In vivo imaging methods to assess glaucomatous optic neuropathy

Abstract

The goal of this review is to summarize the most common imaging methods currently applied for in vivo assessment of ocular structure in animal models of experimental glaucoma with an emphasis on translational relevance to clinical studies of the human disease. The most common techniques in current use include optical coherence tomography and scanning laser ophthalmoscopy. In reviewing the application of these and other imaging modalities to study glaucomatous optic neuropathy, this article is organized into three major sections: 1) imaging the optic nerve head, 2) imaging the retinal nerve fiber layer and 3) imaging retinal ganglion cell soma and dendrites. The article concludes with a brief section on possible future directions.

Introduction

Glaucoma is the most common optic neuropathy and the second leading cause of blindness worldwide.(Quigley and Broman, 2006) Vision loss in glaucoma occurs primarily because retinal ganglion cells (RGCs) die and their axons degenerate, losing the capacity to convey visual information to the brain.(Quigley, 1999, 2011; Quigley et al., 1981; Weinreb and Khaw, 2004) The initiating injury to RGCs and their axons is thought to occur within the optic nerve head (ONH).(Anderson and Hendrickson, 1974; Emery et al., 1974; Howell et al., 2012; Nickells et al., 2012; Quigley, 1999; Quigley et al., 1981; Vrabec, 1976) Indeed, chronic progressive deformation of the ONH tissues (traditionally referred to as “cupping” and/or “excavation”) is considered the hallmark of glaucoma, a clinical sign used to distinguish it from other optic neuropathies.(Danesh-Meyer et al., 2009; Quigley, 1993, 2011; Van Buskirk and Cioffi, 1992; Weinreb and Khaw, 2004) Thus clinical diagnosis and management of glaucoma depend not only on assessment of vision function, such as by measuring sensitivity across the visual field using automated perimetry, but also on assessment of structural integrity of the ONH and RGC axons within the retinal nerve fiber layer (RNFL).

Assessment of ONH and RNFL structural integrity can be achieved most easily by clinical examination using a direct ophthalmoscope, however, appreciation of finer detail is afforded by stereoscopic examination using binocular indirect ophthalmoscopy. Similarly, archival records of ONH and RNFL structure can be achieved in simplest form using a two-dimensional drawing or flash photograph, but simultaneous stereoscopic photographs offer superior capability for discerning changes in ONH surface topopgraphy due to their three-dimensionality.

More recent developments in ophthalmic imaging technology have enabled quantification of structural parameters, such as the width or volume of optic disc rim tissue, width and depth of the optic cup and cup-to-disc ratio, etc. One such technique known as confocal scanning laser tomography (CSLT) has become widely incorporated into both glaucoma research and clinical practice, and has even been used as an ancillary outcome measure for randomized clinical trials.(Alencar et al., 2008, 2005; Sharma et al., 2008; Wollstein et al., 2000; Zangwill et al., 2013; Zangwill et al., 2005) Optical coherence tomography (OCT) enables cross-sectional images of ONH and retinal tissue to be acquired in the living eye and thus quantification of individual retinal layer thicknesses, as well as three-dimensional (3D) visualization and quantification of deeper ONH structures such as the lamina cribrosa (LC).(Inoue et al., 2009; Kagemann et al., 2008; Srinivasan et al., 2008; Strouthidis et al., 2009; Wollstein et al., 2005a) Modern OCT systems provide axial resolution of a few micrometers and the addition of adaptive-optics techniques are able to improve lateral (transverse) resolution to the same level.(Hermann et al., 2004; Kocaoglu et al., 2011; Kocaoglu et al., 2014a; Kocaoglu et al., 2014b; Torti et al., 2009; Zawadzki et al., 2009; Zawadzki et al., 2005; Zhang et al., 2005) These most recent developments hold great promise for helping to elucidate directly from studies performed in clinical settings a more precise understanding of glaucoma risk, the susceptibility of individual eyes and even the sequence of events in glaucoma pathogenesis.

Yet experimental models of glaucoma remain important for these purposes, for testing hypothesis about pathogenesis and for pre-clinical trials of novel therapeutic interventions (reviewed elsewhere this issue and others, e.g., (Burgoyne et al., 2005; Goldblum and Mittag, 2002; Morrison et al., 2008; Pang and Clark, 2007)). Common outcome measures for experimental glaucoma (EG) models include histological counts of RGC soma and/or orbital optic nerve axons, which require euthanasia and are thus only available at a single time point. Outcome measures based on in vivo imaging techniques, such as those used in the clinical setting and mentioned above, enable longitudinal evaluation within animals. Benefits of longitudinal imaging include reduction of the number of animals required to adequately power a scientific study (instead of sacrificing a different set of animals at each time point) and mitigation against errors that can arise when inferences about longitudinal time course and inter-relationships are drawn from cross-sectional data.(Naskar et al., 2002; Prilloff et al., 2010; Sabel et al., 1997; Thanos et al., 2002) Moreover, application of imaging techniques that can be used across the spectrum of laboratory and pre-clinical studies in animals as well as clinical research and patient management in human glaucoma enables similar outcome measures to be assessed in order to better compare the pathophysiological sequence of an experimental model with the human disease. Such approaches may facilitate future translation of novel interventions and to ensure relevance and/or refine any given EG model accordingly. This review article summarizes the most common imaging methods currently applied for in vivo assessment of ocular structure in EG models with an emphasis on translation to clinical studies of the human disease. It is organized into three main sections based on target structure: ONH, RNFL, RGC soma and dendrites. Generally, since most ophthalmic imaging modalities were originally developed for clinical use in human eyes and then later adapted for use in animals, the discussion of each modality and target structure begins with some relevant background on development and application to basic and clinical studies of human eyes, then continues with laboratory studies in animals. It is hoped that this organizational structure will help emphasize the bidirectional translational nature of imaging in glaucoma research.

1. Imaging the ONH

Essentially any imaging modality used in a clinical setting for imaging human eyes can also be applied for imaging the eyes of larger laboratory animals - such as monkeys - with only minor modifications of technique to maintain optimal alignment, focus and corneal hydration since the latter are generally imaged under anesthesia. It is more challenging to image eyes of smaller laboratory animals such as rats and mice, primarily due to their smaller pupil size, higher optical power and aberrations. Thus, fundus photography in rats and mice benefits greatly from specialized equipment arrangements and procedures in order to obtain higher quality fundus images.(Cohan et al., 2003; Hawes et al., 1999; Kocaoglu et al., 2007; Paques et al., 2007) High-resolution images of the mouse or rat fundus can also be obtained by confocal scanning laser ophthalmoscopy (CSLO).(Chauhan et al., 2002; Cordeiro et al., 2004; Paques et al., 2006; Seeliger et al., 2005) Advantages of CSLO for fundus imaging include improved image contrast due the confocal aperture and relative ease of imaging through the smaller pupil due to the narrow diameter of the scanning beam; disadvantages include greater expense of CSLO systems and lack of true color reflectance since the imaging source is monochromatic, though this aspect also contributes to greater image contrast and facilitates fluorescence applications.(Cordeiro et al., 2004; Paques et al., 2006; Seeliger et al., 2005)

