Cloud-based Miniscope Reveals Infraslow Changes in Neurovascular Function in Freely Moving Animals

Introduction: From GoPros to mini cameras, continuous imaging is driving unprecedented advancements in our daily lives. Using an analogous approach for neuroscience research wherein the brain is imaged continuously during natural animal behavior could revolutionize our understanding of brain function. However, conventional functional imaging platforms such as MRI or PET scanners are incapable of imaging freely moving animals. Miniaturized optical imaging microscopes or “miniscopes” offer hope for shattering this barrier by allowing the imaging of brain activity in freely behaving animals. However, despite their utility, miniscopes can conduct continuous imaging for only a few hours due to technical issues. Moreover, most miniscopes can only image neuronal activity or vascular function, but not both. Finally, they cannot be remotely operated as would be necessary for long-term (i.e. >24 h) imaging. To circumvent these hurdles, we developed the “CloudScope”, by combining a robust 3D-printed miniscope design with an internet of things (IoT) network architecture. The CloudScope could successfully break the extant time barrier by enabling continuous 24-hour-long functional neuroimaging in freely behaving mice, while allowing the user to conduct real-time imaging from anywhere in the world. We call this capability “neurosurveillance” and showcase its exciting potential for understanding the infraslow changes ( < 0.1 Hz) in neuronal activity and hemodynamics that occur during normal animal behavior.

Materials and

Methods: Fig. 1a shows the IoT architecture of CloudScopes that can be distributed globally in animal facilities or labs and connect via Wi-Fi to a cloud-based server. This allows a user to log into the server via the internet, set acquisition parameters, initiate image acquisition, and observe experiments from anywhere. Fig. 1b shows a picture of the CloudScope ( < 3.5 g) on a mouse which permits visualizing maps of neuronal activity (NeuACT, via fluorescence or FL, Fig. 1c), changes to cerebral blood volume and oxygenation (CBV and HbSAT via intrinsic optical signals or IOS, Fig. 1d) and cerebral blood flow (CBF, via laser speckle contrast or LSC, Fig. 1d) over a 33 mm2 field of view (FoV) at a spatial resolution of 6 m every 5 seconds. Animal behavior was captured via a behavioral camera and annotated offline to depict the animal’s activity level via a score (e.g. MobSCORE) from 0-2. We interrogated the murine motor and somatosensory cortices (i.e. via a 3 mm diameter FoV centered -2/+2 mm AP/ML), continuously over 24-hours in a cohort (n = 5) of freely behaving mice. Next, for identifying infraslow changes in brain function, we quantified: (i) the relationship between NeuACT, CBF, and the Pearson’s correlation coefficient of CBF vs CBV (RCBV,CBF) with MobSCORE; (ii) peak NeuACT and CBF areas; (iii) the degree of vasomotor synchronization between arteriolar and venular vessel networks; over 15-minute epochs, wherein each epoch approximated the duration of a conventional head-fixed neuroimaging experiment. Data reported as mean ± std.

Results, Conclusions, and Discussions: As shown for a representative animal in Figs. 1f-h, changes to NeuACT and CBF were tightly coupled to animal mobility (Spearman’s ρ = 0.85 ± 0.02 and 0.77 ± 0.06 for NeuACT vs. MobSCORE and CBF vs. MobSCORE, respectively, n = 5 mice). However, the CBF vs. CBV correlation (RCBF, CBV) was significantly diminished during epochs of high animal activity (Fig. 1i, for epochs with average MobSCORE > 0.5 vs. MobSCORE ≤ 0.5, RCBF, CBV = 0.50 ± .09 < 0.73 ± 0.05, p < 0.001). Because RCBV,CBF indicates the degree to which CBF changes due to vasodilations or vasoconstrictions, this suggests that sympathetic activity, e.g. heart rate or blood pressure changes instead of neurovascular coupling may be driving the observed CBF changes during high animal activity. Moreover, as shown in Fig. 1j, cortical regions that exhibited peak NeuACT (top 50%, shaded green) did not colocalize with those having peak CBF (shaded red, overlap coefficients κ = 0.40 ± 0.04 in n = 5 mice) over 24-hours, underscoring the prevalence of CBF changes due to non-neuronal sources during natural behavior. We also quantified how microvessel diameter changes, i.e. vasomotor activity, was synchronized (VesSYNC) within pial arteriolar and venous networks during natural animal behavior. Fig. 1k shows arteriolar (“A”, in red) and venous (“V”, in blue) microvessel segments marked as “nodes”. VesSYNC was computed per each node pair as the correlation between the time series of their diameter changes (Figs. 1l-m show VesSYNC networks during a given 15-minute epoch). As shown for the example in Figs. 1n-o, arteriolar and venous VesSYNC changed substantially during 24 hours (coefficients of variation, CoVs, of 0.20 ± 0.09 and 0.39 ± 0.03, respectively, for median arteriolar and venous VesSYNC, n = 4 mice). The arteriovenous vasomotion transfer gain (GAV), defined as the ratio of linearly related venous to arteriolar vasomotion, was small (Fig. 1p for a representative mouse, 0.14 ± 0.14 in n = 4 mice) and varied substantially (CoV of 1.24 ± 0.85). Collectively these demonstrate how neurosurveillance via cloud-based functional imaging has the potential to usher in a new era of neuroscientific discovery.

Acknowledgements (Optional): This work was supported by NIH grant nos. 2R01CA196701-06A1, 5R01CA237597-04 and and 5R01DE027957-05