Alzheimer's disease and inflammatory biomarkers positively correlate in plasma in the UK‐ADRC cohort

Abstract

INTRODUCTION

Protein‐based plasma assays provide hope for improving accessibility and specificity of molecular diagnostics to diagnose dementia.

METHODS

Plasma was obtained from participants (N = 837) in our community‐based University of Kentucky Alzheimer's Disease Research Center cohort. We evaluated six Alzheimer's disease (AD)‐ and neurodegeneration‐related (Aβ40, Aβ42, Aβ42/40, p‐tau181, total tau, and NfLight) and five inflammatory biomarkers (TNF𝛼, IL6, IL8, IL10, and GFAP) using the SIMOA‐based protein assay platform. Statistics were performed to assess correlations.

RESULTS

Our large cohort reflects previous plasma biomarker findings. Relationships between biomarkers to understand AD–inflammatory biomarker correlations showed significant associations between AD and inflammatory biomarkers suggesting peripheral inflammatory interactions with increasing AD pathology. Biomarker associations parsed out by clinical diagnosis (normal, MCI, and dementia) reveal changes in strength of the correlations across the cognitive continuum.

DISCUSSION

Unique AD–inflammatory biomarker correlations in a community‐based cohort reveal a new avenue for utilizing plasma‐based biomarkers in the assessment of AD and related dementias.

Highlights

• Large community cohorts studying sex, age, and APOE genotype effects on biomarkers are few.

• It is unknown how biomarker–biomarker associations vary through aging and dementia.

• Six AD (Aβ40, Aβ42, Aβ42/40, p‐tau181, total tau, and NfLight) and five inflammatory biomarkers (TNFα, IL6, IL8, IL10, and GFAP) were used to examine associations between biomarkers.

• Plasma biomarkers suggesting increasing cerebral AD pathology corresponded to increases in peripheral inflammatory markers, both pro‐inflammatory and anti‐inflammatory.

• Strength of correlations, between pairs of classic AD and inflammatory plasma biomarker, changes throughout cognitive progression to dementia.

1. BACKGROUND

As the prevalence of amnestic dementia is doubling every 20 years, researchers are racing to better predict, diagnose, and evaluate dementia prognosis. ¹ Recently there have been multiple advances in diagnostic technology including (1) genetic testing for specific genes and mutations known to increase dementia likelihood (ie, apolipoprotein E [APOE], APP, PSEN1), (2) positron emission tomography (PET) imaging for the presence of hallmark pathologies such as Aβ and tau, (3) magnetic resonance imaging (MRI) sequences showing advanced structural damage to vulnerable brain regions (hippocampal sclerosis, cortical thinning), and lastly (4) cerebral spinal fluid (CSF)‐ and plasma‐based protein markers based on biological pathways in the brain and body that change in response to Alzheimer's disease (AD)‐associated pathologies. ² , ³ , ⁴ Each of these techniques have pros and cons related to reliability, time, cost, and invasiveness, with plasma sampling being cheaper, quicker, and less invasive compared to CSF or neuroimaging, but also somewhat less sensitive and specific.

Plasma biomarkers for AD and related pathologies have advanced, which increases diagnostic precision. AD biomarkers for amyloid beta (Aβ) plaque deposition via Aβ40 and Aβ42 isoform levels are increasingly well characterized, leading to the utilization of the Aβ42/Aβ40 ratio as another standardized method of gauging Aβ42 toxic buildup in the brain. ⁵ , ⁶ Contrary to many biomarkers, it is the decrease in Aβ40 and Aβ42 levels in the plasma that signify the increased vascular and parenchymal Aβ plaques deposited in the brain. ⁷ Total tau protein has also been used as a marker for distinguishing increased levels of tau in the brain, and more importantly, variations of phosphorylated tau at particular residues, such as p‐tau181, p‐tau217, and p‐tau231, have been found in the blood of patients with AD. ⁸ , ⁹ , ¹⁰ , ¹¹ However, p‐tau181 in plasma correlates better with Aβ pathology and less with tau pathology in the brain. ¹² With neurodegeneration being the ultimate event that scientists are aiming to prevent, evaluating neurofilament light (NfLight), a protein released into the body after axonal injury and cell death. There are many other important biomarkers for AD that are gaining traction; however, this paper focuses on six frequently used plasma AD biomarkers: Aβ40, Aβ42, Aβ42/Aβ40 ratio, total tau, p‐tau181, and NfLight (Figure 1A,B).

FIGURE 1.

UKY‐ADRC demographics and plasma biomarkers of interest. (A) Schematic of cell type and pathology of origin for the AD and inflammatory biomarkers. (B) Plasma biomarker schematic showing transfer of protein signal from brain to CSF, and categories of AD biomarkers and inflammatory biomarkers. (C) Flow chart of participant selection. Participants were selected based on having a clinical diagnosis within 370 days of their first blood draw, plasma run on the same machines with same settings, excluding participants under 60 years old. Aβ, amyloid beta; AD, Alzheimer's disease; GFAP, glial fibrillary acidic protein; IL, interleukin; NfLight, neurofilament light; TNFα, tumor necrosis factor alpha; UKY‐ADRC, University of Kentucky Alzheimer's Disease Research Center.

