Gene targeting and generation of mutant mice
The generation of Col18a1-/- mice (B6.129S4-Col18a1tm1Hms; http://www.informatics.jax.org/allele/MGI:2179134) has been described [35]. Heterozygous Col18a1+/- mice were backcrossed to the C57BL/6J mouse strain (Jackson Laboratory) for more than 15 generations and the C57BL/6JOlaHsd strain (Harlan) for more than 10 generations and 4 times with the C57BL/6NCrl strain (Charles River) to produce an inbred C57BL/6 Col18a1-/- mouse line. The strain was maintained by means of backcrossing and heterozygous Col18a1+/- matings. Wild-type littermates (Col18a1+/+) were used as controls in all the experimental analyses. The mouse genotypes were confirmed by polymerase chain reaction (PCR) as described previously [36]. Two cohorts were used: one was perfuse-fixed for histological analyses, and the second cohort was used for both histological and q-PCR analyses as described below.
Tissue collection
Tissue preparation for hematoxylin and eosin staining and immunohistochemistry
For tissue preparation, animals were injected s.c. with ketamine (90 mg/kg body weight) and xylazine (18 mg/kg body weight) fentanyl (0,6 mg/kg body weight), midazolam (15 mg/kg body weight) and medetomidin (2,25 mg/kg body weight) in 0.9% NaCl solution and then transcardially perfused with phosphate buffered saline (PBS), followed by 4% paraformaldehyde containing PBS (pump speed of 15 ml/min, duration 240 s each). The following day, the tissue was transferred through PBS washes from PFA to 30% sucrose in PBS for overnight cryoprotection, frozen in methylbutane (Carl Roth GmbH + Co. KG, Karlsruhe, GER) and stored at −80 °C until sectioning. Coronal slices (thickness 30 µm) of the whole brain were prepared using a cryostat and stored in cryo-protection solution (25% ethylene glycol, 25% glycerin, 50% 0.1 M PBS pH 7.4) at 4 °C. From the frontal to the occipital pole, 3 brain slices were taken from each of the 6 sectional planes with a distance of 750 mm between each sectional plane (18 brain slices per animal in total) for hematoxylin and eosin (HE) staining and immunohistochemistry (IHC) analysis. In this study, HE analysis was performed on sections from 7 wild-type (Col18a1+/+) and 5 Col18a1-/- 5-month-old mice as well as from 5 Col18a1+/+ and 6 Col18a1-/- 12-month-old mice. Furthermore, IHC analysis was performed on sections from 6 Col18a1+/+ and 5 Col18a1-/- 5-month-old mice and 5 Col18a1+/+ and 5 Col18a1-/- 12-month-old mice.
Tissue preparation for qPCR
After an overdose with i.p. injection of ketamine (75 mg/kg body weight), xylazine (10 mg/kg body weight) and acepromazine (3 mg/kg body weight), tissue isolation for quantitative PCR (qPCR) was performed by intracardial perfusion with ice-cold PBS (pump speed of 15 ml/min, duration 70 s), quick decapitation, and dissection of the brain into ice-cold PBS. Left hippocampi from the second batch of animals were stored for only RNA extraction and downstream processing. Six Col18a1+/+ and 11 Col18a1-/- 12-month-old mice were used for qPCR analysis.
Histology
For HE staining, brain slices were washed with distilled water twice, incubated with hematoxylin (Carl Roth GmbH + Co. KG, Karlsruhe, GER) for 5 min and washed again, followed by bluing under running tap water for 10 min and another rinse with distilled water. Afterwards, staining was performed with 1% eosin solution (Carl Roth GmbH + Co. KG, Karlsruhe, GER) for 40 s. After dehydration with increasing concentrations of alcohol (Rotisol, Carl Roth GmbH + Co. KG, Karlsruhe, GER), slices were finally placed in Xylene (Carl Roth GmbH + Co. KG, Karlsruhe, GER) and mounted with coverslips using Histomount (Fisher Scientific GmbH, Schwerte, GER).
