Association between anti-inflammatory interleukin-10 and executive function in African American women at risk for Alzheimer’s disease

Abstract

Introduction:

African-Americans (AAs) are 64% more likely to be diagnosed with AD than non-Hispanic Whites. AAs with elevated AD biomarkers exhibit greater neurodegeneration in AD signature regions compared to non-Hispanic Whites with elevated AD biomarkers. This pilot trial examined whether normal or elevated plasma levels of interleukin (IL)-10 are associated with changes in executive function and short-term memory in AA women at risk for developing AD due to parental history.

Method:

Observational study comparing groups with elevated and normal plasma IL-10 levels. Study included 31 AA women (age=58.9±8 years) with parental history of AD. Measures included inflammatory blood biomarkers, executive function and visuospatial short-term memory tests. Multivariate linear regression with adjustment for comorbidities, and Bonferroni corrections for multiple comparisons were used to compare groups. Effect sizes (Cohen’s d) were generated. Using endpoints with moderate-large effects between groups, Pearson correlations determined associations between biomarker levels and cognitive performance.

Results:

The elevated IL-10 group performed worse on the Trail-Making Test proportional score ((B-A)/A) (effect size (d =−0.87 (−1.6, −.1)). Moderate effects with large confident intervals were noted in inhibition, set-switching, and body position spatial memory. Significant differences between groups in levels of other inflammatory markers were noted, including IL-7 (p=0.002) and interferon γ (p=0.02). IL-7 remained significant after Bonferroni correction. Correlation matrices revealed moderate-large, significant correlations (yet with wide confidence intervals) between levels of IL-10 and IL-9 with BPST total correct trials, and between interferon γ and delayed recall.

Conclusions:

Interleukins may incite inflammation, leading to impaired aspects of executive function and short-term memory in this sample of African American women at risk for developing AD. This research provides effect sizes that will be used to power future research that will further investigate the relationship between inflammation, AD biomarkers, and cognitive function in an understudied population.

Introduction

As the average life expectancy of the United States population increases, prevalence of neurodegenerative diseases, such as Alzheimer’s Disease (AD), also increases (McDonough, 2017). In 2010, 4.7 million Americans had AD, which is expected to increase to 13.8 million diagnoses by 2050 (Matthews et al., 2019). Multiple factors influence one’s risk for developing AD. African-Americans (AAs) are 64% more likely to be diagnosed with AD than Whites (Steenland et al., 2016). AAs with elevated AD biomarkers, i.e., amyloid beta (Aβ) and tau deposition, exhibit greater neurodegeneration in AD signature regions compared to Whites with elevated AD biomarkers (McDonough, 2017). By age 85, descendants of AAs with AD are 1.6 times more likely to develop dementia than descendants of Whites (Green et al., 2002). AAs usually present with an earlier age of onset, exhibit greater severity of symptoms, and are less likely than Whites to receive treatments for AD (Barnes & Bennett, 2014). Family history of AD, especially maternal history, is associated with increased AD risk and reduced cerebral blood flow (Chetelat et al., 2013). Sex also impacts AD onset and age-related cognitive decline, and women are at a higher risk for AD than men (O’Hagan et al., 2012). Factors, including longevity and estrogen decline during the pre- and post-menopausal transition, contribute to the increased incidence of AD in women (Giacobini, 1998). Clearly, an individual’s sociodemographic profile influences his or her risk for AD development.

AD causes decline in multiple cognitive domains including executive function. Executive function is need for planning, problem solving, cognitive flexibility, and performing goal-oriented activities (Jahn, 2013). Although memory impairment is requisite in diagnosis of AD, executive function is notably among the first cognitive domains affected in AD. In older women, executive function may be predictive of future declines in memory and global cognition (Carlson et al., 2009). Approximately 49% of older women included in a 9-year longitudinal study developed cognitive impairment and executive function, which preceded overall memory decline. Executive function likely declines more rapidly (Chen et al., 2001) and to a greater extent in those with AD (Albert et al., 2007). In a longitudinal study held over 1.5 years, people diagnosed with AD experienced a decline in executive functioning (Smits et al., 2012). Additionally, individuals with a family history of AD performed poorer on executive function tests than those without a family history (Donix et al., 2012). Versus those without, individuals with family history of AD performed worse on the Wisconsin Card Sorting Test, an executive functioning measure of response inhibition, abstract reasoning, and perseveration (Donix et al., 2012). Tasks assessing characteristics of executive function may be beneficial in early detection of AD in high risk populations based on sex, race, and family history.

Inflammation and cognitive function

Histologically, AD is characterized by the presence of aggregated tau proteins and amyloid beta (Aβ) peptides. These tau “tangles” and Aβ “plaques” which may form secondary to inflammatory processes (Kinney et al., 2018). Aging is associated with chronic elevation of blood cytokines, which are proteins secreted throughout the body that act as signaling molecules allowing for the communication between different cell types (Alvarez-Rodriguez et al., 2012; Pedersen et al., 2000; Wei et al., 1992). This chronic systemic inflammation has been correlated with an increased risk for neurodegeneration and age-related cognitive decline (Holmes, 2013; Michaud et al., 2013; Tegeler et al., 2016; Weaver et al., 2002). Inflammatory markers have been linked to poorer cognitive function in specific groups of people. Higher plasma C-reactive protein (CRP) levels have been linked to worse mental status in women younger than 50 years, and worse attention in older women and AAs (Beydoun et al., 2018). Whether peripheral inflammation is a key force in progression of cognitive decline in AD is unclear.

