Data from: Pathways underlying selective neuronal vulnerability in Alzheimer's disease: Contrasting the vulnerable locus coeruleus to the resilient substantia nigra
Data files
Apr 01, 2025 version files 102.12 GB
-
ABCA1_IHC.zip
13.60 GB
-
HMGCS1_IHC.zip
11.14 GB
-
IHC_Analyzed_ROIs.zip
1.81 GB
-
LDLR_IHC.zip
47.29 GB
-
MYLIP_IHC.zip
11.79 GB
-
README.md
14.92 KB
-
SD1_Metadata_Dryad.xlsx
16.16 KB
-
SD10_DEG_Braak1_uncor.csv
2.36 MB
-
SD11_DEG_Braak2_uncor.csv
2.25 MB
-
SD12_DEG_Braak3_uncor.csv
2.48 MB
-
SD13_Cholesterol_IHC.xlsx
11.27 KB
-
SD2_count_matrix.csv
8.12 MB
-
SD3_cpm_matrix.csv
8.12 MB
-
SD4_PCA_feature_loadings.csv
20.04 MB
-
SD5_DEG_Braak0.csv
2.50 MB
-
SD6_DEG_Braak1.csv
2.35 MB
-
SD7_DEG_Braak2.csv
2.25 MB
-
SD8_DEG_Braak3.csv
2.46 MB
-
SD9_DEG_Braak0_uncor.csv
2.66 MB
-
SREBP2_IHC.zip
16.43 GB
Abstract
Identifying factors underlying selective neuronal vulnerability is crucial for understanding Alzheimer's disease (AD) pathophysiology. The Neuromodulatory Subcortical System (NSS) includes nuclei that exhibit early, but varied vulnerability to tau accumulation and neuronal loss. This varied vulnerability represents a valuable opportunity to explore the underlying mechanisms of AD. In this study, we investigated factors contributing to selective neuronal vulnerability by comparing transcriptomic profiles of two similar NSS nuclei with differing vulnerabilities to AD, the locus coeruleus and substantia nigra. Using paired samples from well-characterized postmortem human tissue from individuals in early Braak stages and free of comorbid neuropathologic diagnoses, we identified pathways related to cholesterol homeostasis and antioxidant pathways response as key potential drivers of vulnerability.
Immunohistochemistry data, supplementary data, and supplementary methods
https://doi.org/10.5061/dryad.sbcc2frgp
Trancriptomic analysis
Description of the data and file structure
The frozen half of the brainstem for selected cases was kept on dry ice during the dissection. A scalpel was used to shave down the midbrain until the pigmented SN was exposed. An additional 3-5mm of midbrain was shaved down around the rostral portion of the SN, with borders defined by the pigmented area. The protruding portion of the SN was sliced off of the shaved-down face of the midbrain and put into RNAlater (AM7020, Invitrogen) to protect RNA in case of thawing during transport for processing. The sample in RNAlater was frozen down at -80C and transported on dry ice.
The LC was isolated by excising a tissue block approximately 5-10mm in length along the rostrocaudal axis and about 10mm in depth from the fourth ventricle near the medial eminence. After removal, the pigmented area of the LC was identified. Tissue outside the pigmented LC border was then discarded. The isolated LC tissue was subsequently preserved in RNAlater and frozen.
The LC and SN samples were sent to a vendor (Novogene Inc., Davis, CA) for RNA extraction and sequencing. Tissue homogenization and cell lysis were performed in TRIzol. Following cell lysis, impurities were removed, and RNAse activity was inhibited. Total RNA was then extracted using a phase separation method to remove cell debris. RNA quality and concentration were assessed using an Agilent Bioanalyzer 2100. Samples with an RNA integrity number (RIN) greater than 4 and at least 0.1 µg of RNA were advanced to library preparation and sequencing. Library preparation, including poly-A enrichment, was conducted using the NEBNext Ultra II RNA Library Prep Kit for Illumina. Sequencing was performed on the Illumina NovaSeq 6000 platform, generating 150 bp paired-end (PE150) reads for each sample.
