1. Spectrally-resolved fluorescence and optical coherence Doppler microscopy for simultaneous imaging of neuronal, astrocytic and microvascular dynamics
Figure 1a illustrates a custom hybrid dual-channel fluorescence and ultrahigh-resolution optical coherence Doppler microscope (fl-ODM) which combines the two imaging modalities into an upright microscope body (FN1, Nikon) via epi-fluorescence cube turrets (C1, C2) for in vivo small animal studies. An ultrahigh-resolution optical coherence angiography (µOCA) and Doppler tomography (µODT) system in the near infrared range (λ=1.3µm, ∆λ=230nm) was integrated through a dichroic mirror (λ1DM=1.1µm) in C1 to provide 3D images of the microvasculature and quantitative cerebral blood flow velocity (CBFv) in vascular networks in the mouse cortex over a large field of view (FOV, e.g., 2.4⋅2⋅1.2mm3) with capillary resolution (e.g., <5µm). The technical details of 3D µOCA/µODT were previously reported23 except a custom high-fidelity 2D confocal laser scanning module to interconnect the µOCT engine to fl-ODM. In parallel, a custom epifluorescence cube (C2) is used for 2-channel spectral-multiplex imaging of the synchronized intracellular calcium fluorescence changes in astrocytes (Ca2+A expressed with GCaMP6f: λEX1=485±12nm, λDM1=495nm, λEM1=520±20nm) and in neurons (Ca2+N expressed with jRGECO1a: λEX2=559±8nm, λDM2=573nm, λEM2≥574nm) over a larger FOV (e.g., 4⋅3mm2). Pulsed high-power narrow-band blue (488nm) and yellowish green (560nm) LEDs from a light engine (Aura III, Lumencor) were used for time-sharing spectral excitation and synchronized with a SCMOS camera (Zyla 5.5, Andor) for sequential fluorescence image acquisition (T=10ms exposure per channel). Ca2+N(t) and Ca2+A(t) imaged at up to 80fps were quantified as the relative florescence changes (∆F/F) vs their baselines to represent neuronal or astrocyte activities, respectively.
To define the mechanisms by which astrocytic and neuronal activities are involved in NGV interactions in vivo, we used viral injection to express genetically encoded Ca2+ indicators in mice in a cell-specific manner. Figure 1b illustrates our approach to express astrocytic Ca2+A and neuronal Ca2+N in the PFC in vivo. The use of GFAP-cre mouse with delivery of two mixed viral vectors including 50% AAV5.CAG.Flex.GCaMP6f.WPRE.SV40 (#100835, Add-gene) and 50% AAV1.Syn.NES-jRGECO1a.WPRE.SV40 (100854, Add-gene) allowed us to express GCaMP6f for Ca2+A fluorescence and jRGECO1a for Ca2+N fluorescence in the cortex. Prior to imaging, a cranial window was implanted above the PFC as illustrated in Fig. 1c (see Method Section for details). Figures 1d1-d3 show simultaneous in vivo images of Ca2+A (d1) and Ca2+N (d2) fluorescence from PFC of a GFAP-cre mouse at ~4wks after viral injection of synapsin jRGECO1a for neurons and cre-GCaMP6f for astrocytes. Ex vivo double staining with GFAP antibody to label astrocytes and NeuN antibody to label neurons confirmed astrocyte- (d1’) and neuron- (d2’) specific Ca2+ expressions. Astrocytes also modulate neuro-vascular coupling (NVC) through their process endfeet forming close interactions with neurons and microvessels3. Our ex vivo images (Figs. 1e-g) show that astrocytes (labeled by GFAP, green) ensheathe neurons (labeled by NeuN, red) and microvessels (red dashed lines – GFAP of astrocyte endfeet) in the mouse cortex (e, g).
