Photochemical damage to the retina occurs for prolonged exposures of intense light. Two action spectra exist for this phenomenon. In rat an action spectrum matching the absorption spectrum of rhodopsin was found. In macaque, the susceptibility for photochemical damage decreased continuously from the UV to long visible wavelengths. Later, such a spectrum was also found in rat. In search for critical parameters that determine the shape of the spectrum we gathered all available data on the damage threshold dose for monochromatic radiation and noted the experimental conditions. The rhodopsin action spectrum was found in two sources; the other 16 sources adhered to the short wavelength spectrum. Comparing the conditions we conclude that the critical parameters for the generation of either action spectrum remain elusive. Experiments are suggested to resolve this issue and fill a few gaps in our knowledge.
Until the first half of the twentieth century it was common knowledge that light could damage the retina only through a thermal mechanism. A small dent in this framework appeared in 1962 when Vos (1) calculated that prolonged staring in the sun, the kind of light damage known since ancient times, caused a local temperature rise of only 2°C. He therefore postulated a new mechanism, “metabolic poisoning.” The breakthrough came a few years later when Noell et al. published an extensive study on what is now known as photochemical damage of the retina (2). Freely moving rats exposed to green filtered light for periods up to 2 days showed extensive functional and histological damage at levels far too low to cause a significant rise in retinal temperature. Part of this study was devoted to the action spectrum of damage. It was found to be similar to the rhodopsin absorption spectrum (RAS). A decade later Ham described a quite different action spectrum in macaques. For visible light (3), and later also for UV radiation (4), susceptibility decreased monotonically from short to long wavelengths (short wavelength action spectrum, SAS). The nomenclature for this kind of damage also includes “blue light hazard,”“actinic damage” and “photic injury.”
The different action spectra were at first attributed to different species, i.e. rats versus macaques, but in the last decade of the twentieth century several papers showed that rats, studied in similar conditions as macaques, also exhibited an action spectrum with decreasing threshold dose towards the short wavelengths (5–8). Ham’s action spectrum served as the base for international standards for the protection against retinal damage by lasers and other light sources (e.g.9–12). Noell’s spectrum was not incorporated because it was thought to apply exclusively to rodents. In light damage research the roles were reversed, with Noell’s spectrum often assumed to be the only one applicable because of rhodopsin’s all important role in the mechanism of light damage, and because this area of research is dominated by rodent animal models (13).
An action spectrum is defined as the efficiency with which electromagnetic radiation produces a photochemical reaction plotted as a function of the wavelength of the radiation. In retinal research the efficiency of the photochemical reaction was operationally defined in various ways such as a specified loss in the amplitude of the electroretinogram a- or b-wave, a just visible funduscopic change, or a just visible microscopic or electron microscopic change in a specific structure of the retina.
Several attempts have been made to define the critical conditions for each action spectrum (14–19), sometimes identified as class 1 and 2. The RAS seemed at least valid for freely moving rodents in long (>24 h) exposures to relatively low retinal irradiances (<1 mW cm−2) while the SAS was generally associated with high irradiance levels, a limited time frame (up to hours) in anesthetized animals. A major problem, though, with such a scheme is that Noell et al. (2) actually measured their action spectrum in anesthetized animals exposed to monochromatic light for only 1.5 h.
Knowledge of the action spectrum is essential for establishing guidelines for safe levels of light in normal daily life as well as ophthalmic practice. Interspecies comparisons are needed to translate experimental data from animal studies to the human situation. This paper is a new attempt to define the crucial parameters for the two action spectra. For that purpose we tried to gather all available monochromatic data on threshold dose as a function of wavelength, thereby significantly extending earlier attempts (e.g.15,17,20).
Selection and Presentation of Data
We searched the literature for in vivo monochromatic data, including lasers and incoherent light with bandwidths up to about 10 nm. Damage criteria included retinal or functional changes. To allow the use of single data points, we required that the retinal dose was provided, or could be calculated from the product of retinal irradiance and exposure time. In some cases corneal irradiance (Ecor) was converted to retinal irradiance (Eret) if pupil diameter and area of the exposed retina were provided:
where D is the diameter of the pupil, Aret the area of the irradiated retina, and τ the ocular media transmission. The latter is very relevant because some species, like primates, have a UV filter in their crystalline lens (cf. review ), while rats and mice, have not (22,23). Media transmission was converted to (22) in some early studies where rat spectral media transmission <450 nm was assumed to be similar to that of primates. When a Ganzfeld was used with a rat eye (approached as half a globe with focal length f as diameter) then the area of the retina was taken as Aret = 2πf2. To minimize interference with thermal effects exposure times of at least 100 s were selected. Photochemical damage was expressed in milliWatt per square centimeter (mW cm−2). It was supposed that the dose of radiant energy, i.e. the product of irradiance and time, sets the threshold for damage, expressed as radiant exposure in Joule per square centimeter (J cm−2).