The quality of sequence files was assessed using the FastQC package before and after trimming steps. Trimmomatic (ILLUMINACLIP: TruSeq3-PE.fa:2:30:10:2:True, LEADING: 3, TRAILING: 3, SLIDINGWINDOW: 4:15, and MINLEN: 36) was used to remove adapter sequences and any sequences with low mean quality scores. Sequences were aligned to GRCh38 using STAR alignment and count matrices were generated using featureCounts. The count matrices were converted to counts per million (CPM) using the edgeR package. Expression levels of positive control genes (DBH, SLC6A2, and SLC6A3) were checked to confirm accurate sampling. Cases with less than 100 CPM of DBH or 10 CPM of SLC6A2 in the LC, or less than 100 CPM or SLC6A3 in the SN were excluded from subsequent analyses.
Principal component analyses (PCA) were performed using the RunPCA command from the Seurat package in R with 20 principal components computed. Differential expression analyses were performed using the edgeR package [34]. Only genes with at least 5 counts present in at least 25% of samples were included in differential expression analyses. Multiple comparison correction was done using the Benhamini-Hochberg method and the upper bound of the expected false discovery rate (FDR) is reported in analyses as the FDR. Differential expression analyses set cutoffs at 0.05 FDR and +/- 0.5 log-fold change (logFC). The design matrix for differential expression analyses is formed from an additive model formula including the case. We ran models that were uncorrected and corrected for RNA Integrity Number.
Gene set enrichment analysis (GSEA) was done using the fgsea package in R. The Hallmark gene set collection was downloaded from the GSEA Molecular Signatures Database. The gene set files were filtered to contain only genes present in the differential expression table (5% FDR and +/- 0.5 log-fold change) and then reformatted to the specifications required by the fgsea package. The differentially expressed genes (DEGs) were ranked based on the negative log10(p-value) such that the genes with the smallest p-values were ranked at the very top of the ranking list. The fgsea function was run on the filtered gene set files with the ranked genes using a minimum gene set size of 10 and maximum of 500.
Scripts used to process and analyze the transcriptomic data are provided here and referenced to in the manuscript as “Supplementary Methods 1”. The raw transcriptomic data are available on GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE273787) and the processed data are provided here.
Scripts used
“Supplementary methods 1” in published manuscript.
File: trim_and_qc.sh
Description: Trimming and quality control script
File: STAR_index_exec.sh
Description: Generation of index genome using STAR
File: STAR_align.sh
Description: STAR alignment script
File: featureCounts_exec.sh
Description: Feature counts execution script
File: featureCounts.R
Description: Feature counts script
File: NSS_Vulerability_pub.R
Description: Statistical analysis script of processed transcriptomic data
R Package Software Versions
| Package | Version |
|---|---|
| gridExtra | 2.3 |
| stringr | 1.5.1 |
| ggpubr | 0.6.0 |
| corrplot | 0.95 |
| data.table | 1.16.4 |
| fgsea | 1.32.0 |
| ggVennDiagram | 1.5.2 |
| ggbeeswarm | 0.7.2 |
| ggrepel | 0.9.6 |
| org.Hs.eg.db | 3.20.0 |
| AnnotationDbi | 1.68.0 |
| IRanges | 2.40.1 |
| S4Vectors | 0.44.0 |
| Biobase | 2.66.0 |
| BiocGenerics | 0.52.0 |
| Seurat | 5.1.0 |
| SeuratObject | 5.0.2 |
| sp | 2.1-4 |
| edgeR | 4.4.1 |
| limma | 3.62.1 |
| tibble | 3.2.1 |
| forcats | 1.0.0 |
| tidyr | 1.3.1 |
| dplyr | 1.1.4 |
| ggplot2 | 3.5.1 |
| ggridges | 0.5.6 |
| readxl | 1.4.3 |
Data and metadata
Raw transcriptomic data from this paper are available at GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE273787).
