Relationship Between Adipokines, Cognition, and Brain Structures in Old Age Depending on Obesity

Abstract

Background

Adipokines such as leptin and adiponectin are associated with cognitive function. Although adiposity crucially affects adipokine levels, it remains unclear whether the relationship between adipokines and cognition is influenced by obesity.

Methods

We enrolled 171 participants and divided them into participants with obesity and without obesity to explore the effect of obesity on the relationship between adipokines and cognition. In addition to plasma levels of leptin and adiponectin, multidomain cognitive functions and brain structures were assessed using neuropsychological testing and magnetic resonance imaging. Association between levels of these adipokines and Alzheimer’s disease (AD) was then assessed by logistic regression.

Results

We found that cognitive function was negatively associated with leptin levels and leptin-to-adiponectin ratio (LAR). Such correlations between leptin and cognitive domains were prominent in participants with obesity but were not observed in those without obesity. Leptin levels were associated with lower hippocampal volumes in participants with obesity. A significant interaction of leptin and obesity was found mostly in the medial temporal lobe. Both leptin and LAR were positively associated with insulin resistance and inflammation markers in all participants. Of note, LAR was associated with a higher risk of AD after adjusting for demographic variables, Apolipoprotein E genotype, and body mass index.

Conclusions

Obesity might be a factor that determines how adipokines affect brain structure and cognition. Leptin resistance might influence the relationship between adipokines and cognition. In addition, LAR rather than each adipokine levels alone may be a better indicator of AD risk in older adults with metabolic stress.

Adipokines, Adiponectin, Alzheimer’s disease, Brain atrophy, Leptin, Obesity

Background

Obesity is associated with a high risk of dementia (1), and this relationship becomes complicated in older age. Although central obesity during midlife is associated with a greater risk of dementia in later life (2), weight loss is often observed during premorbid stages in individuals with Alzheimer’s disease (AD) (3,4). Such changes in body weight in at-risk individuals may reflect deregulation in the homeostatic feedback mechanism associated with neurodegeneration (5,6). In fact, neurodegeneration coincides with changes in the levels of adipokines, which are also known to modulate bioenergetics and metabolism in the brain (7). Therefore, adipokines have emerged as a potential mediator of AD (8).

Adipose tissue has been identified as an important endocrine organ secreting various signaling mediators that regulate metabolism, insulin signaling, and inflammation (9,10). Among such adipocyte-derived hormones (adipokines) are leptin and adiponectin. Leptin acts on hypothalamic neurons to maintain a balance between food intake and energy expenditure (11). This hormone also enhances neurogenesis and synaptic plasticity, thereby indicating its involvement in learning and memory (12). The expression of leptin receptors in the hippocampus and cortex (13) strongly suggests that the action of leptin might play a role beyond systemic energy homeostasis. Adiponectin is another major adipokine that regulates insulin sensitivity, glucose homeostasis, and inflammation (14,15). Low adiponectin levels are associated with the onset of obesity, dyslipidemia, and dementia (16). Moreover, the stimulation of adiponectin receptors in the brain improves cognition and ameliorates pathologies of AD (17).

Recently, the plasma leptin-to-adiponectin ratio (LAR) has been proposed as a better indicator of adipokine dysregulation than measures of either of the adipokines alone. This ratio serves as a useful marker for detecting metabolic syndrome related to insulin resistance (18). Furthermore, the fact that these adipokines can cross the blood–brain barrier (BBB) (19) may provide clues regarding the cross talk between impaired metabolism and cognition.

Despite theoretical relevance, only a few studies have investigated the relationship between plasma levels of these adipokines and the onset of AD in the context of adiposity, and they have shown conflicting results (20). Because leptin reactivity decreases with obesity or increased age, leptin signaling cannot exert a beneficial effect on brain function, even with chronically increased leptin levels, a state which is referred to as “leptin resistance” (21). Therefore, it is plausible to assume that the role of adipokines in cognitive function may differ by adiposity, especially in old age. In this study, we aimed to investigate the relationship between adipokines, cognition, and brain structure in the old age and whether these relationships are affected according to adiposity. We hypothesized that adipokines are differently associated with cognitive impairment by the presence of obesity.

