Giving invertebrates an eye exam: an ophthalmoscope that utilizes the autofluorescence of photoreceptors

One of the most important features of any eye is its ability to correctly focus light onto the retina, a process that requires tight coordination between the optical properties of the lens and the spacing between the lens and retina. Accordingly, many studies exist in vertebrates that investigate how correct focusing is established and maintained (Hung et al., 1995; Kroger and Wagner, 1996; McFadden et al., 2004; Schaeffel et al., 1988; Sivak, 2008). However, few studies have explored how invertebrate eyes are focused, with the most notable studies having been performed on the image-forming eyes of spiders (Land, 1969b), larval stemmata (Mizutani and Toh, 1995; Stowasser and Buschbeck, 2014), insect ocelli (Berry et al., 2007; Taylor et al., 2016) and the ommatidia of fly eyes (Franceschini and Kirschfeld, 1971; Kirschfeld and Franceschini, 1968). However, the majority of refractive state assessments in these animals used indirect methods that are difficult to conduct and suffer from some inherent inaccuracies. In addition, studies on how correct focusing is established are almost entirely absent in invertebrates, though it recently has been shown that in cephalopods (Turnbull et al., 2015), like in vertebrates (Flitcroft, 2013; McBrien and Barnes, 1984; Wallman and Winawer, 2004), this process requires visual input. Considering the many molecular (Cook and Desplan, 2001; Cook et al., 2011; Freund et al., 1996), functional (Land and Nilsson, 2012) and neurological (Sanes and Zipursky, 2010) similarities between vertebrate and invertebrate eyes, there is a strong need for such studies in invertebrates, but these require tools that rapidly and accurately assess refractive properties in very small eyes. Such tools are currently unavailable, as conventional methods do not work on eyes at the scale found in most invertebrates.

Nearly 50 years ago, Mike Land developed an ophthalmoscope (Land, 1969a,b) to directly measure refractive properties of jumping spider eyes. Although this tool was limited to eyes with tapeta or other reflective structures that return incoming light, it has greatly enhanced our understanding of invertebrate optics. In addition to the work on jumping spiders, it has been used to assess optical parameters of the spider Cupiennius salei (Land and Barth, 1992) and to visualize the retina of other spiders (Nilsson et al., 1992). It also was used to measure the resolving power of butterflies (Land, 1984, 1990) and hummingbird hawkmoths (Warrant et al., 1999), to investigate the eye shine of butterflies (Stavenga, 2002) and the pupil reflex mechanism of insect apposition eyes (Nilsson et al., 1992; Nordstrom and Warrant, 2000). However, perhaps because of its restriction to reflective eyes, this method seems to have been largely forgotten since then. To develop a more widely accessible tool, we re-created and modified the ‘Land-ophthalmoscope’ so that it can be applied to a wider range of invertebrate eyes. Relying on the autofluorescence of photoreceptors (Franceschini, 1977; Franceschini et al., 1981; Kruizinga and Stavenga, 1990; Stark et al., 1977; Stavenga, 1983) instead of tapetal reflections, our micro-ophthalmoscope is suitable to measure eyes ranging from the relatively large eyes of jumping spiders to the very small individual ommatidia of compound eyes of flies. The autofluorescence of photoreceptors has previously been used to investigate opsin properties (Franceschini, 1983) and as an in vivo screening tool for Drosophila mutants (Pichaud and Desplan, 2001). However, to the best of our knowledge, this is the first time it has been utilized to measure the refractive state of invertebrate eyes.

To demonstrate the versatility and power of our micro-ophthalmoscope, we show that this tool allows us to independently image and measure the two deepest layers of the anterior medial (AM) eye of Phidippus audax jumping spiders. In fact, our tool is so sensitive that it can even resolve the relatively minor refractive differences across the medial–lateral extent of the first layer, which is organized as a ‘staircase’ (Land, 1969b) so that photoreceptors are staggered between their medial and lateral extent. In addition, our methods allowed us to generate complete high-resolution retinal images of these spiders' anterior lateral (AL) eyes. Rhabdomere tips of flies also have been successfully imaged in the less-reflective compound eyes (Franceschini, 1975). To demonstrate that our imaging method extends to compound eyes, we evaluated the refractive state of individual ommatidia of flesh flies. Our results demonstrate that our micro-ophthalmoscope has great potential to serve as a valuable tool to visualize the retina, and to investigate the refractive state of live invertebrates with different eye types, even in the absence of a reflective structure. Given the large diversity of invertebrate eye types, and the availability of sophisticated molecular tools in Drosophila, we anticipate that this tool will allow researchers to address a large body of open questions relating to invertebrate eye development and function, at both the molecular and the organismal level.

