Background
Dysfunction of brain vasculature and the blood–brain barrier (BBB) has emerged as an important comorbidity and modifier of brain diseases. Furthermore, increased BBB permeability has been reported to occur during aging [
1,
2]. A better understanding of BBB and vascular changes during aging and in brain diseases is essential for the development of vasoprotective therapies, and for facilitating drug delivery into diseased brain. A reductionist approach using loss-of-function mouse models in a single genetic background has been successful in identifying genes important for BBB development [
3]. However, this approach will most likely be ineffective for identifying genes and pathways in humans that influence BBB alterations during aging and in brain diseases. Mouse studies have shown that genotype–phenotype relationships cannot be reliably inferred by studying a single genetic background [
4] because inbred strains fail to mimic the genetic and physiological complexity in humans. The genetic variability of each individual can have profound effects on the presentation of even monogenic diseases [
5]. The identification of genetic modifiers in humans is very difficult due to differences in environmental factors and the vast genetic variability in the natural population. Furthermore, studies on human brain vasculature (and brain in general) are hindered due to the scarcity of tissue available for such analyses. Most studies on BBB development as well as alterations of the BBB in aging and disease (e.g. Alzheimer’s, ischemic stroke) are carried out in a C57BL/6J background, the strain with the lowest number of single nucleotide polymorphisms (SNPs) [
6]. However, this limited genetic variability can be overcome by applying a systems biology approach and investigating several inbred mouse lines and reference populations. This strategy provides a controlled approach to study variation that influences a certain phenotype. Experimental models of cardiovascular disease or adaptation to exercise largely depend on genetic background and differences in endothelial function (e.g. ACh dose–response curves) [
7,
8]. In the CNS, inbred mouse strains show differences in neurogenesis [
9‐
11], in behavioral performance [
12], ischemic stroke volume [
13], regenerative capacities [
14], and cerebral collateral vessel density [
15]. Furthermore, the genetic background of mice influences the deposition of amyloid and myeloid-driven neuroinflammation in murine models of Alzheimer’s disease [
16].
The BBB is a dynamic interface that actively responds to changes in neural tissue and metabolism. The genetic background of mouse models may not only affect adaptive changes but also severity of BBB alterations during aging and brain diseases. Characterization of the phenotypic heterogeneity of the BBB in genetically distinct mouse strains may yield insights into how genetic variation influences the BBB characteristics.
In this study, we investigated the phenotypic heterogeneity of vasculature and endothelial cells (EC) using several inbred mouse strains: 129S1/SvlmJ, DBA/2J, A/J, C57BL6J, NOD/ShiltJ, WSB/EiJ, PWK/PhJ and CAST/EiJ. WSB/EiJ, PWK/PhJ and CAST/EiJ are recently wild-derived (1980-s) and present highest genetic variation whereas 129S1/SvlmJ, DBA/2J, A/J, C57BL6J, NOD/ShiltJ strains present very limited intra-strain polymorphism [
6,
17,
18]. We characterized brain vasculature using three phenotypic arms—analysis of vascular density, arteriovenous (A-V) zonation and pericyte coverage, BBB permeability to sodium fluorescein, and transcriptomic analysis of EC using RNA sequencing. Investigated inbred strains showed similar vessel density and pericyte coverage, endothelial zonation pattern, and permeability to sodium fluorescein in the cortex and hippocampus. Transcriptomic analysis of brain EC revealed sex dimorphism in genes involved in mitochondrial function. Transcriptome of EC from recently wild-derived and long-inbred strains revealed differential expression of genes involved in vascular development and homeostasis, immune system response, and membrane transport. To summarize, we describe the spectrum of variation in intact BBB and brain vasculature in healthy young mice of different strains as well as age-related changes in some long-inbred strains.
Discussion
In this study, we characterized the BBB of eight inbred mouse strains; A/J, C57BL/6 J, CAST/EiJ, DBA/2J, NOD/ShiLtJ, PWK/PhJ, WSB/EiJ, and 129S1/SvImJ using several phenotypic arms. We quantified vessel density, pericyte coverage, BBB permeability to NaF, and characterized endothelial A-V zonation pattern. In addition, we performed genome-wide transcriptional profiling of brain EC.