Quantification
The following CSVD hallmarks were quantified in the HE-stained brain slices: (i) nonocclusive erythrocyte thrombi defined as the accumulation of erythrocytes in the lumen of the vessels (subsequently referred to as erythrocyte thrombi)[37], (ii) small perivascular bleeds/microbleeds defined as leakage of erythrocytes out of the vessel, and (iii) enlarged PVS defined as distended, white, nonstained perivascular areas.
Erythrocyte thrombi and small perivascular bleeds were counted in 25 randomly chosen fields of view (FOV) in different brain regions (retrosplenial cortex (RSC), basal ganglia (BG), hippocampal CA1 region, corpus callosum (CC) and thalamus) in both hemispheres. When quantifying, we additionally differentiated between two different types of vessels: (i) capillaries with a luminal diameter < 10 μm and (ii) small vessels defined through a luminal diameter > 10 μm [38]. For statistical analysis, the mean of all FOVs per brain region was calculated.
For PVS measurements, 18 FOV per brain region (see above) were quantified in each animal. In each FOV, all arterioles and the number of arterioles with PVS in the same FOV were counted. Percentages of arterioles with PVS per FOV were then calculated, and mean values per animal were used for statistical analysis.
RNA extraction, cDNA conversion, and qPCR
A total RNA extract was prepared from the frozen hippocampal region using the EURx GeneMatrix DNA/RNA Extracol kit (Roboklon Cat. No. E3750), following the manufacturer's recommendations [39]. Nano-drops were used to measure the yield, purity, and integrity of RNA. Furthermore, the High-Capacity cDNA Reverse Transcription Kit (Cat. 4368814) was used for the conversion of 1.5 µg of RNA to cDNA, and then real-time (RT)-qPCR was performed using a TaqMan gene expression array (Cat. 4331182) from Thermo-Fisher Scientific using Quant-Studio-5 from Applied Biosystems (Table 1). Overall, 43 genes were analyzed, comprising five chondroitin sulfate proteoglycan (CSPG) genes, namely, aggrecan (Acan), versican (Vcan), neurocan (Ncan), phosphacan (Pcan) and brevican (Bcan), four genes coding for link proteins, hyaluronan and proteoglycan link protein (Hpln2, Hpln3, Hpln4) and tenascin-R (Tnr), five genes for major metalloproteinases degrading neural ECM proteins, namely, matrix metallopeptidase 2 (Mmp2), matrix metallopeptidase 9 (Mmp9), cathepsin S (Ctss), a disintegrin-like and metalloproteinase with thrombospondin type 1 motif, 4 (Adamts4), a disintegrin-like and metalloproteinase with thrombospondin type 1 motif, 5 (Adamts5) and four genes coding for tissue metalloproteinases inhibitors (Timp1, Timp2, Timp3, Timp4). In addition, three genes coding for tight junction proteins - occludin (Ocln), claudin-5 (Cldn5) and zonula occludens-1 (Zo1) - and six genes coding for basement membrane proteins laminin alpha5 (Lama5), nidogen1 (Nid1), collagen III alpha 1 (Col3a1), collagen IV alpha 1 (Col4a1), collagen VIII alpha 1 (Col8a1), vascular cell adhesion molecule 1 (Vcam1) and lysyl oxidase-like 2 (Loxl2) were also analyzed. Additionally, the expression of three astrocyte markers, glial fibrillary acidic protein (Gfap), aquaporin 4 (Aqp4), and S100 calcium-binding protein A10 (S100a10); six microglial genes, allograft inflammatory factor 1 (Iba1), complement component 1 (C1q), myeloperoxidase (Mpo), Cd86, chemokine (C-C motif) ligand 2 (Ccl2); transforming growth factor beta 1 (Tgfb1); two oligodendrocyte/oligodendrocyte progenitor cell (OPC) markers, proteolipid protein (myelin) 1 (Plp1) and chondroitin sulfate proteoglycan 4 (Cspg4); pericyte marker (Pdgfrb); and three interleukins (Il1b, Il6, and Il33) were also quantified and analyzed relative to the expression of glyceraldehyde 3-phosphate dehydrogenase (Gapdh). The Gapdh gene was used to normalize the expression levels of these genes as a housekeeping gene whose expression was not altered in the compared conditions [40].