Systemic inflammation can cause neuroinflammation through pathways such as the vagus nerve (Quan & Banks, 2007). Neuroinflammation, such as microglia activation and elevated levels of pro-inflammatory cytokines, occur in AD (Kinney et al., 2018). Our group recently reported that neuroinflammation was correlated to tau levels in cerebrospinal fluid (CSF), and that this relationship was modified by race (Wharton et al., 2019). In addition to the neuroinflammation induced by systemic inflammation, Aβ plaques and tau tangles can stimulate the production of cytokines (Weisman et al., 2006). This induction of the immune response and whether it would be beneficial or harmful to the brain is debated (Czirr & Wyss-Coray, 2012).

Whether cytokines are beneficial or harmful to neuronal function depends upon the function of the cytokine (either inflammatory or anti-inflammatory) and the concentration of the cytokine. Interleukins (ILs) are a subset of cytokines can be eiether inflammatory or anti-inflammatory. Interleukin (IL)-10 is an anti-inflammatory cytokine. The connection between AD and IL-10 has been reported, but results are mixed. Some level of IL-10 appears to be necessary for proper immune function and for normal cognitive function (Fabregue & Butkowski, 2016). AD patients have been documented to have greater frequency of a nucleotide polymorphism (−1082A) which is associated with low production of IL-10 (Combarros et al., 2009). Lower IL-10 levels may be associated with poorer cognitive function; however, overproduction of IL-10 also seems to be increase cognitive dysfunction.

High IL-10 levels have been associated with lower composite scores of executive function and cognitive speed (Tegeler et al., 2016). In rodent studies, learning and memory were more impaired as IL-10 levels increased in AD mice (Chakrabarty et al., 2015). Excessive IL-10 production can inhibit the proinflammatory response (Iyer & Cheng, 2012) leading to a suppression of microglial Aβ phagocytosis (Chakrabarty et al., 2015). Further, studies have also linked IL-10 and AD with ApoE, such that IL-10 increases the expression of ApoE (Chakrabarty et al., 2015). Race may affect the association between IL-10 and AD. There are different allelic frequencies for IL-10. The IL-10–1082 G gene has been shown to be much lower in Asians, but not whites (Ma et al., 2005). Yet, very few studies of inflammation and cognition have included AA participants (Windham et al., 2014).

Very few studies of inflammation and cognition have included AA participants (Windham et al., 2014). This pilot study examined whether IL-10 and other related inflammatory markers were associated with changes in executive function and short-term memory, in middle aged AA women with a parental history of AD. In this preliminary work in an understudied population, we tested an overall hypothesis that individuals with levels of IL-10 above an established cutoff from the literature would have worse cognitive performance, particularly in executive function, compared to a group with normal IL-10 levels. From this battery of executive function and short-term memory measures in a small sample, we aimed to derive effect sizes that would indicate a test with adequate sensitivity to power future trials. We additionally aimed to identify appropriate secondary cognitive measures and explore relationships with related biomarkers

Methods

Thirty-one middle-aged (45–65 years old) AA women taking part in a larger clinical trial were included in the present study (Hackney et al., 2019). Participants were recruited from the Emory Alzheimer’s Disease Research Center and from ongoing studies of individuals with parental history of AD. The participants’ parents had a diagnosis of probable AD as defined by National Institute of Neurological Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association (NINDS-ADRDA) criteria, which was verified using the validated Dementia Questionnaire and medical records when available (Arabi et al., 2016).

Participants underwent blood draws for inflammatory cytokines after an 8-hour overnight fast by a member of the research team. Four panels of biomarkers were measured in plasma using singleplex or multiplex assays in a Luminex 200 platform. Participants also performed cognitive tasks.

Cognitive testing included evaluation of global and executive function, using the Montreal Cognitive Assessment (MoCA) (including the delayed recall section examined separately), the Serial 3s test, the Trail-Making Test Parts A & B (Trails), the D-KEFS Color Word Interference Task (CWIT), and the D-KEFS Tower of London (TOL) task.

Short term/working memory tests of spatial cognition were included: The Brooks Spatial memory task, the Corsi blocks test and the Body Position Spatial working memory test. For the CWIT and the TOL, several scores have been converted into scaled values based on age group norms and published guidelines (Delis et al., 2001).

Cognitive assessments

The MoCA is a global cognitive screen for patients with mild cognitive impairment that measures visuospatial, executive function, language, attention, concentration and working memory. Visuospatial abilities are assessed through a task where participants are required to draw a clock and a cube. Areas of executive functioning are assessed using the Trail making B task, a fluency task, and a verbal abstraction task. Attention, concentration, and working memory are evaluated through a sustained attention task, serial subtraction task, and digit forwards and backwards task. Language is assessed through a naming task, fluency task, and complex sentences. Knowledge of place and orientation is also tested (Nasreddine et al., 2005). Thirty is the highest possible score on the MoCA, although the definition of MoCA-indicated cognitive impairment has varied across studies (Lim & Loo, 2018). In addition to the total score, the Delayed Recall section was also considered for analysis.

We used the mental processing task, Serial 3s, to assess working memory and mental status. Participants are asked to perform serial subtractions by 3s from a random number ranging from 20–100. Participants were given 15 seconds to perform subtractions. Three trials were administered, and number of errors and correct subtractions were averaged from each of the three trials. Percent correct was calculated from the total number of responses given.

Executive function

A battery of executive function assessments was administered to capture many subdomains within the overall umbrella concept of Executive function, including the subdomains of attention, planning and modification of that plan, goal-directed behavior, inhibition, set-switching, maintaining different trains of thought and cognitive flexibility.