Materials and Methods

Study Population

A total of 171 participants were recruited from the memory clinic of a university-affiliated general hospital and a local dementia center in accordance with the approval of the Institutional Review Board (IRB 4-2021-0261). The inclusion criteria were age 50 years or older and Clinical Dementia Rating (CDR) score of ≤2.0. The exclusion criteria were a history of major psychiatric or neurological illnesses such as schizophrenia, recurrent depressive disorder, stroke, head trauma, or epilepsy. Participants who could not read or write, and those with significant visual/hearing difficulty were also excluded. All the participants were aged 52–95 and the mean age of participants was 74.30 ± 6.63 years. All participants were comprehensively evaluated at baseline to obtain details regarding their medical history, anthropometric measurements (weight and height, body mass index [BMI], waist-to-hip ratio), vital signs, and life style (smoking and alcohol consumption). We categorized the participants into 2 groups according to their BMI, based on the Asia-Pacific cutoff points (22): participants without obesity (BMI < 25 kg/m²) and with obesity (BMI ≥ 25 kg/m²).

Blood Analyses

Overnight fasting blood samples were collected in the morning from the antecubital vein. Blood samples were transported to the laboratory to perform blood cell counting, clinical chemistry, and Apolipoprotein E (APOE) genotyping. The APOE ε4 genotypes were assessed using methods described previously (23). Plasma was immediately prepared by centrifugation (2 000g) for 15 minutes from the whole blood collected in heparin tubes after gentle mixing. The plasma was allocated into several microcentrifuge tubes (0.7 cc in each), and the tubes were immediately stored at −80°C until further analysis. The homeostasis assessment of insulin resistance (HOMA-IR) was calculated using the standard formula (24). The systemic inflammatory state was indexed by the C-reactive protein (CRP) level. Both plasma adiponectin and leptin levels were determined using enzyme-linked immunosorbent assay kits (DRP300, R&D Systems, Minneapolis, MN; DLP00, R&D Systems, respectively).

Assessment of Cognitive Functions

The Korean version of the Consortium to Establish a Registry for Alzheimer’s disease (CERAD-K) was used to assess multiple cognitive domains: attention, language, visuospatial, memory, and executive function (25). The Korean version of the Boston Naming Test was used to measure language function. Attention was assessed using the Trail Making Test A, and frontal/executive function was assessed using the ratio of the Trail Making Test B/A (26). All items in the CERAD-K were standardized as z-scores for age, sex, and educational level using reference values from a Korean-speaking population (25). The global cognitive status was assessed with the Mini-Mental State Examination (MMSE), and dementia severity was measured using the CDR and CDR-sum-of-box (CDR-SB) scores. The diagnosis of AD was made by 2 board-certified psychiatrists according to the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria (27). Functional levels were assessed using the Barthel Activities of Daily Living Index, and depression was screened using the Korean version of the Geriatric Depression Scale-Short Form.

Magnetic Resonance Imaging Acquisition and Processing

Structural brain magnetic resonance imaging (MRI) scans were acquired with a Philips 3.0T scanner (Intera Achieva; Philips Medical Systems, Best, The Netherlands). A standardized imaging protocol consisting of whole-brain axial and coronal fluid-attenuated inversion recovery (FLAIR; TR 11 000 ms, TE 346 ms) and T1-weighted (TR 9 300 ms, TE 4.6 ms) sequences was used.

Image Processing

We used a fully automated preprocessing procedure implemented in CAT12 r1450 (Computational Anatomy Toolbox, Structural Brain Mapping Group, Departments of Psychiatry and Neurology, Jena University Hospital, http://dbm.neuro.uni-jena.de/cat/) to apply a standardized analysis pipeline. First, a spatial-adaptive nonlocal means denoising filter was employed. Segmentation algorithms based on the adaptive maximum, a posterior technique implemented in CAT12, were used to classify brain tissue into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). Additionally, partial volume estimation was performed to create a more accurate segmentation for the 2 mixed classes: GM–WM and GM–CSF. After segmentation, the segmented image of the GM tissue was spatially transformed while preserving the total amount of GM signal in the normalized partitions. In the bilateral hippocampus and entorhinal region, we set the regions of interest (ROIs) defined by the Neuromorphometrics atlas (http://www.neuromorphometrics.com). The GM densities were extracted within the ROIs and proportionally adjusted for total intracranial volume (TIV). To identify white matter hyperintensity lesions (WMHL), a 3D T2-weighted FLAIR sequence was used. The segmentation of WMHL was processed using a previously reported automatic algorithm (28). In the present study, we modified 2 procedures that we previously followed: (a) an optimal threshold of 70, instead of 65 in the original study, was applied for suitability; and (b) diffusion-weighted images were not used in this study because participants with an acute infarct were excluded. The volumes of WMHL were calculated using lobar ROIs in the native space of each participant and corrected to the permillage of TIV to adjust for head size.

