Adaptive optics in two-photon fluorescence microscopy for functional imaging in neuroscience in mice
Dirigée par Laurent Bourdieu (IBENS) et Alexandra Fragola (ISMO).
Two-photon fluorescence microscopy (2PFM) is a reference technique for in vivo recording of neuronal activity in the murine cortex. However, refractive index inhomogeneities in biological tissues generate optical aberrations that distort the wavefront, broaden the PSF and degrade the signal. These aberrations reduce the SNR of calcium transients and create artificial correlations between neighbouring neurons through neuropil contamination. Direct adaptive optics (AO), based on a Shack-Hartmann wavefront sensor (SHWS) and a fluorescent guide star in a closed loop, enables faster and more precise measurement and correction of these aberrations than indirect AO. Its main limitation is the rapid degradation of the SHWS signal-to-background ratio (SBR) with depth, compounded by the lack of guide stars that are simultaneously bright, photostable and biocompatible.
This thesis proposes the use of intravenous CuInSe2/ZnS quantum dots (QDs) as a guide star for direct AO in 2PFM in vivo in mice. A 2P microscope (920 nm laser, 20× NA 1.0 objective) was adapted for in vivo imaging in anaesthetized mice, with a closed-loop AO system (deformable mirror MIRAO 52E, 52 actuators; SHWS HASO). The QDs emit at 700 nm under 920 nm excitation, are spectrally separated from GFP and directed to the SHWS via a descan architecture. Three successive correction levels are defined: System AO (optical system aberrations), Surface AO (cranial window and brain surface) and Full AO (complete correction at the imaging depth of interest).
Before using the QDs as chronic in vivo probes, their pharmacokinetics and toxicity were characterised. The blood circulation half-life of CuInSe2/ZnS QDs (~10 nm, coated with an imidazole-sulfobetaine block copolymer) is 17 h, seven times longer than that of TRITC-dextran (2.4 h), enabling a full day of experiments with a single injection. Toxicological evaluation (body weight, organ weights, leukocyte counts, biodistribution by ICP-MS/MS) revealed no overt inflammatory or toxic effects, validating their biocompatibility for chronic brain imaging.
In GAD65-EGFP mice, Full AO correction markedly improves signal intensity across different structure types and resolution up to 550 µm, with a more pronounced gain on dendrites than on cell bodies. Spatial frequency analysis confirms the increase of high spatial frequency spectral energy between System AO and Full AO. A depth-dependent signal to background ratios (SBR) analysis of the SHWS spot patterns, compared with the data of Liu et al. (2019), indicates that our sub-aperture selection algorithm discards too many peripheral sub-apertures, representing a clear direction for future optimization.
The most significant contribution concerns functional imaging. In mice expressing GCaMP8m, two 8-min recordings (System AO vs Full AO) were compared at 600 µm. Across 23 ROIs, AO produces an increase in the raw fluorescence of ROIs F0 and in the skewness of calcium transients ΔF/F distributions, a reduction in noise, and a reduction in R² correlations, reflecting the decrease in neuropil contamination.
This thesis represents, to our knowledge, the first demonstration of the use of intravenous QDs as a guide star for direct AO in 2PFM in vivo, with demonstrated gains in structural (≤ 550 µm) and functional (≤ 600 µm) imaging. Perspectives include the use of QDs emitting at 1200 nm to extend the accessible depth, an interleaved System AO/Full AO acquisition scheme to isolate the pure effect of AO independently of calcium transients dynamics, and resonant scanning to improve temporal resolution.