Decision letter: Retinal microvascular and neuronal pathologies probed in vivo by adaptive optical two-photon fluorescence microscopy

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract The retina, behind the transparent optics of the eye, is the only neural tissue whose physiology and pathology can be non-invasively probed by optical microscopy. The aberrations intrinsic to the mouse eye, however, prevent high-resolution investigation of retinal structure and function in vivo. Optimizing the design of a two-photon fluorescence microscope (2PFM) and sample preparation procedure, we found that adaptive optics (AO), by measuring and correcting ocular aberrations, is essential for resolving putative synaptic structures and achieving three-dimensional cellular resolution in the mouse retina in vivo. Applying AO-2PFM to longitudinal retinal imaging in transgenic models of retinal pathology, we characterized microvascular lesions with sub-capillary details in a proliferative vascular retinopathy model, and found Lidocaine to effectively suppress retinal ganglion cell hyperactivity in a retinal degeneration model. Tracking structural and functional changes at high-resolution longitudinally, AO-2PFM enables microscopic investigations of retinal pathology and pharmacology for disease diagnosis and treatment in vivo. Editor's evaluation The authors developed a two-photon fluorescence microscope coupled with adaptive optics (AO-2PFM), allowing in vivo imaging of the mouse retinal structure and function. This new imaging system will be important for exploring normal retinal physiology and pathological alterations in retinal disease models. https://doi.org/10.7554/eLife.84853.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Retina is a layered tissue in the back of the eye that transduces light into electrochemical signals to be further processed by the brain for visual perception and cognition (Kandel et al., 2021). As one of the most energy-demanding tissues, the retina is metabolically sustained by an intricate vasculature with several laminar plexuses (Selvam et al., 2018). Vascular and neuronal abnormalities in the retina are associated with both ocular (Schachat et al., 2017) and systemic diseases (Cheung et al., 2017; London et al., 2013; Lechner et al., 2017), underscoring the importance of studying retinal pathology and pharmacology. With well-developed genetics and similar physiology to the human retina, mouse models have been widely utilized for mechanistical studies of retinal diseases. Behind highly transparent mouse eye optics (i.e. cornea and crystalline lens), the retina is uniquely accessible to light and the only part of the nervous system that can be probed non-invasively by optical imaging. Recent advances in mouse genetics have enabled fluorescence microscopy investigations of vasculature (Ivanova et al., 2021) as well as neurons and glial cells (Jo et al., 2018; Martersteck et al., 2017; Eme-Scolan and Dando, 2020) of the mouse retina. Among fluorescence microscopy techniques, two-photon fluorescence microscopy (2PFM) (Denk et al., 1990) utilizing near-infrared (NIR) excitation is particularly suited for retinal imaging. Its intrinsic optical sectioning capability permits depth-resolved three-dimensional (3D) imaging throughout the retina. With the retinal photoreceptors minimally responsive to NIR light, 2PFM is also an ideal tool for functional studies of retina (Euler et al., 2002; Baden et al., 2016). However, as a far-from-perfect imaging system, the optics of the mouse eye introduce severe aberrations to the NIR excitation light, preventing high-resolution visualization of subcellular features in vivo. As a result, the vast majority of microscopy studies have been carried out ex vivo on dissected retinas, preventing longitudinal investigations of retinal pathology under physiological conditions. Adaptive optics (AO) is a collection of technologies that actively measure and correct for optical aberrations (Hampson et al., 2021), and has been applied to optical microscopy for high-resolution imaging of neural tissues (Ji, 2017; Rodríguez and Ji, 2018). It has also been combined with ophthalmological imaging modalities to restore diffraction-limited