Inbred mouse strains differ in collateral vessel density [
15]; however, less is known regarding the extent of differences of brain vascular topology between inbred mouse strains. Recent methodological advances have already yielded brain-wide imaging of vasculature of different strains (C57BL/6J, CD1 and BALB/c) [
33]. Our study complements and extends previous investigations. Although not brain-wide, we provide a high-resolution characterization of A-V zonation pattern, vessel density and pericyte coverage in different young inbred mouse strains in two brain regions. The analyzed strains showed variable vessel density in the hippocampus. Interestingly, wild-derived PWK/PhJ showed decreased vessel density in the cortex compared to NOD/ShiLtJ and greater pericyte coverage in the cortex than DBA/2J. Analysis of vessel density and pericyte coverage in aged (12 month old) A/J, C57BL6/J and DBA/2J mice, which have a different life span, showed reduced vascular density in the hippocampus in old DBA/2J mice. Previous studies have shown that vessel density and pericyte coverage decreases during ageing in C57BL6/J mice, both in the brain and in peripheral organs with low regenerative capacity [
34,
35]. We did not detect reduced vascular density in brains of C57BL6/J mice. However, our analysis was performed in relatively thin (22 µm) z-stack confocal images in two brain regions. Although our images for quantification of vascular parameters were of high resolution, it is likely that we did not reach high enough volumes to detect changes in vessel density and pericyte coverage.
We investigated whether different mouse strains present different baseline permeability to NaF, a small molecular weight tracer. As expected, we did not detect statistically significant differences in permeability to NaF among the investigated mouse strains. However, NOD/ShiLtJ and CAST/EiJ mice showed higher PI in the cortex compared to other strains. Previous studies in C57BL6/J animals have shown a higher vascular permeability to NaF in the neurogenic region along the lateral wall of the lateral ventricle [
25]. Except for A/J, our analysis of BBB permeability did not show an elevated PI index in the hippocampus in C57BL6/J animals and other investigated strains. Thus, if neurogenic zones in a mouse brain have a higher BBB permeability compared to other regions, it might be restricted to a certain zone and/or this characteristic is strain specific. Prolonged isoflurane anesthesia in old C57BL/6J female mice was shown to influence the BBB permeability [
36]. CAST/EiJ were anesthetized with isoflurane before tail vein injection, which could potentially increase BBB permeability. However, PWK/PhJ and WSB/EiJ strains, which show a similar PI as non-anesthetized strains (e.g. C57BL/6J), were also anesthetized before tail vein injection. Therefore, if isoflurane anesthesia had caused an increase in BBB permeability, this would indicate a strain specific effect. In healthy mice, the brain penetration mode of NaF has been described to be passive [
37], however, NaF is a general substrate for transporters of the SLCO family and MRP2 (
ABCC2) [
38,
39]. Several SLCO transporters are expressed by mouse brain EC [
22] and SLCO2B1 is expressed by human brain endothelium [
40]. Interestingly, streptozotocin-induced diabetes led to decreased vascular permeability to fluorescein paralleled by increased expression of Mrp2 in rat brain [
41], a transporter that has been shown to be differentially expressed in cerebral blood vessels in C57BL/6J and FVB strains [
42]. However, our transcriptional analysis did not identify any differences in expression levels of transcripts encoding for Slco1c1, Slco2b1 and Mrp2 within analyzed strains. In accordance with published literature [
2,
27], we detected an increase in BBB permeability during aging, both in the hippocampus and cortex. BBB permeability changes in cortex within three investigated strains showed a different slope. It would be interesting to investigate whether the observed differences are due to the genetic background of animals.