Studies are discussed in chronological order, but occasionally data from the same laboratory published within a few years were clustered. Table 1 summarizes the exposure conditions and retinal endpoints.
Table 1. Summary of literature data on spectral photochemical thresholds. Sequence follows the text in the paragraph Dose Data.
First author, year of publication
Duration Retinal area
Endpoint of retinal damage Time of observation
DA = dark adaptation; LA = light adaptation; FVL = (just) funduscopic visible lesion.
Noell et al. (2) reported two experiments with monochromatic radiation, the first with an exposure time of 1½ h (experiment 1a), the second with 3 h exposures (experiment 2). Only relative irradiances were given but these may be converted to absolute levels by using a statement on p. 460 “… the minimally effective retinal irradiation for several hours is estimated to be about 1 μW per square centimeter for a wavelength of 5000 Å.” Minimally effective was most likely a permanent reduction in the a-wave of about 10%. For the action spectrum a criterion of 50% change in a-wave was used. Going from 10% to 50% change required an increase of log Intensity of the damaging light of 0.5, a factor 3 (fig. 5 of their paper). “Several hours” in this context was 3–4 h. A wavelength of 5000 Å was taken as similar to the 495 nm used in the action spectrum. Thus, we arrived at a retinal dose of 10−6 × (3.5 × 3600) × 3 = 0.038 J cm−2 at 495 nm. It should be noted that these experiments were carried out at elevated body temperature which reportedly (their tables I–III) decreased exposure time (thus threshold dose) by a factor 25. Threshold data for their experiments 1a and 2 coincided and were plotted with one symbol in Fig. 1a, which has all data on rodents.
Lawwill et al. (24) reported that in rabbits for 514.5 nm “the sharp change in dose/response curve falls between 0.02 and 0.03 Watt per square centimeter.” With 4 h exposure this leads to a threshold dose of 4 × 3600 × 0.025 = 360 J cm−2. The data point is plotted in Fig. 1b, which has all data on nonrodents. Follow-up studies in macaques with wavelengths in the range 450–600 nm were published in 1977 (25). Blue light thresholds (458 nm) were found lower than green. Data show a fairly large experimental spread, with the range from minimal to maximal damage occasionally covering more than 3 log units. The histologic thresholds in a 1982 paper (p. 566, 26) based on the same experiments, were somewhat different. We plotted the most recent interpretation.
In 1976 Ham et al. (3) published a study of spectral photochemical damage in the retina of primates. They calculated local retinal temperatures and showed that in the envelope of wavelengths shorter than 514 nm and exposure times longer than 100 s, the raise did not exceed 4°C. Thus, a distinction could be made with thermal injury. The threshold dose monotonically increased from 441.5 to 1064 nm. Data for 100 and 1000 s are plotted in Fig. 1b. At 580 nm thresholds might have been lowered by thermal enhancement. In 1982 Ham et al. (4) extended the wavelength range to 325 nm with three aphakic macaque eyes. In the UV damage was immediately visible. Different wavelengths caused different types of damage (4,27). In the UV damage to the photoreceptors (PRs) was extensive and depigmentation occurred in the retinal pigment epithelium (RPE). In the visible only a few PR showed damage, but the RPE was swollen with many inclusions.
