Neurovascular coupling and oxygenation are decreased in hippocampus compared to neocortex because of microvascular differences

Abstract

The hippocampus is essential for spatial and episodic memory but is damaged early in Alzheimer's disease and is very sensitive to hypoxia. Understanding how it regulates its oxygen supply is therefore key for designing interventions to preserve its function. However, studies of neurovascular function in the hippocampus in vivo have been limited by its relative inaccessibility. Here we compared hippocampal and visual cortical neurovascular function in awake mice, using two photon imaging of individual neurons and vessels and measures of regional blood flow and haemoglobin oxygenation. We show that blood flow, blood oxygenation and neurovascular coupling were decreased in the hippocampus compared to neocortex, because of differences in both the vascular network and pericyte and endothelial cell function. Modelling oxygen diffusion indicates that these features of the hippocampal vasculature may restrict oxygen availability and could explain its sensitivity to damage during neurological conditions, including Alzheimer's disease, where the brain's energy supply is decreased.

Fig. 1. Experimental set-up and example haemodynamic data.

Representative schematic showing the GCaMP6f-positive pyramidal neurons (green) and blood vessels (red) accessible for two-photon imaging after a visual cortical or b hippocampal surgery, with example maximum-projected images across each layer. Scale bars represent 100 µm, and similar z-stack images across layers were taken for nine animals in HC and 11 animals in V1. c Schematic of the imaging set-up. Either the two-photon objective or oxy-CBF probe was used to collect data while the mouse was head-fixed but awake and able to run on the cylinder. d Representative locomotion recorded by the rotary encoder during one imaging session (centimetres per second). Distinct periods of running are indicated by the black bars. A virtual reality maze e or drifting gratings f were presented on the screens in c. Locomotion advanced the mice through the virtual reality maze. The arrows beneath the drifting gratings display show the direction the gratings travelled. g Example, haemodynamic recordings from visual cortex using the oxy-CBF probe during visual stimulation (grey bar represents stimulation, N  =  4 animals, 10 sessions, 202 trials). h The cerebral metabolic rate of oxygen consumption (CMRO₂) is calculated from the haemodynamic parameters collected using the oxy-CBF probe for the data in g (see Methods). All data traces are unsmoothed averages, and error bands represent mean ± SEM.

Fig. 2. Baseline haemodynamics in HC and V1.

Estimated CMRO₂ a, CBF b and oxygen saturation (SO₂) c in HC (purple) and V1 (orange) in stationary mice in the dark (HC: 9 animals across 21 sessions; V1: 9 animals across 19 sessions, 1–5 sessions per mouse; data points are averages for each animal). d FITC-gelatin filled vessels (green) from the HC and neocortex of one NG2-DsRed mouse (pericytes are red). Scale bars represent 100 μm. Example images are projections of the 200 μm Z-stacks compared in e. e Capillary density in these slices was significantly lower in HC (N  =  6 slices, five mice) compared with V1 (N  =  5 slices, four mice). f Single capillaries were imaged in vivo using two-photon microscopy (HC: 55 vessels from 11 mice, V1: 54 vessels from 14 mice), and the average diameter of the vessels scanned did not differ between HC (mean: 4.9 µm) and V1 (mean: 5.2 µm, t test p  =  0.13). g Example line scan trajectory from one in vivo two-photon recording to represent the 99 vessels scanned in f (top; as input into the acquisition software—the actual trajectory will differ from that shown owing to mirror inertia). The scan path of the laser goes along (blue) and across (red) the capillary (labelled with i.v. Texas Red dextran, white. Dark stripes are RBCs). Each row in the corresponding line scan image (bottom) represents a single time point. As RBCs move, their shadow shifts along the vessel, so the angle of the stripes (left, under blue line) shows how fast they are moving. The vessel diameter can be measured from the intensity profile of the Texas Red-labelled lumen (right, under red line). The number of red blood cells per second can be calculated by counting the number of dark stripes (marked with yellow circles) within this time period. The vertical scale bar represents 5 ms, and the horizontal scale bar 5 μm. Despite the same vessel types being scanned in f, the average resting h haematocrit, i RBC flux and j RBC velocity were different between regions (statistical comparisons were made on individual vessels, N specified in f). Data in bar charts represent mean ± SEM, dots are individual vessels or mice, as indicated. P values are from independent sample t tests, or Mann–Whitney U tests (see Statistics Report Table SR1). Source data are provided as a Source Data file.

