample bias that could compromise the generalizability and robustness of the findings. Additionally, statistical corrections were not conducted in our study, which could affect the significance of certain results and increase the risk of false positives, potentially impacting result stability. Nevertheless, these preliminary findings offer valuable insights and warrant further investigation. To bolster the credibility and applicability of the study, future research efforts could focus on enlarging the sample size and integrating an external validation dataset to validate and reinforce our research findings.

 

Conclusion

Analyzing 51 CSF proteins from 474 non-demented participants, APOE, CHI3L1, and SMOC1 emerged as potential candidates for distinguishing Aβ pathological status. Among these, SMOC1 stood out for its comparatively suitable ability to discriminate Aβ pathological status, predict Aβ-PET trajectory, and enhance Aβ42’s diagnostic performance. The study highlights SMOC1’s potential as Aβ42’s auxiliary diagnostic factor for Aβ abnormalities detection. The combination of Aβ42 and SMOC1 offers insights for AD pathological diagnosis and anti-amyloid management. Overall, our study contributes to the growing body of knowledge in the realm of AD biomarkers and provides new avenues for the potential pathogenesis of amyloid pathology.

 

Availability of data and materials: The dataset supporting the conclusions of this article is available in the ADNI repository (ida.loni.usc.edu). Derived data can be obtained from the corresponding author upon request by any qualified investigator subject to a data use agreement.

Ethics approval and consent to participate: All procedures performed in studies involving human participants were conducted following the ethical standards of the institutional and/or national research committee and with the principles of the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. Formed written consent was obtained from all participants at each site of ADNI.

Acknowledgments: Not applicable.

Funding: Not applicable.

Consent for publication: Not applicable.

Competing interests: The authors declare that they have no competing interests.

Author contributions: Data curation, data analysis, and writing original draft, Fanwei Meng; Writing, review, and editing, Xi Zhang. All authors reviewed and approved the final manuscript.

 

SUPPLEMENTARY MATERIAL1

 

SUPPLEMENTARY MATERIAL2

 

