Interaction of insula and hippocampus in memory dysfunction in Alzheimer's disease

Abstract

The insula, classically linked with emotion and perception, has recently been associated with memory function in Alzheimer's disease (AD). However, its role in memory remains unclear. This study investigated whether the insula contributes directly to memory consolidation or is indirectly involved in memory-related brain regions. We systematically increased diagnostic specificity in memory-impaired patients to assess the role of the insula in memory. Gyrification of the insula was associated with memory consolidation in Aβ+ individuals, only in combination with hippocampal atrophy. These findings suggest insular atrophy interacts with degeneration of the hippocampus in memory dysfunction in AD, but not necessarily in other pathologies.

Keywords

Alzheimer's disease amyloid-β gyrification insula memory

Introduction

Alzheimer's disease (AD) typically presents with impaired storage and consolidation of new memories, with pathology centering around the entorhinal cortex and hippocampus.^1,2 Atrophy in these regions strongly correlates with disordered memory consolidation, and measures such as hippocampal volume and entorhinal cortical thinning are well-established magnetic resonance imaging (MRI) markers of memory dysfunction.^1–3 More recently, additional measures of cortical surface morphology have been developed. For example, the degree of cortical folding, quantified by the gyrification index (GI), has been used as an additional morphometric measure of AD. Gyrification has been shown to correlate with memory in aging and disease.^4,5 A higher GI might indicate a larger number of neurons and more efficient cortical communication, potentially enhancing cognition. While GI often declines with age and neurodegeneration, some studies report positive associations with cognitive function.^4–6 Nunez et al.⁴ reported that the folding patterns in AD change depending on the extent of memory impairment: the entorhinal cortex demonstrated increased folding, while the insula had lower GI. Notably, individuals with AD who had better memory function had higher insular GI, and individuals with mild cognitive impairment (MCI) showed an intermediate folding pattern between those of healthy and AD participants. While the insula is known to be associated with salience, emotion and perception,⁷ identification of the insula's role with memory is more recent and is less understood. Recently, using direct electrical brain stimulation, interactions between the insula and hippocampus during memory encoding were reported.⁸ Prior literature also implicates the insula in consolidation of memory.⁹ How and why the insula is involved in memory, independent of emotional context, requires further investigation, however.

The current study sought to replicate the finding of the insula's gyrification association with memory and to investigate how insular gyrification may play a role in memory consolidation along the continuum of memory-impaired individuals. We started by examining a large sample of patients clinically diagnosed with AD to determine if the relationship of the insula with memory consolidation is broadly observed in memory impairment. By narrowing the sample from clinically diagnosed individuals to those diagnosed with the addition of AD biomarkers, we sought to determine if the relationship between the insula and memory is specific to amyloid. Furthermore, by concurrently examining the role of well-established brain regions for memory consolidation, we aimed to determine if the insula is an independent factor in memory consolidation.

Methods

Participants

We retrospectively analyzed clinical data from 659 participants seen in a memory clinic and clinically diagnosed with probable MCI or dementia due to Alzheimer's disease (NIA-AA).¹⁰ Fifty-one individuals made up a healthy participant (HP) group. Diagnosis was established by consensus (neurologists, neuropsychologists, neuroradiologists, psychiatrists, geriatricians). For inclusion in the analysis, the California Verbal Learning Test and a high-resolution 3T MRI were required. A subset was utilized with known amyloid-β (Aβ) biomarker status by amyloid PET or cerebrospinal fluid (CSF) (162 participants Aβ+, 61 Aβ−). Abnormal Aβ₄₂ level or ptau/Aβ₄₂ ratio for CSF findings was based upon the standard Mayo Laboratory cutoffs.¹¹ For PET imaging, using ¹⁸F-florbetaben and following ADNI acquisition parameters, standard uptake value ratio (SUVr) was calculated using the cerebellar gray matter to normalize the standard uptake value. We used a cutoff of 1.18 SUVr for participants with symptom onset before age sixty-five and 1.20 for participants with symptom onset after 65.^12,13 This study was approved by the West Virginia University Institutional Review Board and all participants signed informed consent.

