Pericyte remodeling is deficient in the aged brain and contributes to impaired capillary flow and structure

No difference in capillary landscape in upper cortex of adult and aged mice

We compared capillary structure and pericyte abundance at baseline between adult and aged PDGFRβ-tdTomato mice using in vivo two-photon imaging (Fig. 1a,b, Supplementary Fig. 1a,b). Our analyses focused on capillary networks within the first 10–100 μm from the cortical pial surface, as subsequent pericyte ablation studies were limited to these depths. Capillary networks of aged mice (18–24 months) did not differ in total length, number of junctions, or abundance of pericyte somata when compared to adult mice (3–6 months)(Supplementary Fig. 1c, d, e). Further, the limited capillary regions that lacked pericyte coverage did not differ between ages (Supplementary Fig. 1f). Thus, capillary structure and pericyte coverage in upper layers of cortex are comparable at baseline between adult and aged mice.

Fig. 1: Acute consequence of ablating multiple contiguous pericytes.
figure 1

a In vivo two-photon imaging through a chronic cranial window in anesthetized PDGFRβ-tdTomato mice. b Capillary bed with vessels labeled green with intravenous (i.v.) dye, FITC-dextran (70 kDa), and pericytes genetically labeled in red. The inset shows the structure of capillary pericytes. Representative of 12 PDGFRβ-tdTomato mice in this study. c A capillary pericyte targeted for two-photon ablation encircled in yellow. Ablative line-scanning is restricted to the pericyte soma. Post-ablation image shows loss of fluorescence from targeted pericyte, including its processes, within minutes post-ablation. Neighboring pericytes are unaffected. Representative of >42 pericyte ablations. d Schematic showing metrics collected following ablation of three contiguous pericytes. Left, targeted pericytes are marked with asterisks. Right, metrics include total capillary length uncovered, number of remodeling pericytes, and number of remodeling pericyte processes. e,f Before and 5 min after triple pericyte ablation in adult and aged animal. Arrowheads indicate targeted pericyte somata, and dotted magenta line outlines approximate regions of uncovered endothelium. Inset shows a magnified region with pericyte ablation and uncovered territory 5 min afterwards. Each example is representative of 7 experiments per age group. Comparison of vascular metrics between adult and aged mice after triple pericyte ablation. g Total capillary length uncovered; t(12) = 1.790, p = 0.0988. h Number of fully and partially uncovered capillary segments; t(12) = 0.1936, p = 0.8497. i Range of capillary segment lengths; t(195) = 0.2325, p = 0.8164. All are unpaired t tests (two-sided), for n = 7 triple pericyte ablations (6 adult mice), 7 triple pericyte ablations (6 aged mice) for g and h and = 99 capillary segments over 6 adult mice, n = 98 capillary segments over 6 aged mice for i. Data are shown as mean ± SD. Ns = non-significant. Comparison of pericyte metrics between adult and aged mice after triple pericyte ablation. j Number of remodeling pericytes; t(12) = 0.6030, p=0.5577. k Number of remodeling pericyte processes; t(12) = 0.7206, p = 0.4850. Unpaired t tests for n = 7 triple pericyte ablations (6 adult mice), 7 triple pericyte ablations (6 aged mice). Data are shown as mean ± SD.

Acute consequence of triple pericyte ablation

In vivo optical ablation of capillary pericytes allowed us to observe the response to focal pericyte loss. We ablated three contiguous pericytes from capillary regions in upper layers of cortex in adult and aged mice (Fig. 1c-f). Across both age groups, triple pericyte ablations uncovered capillary lengths of 643.75 ± 26.43 μm (mean ± SEM; 475–809 μm range)(Fig. 1d,g), which encompassed ~15 total capillary segments, i.e., sections between branch-points (Fig. 1h). The affected individual capillary segments varied greatly in length (4.79–236.6 μm)(Fig. 1i). Triple pericyte ablation resulted in loss of pericyte–pericyte contact at 7–10 process termini from an average of eight neighboring pericytes, each with potential to grow into the uncovered territory (Fig. 1d,j,k). We detected no difference in these basal vascular or pericyte metrics between adult and aged mice.

Pericyte remodeling is efficient in the adult mouse brain, but impaired with aging

To measure pericyte remodeling, we re-examined capillary regions with pericyte loss at 3, 7, 14, and 21 days post-ablation (Fig. 2). In adult mice, neighboring pericytes remodeled their processes to achieve contact with nearly the entire uncovered region within 21 days (Fig. 2a,b). In a representative example, endothelial contact was regained through simultaneous growth of ten neighboring pericytes, each extending a single process toward the uncovered region (Fig. 2c). Remodeling led to eventual contact with another pericyte process terminus, i.e., pericyte-pericyte contact, upon which no further growth was possible (Fig. 2d)10. In the aged group, a similar number of remodeling processes contributed to endothelial re-coverage, but a greater amount of capillary length remained uncovered at 21 days (Fig. 2e–h).

