Two-photon Light Field with Neuro-active Sensing for Fast Volumetric Neural Microcircuit Readout

Underlying our every sensation, thought, memory, decision and action are 100 billion neurons communicating through trillions of electrical impulses each second. Over the past century, neuroscientists have explored brain function on primarily two scales, that of single neurons (i.e., by impaling them with electrodes) and that of entire brain regions (i.e. with electroencephalogram, EEG, and functional magnetic resonance imaging, fMRI). However, between these two scales lies a large knowledge gap surrounding how neurons interact in networks to process and store information, form memories and generate actions.

Over the past 10 years, geneticists have developed methods to control and read out brain cell activity with light. They can render neurons sensitive to light to activate or silence them when illuminated with certain wavelengths. In addition, neurons can be made to "glow" or become more fluorescent when active. These "optogenetic" tools make it possible to connect single-neuron properties (i.e., through electrode studies) with functions evolving on the population level (through fMRI and EEG). To achieve this, optical engineers must first overcome a key challenge: the mammalian brain severely scatters and distorts light, resulting in blurry images and thus confusion about which neuron is active. Here we propose to overcome this limitation by utilizing the "optogenetic" ability to activate individual neurons with light in rapid succession. Specifically, we will activate each neuron throughout a brain volume in turn to determine each one's "signature"; that is, the blurry, distorted light pattern it generates when active. We will then use this collection of activity signatures to rapidly and precisely determine which neuron activates and when during subsequent spontaneous activity. We will implement this "collection" approach with a three-dimensional (3D) imaging strategy called "light field." While traditional imaging captures focused images for objects lying in a single plane, "light field" captures perspectives from different angles within a single shapshot. The "light field" approach thus enables us to track neuronal activity simultaneously throughout a volume a brain tissue rather than within a single plane. This novel combination of "light field" imaging with active sensing will significantly increase the speed (10-fold) with which we can track the activity of single neurons throughout a volume. In the near future, development of faster, more sensitive cameras and sensors could increase our instrument's volume capture rates to 100-fold compared to the current state-of-the-art. Moreover, here we will, for the first time, implement "light field imaging" in "two-photon" mode. "Two-photon" is a method to excite fluorescence that is used widely in biomedical research. In contrast to the blue/green wavelengths previously used with "light field," "two-photon" utilizes near-infrared wavelengths that are far less scattered than blue and green, enabling researchers to image deep in scattering tissues. Our new two-photon light field instrument will decrease distortion and thus enable us to image deeper into the brain.

By combining targeted neural activation with 3D light-field imaging, we will overcome a key barrier to understanding how neurons interact in networks. With our new instrument, neuroscientists will at last be able to collect data on how neurons work together to process and store information, make decisions and effectuate actions. A detailed understanding of these network-level processes will inform the design of new therapies for neuronal diseases and disorders, such as Alzheimer's, in which these functions are compromised.

Our goal is to increase the temporal resolution and throughput of neuronal functional fluorescence readout through a novel light field acquisition instrument and analysis strategy for highly scattering mammalian brain tissue. Light field microscopy captures both the angle and position of light emitted from a volume, enabling reconstruction of multiple perspectives and planes from single data frames. Light field enables three dimensional reconstruction of fluorescent objects without scanning, and is thus a promising technology for high speed volumetric fluorescence monitoring as demonstrated by Prevedel et al. (Nature Methods, 2014) in transparent larval zebrafish and C. elegans. Mammalian brain tissue, however, is highly scattering, precluding 3D reconstruction through standard light field algorithms. Here we propose two novel methods to mitigate scattering effects. First, we will excite neuronal fluorescence in two-photon mode using near-infrared wavelengths less scattered than the visible wavelengths used in one-photon fluorescence light field microscopes. Second, we will optogenetically evoke spiking and fluorescence transients in each neuron throughout the volume in rapid succession, recording each one's blurry, distorted light field signature. We will use this collection of signatures to segment and map subsequent spontaneous activity back onto high spatial resolution images of these neurons. Our "neuro-active sensing" acquisition and analysis approach will enable fast light field processing and will leverage signal that would otherwise be lost to scattering. We estimate that our instrument will increase the speed of mammalian live network volumetric imaging by a factor of 10 to 100, thus advancing one of Neuroscience's long-standing goals: to observe and manipulate living mammalian neural microcircuits in real time and in closed-loop.

1. Economic: Our project will lead to the commercial dissemination of a turn-key microscopy and analysis system. Prime commercial partners for this include two UK companies Scientica Ltd. and Cairn Research Ltd. who have well established neurophysiology client bases (10k+ laboratories worldwide). We will ensure that the commercial design features the modularity and alignment tools necessary to simplify upgrades with future advancements in cameras and lasers.

2. Health: A long-term impact of our proposal is to improve basic understanding of neural network function critical to identifying novel, effective theraputic strategies. Neurological diseases such as dementia, epilepsy, stroke and mental health disorders cost the UK 112 billion GBP per year, accounting for medical costs, loss of productivity and forced early retirement. Treatment for diseases impacting cognition, memory, learning and motor control advance slowly and remain primitive due to a fundamental lack of understanding of how these system function. Our technology will fill critical data-gaps in animal models at the network level necessary to understand how these systems function in healthy mammals as well as those suffering from neurological disease.

3. Training the UK's cross-disciplinary technical base: A key element to the project's success will be the training and contribution of the two PDRAs who will work full time to develop, test, and refine the two-photon light field system and rapid analysis strategies. These PDRAs will receive specialized training in communicating and synthesizing effectively across signal processing, photonics, neurophysiology, and computation neuroscience disciplines. This rare cross-disclinary training will well prepare these researchers for neurotechnology development careers both in academic and industrial contexts.

4. Public enthusiasm for STEM approaches to biological research: Our hands-on exhibit at the annual Imperial College Festival and other outreach venues will generate enthusiasm for STEM approaches to brain research and encourage students to consider training in these domains. The exhibit will introduce the basic concepts of neuronal communication (action potentials and synapses) as well as the technical challenges of observing neurons as they interact in large networks (size, scattering, contrast). Key optical concepts such as diffraction, scattering, holography, and light-field photography will be demonstrated. We will show how we have applied these concepts to improve network-level neuronal imaging with videos generated from experiments.