Immunohistochemistry
All sections were washed at room temperature (RT) three times with 120 mM phosphate buffer (PB), pH = 7.2, for 10 min each wash, followed by permeabilization with 0.5% Triton X-100 (Sigma-Aldrich Inc. T9284, St. Louis, MO, USA) in PB for 10 min at RT. The sections were subsequently blocked using a blocking buffer (0.4% Triton X-100 + 10% goat serum (Gibco 16210-064, Amarillo, TX, USA/Thermo Fisher Scientific 16210-064, Waltham, MA, USA) + 0.1% glycine in PB) for 45–60 min at RT. Afterward, sections were either incubated overnight at 37 °C or for 2 days at 4 °C depending on which primary reagents were used (Table 2). Following three washes with PB, sections were incubated with secondary antibodies (Table 2) for 2 h at RT. Labeled sections were washed again three times at RT for 10 min in PB and mounted on glass slides using Fluoromount medium (Sigma-Aldrich F4680, St. Louis, MO, USA) [41].
Image acquisition and histochemical measurements
To analyze the tissue distribution of mouse immunoglobulin G (IgG) in Solanum tuberosum lectin-fluorescein isothiocyanate (STL-FITC)-fluorescent small vessels, the expression of ECM markers, astroglial/microglial cell activation, and synaptic puncta, images were acquired using a Zeiss confocal microscope (LSM 700) and EC Plan-Neofluar 20×/0.50 M27, 40×1.3 oil M27 and 63×/1.40 oil M27 objectives by an experimenter blinded to the experimental groups. The acquisition conditions were maintained throughout all the imaging sessions to compare the fluorescence intensity between samples. Images were further processed and analyzed using the open software ImageJ (Fiji) [42] and MATLAB 2019a (MathWorks, Natick, MA).
Detection and quantification of IgG leakage as an indicator of BBB breakdown
BBB integrity was assessed by measuring the parenchymal abundance of mouse IgG. Briefly, sections containing the RSC, dorsal striatum (DS), CA1, BG, thalamus, and medial prefrontal cortex (mPFC) from young animals as well as the RSC and CA1 from aged animals were immunostained with anti-mouse IgG antibody and STL-FITC. Accordingly, we calculated the vascular leaks as a percentage of overlap between IgG+ area and STL+ relative to the total STL+ area. To accomplish this, we measured the positive area (in µm2) represented in each IHC-stained section by using self-developed MATLAB code based on a thresholding technique. Analyses were performed on nine images from each region of each animal.
Vessel-associated microglia/perivascular macrophages
The association of microglia/perivascular macrophages with blood vessels was quantified in fluorescence microscopy images using Fiji by comparing the vessel area associated with microglia and the total area of blood vessels in FOVs located in the CA1 region. To quantify the associated area, we used color thresholding and binary analysis. Manual thresholds were set according to signal intensities. Then, the selected threshold value was applied to all animals, and the resulting data represent the percentage of signal colocalization. Ten random images from each group were analyzed.
Microglia and C1q analyses
It is widely accepted that there are clear morphological changes in microglia after activation, such as enlargement of soma and a reduction of microglial processes [43, 44]. Ionized calcium-binding adaptor molecule 1 (Iba1) antibody recognizes the microglial protein Iba1, which is commonly used as a microglia/macrophage-specific marker. For the evaluation of the arborization area, somatic area and number of positive (Iba1+) cells, nine images per animal were acquired. Square regions of interest (ROIs) (374 × 374 μm2) were selected in the RSC and CA1. Two different thresholding and size exclusion criteria were applied to measure the total area covered by branching (thresholding 1: autothresholding with the mean dark option) and the soma area (thresholding 2: autothresholding with the triangle dark option) of Iba1+ cells. The soma area was subtracted from the total area to calculate the total area covered by Iba1 branching. Thresholding 2 with a size-exclusion parameter enabled us to accurately count the number of Iba1+ cells. Differences in the averaged arborization area and somatic area per cell were calculated and normalized in each animal. The number of Iba1+ cells was counted manually.