Fig. 3. Vessel responses to local neuronal calcium events.

a Texas Red dextran-filled vessel (red) and GCaMP6f-positive pyramidal neurons (green) from one recording in V1 before (top) and during (bottom) increases in neuronal calcium (to represent the 87 vessels with local neuronal calcium imaged). Yellow outlines show vessel before calcium event. Arrows indicate the largest dilation. Scale bars represent 5 μm. b Neuronal calcium averaged over all cells in the field-of-view of one imaging session and the corresponding vessel diameter. Red dots mark calcium events. c Vessel responses to preceding calcium events were more frequent in V1 than HC (Chi-square test). d Average calcium response and e diameter change in HC (purple, N = 779 trials, 46 vessels, 6 animals) and V1 (orange, N  =  1238 trials, 41 vessels, 7 animals). f Diameters when vessel traces were shuffled 100 times so no longer aligned to calcium events (HC N = 77900 events, V1 N = 123800). g The calcium events were larger in HC compared with V1. h Vessel dilations were larger in V1 than HC when aligned to calcium events, i but not when shuffled (statistical comparisons were made on individual calcium/dilation/shuffled dilation events, N specified in d/e & f). j Calcium events that led to dilations in HC (N = 120 events, six mice) and V1 (N = 313 events, seven mice). k Corresponding diameter changes. l Calcium peaks leading to dilations were higher in HC (N = 120, V1 N = 313). m Diameter peaks were significantly larger in V1 than HC. n A neurovascular coupling index (NVC_index) was calculated by dividing each dilation peak by its corresponding calcium peak. NVC_index was lower in HC than V1. P values are from Mann–Whitney U tests, unless stated (see Statistics Report Tables SR2a–b). Averages and shaded error bands show mean ± SEM. Horizontal lines on violin plots show median and interquartile range. Source data are provided as a Source Data file.

Fig. 4. Vasodilatory second messenger pathways in HC and cortex.

The expression of a Nos1 and b Ptges3 were higher in the pyramidal cells of HC (941 cells) than cortex (398 cells). c Ptges3 expression was also higher in astrocytes in HC (80 cells) than V1 (143 cells). Horizontal lines on violin plots show median and interquartile range. P values represent the result of independent sample t tests, with Holm–Bonferroni corrections for the multiple comparisons presented in each of Supplementary Tables 1–3.

Fig. 5. Wide-field neuronal activity patterns.

a Representative wide field-of-view recording of calcium signals (white; maximum projection across time) in HC (left, 1 of 10 independent recordings) and V1 (right, 1 of 19 independent recordings). Regions of interest (ROIs) used to measure cell activity are displayed in pink. Scale bar represents 100 μm. b Example neuronal calcium trace, averaged across all detected ROIs. Times when net activity peaks (>2 SD above baseline mean) are shown by red dots. Neither c the correlation (Pearson’s R) between individual cells within a field-of-view was different, nor d the size of net calcium peaks between the HC (10 recordings from four animals, 407 cells detected in total) and V1 (19 recordings from three animals, 338 cells in total). Dots in c and d represent separate recording sessions. e Example CMRO₂ fluctuations over time and corresponding total haemoglobin (Hbt). Detected signal peaks are marked in red. f Averaged CMRO₂ peaks per region, and g the Hbt in response to these CMRO₂ peaks (HC data represent 926 peaks taken from nine animals across 37 recordings, V1 data represent 996 peaks taken from 10 animals across 46 recordings). h The magnitude of these CMRO₂ peaks was larger in V1 than HC, and i Hbt increased more in V1 than in HC within 5 s of an increase in CMRO₂. j The NVCindex was calculated by dividing Hbt/CMRO₂, and was significantly lower in HC (n = 926 and n = 996 data points examined for CMRO₂, Hbt and NVCindex statistical comparisons). P values represent the result of Mann–Whitney U tests for all data except d which was a result of independent sample t tests (see Statistics Report Table SR3). Bar charts and traces show mean ± SEM, violin plots show median and interquartile range. Source data are provided as a Source Data file.