Fig. 2: Pericyte remodeling ensures capillary coverage following pericyte loss.
figure 2

a Representative example of pericyte remodeling from an adult mouse (4 months old at pre-ablation), showing the area before triple pericyte ablation, and at select time-points after ablation. Arrowheads indicate pericytes targeted for ablation. The lower panels show mural cells only. Insets show magnified view of region with pericyte ablation followed with remodeling of neighboring pericyte processes. Each example is selected from 7 experiments in adult mice. I.v. dye = intravenous dye. b Schematic of capillary coverage by pericytes on each imaging day. c Schematic representation of growth contributed by different pericyte processes over 21 days. Each color represents an individual pericyte process, with location of original terminus of the process as a dashed line. d Plot of added pericyte process length over time from Day 0. Each color corresponds to an individual pericyte process from c. Closed circles mark the day of pericyte-pericyte contact, after which growth stops (dashed lines). PC-PC contact = pericyte–pericyte contact. e–h Representative example from an aged mouse (21 months old at pre-ablation), with same layout as for adult mouse. Each example is selected from 7 experiments in aged mice.

In the adult brain, 84% of processes had made pericyte-pericyte contact within the 21-day timeframe (Fig. 3a,b). In the aged brain, less than 40% of processes had made contact. To assess the dynamics of pericyte process growth, we measured process length and growth rate for each interval of imaging, i.e., day 3 to 7 (Fig. 3c,d). “Maximum growth rate” was the time interval with the fastest growth, whereas “average growth rate” was the mean rate across all time intervals before pericyte-pericyte contact was achieved. We observed a significant decrease in both maximum and average growth rates in aged mice (Fig. 3e,f). After 21 days, however, the final length of extension achieved and total process length (from soma to process terminus) was not statistically different between age groups (Fig. 3g,h). An age-related effect in these metrics may be masked by the high variance of remodeling between pericytes, and because pericytes were studied broadly across the arterio-venous axis, as will be discussed below.

Fig. 3: Inefficient pericyte remodeling in the aged brain fails to regain complete capillary coverage.
figure 3

a In vivo two-photon images of a pericyte process growing over time (white arrow), until it is inhibited by pericyte–pericyte contact. This is a representative example of pericyte-pericyte contact from 21 pericyte ablations. I.v. dye  =  intravenous dye. b Percent of total pericyte processes making pericyte-pericyte contact on each imaging day across age groups. n = 60 processes from 6 adult mice, n = 57 processes from 6 aged mice. c Pericyte process extension over time in each age group, including all processes examined. n = 60 processes from 6 adult mice, n = 57 processes from 6 aged mice. d Plot showing the extension of pericyte processes, excluding those that have made pericyte-pericyte contact, at each post-ablation imaging day. Mixed-effects analysis with Sidak’s multiple comparisons (two-sided), F(1,115) = 9.979; overall effect **p = 0.002, Day 3, *p = 0.0187; Day 7, **p = 0.0056; Day 14, **p = 0.0087; Day 21, p = 0.6574. N = 60 processes from 6 adult mice, and n = 58 processes from 6 aged mice. Data are shown as mean ± SEM. e, f Maximum and average rate of pericyte process growth in aged compared to adult mice. e Unpaired t tests (two-sided) with Welch’s correction for unequal variances: t(109.1) = 3.313; **p = 0.0013; f t(90.50) = 4.977; ****p < 0.0001. N = 60 processes from 6 adult mice, n = 57 processes from 6 aged mice. Data are shown as mean ± SD. g Maximum process extension achieved in 21 days. Unpaired t test (two-sided), t(115) = 0.9370; p = 0.3507. n = 60 processes from 6 adult mice, 57 processes from 6 aged mice. Data are shown as mean ± SD. h Total process length at day 21. t(112) = 1.066; p = 0.2889. Data in eh shown as mean ± SD. Statistics performed with unpaired t tests (two-sided) for n = 60 processes from 6 adult mice, n = 57 processes from 6 aged mice. Data shown as mean ± SD. i Uncovered vessel length remaining over time. Each line represents one triple pericyte ablation region. Adult, n = 7 triple pericyte ablations; 6 mice; Aged n = 7 triple pericyte ablations from 6 mice. j Comparison of average vessel length remaining on each imaging day across age groups. Two-way ANOVA with Sidak’s multiple comparisons test, F(1, 12) = 16.72; overall effect **p = 0.0015, Day 0, p = 0.4196. Day 3, *p = 0.0118. Day 7, **p = 0.0098. Day 14, **p = 0.0052. Day 21, *p = 0.0490. Adult, n = 7 regions from 6 mice; aged, n = 7 regions from 6 mice. Day 49 is from a subset of triple pericyte ablation experiments in aged animals only (5 regions from 4 mice), not included in statistical analyses. Data are shown as mean ± SEM.