File: SD1_Metadata_Dryad.xlsx
Supplementary data 1: Metadata for cases used in transcriptomics analysis
File: SD2_count_matrix.csv
Supplementary data 2: Processed transcript counts
File: SD3_cpm_matrix.csv
Supplementary data 3: Processed counts per million
File: SD4_PCA_feature_loadings.csv
Supplementary data 4: Feature loadings for principal component analysis
File: SD5_DEG_Braak0.csv
Supplementary data 5: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage 0 cases with RIN correction built into the paired model (~case+RIN+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD6_DEG_Braak1.csv
Supplementary data 6: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage I cases with RIN correction built into the paired model (~case+RIN+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD7_DEG_Braak2.csv
Supplementary data 7: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage II cases with RIN correction built into the paired model (~case+RIN+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD8_DEG_Braak3.csv
Supplementary data 8: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage III cases with RIN correction built into the paired model (~case+RIN+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD9_DEG_Braak0_uncor.csv
Supplementary data 9: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage 0 cases without RIN correction built into the paired model (~case+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD10_DEG_Braak1_uncor.csv
Supplementary data 10: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage I cases without RIN correction built into the paired model (~case+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD11_DEG_Braak2_uncor.csv
Supplementary data 11: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage II cases without RIN correction built into the paired model (~case+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
File: SD12_DEG_Braak3_uncor.csv
Supplementary data 12: Differentially expressed genes between the locus coeruleus and substantia nigra in Braak stage III cases without RIN correction built into the paired model (~case+Nucleus). Fields: logFC, log fold change of differential expression; logCPM, log of the counts per mission of the transcripts; F, F-statistic; PValue, unadjusted p-value; FDR, Benjamini-Hochberg corrected p-value; Ensembl, Ensembl transcript identifier; Gene, Gene name.
Immunohistochemistry analysis
Description of the data and file structure
Formalin-fixed paraffin embedded tissue cut at 8µm was cut from the pons at the level of the LC and the midbrain at the level of the SN for 13 postmortem cases (Table 2). Sections were deparaffinized and underwent antigen retrieval using the Discovery CC1 solution (950-500) for 40 minutes at 95oC. Sections were labeled with HMGCS1 (1:400, ab155787), ABCA1 (1:200, NB400-105), and LDLR (1:200, 10785-1-AP) developed with the Discovery Purple kit (760-229) and counterstained with hematoxylin, MYLIP (1:50, 15455-1-AP) developed with the Discovery Green HRP kit (760-721) and counterstained with hematoxylin, and SREBP2 (1:200, PA1338) developed with the Discovery Purple kit with no counterstaining. The OmniMap anti-Rabbit HRP (760-4311) was used for secondary detection of all primary antibodies. A set of serial LC slides run without primary antibody or counterstain were treated with the anti-Rabbit HRP secondary antibody and purple chromogen to verify SREBP2 labeling in the LC.
Stained sections were scanned on a Zeiss Axioscan slide scanner. Staining was assessed semi-quantitatively by selecting 500x500 µm regions of interest and scoring all the neurons in the frame on a scale from 0 to +++. A score of “+” corresponds to a neuron with staining that differentiated the cell from being completely devoid of signal, but without more than very faint diffuse signal or, in the case of granular staining, several speckles. A score of “++” corresponds to a cell with moderate-to-strong diffuse staining or a speckled granular staining occupying most, but not all, of the cell. A score of “+++” corresponds to a cell with very strong diffuse staining or granular staining occupying the entirety of the cell represented in the section. These scores were converted to numbers (0-3, respectively), the proportions of neurons with each score were determined based on the total neuron numbers counted, and an average score was computed for each nucleus for each case. Scores were analyzed using a paired Wilcox test to compare average scores between nuclei for cases at Braak stage 0-II and Braak stage VI.
Each zipped folder contains scans of files in the Zeiss file format (.czi) which is a slide scan for one region per stain per case. The scores assigned to each case are provided in tabular format.
Antibodies used
- HMGCS1 (1:400, ab155787)
- ABCA1 (1:200, NB400-105)
- LDLR (1:200, 10785-1-AP)
- MYLIP (1:50, 15455-1-AP)
- SREBP2 (1:200, PA1338)
Regions
- LC: Pons section featuring the locus coeruleus
- SN: Midbrain section featuring the substantia nigra
Code/software
Zeiss ZEN lite: https://www.zeiss.com/microscopy/de/produkte/software/zeiss-zen-lite.html
File: IHC analysis script.R
Description: Script used to generate figures in the manuscript
Scores and Files
File: IHC Analyzed ROIs.zip
Description: Analyzed regions of interest for all scanned slides
File: SD13_Cholesterol_IHC.xlsx
Description: Quantification of IHC staining.