Williams and Howell (28) exposed albino rats for 6 h to a number of narrow band (10 nm) wavelengths in the range 440–590 nm, each with a constant corneal flux of 2 × 1015 photons s−1 cm−2. For 485 nm the given photon flux translates to 0.82 mW cm−2. With 6 h exposure the corneal dose varied from 19 J cm−2 at the shortest wavelength to 14 J cm−2 at the longest. To save time the authors first measured the loss of the thickness (compared with that in the control eye) of the outer nuclear layer (ONL) in a sensitive area 1.88 mm superior to the optic nerve. When “no reasonable spectrum was obtained,” they reverted to a more elaborate criterion, integrating the ONL thickness across the vertical meridian. The cause of the variation in the sensitive area was not analyzed and the size of the experimental error was not given. To arrive at a retinal action spectrum we assumed a pupil of 5 mm diameter like in (5) and a retinal field of 8 mm diameter (their fig. 1); the ratio between corneal and retinal irradiance then becomes 0.4. Next, we took the dose of 14.8 × 0.4 = 5.92 J cm−2 at 567 nm as a starting point because it evoked a minimal loss in ONL thickness. Thereafter, the ONL losses at other wavelengths were converted to a log intensity change relative to 567 nm with the help of the dose response curve in fig. 5 of Noell et al. The result of the conversion of William and Howell’s data is plotted in Fig. 1a.
The two data points on aphakic squirrels (29), published by Collier et al. in 1989 should be viewed with some caution because in another publication on the same material, also published in 1989, they state “Light measurements were made to determine the level of energy incident on the cornea” (30), but a few pages later they compare their dose with Ham’s retinal one (4). The histopathology as a function of wavelength resembled that in the macaque (4,27).
In the 1990s two groups (5–7) subjected anesthetized rats to ultraviolet as well as visible radiation and both found highest susceptibility at the shortest wavelengths. Monochromatic data (5) are plotted in Fig. 1a, corrected for proper transmission (22). Additional data were later measured with a different setup and a different strain of pigmented rats (8,31), and with albino (32) rats. Only the albino data from the latter publication were plotted because they nearly coincided with the pigmented data. The histologic data (8) showed two different spectral types, one in the range of 320–440 nm, the other in the range of 470–550 nm, again resembling macaque data (4,27).
The susceptibility of impairment of the blood brain barrier to blue light was studied in rabbits by Putting et al. (33), yielding a dose of about 30 J cm−2 for a funduscopic threshold change 2 days after exposure to 418 nm. Chen (34,35) subjected rats to 404 nm light. A dose of 11 J cm−2 led to temporary changes in cytochrome oxidase activity, but was subthreshold for funduscopic or morphologic changes. A dose of 38 J cm−2 was clearly suprathreshold for funduscopic change. We took 20 J cm−2 as the threshold for funduscopically visible damage, but plotted 60 J cm−2, correcting for the much higher (0.6) than assumed (0.2) media transmission (22). Grimm et al. (36) exposed anesthetized rats to a Ganzfeld illumination at 403 and 550 nm with a corneal irradiance of 3.1 and 8.7 mW cm−2 respectively. We calculated the corresponding retinal irradiances with formula 1 and parameters pupil diameter 5 mm, f = 0.525 cm, τ = 0.58 (403 nm) and 0.88 (550 nm), and Aret = half globe. With a 30 min exposure at 403 nm threshold damage (vesiculation of outer segments) was seen (retinal dose 1.47 J cm−2). An exposure for 120 min at 550 nm resulted in no damage (dose 23.3 J cm−2). With a different (spot) illumination threshold at 403 nm was reached with 1/3 of the dose, but again no damage was seen at 550 nm. Supposing equal light distribution for both wavelengths it may be estimated that a dose of 70 J cm−2 at 550 nm caused no damage, consistent with the SAS threshold levels (Fig. 1a). Because albino eyes also transmit through the sclera, the calculated doses represent minimum values.
Dawson et al. (37) and Lund et al. (38) obtained thresholds in macaques with blue lasers. Their data are in Fig. 1b. Permanent retinal damage in macaque eyes after exposure to 568 nm was observed with high resolution autofluorescence imaging by Morgan et al. (39,40) for retinal irradiance substantially below the ANSI limits (9,41); Hunter et al. (42) from the same group, added data at 488 nm. Kaidzu et al.’s (43) data in rat are not peer reviewed, but they are of interest because the authors, like Noell et al. (2), used the ERG as criterion, be it limited to the UV.
We searched the literature for monochromatic dose data to shed light on the conditions for obtaining the RAS and the SAS. Strict selection criteria were applied, monochromatic irradiation to obtain sufficient spectral resolution and to avoid excitation of multiple chromophores, and absolute dose data to enable comparison between different studies. Yet 18 publications were found with useful data; earlier attempts (15,17) had 4–5 sources. In the following the experimental conditions (italics), pass the review.