To better understand the outcome of synergistic remodeling, we calculated the length of endothelium uncovered over time. In adult mice, uncovered capillary length declined steeply in the first 7 days following ablation, and then slowly approached zero over the next 14 days (Fig. 3i, left). The aged group displayed a slower decline of uncovered territory, with a portion of the endothelium still left uncovered after 21 days (Fig. 3i, right). In a subset of experiments in aged mice, ablation areas were re-visited 49 days post-ablation and sizeable stretches of capillary remained uncovered (Fig. 3j, Supplementary Fig. 2). Together, these data show that the interconnected structure of the capillary network facilitates restoration of endothelial coverage, but the capacity for pericyte remodeling is reduced with aging.

Since tdTomato expression was driven under constitutive pdgfrb promotor activity, new pericytes arising from proliferation or migration would be fluorescently labeled. However, we did not observe the emergence of new pericyte somata in either age group, suggesting that remodeling of existing pericytes is the main reparative strategy at this scale of pericyte loss.

Contributors to the heterogeneity of pericyte process growth

We explored what parameters might contribute to variability in pericyte process growth. Many remodeling processes met one or more capillary bifurcations as they entered the uncovered territory, when the processes would split and extend in both directions, unless the terminus of another pericyte impeded its growth (Supplementary Fig. 3a). A single process could split up to three times, resulting in four growing termini. Given this parallel growth, processes that split had significantly greater final extension lengths than those that did not (Supplementary Fig. 3b). In adult mice, 49% of processes split, as opposed to 35% in aged mice, a difference attributed to their slower growth rather than differences in the number of capillary bifurcations available (Supplementary Fig. 1d). There was no difference in the number of bifurcations within uncovered regions between age groups (7.3 ± 2.7 and 7.9 ± 2.9 bifurcations uncovered in adult and aged mice, respectively; p  =  0.4035, t-test; mean ± SEM).

We further examined if the maximum process length achieved was limited by baseline process length. Baseline lengths did not differ between age groups, consistent with normal pericyte density at baseline (Supplementary Fig. 3c). We, therefore, pooled age groups and plotted baseline process length as a function of length extended post-ablation. This revealed a negative correlation suggesting limited growth potential in individual processes longer than ~150 μm (Supplementary Fig. 3d). It also showed that short basal processes that split during growth had the greatest final added length.

In both age groups, ~25% growing pericyte processes created small vessel distortions (“microbends”) when they tethered to a point further along the capillary and pulled the anchored point closer (Supplementary Fig. 4). Microbends were transient and capillary shape returned to normal once the process terminus had extended further. Processes that created microbends grew at a faster rate and gained more contact distance (Supplementary Fig. 4e,f). Thus, both intrinsic pericyte properties (baseline process length, ability to distort vessel) and extrinsic factors (capillary branching) can influence pericyte growth potential.

Mural cells across microvascular zones can contribute to capillary re-coverage

The cortical microvasculature is divided into arteriole-to-capillary transition, capillary, and venular zones, across which mural cells vary in their transcriptional profile, morphology, and contractile dynamics20,21. Although pericytes were exclusively ablated in the capillary zone, they were often adjacent to arteriole–capillary transition and venular zones, allowing us to assess remodeling of mural cells from these regions into the capillary zone (Fig. 4a).

Fig. 4: Mural cell remodeling capacity varies with age across microvascular zones.
figure 4