Conditions A and B of the Trail-Making Test were administered to participants. Condition A requires participants to draw lines connecting numbers inside circles in numerical order. Condition B requires participants to draw lines connecting alternating numbers and letters in order. This test reflects several cognitive processes including attention, visual search and scanning, sequence and shifting, ability to execute and modify a plan of action, the ability to maintain different trains of thought at once, and fluid cognitive ability (Salthouse, 2011). The difference score between conditions A and B, which is B-A, was used as an outcome variable. Performance of Trail-Making Test Part B relative to performance on part A reflects the ability to switch between two tasks and efficiency in suppression of a task that was previously abandoned. The difference score between parts A and B provides information on task switching independently of motor speed and visual scanning speed (Arbuthnott & Frank, 2000). We also calculated the proportional score ((B-A)/A) to further address issues the TMT test has with processing speed.

The Color Word Interference Task contains color naming, color name reading, inhibition, and switching conditions. Executive function components of inhibition and switching are able to be isolated because results control for color naming and reading, which are measured in participants in the first two conditions of the test (Fjell et al., 2017). Scaled (according to age norms) variables are yielded from this test. For this study of cognitively normal participants, we included variables only from the more challenging conditions. As such, we analyzed: scaled summary and error scores for the Inhibition and Inhibition/switching conditions, and the contrast score that isolates switching from the inhibition/switching condition. For all variables but the contrast scores, higher scores indicate better performance. The contrast scaled score parcels out performance on the key component task (inhibition) from the higher level task (inhibition/switching). The contrast scaled score has a mean of 10 (SD = 3). A contrast scaled score between 8 and 12 reflects a roughly equivalent performance on the higher-level task to the baseline measure, whereas 13 and higher reflects disproportionately better performance on the higher level task relative to that on the baseline measure.

The Tower of London test is designed to measure frontal lobe function and executive function processes (Rainville et al., 2012). Performance on the Tower of London requires coordination of distinct cognitive abilities that fall under executive function: planning an organized series of steps, goal-directed behavior, inhibition of inappropriate move selections and error monitoring. Working memory is also involved in this test (Rainville et al., 2012). Scaled (according to age norms) variables are yielded from this test.

Short-term memory

Tests of visuospatial, short term memory were used in the battery to assay spatial cognition, mental imagery and short-term memory.

Brooks Spatial Memory Task- –

This task is an assessment of spatial cognition and short-term memory that requires participants to use mental imagery in remembering and repeating the placement of three to eight numbers. Participants receive a visual presentation and verbal explanation of a 4 × 4 matrix. The researcher tells the participant the position of the numbers, and the participants must remember and repeat the position of the numbers after each trial (Brooks, 1967; Kline et al., 2014).

Reverse Corsi Blocks Task–

The Corsi Block-Tapping Task assesses short-term visuospatial memory. The task requires maintenance of both a movement sequence and a visuospatial pattern. The task consists of a panel with nine blocks fixed at random positions. The blocks are labeled on one side, such that numbers are only visible to the assessor. The assessor taps several blocks in a certain order, after which the participant must tap the block sequence in the reverse order as presented. The block sequences gradually increase in length up to 9 blocks (Kessels et al., 2008).

Body Position Spatial Memory Task–

Modeled after the Corsi Blocks test, Body Position Spatial Task (BPST) is a test of short term/working memory of the spatial domain. Examiners visually and verbally demonstrate a pattern of side and forward steps and turns, which the participant then repeats. After a practice trial of two moves, the examiner begins with two moves and progresses to a maximum of nine moves. The number of moves increases by one at each subsequent level containing two trials. Participants advance to the next level if they correctly completed at least one of the trials in a level. Once a participant missed both trials in a level, the task ends. Span (length of sequence of moves that were correctly performed) and number of trials performed correctly (consistency of performance) are used for analyses (Battisto et al., 2018).

Stratification into elevated and normal IL-10 groups

IL-10 levels were categorized into 2 groups (elevated and normal) using designated IL-10 cutoffs from clinical literature (Sarris et al., 1999). Sarris et al. derived these cutoffs from a sample with a M = 7.1, SD = 1.5 pg/mL. Based on these values, the upper limit of the normal range of IL-10 was determined by adding two standard deviations to the mean to obtain 10.1 pg/mL. Therefore, IL-10 levels under 10 pg/mL were considered normal while values equal to 10 pg/mL or higher were categorized as elevated (Sarris et al., 1999). A later study, using a cutoff level of 10 pg/ml to divide a group of melanoma patients into normal and elevated groups, revealed significant differences in survival between groups. Mean survival rates of patients with IL-10 levels below 10 pg/ml were significantly higher than patients with IL-10 levels of 10 pg/ml or higher (Nemunaitis et al., 2001). Because of these reports, we used the cutoff of 10 pg/ml for IL-10 levels to categorize the sample into elevated and normal IL-10 groups.

Statistical analysis

Descriptive statistics were calculated and compared between groups using t-tests or Fisher’s exact tests. On cognitive and inflammatory variables, we compared groups using multivariate linear regression, adjusting for the number of chronic diseases, education and income. Significant p values were corrected using the Bonferroni correction for multiple comparisons. Effect sizes (Cohen’s d) were calculated according to the formula:

$${\text{Cohen}}^{\prime }\text{s}d=\left(M_2-M_1\right)/S{D}_{\text{pooled}\text{where}SD\text{pooled}}=\surd \left(\left(SD{1}^2+SD{2}^2\right)/2\right)$$

Confidence intervals were generated. Correlation matrices were calculated with Pearson’s r. The variables of interest (e.g., IL-10, memory variables) and variables that had moderate to large effects as a result of the multivariate analyses were included in the correlational analyses.