2. Cocaine increased neuronal Ca2+N and astrocytic Ca2+A activity and decreased CBFv
To demonstrate the technical capability of fl-ODM for tracking neuronal, astrocytic and vascular changes in real time, we simultaneously imaged the mouse PFC (A/P: +2.5; M/L: 0.5; D/V: -0.5 mm) to detect activations of Ca2+N, Ca2+A fluorescence and local CBFv changes elicited by an acute cocaine challenge (1mg/kg, i.v.). Figures 2a’-c’ show representative Ca2+N, Ca2+A and CBFv images obtained from the PFC of a GFAP-mouse at baseline (before cocaine). To assess the dynamic changes in Ca2+N, Ca2+A and CBFv from time of cocaine injection, five regions of interest (ROIs, i.e., white circles illustrated in a’ and b’) were selected in the cortex within the fluorescence expressing regions and five in vessels in the surrounding area as shown in Fig. 2c’. Figures 2a-c show the time courses of Ca2+N, Ca2+A and CBFv in response to cocaine, respectively. Cocaine increased Ca2+N and Ca2+A activities (Figs. 2a-b), whereas it decreased CBFv in cerebrovascular trees (c). Specifically, cocaine triggered a Ca2+N increase to a maximum of 2.92%±0.20% (m=5) over baseline at tp−N=8.6±0.4min, after which it recovered within tr−N=27.6±3.3min followed by a downshoot. Similarly, cocaine increased Ca2+A to a maximum of 3.87%±0.23% (m=5) at tp−A=10.6±0.4min, peaking slightly behind the tp−N of Ca2+N (p=0.008, m=5). However, Ca2+A did not return to baseline until tr−A=50.6±5.6min (m=5), indicating a longer-lasting effect of cocaine on astrocytes than on neurons (p=0.01, m=5). Meanwhile, cocaine decreased CBFv to -22.8%±5.1% (m=5) below baseline at tp−V=16.6±2.2min followed by a gradual recovery by tr−V=58.6±5.8min, also indicative of a long-lasting CBFv decrease presumably due to vasoconstriction.
In addition, fl-ODM allowed us to concomitantly measure cocaine-induced transient CBFv changes in arteries, veins and capillaries. Figures 2d-e show time-lapse µODT images and the ratio changes over the baseline of a smaller volume (2×0.3×1.2mm3, marked by a dished box in Fig. 2c’) in the mouse PFC acquired to quantify dynamic flow changes (2min/volume) from baseline (t<-4min) to t>50min after cocaine (1mg/kg, i.v., t=0min). The 3D µODT image in Supplemental Fig.S1 shows that the selected ROIs included pial flows in layer 1 and deep flows in layers 4–5. Under isoflurane anesthesia, all three vessel compartments (f) showed flow decreases after cocaine injection, among which arteriolar (AF) and venular (VF) flows dropped −19.8%±6.3% (p<0.001, m=4; t=8-36min) and −29.3%±4.5% (p<0.001, m=4; t=6-36min), respectively, followed by a gradual recovery to their baselines at 49.5±4.6min and 45.5±3.2min, respectively. Although capillary flows (CF) showed an overall decrease to -13.5%±3.0% that peaked at 20±2.9min, individual flow changes varied, with increases over 7% in some capillaries and decreases over −35% in others. The heterogeneity in the capillary responses highlights the importance of measuring flow in multiple capillaries instead of isolated vessels when studying NGV interactions and its responses to cocaine.