imaging performance for the human retina (Liang et al., 1997; Akyol et al., 2021). Because of the severe aberrations of the mouse eye, AO has also been applied to in vivo imaging of the mouse retina (Palczewska et al., 2014; Biss et al., 2007; Alt et al., 2010; Geng et al., 2012; Sharma et al., 2013; Wahl et al., 2016; Wahl et al., 2019; Qin et al., 2020). However, there are disagreements in the reported spatial resolutions (Palczewska et al., 2014; Biss et al., 2007; Alt et al., 2010; Geng et al., 2012; Sharma et al., 2013; Wahl et al., 2016; Wahl et al., 2019; Qin et al., 2020), characteristics and magnitude of aberration (Palczewska et al., 2014; Biss et al., 2007; Alt et al., 2010; Wahl et al., 2016; Wahl et al., 2019; Qin et al., 2020), and the effectiveness of AO (Biss et al., 2007; Alt et al., 2010; Wahl et al., 2016; Wahl et al., 2019; Qin et al., 2020). For example, whereas previous papers reported cellular resolution without AO, a recent AO-2PFM study (Qin et al., 2020) reported extremely large aberrations in the mouse eye and found AO to be required in order to resolve microvasculature and cell bodies in 2D in vivo. These discrepancies have led to uncertainty over the imaging performance achievable with conventional 2PFM and the necessity of AO for microvascular and cellular investigations of retinal physiology. Together with a lack of detailed imaging protocols, they have prevented the routine application of AO-2PFM to disease diagnosis and therapeutic intervention in the retina of mouse models of ocular, cerebral, and systemic diseases. The aims of this work are to provide a resource for in vivo retinal imaging using 2PFM, by optimizing the design of a 2PFM for in vivo imaging of the mouse retina, characterizing mouse ocular aberration, developing a guideline for adaptive optical 2PFM (AO-2PFM) imaging, and demonstrating its applications to retinal pathology and pharmacology. Using a carefully engineered 2PFM and following an optimized sample preparation procedure, we were able to achieve two-dimensional (2D) cellular resolution imaging performance in the mouse retina without AO. For synaptic, subcellular, and three-dimensional (3D) cellular resolution imaging of the mouse retina, AO was essential in improving image brightness, contrast, and resolution. Testing the performance of AO-2PFM in various transgenic mouse lines, we proposed strategies to maximize its impact on image quality improvement. We extended the application of AO-2PFM to mouse retinal pathology and pharmacology by imaging the retinas of two transgenic models with proliferative vascular retinopathy and retinal degeneration, respectively. In our model of proliferative vascular retinopathy, AO enabled us to, for the first time, characterize retinal vascular lesions with sub-capillary details over multiple days and observe cell migration in vivo. In our model of retinal degeneration, AO allowed high-fidelity interrogation of pharmacologically modified hyperactivity of retinal ganglion cells (RGCs), indicating AO-2PFM as a promising tool evaluating retinal pharmacology in vivo. Together, by systematically optimizing and applying AO-2PFM to in vivo mouse retinal imaging, our work represents an important advancement in enabling high-resolution longitudinal studies of retinal pathology and pharmacology for disease diagnosis and treatment. Results Optimized AO-2PFM for in vivo mouse retinal imaging A home-built two-photon fluorescence microscope equipped with a segmented deformable mirror (DM) and a Shack-Hartmann (SH) sensor (Li et al., 2020b) was modified for in vivo mouse retinal imaging by replacing the objective lens with an add-on eye imaging module (Qin et al., 2020; Grulkowski et al., 2018; Figure 1A, Materials and methods). The module consisted of an electrically tunable lens (ETL) whose adaptive surface was conjugated to the DM, a turning mirror, and two lens groups (L7 and L8) that relayed the adaptive surface of the ETL to the pupil of the mouse eye. With this design, the optics of the mouse eye focused 920 nm light onto the retina to excite fluorescent markers and collected the emitted fluorescence for detection. The ETL allowed us to adjust the focal plane in the mouse eye