a In vivo two-photon image of a microvascular network from penetrating arteriole to ascending venule, with branch orders in white. Bottom image highlights mural cell subtypes within zones. This is a representative view from 14 separate ablation experiments in regions containing different microvascular zones. I.v. dye  =  intravenous dye. Venule SMC  =  venular smooth muscle cell. b Example of an ensheathing pericyte from an adult mouse remodeling into the capillary zone. Arrow indicates position of leading process terminus over time. Bottom row shows high-resolution image of the cell at 21 days post-ablation. This is a representative example from 13 processes observed in arteriole-capillary transition zone. c Example of a mesh and thin-strand pericyte from an adult mouse growing within the capillary bed. Arrows show growing terminal ends of processes. This is a representative example from 35 processes observed in capillary zone. d Example of a venule SMC process from an adult mouse growing into the capillary bed, as indicated by arrows. This is a representative example from 7 processes observed in venular zone. e Average process growth by mural cells in different microvascular zones. Two-way ANOVA with Tukey’s multiple comparisons test (two-sided), For age comparison, F(1,102) = 21.45; overall effect, ****p = 0.0001; Arteriole-capillary transition (adult vs. aged) ****p < 0.0001; capillary (adult vs. aged) **p = 0.0016; Venule (adult vs. aged) p > 0.9999. For vessel type comparison, F(2,102) = 0.1838; overall effect, p = 0.8324, For interaction between age and vessel type, F(2,102)  =  6.029, overall effect, **p = 0.0033. Arteriole-capillary transition vs. capillary (adult), p = 0.5833; arteriole-capillary transition vs. venule (adult), p = 0.4125; capillary vs. venule (adult), p = 0.9533. Arteriole-capillary transition vs. capillary (aged), p = 0.4855; arteriole-capillary transition vs. venule (aged), *p < 0.0398; capillary vs. venule (aged), p = 0.4516. Adult: n = 13 arteriole-capillary transition, n = 35 capillary, n = 7 venule from 6 mice; aged: n = 12 arteriole-capillary transition, n = 30 capillary, n = 10 venule from 6 mice. Data are shown as mean ± SEM. f Maximum process extension by mural cells in different microvascular zones. Two-way ANOVA. For age comparison, F(1,97) = 0.1455; overall effect, p = 0.7037. For vessel type comparison, F(2,97) = 2.613; overall effect, p = 0.0784. For interaction between age and vessel type, F(2,97) = 5.532; **p = 0.0053. Arteriole-capillary transition vs. capillary (adult), p > 0.9999; arteriole-capillary transition vs. venule (adult), p < 0.9897; capillary vs. venule (adult), p = 0.3511. Arteriole-capillary transition vs. capillary (aged), p = 0.8750; arteriole-capillary transition vs. venule (aged), **p < 0.0014; capillary vs. venule (aged), **p = 0.0044. Adult: n = 13 arteriole-capillary transition, n = 35 capillary, n = 7 venule from 6 mice; aged: n = 12 arteriole-capillary transition, n = 30 capillary, n = 10 venule from 6 mice. Data are shown as mean ± SEM.

Mural cells from all zones were able to extend processes into the uncovered capillary territory. Interestingly, regardless of the morphological characteristics of their existing processes, mural cells extended into the capillary zone with a “thin-strand” morphology typical of capillary pericytes (Fig. 4b–d, Supplementary Fig. 5a–d). Striking examples are ensheathing pericytes and venular smooth muscle cells (venule SMCs), whose pre-existing processes were more complex (Fig. 4b,d) This suggests that pericyte process morphology is dictated by the endothelial zone it contacts. In further support of this concept, we also ablated ensheathing pericytes and found that remodeling neighbors retained their circumferentially-oriented processes when extending within the arteriole-capillary transition zone (Supplementary Fig. 5e,f).

We found no difference in rate of process growth across microvascular zones in adult mice (Fig. 4e). The aged group exhibited selective impairment in remodeling of ensheathing and capillary pericytes (mesh and thin-strand morphology) of the arteriole-capillary transition and capillary zones, respectively. In contrast, venule SMCs on ascending venules maintained their remodeling capacity (Fig. 4e, Supplementary Fig. 5c,d). Comparing maximal process extension between adult and aged mice for each vascular zone yielded a similar outcome (Fig. 4f). This points to venule SMCs as a retained source of repair capacity within the aged brain and suggests that arteriole-capillary transition and capillary zones may be more susceptible to prolonged loss of pericyte coverage.

Overt BBB leakage is very rare following focal pericyte ablation

Pericytes are well-established as custodians of BBB integrity5. In recent studies, BBB permeability occurring from adult-induced pericyte loss could be detected by extravasation of a 70 kDa dextran dye22 and even larger IgG proteins23. In both adult and aged mice, we detected no extravasation of 70 kDa dye after focal loss of pericyte coverage loss (Supplementary Fig. 6a,b). However, in a rare occurrence in an aged mouse (1 out of 7 triple ablation experiments; 1 of 98 uncovered capillary segments inspected) leakage was observed on days 3 and 7 post-ablation (Supplementary Fig. 6c). We also assessed a low molecular weight dextran dye in aged mice (10 kDa FITC-dextran), which stays within the blood plasma for several minutes following a single i.v. bolus. No increase in dye extravasation was seen after pericyte loss (3 days post-ablation), compared to regions receiving off-target sham irradiations where the ablative laser path was placed away from a pericyte soma, but of similar distance to the vessel wall (Fig. 5a–d).