Data were analyzed with R software and SPSS (version 22).

Results

Thirty-one women with parental history of AD (age 58.9 ± 8 years) were included in the sample. Participants in the normal IL-10 group and the elevated IL-10 group did not differ in education level, number of medications, occupational status, caregiver status, income, or indices of sleep (Table 1).

Table 1.

Characteristics of 31 African American women with parental history of AD.

Characteristics	Whole Sample N = 31 Mean (SD)/N (%)	Normal IL-10 N = 17 (55%) Mean (SD)/N (%)	Elevated IL-10 N = 14 (45%) Mean (SD)/N (%)	p
Age (years)	58.9 ± 8.1	61.2 ± 8.9	55.9 ± 6.0	0.078
Highest Education Level	1 (3.2)	0 (0)	1 (7.1)	0.789
High School Grad/GED	2 (6.5)	1 (5.9)	1 (7.1)
Vocational training	5 (16.1)	3 (17.6)	2 (14.3)
Some	14 (45.2)	8 (47.1)	6 (42.9)
college/Associate’s
Bachelor’s degree	5 (16.1)	2 (11.8)	3 (21.4)
Master’s degree	4 (12.9)	3 (17.6)	1 (7.1)
Doctoral degree
Number of Medications	5.1 (4.28)	4.2 (3.3)	6.1 (5.2)	0.246
Occupational Status	9 (29)	6 (35.3)	3 (21.4)	0.724
Work full-time	4 (12.9)	2 (11.8)	2 (14.3)
Work part-time	12 (38.7)	7 (41.2)	5 (35.7)
Retired	2 (6.5)	1 (5.9)	1 (7.1)
Unemployed	4 (12.9)	1 (5.9)	3 (21.4)
Disabled
Caregiver	19 (61.3)	10 (58.8)	9 (64.3)	1.000
Yes	12 (38.7)	7 (41.2)	5 (35.7)
No
Income	5 (16.1)	1 (5.9)	4 (28.6)	0.092
$19,000 or less	6 (19.4)	4 (23.5)	2 (14.3)
$20,000-$39,000	11 (35.5)	9 (52.9)	2 (14.3)
$40,000-$59,000	2 (6.5)	1 (5.9)	1 (7.1)
$60,000-$79,000	7 (22.6)	2 (11.8)	5 (35.7)
$80,000 or more
Housing	29 (93.5)	16 (94.1)	13 (92.9)	1.000
House/Apt/Condo	2 (6.5)	1 (5.9)	1 (7.1)
Relative’s Home
Use Assist. Device Walking	5 (16.1)	4 (23.5)	1 (7.1)	0.457
Some of the time	26 (83.9)	13 (76.5)	13 (92.9)
Never
Marital Status	5 (16.1)	2 (11.8)	3 (21.4)	0.867
Single	10 (32.3)	6 (35.3)	4 (28.6)
Married/partnered	13 (41.9)	7 (41.2)	6 (42.9)
Separated/divorced	3 (9.7)	2 (11.8)	1 (7.1)
Widowed
Sleep Quality Past Month	2 (6.5)	1 (5.8)	1 (7.1)	0.861
Excellent	6 (19.4)	4 (23.5)	2 (14.3)
Very good	8 (25.8)	5 (29.4)	3 (21.4)
Good	9 (29.0)	5 (29.4)	4 (28.6)
Fair	6 (19.4)	2 (11.8)	4 (28.6)
Poor
Number Chronic Diseases	1 ± 0.86	1.06 ± 0.90	0.91 ± 0.83	0.662
Interleukin-10 levels (pg/mL)	10.8 (4.7)	7.73 (1.4)	14.61 (1.4)	<0.001

Table 1. Participant Baseline Characteristics. Values are presented as Mean± SD for continuous variables, and n (%) for categorical variables. p values were calculated with t-test/ANOVA for continuous variables and Chi-square test for categorical variables. IL = interleukin.

All effects had wide confidence intervals for cytokine analysis. Multivariate analysis with adjustment for number of comorbidities revealed significant differences between groups in levels of IL-7 (p=0.002) and interferon γ (p=0.02). IL-7 was significant after Bonferroni correction (Table 2).

Table 2.

Inflammatory biomarker levels of 31 African American women with parental history of AD.