For microglial morphology analysis, we also used 3DMorph, a MATLAB-based script that analyzes microglial morphology from 3D data as described elsewhere [45]. Briefly, the program uses graphical user interfaces to initially define an image threshold, noise limits, and cell sizes. The program requires the input of threshold levels, cell size expectations, and preferred methods of skeletonization. After these settings were defined, the program made measurements automatically. The output data included cell volume, territorial volume, branch length, number of endpoints and branch points, and the average distance between cells. For 3DMorph analysis, 3 images per animal were taken in the CA1 region using a Zeiss LSM 700 confocal microscope equipped with an EC Plan-Neofluar, 63×/1.40 oil M27 objective (16-bit, 12 optical sections, 0.2 µm intervals between sections, 1,024 × 1,024 pixels, pixel size of 0.625 µm).
For quantitation of C1q expression in the stratum radiatum of the CA1 region and RSC, images were analyzed using Fiji software. Three ROIs (squares) were defined randomly within soma-free ROIs, and the mean pixel intensity per ROI was determined. The mean ROI intensity per animal was obtained by averaging all ROI mean intensities in the sections studied (three sections per region).
Astroglia analysis
Glial fibrillary acid protein (GFAP) antibody has been extensively used to investigate astrocytic activation [46]. To count GFAP-positive (GFAP+) cells, 3 images per animal were taken in the CA1 region using a Zeiss LSM 700 confocal microscope equipped with an EC Plan-Neofluar, 40×1.3 oil M27 objective (16-bit, 12 optical sections, 1.2 µm intervals between sections, 1,024 × 1,024 pixels, pixel size of 0.625 µm). Autothresholding (moments dark option) and size exclusion criteria (size = 20) were applied in Fiji to measure the area and mean intensity of GFAP+ cells on the maximum intensity z-projection images. The number of GFAP+ cells was counted manually.
Neural ECM analysis
To count the perineuronal net (PNN)-associated and parvalbumin (PV)-immunopositive (PV+) cells, 3 images per animal were taken per RSC and the CA1 region using a Zeiss LSM 700 confocal microscope equipped with an EC Plan-Neofluar, 40×1.3 oil M27 objective (16-bit, 12 optical sections, 1.2 µm intervals between sections, 1,024 × 1,024 pixels, pixel size of 0.625 µm). The numbers of PV+ Acan-immunopositive (Acan+) and PV+ Acan-immunonegative (Acan−) cells were counted manually. The soma area and intensity were measured on the maximum intensity z-projection images using Fiji as described below.
Biotinylated Wisteria floribunda agglutinin (WFA) and anti-Aggrecan antibodies were used to label the ECM of PNNs. WFA labels Gal- and GalNAc-terminated glycoepitopes on core ECM proteins, such as lecticans, whereas Aggrecan antibody labels the Acan core protein [47, 48]. High-resolution images for WFA/vesicular GABA transporter (VGAT) analysis were acquired using a Zeiss LSM 700 confocal microscope and the EC Plan Apochromat 63×/1.40 oil M27 objective. For each animal, 3 images of WFA/VGAT cells per animal were acquired in the RSC and CA1 (16-bit, 14 optical sections, 0.170 μm intervals between sections, 1,024 × 1,024 pixels, pixel size of 0.099 μm).
For soma size and mean intensity quantification, the soma of PV+ Acan+ cells were manually outlined, and a band of 1.5 μm was then crated as the ROI to measure the mean intensity and soma area of PV cells and the mean intensity of Acan signals. In addition, the somatic area of Acan+ cells was measured as described. For further analysis, the somatic area and mean intensity of WFA+ cells were manually outlined and analyzed by creating a band of 1.5 μm around cells as the ROI.
For the quantification of the expression of VGAT+ puncta on WFA+ cells, the soma was manually outlined by creating a band of 1.5 μm as the region of interest (ROI). Then, the mean intensities of at least 5 cells were averaged in each region to obtain the mean ROI intensity per animal. At least 5 cells per section were measured and averaged for each mouse (three sections per region).