without translating the mouse (Jian et al., 2013) or optics (McNabb et al., 2019) in the imaging system. For all experiments, system aberrations in the two-photon illumination path were measured with a modal AO method and corrected before image acquisition (Materials and methods; ‘No AO’ images: system aberration correction only). Figure 1 with 3 supplements see all Download asset Open asset AO-2PFM for diffraction-limited imaging of the mouse retina in vivo. (A) Schematics of AO-2PFM. Inset 1: direct wavefront measurement by a Shack-Hartmann (SH) sensor composed of a lenslet array and a camera. Inset 2: wavefront correction with a deformable mirror composed of 163 segments with piston, tip, and tilt controls. Grey dashed box: eye imaging module. Bottom: 3D assembly of eye imaging module. L, lens; D, dichroic mirror; DM, deformable mirror; PMT, photomultiplier tube; ETL, electrically tunable lens. (B) Maximum intensity projections (MIPs) of image stacks (72×72×25 µm3) of RGC axons measured without and with AO, respectively, normalized to AO image. Insets: kXkY spatial frequency representation of the images and corrective wavefront. (C) MIPs of image stacks (132×97×32 µm3) of fine RGC processes measured without and with AO, respectively, normalized to AO image. ‘No AO’ image brightness artificially increased by 10.6× for better visualization. White arrowheads: putative synaptic structures. Inset: corrective wavefront. Bottom: i: lateral signal profiles along white dashed line i; ii: axial signal profiles of process ii (white arrow). Signals in the line profiles were normalized to the maximal value of the AO condition. Representative data from >3 experiments (technical replicates). Figure 1—source data 1 Source image stacks of retinal axons (Figure 1B). ‘1-NoAO_8fAvg_stack.tif’ (No AO) and ‘2-AO_8fAvg_stack.tif’ (AO) Figure 1—source data 2 Source image stacks of retinal dendrites (Figure 1C): ‘1-NoAO_8fAvg_160_100_stack.tif’ (No AO) and ‘2-AO_8fAvg_160_100_stack.tif’ (AO). https://cdn.elifesciences.org/articles/84853/elife-84853-fig1-data1-v1.zip Download elife-84853-fig1-data1-v1.zip Figure 1—source data 2 Source image stacks of retinal neuronal processes (Figure 1C). https://cdn.elifesciences.org/articles/84853/elife-84853-fig1-data2-v1.zip Download elife-84853-fig1-data2-v1.zip To ensure optimal performance, we thoroughly characterized our AO-2PFM. We investigated how ETL current and mouse eye placement (with a longitudinal displacement of up to 4 mm in typical experiments) impacted imaging performance (Figure 1—figure supplement 1). We found that aberrations introduced by the ETL at different control currents minimally affected image quality and that axial focal shift varied linearly with ETL current while field-of-view (FOV) size remained mostly constant. We also optimized sample preparation procedure. We discovered that a custom-designed 0-diopter contact lens (CL; design parameters in Figure 1—figure supplement 2A) in combination with a single application of eye gel between the CL and the cornea reduced aberrations, prevented cataract formation, and improved wavefront sensing and imaging for hours (Figure 1—figure supplement 2). In order to achieve diffraction-limited imaging of the mouse retina in vivo, we measured and corrected ocular aberrations with a direct wavefront sensing method (Wang et al., 2014b; Wang et al., 2015), utilizing the SH sensor for wavefront measurement and the DM for wavefront correction (Figure 1A). Briefly, a 3D-localized fluorescence ‘guide star’ was formed in the retina via two-photon excitation and scanned over a user-defined 2D area with galvanometer scanning mirrors. The emitted fluorescence was collected and, after being descanned by the same pair of scanning mirrors, directed to the SH sensor. The now stationary fluorescence wavefront was segmented by a lenslet array and focused onto a camera, forming an SH image composed of an array of foci (Figure 1A, inset 1). Local phase slopes of wavefront segments were calculated from the displacements of the foci from those taken without aberrations. Assuming spatially continuous aberrations, we computationally reconstructed the wavefront from the phase slopes (Panagopoulou and Neal, 