Transcriptomic, cellular, and functional signatures of white matter damage in Alzheimer’s disease

doi:10.21203/rs.3.rs-4668130/v1

Download PDF

Article

Transcriptomic, cellular, and functional signatures of white matter damage in Alzheimer’s disease

https://doi.org/10.21203/rs.3.rs-4668130/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Anatomical white matter (WM) alterations in Alzheimer’s disease (AD) have been widely reported, but functional WM dysregulation in AD has been rarely investigated. The current study focuses on characterizing WM functional and network properties alterations in participants with AD and mild cognitive impairment (MCI), and on further describing their spatially overlapping gene expression patterns. Both AD and MCI shared decreased functional connectivity, clustering coefficient and local efficiency within WM regions involved in impaired sensory-motor, visual-spatial, language or memory function. Notably, observed AD-specific dysfunction (i.e., AD vs. MCI and cognitively unimpaired participants) was predominantly located in WM, including anterior and posterior limb of internal capsule, corona radiata and left tapetum. This WM dysfunction spatially correlates with gene expression of BCHE and SLC24A4, enriched in multiple biological processes such as brain development and behavior, and mostly active in endothelial cells. These findings may represent a substantial contribution to the understanding of molecular, cellular, and functional signatures associated with WM damage in AD.

Biological sciences/Neuroscience/Diseases of the nervous system/Alzheimer's disease

Health sciences/Biomarkers/Diagnostic markers

Alzheimer’s disease (AD) is the most common cause of dementia, and is confirmed by detecting β-amyloid and tau deposits accumulation^1,2. As the spatial patterns of β-amyloid/tau overlap with anatomical³ and functional networks⁴, existing models of AD neurobiology have hypothesized the active role of brain connectomes in the spread of pathology^5,6. Previously, numerous neuroimaging studies explored the functional connectomics in AD within gray matter (GM)^5,7. Although GM regions directly connected by white matter (WM) fiber bundles always show high functional connectivity (FC), the dysfunction within WM has often been ignored.

Recently, several studies have clarified that the blood-oxygen-level-dependent (BOLD) signals in WM can reflect the intrinsic neural activities in the human brain^8–10. Specifically, the power spectra distributions of functional signals in WM are repeatable and highly consistent with the anatomical configurations¹¹. Using independent component analysis, symmetrical WM functional organizations have been found to overlap spatially with WM anatomical structures¹². Combing the intracranial stereotactic-electroencephalography and functional magnetic resonance imaging (fMRI), Huang and colleagues showed the electrophysiological basis underlying FC in WM⁹. For the macro-scale organization, WM functional modules similar to those present in GM have been identified^13–15. At the molecular genetic level, the common pattern of WM dysfunction in major psychiatric disorders was correlated with synaptic function and excitatory neurons¹⁰. Although various studies have implicated significance of WM degeneration, demyelination and hyperintensities in AD pathology^16,17, few have explored the potential role of WM dysfunction in AD¹⁸. Detailed characterizations of WM functional features would provide a comprehensive understanding of the pathophysiological mechanisms of AD.

As an effective method to capture the topological features of functional brain networks, graph theoretical analysis can describe global and local topological patterns of activity in the human brain¹⁹. Assortativity, a global topological metric, measures resilience, with a larger value indicating a more robust network. Other nodal metrics, such as clustering coefficient and local efficiency, have proven to be effective to identify AD with electroencephalograph functional networks²⁰ and have described key features of WM functional network disorganization in neuropsychiatric disorders such as major depression²¹ and schizophrenia²².

The disrupted brain functions in AD are not only reflected in the macroscopic network configuration patterns, but are also inextricably linked to the disease’s molecular genetic etiology^23,24. Genetic etiology affects the structural and functional attributes in the human brain^25,26, and is associated with the risk of developing AD²⁴. For example, APOEε4 carrier with amnestic mild cognitive impairment (amnestic MCI) showed decreased FC in WM regions²⁷. The abnormal structural connectome changes of AD were linked with synaptic related biological process and neuronal cells²⁸. Moreover, neurotransmitters and gene expression profiles relating to WM dysfunction in schizophrenia, major depressive disorder, bipolar disorder and obsessive-compulsive disorder have been reported¹⁰. Therefore, identification of the genetic influences underlying the variation in WM functional connectome can be crucial to further clarify the multiscale pathological processes in AD.

Here, we hypothesize that the macroscopic WM functional changes in AD align with distinctive microstructural gene transcriptional signatures in the neurotypical adult brain. With this aim, we first characterized the diagnosis-specific patterns of WM functional changes by measuring the FC, topological properties, and regional functional connectivity strength (RFCS) among clinical AD, MCI and cognitively unimpaired (CU) participants. Next, we characterized the spatial relationship between gene expression and regional WM dysfunction using the Allen Human Brain Atlas (AHBA)²⁹. The model was submitted to the enrichment analysis to provide preliminary transcriptional correlates of RFCS. Finally, we identified cell types with spatial distributions overlapping WM functional disconnection.

Overall approach

Current study employed 53 individuals with AD, 90 individuals with MCI, and 100 matched CU individuals shared from the Open Access Series of Imaging Studies (OASIS-3) (https://www.oasis-brains.org/). No significant (P > 0.05) group differences were found in age, sex, or education (Table 1). Age, sex and education were all regressed as covariates in statistical analyses.

Table 1

Demographics and clinical characteristics of participants.
Characteristic	AD	MCI	CU	P value
Number	53	90	100	-
Age	75.89 ± 8.555	74.76 ± 7.668	74.41 ± 8.227	0.4940^a
Sex(M/F)	33/20	48/42	50/50	0.3471^b
Education	14.75 ± 3.204	14.59 ± 3.016	14.32 ± 1.757	0.8438^a
Handedness(L/R)	0/53	0/90	0/100	-
MMSE	22.08 ± 4.057	25.99 ± 2.882	28.86 ± 1.318	< 0.0001^a
Note. F, female; L, left; M, male; MMSE, Mini-Mental State Examination; R, right.
^a Kruskal Wallis Test
^b Chi-Square Test

The data analysis procedure consisted of five main parts (Fig. 1). (a) Preprocessing on 243 participants fMRI data from OASIS-3. (b) Characterizing the diagnosis-specific FC and (c) topological on GM-, WM- and whole-brain- functional networks separately. (d) Calculating the group differences in RFCS. (e) Utilizing partial least squares (PLS) regression to model the relationship between the brain changes of AD and transcriptional profiles, and assigning the first component of PLS (PLS1) gene lists into different cell types.

Altered Functional Connectivity in the whole-brain

We first obtained 45 WM regions of interest (ROIs) and 96 GM ROIs by point-multiplying the JHU ICBM-DTI-81 WM atlas³⁰ and the group-level WM mask, the Harvard-Oxford Atlas³¹ and group-level GM mask, respectively (see Methods, subsection Obtaining the group-level WM and GM masks). Next, we construct the FC matrices in the size of 141 × 141 (i.e., 45 + 96) for the whole-brain by performing Pearson’s correlation between times series of ROI-pairs for each individual. Details on all ROIs see Supplementary Table S1 and S2.