Measures

The cohort was clinically characterized with the Mini-Mental State Examination (MMSE),¹⁴ Wide Range Achievement Test, 4^th edition-word reading subtest (WRAT),¹⁵ Neuropsychiatric Inventory Questionnaire (NPI-Q),¹⁶ and the Functional Activities Questionnaire (FAQ).¹⁷ The California Verbal Learning Test-II Short Form (CVLT II-SF)¹⁸ was used to assess memory, with long-delay free-recall (CVLT LD FR) as the dependent variable to examine memory consolidation.¹⁹

MRI data acquisition and processing

Structural MRI was obtained on a Siemens Prisma 3T scanner. A T1-weighted MPRAGE sequence with a minimum of 1 × 1 × 1 mm³ voxel size was acquired using a 20-channel head coil (TR = 2300 ms, TE = 2.26 ms). A comparable 1 × 1 × 1 mm³ T1-weighted sequence was completed on a GE Architect 3T scanner for some participants (TR = 8.5 ms, TE = 3.3 ms). Image quality rating was measured through CAT12, and was comparable between the Siemens and GE scanners (F < 1). Analysis of cortical thickness included the AD signature areas² (middle frontal, superior frontal, middle temporal, inferior temporal, temporal pole, superior parietal lobule, angular gyrus, supramarginal gyrus, precuneus), and entorhinal cortex, as well as GI of the insula, and inferior temporal gyrus.^4,5 Thickness and gyrification were calculated using the standard CAT12 pipeline²⁰ with the DKT40 atlas. Hippocampal volume was calculated using SLANT (spatially localized atlas network tiles).²¹ All measures were averaged across hemispheres and hippocampal volume was corrected for total intracranial volume based on CAT12. We additionally measured white matter hyperintensity (WMH) volumes using the lesion segmentation tool²² based upon 3D FLAIR acquisitions.

Statistical analysis

SPSS 29.0 was used for statistical analyses. Relationships between brain measures (hippocampal volume, entorhinal thickness, insula GI) and CVLT LD FR were first assessed with Pearson's correlations. Stepwise linear regressions were conducted first with individual brain regions, then with calculated interaction terms for the insula GI with hippocampal volume and entorhinal thickness. After examining the whole sample, regressions were repeated in the biomarker cohort (including both positive and negative for amyloid), and then separately in the amyloid-positive and negative sets. Age, sex, education, scanner type, and WMH volume were entered first in the regression, to control for these effects, before the stepwise procedure. We analyzed if there was any difference between amyloid-positive (Aβ+) and amyloid-negative (Aβ−) groups on the brain measures.

Results

Demographic and clinical information, including hippocampal volume, entorhinal thickness, and gyrification index, of the sample is listed in Table 1. Of the 223 participants clinically diagnosed with AD, 162 were Aβ+ and 61 Aβ−. Overall, the sample included more females than males, was well-educated with average estimated intelligence, and exhibited minimal neuropsychiatric symptoms. WMH were common in the sample.

Table 1.

Demographics and neuroimaging characteristics of the combined set (n = 659).

Variables	Mean (SD)
Age	71.5 (8.9)
Sex	296 M / 363F
Education	14.5 (2.8)
MMSE (n = 655)	25.4 (3.7)
WRAT (n = 646)	101.1 (14.5)
NPI (n = 519)	3.8 (2.8)
FAQ (n = 602)	9.3 (8.1)
MRI descriptives (n = 659)
WMH volume	7.6 (10.6)
Hippocampal volume	0.025 (0.0)
Entorhinal thickness	3.1 (0.4)
Insula GI	24.5 (1.9)

Participants with their average demographics and neuroimaging characteristics.