File: ABCA1_IHC.zip
Description: IHC staining for ABCA1HMGCS1
File: HMGCS1_IHC.zip
Description: IHC staining for HMGCS1
File: LDLR_IHC.zip
Description: IHC staining for LDLR
File: MYLIP_IHC.zip
Description: IHC staining for MYLIP
File: SREBP2_IHC.zip
Description: IHC staining for SREBP2
Human subjects data
This dataset is derived from human research subjects who enrolled in a research study approved by a site-specific IRB. Participants, or a next-of-kin, provided informed consent permitting the use of and sharing of de-identified data. All personally identifiable information are removed when data are shared using technical controls approved by the relevant IRB’s.
We leveraged transcriptomics and immunohistochemistry in paired samples from human postmortem tissue representing a vulnerable and resilient region – the locus coeruleus (LC) and substantia nigra (SN). These regions have comparable anatomical features but distinct vulnerability to AD.
Participant selection and neuropathologic assessment
Cases were sourced from the Biobank for Aging Studies at the University of São Paulo and the Neurodegenerative Disease Brain Bank at the University of California, San Francisco Memory and Aging Center which is an ADRC. Consent for brain donation was obtained from subjects or next of kin following the site-specific protocol approved by the relevant Institutional Review Board and the study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki. In both brain banks, brain tissue was sampled for neuropathological diagnosis following NACC guidelines. Basic histological and immunohistochemical stains were made with the antibodies for phospho-Ser202 Tau (CP13, 1:250, courtesy of P. Davies), TDP-43 (1:2000, 10782-2-AP, Proteintech), β-amyloid (1:500, MAB5206, Millipore), α-Synuclein (1:5000, LB509, courtesy of J. Trojanowsky and V. Lee). Final neuropathologic diagnosis was made by following current guidelines.
For this study, we included cases with Braak stages 0-III that had available frozen LC and SN tissue (Supplementary Data 1). We excluded cases with Lewy body inclusions, TDP-43 proteinopathy, a primary or contributing diagnosis of chronic traumatic encephalopathy, a primary or contributing diagnosis of a primary tauopathy, an Axis I psychiatric disorder, a post-mortem interval over 24 hours, or gross non-neurodegenerative structural neuropathology. Initial case selection sought equal proportions of males and females and informant-reported race within Braak stages, as well as a similar age distribution between Braak stages.
Sample collection and processing
The frozen half of the brainstem for selected cases was kept on dry ice during the dissection. A scalpel was used to shave down the midbrain until the pigmented SN was exposed. An additional 3-5mm of midbrain was shaved down around the rostral portion of the SN, with borders defined by the pigmented area. The protruding portion of the SN was sliced off of the shaved-down face of the midbrain and put into RNAlater (AM7020, Invitrogen) to protect RNA in case of thawing during transport for processing. The sample in RNAlater was frozen down at -80oC and transported on dry ice.
The LC was isolated by excising a tissue block approximately 5-10mm in length along the rostrocaudal axis and about 10mm in depth from the fourth ventricle near the medial eminence. After removal, the pigmented area of the LC was identified. Tissue outside the pigmented LC border was then discarded. The isolated LC tissue was subsequently preserved in RNAlater and frozen.
Tissue homogenization and cell lysis were performed in TRIzol. Following cell lysis, impurities were removed, and RNAse inhibition, total RNA was extracted using phase separation. RNA quality and concentration were assessed using an Agilent Bioanalyzer 2100. Samples with an RNA integrity number (RIN) greater than 4 and at least 0.1 µg of RNA were advanced to library preparation and sequencing. Libraries were prepared from RNA using the NEBNext Ultra II RNA Library Prep Kit for Illumina, which included poly-A enrichment using poly-T oligo-attached magnetic beads. mRNA was fragmented, and first-strand cDNA synthesis was performed using random hexamer primers, followed by second-strand cDNA synthesis. Sequencing was performed on the Illumina NovaSeq 6000 platform, generating 150 bp paired-end (PE150) reads for each sample. Extraction, library preparation, and sequencing was performed by Novogene Inc. (Davis, CA).