Presynaptic and synaptic scaffold protein analysis
Previous findings suggest that a loss of glutamatergic synapses and changes in the expression of pre- and postsynaptic proteins are correlates of cognitive impairment in both CSVD and Alzheimer's disease (AD) [49, 50]. To count synapses, 3 images per animal were taken per RSC and the CA1 region using a Zeiss LSM 700 confocal microscope equipped with an EC Plan-Neofluar, 40×1.3 oil M27 objective (16-bit, 12 optical sections, 1.2 µm intervals between sections, 1,024 × 1,024 pixels, pixel size of 0.625 µm). We analyzed the mean ROI intensity and the numbers of presynaptic vesicular glutamate transporter 1 (VGLUT1)- and vesicular glutamate transporter 2 (VGLUT2)-immunopositive puncta in the neuropil area. The expression levels of VGLUT1 and VGLUT2 in the CA1 stratum radiatum and RSC were quantified using Fiji software. Three ROIs (squares) were defined randomly within soma-free ROIs, and the mean pixel intensity per ROI was determined. To obtain the mean ROI intensity per animal, three ROI mean intensities in each region were averaged.
Furthermore, we analyzed Homer-1-immunopositive puncta because Homer-1 is a postsynaptic scaffolding protein that regulates glutamatergic synapses and spine morphogenesis. The expression levels of Homer-1 in CA1 str. radiatum and RSC were quantified using Fiji. To determine the mean pixel intensity within each ROI, three random ROIs (squares) were defined within soma-free ROIs, and then three ROI mean intensities were averaged in each region to obtain the mean ROI intensity per animal.
Analysis of VGLUT1+ puncta and perisynaptic brevican
To study the synaptic changes in the adult Col18a1-/- mouse RSC and CA1 and correlate these changes to the respective perisynaptic ECM, we developed a semi-automatic open-source Fiji script that we named Analysis of PeriSynaptic Matrix (APSM_v1.1) [51]. The program can be applied to confocal z-stacks and prompts the user to choose the sharply focused plane to be used for quantification. The program detects the clusters of high-intensity pixels as presynaptic boutons in a noisy background in brain sections. It further allows filtering out puncta not representing synaptic boutons based on the intensity, size, and circularity of the structures. In contrast to other thresholding methods available in Fiji, we were able to automatically segment presynaptic boutons using this tool and obtained quantification of individual synaptic puncta and the ECM associated with them. We quantified the expression of these proteins in the CA1 str. radiatum.
Following the manual selection of the sharpest VGLUT1 signal, the program performs a maximum intensity projection of 3 z-planes (the selected central plane + the planes above and below), subtracts the background and applies a Gaussian filter. Next, it implements the Find Maxima algorithm (prominence = 2 (user-defined)), exclusion of maxima on image edges) to locate local maxima that are the brightest VGLUT1+ areas of synaptic boutons and collect the x-y coordinates of these pixels (central pixels).
Furthermore, the program iterates over all the identified central x-y coordinates and finds the pixels nearby to identify surrounding pixels belonging to the same presynaptic bouton according to user-defined criteria of the percentage of intensity of the central pixel within a specified radius. For this study, all neighboring pixels within a 1 µm radius of the central pixel that are at least 75% as bright as the central pixel are considered to be part of the same bouton. Furthermore, it applies the area (>0.05 µm2) threshold to filter out nonwell-defined structures. Finally, it measures the properties of the remaining synaptic puncta and measures the mean intensity signals within perisynaptic bands surrounding the synaptic puncta (width = 0.2 µm in this study) in all channels of the original image. The number of puncta was quantified after filtering particles for size (included from 0.3 to 1 μm2) and circularity (included from 0.5 to 1). Finally, the selection of binarized puncta was used as a mask to measure the fluorescence intensity (FI) of puncta.
Data analysis and statistics
Statistical analysis was performed with Prism GraphPad (v.8.3.0). To compare two groups, the t-test with Welch’s correction was used. To compare multiple groups, ordinary two-way ANOVA with Sidak's multiple comparisons test was used to examine the genotype x age interaction. Two-way RM ANOVA with Sidak's multiple comparisons test was used to assess the genotype x region interaction. All bar graphs represent the mean ± standard error of the mean (SEM), with individual sample values shown as dots. Statistical significance was accepted at p < 0.05.