Visual Ratings for Brain Pathology

The assessment was performed by 2 experienced independent observers blinded to the clinical diagnoses. T1-weighted coronal images were used to assess and rate medial temporal lobe atrophy (MTA), using a 5-grade scale, from 0 (no atrophy) to 4 (severe atrophy), which is based on the width of the coronal fissure, the temporal horn, and the height of the hippocampal formation (29). None of the participants had fourth-degree MTA. WM hyperintensities were rated according to the Fazekas scale (30). Areas of periventricular hyperintensity (PVH) and deep WM hyperintensity (DWMH) were rated from 0 (none) to 3 (severe).

Statistical Analyses

Statistical analyses were conducted with the Statistical Package for the Social Sciences, version 25.0 (SPSS Inc., Chicago, IL), and GraphPad Prism software, version 8.02 (San Diego, CA). Data were expressed as mean (standard deviation) and tested for normality using Shapiro–Wilk’s test. Pearson’s chi-square test was used for testing distribution by groups. Between-group comparisons were conducted by t test or Mann–Whitney U test according to the normality of data. Spearman’s correlation coefficients (r) were used to describe the association between plasma adipokine levels and other continuous variables of interest; the partial correlation coefficient was used to control for the effects of age, sex, education, APOE ε4 genotype, and lifestyle variables (smoking and alcohol consumption). Multiple linear regression analyses were performed with age, sex, education, APOE ε4 genotype, BMI, and lifestyle variables as covariates to test the association of adipokine levels and clinical dementia severity parameters (MMSE, CDR-SB).

To test the influence of leptin resistance on neurocognitive outcomes, we performed stratified analyses in people with obesity and people without obesity. To control Type I error of multiple comparison, the false discovery rate was used to calculate adjusted p value. The voxel-wise General Linear Model (GLM) analysis was used to test the distinct effect of leptin level across the adiposity subgroups (nonobese group and obese group). The GLM fitting was conducted with the following formula: Y = b₀ + b₁(obese) + b₂(nonobese) + b₃(obese ~ leptin) + b₄(nonobese ~ leptin) + b_i(covariates). While the group regressors (b₁, b₂) represented the main effect of group difference, the other 2 regressors of interest (b₃, b₄) represented the slope of the leptin effect in each of the adiposity subgroups. Primarily, the main effect of b₃ (leptin slope in obese group) and b₄ (leptin slope in nonobese group) was tested (p < .05, uncorrected). The interaction effect of obesity × leptin on regional GM was tested by comparing the slope effects of each obesity group (b₃ − b₄). We calculated odds ratios (ORs) and 95% confidence intervals of AD (non-AD vs AD) by adipokines using binary logistic regression, adjusting for age, sex, education, APOE genotype, BMI, and lifestyle variables. Patients with mild cognitive impairment diagnosed according to Petersen’s criteria (31) and the normal cognitive group was considered as non-AD group. Statistical significance was set at p = .05.

Results

Characteristics of the Participants

Table 1 shows the descriptive characteristics of participants with and without obesity. The participants with obesity compared with those without obesity showed significantly higher values in levels of leptin and LAR and lower levels of adiponectin. These participants also had higher levels of glycosylated hemoglobin, HOMA-IR, and CRP. There were no significant between-group differences in age, sex, and education years. There was no significant difference in the levels of adipokines according to the presence or absence of AD (Supplementary Figure S1).

Table 1.

