C3. Adaptive Optics In Vivo Cellular Imaging
C3.1. Photoreceptors and RPE
An important application of AOSLO technology studied in collaboration by several of the BRP teams is cone photoreceptor and RPE cell imaging and quantification. Although only inherited retinal degenerations are illustrated in this report, we have also found visualization and quantification of these cells to also be of value in AMD, central serous choidoretinopathy, infectious/inflammatory diseases, and retinal vascular disease.
C3.2. High Resolution Imaging of Cones in Patients with Retinal Degenerations (UR, UCB-AR, DEI)
Cone Morphology, Quantification, and Functional Correlation in Inherited Retinal Disease: In the first adaptive optics study of eyes with inherited retinal disease, photoreceptor packing and retinal microstructure was measured, and we reported a marked reduction in cone spacing in the cone-preserved regions of a single CRD patient compared to normal eyes (Wolfing 2006).
In a more extensive UCB-AR study (Duncan 2007), AOSLO images of macular photoreceptors were taken in 16 subjects, 5 patients with retinitis pigmentosa (RP), 3 with cone-rod dystrophy (CRD), and 8 normals. One RP patient and one CRD patient had known genetic mutations. The purpose was to investigate macular cone photoreceptor structure and function in retinal degeneration patients. Data from RP and CRD patients were compared with data from normal subjects. Cone spacing at 1 deg from the fovea was compared to standard measures of central visual function including best-corrected visual acuity (BCVA), foveal threshold and multifocal electroretinogram amplitude and timing. Nidek MP1 microperimetry was performed in a subset of subjects. Intervisit variation was studied in 1 RP and 1 CRD patient.
Cone spacing values were significantly different from normal subjects for both RP (P = 0.01) and CRD (P < 0.0001) patients (Fig C10). Cone spacing values were significantly correlated with foveal threshold (P = 0.0001), BCVA (P = 0.02), and mfERG amplitude (P = 0.01). Microperimetry results revealed functional losses that corresponded with regions where cones were not visible or cone spacing was abnormally large. Cone spacing measured during 2 sessions less than 8 days apart showed little variation.
Fig C10: Left panel: The images show a patient with RP (left) and CRD (right) at a location about 0.5 degrees from the fovea (scale bar is 10 minutes of arc). The cone mosaic of this RP patient is contiguous with a spacing that is no different from normal eyes, whereas the CRD patient shows a marked increase in spacing. The right panel shows the cone spacing results from the CRD (triangles) and RP patients (diamonds) compared to a normal distribution (based on AOSLO data from 8 normal eyes).
Fig C11: Photoreceptor defect in asymptomatic patient
Cone morphometry has also been performed on patients with advanced RP at DEI and will be used to characterize, select, and monitor patients being enrolled into upcoming retinal prosthetic trials. In addition to photoreceptor quantification, we have found AOSLO cone imaging to be useful in identifying early, otherwise sub-clinical abnormalities in asymptomatic patients. In a family with several members affected by Stargardt's disease and varying degrees of cone dropout evident on AOSLO, an asymptomatic member with a normal clinical exam (by biomicroscopy and angiography), was found with AOSLO to also have an unexpected defect in the outer segment mosaic (Fig C11). This observation highlights the potential for early detection of disease, improved understanding of early stages of disease pathogenesis, and early intervention afforded by AOSLO.
C3.3. Imaging Patients with Unexplained Vision Loss—Elucidating Disease Mechanisms
Fig. C12: The inset shows the ring shaped disruption of photoreceptors in Coffee and Donut Maculopathy seen here for the first time with the DEI AOSLO.
In a recent series of patients at Doheny complaining of otherwise unexplained focal central visual disturbance, we have observed subtle disruption of the photoreceptor inner segment-outer segment (IS-OS) band on spectral domain OCT. It was unclear whether this disruption of the IS-OS band was significant, and so one of these patients, despite 20/20 acuity, complained of a ring-like scotoma. FA in this patient was unrevealing, but an OCT B-scan demonstrated a subtle IS-OS discontinuity. AOSLO imaging, however, demonstrated a ring-like pattern of loss of the photoreceptor outer segments (Fig C12) which correlated perfectly with the patient's complaints. We believe this patient may have a form of "Coffee and Donut Maculopathy" as described by (Kerrison 2000) and the AOSLO finding may be first demonstration of the anatomic substrate for this poorly understood disorder.
