21 Jun 2023

How the brain’s intricate vasculature deteriorates during Alzheimer’s disease has only recently come into focus, thanks to advances in isolating and studying cerebrovascular cells. In a single-nucleus RNA sequencing analysis published in the June 1 Nature Neuroscience, researchers led by Li-Huei Tsai and Manolis Kellis at Massachusetts Institute of Technology, Boston, reported almost 2,700 differentially expressed genes in vascular cells taken from six brain regions of 220 people with AD. In ApoE4 carriers, the cells expressed yet different transcriptomes. In AD, the cells ramped up genes involved in the immune response and suppressed those needed to maintain blood-brain barrier integrity. AD risk genes played a role, with 125 of the genes being linked to risk variants in previous genome-wide association analyses.

SnRNA-Seq identifies 11 subtypes of neurovascular cells.
In people with Alzheimer’s disease, 2,700 genes dysregulated.
Of these, 125 are known to carry AD risk variants.
Vascular gene expression in ApoE4 carriers is distinct from that in AD.

“This study presents the most detailed transcriptomic analysis thus far of regional differences in blood-brain barrier vasculature in Alzheimer’s disease at the single-cell level,” wrote Berislav Zlokovic, University of Southern California, Los Angeles, and Ruslan Rust, University of Zurich, Switzerland (comments below). Jonathan Kipnis, Steffen Storck and Leon Smyth of Washington University School of Medicine in St. Louis noted that this study’s use of sorting- and enrichment-free sequencing provides an unbiased representation of cell types and an unprecedented resolution of the human brain vasculature.

Still, some were concerned that the number of vascular cells used in the analysis was low. Kipnis, Storck, and Smyth noted that an average of only five cells per cell type and per individual, taken across six different brain regions, were used in the analysis. “This is a big caveat especially given the heterogeneity that exists in human samples and the fact that cell types across individuals are not captured evenly,” they wrote (comments below).

First author Na Sun used snRNA-Seq to capture the transcriptomes of cells from prefrontal, entorhinal, and temporal cortices, the hippocampus, thalamus, and angular gyrus tissue taken postmortem from the 220 people who had had AD and from 208 age-matched controls. All were part of the Religious Orders Study and Rush Memory and Aging Project (ROSMAP) cohort. Most had died in their 80s or 90s, and AD cases ranged from early to late-stage disease.

Sun isolated 22,514 total cerebrovascular cells, which clustered into 11 subtypes based on their transcriptomes: three endothelial subtypes, three fibroblast, two pericyte, two smooth muscle cells, and one cluster of ependymal cells. Cell abundance varied across brain regions (see image below). Gene expression among the same cell types varied across brain regions. For example, endothelial cells in the prefrontal cortex, but no other region, overexpressed genes involved in antigen presentation and cell response to interferons. Sun found a total of 1,745 of such inter-regional differentially expressed genes (DEGs).

Regional Differences. The thalamus (yellow), entorhinal cortex (green), and hippocampus (cyan) contain fewer endothelial cells and pericytes and more fibroblasts and ependymal cells than other brain regions. [Courtesy of Sun et al., Nature Neuroscience, 2023.]

What about changes in AD? While the proportion of cells was the same as in controls, gene expression varied. AD cases up- or downregulated 2,676 genes, most of which were cell-type-specific. Among the top upregulated genes were those encoding Apolipoprotein D in pericytes and the α1 subunit of type IV collagen in endothelial cells. ApoD expression is high in postmortem AD brain tissue, and a fragment of type IV α1 collagen suppresses angiogenesis (Bhatia et al., 2019; reviewed by Mundel and Kalluri, 2007). As for downregulated genes, pericytes stifled expression of the GABA transporter gene SLC6A1 and the growth factor receptor gene PDGFRB. Using immunohistochemistry and in situ hybridization, Sun corroborated that up- and downregulation held true in postmortem tissue samples as well. Endothelial cells suppressed ABCB1, which encodes the efflux pump, P-glycoprotein. It may help flush Aβ from the brain and seems to pump less efficiently in people living with AD (see minireview by Elali and Rivest, 2013; van Assema et al., 2012).