The investigated mouse strains did not show major differences in BBB permeability to NaF in young mice. However, due to differences in the genetic background, we assumed that the transcriptome of EC isolated from different stains would show gene expression changes reflecting normal variation of the transcriptome BBB. Unexpectedly, the largest variation in the obtained RNAseq dataset was not due to the genetic background of mice but sex. Interestingly, we found that EC isolated from female mice express higher levels of genes associated with mitochondrial function. Female and male cells show differences in mitochondrial function (e.g. oxygen consumption, ATP production) in various tissues in both rodents and human [
43‐
45]. In addition, analyses of genetically diverse mice have shown that mitochondrial function is also dependent on genetic background [
46]. Mitochondrial dysfunction is associated with diseases, including cardiovascular diseases and neurodegeneration. Since vascular dysfunction modifies brain diseases, future studies should investigate the impact of mitochondrial dysfunction due to sex differences and its potential impact on pathology and tissue repair. A recent study by Paik et al. [
30] identified higher expression of
Ddx3y, Zbtb16, Gm6981, Mapk6 genes in EC of male brain and
Xist, Tsix, Arglu1, Twf1 and Plp1 genes in EC of female brain [
30]. The authors concluded the existence of brain EC-specific sex-dependent genes. However, many of the identified genes are encoded by X- and Y- chromosomes and used as markers for sex determination (e.g.
Ddx3y,
Xist). Thus, these findings cannot represent a true organ-specific EC difference. Similar to other studies [
30,
47], we identified sexually dimorphic expression of
Lars2, a gene encoding a leucyl-tRNA synthase 2 catalyzing the ligation of leucine to its cognate tRNAs. Mutations in
LARS2 cause Perraut syndrome, a rare autosomal recessive disorder causing sensory sensorineural hearing loss in both genders and ovarioleukodystrophy in females [
48]. It remains to be determined whether Lars2 affects mitochondrial function in a gender dependent manner and whether dysfunction of brain endothelial cells contributes to the ovarioleukodystrophy. A recent proteomic study of rat brain microvessels also identified sexual dimorphism with females expressing more mitochondrial proteins involved in energy production including the OXPHOS pathway [
49].
We compared the EC transcriptome of recently wild-derived with long-inbred strains, which show low inter-strain polymorphism. We identified differential expression of several EC-specific genes. Our results corroborated previous findings that Vwf expression differs between inbred strains (e.g. lower levels in C57BL/6 J mice compared to WSB/EiJ mice) [
50]. Vwf is a key protein for vascular homeostasis and angiogenesis. In addition, several genes implicated in the formation of vasculature or regulating endothelial responses in different pathological settings (
Lrp8, Angpt2, Cxcl12) [
51‐
53] are differently expressed in inbred strains. We found that a brain endothelial specific gene
Lrp8 (Apoer2) is upregulated in EC of recently wild-derived mice. Apoer2 mediating Reelin signaling in EC is necessary for cerebral cortex vascularization and BBB maturation [
53]. Genetic variation in the LRP8 gene in humans was shown to influence myocardial infarction and the early onset of coronary artery disease [
54]. Interestingly, the binding of ApoE3 to ApoER2 stimulates endothelial NO synthase (eNOS) [
55]. It would be interesting to investigate whether differences in the expression of
Vwf, Lrp8 genes and transporters/receptors (e.g.
Slc7a5, Insr, Lepr, Stra6), and immune response genes (e.g.
B2m,
Ifnar1, H2-D) in EC have functional consequences in the brains of adult mice during aging or pathological process. For example, whereas the contribution of Stra6-mediated transport to total retinol uptake by tissues other than the eye is rather modest, Stra6 has been shown to induce endothelial inflammation via circulating RBP [
56,
57]. Interestingly, several ATP-dependent efflux pumps, which could affect bioavailability of drugs in the CNS, show a high polymorphism in human population that affects expression level and function [
58]. Also, our analysis has identified differential expression of transcripts encoding ABC transporters (Abcg1, Abcg4, Abcg3) in wild-derived and long-inbred strains. Further studies are needed to determine whether changes in mRNA level alter protein level and activity.