C3.4. RPE Cell Imaging
Figure C13: Patient with cone-rod dystrophy. RPE cells appear as a regular array surrounding the central cone-preserved region. Microperimetry data are overlayed on the AOSLO image. Numbers indicate the attenuation scale in decibels. Absolute scotomas, indicated by empty red squares and labeled '0', correspond with regions with absent cones and visible RPE cells. Scale bar =1 deg.
The BRP has demonstrated two different methods for imaging single RPE cells in the living human eye based on reflectance (Roorda 2007) or autofluoresence (Gray 2006). In normal eyes, RPE cells are usually not visible using direct scattered light through the confocal pinhole. However, in patients with retinal disease that causes loss of the overlying photoreceptors, the UCB-AR group was able to visualize RPE cells for the first time (Fig C13). Spatial tests were used to compute the average size and regularity of the cells and it was found that they were consistent with histological reports for the same retinal location. Nidek MP1 microperimetry of the same CRD subject shows that areas where RPE cells are visible correlate with complete loss of visual function.
For visualization of RPE cells in the presence of intact photoreceptors, the UR group has successfully applied autofluorescence imaging with adaptive optics. The signal from the autofluorescence imaging at high-resolution is low; to combat this problem we have employed a dual simultaneous imaging and registration technique in order to increase the signal to noise of the RPE images (see C1.4.1). Images of the retina were obtained at 830 nm in reflectance and simultaneously autofluorescence images were collected over a 40 nm bandpass centered at 620 nm. Using a cross-correlation algorithm, the reflectance images were registered and the shifts between successive images were estimated. This motion estimation was then applied to the simultaneously collected low signal autofluorescence images. We have imaged both the macaque monkey (Gray 2006) at the human RPE mosaic (Wolfing 2006). Fig C14a shows the macaque monkey RPE mosaic at approximately 10 deg infero-nasal. Fig C14b shows the RPE mosaic at the fovea in the same monkey. In both images the RPE cell mosaic appears in a honeycomb pattern, where the cell nucleus is a dark spot surrounded by the autofluorescent lipofuscin granules.
Fig C14: RPE cells imaged with AOSLO at two retinal locations in the macaque monkey: (a) at an eccentricity of approximately 10 degrees infero-nasal and (b) centered at the fovea. Both images are the average of 1000 frames using dual registration. Scale bars are 75 µm. We have imaged the human photoreceptor and RPE mosaics simultaneously at several locations.
C3.5. Light-induced Changes in RPE Autofluoresence
The UR group has discovered a previously unknown change in RPE autofluorescence following exposure to light. A square region 0.5 deg on a side near the center of the macaque fovea was exposed for 15 minutes to 150 microwatts of light at 568 nm. This light level is 2.6 times lower than the maximum permissible exposure (MPE) as calculated by the American National Standard for the Safe Use of Lasers. Fig C15a shows the area of the retina that was imaged immediately following the exposure. The appearance of the dark square at the center of image shows that the exposure caused an immediate reduction in RPE cell autofluorescence. Despite this dimming, the mosaic of exposed cells is complete at this time. One hypothesis for this dimming is that it is caused by bleaching of lipofuscin in the RPE. If this hypothesis turns out to be true, we may have an interesting new tool for measuring the kinetics of lipofuscin. Fig C15b shows the same retinal location 11 days after the exposure. There appears to be a complete disruption of autofluorescence in the region exposed to light. Fig C15c shows some recovery of autofluorescence in the exposed area after 26 days. This may be indicative of RPE cell damage followed by RPE cell mosaic remodelling, but the correct interpretation of these images awaits histological confirmation. In any case, these results show that the fluorescence AOSLO instrument can not only image RPE cells but also uncover as of yet unreported changes in autofluorescence in the living eye.
Fig C15: A) The immediate dimming of autofluorescence in monkey RPE caused by extended light exposure. B) Loss in autofluorescence 11 days following exposure. C) Partial recovery of autofluorescence 26 days after exposure.