CSLO has been used to image the LC in vivo and perform quantitative morphometry in healthy and glaucomatous human eyes, measurements that were previously approachable only by histology. One of the earliest of such studies demonstrated that elongation of LC pores is related to the severity of glaucomatous visual field damage.(Fontana et al., 1998) The addition of adaptive-optics (AO) to compensate for optical aberrations(Burns et al., 2007; Liang et al., 1997; Roorda et al., 2002) provides higher resolution in both transverse and axial dimensions for SLO imaging, which is particularly important for morphometric analysis of a complex three-dimensional structure such as the LC.(Akagi et al., 2012; Ivers et al., 2011; Sredar et al., 2013; Vilupuru et al., 2007) Nevertheless, SLO imaging of the LC is ultimately limited to the visible portions of its anterior surface, which is masked across much of the ONH by major blood vessel branches and overlying rim tissue (consisting of axon bundles, astroglia and capillaries, all sources of scatter). Moreover, the LC of the rat ONH has only a few wispy vertical collagenous beams(Morrison et al., 1995) while the mouse ONH has none(May and Lutjen-Drecoll, 2002; Sun et al., 2009) and the major vessel branches also occupy a much greater proportion of the “optic disc” in rats and mice as compared with human and non-human primate (NHP) eyes.(Chauhan et al., 2002; Fortune et al., 2011; Guo et al., 2005; Srinivasan et al., 2006; Zhi et al., 2011; Zhi et al., 2012) Thus the applicability of two-dimensional reflectance imaging of the ONH in rat or mouse EG models is limited. However, Bosco and colleagues recently used fluorescence CSLO imaging in vivo to document early-stage ONH microgliosis in the DBA/2J mouse model of inherited glaucoma.(Bosco et al., 2015) Ho et al., have also used in vivo longitudinal CSLO imaging to document gliotic responses of astrocytes within the ONH after RGC injury.(Ho et al., 2009)

OCT imaging provides much higher axial resolution (~ two orders of magnitude better) than CSLO, which has proven important to 3D visualization of the human LC.(Inoue et al., 2009; Kagemann et al., 2008; Srinivasan et al., 2008) Use of the “enhanced depth imaging” (EDI) mode(Lee et al., 2012; Park et al., 2012a; Park et al., 2012b; Yang et al., 2012a) and related strategies(Kim et al., 2012) in order to mitigate the signal roll-off inherent to spectral domain OCT improves measurement reliability and has revealed focal LC structural alterations that relate to glaucomatous disease severity.(Faridi et al., 2014; Tatham et al., 2014; You et al., 2013) Similarly, postprocessing techniques can be applied to recover contrast lost to signal attenuation(Girard et al., 2015; Mari et al., 2013) and further improve OCT imaging of deep ONH structures. The increased penetration of longer wavelength OCT sources (e.g. center wavelength of 1050 nm), such as those typically used in swept-source OCT may also improve imaging of the LC, peripapillary sclera and deep ONH structures.(Srinivasan et al., 2008; Takayama et al., 2013a; Wang et al., 2013; Yoshikawa et al., 2014) Using polarization-sensitive swept-source OCT imaging at 1μm further enhances visualization of these strongly birefringent connective tissues.(Yamanari et al., 2009) However, one trade-off of the longer wavelength is a slight reduction of transverse resolution and the narrower frequency band available for most swept-source lasers results in reduced axial resolution (as compared with spectral domain OCT). These trade-offs can be overcome using adaptive optics, which enables 3D visualization in vivo of retinal and ONH anatomy at a cellular scale (Kocaoglu et al., 2011; Kocaoglu et al., 2014a; Kocaoglu et al., 2014b; Torti et al., 2009; Zawadzki et al., 2009) and reliable morphometry of LC microarchitecture in the living human eye.(Nadler et al., 2014a; Nadler et al., 2014b) Yet masking by the major vessel branches remains an important constraint on all OCT imaging of the LC.

Spectral domain OCT (SDOCT) has been used to reveal early-stage ONH changes in vivo in a unilateral, inducible model of EG in NHP based on chronic, mild-to-moderate elevation of intraocular pressure (IOP).(He et al., 2014; Strouthidis et al., 2011b; Yang et al., 2014; Yang et al., 2012a) Specifically, SDOCT studies have shown that LC changes such as an increase in the depth of its anterior surface precede peripapillary RNFL thinning, another OCT-derived parameter now commonly used in clinical management of glaucoma.(He et al., 2014; Strouthidis et al., 2011b; Yang et al., 2014) Further evidence that ONH changes precede thinning of the peripapillary RNFL in this NHP EG model has been reported for studies that used CSLT for longitudinal assessment of ONH surface topography.(Fortune et al., 2012; Fortune et al., 2013b; He et al., 2014; Strouthidis et al., 2011b) Collectively, these studies demonstrate clearly that ONH changes occur at a very early stage in the NHP EG model, prior to loss of RNFL thickness measured adjacent to the ONH as is typically done clinically for glaucoma management. What remains to be determined is whether any or all of these early-stage ONH changes, such as changes in the depth of the anterior LC surface, alterations of LC microarchitecture, pore size and shape, or changes in ONH surface topography are specifically predictive of subsequent RGC injury and axon loss such that they should be considered clinical indicators for therapeutic intervention. In contrast, current evidence from the NHP EG model indicates that the first detectable loss of peripapillary RNFL thickness is already associated with 10–15% loss of axons from the orbital optic nerve,(Cull et al., 2012, 2014) suggesting that earlier indicators are needed for timely treatment intervention.

Another exciting question that remains to be answered is whether aspects of ONH structure measured in vivo by OCT or other techniques will serve as specific predictors of glaucoma risk even in the absence of documented change; for example perhaps certain aspects of LC microarchitecture will prove to confer biomechanical susceptibility for a given individual eye. One manner of probing biomechanical behavior of ONH tissues is by acute “compliance test” whereby IOP is manipulated acutely to alter mechanical load and the response of ONH structures is imaged in vivo.(Agoumi et al., 2011; Burgoyne et al., 1995; Coleman et al., 1991; Fortune et al., 2011; Heickell et al., 2001; Strouthidis et al., 2011a) This kind of study is also important for characterizing the degree to which structural parameters are influenced by the ambient IOP level in order to help inform cross-sectional clinical diagnosis.(Fortune et al., 2009) Thus it may be that through a combination of “baseline” ONH structural parameters and acute compliance test results, the biomechanical integrity of ONH tissues will refine predictions of which individual eyes are at greatest risk for developing glaucoma and the most rapid rate of progression.