	Total	Normal IL-10 (n = 17)	Elevated IL-10 (n = 14)	Mean Difference	Raw p	Effect Sizes Cohen’s d (95% Cl) (Raw)	Multivariate Adj M Difference	p (Adj)
TGFα ^	1.22 ± 1.46	0.62 ± 0.74	1.95 ± 1.78	−1.32 (0.25,2.4)	0.019	−1.01 (−1.8, −0.2)	−1.18 (−2.4, 0.1)	0.062
IFNg ^	11.72 ± 6.97	8.65 ± 4.46	15.45 ± 7.77	−6.8 (1.91,11.69)	0.009	−1.1 (−1.9, −0.3)	−6.86 (−12.5,−1.2)	0.020
MDC	932.34 ± 344.26	877.69 ± 359.04	998.7 ± 325.87	−121 (−133.33,375.34)	0.339	−0.35 (−1.1, 0.4)	−60.28 (−405, 284.4)	0.716
IL-9^	0.56 ± 1.01	0.16 ± 0.34	1.06 ± 1.31	−0.91 (0.14,1.67)	0.024	−0.99 (−1.8, −0.2)	−0.62 (−1.3, 0.1)	0.078
IL-7^	4.28 ± 1.25	3.58 ± 0.8	5.13 ± 1.18	−1.55 (0.82,2.28)	<0.001	−1.57 (−2.4, −0.7)	−1.62 (−2.6, −0.7)	0.002*
IL-8	11.83 ± 30.19	3.86 ± 1.56	21.5 ± 43.78	−17.63 (−7.65,42.92)	0.156	−0.6 (−1.4, 0.2)	−2.19 (−5.7, 1.3)	0.203
MCP-1	172 ± 52.83	164.38 ± 53.2	181.26 ± 52.8	−16.89 (−22.25,56.02)	0.385	−0.32 (−1.1, 0.4)	−7.11 (−68.6, 54.4)	0.810
TNFα	15.88 ± 57.72	23.8 ± 78.09	6.26 ± 1.66	17.54 (−57.7,22.61)	0.368	0.3 (−0.4, 1)	10.69 (−51.9,73.3)	0.722
CRP	0.37 ± 0.46	0.48 ± 0.57	0.24 ± 0.24	0.24 (−0.56,0.08)	0.128	0.53 (−0.2, 1.3)	0.13 (−0.4, 0.7)	0.613
SAP	0.27 ± 0.09	0.26 ± 0.09	0.29 ± 0.11	−0.02 (−0.05,0.09)	0.506	−0.24 (−1,0.5)	−0.06 (−0.2, 0)	0.187

Values are presented as Mean ± SD, and the units are pg/mL, except for CRP and SAP which were measured in ng/mL; IL = interleukin; TGFα = Transforming growth factor alpha; IFNg = Interferon gamma; MDC = Macrophage-derived chemokine; MCP-1 = Monocyte chemoattractant protein −1; TNFα = Tumor necrosis factor alpha; CRP = C-reactive protein; SAP = Serum amyloid P-component. Adjusted mean differences between groups obtained by linear regression model adjusting for number of chronic diseases, education and income. ^{^} Large effect size per Cohen convention.

^*Significant finding, Bonferroni corrected p <.0005.

All effects had wide confidence intervals for cognitive assessments. A large effect (d = −0.87 (−1.6, −.1)) was observed on the Trail-Making Test proportional score ((B-A)/A). The elevated IL-10 group performed worse. Moderate effects were noted in CWIT Inhibition scaled score, inhibition errors score (scaled), and the Switching contrast score, body position spatial memory (number of correct trials), and the TOL mean rate per move time (scaled) (Table 3).

Table 3.

Cognitive performance in African American women with parental history of AD.