Characteristics of the Study Population, Stratified by BMI

Variables	Total (n = 171)	Without Obesity (n = 111)	With Obesity (n = 60)	p Value
Age, years†	74.3 (6.63)	74.98 (6.54)	73.03 (6.68)	.07
Female	123 (71.9)	78 (70.3)	44 (73.3)	.67
Education level, years‡	7.76 (4.9)	8.27 (4.93)	6.82 (4.73)	.76
HbA1c‡	6.23 (1.01)	6.08 (0.77)	6.52 (1.31)	.047*
HOMA-IR‡	2.70 (2.05)	2.16 (1.30)	3.74 (2.73)	<.001*
BMI‡	23.91 (3.05)	22.20 (2.11)	27.07 (1.67)	<.001*
Waist-to-hip ratio‡	0.92 (0.08)	0.91 (0.89)	0.93 (0.07)	.018
Waist circumference†	86.65 (0.72)	83.45(8.25)	92.69(7.29)	<.001
Hip circumference†	94.32 (0.60)	91.43 (6.43)	99.79 (6.33)	<.001
APOE ε4 carrier	46 (26.9)	36 (32.4)	10 (16.7)	.15
Total cholesterol, mg/dL†	182.75 (40.40)	182.89 (40.11)	182.50 (41.27)	.95
HDL cholesterol, mg/dL‡	51.23 (17.32)	51.02 (13.82)	51.62 (22.55)	.38
LDL cholesterol, mg/dL†	112.43 (35.12)	112.62(34.71)	112.07(36.16)	.92
Triglyceride, mg/dL‡	123.02 (76.39)	120.31(73.80)	128.03(81.36)	.92
CRP, nmol/L‡	0.31 (1.46)	0.13 (0.20)	0.65 (2.43)	<.002*
MMSE score‡	21.24 (5.41)	21.58 (5.38)	20.60(5.43)	.26
CDR-SB score	3.09 (2.42)	3.12 (2.50)	3.03 (2.27)	.75
Alzheimer’s disease diagnosis	82 (48.0)	52 (46.8)	30 (50.0)	.69
Hypertension, N (%)	84 (49.1)	46 (41.8)	38 (63.3)	.01*
Diabetes mellitus, N (%)	50 (29.2)	27 (24.5)	23 (38.3)	.06
Dyslipidemia, N (%)	40 (23.4)	28 (25.5)	12 (20.0)	.42
Smoking				.91
None	139 (81.3)	91 (87.5)	48 (85.7)
Past	15 (8.8)	9 (8.7)	6 (10.7)
Current	6 (3.5)	4 (3.8)	2 (3.6)
Alcohol use				.26
None	135 (78.9)	91 (87.5)	44 (78.6)
Past	15 (8.8)	7 (6.7)	8 (14.3)
Current	10 (5.8)	6 (5.8)	4 (7.1)
Barthel Activities of Daily Living†	19.50 (1.73)	19.54 (1.86)	19.43 (1.47)	.70
SGDS-K score†	6.23 (3.99)	6.07 (4.04)	6.54 (3.91)	.47
Leptin, ng/mL‡	10.41 (10.11)	6.63 (5.19)	17.41 (12.95)	<.001*
Adiponectin, μg/mL‡	8.74 (6.15)	9.45 (6.64)	7.43 (4.89)	.08
LAR‡	1.95 (2.38)	1.25 (1.54)	3.20 (3.07)	<.001*

Notes: Data are presented as mean (standard deviation) or no. (%). Obesity (BMI ≥ 25 kg/m²) according to the Asia-Pacific Classification. APOE = Apolipoprotein E; BMI = body mass index; CDR-SB = Clinical Dementia Rating-sum-of-box; CRP = C-reactive protein; HbA1c = glycosylated hemoglobin; HDL cholesterol = high-density lipoprotein cholesterol; HOMA-IR = homeostatic model assessment for insulin resistance; LAR = leptin-to-adiponectin ratio; LDL cholesterol = low-density lipoprotein cholesterol; MMSE = Mini-Mental State Examination; SGDS-K = Geriatric Depression Scale-Short Form, Korean version.

^†Data were tested using the Student’s t test.

^‡Data were tested using the Mann–Whitney U test.

*p < .05 was considered statistically significant after the Student’s t test or Mann–Whitney U test.