Compared to CU, both AD and MCI showed decreased FC within WM between the left anterior (ALIC) and the right posterior limb of internal capsule (PLIC) (all t < -2.525, P_FDR < 0.020), and two increased FC between GM and WM (i.e., between the right angular gyrus and right posterior thalamic radiation, all t > 3.104, P_FDR = 0.004; between the left angular gyrus and left tapetum, all t > 3.497, P_FDR < 0.002). Moreover, we also observed decreased FC within GM involving superior parietal lobule, postcentral gyrus, fusiform cortex, and occipital cortex (all t < -2.505, P_FDR < 0.020) (Fig. 2a, 2b, Table S3).

Compared with CU and MCI groups, AD-specific FC decreased within WM regions involving the ALIC, PLIC, the anterior/posterior corona radiata, the splenium of corpus callosum, the posterior thalamic radiation, and left tapetum (all t < -2.763, P_FDR < 0.011), but enhanced between WM and GM (i.e., between right pre/post- central gyrus and left ALIC, between right inferior frontal gyrus and left superior fronto-occipital fasciculus; all t > 2.382, P_FDR < 0.028). However, no significant MCI-specific FC was found (i.e., in any ROI-pair, for comparison of FC between MCI vs. CU or MCI vs.AD, at least one P_FDR > 0.050) (Fig. 2a, 2b, Table S3).

As shown in Fig. 2b, we can observe three patterns: first, the regions showing FC differences between MCI and CU were all detected in the comparison between AD and CU; second, the extra changes in AD versus MCI were mainly related to WM; and third, the patients groups (i.e., MCI and AD) showed weakened intra- WM or GM FC but enhanced inter-WM-GM FC. Moreover, we performed a robust modified Spearman correlation between the FC and clinical score of Mini-Mental State Examination (MMSE). We observed that both the FC between the left PLIC and the splenium of corpus callosum (Spearman’s r = 0.32, uncorrected P = 0.043), and between the right PLIC and left posterior corona radiata (Spearman’s r = 0.33, uncorrected P = 0.048) showed moderate positive correlations with MMSE of AD group (Fig. 2c).

Global topological properties

For WM network, the global clustering coefficient in AD was significantly lower than CU (Dunn’s test, P = 0.0302) (Fig. 3a, 3c). Similarly, the whole-brain network also showed a significantly smaller clustering coefficient in AD than that in CU (Dunn’s test, P = 0.0402) (Fig. 3c). Moreover, in the GM network, the global clustering coefficient (H = 7.706, P = 0.0212) and assortativity (H = 6.116, P = 0.0470) were significantly different among AD, MCI and CU. Specifically, clustering coefficient in AD (Dunn’s test, P = 0.0521) and MCI (Dunn’s test, P = 0.0693) were both marginally significantly smaller compared to CU, and assortativity in AD (Dunn’s test, P = 0.0447) was also lower than CU (Fig. 3b, 3c).

Nodal topological organization

Analysis on graph metrics of WM ROIs revealed that both nodal clustering coefficient within 11 ROIs (all H > 8.842, P_FDR < 0.0492) and nodal local efficiency within 9 ROIs (all H > 9.702, P_FDR < 0.0391) exhibited significant group effects (Fig. 3d). Post hoc analysis on nodal clustering coefficient (Fig. 3d, the top two rows) revealed significantly lower values in both AD and MCI groups than CU group within the left corticospinal tract, the bilateral sagittal stratum, the left superior fronto-occipital fasciculus, and the bilateral uncinate fasciculus (all Z < -2.369, P_FDR < 0.031). Compared with CU and MCI groups, the AD-specific clustering coefficient was reduced in the bilateral superior corona radiata and the posterior thalamic radiation (all Z < -2.263, P_FDR < 0.0390). Moreover, in the right ALIC and left tapetum, only the clustering coefficient in AD was significantly lower than that in CU (all Z < -3.215, P_FDR < 0.008). No significant MCI-specific clustering coefficient was found (i.e., in any ROI, for comparison of clustering coefficient between MCI vs. CU or MCI vs.AD, at least one P_FDR > 0.050). More details are shown in Supplementary Table S4.

Post hoc analysis on local efficiency of WM ROIs (Fig. 3d, the bottom two rows) indicated that the values in both AD and MCI groups were significantly decreased than CU group within the left corticospinal tract, the bilateral sagittal stratum, and the bilateral uncinate fasciculus (all Z < -2.136, P_FDR < 0.050). AD-specific local efficiency was smaller within the bilateral superior corona radiata, the posterior thalamic radiation and the left tapetum (all Z < -2.130, P_FDR < 0.050). No significant MCI-specific local efficiency was found (i.e., in any ROI, for comparison of local efficiency between MCI vs. CU or MCI vs.AD, at least one P_FDR > 0.050). More details are shown in Supplementary Table S5.

For GM, results of nodal clustering coefficient showed significant differences among AD, MCI and CU (all H > 10.73, P_FDR < 0.050; Fig. 3e). Compared to CU, both AD and MCI groups had reduced clustering coefficient in the right parahippocampal gyrus, the right temporal fusiform cortex, the right planum polare, the right Heschls gyrus, the left planum temporale, the left supracalcarine cortex and the left occipital pole (all Z < -2.239, P_FDR < 0.044). The AD group showed additional reduction in clustering coefficient in the left postcentral gyrus and left temporal occipital fusiform cortex than CU (all Z < -3.345, P_FDR < 0.005). Nevertheless, the differences between the nodal clustering coefficient of MCI and AD group (all Z > -1.703, P_FDR > 0.126) were not significant after multiple comparison correction, nor were the difference among the nodal local efficiency of AD, MCI and CU (all H < 12.922, P_FDR > 0.1500). More details are shown in Supplementary Table S6.

Differences in regional functional connectivity strength

Results from WM ROIs (Fig. 4a, Table S7) indicated that AD and MCI exhibited significantly lower RFCS than CU group in the right superior corona radiata (all Z < -2.350, P_FDR < 0.033). AD participants showed additional reduction in RFCS than CU in the left ALIC and right posterior corona radiata (all Z < -3.096, P_FDR < 0.007). MCI group showed additional reduction in RFCS than CU in the right superior corona radiata (all Z < -2.350, P_FDR < 0.033). Compared to MCI and CU, AD-specific RFCS was significantly reduced in wide regions including the splenium of corpus callosum, the bilateral PLIC, the anterior and left posterior corona radiata, the right external capsule and left tapetum (all Z < -2.268, P_FDR < 0.039). But no significant MCI-specific WM RFCS was found (i.e., in any ROI, for comparison of RFCS between MCI vs. CU or MCI vs.AD, at least one P_FDR > 0.050). The Spearman’s rank correlation showed moderate correlation between the WM RFCS and MMSE in the splenium of corpus callosum (Spearman’s r = 0.33, uncorrected P = 0.011; Fig. S1) in AD group.