In addition to misfolded proteins, brain inflammation also plays a critical role in AD pathogenesis. This is supported strongly by the many risk genes that also modulate inflammatory signaling. Since there are both pro‐inflammatory (tumor necrosis factor alpha [TNFα], interleukin 6 [IL6], IL8) and anti‐inflammatory (IL10) cytokines released in AD, it is imperative to establish a broad picture of inflammatory signaling in AD patient plasma. ⁸ , ¹³ We have chosen five inflammatory biomarkers (TNFα, IL6, IL8, IL10, and glial fibrillary acidic protein [GFAP]) that span pro‐ and anti‐inflammatory signals, providing insight into the inflammatory profile in our University of Kentucky ‐ Alzheimer's Disease Research Center (UKY‐ADRC) cohort (Figure 1A,B). Plasma GFAP is elevated in people with preclinical Aβ pathology and AD diagnoses. ¹⁴ , ¹⁵ However, there is still controversy over use of inflammatory biomarkers, as many studies have shown conflicting results. ¹⁶ This discordance could be due to sample size, non‐steroidal anti‐inflammatory drug use, ethnicity, technical factors such as blood draw times, or exclusivity to AD participants instead of all dementias.

While hallmark AD and inflammatory plasma biomarkers are now a focus of the field, it is not known how these biomarkers correlate to 1) key risk demographics and 2) other plasma biomarkers in a community‐based cohort that recruits cognitively normal individuals and follows them longitudinally such as the UKY‐ADRC cohort. It is critical moving forward in plasma biomarker technology to understand the natural correlation between plasma biomarkers and sex, age, clinical diagnosis, as well as other plasma biomarkers, in an all‐encompassing population demographic. This will be key to compare back to, and subset correlations to determine pathology and dementia causes.

2. METHODS

2.1. Participant selection and plasma collection

Participants in this study were enrolled in the University of Kentucky ‐ Alzheimer's Disease Research Center (UKY‐ADRC) cohort between 2012 and 2022 (N = 1302 initially included). Recruitment and visit procedures have been described previously. ¹⁷ All participants signed and consented for their participation and data usage and all protocols have been approved by the University of Kentucky Institutional Review Board. Participants agreed to annual visits to examine physical attributes, cognition, neurological and psychiatric status, medical history, and for blood draws.

Participants with blood draws included 862 unique persons, with 851 blood samples run on the SIMOA platform. Further, we excluded participants under the age of 60 (N = 841). Utilizing data from the UKY‐ADRC cohort, we chose to examine plasma biomarker correlations cross‐sectionally, spanning both sexes and multiple cognitive stages (normal, mild cognitive impairment [MCI], impaired [other], and dementia). Clinical diagnosis was assigned by group consensus according to National Altzheimer Coordinating Center protocols. ¹⁸ , ¹⁹ , ²⁰ Briefly, participants underwent several neuropsychological tests, including but not limited to the following: the Mini‐Mental State Examination, Montreal Cognitive Assessment, animal naming test, and Craft Story 21 Recall tests. ¹⁸ , ²¹ , ²² , ²³ A group comprised of neurologists and neuropsychologists arrived at a consensus diagnosis (“normal,” “MCI,” “Impaired other,” “demented”) for each participant visit. For this analysis, an Impaired other cognitive status was grouped into MCI, as impaired is often used to designate MCI if it is the first visit. Four participants were excluded due to missing clinical diagnosis (N = 837) (Figure 1C). Of note, this cohort is not exclusively Alzheimer's disease participants, as previous autopsy data of our participants have shown ranges of dementia‐inducing pathology. ²⁴ , ²⁵ , ²⁶ Participants in this cohort may have other neurodegenerative comorbidities that may contribute to systemic inflammation.

2.2. Plasma analysis

Blood was collected at each clinic visit by venous puncture using 10‐mL K2 EDTA tubes. The tube was immediately inverted 5 to 10 times, then was spun for 10 min at 2000 × g and plasma was collected in 1‐mL aliquots and stored at −80°C. When ready for quantification, plasma was thawed on ice and centrifuged at 4°C for 10 min at 21,000 × g. Samples were run on the Quanterix Simoa HD‐X platform in duplicate using the same settings as previously described (Table S1). ²⁷

2.3. Statistical analysis

All biomarker quantification values were log transformed and visually inspected for approximate normal distributions in R (v4.1.2). Welch's two‐sample t test was used to determine whether there were differences between the sexes. Analysis of variance (ANOVA) was used to determine mean differences between age groups, APOE status, and cognitive status. Tukey HSD post hoc analyses were used after ANOVAs (dependent variable = biomarker of interest, independent variables = age group, APOE status, cognitive status) to determine the specific groups that were different. The aforementioned tests were completed in R and considered significant if p < 0.05. Biomarker values were scaled, and Pearson correlations for linear association analysis and nominal p‐values using the rcorr command in the Hmisc package (v5.1) were used to understand the correlation relationships between biomarkers. Biomarker heatmaps were all made using data from rcorr. Individual relationships were graphed and Pearson correlations and p‐values calculated using the ggscatter (ggplot2 v3.4.2) and stat_cor (method = “pearson,” v0.6.0) commands. Both ggscatter and rcorr produced identical results. Plasma correlation matrices for each cognitive status were computed by subsetting data based on cognitive status and then computing the correlations and p‐values from rcorr.