2005). We then applied a corrective wavefront, opposite to the measured aberrations, to the DM by controlling the tip, tilt, and piston of each segment (Figure 1A, inset 2; Figure 1—figure supplement 3) so that mouse ocular aberrations could be canceled out, ensuring diffraction-limited focusing of the two-photon excitation light on the mouse retina. All in vivo imaging experiments were conducted in anesthetized mice with dilated pupil (Materials and methods). In most experiments, an area of 19×19 µm2 of the retina was scanned for 3–10 s for wavefront sensing. To estimate the spatial resolution of our AO-2PFM for in vivo mouse retinal imaging, we imaged Thy1-GFP line M transgenic mice that had green fluorescent protein (GFP) expressed in a subset of RGCs (Feng et al., 2000). The image taken without AO showed dim and distorted RGC axons; after aberration correction, we achieved an 8.6× increase in signal and proper visualization of the fine RGC axons (Figure 1B). The spatial frequency space representations of the images indicated that AO enhanced the ability of the imaging system to acquire higher resolution information and led to a lateral resolution that was better than ~0.8 µm (Figure 1B, insets). For some thin RGC processes (Figure 1C), restoring diffraction-limited resolution led to an increase in signal (by 10.6×) and contrast (Figure 1C, i), and, for the first time, enabled in vivo 2PFM visualization of varicosities resembling synaptic structures in the mouse retina (Figure 1C, white arrowheads). From the axial profile of a thin process (Figure 1C, ii), we estimated the axial resolution after AO correction to be 6.7 µm. Both the lateral and axial resolution estimations were close to the theoretical diffraction-limited resolution for a fully-dilated mouse eye with 0.49 numerical aperture (Geng et al., 2011). AO improves in vivo imaging of retinal vasculature Retinal vasculature supports the physiological functions of the retina. Retinal vascular diseases can lead to vision loss. Abnormalities in retinal vasculature morphology and physiology serve as important biomarkers for various cerebral and systemic diseases (Patton et al., 2005; Frost et al., 2013; Ikram et al., 2013; Liew et al., 2008). Therefore, in vivo characterization of retinal vasculature, especially at the microvasculature level, is of great physiological and clinical importance. Utilizing either confocal microscopy (Biss et al., 2007; Wahl et al., 2019) or 2PFM (Qin et al., 2020; Bar-Noam et al., 2016; Wang et al., 2021), previous publications have achieved in vivo visualization of retinal microvasculature through either full correction of the mouse eye aberrations (Biss et al., 2007; Wahl et al., 2019; Qin et al., 2020), partial correction of the anterior optics of the mouse eye (Wang et al., 2021), or stringent selection of imaging lenses (Bar-Noam et al., 2016). These prior demonstration-of-principle experiments suggest that in order to image retinal microvasculature in vivo, mouse eye aberrations need to be corrected, either fully or partially. With our optimized imaging system, we aimed to determine whether aberration correction was indeed essential for visualizing microvasculature. Furthermore, we proceeded to systematically characterize the spatial dependence of mouse eye aberrations and how large a FOV can benefit from a single AO correction. To verify the necessity of AO in resolving mouse retinal microvasculature and characterize mouse eye induced aberrations, we performed in vivo 2PFM angiography by retro-orbitally injecting dextran-conjugated fluorescein isothiocyanate (FITC) into the non-imaged eye. Aberrations were measured with fluorescence emitted from vessels in the superficial plexus (red asterisk, Figure 2A; wavefront sensing area: 19×19 µm2). After AO correction, we observed a 2–10× enhancement in signal (Figure 2B and C). Comparing the line signal profiles (along the orange dashed lines, Figure 2A and B), we found that AO improved signal for all vessels while its impact on signal of smaller capillaries (Figure 2C, black asterisks; 6–10× improvement) was more substantial than on larger vessels (Figure 2C, black circles; 2–3× improvement). Despite the substantial signal improvements