Findings from GM ROIs (Fig. 4b, Table S8) revealed that both AD and MCI showed decreased RFCS than CU within the right precentral gyrus, the right anterior division of superior temporal gyrus, the right superior parietal lobule, and bilateral inferior division of lateral occipital cortex (all Z < -2.490, P_FDR < 0.026). Furthermore, AD showed additional reduction in RFCS than CU within the right temporal pole, the right posterior division of superior temporal gyrus, the right middle temporal gyrus and the right central opercular cortex (all Z < -3.253, P_FDR < 0.005). MCI participants showed additional reduction in RFCS than CU in the right cingulate gyrus (Z = -3270, P_FDR = 0.005). Compared to CU and MCI groups, the AD-specific RFCS was found to be decreased in the right parahippocampal gyrus and the bilateral temporal occipital fusiform cortex (all Z < -2.255, P_FDR < 0.045). No significant MCI-specific GM RFCS was found (i.e., in any ROI, for comparison of RFCS between MCI vs. CU or MCI vs.AD, at least one P_FDR > 0.050).

Functional disconnection correlates with gene expression level

We performed an imaging transcriptomics analysis by using the abagen toolbox³² on the AHBA dataset (http://human.brain-map.org). Since the right hemispheres in the AHBA dataset were incomplete, only 48 GM ROIs and 22 WM ROIs in the left hemisphere were submitted to the analysis of estimating gene expressions matrix. Thus, a 48×15633 transcriptional level matrix of GM ROIs and a 22×15632 transcriptional level matrix of WM ROIs were obtained. Subsequently, we conducted a partial least squares (PLS) regression³³ to characterize for the relationship between the RFCS and the gene transcriptional profiles. The first component (PLS1) for predicting WM RFCS explained 65.72% of the variance (permutation test, 10,000 times, P = 0.0023) while for predicting GM RFCS only explained 30.34% of its variance (permutation test, 10,000 times, P = 0.1815). Accordingly, because the percentage of variance explained for GM RFCS did not reach the significant level, only the PLS regression results relating with the WM RFCS were considered to the further analysis.

We found a strong spatial correlation between the group difference in WM RFCS and PLS1 score (Spearman’s r = 0.61, uncorrected P = 0.011; Fig. 4c, 4d, 4e). The PLS1 score exhibited a regional gradient distribution that decreased from anterior to posterior and from lateral to medial areas (Fig. 4d). In particular, in the anterior corona radiata, the tapetum and the ALIC, both the PLS1 score and the group difference in WM RFCS were high, but in regions including the inferior cerebellar peduncle, the cingulum in the cingulate cortex and medial lemniscus, both the PLS1 score and the group difference in WM RFCS showed lower values.

Biological enrichment pathways of functional disconnection

We identified 1265 genes in the PLS1+ (Z > 3) gene set and 1673 genes in the PLS1- (Z < -3) gene set. By comparing the PLS1 gene lists and the reported AD-related gene lists, we found 8 overlapping genes (i.e., ACHE, BCHE, BLMH, DBN1, KCNIP3, LRP1, PLAU and PSEN1) in comparison to AHBA (https://help.brain-map.org/display/humanbrain/Documentation), and 15 overlapping genes (i.e., BIN1, HAVCR2, TMEM106B, APH1B, CD33, SPDYE3, EPHA1, EED, FERMT2, SLC24A4, BCKDK, WDR12, PLEKHA1, FOXF1, and RBCK1) in comparison to genome-wide meta-analysis studies (GWAS)^24,34. In addition, we observed that BCHE and SLC24A4 gene expression levels significantly correlated with the WM RFCS changes among three groups (BCHE: Spearman’s r = 0.64, uncorrected P = 0.0045, P_FDR = 0.0516; SLC24A4: Spearman’s r = -0.795, P_FDR < 0.0001; Fig. 4f).

We selected gene ontology (GO) biological processes and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways to conduct the enrichment analysis. After correction for multiple comparison and removing the discrete clusters, twelve significant GO biological processes associated with the PLS1- gene list were identified, such as “brain development”, “ribonucleoprotein complex biogenesis”, “tissue morphogenesis”, “gland development”, “heart development”, “behavior” and “synaptic signaling”. No KEGG pathway reached the significant level (Fig. 5a,5b). The multi-gene-list meta-analysis revealed nine enrichment pathways overlapping between PLS1- gene list and GWAS. These pathways included “membrane organization”, “protein catabolic process”, “behavior”, “cell morphogenesis” and “regulation of synapse structure or activity” and so on (Fig. 5c). Moreover, we observed thirteen significant GO biological processes linked with PLS1 + gene list, including “regulation of cellular response to stress”, “cell activation”, “DNA damage response”, “regulation of apoptotic signaling pathway”, “metal ion transport” and so on (Fig. 6a, 6b). No KEGG pathway was significant. Compared with GWAS, several ontology terms were identified, such as the “cell activation”, “regulation of cellular response to stress”, “regulation of apoptotic signaling pathway”, “regulation of membrane potential” and so on (Fig. 6c).

Cellular signatures

For genes in the significant PLS1- list, we observed 68 genes that are genetic markers of microglia, 103 of excitatory neurons, 85 of inhibitory neurons, 109 of endothelial cells, 87 of astrocytes, 18 of oligodendrocyte precursors and 90 genes of oligodendrocytes. Only the number of genes in the endothelial cells (number = 109) reached a significant level (permutation test, 10,000 times, uncorrected P = 0.0169). Functional enrichment results demonstrated five significant pathways, including the “negative regulation of cell population proliferation”, “tissue morphogenesis”, “adrenomedullin receptor signaling pathway”, “mesenchyme development” and “response to wounding” (Fig. 5d). However, as for PLS1 + gene list, none of the number of genes assigned to the seven cell types were significant (permutation test, 10,000 times, uncorrected P > 0.0773).

Combining whole-brain resting-state FC and gene transcriptional profiles, we characterized AD’s dysfunctional connectivity, and diagnosis-related transcriptional and cellular signatures in WM areas. Compared to CU individuals, both AD and MCI groups exhibited decreased intra-WM/-GM FC, increased inter-WM-GM FC, and reduced network properies (i.e., assortativity, global/local clustering coefficient, and nodal local efficiency) in WM (and in GM) functional networks. On the other hand, we observed that the AD-specific dysfunctional connections were mainly distributed in WM regions including the ALIC, the PLIC, the corona radiata and the left tapetum. Moreover, based on the WM RFCS, we constructed PLS regression models and identified genes/cells associated with AD WM damage in a data-drive perspective. The gene expression level of BCHE (or SLC24A4) positively (or negatively) correlated with the RFCS difference among clinical groups. Ontology terms enriched from the PLS1- gene list widely overlapped with terms obtained from GWAS. Finally, we classified PLS1- genes into 7 cell types, identifying endothelial cells distributions as strong predictors of functional WM damage in AD.

Changed functional connectivity in the whole-brain

Decreased FC within GM regions has been widely reported to be correlated with deteriorated cognitive performance of AD patients and detected in regions exhibiting abnormal amyloid deposition^7,35. Results in this study consistently revealed FC was not only weakened within GM, but also within WM. The positive correlation between the intra-WM FC and MMSE suggests that the reduced FC within WM may be partly responsible for cognitive impairment, similarly to the observed GM effect. In contrast, FC between the WM and GM was observed to be enhanced, which suggests an excessive information exchange between these brain structures, which may be due to compensatory mechanisms³⁶ and/or typically observed hyperactivity with AD progression³⁷.

