Contribution of low- and high-flux capillaries to slow hemodynamic fluctuations in the cerebral cortex of mice

Abstract

We employed optical coherence tomography to measure cerebral cortical capillary red blood cell (RBC) flux in mice. The results suggest that baseline-flux weakly depends on cortical depth. Furthermore, under hypercapnia, low baseline-flux capillaries exhibit greater flux increases while the higher ones saturate, resulting in RBC-flux homogenization. Power-spectrum analysis indicates that higher flux capillaries saw greater flux variability in the low-frequency range (0.01–0.1 Hz) both at baseline and during hypercapnia. These results suggest that lower baseline-flux capillaries have more reserve to deliver oxygen with increased blood flow; but higher ones more strongly impact the low-frequency fluctuations associated with BOLD fMRI measurements of resting state functional connectivity.

Introduction

Low-frequency oscillations (0.01–0.1 Hz) of the fMRI blood oxygen level dependent (BOLD) signal play a significant role in assessing brain function.^1,2 However, little microscopic evidence exists to characterize the microvascular origin of this macroscopic phenomenon.^3,4 Cerebral cortical capillaries exhibit highly heterogeneous baseline red blood cell (RBC) flux raising the question of the role that such baseline-flux heterogeneity plays in supporting brain function. Addressing this question is complex due to the need to analyze large ensembles of capillaries with sufficient spatiotemporal resolution.⁵ Techniques such as confocal microscopy and two photon microscopy (TPM) have been employed to capture individual capillaries in rodent cerebral cortex,^6–8 finding that, under stimulation, capillary RBC-flux has a tendency to saturate at a high RBC-flux value.⁶ This observation further leads to the question of how the dynamic RBC-flux heterogeneity correlates with brain oxygen delivery. However, the limited number of capillaries acquired with confocal microscopy or TPM, along with the long acquisition time (e.g., minutes) to monitor multiple capillaries challenge further investigation. Recently, it has been shown that RBC-flux in many capillaries could be measured simultaneously with optical coherence tomography (OCT).⁹ Here, we utilize this OCT technique and extend it to monitor the temporal fluctuations of RBC-flux in multiple capillaries simultaneously at baseline and during hypercapnia. The results extend our previously observed heterogeneous RBC-flux properties by demonstrating that, under hypercapnia, the lower baseline-flux capillaries exhibit greater flux increases than higher ones. Further, power-spectrum of time varying RBC-flux showed that higher baseline-flux capillaries exhibited greater flux variability in the low-frequency range (0.01–0.1 Hz) at baseline; and higher flux was associated with greater flux variability both at baseline and during hypercapnia. These results suggest that lower baseline-flux capillaries have the capacity to deliver more oxygen with increased blood flow; but the higher ones more strongly impact the low-frequency fluctuations associated with BOLD fMRI measurements of resting state (RS) functional connectivity.^1,2

Materials and methods

Spectral domain OCT

Angiograms and RBC-passage B-scans were obtained using spectral domain OCT (SD-OCT) system.^9,10 Briefly, a light source having a central wavelength at 1310 nm with a 170 nm bandwidth was employed. The transverse resolution was either 3.5 or 7 µm when using the 10 × (NA = 0.26) or 5 × (NA = 0.14) objective, respectively. The acquisition speed was 47,000 A-line/s along the depth axis (Z), and ∼667 B-scan/s in the X–Z cross-sectional plane. Typical angiogram and RBC-passage B-scan images are shown in Figure 1(a) and 1(b).

Figure 1.

(a) An en face MIP of an angiogram with the red line indicating the location of (b) consecutive RBC-passage B-scans. (c) The representative RBC-passage time-course (1-s duration) of the capillary denoted by the red circle in (b). (d) A typical RBC-flux time-course (1-min) at rest (blue) and under hypercapnia (red). (e) Histogram of RBC-flux averaged per capillary over the 2-min RBC-flux time-course, at baseline (blue) and during hypercapnia (red). Another angiograms at (f) baseline and (g) its counterpart during hypercapnia showing enhanced capillary perfusion during hypercapnia. The three red arrows denote three sites having enhanced perfusion relative to baseline. (h) Low baseline-flux capillaries exhibited a greater increase during hypercapnia. Scale bar: 100 µm.