Fig. 6. Vascular morphology across brain regions.

a Example in vivo Z-stacks of vasculature in V1 (left) and HC (CA1, right, as used in b). b Left panel: all data points for vessel diameters by depth from the pia in V1 (48458 vessels collapsed from nine stacks/nine animals, orange) or layer SO in HC (21175 vessels collapsed from 8 stacks/6 animals, purple). Right panel: histogram of average ±SEM of these vessel diameters by depth. The overlaying lines are a smoothed trace of the average diameter per bin. c Example arteriole (left) and venule (right) from an in vivo Z-stack of CA1 in a NG2-DsRed mouse with FITC dextran-filled vessels (representative image from N = 3 stacks from different mice per region). Scale bars represent 100 µm. The termination point of an arteriole was the final smooth muscle cell before an ensheathing pericyte (blue arrow). Venule termination points were the final branch before the capillary bed, identified by the presence of distinct pericytes. Dots represent individual vessels (N = 3 stacks from different mice per region). The diameter and length of d arteriole branches were not different between HC (purple, N = 24) and V1 (orange, N = 25). e The diameter of the vessel at the final SMC was significantly smaller in HC (N = 7) vs V1 (N = 8), despite similar vessel lengths. f The precapillary arterioles (with ensheathing pericytes, HC N = 10, V1 N = 9) and g venule branches (HC N = 16, V1 N = 12) showed no differences in length or diameter between regions. h Intersoma distances (ISD) between neighbouring pericytes (taken from fixed tissue) were longer in HC than V1 (89 vessels in HC and 127 in V1, from six mice). Confocal stacks of FITC-gel-filled vessels of NG2-DsRed mice were taken in HC and V1 from six mice. i Examples of ensheathing (EP), mesh (MP) and thin-strand (TSP) pericytes. Scale bars represent 5 μm. j The distribution of cell types was not different across regions (Cochran–Mantel-Haenszel 3D variant of Chi-square test). Numbers inside bars are numbers of vessels. k There were regional differences in vessel diameter, although post hoc comparisons revealed that these were specific to TSP locations in HC. l Ensheathing pericyte lengths were similar between regions, but both mid-capillary pericyte categories were longer in HC than V1. P values are from one-way ANOVAs with Welch’s correction to test for effects of cell type and brain region. P values above the bars are from unpaired t tests, Mann–Whitney U tests or post hoc Mann–Whitney U tests with Holm–Bonferroni correction for multiple comparisons (see Statistics Report Tables SR4a–e). Bar charts are mean ± SEM, and statistical comparisons for d–h & k–l compared data from the individual vessels represented by the dots. Horizontal lines on violin plots show median and interquartile range. Source data are provided as a Source Data file.

Fig. 7. Modelling O₂ diffusion in brain tissue.

a Schematic of radial diffusion model. b Top: example in vivo Z-stack (250 μm³) of fluorescent dextran-filled vasculature in CA1. Bottom: 3D distance map generated from the z-stack. White pixels denote high values and black low values. Red cube shows example 100 μm³ substack used to generate c. c Average histogram of the distance of each pixel from a capillary (four mice per region, the value for each mouse representing the average of five non-overlapping substacks sampled from one Z-stack). Dotted lines mark the 50th and 95th percentiles for HC (purple) and V1 (orange), data are presented as mean ± SEM between animals. d Example heatmap showing oxygen diffusion from a capillary in HC separated from the next capillary by twice the median tissue distance from a capillary (14.4 μm). Simulated time courses from initial conditions of zero [O₂] for HC (purple) and V1 (orange) for tissue at e the median or h 95th centile distance from a capillary, showing that steady-state is reached by 6 s (when [O₂] profiles were extracted). Lines represent different values of V_max as labelled. [O₂] profiles across tissue for capillary separations of twice the tissue f median and i 95th centile distance from a capillary. O₂ consumption rate as a fraction of V_max (VO₂) for g median and j 95th centile capillary spacing conditions, calculated from oxygen profiles shown in e & h. k [O₂] and l VO₂ reached midway between capillaries for different values of V_max in HC (purple) and V1 (orange) at median (solid symbols) or 95th centile (hollow symbols) capillary spacings. m [O₂] (solid symbols) and VO₂ (hollow symbols) reached midway between capillaries for capillaries spaced at twice the 50th, 60th, 70th, 80th, 90th and 95th centiles (from left to right for each range of solid or hollow dots) of tissue distance from a vessel in HC (purple) and V1 (orange). Source data are provided as a Source Data file.