One more important way of characterizing ONH structure is the quantitative measurement of the neural rim tissue. For more than two decades, CSLT was the predominant means by which reliable clinical measurements of ONH structural parameters such as rim width or rim area were obtained.(Alencar et al., 2008, 2005; Sharma et al., 2008; Wollstein et al., 2000; Zangwill et al., 2013; Zangwill et al., 2005) The preceding sections of this review alluded to the capability of CSLT for measuring ONH surface topography, that is, providing a high-resolution 2D map of “surface height” relative to an axial (longitudinal) reference plane. By outlining within that map the outer extent of the optic disc and by determining where the surface dips below the reference plane (i.e. designating the border between the rim and the “optic cup”), the dimensions of the optic disc rim tissue can be calculated with high precision.(2005) For example, Shimazawa et al monitored rim (and cup) parameters longitudinally using CSLT in cynomolgus monkeys with very high IOP over 12–16 weeks and found high correlations with RNFL parameters determined both by scanning laser polarimetry (SLP) and by histological measurements.(Shimazawa et al., 2006) However, there was no means of IOP control at the time of CSLT scanning in order to account for the elastic (in contrast to plastic) components of ONH deformation.(Burgoyne et al., 1995; Coleman et al., 1991; Heickell et al., 2001) Since peripapillary SLP measurements are minimally affected by ambient IOP,(Fortune et al., 2009) correlations may be even higher than reported. Yucel and colleagues reported strong correlations between ONH parameters measured using CSLT and optic nerve axon counts from eyes of macaque monkeys with EG,(Yucel et al., 1998) yet again these correlations may have been even stronger after accounting for the effects of ambient IOP at the time of the CSLT scans.(Burgoyne et al., 1995; Coleman et al., 1991; Heickell et al., 2001; Yucel et al., 1998) Hare et al demonstrated evidence of a protective effect from memantine treatment in monkeys with EG based on longitudinal CSLT measurements of ONH structure.(Hare et al., 2004)

However, it is well known that the lack of a stable, reliable reference plane ultimately hampers both the precision and accuracy of CSLT measurements and their ability to detect ONH structural changes in human glaucoma.(Kamal et al., 2000; Strouthidis et al., 2005a; Strouthidis et al., 2005b) Studies have also shown that agreement between CSLT and expert analysis of stereoscopic photographs for detecting progressive ONH structural change is limited.(O'Leary et al., 2010; Vizzeri et al., 2010; Wollstein et al., 2000) Hence there is growing interest in application of OCT for the measurement of ONH neural rim parameters since OCT provides a more accurate anatomical definition of rim tissue boundaries and more stable anatomical reference plane (for review, see(Chauhan and Burgoyne, 2013)). Specifically, given the advantage of cross-sectional imaging with high axial resolution, OCT is able to provide neural rim measurements that better follow ONH anatomy, such as those based on the minimum distance calculations.(Chauhan et al., 2013; Chen, 2009; Gardiner et al., 2014; Povazay et al., 2007) In addition to neural rim measurements such as minimum rim width, minimum rim area and rim volume, OCT also should be capable of producing 2D surface topography maps with resolution equal to CSLT since increasingly fast acquisition speeds (now into the MHz range for research instruments) will enable high density A-scan patterns with minimal eye movement artifact.

Strouthidis et al first demonstrated in a longitudinal study of NHP EG that minimum rim width and minimum rim area measured using SDOCT had decreased from baseline at the onset of CLST-detected ONH surface topography change. Evaluating entire longitudinal sequences, He et al(He et al., 2014) subsequently demonstrated that neural rim thinning as measured by these SDOCT parameters preceded or coincided with CSLT-detected ONH surface topography change in most NHP EG eyes. Based on SDOCT data from a similar NHP EG model, as well as cross-sectional analyses of clinical data from human eyes, Patel et al also reported evidence that neural rim thinning precedes peripapillary RNFL thinning.(Patel et al., 2014) What remains to be determined is the extent to which ONH rim deformation and thinning relates directly to ONH connective tissue deformation and remodeling since current evidence from NHP EG suggests that early-stage changes of both structural components manifest in concert before any decrease in the total number of optic nerve axon loss occurs. Figure 1 provides an individual example from NHP EG of relationships between SDOCT derived ONH parameters, peripapillary RNFL thickness and complete orbital optic nerve axon counts.

Figure 1. ONH parameters obtained from SDOCT radial scans under manometric IOP control.

Individual example of an NHP EG eye imaged at its first baseline (A, B, C) and at its final time point (D, E, F) just prior to sacrifice. Insets in the lower left corner of A and D show the location of the B-scan as indicated by the bold green line overlaid onto the infrared CSLO reflectance image. Structures delineated in each radial B-scan include the ILM (yellow) and BMO points (red). The green segments connecting BMO points to the ILM represent the pair of MRW measurements made in each radial B-scan. Results for all 80 B-scans are shown projected from 3D for the baseline (B) and final time point (E). Note deformation of the ONH apparent in E, including a deeper “cup” and thinner “rim”. Global average MRW decreased from 330 μm at baseline to 200 μm at the final time point (−40%). MRA decreased from 1.11 mm² at baseline (C) to 0.69 mm² at the final time point (F, −38%). Segmentation of peripapillary circular B-scans to obtain the parameter RNFL thickness are shown for the same eye and time points in G and H. Segmentations are shown for the ILM (yellow) and posterior RNFL boundary (magenta). Global average RNFL thickness decreased from 107.1 μm at baseline (G) to 82.9 μm at the final time point (H, −23%). Optic nerve axon counts for both eyes (the EG eye and fellow control eye) of the same animal (I,J). All axons were counted from 100% of the cross-sectional area of each optic nerve. Shown here are the composited montages of all image tiles used to count axons for the control eye (I) and EG eye (J) of the same NHP shown in panels A–H. Axon density of each tile is color-coded, insets show a magnified view of a single tile. Total axon count was 1,319,268 in the control eye and 1,001,512 in the EG eye (−24.1%). As reported by Hardin et al. (IOVS 2015;56:ARVO E-Abstract 1010) and Fortune et al. (IOVS 2015;56:ARVO E-Abstract 3982), the ONH parameters MRW and MRA are “more sensitive” than RNFL thickness insofar as they discriminate better between EG and control eyes and change at an earlier stage of damage (less axon loss), but peripapillary RNFL thickness is more strongly correlated with axon loss than MRW or MRA. Both observations are most likely caused by ONH connective tissue deformations contributing more to ONH rim thinning than to peripapillary RNFL thickness changes, whereas axon loss dominates the change in peripapillary RNFL thickness.

Another desirable ONH imaging target is blood flow. Longitudinal studies by Lin Wang and colleagues using laser speckle flowgraphy have demonstrated that blood flow through ONH capillary beds actually increases above normal during the earliest stages of NHP EG, followed by progressive decline in proportion to the degree of peripapillary RNFL thinning.(Cull et al., 2013) Further, their studies have revealed that the dynamic (but not the static) autoregulation of ONH capillary blood flow is altered in NHP EG.(Wang et al., 2014b; Wang et al., 2014c) Future studies should determine how ONH capillary bed blood flow changes relate to ONH structural changes such as deformation and remodeling of LC microarchitecture, LC insertion and neural rim thinning.