Animal and experiment

All animal experimental procedures were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care and conducted in accordance with Massachusetts General Hospital guidelines and ARRIVE guidelines. CD1-Elite mice (Charles River Laboratories, n = 8, male, ∼3 months, ∼35 g) were investigated at baseline and during hypercapnia (5% CO₂ in air/O₂ mixture) over separate 2-min periods. To remove potential confounds, baseline and hypercapnic measures were alternated. Animal preparation and procedures were conducted similar to what was described previously under α-Chloralose anesthesia.¹¹ Previous animal studies showed that among other anesthetics, α-Chloralose had a relatively light effect on cardiovascular and respiratory functions.¹² Physiological parameters were monitored: At baseline, averaging over all mice, the mean arterial blood pressure, arterial blood PO₂ and arterial blood PCO₂ were ∼93 mmHg, ∼107 mmHg, and ∼35 mmHg, respectively. During hypercapnia, the arterial blood PCO₂ was increased by ∼51%. Body temperature was maintained at ∼36.9 ℃ throughout all experiments.

Determination of flux

The RBC-passage B-scan images were spatially smoothed with a 2D Gaussian filter (σ = 1). As indicated by the red circle in Figure 1(b), capillaries could be identified in the variance images of consecutive B-scan images, and then manually selected. For each point selected, the pixel intensities at the same coordinate of the consecutive B-scans were extracted to reveal the RBC-passage time-course;⁹ then the RBC-passage time-course was normalized having the range from 0 to 1; finally, a MATLAB function findpeaks was used to count the RBC-passage peaks using a threshold defined as the half maximum intensity of each RBC-passage time-course. To eliminate false positives in counting RBC-passage peaks, we applied a temporal 1D Gaussian filter (σ = 1) to denoise the RBC-passage time-course. The filter performance was evaluated by simulations (not shown). Briefly, speckle noise (estimated from experimental data) was added to simulated RBC-passage time-courses. Before filtering, the ratio of the estimated flux to true flux was ∼179% due to false detections associated with noise, but decreased to ∼101% after filtering. The RBC-flux time-course of each capillary was obtained by determining the number of RBC-passage peaks every second for the 2-min scan.

Results

Angiogram, RBC-passage B-scans, and RBC-flux estimation

For each mouse, volumetric angiography was performed.¹⁰ A representative en face maximum intensity projection (MIP) of the angiogram is shown in Figure 1(a). Multiple lines were then chosen from the angiogram to conduct consecutive RBC-passage B-scans for 2 min. Figure 1(b) displays consecutive B-scan images acquired at the representative red line denoted in Figure 1(a). Here, the B-scan line spans a distance of 100 µm and was repeated every Δt = 1.5 ms. In previous studies with TPM or confocal microscopy, RBC-flux was typically measured with Δt = 2 ms.^6,7 With Δt = 1.5 ms, we expect a greater upper limit for accurately measuring RBC-flux. Indeed, simulations suggest (not shown) that with Δt = 1.5 ms, RBC-flux up to 150 RBCs/s can be measured accurately. The red circle in Figure 1(b) denotes a capillary. The corresponding RBC-passage time-course is shown in Figure 1(c), with RBC-passage peaks identified in red. The number of RBC-passage peaks was averaged every second to provide an RBC-flux time-course. Figure 1(d) displays a typical RBC-flux time-course showing an increased RBC-flux from baseline to hypercapnia. In this study, on average, ∼45 capillaries were selected per mouse. Figure 1(e) shows the histogram of mean RBC-flux averaged for the 2-min scan of each capillary of n = 8 mice. The mean RBC-flux ranged from ∼20 to ∼90 RBCs/s at baseline, and was “right shifted” to greater values during hypercapnia. Averaging over all capillaries from all mice, the baseline-flux was ∼53 RBCs/s, and increased by ∼15% to ∼61 RBCs/s under hypercapnia. The intensity of the capillaries in the angiograms (Figure 1(f) and 1(g)) is an index of RBC flux,¹³ qualitatively showing that the RBC-flux increases during hypercapnia are associated with increased capillary intensity.