MMSE: Mini-Mental State Exam; WRAT: Wide Range Achievement; NPI: Neuropsychiatric Inventory, total number of symptoms; FAQ: Functional Activity Questionnaire; MRI: magnetic resonance imaging; WMH: white matter hyperintensities volume; GI: Gyrification Index.

Correlations between memory performance (CVLT LD FR) and measures of atrophy in the combined sample are presented in Table 2. While all correlations were significant, the strongest correlation was with entorhinal thickness, followed by insula GI and then hippocampal volume. Linear stepwise regression after controlling for age, sex, scanner type, education, and WMH volume revealed the interaction of insula GI with entorhinal thickness accounted for an additional 9% of variance in predicting memory consolidation (medium effect size of 0.26), while the interaction of insula GI and hippocampal volume accounted for an additional 1% of the variance (Table 3, combined set). This last interaction is not likely meaningful.

Table 2.

Correlations between memory performance and brain measures in the combined set (n = 659).

Area	Correlations
Hippocampal volume	0.20**
Entorhinal thickness	0.33**
Insula GI	0.32**

Memory consolidation as assessed by California Verbal Learning Test long-delay free-recall (CVLT LD FR). GI: Gyrification index.

** p < 0.001.

Table 3.

Brain areas predicting memory consolidation in the combined set (n = 659).

Model	R	R2	SEE	Delta R2	Delta F	Significance	f²
WMH, Sex, Scanner, Education, Age	0.43	0.19	2.7	0.18	30.2	<0.001
Insula GI × Entorhinal Thickness	0.53	0.28	2.6	0.09	79.2	<0.001	0.26
Insula GI × Hippocampal Volume	0.54	0.29	2.6	0.01	10.2	0.001	0.07

Linear stepwise regressions predicting memory consolidation (CVLT LD FR) with different morphometric and volumetric measures.

GI: Gyrification index; WMH: white matter hyperintensities volumes; R: multiple correlation coefficient; R²: coefficient of determination; SEE: standard error of the estimate; f²: effect size.

In the biomarker-known set, using the same predictors, the combined effect of age, sex, education, scanner type, and WMH volume accounted for 8% of variance. The results of the linear stepwise regression revealed the interaction of insula gyrification, and hippocampal volume accounted for an additional 12% of the variance in memory consolidation (Table 4) (medium effect size).

Table 4.

Brain areas predicting memory consolidation in the biomarker-known set (n = 223).

Model	R	R2	SEE	Delta R2	Delta F	Significance	f²
WMH, Sex, Scanner, Education, Age	0.29	0.08	2.7	0.08	3.8	0.002
Insula GI × Hippocampal Volume	0.45	0.21	2.5	0.12	33.5	<0.001	0.17

Linear stepwise regressions predicting memory consolidation (CVLT LD FR) with different morphometric and volumetric measures.

GI: Gyrification index; WMH: white matter hyperintensities volumes; R: multiple correlation coefficient; R²: coefficient of determination; SEE: standard error of the estimate; f²: effect size.

In the Aβ+ set, using the same measures and procedures, covariates accounted for 7% of variance in this sample. The significant interaction of hippocampal volume and insula GI accounted for an additional 15% of the variance in memory consolidation (medium effect size) (Table 5).

Table 5.

Brain areas predicting memory consolidation in the amyloid-positive set (n = 162).

Model	R	R2	SEE	Delta R2	Delta F	Significance	f²
WMH, Education, Sex, Scanner, Age	0.27	0.07	2.7	0.07	2.4	0.043
Insula GI × Hippocampal volume	0.47	0.22	2.5	0.15	28.8	<0.001	0.16

Linear stepwise regressions predicting memory consolidation (CVLT LD FR) with different morphometric and volumetric measures.

GI: Gyrification index; WMH: white matter hyperintensities volumes; R: multiple correlation coefficient; R²: coefficient of determination; SEE: standard error of the estimate; f²: effect size.

Finally, in the Aβ− set, the linear stepwise regression failed to produce any significant results (Table 6) beyond the 24% of the variance accounted for by the covariates, and showing a very small effect size 0.07.

