The roadmap for estimation of cell-type-specific neuronal activity from non-invasive measurements

Abstract

The computational properties of the human brain arise from an intricate interplay between billions of neurons connected in complex networks. However, our ability to study these networks in healthy human brain is limited by the necessity to use non-invasive technologies. This is in contrast to animal models where a rich, detailed view of cellular-level brain function with cell-type-specific molecular identity has become available due to recent advances in microscopic optical imaging and genetics. Thus, a central challenge facing neuroscience today is leveraging these mechanistic insights from animal studies to accurately draw physiological inferences from non-invasive signals in humans. On the essential path towards this goal is the development of a detailed 'bottom-up' forward model bridging neuronal activity at the level of cell-type-specific populations to non-invasive imaging signals. The general idea is that specific neuronal cell types have identifiable signatures in the way they drive changes in cerebral blood flow, cerebral metabolic rate of O2 (measurable with quantitative functional Magnetic Resonance Imaging), and electrical currents/potentials (measurable with magneto/electroencephalography). This forward model would then provide the 'ground truth' for the development of new tools for tackling the inverse problem-estimation of neuronal activity from multimodal non-invasive imaging data.This article is part of the themed issue 'Interpreting BOLD: a dialogue between cognitive and cellular neuroscience'.

Keywords: BOLD fMRI; CMRO2; cerebral blood flow; magnetoencephalography; neurometabolic; neurovascular.

Figure 1.

Conceptual relationship between physiological parameters and measurements in mice and humans across scales. Advancing our ability to infer microscopic details of underlying physiology in the human brain from non-invasive methods requires parallel data acquisition in animals (e.g. mice) and humans, and a complementary theoretical effort building computational bridges between the varying scales as well as from animals to humans. Animal experiments are necessary to directly measure and manipulate concrete cellular-level physiological parameters. These measurements, which are only available in model organisms, provide the microscopic ‘ground truth’ needed for the development of theoretical models. In parallel, multimodal non-invasive human data are needed to evaluate translation.

Figure 2.

Two-photon measurements of dilation and constriction. (a)(i) Top-down maximal intensity projection (MIP) of a low-magnification image stack showing surface vasculature. (ii) Three-dimensional projection from the area outlined in red on the MIP. (b) Line-scan (i) and frame-scan (ii) diameter measurements. (c) Mapping the centre of neuronal response to the stimulus using surface potential recordings. (d) The ‘centre-surround’ structure of the arteriolar dilation response. Three examples of stimulus-induced arteriolar diameter change (red) at different distances from the centre of evoked neuronal response with overlaid computational fit to the data (black). The inset shows the functions C (positive) and D (negative) used to fit the data. The sum of the two normalized by the maximum of the sum of their absolute values is shown in red in the inset. (e) The surround region with stronger vasoconstriction experiences stronger inhibition. The plot shows the hyperpolarization ratio (the ratio of peak hyperpolarization to peak depolarization) as a function of distance from the centre obtained using voltage-sensitive dye imaging. Data from four animals are superimposed. (f) Sensory stimulus-induced dilation time courses sorted by depth and peak-normalized. Colour-coded depth categories are indicated on top. (g) Onset (black) and time-to-peak (red) of dilation as a function of depth. Each data point represents a single measurement. For each subject, the data were group-averaged according to depth in 100-μm bins. Error bars represent the mean ± s.e. across subjects for each bin (green). (h) Sensory stimulus-induced dilation time courses sorted by the branching order (colour-coded). The data were averaged from cortical depth corresponding to layers II–V (100–700 µm). DA, diving arteriole; BO, branching order.

Figure 3.

Cell-type specificity of neurovascular coupling. (a)(i) The 473 nm laser beam visualized in fluorescent medium, used for OG modulation. (ii) Schematic of the OG beam centred on a diving arteriole. The full width at half maximum (FWHM = 230 µm) of the beam is superimposed on a vascular mean intensity projection image. Red arrows indicate the direction of flow in the arteriole. (b) Simulated spatial profile of the OG beam in cortical tissue. (i) Colour-coded photon density. (ii) Photon density as a function of depth (z-axis). (c) Diameter change time courses of the diving arteriole in (a) in response to a sensory (forepaw) stimulus and selective OG stimulation of inhibitory cortical neurons. Ten sensory stimulus trials were averaged. (d) LFP recorded from layer II/III during the OG stimulation of inhibitory cortical neurons. Each trace shows a single stimulus trial. Downward deflections in the LFP signal indicate spontaneous bursts of PCs activity. This behaviour is suppressed during the OG stimulus. (e) Averaged dilation time courses grouped by depth. The inset shows an expanded view of the initial 4 s after the stimulus onset. The depth in micrometres is indicated on the left. Error bars indicate s.e. across subjects. (f) Comparison of dilation time courses in response to OG stimulation of cortical inhibitory neurons before (black) and after (red) blocking Y1 receptors for NPY. Error bars represent s.e. across subjects. (g) As in (f) for the sensory response in the surround area. (h) Comparison of the response to OG stimulation of the inhibitory neurons (also shown in f) to OG stimulation of PCs. Blocker of glutamatergic synaptic transduction were present in the PC experiment to ensure specificity.

Figure 4.