As in human glaucoma and the NHP EG model, it is thought that the site of injury initiating subsequent RGC axon degeneration and death in rodent EG models is also located within the ONH)(Chidlow et al., 2011; Howell et al., 2007; Jakobs et al., 2005; Martin et al., 2006; Morrison et al., 1997; Salinas-Navarro et al., 2010; Schlamp et al., 2006; Soto et al., 2008; Soto et al., 2011) Therefore the ONH is an important target for in vivo imaging in small rodents. Chauhan and colleagues reported the first longitudinal study demonstrating ONH changes in vivo in a rodent model of EG.(Chauhan et al., 2002) Using CSLT, Chauhan and colleagues documented progressive “cupping” of the rat ONH in proportion to the degree of IOP elevation and in some cases also a dramatic expansion of the scleral canal confirmed histologically. Guo et al applied longitudinal CSLO and OCT imaging in a similar EG model in rats and confirmed the development of ONH changes in vivo, including expansion of the scleral canal.(Guo et al., 2005) Interestingly, Chauhan et al presented evidence for a “threshold” of peak IOP and IOP integral beyond which these ONH changes observed in vivo became increasingly prominent.(Chauhan et al., 2002) However, they also reported that the proportion of surviving axons exhibited a more linear relationship to peak and integral IOP, suggesting that the sudden, dramatic structural changes in the rat ONH in that EG model may reflect changes in connective tissues like the peripapillary sclera and may be “primarily a mechanical phenomenon”.(Chauhan et al., 2002) The full-field ERG results presented in the study by Chauhan et al suggest retinal ischemia was also present above the same IOP threshold that caused rapid severe cupping and expansion of the scleral canal.(Bui et al., 2013; Bui et al., 2005; Bui and Fortune, 2004; Fortune et al., 2004) Thus it may be that ocular ischemia in combination with significant chronic IOP elevation combine to create a more dramatically altered mechanical environment in the rat (or any) ONH.

Early demonstrations of OCT imaging in small rodents using SDOCT operating at either 830 nm (and bandwidth of 70 nm),(Ruggeri et al., 2007) at 870 nm (bandwidth of 100 nm)(Fischer et al., 2009) or at 900 nm with ultrahigh-resolution (bandwidth of 150 nm)(Srinivasan et al., 2006) revealed the difficulty of imaging details of the deeper ONH structures, which are largely masked in shadows cast by the overlying dense vasculature. Thus few studies have used OCT specifically to assess ONH structure in rodents. Fortune et al documented rapid and reversible deformation of ONH and peripapillary tissue in the rat eye subjected to acute IOP elevation using a commercial SDOCT system (Heidelberg Spectralis operating at 870 nm with 100 nm bandwidth).(Fortune et al., 2011) Based on the limited visibility of deeper ONH structures in that study,(Fortune et al., 2011) a subsequent study was performed in healthy naïve rats, which showed that SDOCT imaging at 1050 nm with the EDI mode enhanced visibility of deeper ONH structures in the rat eye.[Choe et al., Invest Ophthalmol Vis Sci 2012;53: E-Abstract 2820] Images through the ONH of mice shown recently in reports using another commercial SDOCT system (Bioptigen) suggest that assessment of ONH structure may be possible with that system, though the investigators in those studies were focused on the peripapillary retina and did not comment on the ONH.(Liu et al., 2014; Yang et al., 2012b) Using a custom SDOCT system operating at 1300 nm in combination with their ultra-high sensitive optical microangiography technique, Zhi and colleagues have demonstrated extraordinary capabilities to visualize detailed anatomy of the rat ONH in vivo, especially of the microvasculature and responses to acute IOP elevation.(Zhi et al., 2011; Zhi et al., 2012) Polarization-sensitive OCT may further enhance visualization of connective tissues and other birefringent structures in the rat eye, such as the peripapillary sclera as recently demonstrated by Baumann et al.(Baumann et al., 2014) For the rodent ONH, which lacks a collagenous LC, an approach such as polarization-sensitive OCT may enable in vivo detection of changes in the “glial LC”, including density and orientation of astrocyte processes and their actin cytoskeleton.(Lye-Barthel et al., 2013; Sun et al., 2013; Tehrani et al., 2014) Implementation of AO strategies for overcoming challenges inherent to imaging rat and especially mouse eyes by OCT (and SLO) may also become important to ONH assessment in EG models.(Geng et al., 2012; Geng et al., 2009; Jian et al., 2013);Jian et al., Invest Ophthalmol Vis Sci 2014;55: E-Abstract 2083;Zhang et al., Invest Ophthalmol Vis Sci 2014;55: E-Abstract 2085.

2. Imaging the RNFL

Ultimately, glaucoma causes degeneration of RGC axons, which are visible along the innermost layer of the retina by ophthalmoscopy and other techniques. For decades RNFL defects have been considered an important diagnostic sign of glaucoma since they present clinically at an early stage and are predictive of subsequent visual field loss.(Hoyt et al., 1973; Hoyt and Newman, 1972; Sommer et al., 1991) The visibility of axon bundles within the RNFL by clinical ophthalmoscopy and photography depends on their reflectance, which is also known to be highly directional, wavelength dependent, and strongly influenced by normal cytoskeletal integrity.(Huang et al., 2006; Knighton and Huang, 1999; Knighton et al., 1989; Knighton and Qian, 2000; Knighton and Zhou, 1995; Zhou and Knighton, 1997)

However, the minimum loss of RNFL detectable by traditional clinical methods such as ophthalmoscopy and photography may be 50 to 70 μm, or as much as 50% of the normal thickness.(Quigley, 1986; Quigley and Addicks, 1982) Direct clinical examination and photography are also limited by subjectivity, with high intra- and inter-observer variability. More recent development of advanced techniques for imaging the RNFL have enabled reliable, sensitive and quantitative measurements to be obtained clinically with the hope of improving glaucoma detection and management.(Greaney et al., 2002; Medeiros et al., 2004; Townsend et al., 2009; Zangwill and Bowd, 2006) The two techniques that have been used most commonly for this purpose are SLP(Lemij and Reus, 2008; Medeiros et al., 2007; Reus and Lemij, 2004; Weinreb et al., 1990) and OCT,(Schuman et al., 1995; Wollstein et al., 2005b) both of which provide predictive capability for future glaucoma progression including loss of vision.(Lalezary et al., 2006; Mohammadi et al., 2004)

Based on the principle of interferometry, OCT produces high-resolution cross-sectional images of tissue structure,(Costa et al., 2006; Drexler and Fujimoto, 2008; Huang et al., 1991) which can be used to estimate thickness of individual retinal layers such as the RNFL by automatic detection of the relatively steep reflectance transitions at its anterior and posterior boundaries.(Sakamoto et al., 2010) RNFL thickness measurements obtained by OCT have been shown to agree with histomorphometric measurements of RNFL thickness (Schuman et al., 2007) and optic nerve axon counts (Cull et al., 2012, 2014; Fortune et al., 2015) in NHP EG and in rats after optic nerve crush injury.(Nagata et al., 2009) RNFL thickness measurements from OCT also correlate with RGC soma counts obtained in vivo by CSLO in mice and rats after optic nerve injury.(Chauhan et al., 2012; Choe et al., 2014)