In summary, we provide a detailed description of BBB variation in healthy young and aged mice of different inbred strains. Further analysis of complex BBB trait in inbred mice during aging and pathological conditions is likely to yield insight on the influence of sex and genetic diversity on the BBB response during different metabolic conditions (e.g. fasting, high-fat diet), response to injury, and repair processes in mice. Future challenges include the development of methods that will allow quantitative assessment of individual BBB properties (e.g. transport properties, EC junctional organization, inflammatory state etc.) and linking these characteristics with various functional outcomes (e.g. blood flow, neuronal activity).
Materials and methods
Mice
Mice were bred in house and pups were separated at postnatal day 21. The following mouse strains were used for the experiments—A/J, 129S1/SvlmJ, C57BL/6J, NOD/ShiLtJ, PWK/PhJ, WSB/EiJ, CAST/EiJ, DBA/2J. Age of mice was 8 weeks ± 2 weeks or 12 months ± 3 weeks. Both, male and female mice were used. All mice were housed in Type 2L cages (530 cm2, max 5 mice per cage) with individual ventilation under specific-pathogen-free conditions and a 12 h light/dark cycle. Cage environment included a red polycarbonate house and shredded paper nesting material. Water and food (cat #3336, #3436 KLIBA NAFAG) were provided ad libitum. All mice were housed in the same animal house, in the same room. Experimental procedures were approved by the Cantonal Veterinary Office Zurich (ZH151/2017).
Analysis of BBB integrity using sodium fluorescein
Sodium fluorescein (NaF, Sigma F6377) was diluted in 0.9% NaCl (8 mg/ml) and 1.2 mg was injected into the tail vein. The tracer circulated for 2 h. Before restraining and tail vein injection, CAST/EiJ, WSB/EiJ and PWK/PhJ mice were anesthetized with isoflurane using an anesthesia system with a vaporizer and chamber (VetEquip.Inc) for 3–4 min because of difficulties in capturing and holding recently wild-derived strains [
59]. However, mice were conscious during the tracer circulation time. After 2 h, animals were deeply anesthetized using Ketamine (200 mg/kg) and Xylazine (20 mg/kg). After a loss of response to reflex stimulation (toe-pinch), they were transcardially perfused with PBS. Prior to perfusion, cardiac blood was withdrawn and serum was collected for further analysis using a BD microtainer (BD, #365,968). After perfusion with PBS, brains were removed and hippocampi and cerebral cortex were dissected. Dissected tissue was weighed and lysed in 1% Triton X-100 in PBS using a Qiagen TissueLyser LT (2 × 5 min, 50 osc/s) and centrifuged for 10 min at RT, 13,000 rpm. Supernatants were collected for further analysis. Tissue supernatants and serum were analyzed in 96 well plates (Thermo Scientific, 265,301) using a spectrophotometer from Spectramax Paradigm. Samples were excited using wavelength of 492 nm and signal was collected in a bottom read mode in 2.5 mm reading height an emission wavelength of 525 nm (Software-Soft Max Pro 6.2.). Brain permeability to NaF was calculated using a formula (modified from Devraj et al. [
24]).
$$PI=\frac{{tissue (g)fluorescence}_{injected}- \underline{x} ({tissue (g)fluorescence}_{uninjected})}{{serum \left(mL\right)fluorescence}_{injected}-\underline{x} ({serum \left(mL\right)fluorescence}_{uninjected})}$$
Data analysis was performed in Microsoft Excel and GraphPad Prism 8.