RESEARCH IN CONTEXT

1. Systematic review: The authors of this paper have performed extensive literature searches through both published and preprint databases. While much has been published on these biomarkers in various forms of dementia, we have included only a handful of the relevant citations in our work.

2. Interpretation: Our work utilizes a large community‐based cohort to validate and observe associations between classic AD and inflammatory plasma biomarkers. This work sets the stage for utilizing multiple biomarkers to evaluate the biology between the periphery and the brain in dementia, specifically teasing out the relationships between AD biomarkers and inflammation.

3. Future directions: We hope our work provides important validation for AD and inflammatory biomarkers as they change throughout a community‐based cohort and dementia stages and urge others to publish further validation studies. In addition, we hope that others will begin to evaluate biomarker‐to‐biomarker associations as a means of inferring biological processes throughout dementia progression in humans.

3. RESULTS

3.1. Study participant demographics

A cross‐sectional cohort of UKY‐ADRC participants was selected based upon having a blood draw and clinical diagnosis between 2012 and 2022, being analyzed on the same Quanterix Simoa machines and settings, and all being over the age of 60 years (N = 837 finally included) (Figure 1A). Demographics of the population can be found in Table 1. Approximately 60.7% of this subpopulation of the UKY‐ADRC cohort was female. The average age at blood draw was 76.43 (± 7.11) years, with ages showing a normal distribution in groups of 10 years from 60 years old to 100 years old. APOE genotypes spanned all combinations with the majority of the population being APOE ε3/ε3 (43.0)% or APOE ε3/ε4 (27.4%). This population included each of the cognitive stages among the aged (60 years or older) adult population, including cognitively normal (66.8%), MCI and impaired (MCI/Impaired) (18.2%), and dementia (15.1%). The majority of patients did not have a record of diabetes or stroke, with 8.8% having an active diabetic status and only 0.7% with awareness of active stroke. Many participants had recorded hypertension (60.8%) and hyperlipidemia (55.6%).

TABLE 1.

UKY‐ADRC cohort demographics.

	Normal		MCI		Demented		Total
	N	%	N	%	N	%	N	%
Sex
Female	369	44.09%	65	7.77%	74	8.84%	508	60.69%
Male	190	22.70%	87	10.39%	52	6.21%	329	39.31%
Age group
60–69	101	12.07%	17	2.03%	20	2.39%	138	16.49%
70–79	296	35.36%	77	9.20%	48	5.73%	421	50.30%
80–89	147	17.56%	49	5.85%	50	5.97%	246	29.39%
90–100	15	1.79%	9	1.08%	8	0.96%	32	3.82%
APOE status
ε2/ε2	2	0.24%	0	0.00%	0	0.00%	2	0.24%
ε2/ε3	66	7.89%	10	1.19%	6	0.72%	82	9.80%
ε2/ε4	19	2.27%	4	0.48%	7	0.84%	30	3.58%
ε3/ε3	274	32.74%	56	6.69%	30	3.58%	360	43.01%
ε3/ε4	136	16.25%	40	4.78%	53	6.33%	229	27.36%
ε4/ε4	11	1.31%	10	1.19%	11	1.31%	32	3.82%
Unknown	51	6.09%	32	3.82%	19	2.27%	102	12.19%
Race
White	485	57.95%	118	14.10%	108	12.90%	711	84.95%
Black/African American	68	8.12%	29	3.46%	16	1.91%	113	13.50%
Asian	4	0.48%	4	0.48%	1	0.12%	9	1.08%
Other, Hispanic	0		1	0.12%	1	0.12%	2	0.24%
American Indian or Alaskan Native	1	0.12%	0	0.00%	0	0.00%	1	0.12%
Unknown	1	0.12%	0	0.00%	0	0.00%	1	0.12%
Active comorbidities
Diabetes	36	4.30%	21	2.51%	17	2.03%	74	8.84%
Stroke	1	0.12%	2	0.24%	3	0.36%	6	0.72%
Hypertension	329	39.31%	100	11.95%	82	9.80%	511	61.05%
Hyperlipidemia	317	37.87%	83	9.92%	68	8.12%	468	55.91%
NSAID use	296	35.36%	74	8.84%	65	7.77%	435	51.97%

Abbreviations: APOE, apolipoprotein E; MCI, mild cognitive impairment; NSAID, non‐steroidal anti‐inflammatory drug; UKY‐ADRC, University of Kentucky Alzheimer's Disease Research Center.

3.2. UKY‐ADRC plasma biomarkers vary by sex, age, APOE genotype, and clinical diagnosis

We first wanted to establish baseline differences that result from the greatest risk factors for dementia: sex, age, and APOE genotype, utilizing our community‐based cohort and not excluding for comorbidities (Table S2, Figure 2).

FIGURE 2.