Table 6.

Brain areas predicting memory consolidation in the amyloid-negative set (n = 61).

Model	R	R2	SEE	Delta R2	Delta F	Significance	f²
WMH, Sex, Education, Scanner, Age	0.49	0.24	2.31	0.24	3.4	0.010
Insula GI × Hippocampus Volume	0.54	0.29	2.3	0.05	4.1	0.05	0.07

Linear stepwise regressions predicting memory consolidation (CVLT LD FR) with different morphometric and volumetric measures.

GI: Gyrification index; WMH: white matter hyperintensities volumes; R: multiple correlation coefficient; R²: coefficient of determination; SEE: standard error of the estimate; f²: effect size.

Group comparisons of brain measures (hippocampal volume, entorhinal thickness, and insula GI) within the biomarker-known set revealed no significant differences between Aβ+ and Aβ− patients (p = 0.09, 0.91, 0.11, respectively). Memory performance was significantly better in the Aβ− compared to Aβ+ (p < 0.001).

Discussion

This study aimed to replicate and further investigate the association of the insular cortex in memory performance across the AD continuum. First, we replicated prior findings as insula GI was significantly correlated with delayed memory recall. Next, after controlling for age, sex, scanner type, education, and WMH, we observed the most significant contributor to memory consolidation involved the interaction of insula GI and entorhinal thickness. However, with each successive refining of the sample, the interaction of insula GI and hippocampal volume accounted for the most variance. Thus, insula GI interactions with the well-known brain regions are associated with memory consolidation. Like previous reports,⁵ this data suggests that the brain changes in AD should be studied with various morphometric and volumetric measures (volume, thickness, GI) to obtain a more comprehensive view of the pathologic or adaptive changes.

This interaction of hippocampal volume and insular GI accounted for the most variance in predicting memory consolidation in the amyloid-positive individuals. The moderate size effect (f² = 0.17) improves mechanistic understanding, though its clinical impact is likely modest given the multifactorial nature of memory. This relationship was not present in Aβ− individuals. Taken together, this indicates that insular atrophy, as measured by GI, is related to memory only as a function of the degree of hippocampal atrophy and when amyloid pathology is present. While the hippocampus has been well-established as relating to AD,¹ the insula, particularly its folding pattern, is a more novel association observed in memory dysfunction in AD. Importantly, post-hoc analysis did not reveal significant differences in hippocampal volume, entorhinal thickness, or insula GI between clinically diagnosed AD patients with and without amyloid pathology. However, the Aβ− group exhibited better memory performance, suggesting distinct pathological mechanisms or compensatory processes. This could represent differences in the impact of amyloid deposition, considering the frontal area and thus insula has earlier amyloid deposition.²³

Longitudinal studies comparing AD, MCI, and healthy participants without biomarker data have reported region-specific changes in GI over time. In patients with AD, decreases in GI were observed in the left superior and middle temporal gyri and the right postcentral gyrus, while increases were noted in the left medial occipital and insula regions. Conversely, individuals with MCI showed no significant GI changes compared to healthy controls.²⁴ Cognitive decline in MCI correlated with GI reductions in the insula region. Notably, in MCI, those who progressed to AD dementia within two years exhibited extensive cortical thinning but reduced insula GI compared to non-converters.²⁴ We observed a positive relationship between insula GI and memory consistent with these prior reports of declines in GI being associated with worse memory. Functional connectivity further implicates the insula in memory. Frontal regions, including the insula, are activated during memory encoding tasks and deactivation is associated with successful encoding, suggesting that excessive activity in this region may be detrimental.²⁵ Additionally, there is an interplay between the default mode network and salience network, which includes the anterior insula, with increased connectivity in amyloid-positive individuals with low neocortical tau but declining connectivity as tau pathology progresses.²⁶ Insula connectivity with the limbic system may play a role in our observed interaction between the hippocampus and insula. Insula atrophy may disrupt the salience and default mode networks and impact the extent of memory consolidation problems when the hippocampus has atrophied. Future work linking of salience and default mode networks dysfunction to insular atrophy is needed to determine if this is a viable biomarker in AD.