Cytokines, including interleukin-17, were prevalent among the upregulated genes of pericytes, endothelial cells, and fibroblasts, indicating responses linked to inflammation. According to gene ontology analysis, pericytes downregulated synaptic transmission processes, which for these cells equates to poorly sensing signals from neurons, the authors suggest. Cytoskeleton remodeling also took a hit, which might reflect weak control of blood flow by these cells.

Vascular cells altered the way they communicated with neurons and glia in AD in other ways, as well. By grouping DEGs according to cell types that express them, then focusing on ligand-receptor pairs, Sun and colleagues found that endothelial cells and pericytes appeared to signal more to neurons, microglia, and astrocytes in AD, and that astrocytes and neurons communicated less with endothelial cells and fibroblasts. Among the tempered signaling pathways were those involving multiple types of collagens and laminins and their cognate integrin binding partners, key components of the extracellular matrix supporting the basement membrane of the BBB, hinting that that barrier might be compromised.

The researchers wondered if the DEGs were related to AD risk genes. They integrated their transcriptomics analysis with previous genome-wide association study data (Mar 2019 news; Kang et al., 2021). For a third of the 2,676 DEGs, a GWAS variant fell within, or near, an intron, lay within a gene for a transcription factor that likely regulated the DEG, or landed within a gene for a binding partner. For example, 197 GWAS variants fell in or near introns of 125 DEGs (see image below). Among the 197 were four variants in the cholesterol efflux gene ABCA1, a gene Sun found to be upregulated in AD pericytes.

**GWAS and the Vasculature.** Sun identified 125 genes containing AD risk variants that are up- (red dotted lines) or downregulated (blue dotted lines) in AD tissue. Expression changes occur in endothelial cells (yellow, tan, and brown), fibroblasts (light, medium, and dark green), pericytes (light and dark blue), smooth muscle cells (pink and mauve), and ependymal cells (purple). *[Courtesy of Sun et al., Nature Neuroscience, 2023.]*

APOE4, the strongest genetic risk factor for sporadic AD, also influenced vascular gene expression. Comparing snRNA-Seq data from 108 people who carried one or two copies of APOE4 to transcriptomes of 251 who carried two copies of APOE3, Sun found 2,482 DEGs, of which just 4 percent overlapped with AD DEGs. Among the most downregulated were those involved in cell junction maintenance, transport of molecules across the BBB, and angiogenesis. These perhaps reflect poor cerebral blood flow seen in older E4 carriers, the authors speculate (Wierenga et al., 2013; Kim et al., 2012).

Certain DEGs in APOE4 carriers also correlated with global cognitive decline. These were genes up-/downregulated only in fibroblasts, endothelial cells, and pericytes. Among the latter two, upregulated genes included those involved in lipid and cytokine signaling and apoptosis. Suppressed genes were involved in BBB maintenance, blood vessel development, and neurotransmitter transport. “This supports BBB breakdown contributing to APOE4-associated cognitive decline independent of Aβ and tau pathology,” wrote Nela Fialova and Axel Montagne, University of Edinburgh (comments below).

This study adds to the growing body of evidence that helps scientists paint a better picture of what goes awry in the vasculature during AD. Scientists led by Tony Wyss-Coray at Stanford University, Palo Alto, California, used snRNA-Seq to create an atlas of vascular and perivascular gene expression based on transcriptomes of almost 150,000 cells, including arterial, capillary, and venule endothelial cells and arterial versus venule smooth muscle cells. Wyss-Coray also found massive expression changes in AD brain samples, with vascular cells highly expressing dozens of AD risk genes (May 2021 news). In another snRNA-Seq analysis of endothelial cells, scientists led by Rachel Bennett at Massachusetts General Hospital, Charlestown, reported regional variation in gene expression across five cortical regions. Taking it a step further, Bennett spotted dramatic transcriptional changes in the presence of amyloid plaques and cerebral amyloid angiopathy (Mar 2023 news).—Chelsea Weidman Burke

Comments

- Berislav Zlokovic
  University of Southern California
- Ruslan Rust
  University of Southern California
- Posted: 22 Jun 2023
Sun et al. performed a thorough, single-nuclei analysis of the blood-brain barrier (BBB) vasculature from six brain regions using a large and well-described cohort of Alzheimer’s disease (AD) patients and controls from Rush Memory and Aging Project (ROSMAP) at Rush University Medical Center. By employing state-of-the art computational methods, the researchers were able to elucidate cell-to-cell interactions and establish links between molecular and cellular changes, genetic variants, and ApoE genotypes.