Furthermore, we noticed that ALIC, PLIC, the corona radiata and the left tapetum were involved in most significant group differences. The ALIC has been proposed to be a target for deep brain stimulation in AD since its full connection with all neuromodulation targets³⁸. The PLIC were previously observed to be involved in the FC aggravated in amnestic APOEε4 carriers, the major genetic risk for developing AD²⁷. The corona radiata presents decreased fractional anisotropy accompanied with lower sphingomyelin, which could facilitate in vivo AD prediction³⁹. The tapetum presents lower fractional anisotropy in individuals with family history of AD⁴⁰. Taken together, these results are consistent with previous findings and provide new evidence for the significant role that WM FC plays in AD development.

Abnormalities in gray matter and white matter functional network topology

For topological metrics, assortativity characterizes the resilience of a network. A network with high assortativity contains multiple nodes with similar degrees, thereby performed better in coping with nodal dysfunction. Global clustering coefficient is an index of functional segregation, and denotes the prevalence of connection clusters¹⁹. In the present study, we observed reduced assortativity coefficient and global clustering coefficient within GM functional network in AD patients, which is highly consistent with previous studies^7,41. Moreover, the lower global clustering coefficient in AD and MCI patients was observed within WM functional networks and within the whole brain network (i.e., including GM and WM) for the first time. These results indicate the vulnerability and disrupted specialized functioning of patients’ global-wise brain network, both in the entire GM and WM.

Nodal clustering coefficient and local efficiency are the two main indices of regional functional specificity. Local clustering coefficient reflects the prevalence of clustered connectivity around each node, while the local efficiency measures how efficiently a network exchanges information¹⁹. In the current study, nodal clustering coefficient and local efficiency were observed to be significantly decreased for AD or MCI group within GM or WM regions. Specifically, these regions are primarily engaged in four functions: sensory-motor (i.e., postcentral gyrus⁴² and corticospinal tract⁴³), visual-spatial (i.e., parahippocampal gyrus⁴⁴, bilateral sagittal stratum⁴⁵, temporal fusiform cortex⁴⁶, temporal occipital fusiform cortex, supracalcarine cortex⁴⁷, occipital pole, posterior thalamic radiation⁴⁸, and left tapetum⁴⁹), language (i.e., planum polare, heschls gyrus, planum temporale, and bilateral uncinate fasciculus⁵⁰), and memory-executive function (i.e., parahippocampal gyrus⁴⁴, anterior limb of internal capsule⁵¹). Noticeably, the deficits in these functions are the clinical phenotypes of Alzheimer’s disease⁵². These findings provide new evidence that the two nodal metrics of WM are effective in capturing the pathological features typically observed in AD. Furthermore, it had been reported that reduced clustering coefficient and local efficiency in GM were accompanied with more tau burden in AD⁵³. GM functional network topology can capture the tendency of abnormal deposition of pathological processes in a nonlinear way⁵. The relationship between WM functional network reconfiguration and amyloid and tau burden remains to be explored.

AD-risk genes related to the functional connectivity changes

Using the PLS regression model, we identified the combination of genes expression weights in which the mapping profile of gene expression level on the WM was most similar to the distribution of changes in WM RFCS. In the PLS1 gene list, we noted a strong-positive (or negative) correlation between the BCHE (or SLC24A4) gene expression level and the changes of WM RFCS. The BCHE gene was over-expressed in regions showing decreased RFCS in patients. BCHE encodes butyrylcholinesterase (BuChE), which is highly expressed in WM and has been associated with amyloid protein plaques accumulation and AD progression⁵⁴. Previous inhibition of BuChE has brought potential benefits for patients with AD⁵⁵. In addition, the SLC24A4 gene has been considered as a promising key biomarkers for AD diagnosis due to its involvement in brain amyloidosis and tauopathy⁵⁶. Our novel findings support the strong relationship between BCHE (or SLC24A4) and WM RFCS. Additional direct and causal analyses are needed to further clarify the underlying biological mechanisms.

Biological processes enriched with PLS1-/PLS1 + gene list

We identified ontology terms enriched on the PLS1- gene list, which contained 9 of 16 pathways enriched from genes in GWAS of AD^24,34. These pathways cover a wide range of biological functions including the behavior, protein, synapse, and cellular processes. The “membrane organization” can affect the response to ani-Alzheimer treatments⁵⁷. Disorders of “protein catabolic process” can cause the accumulation of unwanted proteins and contribute to AD⁵⁸. The process of “cell morphogenesis” can be influenced by tau and is a widely reported enrichment pathway related to AD⁵⁹. Moreover, the synapse-related pathways (i.e., “synaptic signaling” and “regulation of synapse structure or activity”) were detected. Synaptic loss and damage underlies the symptoms of neurodegenerative disease⁶⁰. Trillions of synapses in the healthy brain formed the path of information flow. Whereas in AD patients, the accumulation of β-amyloid may induce the decrease in synaptic signaling and damage the communication at synapses¹.

For biological processes enriched on the PLS1 + gene list, the pathway “regulation of cellular response to stress” was identified. Cellular stress response was another pathological feature of AD, and can be triggered by the aggregation of misfolded proteins⁶¹. Modifying stress response could be a promising therapy for AD⁶¹. The “cell activation” may be related with the microglial activation, which was capital for the AD-initiating⁶². The “regulation of apoptotic signaling pathway” can also contribute to AD. Apoptosis can be triggered by nonvascular β-amyloid and phosphorylated tau deposition, and affects the cognition-disruption related regions in AD⁶³. These replicated pathways validate the effectiveness of the gene lists obtained based on the WM RFCS.

We also revealed additional AD-related pathways that could not be obtained from genes in GWAS. The abnormal “ribonucleoprotein complex biogenesis” was accompanied by the improved protein level of amyloid precursor protein⁶⁴. The “heart development” pathway may indicate the link between the heart failure and AD. Myocardial diastolic dysfunction was found in AD patients⁶⁵. Not only heart failure shared similar genetic profiles with AD, but also the failing heart may contribute to the aggregate pathology in AD individuals⁶⁵. The “gland development” process could involve the thyroid gland and the pineal gland⁶⁶. The thyroid gland regulates the development of central nervous system, while the pineal gland secretes melatonin which inhibits the formation of β-amyloid protein⁶⁷.

Cellular features

Finally, we observed that the highest portion of the genes in the PLS1- list were related to endothelial cells. Endothelia cells are the site of the blood-brain barrier, which is essential for maintaining the microenvironment of the central nervous system⁶⁸. Endothelia cells also constitute a major component of the neurovascular system. In AD patients, neurovascular system impairment were detected, which affected the cerebral hypoperfusion and blood-brain barrier^65,69. The dysregulated angiogenesis in endothelia cells may contribute to AD pathogenesis and thus the endothelial cells could be a potential cellular target for therapy⁶⁹. In addition, endothelial dysfunction could be one of the key cellular mechanisms to hypertension, a disease which may induce WM hyperintensities in AD patients⁷⁰. The observed links between endothelial cells and WM functional dysregulation may well represent an important topic of future AD-pathophysiological research.