In contrast, SLP measures the relative phase retardance of orthogonally polarized states of its imaging source after a double pass through the tissue sample.(Weinreb et al., 1990) The relatively high phase retardance exhibited by the RNFL reflects form birefringence, an optical property thought to be due to the orderly parallel array of thin cylindrical cytoskeletal components within these axons, including microtubules and neurofilaments.(Huang and Knighton, 2002, 2005; Zhou and Knighton, 1997) Empirical evidence published over the past decade supports this theoretical framework.(Fortune et al., 2008b; Huang and Knighton, 2005; Pocock et al., 2009) Thus measurements of RNFL birefringence might well provide a sensitive biomarker of compromised cytoskeleton integrity within RGC axons.(Fortune et al., 2008b; Huang et al., 2004; Huang and Knighton, 2005) The importance of this idea is underscored by evidence of axonal cytoskeletal changes observed in models of EG models, including their earliest stages;(Balaratnasingam et al., 2008; Balaratnasingam et al., 2007; Fortune et al., 2014a; Huang et al., 2011a; Huang and Knighton, 2009; Kashiwagi et al., 2003) some of which may represent mechanisms of further susceptibility.(Beirowski et al., 2010; Coleman, 2005; Soto et al., 2011; Vickers et al., 1995)

RNFL birefringence can be assessed in a clinical setting either directly, such as by polarization-sensitive OCT(Braaf et al., 2014; Cense et al., 2002, 2004; Cense et al., 2007; Dwelle et al., 2012; Elmaanaoui et al., 2011; Gotzinger et al., 2005; Pircher et al., 2011; Rylander et al., 2005; Zotter et al., 2013) or it can be inferred by comparing SLP measurements of RNFL retardance to OCT measurements of RNFL thickness. For example, a series of experiments by Fortune and colleagues provides clear evidence of an early stage of RGC degeneration when both axonal cytoskeletal abnormalities and RGC functional abnormalities are found in the absence of significant thinning of axon bundles within the RNFL in an NHP model of EG and after optic nerve transection.(Fortune et al., 2013a; Fortune et al., 2012; Fortune et al., 2008a) (Fortune et al., 2015; Fortune et al., 2014a). As mentioned, the reflectance of axons within the RNFL is also thought to depend on the integrity of their cytoskeleton.(Huang et al., 2006; Knighton et al., 1998) Thus it is interesting that both clinical(van der Schoot et al., 2012; Vermeer et al., 2012) and laboratory studies (Huang et al., 2011b) have demonstrated altered reflectance (and/or attenuation) of the glaucomatous RNFL, suggesting that this phenomenon might be another sensitive indicator of early axonal damage and a target for novel clinical imaging strategies.(Zhang et al., 2011)

In a laboratory setting, the visibility of RNFL axons for in vivo imaging can be further enhanced by introduction of exogenous fluorescent contrast agents(Abbott et al., 2013; Kanamori et al., 2012; Kanamori et al., 2010) or genetically-expressed fluorescent reporters.(Leung et al., 2011; Walsh and Quigley, 2008) The addition of AO to compensate for optical aberrations enables imaging of RNFL structure in exquisite detail, with or without fluorescent contrast agents by SLO, or by OCT, including in the shorter eyes of small rodents.(Geng et al., 2012; Geng et al., 2009; Gray et al., 2008; Jian et al., 2013);[Jian et al., Invest Ophthalmol Vis Sci 2014;55: E-Abstract 2083;Zhang et al., Invest Ophthalmol Vis Sci 2014;55: E-Abstract 2085] AO-SLO and AO-OCT have recently been used to image RNFL bundles in human eyes, revealing fine detail in both healthy and damaged areas.(Chen et al., 2015; Huang et al., 2014; Kocaoglu et al., 2011; Kocaoglu et al., 2014b; Takayama et al., 2012; Takayama et al., 2013b) One interesting finding common to several of these recent AO studies is that reflectivity of RNFL bundles is often discrete, certainly more so than can be appreciated without AO (e.g., see Figure 2). It is possible that these more discrete reflectivity profiles arise from the varicosities known to be present along RGC axons within the RNFL, most commonly at the site of internal mitochondria.(Wang et al., 2003) Mitochondria are known to be actively transported along these unmyelinated axon segments and that this process is disturbed in a rodent EG model.(Brown, 2003)[Takihara et al., Invest Ophthalmol Vis Sci 2013;54: E-Abstract 756 and 2014;55: E-Abstract 2408] These findings lead to the sound prediction that some aspects of RNFL bundle reflectivity might be dynamic and related to transport activity. Huang and colleagues have recently demonstrated in a rat retinal explant preparation that the discrete speckle pattern of RNFL reflectance was dynamic and related to activity.(Huang et al., 2013) It should be possible to extend these developments to investigations of RGC axonal activity in the living eye, in humans and in EG models, given 3D cellular-level resolution and MHz speed of recent AO-OCT systems.(Kocaoglu et al., 2014b) Figure 3 shows preliminary supporting evidence from recent work by Liu, Kocaoglu, Miller and colleagues using MHz AO-OCT.(Kocaoglu et al., 2014b)

Figure 2.

Imaging axonal activity in a human eye. (Left) MHz AO-OCT en face projection of RNFL bundles centered 3.5° nasal to the fovea. I mage is from a single volume acquired at a rate of 1 million A-lines/s (1 MHz), covering 3.6°x3.6° with 1 μm/pixel lateral sampling.(Kocaoglu et al., 2014b) (Right) Time sequence of a subsection of RNFL bundles from the red dashed region of the left image. Colored labels denote bundle locations at which reflectance was measured every minute over a 3 minute period. Bottom image of column is an average of the four time points. Red arrowheads at t=0 point to individual RNFL bundles. (Bottom) Cross section of RNFL bundles extracted at the location of the yellow dashed line in the left image. Figure courtesy of Zhuolin Liu, Omer Kocaoglu and Don Miller, Indiana University.

Figure 3.

Measurement of axonal activity. Correlation coefficient vs. time for RNFL bundle scatter dynamics of the labeled squares in the RNFL bundle image (inset and in Fig. 1). Exponential decay with 0.34 min time constant (blue curve) is consistent with results reported by Huang et al.(Huang et al., 2013) for in vitro rat retina. Figure courtesy of Zhuolin Liu, Omer Kocaoglu and Don Miller, Indiana University.

Meanwhile, to date, the application of imaging methods such as SLO and OCT to perform longitudinal in vivo assessment of RNFL damage in models of optic nerve injury including EG has been primarily focused on quantification of RNFL (or total retinal) thickness. As mentioned above, progressive loss of RNFL thickness is a well-established outcome measure in NHP models of EG.(Cull et al., 2013; Cull et al., 2012; Fortune et al., 2013a; Fortune et al., 2012; Fortune et al., 2015; Fortune et al., 2008a; Fortune et al., 2014b; Fortune et al., 2013b; He et al., 2014; Patel et al., 2014; Strouthidis et al., 2011b; Yang et al., 2014) Such measurements are also becoming increasingly applied in rodent models. Kawaguchi et al inferred the thickness of the RNFL from changes in the focal plane setting of an SLO and documented a specific decline after optic nerve crush in rats.(Kawaguchi et al., 2006) Nagata et al showed that OCT measurements of RNFL thickness declined progressively after optic nerve crush in rats and were strongly correlated to histological measurements of RNFL thickness.(Nagata et al., 2009) Gabriele et al measured total retinal thickness by SDOCT and found an initial increase, followed by progressive decline after optic nerve crush in mice.(Gabriele et al., 2011) Chauhan et al measured both RGC density and total retinal thickness in vivo, by CSLO and SDOCT respectively, after optic nerve transection in mice and showed progressive decline of both parameters.(Chauhan et al., 2012) Choe et al measured RNFL thickness and RGC density in vivo by SDOCT and CSLO, respectively, and found a strong correlation between them over 4 weeks of longitudinal follow up after optic nerve transection in rats.(Choe et al., 2014) Though Choe et al also reported that RNFL thinning was delayed relative to the onset of declining RGC soma, similar to what Nakano et al reported earlier for NMDA-induced injury.(Nakano et al., 2011) Liu et al used longitudinal SDOCT imaging to document a progressive decline in the combined thickness of the RNFL, RGC layer and inner plexiform layer (IPL) after optic nerve crush in mice and also found that this combined layer thickness measurement was correlated to density of soma in the RGC layer and progressive functional loss measured by ERG.(Liu et al., 2014)