Sex, age, APOE status, and clinical diagnosis effects across the UKY‐ADRC cohort. (A) Plasma Aβ42 concentration between males and females was significantly different with a higher concentration in males (diff = 0.157, p = 0.03986). (B) Plasma GFAP showed increased levels in females (diff = 0.258, p = 6.6e‐07). (C) Plasma total tau showed a significant difference between ages (ANOVA: p = 4.38e‐05) specifically between 60s to 70s (Tukey: diff = 0.4728, p = 6.8e‐05) and 60s to 80s (Tukey: diff = 0.512, p = 6.41e‐05). (D) Plasma IL8 showed an overall increase with age (ANOVA: p = 3.88e‐08) with increases between 60s to 70s (Tukey: diff = 0.557, p = 3.83e‐04), 60s to 80s (Tukey: diff = 0.799, p = 7.62e‐07), 60s to 90s (Tukey: diff = 1.33, p = 1.13e‐05), and 70s to 90s (Tukey: diff = 0.775, p = 1.52e‐02). (E) Plasma Aβ42/Aβ40 ratio levels were significantly different between APOE genotypes (p = 0.02951). Tukey post‐hoc test revealed a significant difference between ε2/ε3 and ε4/ε4 (Tukey: diff = 0.508 p = 0.0196). (F) Plasma GFAP levels showed significant differences between APOE genotypes (ANOVA: p = 0.0019). Tukey post‐hoc tests revealed a significant increase in APOE ε3/ε4 when compared to APOE ε3/ε3 (Tukey: diff = 0.251, p = 0.000664). (G) Plasma NfLight showed a difference between cognitive status (ANOVA: p = 1.26e‐08) with an increase in dementia compared to normal (Tukey: diff = .561, p −4.99e‐09), and dementia compared to MCI/impaired (Tukey: diff = 0.442, p = 2.34e‐04). (H) Plasma IL10 showed differences between cognitive status (ANOVA: p = 2.41e‐04). Tukey post hoc tests revealed a decrease of IL10 in MCI/impaired plasma from normal (Tukey: diff = 0.377, p = 0.0001347), and from dementia (Tukey: diff = .0.314, p = 0.024). Aβ, amyloid beta; ANOVA, analysis of variance; APOE, apolipoprotein E; GFAP, glial fibrillary acidic protein; IL, interleukin; NfLight, neurofilament light; UKY‐ADRC, University of Kentucky Alzheimer's Disease Research Center.

3.3. Sex

Only Aβ40 and Aβ42 were significantly different between the sexes for AD biomarkers, with plasma levels being greater in males (Aβ40 diff = −0.274, p = 0.0021; Aβ42 diff = −0.157, p = 0.0398) (Figure 2A, Figure S1, Table S2). Interestingly, the Aβ42/Aβ40 ratio did not show a significant effect of sex (Table S2). For inflammatory biomarkers, females had higher levels of GFAP in the plasma (diff =0.258, p = 6.6e‐07) (Figure 2B, Figure S2, Table S2).

3.4. Age

Aβ40, Aβ42, and total tau showed a general decrease with age, signifying a corresponding increase in the brain, while Aβ42/Aβ40 and NfLight increased with age (Figure 2C, Figure S3, Table S2). Specifically, post hoc testing showed a significant reduction in Aβ40, Aβ42, and total tau, as well as an corresponding increase in Aβ42/Aβ40 ratio levels, occurred during the decades of the 60s, 70s, and 80s (Table S2). The neurodegeneration biomarker NfLight significantly increased across all ages, except for the 80s to 90s (Table S2). Inflammatory biomarkers IL6, IL8, IL10, and GFAP all increased with age in the plasma (Figure 2D, Figure S4, Table S2). The elevations in inflammatory biomarkers were seen between most decade comparisons, indicating a continuous increase (Table S2).

3.5. APOE genotype

APOE genotype only affected Aβ42/Aβ40 plasma levels with regards to AD biomarkers, specifically showing a difference between APOE ε2/ε3 and APOE ε4/ε4 (Figure 2E, Figure S5, Table S2). Similarly, only GFAP levels differed by APOE genotype for inflammatory biomarkers, with a statistically significant difference between APOE ε3/ε3 and APOE ε3/ε4 (Figure 2F, Figure S6, Table S2).

3.6. Clinical diagnosis

We also evaluated the differences in plasma biomarkers across cognitive decline stages. Aβ40, Aβ42, and total tau all showed elevated plasma levels in MCI participants compared to plasma from normal cognition participants, with Aβ40 and total tau increased in dementia compared to normal participants as well (Figure S7, Table S2). NfLight was the only AD biomarker that was increased in the dementia participants compared to normal and MCI groups (Figure 2G, Figure S7, Table S2). Inflammatory biomarkers IL8 and IL10 decreased in MCI compared to normal participants; however, IL10 returned back to normal cognition levels in dementia participants (Figure 2H, Figure S8, Table S2). GFAP was significantly elevated in the dementia population compared to both normal and MCI participants (Figure S8, Table S2).