enabled by AO, we found that most capillaries, due to their size and sparse distribution in space, could be resolved in 3D without AO by our optimized 2PFM, albeit at reduced contrast and resolution (Figure 2D and E). Our results indicate that a properly designed 2PFM is capable of acquiring retinal angiograms at the level of individual capillaries. Figure 2 with 1 supplement see all Download asset Open asset In vivo imaging of mouse retinal vasculature with AO-2PFM. (A,B) MIPs of image stacks (580×580×128 µm3) of vasculature measured (A) without and (B) with AO, respectively, normalized to AO image. Red asterisk: center of 19×19 µm2 wavefront sensing (WS) area. Gamma correction: 0.7. Representative data from >25 experiments (technical replicates). (C) Lateral line profiles along orange dashed lines in A and B. Black circles: large vessels; black asterisks: capillaries. (D) Single image planes at 0, 23, 40, and 53 µm below the superficial vascular plexus acquired without and with AO correction performed at the superficial plexus (0 µm), normalized to AO images. (E) Axial profiles of capillary structures (i-iv in D). Red dashed lines: depth of wavefront sensing area. (F) Left: MIPs of image stacks (580×580×110 µm3) acquired with WS performed at different locations in the FOV (red asterisks). Middle: AO/No AO pixel ratio maps. Right: radially averaged profiles of pixel ratio maps, centered at WS sites. Insets: corrective wavefronts. MIPs and pixel ratio maps individually normalized. Figure 2—source data 1 Source image stacks of retinal vasculature (Figure 2A, B and D). ‘1-NoAO_8fAvg_stack.tif’ (No AO) and ‘2-AO_8fAvg_stack.tif’. https://cdn.elifesciences.org/articles/84853/elife-84853-fig2-data1-v1.zip Download elife-84853-fig2-data1-v1.zip Figure 2—source data 2 Source image stacks with AO measured at different locations (Figure 2F). ‘AO [1]’ stack: ‘1-AO[1]_Location_0_0_stack-9fAVG.tif’, ‘AO [2]’ stack: ‘2-AO[2]_Location_N15_N18_stack-9fAVG.tif’ ‘AO [3]’ stack: ‘3-AO[3]_Location_P4_P25_stack-9fAVG.tif’. https://cdn.elifesciences.org/articles/84853/elife-84853-fig2-data2-v1.zip Download elife-84853-fig2-data2-v1.zip We further evaluated how the mouse ocular aberrations varied with imaging depth and field position. We found that AO performed at the superficial plexus was beneficial for imaging deeper layers, with the correction at superficial depth improving signal, resolution, and contrast of deeper vasculature (Figure 2D and E). This result indicated that most aberrations of the mouse eye arose from cornea and crystalline lens, instead of retina. Because the crystalline lens of the mouse eye has a gradient refractive index distribution (Campbell and Hughes, 1981; Remtulla and Hallett, 1985), ocular aberrations should also be field dependent (Wang and Ji, 2012; Wang and Ji, 2013). Field-dependent aberrations might also be introduced when the mouse eye was positioned off-axis with respect to the eye imaging module. We therefore examined how aberrations varied with FOV position and characterized the area within which a single correction led to substantial signal improvement. We performed AO at different locations of the superficial plexus in the FOV (Figure 2F, left column, red asterisks; Figure 2—video 1) and compared their performance. The ‘AO/No AO’ pixel ratio maps (Figure 2F, middle column) exhibited field-dependent signal increase with larger gain achieved at pixels closer to the locations of aberration measurements. We quantified the effective area of AO in terms of signal improvement by calculating the radially averaged profiles of these pixel ratio maps (Figure 2F, right column; origins at the wavefront sensing locations). We found signal improvement (‘AO/No AO’ pixel ratio ≥1) within a radius of ~216 µm when AO was performed at the FOV center of this mouse (Figure 2F, [1]). For off-center locations, this radius was slightly smaller (Figure 2F, [2] and [3]). AO enables 3D cellular resolution imaging of neurons in the mouse retina The mouse retina consists of multiple layers of neurons with different cell types and distinct physiological properties. In the early stage of retinal diseases, abnormal morphology and function are usually confined to specific cell