	Total	Normal IL-10 (Raw)	Elevated IL-10 (Raw)	Mean Difference	Raw p	Effect Sizes (95% Cl) (Raw)	Multivariate M difference (Adj)	p
Serial 3s % Correct (%)	90.41 ± 16.3	92.52 ± 11.4	87.85 ± 21.0	4.67 (−17.75,8.41)	0.464	0.28 (−0.5, 1)	−4.33 (−16.5,7.8)	0.461
Montreal Cognitive Assessment (/30)	25.74 ± 2.71	25.88 ± 3.0	25.57 ± 2.3	0.31 (−2.34,1.72)	0.756	0.11 (−0.6, 0.9)	0.54 (−2.4, 3.4)	0.701
Delayed Recall (MOCA)	3.61 ± 1.1	3.53 ± 1.3	3.71 ± 0.8	−0.18 (−0.65,1.02)	0.654	−0.16 (−0.9, 0.6)	−0.12 (−1.4, 1.1)	0.847
Brooks Spatial Memory (% Correct)	68.84 ± 18.0	68.59 ± 15.1	69.14 ± 21.5	−0.55 (−12.92,14.03)	0.933	−0.03 (−0.8, 0.7)	−2.16 (−23.4, 19.1)	0.832
BPST Span	3.74 ± 1.0	3.59 ± 1.1	3.93 ± 0.9	−0.34 (−0.4,1.08)	0.354	−0.34 (−1.1, 0.4)	−0.7 (−1.8, 0.4)	0.195
BPST Total Correct Trials	4.29 ± 1.4	4.00 ± 1.3	4.64 ± 1.6	−0.64 (−0.41,1.7)	0.223	−0.45 (−1.2, 0.3)	−0.98 (−2.3, 0.4)	0.140
Corsi Span	4.23 ± 1.0	4.18 ± 0.6	4.29 ± 1.3	−0.11 (−0.67,0.89)	0.772	−0.11 (−0.9, 0.6)	−0.18 (−1.3, 0.9)	0.738
Corsi Total Correct Trials	5.42 ± 1.5	5.47 ± 1.1	5.36 ± 1.9	0.11 (−1.32,1.09)	0.846	0.07 (−0.7, 0.8)	0.06 (−1.6, 1.7)	0.939
TMT Difference (B-A) (s)	43.59 ± 24.6	37.26 ± 22.7	51.27 ± 25.3	−14.01 (−3.64,31.67)	0.115	−0.59 (−1.3, 0.2)	−17.92 (−42.4, 6.6)	0.140
TMT Proportion (B-A)/A	1.84 ± 0.9	1.51 ± 0.7	2.24 ± 1.0	−0.73 (0.11,1.34)	0.022	−0.87 (−1.6, −0.1)	−0.8 (−1.6, 0) *	0.050
CWIT Inhibition scaled score (/19)	10.23 ± 2.6	10.76 ± 1.8	9.57 ± 3.3	1.19 (−3.23,0.84)	0.235	0.47 (−0.3,1.2)	1.21 (−0.9, 3.4)	0.250
CWIT Inhibition/switching scaled score (/19)	10.48 ± 2.1	10.65 ± 1.7	10.29 ± 2.6	0.36 (−1.96,1.24)	0.648	0.17 (−0.6, 0.9)	0.4 (−1.5, 2.3)	0.664
CWIT Contrast switching scaled score*	10.26 ± 1.9	9.88 ± 1.9	10.71 ± 1.9	−0.83 (−0.58,2.25)	0.239	−0.43 (−1.2, 0.3)	−0.81 (−2.8, 1.2)	0.399
CWIT Inhibition total errors scaled score (/12–13)	10.68 ± 2.3	11.12 ± 1.5	10.14 ± 3.0	0.97 (−2.81,0.86)	0.279	0.43 (−0.3,1.2)	0.14 (−1.5, 1.8)	0.862
CWIT Inhibition/switching total errors scaled score (/12–13)	10.58 ± 2.2	10.94 ± 2.3	10.14 ± 2.0	0.8 (−2.39,0.79)	0.313	0.37 (−0.4,1.1)	0.68 (−1.4, 2.8)	0.499
TOL Total achievement scaled score (/19)	9.68 ± 1.9	9.71 ± 2.1	9.64 ± 1.8	0.06 (−1.51,1.38)	0.929	0.03 (−0.7, 0.8)	−0.24 (−2.7, 2.3)	0.841
TOL Mean first move time scaled (/17–19)	9.35 ± 2.5	9.53 ± 3.0	9.14 ± 2.0	0.39 (−2.29,1.51)	0.680	0.15 (−0.6, 0.9)	−1.11 (−4, 1.8)	0.427
TOL Time per move ratio scaled (/17–18)	9.06 ± 3.3	8.47 ± 3.5	9.79 ± 2.9	−1.32 (−1.11,3.74)	0.276	−0.40 (−1.1, 0.3)	−2.76 (−6.5, 1)	0.135
TOL Move accuracy ratio scaled (/18)	8 ± 2.7	7.82 ± 2.9	8.21 ± 2.6	−0.39 (−1.64,2.42)	0.696	−0.14 (−0.9, 0.6)	0.92 (−2, 3.9)	0.227
TOL Total rule violations cumulative percentile rank (/100)	57.97 ± 32.9	60.29 ± 33.64	55.14 ± 33.1	5.15 (−29.8,19.49)	0.672	0.15 (−0.6, 0.9)	16.61 (−23,56.3)	0.632

Elevated = interleukin-10 ≥ 10 pg/ml; Normal = interleukin-10, 3–10 pg/ml; Values are presented as Mean ± SD for continuous variables, and n (%) for categorical variables. Adjusted mean differences presented in Mean ± SD and obtained by linear regression model adjusting for number of chronic diseases, education, and income. Alpha was originally set at p <.05 and Bonferroni corrected for multiple comparisons to 0.0025.- TMT: Trails Making Test CWIT: Color Word Interference Task; TOL: Tower of London. ^ Moderate-Large effect sizes per Cohen conventions. All scaled and contrast measures have normative mean values of 10. “Scaled” scores are scaled according to age norms.

^*CWIT contrast measures are better the closer they are to 10.

Table 4 shows correlations between levels of IL-10, IL-9, IL-7, TGFα and Interferon γ and the cognitive variables with moderate to large effect sizes with wide confidence intervals determined from the multivariate regression analysis. The correlation matrix revealed moderate-large and significant correlations between continuous levels of both IL-10 and IL-9 and BPST total correct trials, and between interferon γ and delayed recall section on the MoCA.

Table 4.

Correlations between inflammatory biomarkers and cognitive performance in 31 African American women with parental history of AD.

	IFNg	TGFα	IL_7	IL_9	IL_10
Interferon gamma	1.0	0.82*	0.84*	0.71*	0.68*
Transforming Growth Factor alpha	0.82*	1.0	0.75*	0.9*	0.78*
IL_7	0.84*	0.75*	1.00	0.61*	0.67*
IL_9	0.71	0.9*	0.61*	1.00	0.83*
IL_10	0.68*	0.78*	0.67*	0.83*	1.00
TMT Proportion (B-A)/A	0.01	−0.01	0.11	0.11	0.22
Delayed Recall (MoCA)	0.40*	0.24	0.22	0.17	0.21
Brooks Spatial Memory (% Correct)	0.08	−0.01	−0.08	0.06	0.04
BPST Total Correct Trials	0.24	0.32	0.21	0.41*	0.42*
CWIT Inhibition Scaled Score	0.04	−0.09	−0.12	−0.21	−0.34
CWIT Switching Contrast	−0.01	0.11	0.20	0.11	0.27
CWIT Inhibition Total Errors Scaled Score	0.01	−0.08	0.03	−0.03	−0.12
TOL Time/Move Ratio Scaled	0.09	−0.06	−0.03	−0.04	0.14

IL = interleukin. TMT: Trails Making Test; BPST: Body Position Spatial Task; CWIT: Color Word Interference Test; TOL: Tower of London;

^*Significant correlation at alpha < 0.05.