The micro-ophthalmoscope is composed of two achromatic lenses (top lens: f=150 mm, AC254-150-A; accessory lens: f=200 mm, AC254-200-A; both Thorlabs, Newton, NJ, USA), an objective (4× UplanFL, 10× UPlanFl in the case of the jumping spider, 40× UPlanFl in the case of the flesh fly and 50× UMPlanFl in the case of Drosophila melanogaster; Olympus, Central Valley, PA, USA), and a light-sensitive camera (ORCA-Flash 4.0 V2 Digital CMOS, C11440-22CU, Hamamatsu, Hamamatsu City, Japan) with imaging software (HCImageLive 4.3.1.30, Hamamatsu). The accessory lens is mounted on an optical linear scale with digital readout (EL10 DRO and GS212 scale, Electronica Mechatronic Systems Pvt Ltd, Pune, Maharashtra, India) the display of which is mirrored into the video frame. Illumination consists of a mercury light illuminator for fluorescence imaging (ULH100HG and BX-RFA, Olympus) equipped with an Olympus Texas Red filter cube (Olympus, U-N31004) and a DAPI filter cube (Olympus, U-N31000), each containing a dichroic mirror. Although we did not specifically optimize these filter cubes for our organisms, it has been described for a variety of invertebrate photoreceptors that their metarhodopsin leads to red emission when excited with blue or green light (Cronin and Goldsmith, 1981; Franceschini et al., 1981; Juse et al., 1989; Stavenga, 1983).

The illumination unit also contains an adjustable aperture and field stop (modified to close the iris further) and neutral density filters (Fig. 1A,B). A motorized micromanipulator (DC3K with MS 314, Märzhäuser, Wetzlar, Germany) allows for the precise positioning of the animals.

The microscope portion of the ophthalmoscope (Fig. 1A) is set up so that an object placed in the focal plane of the objective is in focus of the camera (initially the lens of an animal eye). The accessory lens is then added (Fig. 1B) which allows imaging of the retina. For each objective, the accessory lens has a zero reference position, which yields a focused image of an object at infinity, and thus it is the accessory lens position that yields a focused image of the retina if an eye is focused for infinity (i.e. emmetropic). If the accessory lens needs to be shifted towards the objective to visualize the retina, the eye is myopic, and if it has to be shifted away from the objective, it is hyperopic. To perform quantitative assessment of the refractive error, the correlation between accessory lens shift (the shift from the zero reference position) and in-focus object distance needs to be established for each objective (Fig. 1C). Resulting values are used to fit the function f(x)=f²/x where f is the focal length of the objective, which also describes the relationship in the myopic domain. This correlation in conjunction with an independent focal length measurement of the eye's lens then allows calculation of the refractive state of the eye (see below). Fig. 1D illustrates how the 10× and 40× objectives lead to differently sized sharp images of an object (the sailboat image) placed at effective infinity (∼1.5 m away), with the accessory lens at its respective zero positions.

To test the ophthalmoscope, we purchased juvenile jumping spiders, Phiddipus audex (Hentz 1845), from Phids.net (West Palm Beach, FL, USA) and flesh fly, Sarcophaga bullata (Parker 1916), pupae from Carolina Biological Supply Company (Burlington, NC, USA). We imaged each spider around maturation (N=5, cephalothorax length was between 2.7 and 4.2 mm) and each flesh fly post-eclosure (N=7). To demonstrate that the same technique also can be applied to D. melanogaster, we also imaged selected ommatidia of that species. For each measurement, the animal's lens was first brought into focus without the accessory lens, then the accessory lens was inserted and an image series of the retina was taken while shifting the accessory lens in both directions beyond the range of focus.