Thanks to the dual wavelength imaging capability developed in the BRP to average frames at very low light levels as well as our ability to control exposure duration, we are confident that excellent images of the RPE cell mosaic can be obtained using light levels that do not place the patient at risk. The light levels required to obtain images of RPE cells can be several orders of magnitude dimmer than those used here that produce changes in autofluorescence over time. Using our monkey model, we plan to undertake experiments this year to determine at what light levels these phenomena occur so that they can be avoided in human imaging experiments. Special care will be taken in patient populations, especially in light of the finding that rhodopsin mutant dogs may be especially sensitive to light exposure (Cideciyan 2005).
C3.6. In vivo images of single ganglion cells with fluorescent biomarkers in monkey (UR)
Fig. C16: In vivo ganglion cell imaging compared to histology using confocal microscopy. A, B show photofilled ganglion cells, dendrites and axons in vivo using the fluorescence adaptive optics scanning laser ophthalmoscope. C shows direct comparison of B imaged with a 20x objective in a confocal microscope.
A major goal of the BRP is to develop methods to image most retinal cell classes. The ganglion cells are especially difficult to image because of their transparency. Nonetheless, we have shown that they can be resolved, including their dendritic structure using fluorescence imaging in the living monkey eye. Technical developments which made this imaging possible include: (a) construction of a gimbal mount for rotating the monkey around the nodal point of the eye during imaging and (b) development of custom hard contact lenses with base curves and diameter optimal for monkey imaging, (c) development of procedures for intravitreal, intravenous and intracranial injection of contrast agents, including fluorescent dyes (rhodamine, fluorescein, Alexa 594) and GFP-expressing viral (AAV) vectors. Once filled with rhodamine or Alexa dyes, retinal ganglion cells remain visible for approximately 4 to 5 months. Light exposure causes intensification of rhodamine fluorescence, (Dacey 2003) thereby permitting visualization of ganglion cell axons and dendrites (Fig. C16). We are now comparing imaging with histology to determine if different classes of ganglion cell can be reliably discriminated with in vivo imaging. Fig. C16 shows that we have the capability to image cells in the intact animal and then locate them histology. This capability will be critical for develop new methods of marking and identifying specific cell classes in the next funding period.
C3.7. Imaging Ganglion Cells in vivo Without the Need for Fluorescent Markers (UR)
Fig C17: Reflection Nomarski DIC image of primate retinal ganglion cell mosaic with overlay of fluorescence image of labeled ganglion cells.
A new instrument is being developed at UR that will be capable of implementing a variety of modes for phase contrast and Nomarski differential interference contrast (NDIC) imaging. In vitro experiments in fixed tissue have shown that the ganglion cell mosaic can be resolved in reflection NDIC (Fig C17). The numerical aperture used in the microscope was the same as the numerical aperture available with our AOSLO. To confirm that these cells were indeed ganglion cells a subset were labelled with rhodamine dextran injected in the LGN for retrograde transport to the retina. Previous in vitro studies (Curcio 1990) used transmission NDIC, because it provides much greater contrast, but these data show that it is also possible in reflection NDIC. Such images from AOSLOs using low coherence sources should be superior to AOOCT images, which are degraded by speckle. This provides encouraging evidence that ganglion cell bodies will be resolved in vivo when our new instrument is complete, a major goal of the next funding period.
C3.8. Rodent Models of Retinal Pathology
The UR, UCB-JF, and IU labs are developing AOSLO methodology to acquire in vivo fluorescence images of the rodent retina. Given the large number of transgenic rodent (especially mouse) models of retinal disease, the ability to perform cellular imaging in these eyes will be of great value in understanding disease mechanisms, a potential which will also be exploited in some of the scientific aims of this proposal.