Discussion

In this preliminary study in an underserved population, we present initial evidence that in a small sample of middle aged AA women with a parental history of AD, individuals with IL-10 levels below an established cutoff in the literature, performed better in some aspects of executive function as per some surprisingly moderate-large effects. Although it must be emphasized that there were no statistically significant effects, the CIs for the effects are very large and some significant findings (e.g., the correlations) were contrary to hypotheses. The wide confidence intervals are likely due to the small sample size and the variability of the sample’s performance, both of which tend to widen CIs.

Differences were noted between groups in other inflammatory marker levels that are related to the function of IL-10. IL-7 remained statistically significant after correction for multiple comparison. Further, significant correlations of moderately strong effect sizes were noted between normal levels of IL-10 and better performance on aspects of executive function (switching, inhibition, planning). Interestingly, the opposite was noted with short-term spatial memory, which was significantly associated with higher levels of IL-10. In concert with this finding, a non-significant small correlation (r = 0.21) was noted between delayed recall and higher levels of IL-10, while a significant correlation was noted between delayed recall and interferon γ levels.

It must be acknowledged that analysis reveals cognitive performance of normal versus elevated IL-10 groups was not consistently in the same or predicted directions. In Table 3, of the eight measures with moderate-large effect sizes, performance was better in the elevated IL-10 group on the BPST variables (span and number of correct trials) and TOL Time per move ratio scaled score, which goes against the study’s hypothesis. Conversely, participants with elevated IL-10 levels showed a large effect and poorer performance on the Trail-Making Test compared to participants with normal IL-10 levels. At a mean of 37.3 seconds, individuals in the normal IL-10 group performed at normative levels (Tombaugh, 2004). Individuals in the elevated IL-10 group needed approximately 13 more seconds to complete the task. This large effect was the only one of its size found in the executive function battery, although moderate effects suggesting worse performance among those with elevated IL-10 were also noted in inhibition and switching (CWIT) and planning (TOL). The effects generated from this pilot work will be used to power a future trial that can more definitively provide conclusions; therefore, all results presented here should be interpreted with extreme caution given the small sample.

Although our results are mixed, the largest effect of reduced performance on the Trails proportion score provides initial evidence for the overall hypothesis that increased IL-10 levels may result in decreased executive function performance, particularly in inhibition/switching subdomains. This finding is in keeping with other work by Rasmusson et al. (1998), who used dementia cutoff scores on the Trail Making Test Part B, to correctly classify 89.4% of n (Rasmusson et al., 1998).

Researchers have examined the predictive ability of executive functioning for daily functioning. Executive functioning may influence daily functioning because executive functioning controls other aspects of cognition (Mitchell & Miller, 2008). Significant correlations have been measured between executive function and instrumental activities of daily living as well as leisure and self-development activities (Cornelis et al., 2019). In a study including 50 older adults, investigators measured executive function with the Controlled Oral Word Association test, the Wisconsin Card Sorting Test, and Trail Making Test Part B. The latter test, (TMT-B) and education were the only significant predictors of functional ability (Bell-McGinty et al., 2002). Trail Making Test Part B was also the best predictor of daily functioning compared to the Tower of London test and measures of fluency (Mitchell & Miller, 2008). Performance on the Trail Making Test has also been significantly correlated with impaired driving by older drivers on road tests (Papandonatos et al., 2015). In a sample consisting of normal controls, individuals with mild cognitive impairment, and mild AD patients, the Trails difference score was significantly associated with instrumental activities of daily living (Marshall et al., 2011). All this said, the correlations between TMT and the biomarkers of interest were not strong, suggesting this variable works well with stratified groups on specified cutoffs in this instance.

In this study, set shifting was isolated with the Trail Making Test proportional score. The moderate effect noted in the CWIT Switching contrast score, on which the normal IL-10 group did best, also provides a measure of set-shifting. Previous studies have suggested a relationship between set-shifting and functional status. Nguyen et al. (2019) measured components of executive function including set shifting, verbal abstraction, nonverbal abstraction, and verbal fluency. Only set shifting and verbal abstraction significantly contributed to the functional status (measured with the Texas Functional living scale) (Nguyen et al., 2019). Processing speed may play an explanatory role in the connection between the Trail Making Test and activities of daily living. Trail Making Test Part B is highly associated with measures of processing speed, which if slow and impaired, can limit independent activities of daily living (Karzmark & Deutsch, 2018).

Memory was examined with an extremely limited assessment: only the delayed recall item from MoCA was evaluated. Although this item showed no moderate or greater effects in stratified groups, as mentioned above, better delayed recall had a nonsignificant correlation with increased IL-10 levels, as well as a significant correlation with interferon gamma levels. The latter finding is deserving of further inquiry in this population. Future studies will include additional measures of memory and language. Short term/working memory was examined with a limited battery of spatial working memory measures. The measure, BPST, involved remembering instructions for a series of steps and turns. This measure had moderate effects in the stratified groups and significant associations with both IL-10 and IL-9 levels. Interestingly, in correlational analysis, higher levels of IL-10 led to better performance, not worse, on BPST, a test of motor-cognitive integration involving short term memory, in contrast to the lesser performance noted on tests of executive function. More research is necessary to verify biomarker relationships with aspects of memory to determine causality.