This study presents the most detailed transcriptomic analysis thus far of regional differences in blood-brain barrier (BBB) vasculature in Alzheimer’s disease at single-cell level. The study findings support the mounting evidence of early vascular involvement in AD and this work could potentially guide future treatments aimed at mitigating BBB dysfunction in AD.

The study revealed profound transcriptomic heterogeneity in the BBB vasculature. The identified multiregional gene expression differences and pathway enrichments could contribute to understanding why certain brain regions, such as hippocampus and medial temporal lobe, are more susceptible to BBB breakdown, particularly in ApoE4 carriers, as we have reported earlier (see Montagne et al., 2020).

By employing innovative computational methods, the researchers were able to elucidate a shift in cellular interaction in AD, with increased communication from capillary endothelial cells to neurons, microglia, and astrocytes, contrasted by a reduced reciprocal interaction. This asymmetry could provide important cues for potential interventions on a multicellular level for future therapies.

Future research could extend these findings by exploring these changes at a proteomic level and by aiming to understand the causal relationships among ApoE genotype, BBB functionality, and cognitive impairment in AD.

References:

Montagne A, Nation DA, Sagare AP, Barisano G, Sweeney MD, Chakhoyan A, Pachicano M, Joe E, Nelson AR, D'Orazio LM, Buennagel DP, Harrington MG, Benzinger TL, Fagan AM, Ringman JM, Schneider LS, Morris JC, Reiman EM, Caselli RJ, Chui HC, Tcw J, Chen Y, Pa J, Conti PS, Law M, Toga AW, Zlokovic BV. APOE4 leads to blood-brain barrier dysfunction predicting cognitive decline. Nature. 2020 May;581(7806):71-76. Epub 2020 Apr 29 PubMed.

View all comments by Berislav Zlokovic View all comments by Ruslan Rust

- Jonathan Kipnis
  Washington University in St. Louis, School of Medicine
- Steffen Storck
  Washington University of St. Louis School of Medicine
- Leon Smyth
  Washington University in St. Louis
- Posted: 22 Jun 2023
Studies suggest that vascular alterations precede the classical hallmarks of Alzheimer’s disease (AD), such as amyloid accumulation, tau pathology, and cognitive decline. This study by Sun et al. used sorting-free and enrichment-free single nuclei sequencing of 428 individuals (220 AD and 208 controls) for an unbiased representation of cell types to characterize cell-type-specific changes of vascular cells in the AD brain. This study follows three other reports from last year, which provide the first comprehensive, highly powered view of the human brain vasculature at single cell resolution (Garcia et al., 2022; Winkler et al., 2022; Yang et al., 2022).

Like Yang et al., the authors show the presence of heterogeneity in pericytes in the human brain, something not found in mice. Alongside dissection of zonation and heterogeneity of vascular cell types, the authors show regional heterogeneity of cells associated with the blood-brain barrier. Capillary endothelial cells, the majority of endothelial cells (ECs) in the brain, showed the highest number of differentially expressed genes (DEGs) in AD, which included cell junction and adhesion-associated genes as well as genes involved in BBB transport. DEGs in arterial ECs included tight junction components. All these changes suggest that important transport and barrier properties are altered at the BBB in AD.

Interestingly, the authors found ABCB1, the transcript that encodes for P-glycoprotein, downregulated in AD individuals. P-gp, an amyloid-transporter on the blood side of BBB endothelial cells (Cirrito et al., 2005) has been suggested to efflux brain amyloid-β in concert with LRP1 receptor (Storck et al., 2016; Storck et al., 2018). Loss of P-gp will result in a decreased efflux of amyloid-β across the BBB.

The authors also found immune activation (cytokines, cytokine receptors, interleukin-17 signaling, and inflammatory response) of endothelial cells, pericytes, and fibroblasts, highlighting the role of the immune system in AD. We were pleased to see that the authors included perivascular cells, such as fibroblasts, in their dataset, since their role in neurological disease is just beginning to be uncovered, and little is known about their contribution to AD (Dorrier et al., 2022; Drieu et al., 2022).