Adipokines and Cognitive Function

In all participants, significant correlations were observed between the levels of adipokines and scores from each domain of cognition, and these associations were found mainly for the levels of leptin and LAR (Table 2). Regarding adiponectin, such associations were not clearly observed. Among the 5 domains within CERAD-K, attention, visuospatial function, and executive function were negatively correlated with leptin levels, whereas the memory domains showed no correlation with any of the adipokines (Table 2). When total participants were stratified by adiposity (111 participants without obesity vs 60 with obesity), significant associations between leptin and cognitive domains were strengthened in those with obesity. These associations were either weakened or not maintained in those without obesity. Because patient with AD were included in both obese and nonobese group, we further performed subgroup analysis by excluding patients with global CDR > 1 corresponding to moderate to severe AD. The main finding in this subgroup analysis remains consistent (Supplementary Table S1). These results suggest that the relationship between adipokines and cognitive function may be dependent on the presence of obesity.

Table 2.

Correlation Between Plasma Adipokine Levels and Cognitive Outcomes by Obesity Status

	Total (n = 171)			Without Obesity (n = 111)			With Obesity (n = 60)
	r	p Value	Adjusted p Value*	r	p Value	Adjusted p Value*	r	p Value	Adjusted p Value*
Leptin
Attention	−0.253	.001	.004*	−0.167	.080	.320	−0.493	<.001*	<.001*
Visuospatial function	−0.319	<.001	<.001*	0.225	.018	.144	−0.400	.002*	.008*
Executive function	−0.171	.027	.072	−0.030	.756	1.008	−0.376	.003*	.008*
Language	−0.102	.184	.294	−0.022	.816	.933	−0.223	.087	.116
Immediate recall	0.056	.464	.530	0.122	.201	.536	−0.019	.883	.883
Delayed recall	−0.010	.900	.900	−0.030	.752	1.203	−0.081	.539	.616
MMSE score	−0.110	.153	.306	0.014	.882	.882	−0.260	.045*	.072
CDR-SB score	0.064	.404	.539	−0.059	.537	1.074	0.313	.015*	.030*
Adiponectin
Attention	−0.130	.090	.360	−0.146	.126	.252	−0.111	.399	1.064
Visuospatial function	−0.033	.664	1.062	−0.019	.847	.847	−0.061	.644	1.030
Executive function	−0.018	.817	.817	−0.038	.698	.797	0.005	.973	.973
Language	−0.154	.044	.352	−0.140	.144	.230	−0.197	.132	1.056
Immediate recall	0.036	.782	.894	−0.210	.027	.216	0.074	.575	1.150
Delayed recall	0.210	.107	.285	−0.115	.231	.308	0.130	.323	1.292
MMSE score	0.056	.673	.897	0.155	.105	.280	0.018	.892	1.019
CDR-SB score	0.050	.518	1.036	−0.181	.058	.232	0.021	.874	1.165
LAR
Attention	−0.133	.082	.218	−0.026	.788	.900	−0.343	.007*	.028*
Visuospatial function	−0.230	.002	.016*	−0.152	.112	.448	−0.265	.041*	.082
Executive function	−0.135	.081	.324	0.001	.991	.991	−0.336	.009*	.024*
Language	−0.005	.953	.953	0.065	.496	.793	−0.015	.911	1.041
Immediate recall	0.101	.189	.378	0.204	.032	.256	−0.047	.723	.964
Delayed recall	0.005	.947	1.082	0.040	.680	.906	0.005	.970	.970
MMSE score	−0.037	.628	.837	0.118	.217	.578	−0.225	.084	.134
CDR-SB score	0.042	.584	.934	−0.118	.217	.434	0.359	.005*	.040*

Notes: Participants were stratified into 2 groups: without obesity (BMI < 25 kg/m²) and with obesity (BMI ≥ 25 kg/m²) according to the Asia-Pacific Classification. BMI = body mass index; CDR-SB = Clinical Dementia Rating-sum-of-box; LAR = leptin-to-adiponectin ratio; MMSE = Mini-Mental State Examination.

*The asterisk identifies the false discovery rate-adjusted p values that are significant at the .05 level.