Several inflammatory factors were significantly different in levels between groups, with large effect sizes (d > 1.0). The largest effect size (d = 1.57) was found in IL-7, the only inflammatory marker to remain significant after adjustment and multiple comparison correction. IL-7 is involved in B and T cell development and differentiation. Increased IL-7 levels have been linked to increased depressive symptoms and falls in older adults (Britton et al., 2019), events which have a relationship to executive function. However, other work has shown no link between IL-7 levels and dementia status (Kilic et al., 2018). No strong correlations were noted between IL-7 and cognitive performance. Interferon γ (d = 1.1), a cytokine secreted by activated T cells and natural killer cells, promotes macrophage activation, coordinates innate immune system activation, and controls cellular apoptosis (Tau & Rothman, 1999). Interferon γ also had a significant moderately strong correlation with the delayed recall item on the MoCA. Another with a large effect size (d = 1.0), transforming growth factor α binds to epidermal growth factor receptors, induces cell proliferation events such as wound healing and embryogenesis (McInnes et al., 1998), and indirectly stimulates luteinizing hormone-releasing hormone (Ojeda et al., 1997). IL-9 may be an important marker for this population given that our work has previously shown that AD diagnosis correlated with increased IL-9 levels in AAs but not in Whites (Wharton W, Kollhoff AL, Hu W et al., Ann Neurol 2019). IL-9 was strongly correlated with IL-10 levels and was moderately-strongly correlated with BPST, as was IL-10. IL-10 inhibits production of interferon γ, (Oral et al., 2006), yet the correlation between IL-10 and interferon gamma was large. Macrophage derived chemokine, CRP and SAP were not found to be significantly different between groups, which may be explained by their involvement in acute phase inflammation (Pepys & Hirschfield, 2003), as opposed to interleukins which are involved in leukocyte communication (Akdis et al., 2016).

Regarding the correlations, which should be interpreted very cautiously considering the small sample size, while the executive function and memory (spatial and delayed recall) directions of better performance seem conflicting, these results align with previous studies on IL-10 levels and cognitive function, which largely present equivocal findings. Tegeler et al. (2016) previously reported significant inverse correlations between IL-10 levels and executive function (Tegeler et al., 2016). However, the average age of Tegeler’s sample was 68 ± 3.6 years, a decade older than the present study’s sample (58.9 ± 8 years). The role of inflammation may not yet have fully been expressed in this younger sample.

In summary, contrary to the study hypothesis, it must be acknowledged it was not always the case that the group with “normal” IL-10 performed better than the elevated IL-10 group. Better performance may have been observed in the elevated IL-10 group on the memory measures, and some but not all aspects of the CWIT and TOL. The correlations also pointed to increased performance with elevated IL-10 levels in body position spatial memory (significant correlation) and MoCA (non-significant). The inconsistent pattern indicates that considerably more work and replication are necessary to understand the complex relationship between IL-10 and other biomarker levels and higher order cognitive performance in AA women at risk of developing AD. At this stage, TMT findings may be the least ambiguous, at least in the present context.

Limitations

This study had several limitations, among them design weakness, which mandates that all these findings be interpreted cautiously. As pilot work, the study is underpowered to detect effects with certainty due to the small sample size in both groups. The ability to detect significant correlations was limited in this study given the small sample. Although the number of comorbidities was used as covariate, diabetes and hypertension affected a proportion of participants in this cohort, and both affect biomarker levels. These comorbidities may have impacted the results of this study in unknown ways. The effects of common medications such as statins also affects inflammatory levels, which was not considered here but will be in future studies. Although these adult women were children of parents with AD and some of whom regularly provided some or considerable caregiving, caregiver status was not considered in these analyses. Many studies have suggested a link between family caregiving and adverse mental and physical health consequences. Other studies show family caregiving may benefit health and well-being, given that several population-based studies found longer lifespans for family caregivers compared to individuals who did not have family caregiver responsibilities (Roth et al., 2019). Regardless, AD caregiving is associated with higher levels of subjective stress measures, as well as elevated levels of C-reactive protein and inflammatory cytokines (Aschbacher et al., 2007; Gouin et al., 2012). Future studies in adult children with parental history should evaluate the effects of caregiving. Another limitation is having examined only aspects of executive function, short term/working memory and visuospatial function in this cohort. Importantly, memory and language testing must be included in future studies. Further, genetic data were available in only 2 participants, allowing minimal conclusions to be made about the importance genetic components to developing AD and cognitive performance. ApoE status will be taken into consideration in a future trial. The analyses used in this study were all linear. Potentially there is a non-linear relationship between IL-10 and cognition in this population. Non-linear analyses could be explored in future trials. Lastly but importantly, insufficient evidence supports whether or not cultural biases exist for the neuropsychological battery performed. However, Aiken Morgan et al., found little evidence of test bias in an African American sample participating in the large ACTIVE trial. Because only African Americans were included in the present study, the study is less vulnerable to effects of any cultural biases if they exist (Aiken Morgan et al., 2010).

In conclusion, we present preliminary evidence that may support reduced cognitive performance in executive function, specifically set switching, in a female cohort with elevated IL-10 levels at risk for AD. Other biomarkers, including IL7 and interferon gamma also appear to be candidates deserving of further study in this population, in connection to cognitive performance. We also presented evidence that continues to make conclusions equivocal regarding elevated IL-10. For example, spatial memory may be beneficially affected by higher levels of IL-10. The findings of this study emphasize the necessity for further studies on inflammation, executive function, and memory in African American individuals. Future, properly powered studies are needed. Effect sizes generated from this pilot study will be used to power future trials with adequate sample sizes to make more definitive conclusions about the relationship between inflammatory biomarker levels, AD, race, sex and cognitive performance.

Data availability

This trial is registered under Clinical Trials Registry: NCT03269149. Data will be made available upon reasonable request from qualified investigators.