For each spider, we took image series of the retina of: (a) the AM eyes with the 10× objective (which was used to assess the refractive state) and with the 4× objective (for high magnification images of the retina) while the animal was anesthetized with CO₂ to prevent its retina from moving, and (b) the AL eyes with the 10× objective. To construct retinal maps of entire AL eyes (N=3), we took a series of images (at least 40 images per map) with the 10× objective while tilting and rotating the spider. Later, these images were assembled in Photoshop (Adobe, San Jose, CA, USA).

Imaging flies is somewhat tricky, because if more than one ommatidium is illuminated then the position of the deep pseudopupil is measured rather than the refractive properties of the ommatidia. To demonstrate this, we took image series (with the 40× objective) of the retina of S. bullata in the frontal region of their compound eyes while illuminating: (a) a single ommatidium and (b) multiple ommatidia. To recover autofluorescence, the eye was intermittently illuminated with UV light from an Olympus filter cube typically used for DAPI imaging. To demonstrate that such measurements also are possible in the very small D. melanogaster, we took images of their retina with the 50× objective.

To find the range of accessory lens positions that yielded focused images of the retina, each image series was blindly evaluated for best focus by two to three people. Corresponding object distances were calculated from the established correlation between accessory lens shift and object distance. These object distances, in conjunction with independent focal length assessment (see below), allowed us to determine the eye's refractive state quantitatively in terms of the distance between the surface of the photoreceptors and the focal plane of the eye's lens. This relationship is expressed as Δf′=(f′)²/(n′o−f′), where Δf′ is the distance between the surface of the photoreceptors and the focal plane of the eye, f′ is the focal length of the animals' lens inside the eye (measured from the principal plane), n′ is the refractive index behind the lens (assumed to equal that of water; Land, 1969b; Williams and McIntyre, 1980) and o is the object distance (after Land, 1969b). Note that the focal length (f′) in the denominator can be omitted if the object distance is much larger than the focal length, which is generally true unless the eye is very myopic or hyperopic. This equation was derived from the equation n′/f′=n/o+n′/i, where n is the refractive index of the medium in front of the lens (in our case air), n′ is the refractive index behind the lens (in our case assumed to equal that of water), f′ is the focal length behind the lens, o is the object distance and i is the image distance.

The focal length of the animal's lens was calculated from the image magnification (Land, 1969b; Land and Nilsson, 2012) following ‘hanging-drop’ procedures previously described (Homann, 1924; Stowasser et al., 2010). In brief, the isolated lenses were mounted with wax on top of Ringer solution (O'shea and Adams, 1981). To determine the image magnification, we used a compound scope for which we removed the condenser and took images through the animal's lenses (using a 10× objective for spiders, and a 20× objective for flies) of an object (USAF 1951 negative test target from Edmund Optics, Barrington, NJ, USA) that was placed at a distance from the animal's lens of 12.2 cm for spiders and 4.3 cm for flies. From the image magnification of the best-focused image, we then calculated the focal length behind the animal's lens (f′). For spiders, one of the AM and AL eyes was measured with both green and red light to evaluate chromatic aberration. For each flesh fly, we established the focal length of five randomly chosen ommatidia of the frontal region of one eye and used the other side of the head for histology.

To demonstrate the applicability of the micro-ophthalmoscope, we focused our analysis on relatively well-understood optical systems to show that this method can replicate and even expand upon existing knowledge on their optics. The eyes we tested varied greatly in size and in the focal length of their lenses.