Immunohistochemistry
Mice were deeply anesthetized using Ketamine (200 mg/kg) and Xylazine (20 mg/kg) after a loss of response to reflex stimulation (toe-pinch) were transcardially perfused with PBS and 4% PFA in PBS, pH 7.2. After perfusion, brains were removed and kept in 4% PFA in PBS, pH 7.2 at 4 °C for 5 h for post fixation. Seventy µm thick sagittal sections of brains were sectioned with a Leica VT1000S vibratome. Free floating brain sections were blocked overnight at 4 °C using blocking solution (1% BSA, 0.1% Triton X-100 in PBS). Primary antibodies were incubated for two days at 4 °C. Before adding secondary antibodies, sections were washed 3 × 10 min with 0.5% BSA, 0.05% Triton X-100 in PBS. The following primary antibodies were used: rabbit anti-mouse collagen-IV (Bio-Rad, 2150–1470, 1:300), goat anti-mouse CD13 (R&D Systems, AF2335, 1:100), mouse anti-human ASMA-FITC (Sigma, F37777, 1:100), goat anti-mouse podocalyxin (R&D Systems, AF1556, 1:100), rabbit anti-mouse Slc16a1 (Origene, TA321556, 1:100), rabbit anti-mouse Vwf (DAKO, A0082, 1:100),. Secondary antibodies were incubated overnight at 4 °C, followed by a wash 3 × 10 min in 0.5% BSA, 0.05% Triton X-100 in PBS. All fluorescently labeled secondary antibodies suitable for multiple labeling (donkey anti-goat IgG (H + L) Cy3 (#705-165-147), donkey anti-goat IgG (H + L) AlexaFluor647 (#705–605-147), donkey anti-rabbit IgG (H + L) AlexaFluor488 (#711-545-152), donkey anti-rabbit IgG (H + L) Cy3 (#711-165-152) were purchased from Jackson Immunoresearch. After 7 min incubation in a 1:10,000 4′,6-diamidino-2-phenylindole (DAPI) (Sigma, 10236276001) solution, sections were washed with PBS and mounted in Prolong Gold Antifade Reagent (Invitrogen, P36930). Sections were imaged using a confocal microscope (Leica SP5, objective 20 × and numerical aperture 0.7). Imaris software (Bitplane) and ImageJ were used for image processing and analysis. Salt-and-pepper noise was removed using a median filter with a radius of 1 pixel (ImageJ) or a median filter of 3 × 3x1 (Imaris).
Quantification of vessel density and pericyte coverage
Sagittal Sections (70 µm thick) were stained with antibodies against collagen IV (COL-IV) and aminopeptidase N (CD13). Images of 22 µm z-size were analyzed using FIJI version 11 [
60]. For vessel density, a slightly modified vessel analysis plugin (version 1.1) in FIJI was used, where 3 regions of interest (ROI) were measured in images taken from hippocampus or from cortex. To determine vessel pericyte coverage, images were converted to binary and 6 (150 μm × 150 µm) regions of interest were analyzed in COL-IV and CD13 channel using the area measurement tool. The percentage of signal overlay was calculated to determine the vessel pericyte coverage.
Isolation of brain endothelial cells using flow cytometry
Two months old mice were deeply anesthetized using 25% Ketamine (50 mg/mL), 10% Xylazine (20 mg/mL) and after a loss of response to reflex stimulation (toe-pinch) the brain was removed. Hippocampus and cortex were dissected, coarsely mechanically dissociated using scissors, followed by enzymatical dissociation with 500 µl per brain region of a dissociation solution containing 1.55 g/L glucose, 0.08 W U/ml Liberase DH (Sigma, 5401054001), 100 U/ml DNaseI (Sigma, D4263-5VL) in HBSS (Gibco, 14065-049), at 37 °C for 20 min. After 10 min, tissue was gently mixed using a glass Pasteur pipette. After enzymatic dissociation, tissue was gently pushed 20 times through a 20G needle. For size exclusion, cells were passed through a 40 µm mesh and washed with 10 ml PBS. After washing and pelleting (300 g, 10 min, 4 °C), cells were prepared for myelin removal using myelin removal beads (130–096-731) and LS columns (130-042-401) according to the manufacturer`s protocol (Miltenyi Biotec). After myelin removal, cells were washed with FACS buffer (2% FBS, 0.01% NaN3 in PBS) and subsequently stained with directly conjugated antibodies for 30 min in 100 µl antibody solution. The following antibodies were used: anti-mouse CD45-PE (BD Bioscience, 553081, 1:20), anti-mouse CD11b-APC-Cy7 (BD Biosciences, 557657, 1:20) and anti-mouse CD31-APC (BD Biosciences, 551262, 1:20). Cells were washed with FACS buffer and kept in 200 µl FACS buffer until analysis. Live and dead staining using 7AAD-PerCP-Cy5.5 (559925, BD Biosciences) was performed directly before sorting. All procedures after tissue dissociation were performed on ice or 4 °C. OneComp eBeads™ (ThermoFischer Scientific, 01.1111.42) diluted in PBS were used for compensation. For each compensation, 0.5 µl antibody was added to the compensation beads. Live and dead staining was compensated by using a small fraction of stained sample for live and dead staining only.