It would have been interesting to see any changes to perivascular macrophages in these datasets because they regulate cerebrospinal fluid dynamics, waste clearance, and Aβ-mediated neurovascular dysfunction (Park et al., 2017; Drieu et al., 2022) and, just like fibroblast, are situated in close proximity to the vessel in the perivascular space.

A major limitation is the number of cells per individual captured in the study. The authors sequenced 22,514 vascular cells from 428 individuals (220 AD and 208 controls) so an overall of 53 vascular cells per individual was analyzed. They further annotated 11 different cell types (three types of endothelial cells, two types of pericytes, three types of fibroblasts, ependymal cells, and two types of vascular smooth muscle cells), so an average of five cells per cell type and individual across six different brain regions were analyzed.

This is a big caveat especially given the heterogeneity that exists in human samples and the fact that cell types across individuals are not captured evenly. With this in mind, we must carefully consider how well those few cells, captured across different brain regions, actually reflect the overall cell states in that individual’s brain.

Together with Yang et al., Garcia et al., and Winkler et al., Sun et al. provides unprecedented resolution of the human brain vasculature, and strong evidence that multiple cell types in the brain vasculature become dysfunctional in neurological diseases.

References:

Cirrito JR, Deane R, Fagan AM, Spinner ML, Parsadanian M, Finn MB, Jiang H, Prior JL, Sagare A, Bales KR, Paul SM, Zlokovic BV, Piwnica-Worms D, Holtzman DM. P-glycoprotein deficiency at the blood-brain barrier increases amyloid-beta deposition in an Alzheimer disease mouse model. J Clin Invest. 2005 Nov;115(11):3285-90. PubMed.

Drieu A, Du S, Storck SE, Rustenhoven J, Papadopoulos Z, Dykstra T, Zhong F, Kim K, Blackburn S, Mamuladze T, Harari O, Karch CM, Bateman RJ, Perrin R, Farlow M, Chhatwal J, Dominantly Inherited Alzheimer Network, Hu S, Randolph GJ, Smirnov I, Kipnis J. Parenchymal border macrophages regulate the flow dynamics of the cerebrospinal fluid. Nature. 2022 Nov;611(7936):585-593. Epub 2022 Nov 9 PubMed.

Dorrier CE, Jones HE, Pintarić L, Siegenthaler JA, Daneman R. Emerging roles for CNS fibroblasts in health, injury and disease. Nat Rev Neurosci. 2022 Jan;23(1):23-34. Epub 2021 Oct 20 PubMed.

Garcia FJ, Sun N, Lee H, Godlewski B, Mathys H, Galani K, Zhou B, Jiang X, Ng AP, Mantero J, Tsai LH, Bennett DA, Sahin M, Kellis M, Heiman M. Single-cell dissection of the human brain vasculature. Nature. 2022 Mar;603(7903):893-899. Epub 2022 Feb 14 PubMed.

Park L, Uekawa K, Garcia-Bonilla L, Koizumi K, Murphy M, Pistik R, Younkin L, Younkin S, Zhou P, Carlson G, Anrather J, Iadecola C. Brain Perivascular Macrophages Initiate the Neurovascular Dysfunction of Alzheimer Aβ Peptides. Circ Res. 2017 Jul 21;121(3):258-269. Epub 2017 May 17 PubMed.

Storck SE, Meister S, Nahrath J, Meißner JN, Schubert N, Di Spiezio A, Baches S, Vandenbroucke RE, Bouter Y, Prikulis I, Korth C, Weggen S, Heimann A, Schwaninger M, Bayer TA, Pietrzik CU. Endothelial LRP1 transports amyloid-β(1-42) across the blood-brain barrier. J Clin Invest. 2016 Jan;126(1):123-36. Epub 2015 Nov 30 PubMed.

Storck SE, Hartz AM, Bernard J, Wolf A, Kachlmeier A, Mahringer A, Weggen S, Pahnke J, Pietrzik CU. The concerted amyloid-beta clearance of LRP1 and ABCB1/P-gp across the blood-brain barrier is linked by PICALM. Brain Behav Immun. 2018 Oct;73:21-33. Epub 2018 Jul 21 PubMed.

Winkler EA, Kim CN, Ross JM, Garcia JH, Gil E, Oh I, Chen LQ, Wu D, Catapano JS, Raygor K, Narsinh K, Kim H, Weinsheimer S, Cooke DL, Walcott BP, Lawton MT, Gupta N, Zlokovic BV, Chang EF, Abla AA, Lim DA, Nowakowski TJ. A single-cell atlas of the normal and malformed human brain vasculature. Science. 2022 Mar 4;375(6584):eabi7377. PubMed.