Several limitations of the current study need to be noted. First, these results are based on a cross-sectional design, which is unable to adequately delineate the relationship between the WM functional changes and the disease course. Second, the information on illness duration or antipsychotics medications was unavailable, being subsequently difficult to remove potential medication impacts. Third, no behavioral cognitive task was performed. Future research on the combination of WM functional signal and behavior will contribute to the understanding of pathologies of AD and facilitate clinical applications. Fourth, in order to minimize the nuisance effects on WM from nearby GM, we adopted a practical anatomical WM atlas to obtain WM regions. However, the inconsistent region sizes in the WM atlas inevitably have an impact on the mean functional signal. How to better segment WM tract for functional signal extraction could be an important topic for future work. Last but not least, the gene data were derived from the neurotypical templates, which were based on postmortem tissue from cognitively normal individuals. Although the identified transcriptional patterns could be not representative enough for the whole-brain molecular signatures of patients, it remains relevant in terms of capturing vulnerability to disease factors as proved by previous studies¹⁰.

In summary, the current study characterized functional dysregulation in WM (and in GM) of the AD/MCI brain, and its spatial gene expression and cellular correlates for the first time. AD and MCI individuals showed common decreased FC and clustering coefficient in both GM and WM regions involved in disturbed cognitive functions. Using the PLS model and pathway enrichment analyses, we investigated the gene transcriptional signature associated with WM functional changes, and obtained the related biological processes and cell type in AD. The identified ontology terms shared considerable overlap with terms obtained from GWAS and provided new evidence for the significance of WM functional changes in AD. These findings provide insight into the association between the transcriptional data and WM dysfunction, and further advance our understanding of the pathophysiological mechanisms of AD.

Participants

Current study employed the dataset shared from the Open Access Series of Imaging Studies (OASIS-3) (https://www.oasis-brains.org/). Evaluated by the clinical dementia rating (CDR) score, all participants were categorized as 53 AD (CDR = 1), 90 MCI (CDR = 0.5), and 100 age-, sex-, and education-matched CU (CDR = 0). All participants provided written informed consent to the research procedures. The MMSE were used to evaluate clinical symptoms.

MRI acquisition

High-resolution T1-weighted images, resting-state fMRI were acquired in a 16-channel head coil of Siemens TIM Trio 3T scanner. Since the detailed scanning parameters were described elsewhere⁷¹, we have only briefly described it here. The main scanning parameters of resting-state fMRI were as follows: TR = 2200 ms, TE = 27 ms, flip angle = 90°, number of slices = 33, slice thickness = 4 mm, voxel size = 4 × 4 × 4 mm³.The scanning parameters of T1-weighted anatomical images were as follows: repetition time (TR) = 2400 ms, echo time (TE) = 3.16 ms, flip angle = 8°, voxel size = 1 × 1 × 1 mm³, and slice thickness = 1 mm.

Pre-processing steps of functional images

The pre-processing of neuroimaging data was performed using the Statistical Parametric Mapping (SPM12, https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and Data Processing Assistant for Resting-State fMRI (DPARSF, V4.4_180801, http://rfmri.org/dparsf). The first 10 volumes of functional image were discarded to prevent magnetization disequilibrium for each participant. We performed the realignment to the mean functional images, and obtained the 6-parameter rigid body linear transformation of head motion. All participants have maximum motion < 3 mm and 3°. The mean functional image was co-registered with the T1 image, and then T1 structural images were segmented into GM, WM and cerebrospinal fluid (CSF). Nuisance signals (i.e., the 24-parameters and the mean signal of CSF) were regressed. Temporal scrubbing regressor was also utilized to reduce the effect from head motion. The linear trend was removed for the signal drift correction. The band-pass filtering (0.01 ~ 0.1 Hz) was used to reduce the effect of low-frequency drift and high-frequency physiological noise. Spatial smoothing was performed separately within GM and WM template with a 4 mm FWHM. Finally, the smoothed functional images were normalized into the Montreal Neurological Institute (MNI) space and resampled to 3 × 3 × 3 mm³.

Obtaining the group-level white matter and gray matter masks

To avoid the confusion between GM and WM signals, we first created the group-level GM and WM mask in line with previous study^13,14. In specific, by determining the maximum probability of whether each voxel belongs to the GM, WM or CSF, we can get binarized GM and WM masks for each participant. Then, the individual masks were compared across all participants. Voxels that account for more than 60% of the total number of individual WM templates were classified as WM. Noticeably, a lenient threshold with greater than 20% across participants were adopted for GM, and any voxels identified in the WM mask have been removed. In addition, the resulting GM and WM masks were compared to the functional images. Those voxels that were identified as GM or WM, but accounted for less than 80% of functional data across all participants were also excluded. Finally, the Harvard-Oxford Atlas³¹ were used to mark and exclude the deep regions from group-level WM mask, including the thalamus, caudate nucleus, putamen, globus pallidus, and nucleus accumbens.

Constructing functional connectivity matrices in white matter and gray matter

Before performing the network properties analysis, we conducted the point multiplication on the group-level WM mask and the JHU ICBM-DTI-81 WM atlas³⁰ to obtain a new WM mask with 48 regions of interest (ROIs). We excluded 3 ROIs with no voxels or the number of voxels was one. Eventually, 45 ROIs were retained as the final WM mask. Additionally, the group-level GM mask was point multiplied with the Harvard-Oxford Atlas³¹ to obtain the final GM mask with 96 ROIs. Details on ROIs see Supplementary Table S1 and S2.

The averaged time series of each WM/GM ROI was separately extracted from the functional images for each individual. By conducting the Pearson’s correlation analysis between each pair of WM masks, we obtain a 45 × 45 WM correlation matrix for each participant, then transform the correlation coefficient into Fisher’s Z score. Similarly, we obtained a 96 × 96 GM connection matrix and a 141 ×141 whole-brain connectivity matrix for each individual.

Network topological property analysis

Graph theoretical analysis was carried out using the Graph Theoretical Network Analysis (GRETNA) Toolbox (http://www.nitrc.org/projects/gretna/). The sparsity threshold values were identified in following steps by referring to previous study²². The maximal threshold was defined as the correlation coefficient value which ensured all connections were significant (P < 0.05). The minimum threshold was defined as follows: (1) the mean degree over all nodes of each network was > 2 × log(N) (N expresses the number of nodes, N = 45 in WM network, N = 96 in GM network, and N = 141 in the whole brain network); (2) the number of nodes in the largest component was > 70% × N. Accordingly, the range of sparsity threshold was 0.23 ~ 0.46 for WM functional network, 0.11 ~ 0.50 for GM functional network, and 0.09 ~ 0.47 for the whole-brain network with a common step of 0.01.

To characterize the topological properties of brain networks, we calculated 2 global network metrics (i.e., assortativity and global clustering coefficient) and 2 nodal parameters (i.e., nodal clustering coefficient and nodal local efficiency) on the unweighted, undirected graphs.