RBC-flux homogenization during hypercapnia

Figure 1(h) shows that lower baseline-flux capillaries saw a greater RBC-flux increase than the higher ones during hypercapnia¹¹ (|slope| ≈ 2.5% RBCs⁻¹·s). And suggested by Figure 1(e), the RBC-flux distribution was compressed during hypercapnia. The coefficient of variation of the RBC-flux, defined as the ratio of the standard deviation to the mean,⁸ decreased from ∼27.1% (baseline) to ∼19.6% (hypercapnia). This phenomenon could be potentially explained by the observation that smaller capillaries dilate more than larger ones resulting in a reduced heterogeneity in capillary caliber.^7,14

Depth dependence of capillary RBC-flux

We observed a weak dependence of the baseline-flux on cortical depth. Figure 2(a) shows that baseline-flux decreases with depth (|slope| ≈ 0.02 RBCs·s⁻¹·µm⁻¹). This observation is supported by a previous study showing that RBC-speed was proportional to RBC-flux and decreased with cortical depth.⁶ Furthermore, Figure 2(b) shows that, under hypercapnia, the capillaries located deeper saw a slightly greater RBC-flux increase (rFlux) than those lying superficially (|slope| ≈ 0.05% µm⁻¹).

Figure 2.

(a) The baseline-flux decreases with cortical depth. (b) RBC-flux increases more with depth during hypercapnia. (c) The RBC-flux variability at baseline weakly depends on cortical depth. (d) The relative changes of fluctuation power increase slightly with depth during hypercapnia. (e) From baseline (blue) to hypercapnia (red), the low-frequency flux fluctuations of the lower baseline-flux capillaries (RBC-flux ≤ 56 RBCs/s) significantly increases. (f) Hypercapnia induced no significant change of flux variability in the higher baseline-flux capillaries (RBC-flux > 56 RBCs/s). (g) Flux variability is proportional to flux, both at baseline (blue) and during hypercapnia (red). (h) Lower baseline-flux capillaries saw a greater increase in fluctuation power than the higher ones during hypercapnia.

Low-frequency oscillations of time varying RBC-flux

We aim to characterize the capillary RBC-flux fluctuations and their contribution to the low-frequency signals associated with RS functional connectivity studies performed with BOLD fMRI.^1,2 For each capillary, the RBC-flux time-course was extracted at baseline and during hypercapnia over the 2-min B-scans. Next, the power-spectrum of RBC-flux time-course was retrieved; and then the power-spectrum from 0.01 to 0.1 Hz was averaged over all capillaries/mice. A typical power-spectrum is shown in Figure 2(e). The unit of Y-axes in Figure 2(c, e, f, and g) was converted from power-magnitude to RBCs/s by

$$(\mathrm{mean}\hspace{1em}\mathrm{flux})\times \frac{(\mathrm{power}\hspace{1em}\mathrm{spectrum})}{(\mathrm{DC}\hspace{1em}\mathrm{component}\hspace{1em}\mathrm{of}\hspace{1em}\mathrm{power}\hspace{1em}\mathrm{spectrum})}$$

permitting a physiological interpretation of the power-spectrum as flux variability (ΔFlux). Figure 2(d) and (h) presents relative power changes (rPower), so the unit was not changed.

The mean flux variability for each capillary was calculated by averaging the value of flux variability over the frequency of 0.01–0.1 Hz. Analogous to Figure 2(a), Figure 2(c) shows that the flux variability at baseline weakly depends on cortical depth (|slope| ≈ 0.1 RBCs·s⁻¹·µm⁻¹). Figure 2(d) reveals that the flux variability (rPower) of the capillaries located deeper saw a slightly greater increase than those lying superficially (|slope| ≈ 0.03% µm⁻¹).