3.7. Inflammatory biomarkers correlate with hallmark AD biomarkers in plasma

Given that there seemed to be multifactorial influences of sex, age, and APOE genotype on hallmark AD and inflammatory plasma biomarkers in isolation, we next sought to evaluate how the AD plasma biomarkers correlated with the inflammatory biomarkers. To assess this, we performed Pearson correlation analyses across each pairing of biomarkers (Figure 3A). As expected, we saw high and significant positive correlation results between plasma biomarkers Aβ40, Aβ42, and total tau (Figure 3B,C, Figure S9). Interestingly, we also noted significant and strong positive correlations between pairings of inflammatory markers with each other, including pro‐ and anti‐ inflammatory pairs (IL6–IL10, IL8–IL10), suggesting a general upregulation of immune system signaling (Figure 3A). Further, we assessed the patterns of correlations between hallmark AD biomarkers, showing either a negative (blue: Aβ40, Aβ42, total tau) or positive association (red: Aβ42/Aβ40 ratio, p‐tau181, NfLight) with the inflammatory biomarkers (Figure 3A, Figures S10–14). IL6, IL8, IL10, and GFAP all negatively correlated with Aβ40, Aβ42, and total tau, with IL8 and IL10 having the greatest association and GFAP having the weakest association (Figure 4, Figures S10–14). Conversely, IL6, IL8, IL10, and GFAP were positively correlated with the Aβ42/Aβ40 ratio, p‐tau181, and NfLight, with the strongest associations between NfLight and these inflammatory biomarkers (Figure 4, Figures S10–14). Interestingly, TNFα positively correlated with most hallmark AD biomarkers (Aβ42, Aβ42/Aβ40 ratio, p‐tau181, total tau, and NfLight) (Figure S10). These results suggest that as hallmark AD plasma biomarkers show evidence of advancing AD pathology, there is a strong corresponding association with inflammatory markers IL6, IL8, IL10, and GFAP.

FIGURE 3.

Biomarker‐to‐biomarker plasma correlation shows association between AD and inflammatory markers. (A) Pearson correlation matrix for AD biomarkers and inflammatory biomarkers. Correlation coefficient, r, is the top number, with p‐value in parenthesis below. Red square coloring indicates positive correlation; blue square coloring indicates negative correlation. Bolded squares indicate significant results. (B). Scatterplot of the correlation of Aβ42 and Aβ40 in participant plasma shows a strong positive correlation, r = 0.79 (p < 2.2e‐16). (C) Scatterplot of the correlation of total tau and Aβ42 shows a strong positive correlation, r = 0.65 (p < 2.2e‐16). Aβ, amyloid beta; AD, Alzheimer's disease; GFAP, glial fibrillary acidic protein; IL, interleukin; NfLight, neurofilament light.

FIGURE 4.

AD and inflammatory biomarkers correlate in plasma. (A‐C) Plasma IL10 negatively correlates with (A) Aβ40, r = −0.47 (p < 2.2e‐16), (B) Aβ42, r = −.36 (p < 2.2e‐16), and (C) total tau, r = 0.46 (p < 2.2e‐16). (D‐F) Plasma IL10 positively correlates with (D) Aβ42/Aβ40 ratio, r = 0.29 (p < 4.4e‐15), (E) p‐tau181, r = .19 (p < 8.5e‐07), and (F) NfLight, r = 0.31 (p < 2.2e‐16). Aβ, amyloid beta; AD, Alzheimer's disease; IL, interleukin; NfLight, neurofilament light.

3.8. Cognitive state modulates the association between AD and inflammatory plasma biomarkers

We investigated whether hallmark AD and inflammatory plasma biomarker correlations were altered by cognitive status. Pearson correlation analysis per each cognitive status revealed changes in strength of some of the AD–inflammatory biomarker pairings (Figure 5A).

FIGURE 5.

Correlation strength between biomarkers differs by clinical diagnosis. (A) The correlation matrix of AD and inflammatory biomarkers broken out by clinical diagnosis: normal (n = 559), MCI (n = 152), and demented (n = 126). Correlation coefficient, r, is the top number, with the p‐value in parentheses below. Red square coloring indicates positive correlation; blue square coloring indicates negative correlation. Bolded squares indicate significant results. (B) Strength of correlation for plasma biomarker pairs that show increasing association with worsening cognitive decline. Aβ, amyloid beta; AD, Alzheimer's disease; IL, interleukin; MCI, mild cognitive impairment; NfLight, neurofilament light.

Interestingly, some biomarker pairings had stronger concordance in dementia participants (eg, NfLight–GFAP, Aβ42/Aβ40–IL10). Upon further evaluation, two patterns of interest emerged informing biomarker kinetics in cognitive decline. The first pattern showed a steady increase in AD and inflammatory biomarker correlation with worsening cognitive decline, suggesting that the degree of change between the AD biomarker and inflammatory biomarker were similar (Figure 5B). These changes in correlation strength throughout dementia progression may indicate biological shifts occurring with AD pathology and the immune system. The second pattern revealed the greatest increase in association between normal and MCI, with either a decrease in association when progressing to dementia or steady correlation. This normal‐to‐MCI increase in association implies that during the earliest stages of cognitive decline, these specific pairings of biomarkers are being regulated together or similarly.