References

1. 2023 Alzheimer’s disease facts and figures. Alzheimers Dement. 2023/03/16 2023;19(4):1598-695. doi: 10.1002/alz.13016.
2. Nandi A, Counts N, Chen S, Seligman B, Tortorice D, Vigo D, et al. Global and regional projections of the economic burden of Alzheimer’s disease and related dementias from 2019 to 2050: A value of statistical life approach. EClinicalMedicine. 2022/07/29 2022;51:101580. doi: 10.1016/j.eclinm.2022.101580.
3. Dubois B, Hampel H, Feldman HH, Scheltens P, Aisen P, Andrieu S, et al. Preclinical Alzheimer’s disease: Definition, natural history, and diagnostic criteria. Alzheimers Dement. 2016/03/26 2016;12(3):292-323. doi: 10.1016/j.jalz.2016.02.002.
4. Jack CR, Jr., Holtzman DM. Biomarker modeling of Alzheimer’s disease. Neuron. 2013/12/24 2013;80(6):1347-58. doi: 10.1016/j.neuron.2013.12.003.
5. Jack CR, Jr., Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, et al. Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers. The Lancet Neurology. 2013/01/22 2013;12(2):207-16. doi: 10.1016/s1474-4422(12)70291-0.
6. van der Kant R, Goldstein LSB, Ossenkoppele R. Amyloid-β-independent regulators of tau pathology in Alzheimer disease. Nature reviews Neuroscience. 2019/11/30 2020;21(1):21-35. doi: 10.1038/s41583-019-0240-3.
7. Tan MS, Ji X, Li JQ, Xu W, Wang HF, Tan CC, et al. Longitudinal trajectories of Alzheimer’s ATN biomarkers in elderly persons without dementia. Alzheimer’s research & therapy. 2020/05/13 2020;12(1):55. doi: 10.1186/s13195-020-00621-6.
8. Ahmad F, Karan A, Sharma R, Sharma NS, Sundar V, Jayaraj R, et al. Evolving therapeutic interventions for the management and treatment of Alzheimer’s disease. Ageing Res Rev. 2024/02/17 2024;95:102229. doi: 10.1016/j.arr.2024.102229.
9. Seppälä TT, Nerg O, Koivisto AM, Rummukainen J, Puli L, Zetterberg H, et al. CSF biomarkers for Alzheimer disease correlate with cortical brain biopsy findings. Neurology. 2012/04/21 2012;78(20):1568-75. doi: 10.1212/WNL.0b013e3182563bd0.
10. Wolk DA, Grachev ID, Buckley C, Kazi H, Grady MS, Trojanowski JQ, et al. Association between in vivo fluorine 18-labeled flutemetamol amyloid positron emission tomography imaging and in vivo cerebral cortical histopathology. Archives of neurology. 2011/07/13 2011;68(11):1398-403. doi: 10.1001/archneurol.2011.153.
11. Hansson O. Biomarkers for neurodegenerative diseases. Nature medicine. 2021/06/05 2021;27(6):954-63. doi: 10.1038/s41591-021-01382-x.
12. Popp J, Oikonomidi A, Tautvydaitė D, Dayon L, Bacher M, Migliavacca E, et al. Markers of neuroinflammation associated with Alzheimer’s disease pathology in older adults. Brain, behavior, and immunity. 2017/02/06 2017;62:203-11. doi: 10.1016/j.bbi.2017.01.020.
13. Dhiman K, Blennow K, Zetterberg H, Martins RN, Gupta VB. Cerebrospinal fluid biomarkers for understanding multiple aspects of Alzheimer’s disease pathogenesis. Cellular and molecular life sciences : CMLS. 2019/02/17 2019;76(10):1833-63. doi: 10.1007/s00018-019-03040-5.
14. Nordengen K, Kirsebom BE, Henjum K, Selnes P, Gísladóttir B, Wettergreen M, et al. Glial activation and inflammation along the Alzheimer’s disease continuum. Journal of neuroinflammation. 2019/02/23 2019;16(1):46. doi: 10.1186/s12974-019-1399-2.
15. Haque R, Watson CM, Liu J, Carter EK, Duong DM, Lah JJ, et al. A protein panel in cerebrospinal fluid for diagnostic and predictive assessment of Alzheimer’s disease. Science translational medicine. 2023/09/06 2023;15(712):eadg4122. doi: 10.1126/scitranslmed.adg4122.
16. Leuzy A, Mattsson-Carlgren N, Cullen NC, Stomrud E, Palmqvist S, La Joie R, et al. Robustness of CSF Aβ42/40 and Aβ42/P-tau181 measured using fully automated immunoassays to detect AD-related outcomes. Alzheimers Dement. 2023/01/22 2023;19(7):2994-3004. doi: 10.1002/alz.12897.
17. van der Kroft G, Wee L, Rensen SS, Brecheisen R, van Dijk DPJ, Eickhoff R, et al. Identifying radiomics signatures in body composition imaging for the prediction of outcome following pancreatic cancer resection. Frontiers in oncology. 2023/08/28 2023;13:1062937. doi: 10.3389/fonc.2023.1062937.
18. Jiang W, Wu R, Yang T, Yu S, Xing W. Profiling regulatory T lymphocytes within the tumor microenvironment of breast cancer via radiomics. Cancer medicine. 2023/12/12 2023;12(24):21861-72. doi: 10.1002/cam4.6757.
19. Dammer EB, Ping L, Duong DM, Modeste ES, Seyfried NT, Lah JJ, et al. Multi-platform proteomic analysis of Alzheimer’s disease cerebrospinal fluid and plasma reveals network biomarkers associated with proteostasis and the matrisome. Alzheimer’s research & therapy. 2022/11/18 2022;14(1):174. doi: 10.1186/s13195-022-01113-5.
20. Li Y, Chen Z, Wang Q, Lv X, Cheng Z, Wu Y, et al. Identification of hub proteins in cerebrospinal fluid as potential biomarkers of Alzheimer’s disease by integrated bioinformatics. Journal of neurology. 2022/11/18 2023;270(3):1487-500. doi: 10.1007/s00415-022-11476-2.