Guo et al were the first to report SDOCT measurements of retinal layer thicknesses in a rodent EG model based on chronic IOP elevation.(Guo et al., 2010) They found a small decrease in RNFL thickness 3- and 8-weeks after IOP elevation was induced, but no change in the thickness of the IPL and a rather substantial progressive loss of outer retinal thickness that is not observed in human glaucoma(Ishikawa et al., 2005; Tan et al., 2009; Wang et al., 2009) or in NHP models.[Wilsey et al., Invest Ophthalmol Vis Sci 2015;56: E-Abstract 638] Yang et al also reported that in a mouse EG model based on chronic IOP elevation, loss of the combined RNFL, RGC and IPL thickness measured by SDOCT was related to RGC density sampled histologically.(Yang et al., 2012b) Gramlich et al found after chronic recurrent episodes of a short-duration IOP “spike” (1 hour each to 30–35 mmHg), there was a decline in total retinal thickness measured by SDOCT and RGC density measured histologically as well as an increase in the optic nerve axon damage grade.(Gramlich et al., 2014) Using longitudinal SDOCT imaging, Abbott et al found that RNFL loss occurred earlier and was more severe in older rats age 9.5-months as compared with younger rats aged 2-months despite similar exposure to chronic IOP elevation over 3-weeks. Abbott et al also reported that axonal transport disruption and optic nerve damage were substantially worse than RNFL thinning, especially in the older rats and confirmed by post mortem microscopy of retinal whole mounts that there was relative preservation of the intraretinal portion of RGC axons. This underscores both the usefulness and limitations of in vivo imaging and the need to continue characterizing the relationship between RNFL thickness and the integrity of distal axon segments.(Crish et al., 2010; Cull et al., 2012; Howell et al., 2007; Huang et al., 2011b; Schlamp et al., 2006)

Figure 4 shows an example of longitudinal imaging by CSLO and SDOCT (Spectralis HRA+OCT, Heidelberg Engineering, GmbH) in a pigmented rat during baseline and four weekly follow-ups after unilateral optic nerve transection.(Choe et al., 2014) Infrared reflectance images obtained by CSLO are shown at one baseline time point in panel A and at follow-up time points 1, 2 and 4 weeks after transection in panels B, C and D, respectively. The RNFL bundle reflectance has nearly completely disappeared by 4-weeks while the contrast of the major retinal vessels and capillaries has become more prominent against the fundus background. The horizontal yellow line in panel A shows the location of one of 290 raster lines from the SDOCT volume, the corresponding B-scan is shown in panel E for baseline and E’ 4-weeks after transection; similarly, the vertical yellow line shown in panel A indicates the position of an interpolated B-scan, which is shown in panel F for baseline and F’ 4-weeks after transection. Note the nearly complete thinning of the RNFL 4 weeks after transection as compared with baseline (indicated by the green arrows). En face projections from the SDOCT volumes obtained for the entire longitudinal series are shown for the transected eye in row G and the fellow control eye in row H. The RNFL progressively disappears over weeks 2–4 in the transected eye. In EG models, at least in those where the optical media remain clear enough to obtain image series like these, it will be interesting to test hypotheses about longitudinal changes in RNFL reflectance and thickness relative to the underlying changes in axon ultrastructure.(Fortune et al., 2014a)

Figure 4.

Example of longitudinal imaging by CSLO and SDOCT (Spectralis HRA+OCT, Heidelberg Engineering, GmbH) in a pigmented rat during baseline and four weekly follow-ups after unilateral optic nerve transection.(Choe et al., 2014) (A) Infrared reflectance image obtained by CSLO (30 x 30 deg) at one baseline time point; (B) follow-up time point 1 week after transection; (C) 3 weeks after transection; (D) 4 weeks after transection. RNFL bundle reflectance has nearly completely disappeared by 4-weeks while the contrast of the major retinal vessels and capillaries has become more prominent against the fundus background. SDOCT volumes consisting of 290 horizontal B-scans (15 x 15 deg) were obtained weekly during baseline and at each of the 4 follow-ups. (E) An example B-scan from the position indicated by the horizontal yellow line in A. (F) An interpolated slice through the SDOCT volume at the position shown by the vertical yellow line in A. The healthy RNFL at baseline (E and F, green arrows) is nearly completely thinned 4 weeks after transection E’ and F’. (G and H) En face projections from the SDOCT volumes across the entire longitudinal series are shown for the transected eye in row G and the fellow control eye in row H. The RNFL progressively disappears over weeks 2–4 in the transected eye. Each panel shows the sum voxel projection for a 10-pixel slab beginning at the internal limiting membrane (e.g. downward arrows in E and F).

3. Imaging the RGC Soma and Dendrites

As first demonstrated by Zeimer and colleagues, in vivo imaging of the macula offers an important, complementary approach to glaucoma clinical care and research.(Zeimer et al., 1998) Since loss of RGCs will result not only in thinning of the RNFL, but also of the inner retina, imaging the clinical macula by OCT has gained attention for its potential diagnostic capability.(Greenfield et al., 2003; Hood et al., 2013; Hood et al., 2014; Ishikawa et al., 2005; Leung, 2014; Lisboa et al., 2013; Moreno et al., 2011; Rao et al., 2013; Raza et al., 2011; Tan et al., 2009; Wang et al., 2009) Although most studies show that diagnostic power is not any better than that achieved using a much simpler measurement of peripapillary RNFL thickness(Lisboa et al., 2013; Moreno et al., 2011; Rao et al., 2013; Tan et al., 2009) – and notwithstanding that macular measurements will generally require longer scan times and more involved segmentation – there is likely an important role for macular imaging in clinical glaucoma management, at least in detection of damage missed by standard visual field and OCT testing or confirmation of otherwise questionable defects.(Hood et al., 2013; Hood et al., 2014; Leung, 2014) Like that of the human retina, RGC density also increases dramatically near the fovea of nearly all other primates, so OCT imaging of the macula also proves useful for detection of glaucomatous damage in NHP EG models.(Luo et al., 2014)[Wilsey et al., Invest Ophthalmol Vis Sci 2015;56: E-Abstract 638] Luo and colleagues showed that localized measurements of macular inner retinal thickness (combined RGC+IPL) were correlated with multifocal ERG measurements of RGC function in a NHP EG model.(Luo et al., 2014) Similarly, our group has reported that macular retinal structural and functional losses are correlated and specific to RGCs over a wide range of severity in NHP EG (Figure 5). Consistent with an earlier observation by Ishikawa et al,(Ishikawa et al., 2005) we also found that outer retinal thickness in the macular was slightly increased in NHP EG, though that subtle effect was only weakly related to EG severity.[Wilsey et al., Invest Ophthalmol Vis Sci 2015;56: E-Abstract 638]

Figure 5. Macular retinal layer thickness parameters obtained from SDOCT radial scans under manometric IOP control.