Assortativity quantified the resilience of a network. The formula is as follows:

$$r= \frac{{l}^{-1}\sum _{(i,j)\in L}{k}_{i}{k}_{j}-{\left[{l}^{-1}\sum _{\left(i,j\right)\in L}\frac{1}{2}\left({k}_{i}+{k}_{j}\right)\right]}^{2}}{{l}^{-1}\sum _{(i,j)\in L}\frac{1}{2}\left({{k}_{i}}^{2}+{{k}_{j}}^{2}\right)-{\left[{l}^{-1}\sum _{\left(i,j\right)\in L}\frac{1}{2}\left({k}_{i}+{k}_{j}\right)\right]}^{2}}$$

${k}_{i}$ , ${k}_{j}$ are the degrees of the nodes at the ends of the edge $i$, with $i=1\dots l$.

Clustering coefficient of a node $i$ characterizes prevalence of clustered connectivity around individual nodes, and is defined as follows:

$$C\left(i\right)=\frac{2{t}_{i}}{{k}_{i}({k}_{i}-1)}$$

$$C=\frac{1}{n}\sum _{i\in N}C\left(i\right)$$

$C\left(i\right)$ is the clustering coefficient of a node $i$, ${t}_{i}$ is the number of edges in the sub-graph ${N}_{i}$, and ${k}_{i}$ represents the number of nodes connected to node $i$.

Local efficiency measures how efficient the communication is between the first neighbors of a node $i$ when $i$ is removed. The formula is as follows:

$${E}_{loc}\left(i\right)=\frac{1}{{k}_{i}({k}_{i}-1)}\sum _{j,h\in N,j\ne i}\frac{{a}_{ij}{a}_{ih}}{{d}_{jh}\left({N}_{i}\right)}$$

${k}_{i}$ is the number of nodes in the subgraph ${N}_{i}$ that includes all neighboring nodes of node i, ${d}_{jh}\left({N}_{i}\right)$ is the shortest path lengths between node j and node h in subgraph ${N}_{i}$.

Statistical analyses

To investigate the differences among AD, MCI and CU groups, we conducted statistical analysis on FC, graph theoretical metrics and RFCS separately. We perform non-parametric permutation test on FC (10,000 times) and utilized the Kruskal-Wallis test on RFCS and the area under curve (AUC) of topological metrics. In the post-hoc analysis, Mann-Whitney U test (corrected by FDR) were used for all measures except for topological global metrics, which were examined using the Dunn’s test. Among all the above analyses, the effect of age, sex and education were all regressed as covariates. Finally, a robust Spearman’s rank correlation with a bootstrap of 1,000 times was performed to examine the association between clinical characteristics and FC. Unless otherwise specified, all P-values were adjusted for multiple comparisons using the false discovery rate (FDR) procedure with a significant threshold of 0.05.

Relating gene expression to functional changes in brain region

We used the abagen toolbox³² on the AHBA dataset (http://human.brain-map.org) to make the transcriptional matrix. To relate the FC of each brain region to the transcriptional values, we calculated the RFCS by averaging its FC with all other 44 WM or 95 GM regions respectively. Group differences in RFCS were examined by Kruskal-Wallis test (see Methods, subsection Statistical analysis). Because the right hemispheres in the AHBA dataset were incomplete (i.e., two of the six right hemispheres were collected)⁷², only 48 GM ROIs and 22 WM ROIs in the left hemisphere were used for gene expressions matrix estimation.

Two PLS regression models³³ were built by using the principal components of imaging-transcription matrix as the independent variables to predict the group-differences in GM or WM RFCS respectively. Then the permutation test (10,000 times) was used to evaluate the significance of the variance. After that, we performed bootstrapping to estimate the PLS1 weights for each gene, and normalized the weight scores with a threshold of Z < -3 and Z > 3⁷³.

Enrichment analysis

To figure out the molecular features of the thousands of genes, we submitted the PLS1+ (Z > 3) gene list and the PLS1- (Z < -3) list to the Metascape⁷⁴ (http://metascape.org). Two core default ontologies (i.e., GO biological processes and KEEG pathways) were selected. The significant threshold with 0.05 was set for all pathways and FDR correction was used for multiple comparison. Moreover, to compare the gene list obtained by data-driven analysis versus genes derived through meta-analysis, we conducted the multi-gene-list meta-analysis between the PLS1 + or PLS1- gene list and polygenic risk for AD based on the genome-wide meta-analysis studies (GWAS)^24,34.

Cell types

We assigned the gene lists into 7 types of cells, including the astrocytes, endothelial cells, microglia, excitatory, and inhibitory neurons, oligodendrocytes, and oligodendrocyte precursors, by overlapping the PLS1+/- genes lists with previously reported gene sets under each of the seven cell types⁷⁵. Then, a permutation test was conducted on the number of the genes belonging to each cell type. Finally, we performed functional enrichment analysis based on gene list within each cell type to investigate the cellular signatures associated with the abnormal WM RFCS.

Data availability

Imaging data supporting this study are publicly available dataset from the OASIS-3 dataset (https://central.xnat.org). Human gene expression data in this study are available from the Allen Brain Atlas (“Complete normalized microarray datasets”, https://human.brain-map.org/static/download). Cell-specific gene set list can be obtained at https://static-content.springer.com/esm/art%3A10.1038%2Fs41467-020-17051-5/MediaObjects/41467_2020_17051_MOESM8_ESM.xlsx.

Code availability

All analytical procedures in current study are based on publicly available software. The neuroimaging preprocessing, functional connectivity and graph theoretical analysis is performed using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/), DPARSF (V4.4_180801, http://rfmri.org/dparsf) and GRETNA (v2.0.0, http://www.nitrc.org/projects/gretna/). Transcriptional level matrices were obtained by using the abagen toolbox (v.0.1.4, https://github.com/rmarkello/abagen) on the AHBA dataset (http://human.brain-map.org). Codes for PLS can be found at https://github.com/SarahMorgan/Morphometric_Similarity_SZ. Gene enrichments were performed at https://metascape.org/gp/#/main/step1. The brain maps were visualized with the BrainNet Viewer (v1.7, https://www.nitrc.org/projects/bnv/).

Acknowledgements

This work was supported by grants from the National Natural Science Foundation of China (grant number: NSFC82250410380, NSFC62171101), Natural Science Foundation of Sichuan Province (grant number: 24NSFSC6257) and the China MOST2030 Brain Project (grant number: 2022ZD0208500). The cell cartoon components in Fig. 1 were from Servier Medical Art, provided by Servier, licensed under a Creative Commons Attribution 3.0 unported license (https://creativecommons.org/licenses/by/3.0/).

Author contributions

B.B. and P.W. led the project. Y.L., P.W. and B.B. were responsible for the study concept and the design of the study. Y.L. and P.W. analyzed the neuroimaging data. J.P. analyzed the transcriptional data. Y.L., J.P., L.L and F.Z. created the figures. Y.L. wrote the manuscript. Y.L., P.W., Y.I.-M., D.Y. and B.B. revised the draft. All authors approved the final version to be submitted for publication and agree to be accountable for all aspects of this work.

Competing interests

The authors declare no potential conflict of interest.