Two studies (2,28) fit the RAS. Both used the albino rat as animal model. Their actual dose levels were wide apart, but when Noell et al.’s (2) data are multiplied by a factor 25 to correct for the elevated body temperature, and Williams and Howell (28) are converted as outlined in the previous paragraph, they come quite close. Whether these conversions are allowed without change in spectral shape, remains an open question. The data from the other 16 sources seem to better fit the SAS. They include four different species (monkey, rat, rabbit and squirrel), and several subspecies and strains (Table 1). Most of the data at different wavelengths are within an order of magnitude of each other, despite vast differences in experimental conditions, like state of dark adaptation, retinal location, age and previous light history. Uncertainties in the calculation of retinal dose (44) undoubtedly added variance, but that holds for all data in Fig. 1. The course of dose as a function of wavelength seems steeper for rodents (Fig. 1a) than for macaque and rabbit (Fig. 1b), but a direct comparison in identical conditions is lacking. In the UV macaque data originate from only one source based on three macaques (4); rat data have more sources and seem to point to a higher vulnerability. That melanin pigmentation does not influence retinal vulnerability is long known (32,45,46). One of the surprises of this review is that an attempt to measure a mouse action spectrum has, to our knowledge, never been made, despite the fact that this species serves in many light damage experiments.
An important observation is that all data were obtained under anesthesia, most often pentobarbital. This is not surprising since a collimated beam directed at an immobilized eye is nearly the only way to obtain sufficient monochromatic retinal irradiance. Consequently, an action spectrum in a condition with freely moving animals, commonly assumed to be the RAS, is lacking. A number of early studies with broad band filters were all indicative for the RAS (2,47,48).
Field size was large in the RAS data; in the SAS data it went from 1° to Ganzfeld. Exposure times for the RAS data varied between 1.5 and 6 h; SAS data went up to 5 h. White light experiments indicate that the product of irradiance and time is constant up to 12 h (49). A unique feature in the study of Noell et al. (2) was the use of elevated body temperature, as was the prolonged dark adaptation, but Williams and Howell (28) had physiological temperatures and only 1 h of dark adaption; SAS data varied from light adaptation to up to 16 h of dark adaptation (Table 1).
The choice of damage criterion may influence the shape of the action spectrum. For the SAS data, histology as a function of wavelength indicates that it is an envelope of (at least) two spectral damage types. Ignoring small interspecies differences, in the macaque (4,27), squirrel (29,30) and rat (8,32) threshold irradiations in the violet and ultraviolet (320–440 nm) predominantly damaged the PR cells, blue and green lights (440–550 nm) clearly damaged the RPE, but involved also a few PR cells. If, for instance, damage to the RPE is taken as threshold criterion, the susceptibility in the violet would be somewhat lower than if damage to the PR were the criterion. In practice, these differences are small because the range from just observable changes to major destruction of the outer retina is only 1 log unit (2,7,8,37). Choice of damage criterion is therefore unlikely to explain the large discrepancy between the RAS and the SAS.
Given the strong indications that visual pigments play a role in photochemical light damage (2,13,25,50,51) their bleaching level might influence the action spectrum. To estimate bleaching level retinal irradiance and exposure time should be considered separately. The picture that emerges does not point to a simple distinction like, not fully bleached yields the RAS and fully bleached the SAS. Indeed, rhodopsin was only partially bleached in Noell et al.’s (2) study (irradiance 0.03 mW cm−2), but probably fully bleached in Williams and Howell’s (28) conditions (at 485 nm almost total loss of the ONL after 6 h of 0.82 × 0.4 = 0.33 mW cm−2 at the retina; cf. paragraph Dose Data). In the SAS the bleach level was 100% in the visible with irradiance often at levels 10–100 mW cm−2, but only partial in the UV, as was shown in a study with broader spectral bands (6). Grimm et al. (36,51) have shown that nonmetabolic regeneration of rhodopsin through photoreversal of bleaching by 403 nm light is a very effective way to increase the concentration of rhodopsin, thus offering an explanation for the increase in susceptibility of the retina at that wavelength. Quite likely this mechanism holds for the near UV too, but a quantitative analysis is lacking.