Fig. 5: Lack of overt BBB disruption or perivascular inflammation with focal pericyte loss.
figure 5

a Imaging time course to examine for extravasation of intravenous (i.v.) 10 kDa FITC-dextran dye in aged mice. Pre-ablation image shows location of targeted pericytes and sham irradiation controls. At day 3, T=0 shows regions of interest (ROI) from which fluorescence intensity measurements were collected over time. This is a representative example from 8 ablation experiments. b Intensity of FITC-dextran fluorescence within the capillary lumen as a function of time post-injection. Data shown as mean ± SEM. Fluor. Intensity = fluorescence intensity. c Intensity of FITC-dextran fluorescence in parenchymal ROI between ablation and sham regions. Data are shown as mean ± SEM. d Rate of fluorescence intensity change between ablation and sham ROI regions. Wilcoxon rank sum test (two-sided), p = 0.904. N = 7 ablation regions and n = 7 sham regions from 4 aged mice. Data shown as mean ± SEM. In vivo two-photon images from adult Pdgfrβ-tdtomato;Cx3Cr1-GFP mice, showing microglia reaction to e pericyte ablation or f sham irradiation after 30 min and 3 days. ROIs are drawn in the region directly exposed to laser ablation or irradiation (focal point), or the capillary segments covered by pericyte processes (territory). This is a representative example from 3 ablation experiments in adult mice. GFP intensity in focal point (g) and territory (h) after pericyte ablation and sham in adult mice, shown in arbitrary units (a.u.). Two-way repeated measures ANOVA with Sidak’s multiple comparisons (two-sided). Focal point: F(1.45, 7.252) = 24.41, overall effect ***p = 0.0009. Pericyte ablation vs. sham irradiation *p = 0.0216 at 30 min; p = 0.5504 at 3 days. Territory: F(1.149, 5.744) = 3.170, p = 0.1263. GFP intensity in focal point (i) and territory (j) in aged mice. Two-way repeated measures ANOVA. Focal point: F(1.03, 7.183) = 3.214, overall effect p = 0.1146. Territory: F(1.761, 12.33) = 0.2922, p = 0.7248. For g and h, n = 3 pericyte ablations, 4 sham irradiations from 5 mice between 3 and 6 months of age. For i and j, n = 5 pericyte ablations, 4 sham irradiations from 3 mice between 18 and 24 months of age. Data presented as mean ± SEM.

We considered the possibility that BBB leakage was too slow or subtle to be detected by imaging of dye extravasation. Since microglia are highly sensitive to vascular pathology24,25, cluster around leaky vessels26, and react to BBB leakage in models with more extensive pericyte loss22, we reasoned that they would be sensitive indicators of any BBB leakage. In adult and aged double transgenic mice with co-labeled mural cells and microglia, microglia reacted within minutes to engulf the soma of the ablated pericyte (Fig. 5e,g,i). Critically, this reaction was wholly focused on the soma of the dying pericyte, and not the broader capillary regions previously contacted by their processes. Following off-target sham irradiations, microglial processes occasionally extended to investigate the area in the 30 min following laser exposure, but this reaction was comparatively mild (Fig. 5f,h,j). Three days post-ablation or off-target irradiation, we observed no aggregation of microglia or their processes along peri-vascular regions suggesting no delayed BBB leakage.

We further examined endothelial cell junctions after focal pericyte ablation. Using Claudin5-eGFP fusion protein mice, prior studies showed the development of protrusions and small gaps within tight junction strands during stroke-induced BBB leakage, indicative of endothelial remodeling and BBB damage, respectively27. After pericyte ablation in adult mice with co-labeled pericytes and tight junctions, we observed a transient but non-significant increase in small protrusions, and no increase in tight junction gaps (Supplementary Fig. 7). We also used adeno-associated virus to label astrocytes in aged mice. No overt changes in astrocyte endfoot apposition to the capillary wall following focal pericyte ablation were seen (Supplementary Fig. 8). Thus, the degree of pericyte loss in our paradigm does not reach the threshold needed to cause overt BBB disruption, except with one rare observation in an aged mouse.

Aberrant capillary dilation with loss of pericyte coverage is exacerbated with aging

In prior studies, we reported that loss of pericyte coverage led to abnormal capillary dilation in adult mice, indicating a role in regulation of basal capillary tone6,10. We compared this effect between adult and aged mice with triple pericyte ablations. At baseline, we detected no difference in the diameter of capillaries across age groups (Fig. 6a). For each ablation experiment, we measured the diameters of 4–5 capillaries at the following time-points: i) Pre-ablation, when the capillary was covered, ii) 3 days post-ablation, when the endothelium was uncovered, and iii) 14–21 days post-ablation, when the endothelium was re-covered by process growth from neighboring pericytes (Fig. 6b). In adult mice, we observed average dilations of 0.6 μm (21% increase from baseline) when the capillary was uncovered (Fig. 6c,d,g). Dilation occurred exclusively where pericyte contact was lacking (Supplementary Fig. 9). With coverage regained by pericyte remodeling, capillary diameters returned to levels no different from baseline (Fig. 6h).