types within a single layer (Hoon et al., 2014). Therefore, for microscopic investigations of retinal physiology and pathology, it is essential to resolve cells in 3D. We evaluated whether our optimized 2PFM was capable of 3D cellular resolution imaging without correcting the severe aberrations of the mouse eye. For this purpose, we imaged the densely fluorescent Thy1-YFP-16 mouse retina in vivo, where all bipolar cells, amacrine cells, and retinal ganglion cells were labeled with yellow fluorescence protein (Feng et al., 2000) (YFP). A single AO correction acquired by scanning a 19×19 µm2 area centered on the red asterisk in (Figure 3A) substantially improved signal and resolution (Figure 3A and B; Figure 3—video 1). 2D Fourier transforms of these retinal images indicated that AO recovered higher spatial frequency information (i.e. farther away from the center of Figure 3C and D) thus improved both lateral and axial resolution. The resolution enhancement was especially striking along the axial direction, allowing retinal layers to be more clearly differentiated by better resolving neurons at different depths (Figure 3A and B, XZ images). This improvement in axial resolution is especially important for functional imaging, because it minimizes neuropil contamination and ensures accurate characterization of the functional properties of neurons (Ji et al., 2012; Wang et al., 2014a; Sun et al., 2016). Therefore, AO was necessary for 3D cellular resolution imaging of retinal neurons in vivo. In the lateral image planes, our optimized 2PFM design and mouse preparation allowed the identification of individual neurons without AO, albeit at lower signal and poorer resolution than those achieved with AO, for inner nuclear layer, inner plexiform layer, and ganglion cell layer (Figure 3E). In contrast, subcellular processes could not be clearly visualized without aberration correction (e.g. processes in the inner plexiform layer, Figure 3E, white boxes in the middle column; more examples in Figure 3—figure supplement 1). Figure 3 with 2 supplements see all Download asset Open asset In vivo imaging of mouse retinal neurons with AO-2PFM. (A,B) MIPs of image stacks (580×580×80 µm3) of a Thy1-YFP-16 retina, measured (A) without and (B) with AO, respectively, normalized to AO images. Red asterisk: center of a 19×19 µm2 WS area. Top: lateral (XY) MIPs. Bottom: axial (XZ) MIPs; ‘No AO’ image brightness artificially increased by 2.9× for visualization. Representative data from>10 experiments (technical replicates). (C,D) kXkY and kXkZ spatial frequency space representation of images in (A,B). (E) Images of different retinal layers within the red dashed box in A acquired (top) without and (bottom) with AO, respectively, normalized to AO images. INL: inner nuclear layer; IPL: inner plexiform layer; GCL: ganglion cell layer. INL/GCL: MIPs of 4.9/7.8-µm-thick image stacks; IPL: single image plane. ‘No AO’ image brightness artificially increased for visualization (gains shown in each image). White boxes: zoomed-in views. (F) Single image planes in GCL at FOV edge (blue dashed box in A) acquired (top) without AO, (middle) with central AO (WS area centered at red asterisk in A), and (bottom) with local AO (WS area centered at blue asterisk in A), respectively. Images normalized to local AO image. ‘No AO’ image brightness artificially increased by 2.5× for visualization. (G) AO/No AO pixel ratio map. (H) Radially averaged profile of pixel ratio map, centered at red asterisk in A. Figure 3—source data 1 Source image stacks of retinal neurons (Figure 3A and B). ‘1-NoAO_60_84_8fAvg.tif’ (No AO) and ‘2-AO_60_84_8fAvg.tif’ (AO). https://cdn.elifesciences.org/articles/84853/elife-84853-fig3-data1-v1.zip Download elife-84853-fig3-data1-v1.zip Figure 3—source data 2 Source image stacks of retinal neurons (Figure 3E). ‘1-NoAO_8fAvg_stack.tif’ (No AO) and ‘2-AO_8fAvg_stack.tif’ (AO). https://cdn.elifesciences.org/articles/84853/elife-84853-fig3-data2-v1.zip Download elife-84853-fig3-data2-v1.zip Figure 3—source data 3 Source image stacks of retinal neurons (Figure 3F). ‘1-central_AO_20_28_8fAvg.tif’ (Central AO) and ‘2-local_AO_20_28_8fAvg.tif’ (Local AO). https://cdn.elifesciences.org/articles/84853/elife-84853-fig3-data3-v1.zip