First, we investigated jumping spiders, the AM eyes of which are known to have multiple layers (Land, 1969b). We could independently image their layers I and II and even resolve the staircase organization of layer I (Fig. 2A,B), a schematic drawing of which is shown in Fig. 2C, inset. The staircase is discernible in that, in layer I, the lateral portion of the boomerang comes into focus before the medial portion as the accessory lens is adjusted. A higher magnification image illustrates how individual photoreceptors are well resolved in the periphery of the boomerang (Fig. 2B). We also imaged the retina of the AL eye and could generate entire retinal maps (Fig. 2C). In conjunction with focal length measurements of the lenses of AM and AL eyes, we quantified the refractive state of these eyes. In agreement with what has already been demonstrated for jumping spiders Plexippus (Blest et al., 1981), the difference between the focal length measured with red light (mean±s.d., AM 394.3±26.0 μm, AL 180.5±14.2 μm) and green light (AM 393±29.8 μm, AL 179.2±12.7 μm) was negligible (N=5) and hence we proceeded with optical calculations without correcting for chromatic aberration. For other lenses, calculations may have to be corrected based on the color-specific focal length measurements as established above, as photoreceptor autofluorescence typically produces red emission. Note that the optical components of the ophthalmoscope itself are very well corrected for chromatic aberration, as we could find no noticeable focus difference with different light illuminations. Consistent with previous findings (Land, 1969b), our measurements suggest that layer II of AM eyes is hyperopic, and layer I of the spider's AM eyes is situated around emmetropia. Specifically, we found that it was the medial portion of the staircase of layer I that was positioned around emmetropia (Fig. 2D), and that, in the same set of spiders, the range of focus of AM eyes was quite narrow when compared with that of their AL eyes (Fig. 2D). Our analysis further suggests that the lateral portion of the staircase is distinct from that of the medial portion, and provides sharp images into the near-field and striking range of the animal. It thus appears that the P. audax AM is well set up to obtain object distance-dependent photoreceptor activation within the staircase, which could be an important distance vision mechanism. Note that in Hasarius adansoni, image defocus is already thought to be important in that regard (Nagata et al., 2012). The central region of the retina of AL eyes was essentially emmetropic (Fig. 2E), with a comparatively wide range of depth of focus. Taken together, in regards to the spiders, we could easily resolve the refractive state of distinct retinal layers, and could even discern how portions of layer I are differentially focused.

For flies (Fig. 3F), visualization of the retina of both individual ommatidia (Fig. 3A,B,I) and multiple ommatidia (Fig. 3C,D) resulted in a clear image of the photoreceptor trapezoid. In the latter case, images of multiple retinas overlapped and hence we imaged the deep pseudopupil located in the center of the curvature of the eye (Fig. 3E), rather than individual receptors. The average focal length of S. bullata ommatidia was 50±7.4 μm (mean±s.d., N=7). The single ommatidium was essentially emmetropic and had a wide range of focus (Fig. 3G,H). This confirms that fly ommatidia are focused at infinity (Kirschfeld and Franceschini, 1968). Beyond what had been known, we found that (as might be expected for optimal sampling) the center of focus in S. bullata lies a few micrometers below the distal tips of the photoreceptors (Fig. 3H), but its depth of focus is large enough to allow ommatidia to see into infinity (Fig. 3B,G). In contrast, when we imaged the deep pseudopupil, the perceived depth of focus was extremely tight (Fig. 3D,G,H). This, for the first time, confirms through direct measurements that the pseudopupil lies near the center of curvature of the eye, as had already been inferred by indirect methods and theoretical considerations (Franceschini, 1972).

Our results show low variation and the method is a direct in vivo assessment that allows animals to survive seemingly unharmed. In addition to measuring the refractive state, our methods allowed us to evaluate the depth of focus, an important optical property that has received little attention despite its potentially profound functional implications. Our method takes all optical components into account, and thus promises to be more reliable than any of the indirect methods, opening the possibility to investigate many heretofore-inaccessible questions. These include, for example, molecular-genetic manipulations aimed at elucidating mechanisms that are involved in allowing eyes to establish correct focusing, or whether and how focusing is related to the circadian rhythm (Eckert, 1968; Horridge and Giddings, 1971; Menzi, 1987). As has been previously demonstrated (Land, 1984, 1969b), the ophthalmoscope also can be used to measure the receptor resolution by projecting an object of known angular extent (such as the aperture stop of the illuminator) onto the retina. As the method is not invasive, it can even be applied to longitudinal studies. Hopefully, the versatility of this new tool will facilitate many additional discoveries in regards to the amazingly diverse optics (Land and Nilsson, 2012) of the world of invertebrate eyes.

We thank members of the Buschbeck laboratory for helpful discussions and for feedback on the manuscript, and Winter Partin for assisting us with ophthalmoscope calibrations.