C3.8.1. Technical Challenges for in vivo Rodent Retinal Imaging
Although the rat eye axial length is 6.29 mm smaller, retinal cells in the rat have very similar dimensions and packing density to that of the human. The major difference is in total numbers of retinal neurons, e.g. the rat retina has ~15 million photoreceptors, in contrast to a human photoreceptor complement of ~ 120 million. Although the theoretical diffraction-limited resolution or the rat is nearly 2x better than in humans, it has proven very challenging to image fine cellular structure in the rat in vivo. We have solved most of the major technical hurdles, and are now acquiring high quality images of GFP-labeled retinal cells in the rat. We have modified an isofluorane anesthesia regimen to allow in vivo imaging of rodents for 1-5 hours with minimal eye movement artifacts. The short axial length and high lens power of the rat eye presents challenges for achieving near diffraction-limited optical correction. The overall power of the rat eye is ~300 D (5X that of the human), in addition, the spherical refractive power and corneal curvature of rat eyes changes with age, requiring a spherical correction of +20-25 D for 1-2 month old rats which decreases to +10 D by 4 months. We have developed contact lenses to maintain corneal hydration and correct for defocus.
Monochromatic aberrations: We have used adaptive optics to partially correct the higher order aberrations in the rat eye. Typically, the RMS wavefront error is reduced from ~0.6 µm to 0.1 µm over a 2.65-mm pupil. It is important to minimize the large values of sphere and cylinder before trying higher order corrections with AO. Large amounts of residual cylinder (~5-7 D) lead to little higher order correction with no noticeable improvement in image quality. We plan to evaluate larger stroke mirrors to reduce both residual lower (defocus and astigmatism) and higher order aberrations and yield better AO corrected RMS errors. After AO correction, the reflectance and fluorescence images increase in brightness, but cellular features remain unclear in post-processed images. We are investigating if lack of image quality is related to scattering from the thick rat retina, speckle effects, insufficient optical correction, errors due to eye movements that are not addressed by registration algorithms, or small isoplanatism due to the thickness of the optics compared to the axial length of the eye.
C3.8.2. Rat Ganglion Cell Imaging
Figure C18: Fluorescence montage (488 nm) of the eGFP labeled ganglion cell bodies, dendrites, and axons from an albino rat retina taken in vivo. Scale bar represents 10 µm on the rat retina.
RGC- specific AAV vectors were used to express eGFP in ganglion cells in normal rat retinas and found that with adaptive optics and dual registration we could successfully image fluorescently labeled ganglion cell bodies, axons, and dendrites. The Flannery lab has also engineered a number of additional viral vectors to permanently express fluorescent proteins (with various excitation and emission wavelengths) in specific classes of retinal cells by engineering combinations of cell–specific promoters and viral tropism. This cell–specificity is essential for verification of AO images, to correlate and register fundus images with cell types and layers. The ability to unequivocally identify cell types is critical to our goal of using AO fundus imaging for retinal cell counts to document disease progression, and cell "rescue" when evaluating the efficacy of therapeutics. We now have RGC, Müller glial cell, RPE, rod and cone cell-specific adeno-associated virus (AAV) and lentiviral vectors that reliably express fluorescent proteins in these cells for extended periods. These vectors are now being used in ongoing AOSLO imaging studies in primates and rodents.
C3.8.3. In vivo Fluorescence Imaging of Mouse Retinal Microglia and Vasculature (IU)
Working in collaboration with another BRP ("Live microscopy and cytometry in vascular biology," Charles Lin, PI), an AOSLO for mouse imaging was also constructed at SERI (Biss 2006). The AOSLO BRP provided support for software design and system implementation. This system allowed visualization of GFP expressing microglia, demonstrating the ability to image subcellular structures in the mouse eye. Additionally, we were able to see small capillaries, filled with Evans Blue. Similar to the rat, a number of obstacles related to the small eye size and marked variability in dioptric correction were overcome to achieve these images. First, a special holder system that positioned the mouse pupil at the center of a goniometer was developed, allowing the mouse body to be rotated around its pupil. Second, the mouse and an ophthalmoscopic lens were mounted on an optical slide, making a versatile Badal optometer with a dynamic range of > 200 diopters. Finally, the size of the mouse eye required that we scan the retina over larger angles, than are required for a human AO system. In summary, we have found that an AOSLO optimized for mouse imaging is fundamentally different than a scaled down version of a human AO system.
Fig C19: Fluorescence image of microglia in a mouse retina without (left) and with AO (center).
→ C4. Functional Imaging← C2. Clinical and Scientific Applications