Download elife-84853-fig3-data3-v1.zip Similar to our vascular imaging results, the Thy1-YFP-16 mouse eye exhibited field-dependent aberrations. For areas away from the AO measurement location (e.g. blue dashed box in Figure 3A), although resolution improvement remained, the correction acquired at the FOV center (Figure 3F, Central AO) did not increase signal strength as much as the locally acquired correction (centered on the blue asterisk in Figure 3A; Figure 3F, Local AO). For the Thy1-YFP-16 mouse, the effective area of AO performed at the FOV center was estimated from the ‘AO/No AO’ ratio map (Figure 3G) to have a radius of ~185 µm (Figure 3H). Strategy for enlarging the effective area of AO correction for 3D cellular resolution imaging Imaging retinal vascular and neuronal structures, we found that the spatially varying aberrations of the mouse eye limited the effective area for AO correction that was acquired by sensing wavefront from a small region of the retina (e.g. 19×19 µm2 for Figures 1—3). Although this approach succeeded in resolving varicosities (Figure 1C) and neuronal processes (Figure 3E and F), for applications requiring 3D neuronal population imaging, synaptic resolution can be sacrificed in favor of cellular resolution imaging capability over larger FOVs. The latter can be achieved by correcting only for global mouse eye aberrations measured by scanning a larger retinal region for wavefront sensing. As a demonstration, for a 580×580 µm2 FOV, we measured aberrations from areas of 19×19, 95×95, 190×190, and 380×380 µm2 (Figure 4A, i-iv, yellow dashed boxes) and obtained differing corrective wavefronts resulting from the spatially varying aberrations. Quantifying and comparing AO effectiveness by their ‘AO/No AO’ pixel ratio maps (Figure 4A), we found that correcting aberrations from smaller areas provided greater local signal improvement but exhibited faster decay in signal improvement over distance (Figure 4B, i and ii). This was because the corrective wavefront acquired from a small FOV completely cancelled out the local aberrations and led to diffraction-limited imaging of local structures. For structures away from the wavefront sensing region and thus experiencing different aberrations, however, the same corrective wavefront led to substantial residual aberrations that degraded AO performance. In contrast, correcting aberrations from a larger area reduced signal improvement in the center of the area but enlarged the overall area within which signal was enhanced, which now extended over the entire imaging FOV (Figure 4B, iii and iv). Here, the wavefront measured from scanning the guide star over a larger FOV averaged out the local variations and represented the wavefront distortions common to all field positions. As a result, even though the improvement at the center of the wavefront sensing area was not as large, by removing the common aberrations from the entire FOV, this approach led to a larger effective area for AO correction. Figure 4 Download asset Open asset Larger WS areas enlarges the effective region of AO correction for 3D cellular resolution imaging. (A) Top: AO/NoAO pixel ratio maps for corrections with differently sized WS areas (yellow dashed boxes; i, 19×19 µm2; ii, 95×95 µm2; iii, 190×190 µm2; iv, 380×380 µm2). Bottom: (for [i]) corrective wavefront and (for [ii-iv]) difference in wavefronts between [ii-iv] corrective wavefronts and [i] corrective wavefront. (B) Radially averaged profiles of pixel ratio maps in (A). Insets: zoomed-in views of shaded areas. (C) Single image planes acquired (top) from INL and (bottom) GCL without and with AO using corrective wavefronts [i-iv], respectively. Insets: zoomed-in views of areas at FOV (a) center and (b,c) edge. All images normalized to AO images (AO [i] for inset a; AO [iv] for inset b,c). (D) Lateral (along red dashed lines) and axial (at the center of the neurons indicated by orange arrows) profiles of neurons in the (a) central and (b) edge regions. Figure 4—source data 1 Source image stacks with AO by measuring aberrations with different wavefront sensing areas. ‘No AO’ stack: ‘1_NoAO_9fAvg_stack.tif’‘AO [i]’ stack: ‘2_AO[i]_2_2.8_9fAvg_stack.tif’ ‘AO [