Yang AC, Vest RT, Kern F, Lee DP, Agam M, Maat CA, Losada PM, Chen MB, Schaum N, Khoury N, Toland A, Calcuttawala K, Shin H, Pálovics R, Shin A, Wang EY, Luo J, Gate D, Schulz-Schaeffer WJ, Chu P, Siegenthaler JA, McNerney MW, Keller A, Wyss-Coray T. A human brain vascular atlas reveals diverse mediators of Alzheimer's risk. Nature. 2022 Mar;603(7903):885-892. Epub 2022 Feb 14 PubMed.

View all comments by Jonathan Kipnis View all comments by Steffen Storck View all comments by Leon Smyth

- Nela Fialova
  University of Edinburgh
- Axel Montagne
  University of Edinburgh
- Posted: 22 Jun 2023
In recent years, several publications have focused on creating an atlas of mouse and human brain vasculature using vascular enrichment protocols (Vanlandewijck et al., 2018; Garcia et al., 2022; Yang et al., 2022). This study by Sun et al. explored the regional differences of vascular transcriptome in healthy and Alzheimer’s disease (AD) brains using an unbiased single-nucleus RNA-sequencing approach. Given the importance of vascular contribution to AD, identifying the associated transcriptomic changes is of the utmost importance to expand our knowledge of the underlying mechanisms.

Looking at six postmortem brain regions in controls and in AD individuals, Sun et al. identified 11 vascular cell types including three types of endothelial cells (aEndo, cEndo, vEndo) and two types of pericytes (PER1 and PER2). These findings are in agreement with previously identified cell types, following a vessel enrichment protocol (Yang et al., 2022), however, it seems that a higher number of fibroblast cells was sourced using the unbiased approach. Unfortunately, Sun et al. do not expand further on mural cell type function and zonation as compared to Garcia et al. (2022) or Yang et al. (2022), who identified the transport and matrix subtypes based on the genetic signatures. Furthermore, an extensive heterogeneity of the blood-brain barrier was observed with different gene expressions among the cells at different brain regions expanding the findings of Yang et al. (2022).

Moving on to the AD samples, Sun et al. did not observe any significant change in vascular cell numbers in AD individuals compared to controls. Instead, they observed downregulation of various cell marker genes, such as the PDGFRb transcripts in pericytes, which is in agreement with recent literature highlighting the role of pericyte and BBB integrity in the disease (Sweeney et al., 2019; Montagne et al., 2015). This decrease in canonical cell marker genes could lead to lower proportions of cells in publications that utilize vascular enrichment for single-cell isolation and immunohistochemistry experiments, making this unbiased approach advantageous. Furthermore, upregulation of a high-density lipoprotein component gene APOD was observed in pericytes. This highlights the functional role of pericytes in neurovascular unit lipid transport, which should be investigated further in line with the observed dysregulation of insulin signalling genes in multiple vascular cell types in AD.

Interestingly, based on the link between AD genome-wide association study (GWAS) loci and differentially expressed vascular genes (DEGs), Sun et al. suggest AD genetic risk factors contribute to pathology via intracellular dysfunction and intercellular communication of the vasculature and glial, and the microglial and neural cells. Additionally, APOE4 samples showed more cognitive-decline-related DEGs in capillary endothelia, pericyte 1, and fibroblast type 1 cells compared to individuals with APOE3, proposing a cerebrovascular mediation of the APOE ε4 allele effect on cognitive decline. This supports the findings of BBB breakdown contribution to APOE4-associated cognitive decline independent of Aβ and tau pathology (Montagne et al., 2020), which we linked to cyclophilin A-matrix metalloproteinase-9 BBB-degrading pathway activation in pericytes (Montagne et al., 2021). It would be interesting to see if there are any brain regional differences in the association between the AD DEGs and AD risk genes or APOE4 genotype. This was not discussed in the study, even though multiple brain areas were investigated. Furthermore, the spatial relationship between vascular molecular changes and Aβ plaques/tau tangles was not investigated, limiting our understanding of how distinct genetic expression relates to pathology accumulation along the vascular axis.