Fig. 6: Pericyte coverage loss leads to augmented capillary dilations in the aged brain.
figure 6

a Comparison of baseline capillary diameters across age. Unpaired t test (two-sided), t(60) = 0.7245; p = 0.4716. N = 30 capillaries from 6 adult mice, n = 32 capillaries from 6 aged mice. Data shown as mean ± SD. Ns  =  non-significant. b Capillary dilation occurs in region with loss of pericyte coverage, but not in adjacent capillaries that maintain pericyte coverage. Arrowhead points to pericyte targeted for ablation. I.v. dye  =  intravenous dye. c Example capillary region from adult mouse following triple pericyte ablation. Arrowheads show two targeted cells. Bottom row shows i.v. dye alone, with capillary segments measured over time identified by blue lines. d Plot of capillary diameters over time from adult mice. Repeated measures one-way ANOVA with Tukey’s multiple comparisons test (two-sided), F(1.824, 45.60) =  35.95, ****p < 0.0001. Baseline vs uncovered, ****p < 0.0001; baseline vs. re-covered, p = 0.7393; uncovered vs. re-covered, ****p < 0.0001. N  =  26 capillaries from 6 mice. Data are shown as mean ± SD. e Example capillary region from aged mouse following triple pericyte ablation. Arrowheads point to targeted pericytes. Bottom row shows i.v. dye alone, with capillary segments measured over time identified by purple lines. f Plot of capillary diameters over time from aged mice. Repeated measures one-way ANOVA with Tukey’s multiple comparisons test (two-sided), F(1.914, 53.59) =  111.3, ****p < 0.000. Baseline vs uncovered, ****p < 0.0001; baseline vs. re-covered, **p = 0.0018; uncovered vs. re-covered, ****p < 0.0001. N = 29 capillaries from 6 mice. Data shown as mean ± SD. g Change in diameter from baseline at 3 days post ablation, when capillaries lack pericyte coverage. Unpaired t test (two-sided), t(54) = 3.501, ***p = 0.0009. N = 26 capillaries in 6 adult mice, n = 30 capillaries in 6 aged mice. Data are shown as mean ± SD. h Difference between baseline and re-covered capillary diameter detected in adult and aged groups. Adult: One sample t test (two-sided), t(25) = 0.7451; p = 0.4631, for n = 26 from 6 mice. Aged: One sample t test (two-sided), t(33) = 4.172; ***p = 0.0002, for n = 34 capillaries from 6 mice. Difference in re-covered diameter between adult and aged mice. Unpaired t test with Welch’s correction (two-sided), t(41.40) = 2.751; **p = 0.0088. Data shown as mean ± SD. i Dilations persist for capillaries that do no regain pericyte coverage at 21 days in aged mice. Repeated measures one-way ANOVA with Tukey’s multiple comparisons test (two-sided), F(1.967, 19.67) = 31.35; overall effect ****p < 0.0001. Pre-ablation vs 3 days, ****p < 0.0001; pre-ablation vs 21 days, ***p = 0.0002; 3 days vs 21 days, p = 0.5976. N = 11 capillaries from 6 mice. Data are shown as mean ± SD.

In aged mice, uncovered capillary segments dilated to a significantly greater extent than in adult mice, i.e., an average of 1.1 μm (36% increase from baseline), with some individual capillaries dilating as much as ~1.6  μm (Fig. 6e–g). Further, slight dilations persisted even after pericyte coverage had returned (Fig. 6h), suggesting inefficient re-establishment of vascular tone. Capillaries that continued to lack coverage beyond 21 days post-ablation maintained their dilated states (Fig. 6i). We confirmed that the dilations were not due to the laser damage alone, as off-target sham irradiations led to no change in capillary diameter (Supplementary Fig. 10). Further, these results were not an effect of isoflurane anesthesia, as dilations persisted when mice were imaged in the awake state (Supplementary Fig. 11). Altogether, this indicates that pericyte contact is key to maintaining basal capillary tone in vivo, and dilation with pericyte loss is exacerbated in the aged brain.

Local dilations alter flow distribution and increase flow heterogeneity in capillary networks

The flow and oxygenation of red blood cells (RBCs) in brain capillaries is heterogeneous at rest6,28. Homogenization of blood flow among capillaries is necessary to efficiently extract and distribute oxygen (O2) to the tissue29,30,31,32. Therefore, increased flow heterogeneity is a barrier to achieving flux homogenization and O2 extraction. To examine whether focal capillary dilations increased flow heterogeneity, we performed triple pericyte ablations (Fig. 7a) and used line scans to measure blood cell flux (cells/s) and flow directionality in capillaries both within and immediately surrounding the uncovered region (Fig. 7b,c). Flux was measured in the same capillaries before and 3 days post-ablation, when pericyte remodeling had not yet restored coverage and tone.