Adipokines and Brain Structural Changes

In all participants, there were no correlations of the levels of leptin and LAR with total hippocampal volume, while the levels of adiponectin were correlated (r = −0.176, p = .021). When stratified by obesity, the levels of leptin were not correlated with total hippocampal volume in participants without obesity, but its levels were inversely correlated with the volume in those with obesity (r = −0.260, p = .045). The levels of adiponectin showed negative correlations with the hippocampal volume only in participants without obesity. LAR was positively associated with the total hippocampal volume in participants without obesity, but not in those with obesity (Figure 1). Leptin and adiponectin levels were negatively associated with the volume of the entorhinal cortex in participants with obesity, while not in those without obesity (Figure 2). We further explored potential relationships between the levels of adipokines and brain structural changes such as MTA, PVH, and DWMH, which are established imaging biomarkers for neurodegeneration. We found significant interactions between leptin and BMI (β = 1.239, p = .040) and between LAR and BMI (β = −0.959, p = .040) and MTA, mostly in the right hemisphere. Significant associations between the levels of leptin and MTA became evident only in participants with obesity after adjusting for age, sex, education, APOE ε4, and lifestyle variables (right MTA: r = 0.365, p = .009; left MTA: r = 0.294, p = .038). However, these associations were not found in participants without obesity (Supplementary Table S2). We used whole-brain voxel-based morphometry analysis to further explore the relationship between the levels of leptin and GM volume. There was a different pattern associated with obesity when applied with a lenient statistical threshold for visualization purposes (p < .05, uncorrected, Supplementary Figure S2). Of note, the interaction between obesity and leptin showed a greater effect on the medial temporal lobe than the whole brain, and the effect in participants with obesity exceeded the effect in participants without obesity, p < .005, k > 100 (Figure 3). These findings suggest that leptin might be associated with brain structural changes in an adiposity-dependent manner.

Figure 1.

Higher leptin levels are associated with lower hippocampal volumes in participants with obesity. Bilateral hippocampal regions of interest overlaid onto a T1 image (A). The right and left hippocampal regions are marked in green and red, respectively. (B and C) Relationship between plasma adipokines and hippocampal volume according to obesity status. BMI = body mass index.

Figure 2.

Relationship between the entorhinal cortex and adipokine levels. Bilateral entorhinal regions of interest overlaid onto a T1 image (A). The right and left entorhinal cortices are marked in green and red, respectively. (B and C) Relationship between plasma adipokines and entorhinal cortex volume according to obesity status. BMI = body mass index.

Figure 3.

Brain regions showing the interactive effect of leptin and obesity on gray matter volumes (obese group effect > nonobese group effect, p < .005, k > 100).

Adipokines, Insulin Resistance, and Inflammation

Inflammation and metabolic dysfunction are key risk factors for age-associated cognitive impairment and AD. Hence, we further explored the relationship between adipokines and the metabolic and inflammatory index. Leptin and adiponectin levels and their ratio (LAR) were significantly associated with insulin resistance in all participants: HOMA-IR and fasting C-peptide (Supplementary Table S3). When stratified by obesity, both leptin levels and LAR were positively associated with higher CRP levels in participants with obesity (r = 0.450, p < .001; r = 0.445, p < .001, respectively). In participants without obesity, only LAR was associated with CRP levels. Statistical significances of these findings were maintained after adjusting for age, sex, and APOE ε4 genotype.

Adipokines and AD

As we found that the levels of leptin and LAR were closely associated with brain structure as well as with insulin resistance and inflammation, which are common pathological features related to AD development, we conducted logistic regression to determine the association of AD with adipokine levels. Higher LAR was associated with higher odds of AD diagnosis (non-AD vs AD, adjusted OR = 1.41, p = .042, Supplementary Table S4) after adjustment of age, sex, education, APOE ε4 genotype, BMI, and lifestyle variables. No significant association was noted between AD and adiponectin or leptin levels. Multiple linear regression models including demographic variable, APOE ε4 genotype, lifestyle variables also showed significant association with LAR with CDR-SB (β = 0.382, p = .002), but not with MMSE score (β = −0.191, p = .079). These findings suggest that LAR might be a potential marker for indicating the risk of AD better than each adipokine level alone in the elderly.

Discussion

We found that higher leptin levels and LAR were significantly associated with poor cognition and neuroimaging markers of neurodegeneration, and such associations were more evident in participants with obesity rather than in those without obesity. Furthermore, a higher LAR was associated with higher odds of AD. Consistently, leptin and LAR were significantly associated with insulin resistance and inflammation, which are key pathogenic features of AD (32). To the best of our knowledge, this is the first study to reveal an interactive relationship between adipokines and cognitive function as well as structural changes of the brain according to the presence of obesity. These results suggest that leptin and LAR could be an important factor for AD risk in conditions of metabolic stress such as obesity.