4. DISCUSSION

While plasma biomarkers are still being evaluated as predictors for sensitivity and specificity with regard to their targets, they are far more cost effective, accessible, and less invasive than other biomarker assays (MRI, CSF) and perhaps most likely to have utility in a first‐pass use, for example, to screen individuals for a need to perform PET or CSF testing. As plasma biomarkers gain traction in the diagnosis and prognosis of dementias, it is imperative to understand their reliability in various dementia populations, and especially to understand how factors such as comorbid conditions, race, sex, and ethnicity impact plasma biomarker readouts. As correlations between biomarkers (plasma with CSF, CSF with MR/PET imaging, plasma with imaging) are the future, we evaluated how the plasma biomarkers correlate in our entire cohort and across cognitive stages. We believe that the correlation of these biomarkers in our cohort will instigate evaluation in other cohorts, as well as help identify AD‐immune plasma profiles that with future research will elucidate more predictive modeling. With these goals in mind, we sought to validate several findings of how each biomarker changes with AD risk factors: sex, age, APOE genotype, and cognition status. Many previous papers evaluate these associations; however, many have lower patient numbers due to difficulty in recruitment or locations. ²⁸ , ²⁹

As has been established, sex differences exist between circulating biomarker levels. ³⁰ , ³¹ Our analysis revealed sex effects on plasma biomarkers Aβ40 and Aβ42, with elevated levels in males compared to females, and the astrocytic reactivity marker, GFAP, was elevated in females compared to males. Similarly, Stevenson‐Hoare and coauthors showed increased plasma GFAP in females, along with elevated NfLight and p‐tau181, which were not significant in our cohort. ³¹ Other groups have reported similar findings to ours, showing no sex‐specific effects on p‐tau181 levels in the plasma across a larger cohort. ³² It has been established previously that from a general perspective, females tend to have more immune‐based comorbidities, whereas males tend to have more vascular‐related risks and comorbidities, ³³ , ³⁴ and these trends are supported by our findings.

We also describe multiple biomarker changes associated with chronological age. Hallmark AD biomarkers (Aβ40, Aβ42, Aβ42/Aβ40 ratio, total tau, NfLight) were all significantly affected by age. However, this does not demonstrate the biomarkers’ ability to discern an AD versus control diagnosis, but simply that the populations studied must be age‐matched between controls, MCI, and dementia in order to properly assess the biomarker changes. ³¹ Lue and colleagues demonstrated a negative association of plasma Aβ42 with age, and a positive association of total tau and age; however, their study was conducted with all cognitively normal participants. ³⁵ While we and others show an age effect across multiple biomarkers, there are also papers suggesting no relationship between biomarkers and age.

In investigating whether our biomarkers showed similar results as other cohorts, we found mixed results. One population‐based study found increases of Aβ40 with increasing age, while our data showed a general decrease. ³⁶ The same study found plasma Aβ42 decreased before conversion to a dementia diagnosis, and we noted a general decrease between the 60s, 70s, and 80s, as well as a general decrease with congition. ³⁶ Another study looking at plasma total tau saw a general increase with age, the opposite finding of our present work. ³⁷ However, some of these studies have far fewer participants, which may cause variations between these results.

There was a general elevation in most inflammatory markers across ages as well (IL6, IL8, IL10, GFAP). While the elevation of biomarkers indicating AD pathology and immune response both increase with age, we also identified the decades in which changes occur in our cohort through post hoc tests. We show the most significant differences in hallmark AD biomarkers occurring at the younger end of the elderly adult spectrum (60s to 70s), validating our cohort with previous literature in plasma and CSF.

It was initially expected that there would be a linear effect of cognition on AD biomarkers; however, our data suggest that some biomarkers in the MCI phase are more highly affected. The literature suggests that many of these plasma biomarkers are most useful in determining the early stages of disease, before other contributors (ie, blood–brain barrier breakdown, neurodegeneration, chronic unchecked inflammation) come into play. ¹⁹ It also has been shown that these Aβ biomarkers may be elevated in a cognitively normal cohort that have Aβ plaque pathology without cognitive dysfunction, and may play a role in our findings, as our study did not test or exclude for individuals with Aβ PET+ scans. ³⁸

As more research is being conducted into the best combinations of plasma biomarkers to help minimize tests being run and to improve predictability, we set out to examine the relationships between plasma biomarkers. Although AD–inflammatory relationships are not novel in concept, the correlations between AD and inflammatory plasma biomarkers have not been fully explored. ⁵ , ³¹ , ³⁹ , ⁴⁰ We sought to evaluate patterns of correlation across a community‐based population between relevant and common hallmark AD and inflammatory biomarkers. Correlation analyses revealed tight associations between Aβ40, Aβ42, and total tau in the plasma, which corresponds to previous findings. ³¹ Surprisingly, the correlations between inflammatory biomarkers were also positively associated, suggesting a general activation of the systemic immune response.