Astrocytic Ca⁺⁺ response and vasodilation. (a) Representative field of view (FOV) including a perivascular astrocyte labelled with sulforhodamine 101 (SR101) and Ca⁺⁺ indicator Oregon Green 488 BAPTA-1 AM (OGB1), a diving arteriole labelled by intravascular injection of fluorescein isothiocyanate (FITC)-labelled dextran, and a number of neuronal cell bodies labelled with OGB1, imaged 150 µm below the cortical surface. ROIs used for extraction of time courses are shown on the right. Scale bars are 10 µm. (b) Time courses extracted from the ROIs shown in (a). Astrocytic (red) and neuronal (green) Ca⁺⁺ signal changes are expressed as ΔF/F. Vasodilation is expressed as per cent diameter change relative to the baseline diameter, Δd/d. The diameter change was extracted from the expansion of FITC-labelled intravascular lumen, indicated by ‘v’ in (a). The black bars indicate stimulus duration. (c) ‘The best’ example of astrocytic Ca⁺⁺ response to multiple consecutive stimulus trials. Left: FOV imaged 190 µm below the cortical surface. Middle: ROIs used to extract time courses. Astrocytic Ca⁺⁺ signal time courses. Black bars indicate stimulus duration. Scale bars are 20 µm. (d) An example FOV including two astrocytes, one (a1) with a connected endfoot (EF), imaged 260 μm below the cortical surface. The EF and a1 contours are overlaid. Each image was computed as an average of 10 consecutive ratio frames. The corresponding time windows (in seconds relative to the stimulus onset) are indicated above the images. Scale bar, 10 µm. (e) Time-courses of astrocytic Ca⁺⁺ change (top) and arteriolar diameter change in control (middle) and in inositol 1,4,5-triphosphate IP₃ type-2 receptor knock-out (IP₃R2-KO) mice, in which the primary mechanism of astrocytic Ca⁺⁺ increase—the release of Ca⁺⁺ from intracellular stores following activation of an IP3-dependent pathway—is lacking (bottom). An average is superimposed on each panel (thick lines). The stimulus onset is indicated by the grey vertical line. Peak-normalized averaged control arteriolar diameter change (black) and astrocytic Ca⁺⁺ response (red) are superimposed in the inset.

Figure 5.

Two-photon imaging of pO₂ and estimation of CMRO₂. (a) A reference vascular image with an arteriole (labelled by a magenta cross). The fluorescent contrast is due to intravascular FITC. Measured pO₂ values are superimposed. The red contour indicates the segmented arteriolar territory. (b) The pO₂ values from (a) plotted as a function of the radial distance from the arteriole. (c,d) Stimulus-evoked time courses of pO₂ change (d) extracted from each of the measurement points in (c). The thick line shows the average. (e) Intravascular pO₂ measurements overlaid over a reconstructed microvascular network connecting a diving arteriole and a neighbouring surfacing venule. (f) pO₂ distribution measured inside diving arterioles. (i) MIP of a two-photon image stack. Scale bar, 50 µm. (ii) PO₂ map inside the arteriole (labelled by the red arrow in the image on top) 100 µm below the cortical surface. Scale bar, 20 µm. (iii) Radial intra-arteriolar pO₂ profiles (radial distance calculated from the vessel axis to the vessel wall) from four diving arterioles similar to the example vessel presented on the left. (g) Schematic of the Krogh model parameters.

Figure 6.

Bottom-up modelling of the haemodynamic response. (a) Reconstructed microvascular network with segmented arterioles, capillaries and venules. (b) Simulated CBF. (c) Simulated (i) and experimental (ii) measurements of pO₂. (d) Experimentally obtained dilation time courses averaged across subjects sorted by the cortical depth and branching order. (e) Simulation of nuclear spins diffusing in the magnetic field perturbation volume. (f) Time series of the simulated extravascular signals obtained with the gradient echo (GRE) and spin echo (SE) fMRI pulse sequences. The spatial gradients applied to simulate GRE and SE signals are shown in the inset.

Figure 7.

Synaptic projection domains in the generation of LFP and CSD. (a,b) Predicted LFP and CSD profiles for a population of layer 5 PC due to synaptic inputs. The vertical axis shows the depth profile as a function of the vertical position of synaptic inputs (the horizontal axis). The profiles are analogous to the plots in fig. 13a in [142], but here the cortical surface is assumed to be covered with highly conductive saline. (c) Inhibitory projection domains of different types of inhibitory neurons to layer V PC. The scatter plot showing the horizontal and vertical distance of the synapses made by seven groups of layer II/III inhibitory neurons from the soma of layer V PC is reproduced with permission from [143]. The plot is overlaid on an example layer V PC morphology obtained from NeuroMorpho.org [144] (ID = 070123-z1). The indicated correlation between morphological classification of the inhibitory neurons and expression of neuropeptides and NO is based on the authors' interpretation of [,,–149].

Figure 8.

The current dipole moment in mice and humans. (a,b) Human MEG measurements in response to the median nerve stimulation: cortically constrained source estimate [10] (current dipole) for the SI ROI (a) and the corresponding SEF contour plot (b). (c,d) Rat dipole waveform (c) calculated from laminar LFP (d) in response to weak electrical forepaw stimulation. Contra, contralateral stimulus; ipsi, ipsilateral stimulus. (e) Schematic of the current dipole and surface potential signals due to excitatory (S_e) and inhibitory (S_i) inputs.