Limitations of this study include a lack of racial diversity in the sample; however, the sample represents the often-understudied Appalachian population. The cross-sectional design limits causal inference of hippocampal-insula associations over time. Use of different MRI scanners and the CAT12 pipeline may affect generalizability, though image quality and processing were controlled. CAT12 is widely used for GI^4,5; however, the findings should be replicated using different imaging pipelines. Finally, stepwise regression may overfit the data, however, this was mitigated by theory-driven approach with refined diagnostic grouping.

In conclusion, insular atrophy contributes only as a function of hippocampal volume loss in memory consolidation, specific to symptomatic amyloid-positive patients. Further studies should investigate how this relationship may change over time and contribute to cognitive decline.

Footnotes

Acknowledgements

We thank all the Memory Health Clinic patients and their loved ones, along with staff members of the Memory Health Clinic and the Rockefeller Neuroscience Institute.

ORCID iDs

Michelle M Coleman

Rashi I Mehta

Nafiisah Rajabalee

Ethical considerations

This study was reviewed by the West Virginia University Institutional Review Board.

Consent to participate

All participants gave informed consent.

Consent for publication

Not applicable.

Author contribution(s)

Michelle M Coleman: Conceptualization; Data curation; Formal analysis; Investigation; Writing – original draft; Writing – review & editing.

Marc W Haut: Conceptualization; Data curation; Formal analysis; Investigation; Writing – original draft; Writing – review & editing.

Camila Vieira Ligo Teixeira: Conceptualization; Formal analysis; Writing – original draft; Writing – review & editing.

Cierra M Keith: Conceptualization; Investigation; Writing – review & editing.

Rashi I Mehta: Investigation; Writing – review & editing.

Holly E Phelps: Investigation; Writing – review & editing.

Patrick Worhunsky: Formal analysis; Investigation; Writing – review & editing.

Melanie Ward: Investigation; Writing – review & editing.

Joseph Malone: Investigation; Writing – review & editing.

Mark Miller: Investigation; Writing – review & editing.

R Osvaldo Navia: Investigation; Writing – review & editing.

Gary D Marano: Investigation; Writing – review & editing.

Stephanie Pockl: Investigation; Writing – review & editing.

Nafiisah Rajabalee: Investigation; Writing – review & editing.

William T McCuddy: Data curation; Writing – review & editing.

Pierre-Francois D’Haese: Data curation; Software; Writing – review & editing.

Ali Rezai: Formal analysis; Resources; Writing – review & editing.

Kirk Wilhelmsen: Conceptualization; Formal analysis; Investigation; Writing – review & editing.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data presented here will be made available upon reasonable requests to qualified investigators.

References

Peng

Hong

, et al. Gray matter deterioration pattern during Alzheimer’s disease progression: a regions-of-interest based surface morphometry study. Front Aging Neurosci 2021; 13: 593898.

Dickerson

Bakkour

Salat

, et al. The cortical signature of Alzheimer’s disease: regionally specific cortical thinning relates to symptom severity in very mild to mild AD dementia and is detectable in asymptomatic amyloid-positive individuals. Cereb Cortex 2009; 19: 497–510.

Busovaca

Zimmerman

Meier

, et al.

Is the Alzheimer’s disease cortical thickness signature a biological marker for memory?

Brain Imaging Behav 2016; 10: 517–523.

Núñez

Callén

Lombardini

, et al. Different cortical gyrification patterns in Alzheimer's disease and impact on memory performance. Ann Neurol 2020; 88: 67–80.

Coleman

Keith

Wilhelmsen

, et al. Surface-based correlates of cognition along the Alzheimer’s continuum in a memory clinic population. Front Neurol 2023; 14: 1214083.

Lamballais

Vinke

Vernooij

, et al. Cortical gyrification in relation to age and cognition in older adults. Neuroimage 2020; 212: 116637.