3. Astrocytic Ca2+A, but not neuronal Ca2+N, correlated with cocaine-induced CBFv decreases
Figure 2 illustrates the increases in Ca2+N and Ca2+A and the decreases in CBFv triggered by cocaine in the mouse PFC. Unlike the short-lasting dynamic change in Ca2+N fluorescence (e.g., <30min), cocaine induced long-lasting changes in Ca2+A and CBFv (e.g., 50-58min). To analyze the temporal relationship among these cellular and vascular changes, we studied 6 additional mice to examine the effects of acute cocaine (1mg/kg, i.v.) on neuronal Ca2+N, astrocytic Ca2+A and microvascular CBFv dynamic responses in the PFC. Figure 3 summarizes the cocaine-induced mean changes in Ca2+N, Ca2+A and CBFv within different vascular compartments (e.g., arteries, veins and capillaries) acquired from seven animals. The mean time-course changes in Fig. 3a indicate that cocaine induced a ΔCa2+N increase of 2.46±0.88% at tp−N=8.2±2.1min followed by recovery to baseline at tr−N=28.8±3.5min. Similarly, ΔCa2+A increased to 2.97±0.43% at tp−A=12.1±2.2min, but the effect was long-lasting and did not return to baseline until tr−A=59.5±8.0min. In parallel cocaine decreased mean vascular ΔCBFv to -25.1±4.9% at tp−V=21.0±2.9min, which slowly recovered to baseline at tr−V=64.0±7.5min, with a similar duration to that of ΔCa2+A. Statistical comparisons of the response duration to cocaine between ΔCa2+N, ΔCa2+A and ΔCBFv are summarized in Fig. 3b, which indicates that the response duration of Ca2+A and CBFv to cocaine was significantly longer (~2 folds) than that of Ca2+N (P*=0,005 and P*=0.002, respectively, n=7). No significant difference was found between tr−A and tr−V (P=0.62, n=7). The temporal correlations between cocaine-induced ΔCa2+A(t), ΔCa2+N(t) and ΔCBFv(t), were computed and the data derived from all 7 mice is illustrated in Supplemental Fig.S2. The results in Fig. 3c show a strong correlation between ΔCa2+A and ΔCBFv (0.765±0.048, n=7) that was significantly higher than the significant but weaker correlation between ΔCa2+N and ΔCBFv (0.389±0.115, P*=0.011, n=7).
4. Activation of GFAP-DREADDs(Gi) inhibited astrocytic Ca2+Aand induced vasodilation and increased CBFv during baseline
DREADDS is a chemogenic approach that enables subtype selective activation (Gq) or silencing (Gi) of cellular signaling (e.g., astrocytes) via clozapine activation25. A cocktail of two viruses consisting of 0.4µl AAV5.CAG.Flex.GCaMP6f.WPRE.SV40 and 0.4µl AAV5.GFAP.hM4D(Gi)- mCherry was injected into the PFC of GFAP-cre mice. Figure 4 shows in vivo results using Gi-coupled DREADDS(hM4Di) expression in astrocytes (referred as GFAP-DREADDS(Gi)) into the mouse PFC. After 4-6wks from injection, time-lapse images of Ca2+A fluorescence and cerebrovascular networks in the PFC were continuously acquired before and after clozapine injection (0.8mg/kg, i.p.) at t=0min for over 40min. Clozapine instead of clozapine-N-oxide (CNO) was used to activate DREADDS(Gi) because a recent study showed that clozapine (the CNO metabolite) rather than CNO itself stimulates the DREADDS receptor25. Figure 4b is a representative ratio ΔCa2+A image of a mouse PFC post clozapine (t=25min) over its baseline (t=-5min), showing Ca2+A decreases within the blue region of brain tissue along with vasodilation (red tracks in vascular trees, Fig. 4a). Details of vascular responses to clozapine activation of DREADDS(Gi) can be visualized in Supplemental movie VS1 and ratio image shown in Supplemental Fig.S3. Figure 4d summarizes the mean changes in vascular diameters and CBFv as a function of time post clozapine administration across animals (m=5 ROI/parameter/animal, n=5 mice). DREADD(Gi) activation resulted in vasodilation, with an increase of Δφ=9.17%±0.96% in mean vessel diameters (black trace) from baseline (t=-5min, Δφ=0.22%±0.35%) and an increase in ΔCBFv of 10.51%±2.95% at 25min post clozapine compared to baseline (t=-5min, ΔCBFv=0.21%±1.18%). The time course of ΔCa2+A in response to astrocyte inhibition is shown in Fig. 4e, revealing a significant decrease after 5min following clozapine injection (p<0.05); at t=25min post clozapine, ΔCa2+A decreased −3.0%±1.0% over its baseline (t=-5min, ΔCa2+A=0.02%±0.16%). These results provide evidence that astrocytic signaling modulates vascular tone at baseline, thus adjusting CBFv within the brain.