21. Bai B, Wang X, Li Y, Chen PC, Yu K, Dey KK, et al. Deep Multilayer Brain Proteomics Identifies Molecular Networks in Alzheimer’s Disease Progression. Neuron. 2020/01/14 2020;105(6):975-91.e7. doi: 10.1016/j.neuron.2019.12.015.
22. van der Ende EL, In ‘t Veld S, Hanskamp I, van der Lee S, Dijkstra JIR, Hok AHYS, et al. CSF proteomics in autosomal dominant Alzheimer’s disease highlights parallels with sporadic disease. Brain : a journal of neurology. 2023/06/23 2023;146(11):4495-507. doi: 10.1093/brain/awad213.
23. Montgomery MK, Bayliss J, Devereux C, Bezawork-Geleta A, Roberts D, Huang C, et al. SMOC1 is a glucose-responsive hepatokine and therapeutic target for glycemic control. Science translational medicine. 2020/09/04 2020;12(559). doi: 10.1126/scitranslmed.aaz8048.
24. Awwad K, Hu J, Shi L, Mangels N, Abdel Malik R, Zippel N, et al. Role of secreted modular calcium-binding protein 1 (SMOC1) in transforming growth factor β signalling and angiogenesis. Cardiovascular research. 2015/03/10 2015;106(2):284-94. doi: 10.1093/cvr/cvv098.
25. Li T, Li D, Wei Q, Shi M, Xiang J, Gao R, et al. Dissecting the neurovascular unit in physiology and Alzheimer’s disease: Functions, imaging tools and genetic mouse models. Neurobiology of disease. 2023/04/07 2023;181:106114. doi: 10.1016/j.nbd.2023.106114.
26. Eisenmenger LB, Peret A, Famakin BM, Spahic A, Roberts GS, Bockholt JH, et al. Vascular contributions to Alzheimer’s disease. Translational research : the journal of laboratory and clinical medicine. 2022/12/19 2023;254:41-53. doi: 10.1016/j.trsl.2022.12.003.
27. Fisher RA, Miners JS, Love S. Pathological changes within the cerebral vasculature in Alzheimer’s disease: New perspectives. Brain pathology (Zurich, Switzerland). 2022/03/16 2022;32(6):e13061. doi: 10.1111/bpa.13061.
28. Yamazaki Y, Zhao N, Caulfield TR, Liu CC, Bu G. Apolipoprotein E and Alzheimer disease: pathobiology and targeting strategies. Nature reviews Neurology. 2019/08/02 2019;15(9):501-18. doi: 10.1038/s41582-019-0228-7.
29. Palmqvist S, Insel PS, Stomrud E, Janelidze S, Zetterberg H, Brix B, et al. Cerebrospinal fluid and plasma biomarker trajectories with increasing amyloid deposition in Alzheimer’s disease. EMBO molecular medicine. 2019/11/12 2019;11(12):e11170. doi: 10.15252/emmm.201911170.
30. Prins S, de Kam ML, Teunissen CE, Groeneveld GJ. Inflammatory plasma biomarkers in subjects with preclinical Alzheimer’s disease. Alzheimer’s research & therapy. 2022/08/04 2022;14(1):106. doi: 10.1186/s13195-022-01051-2.
31. Ferrari-Souza JP, Ferreira PCL, Bellaver B, Tissot C, Wang YT, Leffa DT, et al. Astrocyte biomarker signatures of amyloid-β and tau pathologies in Alzheimer’s disease. Molecular psychiatry. 2022/08/11 2022;27(11):4781-9. doi: 10.1038/s41380-022-01716-2.
32. Querol-Vilaseca M, Colom-Cadena M, Pegueroles J, San Martín-Paniello C, Clarimon J, Belbin O, et al. YKL-40 (Chitinase 3-like I) is expressed in a subset of astrocytes in Alzheimer’s disease and other tauopathies. Journal of neuroinflammation. 2017/06/11 2017;14(1):118. doi: 10.1186/s12974-017-0893-7.
33. Simrén J, Weninger H, Brum WS, Khalil S, Benedet AL, Blennow K, et al. Differences between blood and cerebrospinal fluid glial fibrillary Acidic protein levels: The effect of sample stability. Alzheimers Dement. 2022/09/15 2022;18(10):1988-92. doi: 10.1002/alz.12806.
34. Guo Y, Shen XN, Wang HF, Chen SD, Zhang YR, Chen SF, et al. The dynamics of plasma biomarkers across the Alzheimer’s continuum. Alzheimer’s research & therapy. 2023/02/09 2023;15(1):31. doi: 10.1186/s13195-023-01174-0.
35. Young-Pearse TL, Lee H, Hsieh YC, Chou V, Selkoe DJ. Moving beyond amyloid and tau to capture the biological heterogeneity of Alzheimer’s disease. Trends in neurosciences. 2023/04/06 2023;46(6):426-44. doi: 10.1016/j.tins.2023.03.005.
36. Iturria-Medina Y, Adewale Q, Khan AF, Ducharme S, Rosa-Neto P, O’Donnell K, et al. Unified epigenomic, transcriptomic, proteomic, and metabolomic taxonomy of Alzheimer’s disease progression and heterogeneity. Science advances. 2022/11/19 2022;8(46):eabo6764. doi: 10.1126/sciadv.abo6764.
37. Yadollahikhales G, Rojas JC. Anti-Amyloid Immunotherapies for Alzheimer’s Disease: A 2023 Clinical Update. Neurotherapeutics : the journal of the American Society for Experimental NeuroTherapeutics. 2023/07/25 2023;20(4):914-31. doi: 10.1007/s13311-023-01405-0.
38. Imbimbo BP, Ippati S, Watling M, Imbimbo C. Role of monomeric amyloid-β in cognitive performance in Alzheimer’s disease: Insights from clinical trials with secretase inhibitors and monoclonal antibodies. Pharmacological research. 2023/01/01 2023;187:106631. doi: 10.1016/j.phrs.2022.106631.

© Serdi 2024