Sorting of endothelial cells was performed using an ARIA III 5L cell sorter (BD) with 85 µm nozzle size, 1000 events/s with a flow rate of 1. Gating analyses included single cell gating, live cell gating (7AAD), negative gating for immune cells (CD45+, CD11b+) and positive gating for endothelial cells (CD31+). Cells were sorted using a 4-way purity mask directly into the RLT lysis buffer (Qiagen AllPrep Mini Kit, 80004) with β-mercaptoethanol (Sigma, M6250). After sorting, total RNA was isolated using the RNeasy Micro-Kit (Qiagen AllPrep Mini Kit, 80004). RNA quality was checked using an RNA Nano-Kit for BioAnalyzer 6000 (Agilent).
RNA sequencing
RNA sequencing of samples was performed at the Functional Genomics Center (FGCZ) at the University of Zürich and ETH. Total RNA was depleted from rRNA and the library was prepared using the SMARTer Stranded Total RNA-seq Pico Input Mammalian Kit (Cat. 634839, Clontech Laboratories). Samples were sequenced with an Illumina HiSeq4000, single end, with a read length of 125 bp. Sequencing depth was 30–50 Mio reads.
Raw sequencing read alignment and gene count summarization were performed using the bcbio-nextgen pipeline (v.19.03,
https://github.com/bcbio/bcbio-nextgen). Quality encoding of sequencing reads was converted to Sanger format using seqkt (
https://github.com/lh3/seqtk). Sequencing reads were aligned to the
Mus musculus genome (mm10) using HISAT2 (v.2.1.0, [
61]). Sequence alignment map files were converted to binary alignment map files and indexed, and alignment statistics were determined using samtools (v.1.9, [
62]). Read count summarization at the gene level was performed using the featureCounts method of the Subread package (v.1.4.4, [
63]) using the mm10 genome transcriptome annotation (v. 2018-10-10_92). Gene body coverage estimation was performed using the RSeQC package (v.4.0.0, [
64]).
Analysis of the pre-processed raw counts was performed using R packages included in the online web application iDEP.92 [
29]. First, raw counts were converted to log
2 (counts per million (CPM) + 1) using the edgeR package. Genes were filtered to remove the ones expressed at less than 0.5 cpm (counts per million) in at least 1 sample. A principal component analysis (PCA) was performed using the log
2(cpm + 1) data of all retained genes. Genes were centered and the heatmap.2 function of the gplots package was used to generate a heatmap using the 1000 most variable genes. Distance metrics for the sample and gene dendrograms were based on 1-Pearson’s correlation coefficient, followed by hierarchical clustering using the average linkage procedure. Differential gene expression analysis was performed using DESeq2 [
65]. Testing for differential gene expression between different groups (male vs female, cortex vs hippocampus, long inbred vs recently wild-derived) was performed separately with a Benjamini–Hochberg FDR corrected value cutoff of 0.05. For each pairwise comparison, over-representation analysis of Gene Ontology (GO) biological process gene sets was performed for up-regulated and down-regulated genes separately, using hypergeometric tests followed by FDR value adjustment. Expression values of cell type marker genes were calculated as follows: Raw counts were normalized and converted to log
2 using DESeq2. Average expression levels of cell type marker genes (
22) were calculated and visualized with boxplots. RNA sequencing data, both raw data and gene-by-sample matrix of raw counts, were deposited in Gene Expression Omnibus (GEO) under accession number GSE173793.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.