Zhao

Shao

. Altered local gyrification index and corresponding functional connectivity in Alzheimer disease. Alzheimers Dement 2021; 17: e052771.

Huang

Lyu

Stieger

, et al. Direct interactions between the human insula and hippocampus during memory encoding. Nat Neurosci 2025; 28: 1763–1771.

Bermudez-Rattoni

Okuda

Roozendaal

, et al. Insular cortex is involved in consolidation of object recognition memory. Learn Mem 2005; 12: 447–449.

10.

Albert

DeKosky

Dickson

, et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 2011; 7: 270–279.

11.

Scarisbrick

Keith

Vieira Ligo Teixeira

, et al. Executive function and cortical thickness in biomarker aMCI. Appl Neuropsychol Adult 2024. Epub ahead of print 14 Aug 2024. DOI: 10.1080/23279095.2024.2389255.

12.

Villeneuve

Rabinovici

Cohn-Sheehy

, et al. Existing Pittsburgh compound-B positron emission tomography thresholds are too high: statistical and pathological evaluation. Brain 2015; 138: 2020–2033.

13.

Touroutoglou

Katsumi

Brickhouse

, et al. The sporadic early-onset Alzheimer’s disease signature of atrophy: preliminary findings from the Longitudinal Early-onset Alzheimer’s Disease Study (LEADS) Cohort. Alzheimers Dement 2023; 19: S74–S88.

14.

Folstein

McHugh

. Mini-mental state. J Psychiatr Res 1975; 12: 189–198.

15.

Sayegh

Arentoft

Thaler

, et al. Quality of education predicts performance on the Wide Range Achievement Test-4th Edition Word Reading Subtest. Arch Clin Neuropsychol 2014; 29: 731–736.

16.

Reitan

Wolfson

. The Halstead–Reitan neuropsychological test battery: theory and clinical interpretation. 2nd ed. South Tucson, AZ: Neuropsychology Press, 1993.

17.

Pfeffer

Kurosaki

Harrah

, et al. Measurement of functional activities in older adults in the community. J Gerontol 1982; 37: 323–329.

18.

Delis

Kramer

Kaplan

, et al. California verbal learning test. 2nd ed. Adult version. Manual. San Antonio, TX: The Psychological Corporation, 2000.

19.

Wright

Woo

Schmitter-Edgecombe

, et al. The Item-Specific Deficit Approach to evaluating verbal memory dysfunction: rationale, psychometrics, and application. J Clin Exp Neuropsychol 2009; 31: 790–802.

20.

Gaser

Dahnke

Thompson

, et al. CAT: a computational anatomy toolbox for the analysis of structural MRI data. Gigascience 2024; 13: giae049.

21.

Huo

Xiong

, et al. 3D Whole brain segmentation using spatially localized atlas network tiles. Neuroimage 2019; 194: 105–119.

22.

Schmidt

Gaser

Arsic

, et al. An automated tool for detection of FLAIR-hyperintense white-matter lesions in multiple sclerosis. Neuroimage 2012; 59: 3774–3783.

23.

Collij

Mastenbroek

Salvadó

, et al. Regional amyloid accumulation predicts memory decline in initially cognitively unimpaired individuals. Alzheimers Dement (Amst) 2021; 13: e12216.

24.

Bachmann

Schroeter

Chen

, et al. Longitudinal changes in surface based brain morphometry measures in amnestic mild cognitive impairment and Alzheimer’s disease. Neuroimage Clin 2023; 38: 103371.

25.

Vockert

Machts

Kleineidam

, et al. Cognitive reserve against Alzheimer’s pathology is linked to brain activity during memory formation. Nat Commun 2024; 15: 9815.

26.

Schultz

Chhatwal

Hedden

, et al. Phases of hyperconnectivity and hypoconnectivity in the default mode and salience networks track with amyloid and tau in clinically normal individuals. J Neurosci 2017; 37: 4323–4331.