Alzheimer’s disease facts and figures. Alzheimer’s & Dementia 19, 1598–1695 (2023). (2023)
Jack, C.R., et al.: NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimer’s Dement. 14, 535–562 (2018)
Iturria-Medina, Y., Evans, A.C.: On the central role of brain connectivity in neurodegenerative disease progression. Front. Aging Neurosci. 7, 1–10 (2015)
Pereira, J.B., et al.: Amyloid and tau accumulate across distinct spatial networks and are differentially associated with brain connectivity. eLife. 8, e50830 (2019)
Mijalkov, M., et al.: Nonlinear changes in delayed functional network topology in Alzheimer’s disease: relationship with amyloid and tau pathology. Alz Res. Therapy. 15, 112 (2023)
Vogel, J.W., et al.: Connectome-based modelling of neurodegenerative diseases: towards precision medicine and mechanistic insight. Nat. Rev. Neurosci. 24, 620–639 (2023)
Wang, J., et al.: Disrupted Functional Brain Connectome in Individuals at Risk for Alzheimer’s Disease. Biol. Psychiatry. 73, 472–481 (2013)
Ding, Z., et al.: Detection of synchronous brain activity in white matter tracts at rest and under functional loading. Proc. Natl. Acad. Sci. U.S.A. 115, 595–600 (2018)
Huang, Y., et al.: Intracranial electrophysiological and structural basis of BOLD functional connectivity in human brain white matter. Nat. Commun. 14, 3414 (2023)
Ji, G.-J., et al.: White matter dysfunction in psychiatric disorders is associated with neurotransmitter and genetic profiles. Nat. Mental Health. 1, 655–666 (2023)
Li, M., Gao, Y., Ding, Z., Gore, J.C.: Power spectra reveal distinct BOLD resting-state time courses in white matter. Proc. Natl. Acad. Sci. U.S.A. 118, e2103104118 (2021)
Huang, Y., et al.: Detection of functional networks within white matter using independent component analysis. Neuroimage. 222, 117278 (2020)
Peer, M., Nitzan, M., Bick, A.S., Levin, N., Arzy, S.: Evidence for Functional Networks within the Human Brain’s White Matter. J. Neurosci. 37, 6394–6407 (2017)
Wang, P., et al.: Structural and functional connectivity mapping of the human corpus callosum organization with white-matter functional networks. Neuroimage. 227, 117642 (2021)
Wang, P., et al.: White Matter Functional Connectivity in Resting-State fMRI: Robustness, Reliability, and Relationships to Gray Matter. Cereb. Cortex. 32, 1547–1559 (2022)
Nasrabady, S.E., Rizvi, B., Goldman, J.E., Brickman, A.M.: White matter changes in Alzheimer’s disease: a focus on myelin and oligodendrocytes. ACTA NEUROPATHOL. COM. 6, 22 (2018)
Taylor, A.N.W., et al.: Tract-specific white matter hyperintensities disrupt neural network function in Alzheimer’s disease. ALZHEIMERS Dement. 13, 225–235 (2017)
Gao, Y., et al.: Functional connectivity of white matter as a biomarker of cognitive decline in Alzheimer’s disease. PLoS ONE. 15, e0240513 (2020)
Rubinov, M., Sporns, O.: Complex network measures of brain connectivity: uses and interpretations. Neuroimage. 52, 1059–1069 (2010)
Yu, H., Lei, X., Song, Z., Liu, C., Wang, J.: Supervised Network-Based Fuzzy Learning of EEG Signals for Alzheimer’s Disease Identification. IEEE Trans. Fuzzy Syst. 28, 60–71 (2020)
Li, J., et al.: White-matter functional topology: a neuromarker for classification and prediction in unmedicated depression. Transl Psychiatry. 10, 365 (2020)
Jiang, Y., et al.: Characteristics of disrupted topological organization in white matter functional connectome in schizophrenia. Psychol. Med. 52, 1333–1343 (2022)
Adewale, Q., Khan, A.F., Carbonell, F., Iturria-Medina, Y.: Alzheimer’s Disease Neuroimaging Initiative. Integrated transcriptomic and neuroimaging brain model decodes biological mechanisms in aging and Alzheimer’s disease. eLife. 10, e62589 (2021)
Bellenguez, C., et al.: New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nat. Genet. 54, 412–436 (2022)
Fornito, A., Arnatkevičiūtė, A., Fulcher, B.D.: Bridging the Gap between Connectome and Transcriptome. Trends Cogn. Sci. 23, 34–50 (2019)
Hibar, D.P., et al.: Common genetic variants influence human subcortical brain structures. Nature. 520, 224–229 (2015)
Lin, H., et al.: Disrupted white matter functional connectivity in aMCI APOEε4 carriers: a resting-state study. Brain Imaging Behav. 15, 1739–1747 (2021)
Yu, H., et al.: Morphological connectivity differences in Alzheimer’s disease correlate with gene transcription and cell-type. Hum. Brain. Mapp. 44, 6364–6374 (2023)
Hawrylycz, M.J., et al.: An anatomically comprehensive atlas of the adult human brain transcriptome. Nature. 489, 391–399 (2012)
Mori, S., et al.: Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage. 40, 570–582 (2008)
Desikan, R.S., et al.: An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage. 31, 968–980 (2006)
Markello, R.D., et al.: Standardizing workflows in imaging transcriptomics with the abagen toolbox. eLife 10, e72129 (2021)
Abdi, H.: Partial least squares regression and projection on latent structure regression (PLS Regression). WIREs Comp. Stat. 2, 97–106 (2010)
Wightman, D.P., et al.: A genome-wide association study with 1,126,563 individuals identifies new risk loci for Alzheimer’s disease. Nat. Genet. 53, 1276–1282 (2021)
Sheline, Y.I., Raichle, M.E.: Resting State Functional Connectivity in Preclinical Alzheimer’s Disease. Biol. Psychiatry. 74, 340–347 (2013)
Jones, D.T., et al.: Cascading network failure across the Alzheimer’s disease spectrum. Brain. 139, 547–562 (2016)
Targa Dias Anastacio, H., Matosin, N., Ooi, L.: Neuronal hyperexcitability in Alzheimer’s disease: what are the drivers behind this aberrant phenotype? Transl Psychiatry. 12, 257 (2022)
Cheyuo, C., et al.: Connectomic neuromodulation for Alzheimer’s disease: A systematic review and meta-analysis of invasive and non-invasive techniques. Transl Psychiatry. 12, 490 (2022)
Gonzalez, C.E., et al.: Peripheral sphingolipids are associated with variation in white matter microstructure in older adults. Neurobiol. Aging. 43, 156–163 (2016)
Bendlin, B.B., et al.: White matter is altered with parental family history of Alzheimer’s disease. Alzheimer’s Dement. 6, 394–403 (2010)
Fathian, A., et al.: The trend of disruption in the functional brain network topology of Alzheimer’s disease. Sci. Rep. 12, 14998 (2022)
Mastria, G., et al.: Morphology, Connectivity, and Encoding Features of Tactile and Motor Representations of the Fingers in the Human Precentral and Postcentral Gyrus. J. Neurosci. 43, 1572–1589 (2023)
Feng, W., et al.: Corticospinal tract lesion load: An imaging biomarker for stroke motor outcomes. Ann. Neurol. 78, 860–870 (2015)