According to a hypothesis by Kremers and van Norren (15) the RAS occurs at nonfully bleaching retinal irradiances of 12 h or more (their class I), whereas the SAS occurs at higher irradiances and shorter durations (class II). They speculated that at low irradiances rhodopsin, being the most sensitive chromophore, would mediate damage and the action spectrum would follow its absorption curve. At high retinal irradiances, rhodopsin is bleached away and other, less sensitive chromophores mediate damage, giving rise to the SAS. This hypothesis needs firmer experimental evidence, if only because the action spectrum at very long exposures (with freely moving animals) is unknown. With regard to the RAS data that we gathered, in Noell et al.’s (2) study damage was obtained not in >12 h but already in 1.5 h. However, unique aspects in their experimental setup such as increased body temperature and/or prolonged dark adaptation may have allowed for this reduction in exposure time. The study of Williams and Howell (28) is more difficult to fit into the hypothesis. Exposure was 6 h and retinal irradiances were probably fully bleaching.
We conclude that, taking all sources into account, critical conditions for either spectrum do not emerge.
Conclusions and Future Experiments
Monochromatic dose data of photochemical retinal damage adhering to the RAS were found in two sources; in 16 sources they adhered to the SAS. Of the eight conditions considered (animal model, use of anesthesia, field size, exposure time, body temperature, length of dark adaptation, damage criterion and bleach level) none proved to be both necessary and sufficient to predict the generation of either spectrum. Fresh data seem therefore useful in the following areas.
1 Given the limited sources for the RAS an attempt to obtain additional experimental data on this spectrum with special attention to the intertwinement of body temperature, bleach level and length of dark adaptation, might increase our insight in the intriguing difference with the SAS. A simple paradigm could involve exposure of anesthetized rats to monochromatic radiation of close to 500 and 400 nm, equated for rhodopsin absorption. When conditions are manipulated such that damage at 500 nm is at threshold, according to the RAS damage near 400 nm should be at threshold too. The SAS, however, predicts at that wavelength total destruction of the outer retina, given a susceptibility that is two orders of magnitude higher. If possible, the experiments should involve both fully bleaching, and partially bleaching irradiances. Similar experiments with free roaming rats are more complicated because high level monochromatic radiation is difficult to obtain. Modern LEDs might open a new possibility, be it with 30 nm bandwidths. An experiment could involve two, or even three cages, one surrounded by green (λmax 525 nm), one by blue (470 nm) and one by UV LEDs (395 nm), enabling a fresh indication whether the action spectrum in such conditions is indeed the RAS.
2 Spectral data are available for only four species, rat, macaque, squirrel and rabbit. Missing are, in particular, data on the mouse, the most popular animal in biomedical research. By extending the number of species the intriguing questing might be answered whether there exists a common mechanism for light damage in all mammalian eyes, and perhaps even in birds (52–54).
3 The similarity in threshold data across species in the realm of the SAS stands out, but the spread is fairly large. Possible inherent species differences, like those that have been established between rat strains (55,56), should be tested in one set up with one damage criterion, with proper correction for eye dimensions and media absorption.
4 Additional data for primates might provide a firmer base for aphakic protection standards in the UV.
[ Dirk van Norren ]
Dirk van Norren (1942) completed his studies in experimental physics, with mathematics and meteorology at the Free University Amsterdam early 1968. He took up a job as researcher at TNO Human Factors, Soesterberg, the Netherlands. In 1977 the Department of Ophthalmology of the University Medical Center Utrecht asked him to head a small research group; later he was appointed professor in Ophthalmic Physics. He held this 1-day-a-week job until retirement in 2008.
In 1994 he became director of the Netherlands Aeromedical Institute, but returned to TNO Soesterberg to become its director in 1999, until retirement at the end of 2004. He is a (co-)author of about 130 scientific publications. Because of his volunteer work in the fields of human rights and prevention of blindness he was named Knight in the Order of Orange-Nassau in 2010.
[ Theo GMF Gorgels ]
Theo GMF Gorgels studied biology at the Radboud University (Nijmegen, NL). After obtaining his PhD in 1990, he moved to Utrecht University, to study photochemical damage of the retina, with Prof. Dirk van Norren. Next, he investigated the role of DNA damage in eye pathology with Prof. Jan Hoeijmakers at the Erasmus University Rotterdam. Currently, he is senior researcher in the group led by Prof. Arthur Bergen at the Netherlands Institute for Neuroscience (Amsterdam, NL). His research interests include the molecular mechanisms of retinal degenerations, in particular of age-related macular degeneration and pseudoxanthoma elasticum.
Acknowledgements— Encouragement, discussion and comments on earlier versions of this paper from professor Charlotte Remé and Dr. Francois Delori were much appreciated.