Correlation analyses indicated synergy, or similar relationships, between inflammatory and AD biomarkers in the plasma that may contribute to AD–inflammatory patterns. To our knowledge, inflammatory biomarkers have not previously been compared to each other in plasma across dementia diagnoses. Our finding, showing significant positive correlations between pro‐ and anti‐ inflammatory markers, may suggest (1) overactivation of multiple immune system regulators, (2) multiple insults occurring at once creating opposing inflammatory effects, (3) dysregulation of the communication between inflammatory signals, and (4) multiple temporal inflammatory signals occurring simultaneously.

Interestingly, the greatest association was between IL8 and IL10, both having been shown to have antagonistic effects on the immune system. ⁴¹ Further, TNF𝛼 and IL6 are pro‐inflammatory, and IL10 is anti‐inflammatory, leaving one to hypothesize that these pro‐ and anti‐inflammatory markers would be oppositely correlated with AD biomarkers. Pro‐inflammatory IL‐8 typically acts as a chemotactic signal, encouraging phagocytic cell mobilization, while anti‐inflammatory IL10 suppresses pro‐inflammatory interferon gamma signaling and T‐cell recruitment. ⁴² , ⁴³ However, studies in viral and infection biology suggest that IL10 may have a mediating role on IL8 in some cases. ⁴⁴ , ⁴⁵ It has been shown that IL10 can also specifically act to inhibit IL6 and TNF𝛼 signaling as a feedback mechanism. ⁴⁶ , ⁴⁷ Nevertheless there are several points that play a role in why the initial hypothesis may be wrong. It was unexpected to see these biomarkers align in correlation directionality with AD biomarkers; however, given the age and comorbidities in our population, there are likely several sources of these inflammatory cytokines. The processes releasing the inflammatory and anti‐inflammatory substrates may be from completely different organs and cell types, and we agree that further study is warranted to understand how systemic comorbidities relate to dementia (in particular central nervous system) release of all cytokines. Further, we want to note that mouse‐based studies on IL10 suggest that increasing levels of IL10 worsen Aβ deposition and cognition. ⁴⁸ However, we believe this may be an oversimplification when extrapolating these results to human plasma across multiple dementias as there are many more factors at play, thus requiring further research and reports on human plasma biomarkers.

Although Aβ40, Aβ42, and total tau appear negatively correlated with inflammatory biomarkers IL6, IL8, IL10, and GFAP, the reduction of these AD biomarkers in plasma indicate increased corresponding pathology in the brain, suggesting an ultimate positive correlation between brain pathology and systemic inflammation. Interestingly, the correlation coefficients indicated between inflammatory biomarkers and AD biomarkers Aβ40, Aβ42, and total tau, were generally stronger than the association between the inflammatory biomarkers and the Aβ42/Aβ40 ratio, p‐tau181, and NfLight. AB40, AB42, and total tau are not as sensitive or reliable as p‐tau181 and p‐tau217 (not tested in this battery) for Aβ pathology. ⁴⁹ One study showed that the Aβ42/Aβ40 ratio is higher in patients with AD pathology and other comorbid dementia pathologies, suggesting other dementia pathologies influence these biomarkers. ⁵⁰ Additionally, while Aβ40 is often used alone to create the Aβ42/Aβ40 ratio, there are many other works that point towards the usefulness of Aβ40. Besides Aβ40 being the more prominent isoform in the CSF, it has been interpreted that it may be a better measure of overall Aβ concentration among multiple dementias, given that in AD specifically, there is an increase in pathological Aβ42, creating the Aβ42/Aβ40 ratio. ⁵¹ , ⁵² , ⁵³ Others have shown that CSF Aβ40 was decreased in AD dementia patients and also non‐AD dementia patients, when compared to non‐dementia controls, as well as improving diagnostics via a biomarker‐based decision tree, indicating that overall Aβ40 may provide useful insight into dementia biomarker progress. ⁵⁴ While the aforementioned findings were in CSF, plasma‐based biomarkers research is following suit and more studies need to be conducted and reported to evaluate the usefulness of each plasma biomarker.

When measuring forms of tau, p‐tau181 particularly correlates best with Aβ pathology in the brain, while total tau has been used as a measure of multiple isoforms of tau, as well as a proxy for neuronal damage. ⁵⁵ , ⁵⁶ Total tau has been reported to increase after events such as stroke or traumatic brain injury (TBI), suggesting it may be altered due to other (comorbid) sources. ⁵⁶ As our cohort is not AD specific, other disorders may be present and influence total tau in the plasma (ie, Lewy body dementia, TBI, and limbic‐predominant age‐related TDP‐43 encephalopathy). Interestingly, p‐tau181 concentrations showed differences between cognitive status, with a lower concentration in MCI compared to normal, while also showing no difference between age groups, suggesting this is a cognitive status‐specific effect.

We further evaluated the correlations between hallmark AD biomarkers and inflammatory biomarkers in the plasma as they vary across dementia stages (normal, MCI, dementia). While we saw similar patterns and no overt changes in correlation between AD pathology and immune signaling, we did observe changes in correlation strength, indicating that different pairings of biomarkers may synergize at different timepoints in disease. Some pairings of biomarkers showed significant associations in a particular cognitive stage (eg, MCI: NfLight–TNFα, p‐tau181–TNFα; normal: total tau–GFAP), which may indicate biological regulation of AD through a specific inflammatory mediator. In the NfLight–IL10 pair, we saw changes that noted the strongest correlation coefficient in MCI participants. Further, some associations between biomarkers lose significance during the MCI stage, which may suggest dysregulation during early decline where normally there is tighter co‐regulation (total tau–TNFα, Aβ42–IL6). Further, pairings of biomarkers may provide additional measures for screening, diagnosis, and prognosis of disease pathology.