Individual example of an NHP EG eye imaged at baseline (A, A’) and at its final time point (B, B’) just prior to sacrifice when glaucomatous damage was severe (as determined by complete orbital optic nerve axon counts). One of 48 B-scans (vertical orientation) from a radial pattern centered on the fovea is shown without segmentation in A and B, with layer segmentations in A’ and B’. This example is representative of the pattern of results observed across N=21 NHP in this study: there was clear, consistent loss in EG eyes of macular nerve fiber layer, ganglion cell layer and inner plexiform layer thickness, along with a small but significant increase in this thickness of both the outer plexiform layer and a layer defined as the outer nuclear (photoreceptor cell) layer plus inner segments and Henle’s fibers, with no change in outer segment length (defined as being from the reflectivity band at the inner segment-outer segment junction to the band at the cone outer segment tips, “COST line”). Macular inner retinal layer thickness losses were strongly correlated with peripapillary RNFL thickness in EG eyes but the subtle increase of outer retinal layer thickness was only weakly correlated with loss peripapillary RNFL thickness. Macular inner retinal layer losses were correlated with loss of function specific to retinal ganglion cells derived from multifocal electroretinography, while function of distal retinal elements was unchanged relative to fellow control eyes and un-related to peripapillary RNFL thickness loss. From Wilsey et al., Invest Ophthalmol Vis Sci 2015;56: E-Abstract 638.

Because layer-by-layer segmentation is more challenging for the thinner inner retina of small rodents, it has been common for investigators to evaluate either total retinal thickness or the combined RNFL+RGC+IPL thickness (often referred to as the ganglion cell complex, GCC) using OCT measurements in these species.(Fortune et al., 2011; Gabriele et al., 2011; Gramlich et al., 2014; Guo et al., 2010; Liu et al., 2014; Lozano and Twa, 2013; Nakano et al., 2011; Yang et al., 2012b) OCT measurements of specific inner retinal layer thickness (such as RNFL) in rodents has generally been limited to isolated spatial samples or to single circular B-scans around the ONH.(Abbott et al., 2014a; Abbott et al., 2014b; Chauhan et al., 2012; Choe et al., 2014; Fortune et al., 2011; Guo et al., 2010; Nagata et al., 2009)

However, anticipating improved accuracy and speed of forthcoming automated segmentation algorithms, such as recently demonstrated by Antony et al(Antony et al., 2013) and Srinivasan et al,(Srinivasan et al., 2014) it should be soon be feasible to address persisting questions about glaucoma pathogenesis in EG models (perhaps also clinically in humans) using in vivo imaging. For example, one aspect of glaucoma pathogenesis receiving renewed attention is RGC dendritic atrophy.(Chong and Martin, 2014; Leung et al., 2011; Li et al., 2011; Morgan, 2012) Weber and colleagues(Weber et al., 1998) first reported on the basis of cross-sectional data that degenerative retraction of the dendritic arbor preceded thinning of axon caliber or shrinkage of the RGC soma in a NHP EG model, though these findings all occurred well after deformation of the ONH, consistent with the hypothesis that initial RGC injury occurs within the ONH with subsequent degenerative morphological changes. Shou et al subsequently reported similar cross-sectional observations of dendritic atrophy in a feline model of EG.(Shou et al., 2003) More recently, Williams et al(Williams et al., 2013) reported evidence that RGC dendritic atrophy occurs prior to significant axon degeneration in the DBA/2J mouse model of heritable glaucoma, again on the basis of cross-sectional data. If dendritic atrophy occurs during early stages of the glaucomatous RGC injury in humans, its detection by clinical imaging could serve as a warning sign to advance therapy (IOP lowering or other novel mechanism). Similarly, as with the other target structures discussed in this review, in vivo imaging also can be applied in EG models for hypothesis testing and for refining relevance to human glaucoma. Particularly interesting within the context of imaging RGC dendritic integrity in vivo is the recent demonstration by Tanna and colleagues(Tanna et al., 2010) that the improved axial resolution afforded by a commercial broadband SDOCT instrument was sufficient to detect sublamina within the IPL. Meanwhile, evidence based on current clinical OCT imaging is equivocal in terms of whether the phenomenon of dendritic atrophy occurs in human glaucoma and/or whether it can be detected prior to thinning of RGC axon bundles or loss of RGC soma.(Hood et al., 2013; Leung, 2014; Lisboa et al., 2013; Moreno et al., 2011; Rao et al., 2013; Tan et al., 2009).

Nakano et al found by SDOCT thinning of the GCC was delayed relative to the decrease in RGC soma density after NMDA-induced injury in mice. Choe et al reported a similar delay between RNFL thickness and RGC density in the first week after optic nerve transection in rats.(Choe et al., 2014) Using longitudinal CSLO imaging in vivo, Leung et al have shown numerous examples of RGC somas persisting longer than their axons after optic nerve crush.(Leung et al., 2011) The report by Leung et al also demonstrated clearly that dendritic tree shrinkage was consistently the earliest morphological change detectable by in vivo imaging after crush injury.(Leung et al., 2011) Few studies of EG based on chronic IOP elevation have specifically addressed this question. Guo et al observed greater loss of outer retinal layer thickness than for RNFL, IPL or GCC in a rat EG model. Our group has observed in a NHP EG model that loss of macular inner retinal layer thickness is correlated with loss of peripapillary RNFL thickness, but a slight increase in macular outer retinal layer thickness was unrelated to EG severity (Fig. 5). Future studies aim to evaluate the temporal sequence of inner retinal layer changes as compared with ONH changes and axon loss.