Transcriptomic analysis
The quality of sequence files was assessed using the FastQC package before and after trimming steps. Trimmomatic (ILLUMINACLIP: TruSeq3-PE.fa:2:30:10:2:True, LEADING: 3, TRAILING: 3, SLIDINGWINDOW: 4:15, and MINLEN: 36) was used to remove adapter sequences and any sequences with low mean quality scores. Sequences were aligned to GRCh38 using STAR alignment and count matrices were generated using featureCounts. The count matrices were normalized using the counts per million (CPM) function in the edgeR package. Expression levels of positive control genes (DBH, SLC6A2, and SLC6A3) were checked to confirm accurate sampling. Cases with less than 100 CPM of DBH or 10 CPM of SLC6A2 in the LC, or less than 100 CPM or SLC6A3 in the SN were excluded from subsequent analyses.
Principal component analyses (PCA) were performed using the RunPCA command from the Seurat package in R with 20 principal components computed. Differential expression analyses were performed using the edgeR package. Only genes with at least 5 counts present in at least 25% of samples were included in differential expression analyses. Multiple comparison correction was done using the Benhamini-Hochberg method and the upper bound of the expected false discovery rate (FDR) is reported in analyses as the FDR. Differential expression analyses set cutoffs at 0.05 FDR and +/- 0.5 log-fold change (logFC) to balance statistical significance with the need to detect changes limited to small cell populations in the bulk sample. The design matrix for differential expression analyses is formed from an additive model formula including the case. Differential expression analyses were done both with (corrected) and without (uncorrected) RIN as a covariate in the design formula.
Gene set enrichment analysis (GSEA) was done using the fgsea package in R. The Hallmark gene set collection was downloaded from the GSEA Molecular Signatures Database. The gene set files were filtered to contain only genes present in the differential expression table (5% FDR and +/- 0.5 log-fold change) and then reformatted to the specifications required by the fgsea package. The differentially expressed genes (DEGs) were ranked based on the negative log10(p-value) such that the genes with the smallest p-values were ranked at the very top of the ranking list. The fgsea function was run on the filtered gene set files with the ranked genes using a minimum gene set size of 10 and maximum of 500.
Immunohistochemistry (IHC)
Formalin-fixed paraffin embedded tissue cut at 8µm was cut from the pons at the level of the LC and the midbrain at the level of the SN for 13 postmortem cases (Table 2). Sections were deparaffinized and underwent antigen retrieval using the Discovery CC1 solution (950-500) for 40 minutes at 95oC. Sections were labeled with HMGCS1 (1:400, ab155787), ABCA1 (1:200, NB400-105), and LDLR (1:200, 10785-1-AP) developed with the Discovery Purple kit (760-229) and counterstained with hematoxylin, MYLIP (1:50, 15455-1-AP) developed with the Discovery Green HRP kit (760-721) and counterstained with hematoxylin, and SREBP2 (1:200, PA1338) developed with the Discovery Purple kit with no counterstaining. The OmniMap anti-Rabbit HRP (760-4311) was used for secondary detection of all primary antibodies. A set of serial LC slides run without primary antibody or counterstain were treated with the anti-Rabbit HRP secondary antibody and purple chromogen to verify SREBP2 labeling in the LC.
Stained sections were scanned on a Zeiss Axioscan slide scanner. Staining was assessed semi-quantitatively by selecting 500x500 µm regions of interest and scoring all the neurons in the frame on a scale from 0 to +++. A score of “+” corresponds to a neuron with staining that differentiated the cell from being completely devoid of signal, but without more than very faint diffuse signal or, in the case of granular staining, several speckles. A score of “++” corresponds to a cell with moderate-to-strong diffuse staining or a speckled granular staining occupying most, but not all, of the cell. A score of “+++” corresponds to a cell with very strong diffuse staining or granular staining occupying the entirety of the cell represented in the section. These scores were converted to numbers (0-3, respectively), the proportions of neurons with each score were determined based on the total neuron numbers counted, and an average score was computed for each nucleus for each case. Scores were analyzed using a paired Wilcox test to compare average scores between nuclei for cases at Braak stage 0-II and Braak stage VI.