Previously, others utilized the UK‐ADRC cohort in similar fashion to evaluate neurodegenerative, immune, and vascular biomarkers across cognitively normal participants over 5 years. ⁵⁷ Interestingly, in cognitively normal plasma there was a decrease in the Aβ42/Aβ40 ratio and total tau with age; however, that population was slightly older (mean baseline age = 82.6 years). ⁵⁷ Similar to our findings, the APOE ε4 allele influenced the Aβ42/Aβ40 ratio, showing a decrease in this biomarker by ≈19% with one ε4 allele. Our data revealed reduced means in the Aβ42/Aβ40 ratio in APOE ε3/ε4 and APOE ε4/ε4 compared to APOE ε3/ε3 carriers; however, these differences were not statistically significant by post hoc Tukey tests. These findings provide evidence for continued research into pairings of AD and Inflammatory plasma biomarkers across cognitive status in large‐scale community‐based cohorts to evaluate biological mechanisms and improve normal and dementia diagnosis and prognosis.

It is clear that more research needs to be done surrounding the inflammatory biomarkers in AD and related dementias. While our study sets the stage for utilization of correlations of biomarkers throughout cognitive changes, there are a few limitations. We are unable to determine brain cell type of origin for the various analytes. Further workup associating plasma inflammatory effects with protein levels in the CSF and autopsy brain would help elucidate the contribution of cerebral inflammation to these plasma signals. Studies targeting the sources of these cytokines and the underlying biology of brain‐to‐blood transmission would aid in our understanding of the true causes of differential biomarker abnormalities. Additionally, studies could utilize exosomes from specific cell populations to help identify the cell or organ of origin. Further, life course studies, such as the Framingham Heart study, could help identify various health conditions and how inflammatory plasma biomarkers are marked by life events, helping to pinpoint biomarker use specific to AD or other related dementias and not a result of other comorbid symptoms.

While including races of all types in this study, we agree there is an underrepresentation of racial minorities in our work and we therefore did not have the power to address racial differences in biomarker patterns. Future studies pooling multiple ADRC plasma biomarkers may provide key insight into how racial background influences AD and inflammatory biomarker trends. Together, longitudinal studies are needed in diverse populations to determine temporal changes and understand whether circulating inflammation markers are a response or occur prior to AD pathology. ⁵⁸ Further, our list of inflammatory biomarkers was not exhaustive, as there are many other immune‐based plasma biomarkers showing promise in dementia detection. ⁵⁹ Studies on groups of plasma biomarkers, such as complement activation, and their association to AD biomarkers in plasma, may provide insight into the progression of dementia in AD and other neurodegenerative diseases. Additionally, in this study we did not include vascular biomarkers, which have been shown to change earlier in the cognitive decline process indicating the presence of cerebrovascular pathology. ⁶⁰ Future studies encompassing vascular and immune corre‐ lations would be pertinent to including subjects with specific vascular comorbidities and further evaluating how these vascular biomarkers correlate with inflammatory biomarkers in this subgroup.

This work, completed in a large community‐based cohort, leverages the use of plasma biomarkers across a spectrum of AD and associated dementia‐related pathologies and immune response factors to provide insight into how AD and inflammatory biomarkers are changing with regard to high‐risk factors for dementia, such as sex, age, and APOE genotype, as well as how biomarkers correlate throughout dementia progression. Our findings indicate correlative effects of sex, age, and APOE genotype for AD and inflammatory biomarkers, which should be validated in other community‐based cohorts, to establish a standard collective understanding of these plasma biomarkers’ kinetics. Our work also found associations between classic AD and inflammatory biomarkers in plasma, indicating increased immune response with increasing AD brain pathology. However, some pairings of AD–inflammatory biomarkers showed increased strength in association during MCI, and some later in dementia stages. These variations in association between biomarkers may reveal key inflammatory pathways that are activated in the periphery due to different brain pathology, yet more research is necessary.

AUTHOR CONTRIBUTIONS

All blood was collected by BJM, and plasma was processed and analyzed by TLS. KEF performed all analyses. KEF and DMW conceived and designed this project and wrote the manuscript. All authors approved the final manuscript.

CONFLICTS OF INTEREST STATEMENT

Donna M. Wilcock is the Editor‐in‐Chief of Alzheimer's and Dementia but was not involved in the handling, processing, or reviewing of this manuscript. The authors have nothing further to disclose.

CONSENT STATEMENT

All participants signed and consented for their participation and data usage and all protocols have been approved by the University of Kentucky Institutional Revies Board.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Foley KE, Winder Z, Sudduth TL, et al. Alzheimer's disease and inflammatory biomarkers positively correlate in plasma in the UK‐ADRC cohort. Alzheimer's Dement. 2024;20:1374‐1386. 10.1002/alz.13485