To evaluate whether astrocytic inhibition would affect neuronal activity at baseline and in response to cocaine (see subheading below), we injected a mixture of viruses to express GCaMP6f into neurons (0.4ul AAV5.Syn.GCaMP6f.WPRE.SV40) and to express GFAP-DREADDs(Gi) (0.4µl AAV5.GFAP.hM4D(Gi)-mCherry) into astrocytes in the PFC of mice (n=5). Figure 4c shows a representative ratio image of neuronal ΔCa2+N fluorescence of a mouse PFC at t=25min after GFAP-DREADDs(Gi) activation by clozapine over the baseline (t=-5min), exhibiting no neuronal Ca2+N fluorescence changes. Figure 4f shows the mean ΔCa2+N time courses across animals, indicating no significant changes before (-0.08%±0.06%, t=-5min) and after (-0.64%±0.43%, t=25min) GFAP-DREADDs(Gi) activation (p>0.05, m=5/animal, n=5). This result indicates that clozapine activation of GFAP-DREADDs(Gi) to inhibit astrocyte activity (Ca2+A) did not influence neuronal activity (Ca2+N) at baseline. The specificity of the DREADDs(Gi) delivery into astrocytes was corroborated with immunostained brain sections that indicated AAV5.GFAP.hM4D(Gi)-mCherry expression uniquely in astrocytes (Supplemental Fig.S4).
Figure 4g plots the temporal correlations between ΔCa2+A(t) and ΔCa2+N(t) with Δφ(t), and Fig. 4h indicates that the cross-correlation between ΔCa2+A and Δφ (r=0.651±0.07) was significantly higher than that between ΔCa2+N and Δφ (r=0.411±0.10, P*=0.01). A similar correlation analysis between ΔCa2+A(t) and ΔCa2+N(t) with ΔCBFv(t) shows that the correlation between ΔCa2+A and ΔCBFv (r=0.535±0.03) was significantly higher than that between ΔCa2+N and ΔCBFv (r=0.347±0.05, P*=0.008) (Supplemental Fig.S5). These results indicate that inhibition of astrocytes resulted in vasodilation and CBFv increases, demonstrating the modulatory role of astrocytes in setting baseline vascular tone and blood flow in the brain. Though the correlations between decreases in neuronal activity and vasodilation and CBFv increases were also significant, the effect was significantly smaller.
5. GFAP-DREADD (Gi) inhibition of astrocytic Ca2+A activation attenuated the decreases in CBFv and the neuronal activation triggered by acute cocaine
To examine whether inhibition of astrocytic signaling (Ca2+A) could block the CBFv decrease due to cocaine’s vasoconstricting effects, we imaged the vascular responses to cocaine in the PFC without and with GFAP-DREADDS(Gi) activation by clozapine. Animal preparations were as described in the experiment shown in Fig. 4 but extended to assess the effects of acute cocaine. For this experiment, two sets of imaging sessions were conducted with at least 2hr separation between two sequential cocaine injections: images were acquired from 10min prior (baseline) to 60min after the first cocaine injection (1mg/kg, i.v.); this was followed by a>40min no intervention period after which a clozapine injection (0.1mg/kg, 0.16ml) was given and 30min later the second in vivo imaging session was initiated and included a 10min baseline prior to and 60min following a second cocaine injection (1mg/kg, iv). Astrocytic Ca2+A fluorescence and CBFv images were continuously recorded prior to and following each cocaine injection (n=3). In an additional group of animals (n=4) we assessed whether inhibition of astrocytes influenced the neuronal response to cocaine by imaging neuronal Ca2+N and astrocytic Ca2 fluorescence and CBFv changes in response to cocaine without and with GFAP-DREADD(Gi) activation by clozapine.