Fig. 7: Blood flow changes in regions of pericyte ablation and surrounding capillaries.
figure 7

a Image showing capillaries uncovered by a triple ablation (purple). These data are from a single triple pericyte ablation experiment, performed in an aged mouse. I.v. dye = intravenous dye. b Three days post-ablation, changes in red blood cell (RBC) flux from baseline are color-coded in a subset of vessels. c Overlay of blood flow direction on vascular image. Ascending venules are marked in blue, and penetrating arterioles in red. d Blood cell flux before and 3 days post ablation for segments in and immediately around the ablation area. Paired t test, t(24)=3.880; ***p = 0.0007 for n = 25 capillaries from an aged mouse. Significantly higher variance was detected at 3 days compared to baseline, F test to compare variances (two-sided), F(24, 24) = 3.545; ***p = 0.0029. Data shown as mean ± SD. e Proportion of vessels in the area that experienced an increase vs. decrease in flux on Day 3. f Blood cell flux in capillaries that increased in flux from baseline to 3 days post-ablation. Paired t test (two-sided), t(17) = 6.273; ****p < 0.0001, for n = 18 capillaries from an aged mouse. g A subset of capillaries decreased in flux following pericyte ablation. Paired t test (two-sided), t(6) = 3.229; *p = 0.0179, n = 6 capillaries from an aged mouse. Data in d, f, g are shown as mean ± SD. h–j A divergent bifurcation that loses pericyte coverage in one downstream branch, while the alternate route remains covered. RBCs passing through i.v. dye captured by line scans are shown for each vessel (i, ii, iii) over time. k–m A convergent bifurcation that loses pericyte coverage to one upstream vessel. RBCs passing through i.v. dye captured by line scans are shown for each vessel (i, ii, iii) over time.

In an aged mouse, the already high variance of basal flux among capillaries, 18 to 231 cells/s, further broadened to 37 to 453 cells/s post-ablation (Fig. 7d). The increased flux was predominantly observed in dilated capillary segments lacking pericyte coverage (Fig. 7b,e,f). However, a fraction of sampled capillaries surrounding these segments decreased in flux (Fig. 7b,e,g). When pericyte contact was lost in only one of the daughter branches of a divergent bifurcation, blood cells flowed preferentially through the dilated branch. This left the alternate branch under-perfused relative to baseline, even when overall flow entering the bifurcation was higher post-ablation (Fig. 7h-j). This non-uniform partitioning of blood cell flow at capillary bifurcations reflects the Zweifach-Fung effect33, where increased flow rate in a dilated capillary pulls a larger fraction of RBCs into the dilated daughter branch. These local upstream changes can have propagating effects within the capillary network34, as we also observed decreased flow at downstream, convergent bifurcations (Fig. 7k–m).

Focal capillary dilation increases flow heterogeneity in silico

To understand capillary flow changes on a broader network level, we performed blood flow simulations in realistic microvascular networks derived from mouse parietal cortex (Fig. 8a)35. Four cases were studied in two microvascular networks, each involving dilation of 13–14 contiguous capillary segments with similar characteristics to those targeted during triple pericyte ablations in vivo (Fig. 8b)(Supplementary Table 1). The capillaries were each dilated by 0.6 μm or 1.1 μm from their baseline diameters to mimic average dilations measured in the adult and aging brain, respectively (Fig. 6g). An additional group with 1.6 μm dilation was included to examine the effect of the most extreme dilations observed in aged mice. Consistent with observations in vivo, focal dilations produced both increases and decreases in capillary flow (defined as >10% change from baseline)(Fig. 8c). Flow perturbations were detected in hundreds of capillaries surrounding the dilated region. The number of affected capillaries increased commensurately with the extent of dilation, with ~1/3 of the affected capillaries decreasing in flow irrespective of the extent of dilation. Capillaries with increased flow localized generally to the epicenter of dilation, while those with decreased flow resided in surrounding areas (Fig. 8d).

Fig. 8: Impact of local capillary dilations examined in silico.
figure 8

a Microvascular network 1 (MVN1) from mouse parietal cortex. b Example of a typical region affected by pericyte ablation in MVN1. The color bar shows the flow rates in individual capillaries pre-ablation. c Relative flow changes >10% of baseline in the entire MVN1 for an increasing extent of dilation. Vessels with a relative flow change between −10% and +10% are not colored and depicted by the light gray lines. d Relative flow change >10% of baseline at different distances to the center of the dilated capillaries. n(0.6)  =  808 vessels, n(1.1)  =  1114 vessels, n(1.6) = 1367 vessels. e Relative flow changes in the dilated vessels and their direct neighbors (Gen1-Neighbors and Gen2-Neighbors, see schematic). n(0.6, 1.1, 1.6) = 267 vessels each. f Flow steal at divergent bifurcations with one dilated outflow capillary. n(0.6, 1.1, 1.6) = 24 vessels each. Plots df show the relative flow change of individual vessels. The data of all four cases have been combined. For box plots, center = median, box bounds = upper (Q3) and lower (Q1) quartiles, whiskers = last data point within Q1 − 1.5*(Q3 − Q1) and Q3 + 1.5*(Q3 − Q1). g Distribution of standard deviations for numerous undilated capillary sets in upper cortex. The expected baseline heterogeneity of capillaries is defined as the median of all standard deviations. Left: MVN1 (n(sets)= 170, average set size = 14.5 ± 1.6), Right: MVN2 (n(sets)= 346, average set size = 14.6 ± 1.8). h Standard deviation of flow in the dilated capillaries at baseline and with increasing extent of dilation. Dashed lines show expected baseline heterogeneities at 0.50 and 0.75 quantile (see g).