AD is a multifactorial disorder with various risk factors, among which metabolic dysfunction such as obesity has long been investigated as a modifiable target for aging-associated cognitive decline (33). Indeed, obesity in middle age is associated with cognitive decline in later life. Given that obesity is associated with leptin resistance, dysfunctional action of leptin in the brain may be one of attributes for the obesity-related increase in the risk of AD. Leptin modulates cognition by regulating hippocampal synaptic plasticity (34). Previous animal studies also demonstrated the neuroprotective effects of leptin on various neurodegenerative processes; it minimizes the neuronal damage induced by ischemia (35) and reduces Aβ toxicity by suppressing β-secretase activity (36). However, epidemiological studies regarding leptin have often reported conflicting results (37,38), and the role of leptin in AD pathogenesis is not fully understood. Adiponectin also affects cognitive function by modulating insulin sensitivity and inflammation. Contrary to the expectation that adiponectin may play a protective role in cognition (39), previous epidemiological studies have also shown contradictory results regarding the relationship between adiponectin and AD (40,41). These conflicting results may stem from heterogeneity in the study populations such as different degrees of adiposity.

A recent study has shown that leptin regulates adiponectin expression (42). Therefore, an increase in leptin alone without accompanying adiponectin increase (ie, high LAR) may suggest the presence of impaired leptin signaling and subsequent adipokine dysregulation. LAR is a clinically useful marker indicating insulin resistance and atherogenic risk (43). However, only a few studies have evaluated LAR as a potential indicator of cognitive impairment. In this study, we observed that LAR was positively correlated with hippocampal volume in participants without obesity. In addition, leptin levels and LAR were positively associated with dementia severity scales only in participants with obesity. Another study recently reported that inefficient leptin signaling could partly contribute to cognitive decline through alteration in the hippocampal structure (44). Neuronal loss and pathological protein formation such as neurofibrillary tangles and amyloid plaques are observed primarily in the entorhinal cortex in early AD and gradually project to the hippocampus. In participants with obesity, both leptin and adiponectin levels were significantly associated with the volume of the entorhinal cortex, whereas in the nonobese group, there was no association. Thus, the impact of these adipokines on the medial temporal lobe may differ by leptin resistance. Our results are supported by previous findings that high leptin levels are associated with an increased risk of dementia in women with obesity, while leptin lowers the risk of dementia in women with normal BMI (45).

Leptin resistance is found in obesity, with a decrease in the relative transport of circulating leptin across the BBB and a decrease in intracellular signaling downstream of the leptin receptor (21). Given that leptin is involved with hippocampal-dependent learning and memory (46), leptin resistance may be associated with cognitive decline. A recent European study reported that the association between leptin and cognitive impairment was observed only in the nonobese group (BMI < 30 kg/m²), which contradicts our findings (47). However, another study in the United States showed that leptin was significantly associated with poor cognitive outcomes in the obese or overweight group, which is consistent with our findings (48). We additionally confirmed, through MRI, that plasma leptin levels were associated with lower hippocampal volume in participants with obesity. On the contrary, the negative correlation between adiponectin levels and hippocampal volume, which was also reported previously (41), was remarkable in those without obesity. According to these findings, although the obesity-dependent pattern in the brain was not reflected as clearly in the neuropsychological tests, the relationship among adipokines, cognitive function, and brain structure appears different depending on the adiposity of individuals. It is also plausible to assume that adipokines may have a hormonal effect on neurocognitive states rather than directly causing structural changes in the brain.

Although we found several cognitive domains that were associated with leptin/LAR, there still were other domains that were not. This suggests that adipokines’ contribution to brain pathology might be in a region-specific manner. Thus, even if it is true that adipokines affect general cognitive/brain health status, specific relations might be identified in several, if not all, cognitive domains: In our study, we found significant association between leptin/LAR and the frontal/executive domain. This finding is consistent with previous reports showing that leptin levels were associated with volumes of frontal lobe structures (49) or executive function in the elderly (50). However, our relatively small sample size should be counted as the reason for our findings of correlations between adipokines and only several cognitive domains.