LaFlamme, E.M., Waguespack, H.F., Forcelli, P.A., Malkova, L.: The Parahippocampal Cortex and its Functional Connection with the Hippocampus are Critical for Nonnavigational Spatial Memory in Macaques. Cereb. Cortex. 31, 2251–2267 (2021)
Berro, D.H., Herbet, G., Duffau, H.: New insights into the anatomo-functional architecture of the right sagittal stratum and its surrounding pathways: an axonal electrostimulation mapping study. Brain Struct. Funct. 226, 425–441 (2021)
Spagna, A., Hajhajate, D., Liu, J., Bartolomeo, P.: Visual mental imagery engages the left fusiform gyrus, but not the early visual cortex: A meta-analysis of neuroimaging evidence. Neurosci. Biobehavioral Reviews. 122, 201–217 (2021)
Zhang, X., et al.: Association of Visual Health With Depressive Symptoms and Brain Imaging Phenotypes Among Middle-Aged and Older Adults. JAMA Netw. Open. 5, e2235017 (2022)
Zarkali, A., et al.: Fiber-specific white matter reductions in Parkinson hallucinations and visual dysfunction. Neurology. 94, e1525–e1538 (2020)
Lee, S.W., et al.: White matter abnormalities of the tapetum and their associations with duration of untreated psychosis and symptom severity in first-episode psychosis. Schizophr. Res. 201, 437–438 (2018)
Zhang, J., et al.: Correlations between Dual-Pathway White Matter Alterations and Language Impairment in Patients with Aphasia: A Systematic Review and Meta-analysis. Neuropsychol. Rev. 31, 402–418 (2021)
Benito-León, J., et al.: Diffusion tensor imaging in orthostatic tremor: a tract‐based spatial statistics study. Ann. Clin. Transl Neurol. 6, 2212–2222 (2019)
Graff-Radford, J., et al.: New insights into atypical Alzheimer’s disease in the era of biomarkers. Lancet Neurol. 20, 222–234 (2021)
Cope, T.E., et al.: Tau burden and the functional connectome in Alzheimer’s disease and progressive supranuclear palsy. Brain. 141, 550–567 (2018)
Li, Q., Yang, H., Chen, Y., Sun, H.: Recent progress in the identification of selective butyrylcholinesterase inhibitors for Alzheimer’s disease. Eur. J. Med. Chem. 132, 294–309 (2017)
Xing, S., et al.: Structure and therapeutic uses of butyrylcholinesterase: Application in detoxification, Alzheimer’s disease, and fat metabolism. Med. Res. Rev. 41, 858–901 (2021)
Tan, M.-S., et al.: Associations of Alzheimer’s disease risk variants with gene expression, amyloidosis, tauopathy, and neurodegeneration. Alzheimer’s Res. Therapy. 13, 15 (2021)
Colin, J., et al.: Maintenance of membrane organization in the aging mouse brain as the determining factor for preventing receptor dysfunction and for improving response to anti-Alzheimer treatments. Neurobiol. Aging. 54, 84–93 (2017)
Layfield, R., Alban, A., Mayer, R.J., Lowe, J.: The ubiquitin protein catabolic disorders. Neuropathol. Appl. Neurobiol. 27, 171–179 (2001)
Jiang, Z., et al.: Association between chronic periodontitis and the risk of Alzheimer’s disease: combination of text mining and GEO dataset. BMC Oral Health. 21, 466 (2021)
Griffiths, J., Grant, S.G.: N. Synapse pathology in Alzheimer’s disease. Semin. Cell Dev. Biol. 139, 13–23 (2023)
Mallucci, G.R., Klenerman, D., Rubinsztein, D.C.: Developing Therapies for Neurodegenerative Disorders: Insights from Protein Aggregation and Cellular Stress Responses. Annu. Rev. Cell. Dev. Biol. 36, 165–189 (2020)
Gjoneska, E., et al.: Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer’s disease. Nature. 518, 365–369 (2015)
Kumari, S., Dhapola, R., Reddy, D.H.: Apoptosis in Alzheimer’s disease: insight into the signaling pathways and therapeutic avenues. Apoptosis. 28, 943–957 (2023)
Bai, B., et al.: U1 small nuclear ribonucleoprotein complex and RNA splicing alterations in Alzheimer’s disease. Proc. Natl. Acad. Sci. U.S.A. 110, 16562–16567 (2013)
Tublin, J.M., Adelstein, J.M., Monte, D., Combs, F.: Wold, L. E. Getting to the Heart of Alzheimer Disease. Circ. Res. 124, 142–149 (2019)
Ge, F., Zhu, D., Tian, M., Shi, J.: The Role of Thyroid Function in Alzheimer’s Disease. J. Alzheimers Dis. 83, 1553–1562 (2021)
Tzoneva, R., et al.: The Role of Melatonin on Behavioral Changes and Concomitant Oxidative Stress in icvAβ1–42 Rat Model with Pinealectomy. Int. J. Mol. Sci. 22, 12763 (2021)
Liebner, S., et al.: Functional morphology of the blood-brain barrier in health and disease. Acta Neuropathol. 135, 311–336 (2018)
Lau, S.-F., Cao, H., Fu, A.K.Y., Ip, N.Y.: Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in Alzheimer’s disease. Proc. Natl. Acad. Sci. U.S.A. 117, 25800–25809 (2020)
Ungvari, Z., et al.: Hypertension-induced cognitive impairment: from pathophysiology to public health. Nat. Rev. Nephrol. 17, 639–654 (2021)
LaMontagne, P.J., et al.: OASIS-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset for Normal Aging and Alzheimer Disease. medRxiv (2019). 10.1101/2019.12.13.19014902
Arnatkevic̆iūtė, A., Fulcher, B.D., Fornito, A.: A practical guide to linking brain-wide gene expression and neuroimaging data. NeuroImage. 189, 353–367 (2019)
Morgan, S.E., et al.: Cortical patterning of abnormal morphometric similarity in psychosis is associated with brain expression of schizophrenia-related genes. Proc. Natl. Acad. Sci. U.S.A. 116, 9604–9609 (2019)
Zhou, Y., et al.: Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat. Commun. 10, 1523 (2019)
Seidlitz, J., et al.: Transcriptomic and cellular decoding of regional brain vulnerability to neurogenetic disorders. Nat. Commun. 11, 3358 (2020)

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Transcriptomic, cellular, and functional signatures of white matter damage in Alzheimer’s disease

Status:

Version 1

Abstract

Figures

Introduction

Results

Overall approach

Altered Functional Connectivity in the whole-brain

Global topological properties

Nodal topological organization

Differences in regional functional connectivity strength

Functional disconnection correlates with gene expression level

Biological enrichment pathways of functional disconnection

Cellular signatures

Discussion

Changed functional connectivity in the whole-brain

Abnormalities in gray matter and white matter functional network topology

AD-risk genes related to the functional connectivity changes

Biological processes enriched with PLS1-/PLS1 + gene list

Cellular features

Method

Participants

MRI acquisition

Pre-processing steps of functional images

Obtaining the group-level white matter and gray matter masks

Constructing functional connectivity matrices in white matter and gray matter

Network topological property analysis

Statistical analyses

Relating gene expression to functional changes in brain region

Enrichment analysis

Cell types

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1