While OCT measurements of individual retinal layer morphology and thickness can be done both in the clinic and laboratory, it is not yet possible to image individual RGCs in human eyes. In the laboratory, however, various techniques for visualizing RGCs in vivo have evolved over the past two decades. Generally such techniques involve imaging by either epifluorescence microscopy,(Naskar et al., 2002; Thanos et al., 2002) fundus photography(Murata et al., 2008), intravital confocal microscopy (Sabel et al., 1997; Thanos, 1991; Walsh and Quigley, 2008) or CSLO(Abbott et al., 2013; Chauhan et al., 2012; Choe et al., 2014; Cordeiro et al., 2004; Geng et al., 2012; Geng et al., 2009; Gray et al., 2008; Higashide et al., 2006; Kanamori et al., 2012; Leung et al., 2008a; Leung et al., 2008b; Leung et al., 2011; Nakano et al., 2011) after introduction of a fluorescent tracer via retrograde transport from the midbrain(Abbott et al., 2013; Choe et al., 2014; Gray et al., 2008; Higashide et al., 2006; Kanamori et al., 2012; Murata et al., 2008; Prilloff et al., 2010; Sabel et al., 1997; Thanos, 1991; Thanos et al., 2002) or a fluorescent reporter molecule whose expression is driven by a promoter that is relatively specific to RGCs.(Chauhan et al., 2012; Geng et al., 2012; Geng et al., 2009; Leung et al., 2008a; Leung et al., 2008b; Leung et al., 2011; Murata et al., 2008; Nakano et al., 2011; Walsh and Quigley, 2008) In some transgenic lines such as the Thy-1 YFP mouse, expression is limited to a small enough proportion of RGCs to enable visualization in vivo of even fine dendritic structure by CSLO.(Leung et al., 2011; Walsh and Quigley, 2008) Use of SLO with AO to compensate for optical aberrations has also enabled visualization in vivo of RGC fine dendritic structure.(Geng et al., 2012; Geng et al., 2009; Gray et al., 2008) Though several of these previous studies have demonstrated decline in RGC density, changes in RGC soma size, dendritic arbor and axonal integrity, or RGC apoptosis after optic nerve injury,(Chauhan et al., 2012; Choe et al., 2014; Cordeiro et al., 2004; Higashide et al., 2006; Kanamori et al., 2012; Leung et al., 2008b; Leung et al., 2011; Murata et al., 2008; Nakano et al., 2011; Naskar et al., 2002; Prilloff et al., 2010; Sabel et al., 1997; Thanos et al., 2002) none has yet to evaluate longitudinal changes in an EG model that adequately recapitulates the human disease. It is anticipated that such studies will emerge in the near future.

4. Future Directions

As reviewed above, one distinct advantage of OCT over other imaging modalities is its superior axial resolution. However, one advantage that SLO (confocal or flood illuminant) imaging has over OCT is the capability for fluorescence detection, which can be combined with molecular tags or reporters to reveal fine structure or pathological events. During the past decade, interest has grown in the development of contrast agents for enhancement of OCT signals with the goal of targeting specific molecular markers that could transform diagnostic and/or therapeutic approaches.(Boppart et al., 2005; Gordon and Jayagopal, 2014; Jung et al., 2011; Troutman et al., 2007) Gold nanorods (GNRs) have emerged as an excellent candidate due to the biocompatible nature of gold, the ability to functionalize gold particles for covalent binding to specific ligands for molecular targeting and to their inherent physical property of surface plasmon resonance, which can be readily tuned to a specific frequency by controlling rod dimensions during synthesis.(Troutman et al., 2007) Safe and effective use of GNRs in vivo as a contrast agent for OCT has already been demonstrated using photothermal OCT.(Jung et al., 2011; Tucker-Schwartz et al., 2014; Tucker-Schwartz et al., 2012) Wang et al reported use of GNRs in an ex vivo ocular perfusion system to visualize otherwise silent flow fields within the anterior chamber by Doppler OCT.(Wang et al., 2014a) De la Zerda et al reported visualization of signal enhancement by reflectance OCT in vivo after injecting GNRs into the corneal stroma in mice.(de la Zerda et al., 2014) Sandrian et al(Gabriele Sandrian et al., 2012) attempted to introduce into the retina GNRs conjugated to the Thy1 antibody for targeted binding to RGCs, but found that intravitreally injected GNRs became trapped within inflammatory cells and networks within the vitreous; GNRs were not observed within RGCs and the retinal OCT signal was severely attenuated. Thus an important obstacle toward realization of this goal is the ability to get GNRs to cross the retinal internal limiting membrane (ILM) in order to achieve targeted retinal cell binding. Once realized, multimodal imaging approaches such as combined AO-SLO and AO-OCT(Felberer et al., 2014; Zawadzki et al., 2011) could be used in conjunction with combined constructs for contrast enhancement and molecular tagging (such as GNRs conjugated with fluorescent antibodies) to visualize pathophysiological changes in vivo at the molecular level. Other recent exciting work has demonstrated that light responses from individual RGCs can be quantified by in vivo imaging in the living mouse and macaque eye.(Yin et al., 2013; Yin et al., 2014) Other multimodal approaches such as Multifunctional OCT(Hong et al., 2014) or combining polarization-sensitive OCT with AO(Cense et al., 2009) offer further promise for hypothesis testing and clinical diagnostics. Finally, the speed of OCT acquisition continues to increase rapidly, reaching extraordinary capacity such as that recently demonstrated by Klein et al.(Klein et al., 2013) These and other developments should surely facilitate future discovery of key pathogenic events in glaucoma and other diseases of the eye as well as targets for novel therapeutic intervention.

5. Summary and Conclusions

The phrase “bench-to-bedside” is often used to describe the translation of laboratory discovery into clinical care, for the enterprise of improving understanding of human health and disease through the use of experimental models. Technologies for in vivo ocular imaging are one important aspect of conducting such studies. They provide outcome measures such as anatomical parameters that can be used in both settings: laboratory studies in animal models and clinical research in human patients. Indeed, it is possible that failure to deliver a viable option for neuroprotection in glaucoma is due in part to the lack of commonality between the outcome measures used in pre-clinical studies in animals (typically post mortem histological measures such as RGC counts) and those acceptable for clinical research and randomized clinical trials. For example, maintaining RGC soma counts in an animal model may not translate to preservation of vision if surviving RGCs do not each maintain a viable, functioning axon that correctly conveys signals to its central target.(Raff et al., 2002; Whitmore et al., 2005) Moreover, without common outcome measures it is difficult to compare the pathophysiological sequence of an experimental model with the human disease, to ensure relevance and/or refine any given model accordingly. Thus by incorporating in vivo imaging into laboratory studies, the glaucoma research community can not only reduce the number of animals needed to achieve robust outcomes, but also refine and prove relevance of EG models to hasten the translation of discovery from bench-to-bedside. Ironically, most of the recent spectacular achievements in ophthalmic imaging have been developed in for use in human beings and need to be adapted for use in laboratory animals, rather like bedside-to-bench translation. Yet in order to extend the limitations of current imaging paradigms to provide adequate contrast, spatial and temporal resolution for assessment of cellular and subcellular structure and function of RGCs in vivo and to identify specific cellular and molecular targets for imaging in vivo, future translational research will need to occur in both directions.

• The most common techniques in current use for in vivo assessment of ocular structure in animal models of experimental glaucoma include optical coherence tomography and scanning laser ophthalmoscopy.

• These and other imaging modalities can be used in clinical studies as well as in laboratory animal studies of glaucoma models to quantify structural parameters of the optic nerve head, retinal nerve fiber layer, retinal ganglion cell soma and dendrites.

• Incorporating in vivo imaging into laboratory studies, the glaucoma research community can not only reduce the number of animals needed to achieve robust outcomes, but also refine and prove relevance of EG models to hasten the translation of discovery from bench-to-bedside.

• Recent developments hold extraordinary promise for revealing important events in glaucoma pathogenesis in living eyes as well as for identification of specific cellular and molecular targets for imaging in vivo.