Chronic inflammation observed in obesity is also associated with cognitive impairment. In the present study, we demonstrated that a higher LAR was significantly associated with CRP and HOMA-IR. In addition, LAR was associated with increased odds of AD, and its estimated risk was higher than that associated with either leptin or adiponectin levels alone. Previous studies have suggested that LAR may function as a key indicator in assessing different physiological conditions in addition to insulin resistance (18,43). The balance in leptin and adiponectin levels in individuals, rather than leptin or adiponectin levels alone, may reflect physiological changes preceding neurodegeneration. Although the biological mechanisms linking elevated LAR with AD remain unclear, elevated LAR in patients with AD may reflect adipokine dysregulation related to AD development. Further large-scale, prospective studies are needed to determine whether the dysregulation of adipokines may occur before neurodegenerative changes.

Our study has several limitations. This is a cross-sectional study with measurement of adipokines only once at baseline. Although previous studies have demonstrated that a single measurement of circulating leptin and adiponectin levels is a reliable and sufficient indicator of long-term adipokine levels (51), the current findings warrant replication studies with larger sample size on a serial basis. This is a cross-sectional study with no dimension of time; therefore, it may not suitable to support conclusion on AD risk, nor on casual relationships between adipokines and dementia. However, our findings encourage future studies that adiposity should be considered in prospective studies to explore the possible role of adipokine as a biomarker of cognitive decline. Second, the association of LAR with AD was not fully investigated. Thus, the interpretation of these findings must be made carefully. Given that the correlation of LAR with the memory scores was not significant, it is likely that LAR as a marker for the dysregulation of adipokines may be an indicator related to overall brain degeneration, perhaps not specific to AD. To explore the relationship between LAR and AD, additional studies are needed with measurements of AD-specific biomarkers, such as through amyloid imaging and CSF Aβ and tau assays. Third, we used BMI to define obesity. Obesity would reflect a too complicated condition to apply this simple definition to. Several studies examining the effects of BMI on health outcomes in later life have shown their inverted u-shape relationship (52,53), further questioning the validity of our dichotomous distinction of participants using BMI. In addition, we did not consider overweight, which has a narrower range (23 ≤ BMI < 25 kg/m²) in Asia-Pacific definition than that (25 ≤ BMI < 30 kg/m²) of World Health Organization. These participants remain to be elucidated in future studies with sufficiently large sample size. Fourth, the mean CDR-SB of the participants in our study was relatively high (3.09 ± 2.42). Because the participants were recruited from a local dementia center and a hospital setting, selection bias in population may affect the generalizability of the results. Therefore, replication is required in a large-scale cohort that contains a significant proportion of a well-defined and validated cognitively normal population. Lastly, residual unidentified confounders may influence the relationship between adipokines and cognition. For example, adults with obesity have shown to have decreased levels of brain-derived neurotrophic factor, expression of which is deeply involved in cognitive function (54). Homocysteine could be another factor to consider that has been linked to both obesity and increased risk of dementia (55,56). Therefore, further studies in an independent cohort that are necessary to widen our understanding of complexity by which adipokines and other factors influence brain function.

This study demonstrated that adipokines are associated with cognitive impairment, and this relationship is affected by the presence of obesity. LAR is also associated with an increased risk of AD after adjusting for BMI, suggesting that LAR may reflect the dysregulation of adipokines to which neurodegenerative diseases are predisposed. Additional molecular and epidemiological research is required to elucidate why the role of adipokines in cognition differs by obesity.

Funding

This work was supported by a Research Grant from the Yonjung Association, in part of Institute of Behavioral Science in Medicine at Yonsei University College of Medicine. None of the funding sources played a role in the design, collection, analysis or interpretation of the data, or in the decision to submit the manuscript for publication.

Conflict of Interest

The authors declare that they have no conflict of interest.

Acknowledgments

The authors thank all the participants who voluntarily accepted to take part in this study.

Author Contributions

E.K. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. J.H., S.Y.C., K.Y.K., and H.K. contributed to the concept and design; S.K and J.-Y.L. contributed to the acquisition, analysis, or interpretation of data; J.H., S.K., and E.K. drafted the manuscript; statistical analysis was performed by J.H., and S.K.; and E.K. and J.-Y.L. were responsible for supervision.

Declarations

Ethics approval and consent to participate: The study was approved by the institutional review board of Yonsei University College of Medicine (IRB 4-2021-0261).

Consent for publication: Not applicable.

Data Availability

The data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Author notes

These authors are co-corresponding authors.