Contribution of low and high baseline-flux capillaries to the low-frequency signals

Capillaries were divided into two groups having low and high baseline-flux, separated by the median RBC-flux value (56 RBCs/s). The flux variability corresponding to 0.01–0.1 Hz was averaged over all capillaries in each group separately. Figure 2(e) demonstrates that from baseline (blue) to hypercapnia (red), the flux variability of the lower baseline-flux capillaries significantly increases (P < 0.05, Student’s t-test); while no significant change happens to the higher ones (Figure 2f). This is consistent with the notion that the higher baseline-flux capillaries have less reserve to increase their fluctuations.

Figure 2(g) shows that the RS flux variability (blue) is proportional to baseline-flux (|slope| ≈ 0.13); and the flux variability during hypercapnia is proportional to hypercapnic flux with the slope remaining the same. Figure 2(h) indicates that lower baseline-flux capillaries saw a greater increase of flux variability than the higher ones during hypercapnia (|slope| ≈ 2% RBCs⁻¹·s). This observation demonstrates that low-frequency oscillations associated with RS functional connectivity studies are proportional to baseline-flux; but, during hypercapnia, the lower baseline-flux capillaries saw a greater increase in the low-frequency oscillations.

Discussion and conclusion

We investigated cerebral capillary RBC-flux in mice. With SD-OCT, many capillaries were monitored simultaneously for 2 min. Analyzing the selected capillaries, the mean baseline-flux of all capillaries of n = 8 mice was ∼53 RBCs/s, and increased by ∼15% under hypercapnia. Such an increase is consistent with previously published TPM and confocal microscopy data.^6–8 In previous studies with Δt = 2 ms, RBC-flux > 150 RBCs/s was rarely observed.^6,7 To ensure that a ceiling effect was not introduced by temporal sampling, we increased the sampling rate to Δt = 1.5 ms. Our simulations (not shown) indicate that Δt = 1.5 ms could raise the upper limit of measuring RBC-flux up to ∼ 150 RBCs/s.

Further analysis revealed that (1) baseline-flux, flux variability and their relative changes had weak dependence on cortical depth. (2) With increasing demand for oxygen, RBC-flux increases might preferentially happen in the lower baseline-flux capillaries to homogenize RBC-flux; while the highest baseline-flux capillaries in fact showed no increase. (3) Lower baseline-flux capillaries might have larger reserves to increase their low-frequency fluctuations. (4) Flux variability is proportional to absolute flux, both at baseline and during hypercapnia.

Limitations exist in this study. First, although animal physiology was controlled in the experiments, potential effect of anesthesia remained confound. Second, the intensity of the back-scattered OCT signal attenuated with depth, reducing the sensitivity and the number of observed capillaries with depth. However, for those capillaries identified, the analysis procedure remained the same and was able to unequivocally measure the RBC-flux. Third, the algorithms for estimating RBC-passage flux and noise reduction should be further optimized. Specifically, although constant filter parameters were used to avoid biasing, “dual-peak” RBC-passages induced by two adjacent RBCs could possibly be masked resulting in underestimating the RBC-flux.

In conclusion, this work characterized the behavior of cerebral cortical capillary RBC-flux. Specifically, we found that lower baseline-flux capillaries might have the capacity to more significantly increase their oxygen delivery with increased blood flow. This RBC-flux homogenization is reminiscent of the notion of capillary transit time homogenization improving the oxygen delivery efficiency.¹⁵ Furthermore, power-spectrum analysis shows flux variability is proportional to absolute flux, both at baseline and during hypercapnia. Further, higher baseline-flux capillaries saw greater RS low-frequency fluctuations indicating that, although lower baseline-flux capillaries have more reserves to increase their oxygen delivery to the tissue, the higher baseline-flux capillaries appear to contribute more to the low-frequency fluctuations associated with BOLD fMRI measurements of RS functional connectivity.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: CIHR (MOP-130253) and NIH (R01-EB021018 and P01-NS055104).

Declaration of conflicting interests

The authors declare no conflict of interest.

Author contributions

BL conducted the experiments, analyzed the data and drafted the manuscript. JL provided comments on the early version of this manuscript, and shared basic code of processing SD-OCT angiogram data. DB and FL contributed conception of interpreting the results, and revised the article.