To understand how capillary architecture related to blood flow change, we categorized capillaries into three groups: Dilated, Gen1-Neighbors (1 branch from dilated), and Gen2-Neighbors (2 branches from a dilated segment). Consistent with in vivo observations, dilated capillaries generally increased in flow, while Gen1 and Gen2 neighbors showed both increases and decreases from baseline (Fig. 8e). Further, capillary steal was observed at divergent capillary bifurcations, where one daughter branch was selectively dilated and the undilated branch experienced decreased flow (Fig. 8f). As predicted by the Zweifach-Fung effect33, the steal effect is even more pronounced with RBC flux (Supplementary Fig. 12a). Consistent with in vivo data, the variance in absolute capillary flow increased substantially with focal capillary dilation (Supplementary Fig. 12b). To assess whether capillary flow heterogeneity deviated from a normal range, we defined the expected baseline flow heterogeneity among a multitude of undilated capillaries within the two microvascular networks (Fig. 8g). This revealed that capillary dilation in the adult brain led to abnormally elevated regional flow heterogeneity, and that heterogeneity further increases with magnitudes of dilation seen in the aged brain (Fig. 8h).

Local dilations produce blood flow stalls and capillary regression in vivo

In surveying volumetric data collected across all in vivo triple ablation experiments, we found a higher frequency of blood flow stalls in experiments involving pericyte ablation, compared to off-target irradiations (Fig. 9a-d, Supplementary Fig. 13). These stalls appeared as vessels with no moving blood cell shadows within the dye-labeled plasma, and could be detected in images without line-scanning36. Stalls occurred with greatest likelihood one branch-point away from an uncovered capillary segment (Fig. 9e), suggesting blood steal by an adjacent dilated capillary. Many of the stalls occurred at divergent bifurcations, which occur closer to the arteriole-capillary transition zone (Supplementary Fig. 13b).

Fig. 9: Blood flow stalls and regression in capillaries neighboring dilation.
figure 9

a A divergent bifurcation close to the arteriole-capillary transition zone. One branch of the bifurcation is dilated from pericyte loss 3 days post-ablation, and the resulting re-direction of flow creates a stall in the alternate, covered branch. Images are representative examples from 14 ablation experiments across adult and aged mice. I.v. dye = intravenous dye. b Example of blood flow defects at a capillary junction near the arteriole-capillary transition zone. At 21 days post-ablation, a central capillary is uncovered, dilated, and flowing. The other branches have stalled in flow or regressed. c Occurrence of stalls in triple off-target sham irradiation and triple pericyte ablation experiments. Unpaired t test with Welch’s correction for unequal variances (two-sided), t(16.10) = 2.698; overall effect *p = 0.0158, for n = 6 off-target irradiation experiments (3 in adult mice, 3 in aged mice), n = 14 pericyte ablation experiments (7 in adult mice, 7 in aged mice). Data are shown as mean ± SD. d Additional example of a capillary regression occurring after a prolonged stall. e Plot of the relative location of stall and regression events following pericyte ablation, by branch order from a dilated segment. Observations from n = 14 ablation experiments (7 in adult mice, 7 in aged mice). f Schematic summarizing findings of blood flow interruption following pericyte coverage loss.

Most capillary stalls were transient, and presumably resolved by restoration of capillary tone in dilated neighbors. Transient stalls were defined as capillary segments with no blood flow on one day of imaging, but with flow observed in a subsequent imaging session. We could not determine the precise duration over which flow was stalled, due to limited sampling frequency. However, we suspect that some capillaries experienced prolonged loss of flow, as a subset of stalling events were later associated with regression of the non-flowing capillary segment (13% of total stalls). Regressions left behind a pericyte bridge without an endothelial lumen, reminiscent of a string capillary (Fig. 9b,d,f, Supplementary Fig. 13c). These outcomes were based on pooling data across adult and aged mice, as their frequency in focal ablation experiments was insufficient for age comparison. Additionally, we did not detect increases in stalled or low-flow capillaries in the in silico data (Supplementary Fig. 12c,d). This is not surprising since biological processes such as blood cell adhesion to the endothelium and potential vasoconstrictive events were not captured in our modeling. Collectively, these data reveal how altered capillary flow distribution due to pericyte loss can lead to cessation of blood flow in capillary segments, and lead to enduring loss of capillary structure.

Source link

Back to top button