Figure 5(a-b) show the temporal responses of mean astrocytic Ca2+A to cocaine before and after DREADD(Gi) activation, in which the insets illustrate the corresponding representative ratio images of Ca2+A fluorescence at t=30min after cocaine vs baseline (t=-2min), indicating that cocaine-induced mean astrocytic Ca2+A increase (ΔCa2+A (a) was inhibited after GFAP-DREADD(Gi) activation (b)). Multiple ROIs (m=5) within the PFC from each animal (n=3) were selected to track temporal Ca2+A fluorescence changes (ΔCa2+A/Ca2+A) after cocaine and the data averaged to derive mean ΔCa2+A(t) curves in Fig. 5(a, b). For statistical comparison, the ΔCa2+A rate, defined as the averaged per minute Ca2+A increase post cocaine during t=0-30min was blocked from 5.55%±0.83%/min to 0.35%±0.32%/min (p*=0.004) after GFAP-DREADD(Gi) activation (Fig. 5c), consistent with inhibition of cocaine-induced Ca2+A increase in astrocytes. Similarly, Fig. 5(d-e) show the mean temporal CBFv responses (ΔCBFv/CBFv) to cocaine before and after GFAP-DREADD(Gi) activation (n=7, including animals both with astrocytic-GCaMP6f or neuronal-GCaMP6f and GFAP-DREADD(Gi) expressions) as well as their corresponding representative ratio images to illustrate the blunting of cocaine-induced ΔCBFv decrease. The CBFv decrease from −0.72%±1.42% at baseline (t=-2min) to -14.64%±4.75% at t=30min after cocaine (Fig. 5d) was reduced to -1.43%±5.02% (t=30min) from its baseline −0.47%±1.44% (t=-2min) with GFAP-DREADD(Gi) activation (Fig. 5e). Figure 5(f) indicates that cocaine-induced ΔCBFv decrease was reduced significantly from −10.1%±2.1%/min to -2.03%±1.6%/min after GFAP-DREADD(Gi) activation (p*=0.01, n=7). Demonstration of cocaine-induced vascular response without and with astrocytes’ inhibition via DREADDs(Gi) activation can be visualized in Supplemental movies VS2 and VS3, respectively. It showed that cocaine induced vasoconstriction, but this effect was eliminated by astrocytes’ inhibition after DREADDs(Gi) activation (Supplemental Fig.S6). These results support our hypothesis that blocking Ca2+A increases would abolish cocaine-induced vasoconstriction and prevent CBFv decreases.
While GFAP-DREADD(Gi) targets astrocytes and inhibits cocaine-induced Ca2+A increase (Fig. 5a), it was unclear whether it could indirectly modify neuronal responses to cocaine. Figure 5(g-h) shows the temporal responses of mean neuronal Ca2+N to cocaine before and after DREADD(Gi) activation to inhibit Ca2+A. Unlike the long-lasting response of astrocytes to cocaine (>60min shown in Fig. 5a) the neuronal Ca2+N increase returned to baseline at 30min after cocaine as illustrated in Fig. 5g. The full-width-half-maximum duration of cocaine-induced Ca2+N increase was reduced from τg=11.02±1.85min to τh=2.24±0.62min after DREADD(Gi) activation. The ΔCa2+N rate decreased from 1.11%±0.29%/min to 0.11%±0.1%/min after GFAP-DREADD(Gi) activation (Fig. 5i), thus indicating a significant reduction of cocaine-induced neuronal activation (p*=0.02, n=4). Taking together with the effect of DREADDS(Gi)’s blockade on cocaine-induced CBFv decreases (Fig. 5f), this result indicates that cocaine’s effects on CBFv and neuronal activity (Ca2+N) were modulated by Ca2+A signaling. Thus, inhibition of astrocytic activity might help alleviate cocaine associated PFC dysfunction resulting from improper tissue perfusion and the decrease in neuronal reactivity might help reduce compulsive drug taking,