Sun et al. bring a unique approach to vasculature transcriptome analysis, omitting the vessel enrichment protocol, instead sorting for vasculature in silico, post-sequencing, providing an unbiased dataset from a technical point of view. However, with so many isolated cells and transcripts, the depth of sequencing can be compromised, leaving out some nuanced gene changes, which would be even more difficult to detect with larger areas, or whole brains. Furthermore, even though a couple of DEGs were validated via RNA in situ hybridization and immunohistochemistry, the majority of the AD DEGs, APOE DEGs, signalling pathways, and cell-cell interactions were identified via a computational prediction and need to be further supported experimentally in the future.

Overall, this study confirms the importance of investigating the vascular contribution to AD. It sheds light on how upstream regulators control AD differential genes and highlights the dynamics of the cell-cell communication in the neurovascular unit, which may lead to an easier selection of therapeutic targets in the future. Lastly, in agreement with the authors, a thorough investigation of the causal relationship of cognitive decline, APOE genotype, and BBB function is necessary to further tease out the role of vascular cells in AD pathobiology.

References:

Garcia FJ, Sun N, Lee H, Godlewski B, Mathys H, Galani K, Zhou B, Jiang X, Ng AP, Mantero J, Tsai LH, Bennett DA, Sahin M, Kellis M, Heiman M. Single-cell dissection of the human brain vasculature. Nature. 2022 Mar;603(7903):893-899. Epub 2022 Feb 14 PubMed.

Montagne A, Barnes SR, Sweeney MD, Halliday MR, Sagare AP, Zhao Z, Toga AW, Jacobs RE, Liu CY, Amezcua L, Harrington MG, Chui HC, Law M, Zlokovic BV. Blood-brain barrier breakdown in the aging human hippocampus. Neuron. 2015 Jan 21;85(2):296-302. PubMed.

Montagne A, Nation DA, Sagare AP, Barisano G, Sweeney MD, Chakhoyan A, Pachicano M, Joe E, Nelson AR, D'Orazio LM, Buennagel DP, Harrington MG, Benzinger TL, Fagan AM, Ringman JM, Schneider LS, Morris JC, Reiman EM, Caselli RJ, Chui HC, Tcw J, Chen Y, Pa J, Conti PS, Law M, Toga AW, Zlokovic BV. APOE4 leads to blood-brain barrier dysfunction predicting cognitive decline. Nature. 2020 May;581(7806):71-76. Epub 2020 Apr 29 PubMed.

Montagne A, Nikolakopoulou AM, Huuskonen MT, Sagare AP, Lawson EJ, Lazic D, Rege SV, Grond A, Zuniga E, Barnes SR, Prince J, Sagare M, Hsu CJ, LaDu MJ, Jacobs RE, Zlokovic BV. Author Correction: APOE4 accelerates advanced-stage vascular and neurodegenerative disorder in old Alzheimer's mice via cyclophilin A independently of amyloid-β. Nat Aging. 2021 Jul;1(7):624. PubMed.

Sweeney MD, Kisler K, Montagne A, Toga AW, Zlokovic BV. The role of brain vasculature in neurodegenerative disorders. Nat Neurosci. 2018 Oct;21(10):1318-1331. Epub 2018 Sep 24 PubMed.

Vanlandewijck M, He L, Mäe MA, Andrae J, Ando K, Del Gaudio F, Nahar K, Lebouvier T, Laviña B, Gouveia L, Sun Y, Raschperger E, Räsänen M, Zarb Y, Mochizuki N, Keller A, Lendahl U, Betsholtz C. A molecular atlas of cell types and zonation in the brain vasculature. Nature. 2018 Feb 14; PubMed.

Yang AC, Vest RT, Kern F, Lee DP, Agam M, Maat CA, Losada PM, Chen MB, Schaum N, Khoury N, Toland A, Calcuttawala K, Shin H, Pálovics R, Shin A, Wang EY, Luo J, Gate D, Schulz-Schaeffer WJ, Chu P, Siegenthaler JA, McNerney MW, Keller A, Wyss-Coray T. A human brain vascular atlas reveals diverse mediators of Alzheimer's risk. Nature. 2022 Mar;603(7903):885-892. Epub 2022 Feb 14 PubMed.

View all comments by Nela Fialova View all comments by Axel Montagne