Assessing recovery of Drosophila melanogaster photoreceptors with different wavelengths of red and infrared light

Light interacts with life in many aspects, and while vision and optics might come to mind as the most prominent examples, there are other ways in which light affects biological systems. Perhaps one of the most well-known examples is UV light, which is essential for the biosynthesis of vitamin D (Young et al., 2021). Light exposure can also act as a stressor, with damaging repercussions for overexposure, as exemplified by sunburn and DNA damage that can even lead to cancer (de Gruijl, 1999). Another example is that overexposure to blue light has been shown to adversely affect circadian rhythm in vertebrates (Hatori et al., 2017) and invertebrates (Drosophila) (Hall et al., 2018). This concept of light as both a damaging element and something necessary for many physiological processes is not a new one. While we need light to see, both chronic and acute exposure to bright light can have dire consequences for both vertebrate and invertebrate photoreceptors (Albarracin et al., 2011; Hall et al., 2018; Lee and Montell, 2004). In this study, we used such light exposure to induce damage in the eyes of Drosophila melanogaster, which were then exposed to longer wavelengths of light that may be beneficial for their recovery.

Recent studies have suggested that the application of long-wavelength light can be used as a therapeutic tool for photobiomodulation. While there is a boom in commercialized devices for light therapies of all kinds, the science to support such devices greatly lags behind, with a strong need for well-controlled studies. Still, evidence has emerged that benefits include neuroprotection, wound healing and recovery from stroke (Fitzgerald et al., 2013). At the mechanistic level, much obscurity remains, but multiple studies have demonstrated that near-infrared light can affect the metabolic activity of tissues, typically by altering ATP production (de Freitas and Hamblin, 2016) and the generation of reactive oxygen species (ROS) (Begum et al., 2015; Sanderson et al., 2018; Weinrich et al., 2017). One hypothesized mechanism for this interaction is based on the photosensitivity of cytochrome c oxidase (COX) which has four metal centers (two copper and two heme groups) that absorb light from ∼600 to 1000 nm (Karu et al., 2005). The activity of this enzyme is the rate-limiting step in the electron transport chain, and hence it is thought to directly influence ATP and ROS production, possibly in a wavelength-specific way (Fuchs et al., 2021; Pruitt et al., 2022; Sanderson et al., 2018).

Ambiguity continues to exist regarding which wavelengths are most beneficial for specific applications (Fuchs et al., 2021). Among the most studied wavelengths is 670 nm light, which enhances cell proliferation in rats (Eells et al., 2003), improves embryonic development in birds (Yeager et al., 2005) and increases longevity in flies (Begum et al., 2015). Exposure to 670 nm light is also neuroprotective, enhances insecticide resistance in bees (Bombus terrestris) (Powner et al., 2016), enhances resistance to light-induced retinal damage in rats (Albarracin et al., 2011) and can protect from age-related macular degeneration in mice and rats (Begum et al., 2013; Kaynezhad et al., 2021; Sivapathasuntharam et al., 2017). Mechanistically, 670 nm light increased COX expression in a macular degeneration model for mice, which was also related to reduced inflammation in the retina (Begum et al., 2013). Longer wavelengths, such as near-infrared (>800 nm) are relatively understudied in terms of their physiological effects. In an analysis of wavelength-specific effects on mitochondrial activity, it was shown that by decreasing COX activity with exposure to 750 or 950 nm light, there was a significant drop in mass cell death following ischemic reperfusion injury, likely due to a reduction in the buildup of ROS that is typical in reperfusion injury (Sanderson et al., 2018). In contrast, in the same study, 810 nm light increased COX activity and did not alleviate the damage from reperfusion injury. It was also found that both 685 and 830 nm light were effective in aiding the recovery of cutaneous wounds in rats (Mendez et al., 2004), with highest efficiency in rats exposed to the two wavelengths simultaneously. Exposure to 810 nm light has also been shown to improve outcomes following traumatic brain injury (TBI) in mice (Ando et al., 2011). According to one study, mice had better outcomes following TBI after exposure to 665 nm and especially 810 nm light; however, 730 and 980 nm light exposure did not result in improvements (Wu et al., 2012).

Despite these many findings for photobiomodulation showing great promise as a therapeutic tool (Fitzgerald et al., 2013; Zein et al., 2018), the effective dosage of light for this purpose remains unclear. Studies have used a wide range of doses, and even though some important parameters are starting to emerge (for review, see Zein et al., 2018) there is not a clear consensus on how to optimally dose light exposure or which wavelengths work best when applied in the same way.

Compounding this issue is the inconsistency with which light intensities are reported (Johnsen, 2012), including units that are not directly comparable. In some cases illuminance (lux) is reported, which is inherently biased toward human visual sensitivity, and hence is a poor measure for comparisons that involve other animals. Other measurements focus on energy, which accounts for shorter wavelengths having higher energy photons. Conceptually, the biologically most relevant method of reporting light intensity is in terms of photon flux (photons cm⁻² s⁻¹) as individually absorbed photons can trigger a change in a biological structure (Johnsen, 2012), as is well understood for phototransduction in vision (Wang and Montell, 2007). In addition, to account for a more complete picture of energy transfer, Zein et al. (2018) have suggested that reporting the fluence (total energy received per unit of surface area, integrated over a given period of time, i.e. energy density) of a light treatment is of importance.

Considering the complex array of reported wavelengths, intensities, stimulation protocols and units that have been utilized in light studies, comparisons between existing studies are relatively complicated and, in many cases, impossible, therefore emphasizing the need for well-controlled direct comparisons. In this paper, we therefore aimed to assess how different wavelengths of red to near-infrared light may modulate the recovery of the retina of D. melanogaster that previously had been damaged through excessive light exposure. We focused on the white mutant (w1118) that is known for its susceptibility to neurodegeneration (Ferreiro et al., 2017), and in which light can influence photoreceptors particularly well because of the lack of screening pigment. The Drosophila visual system makes for an excellent medium in which to study this question given that this species is an important model for the study of neurodegeneration (e.g. Smylla et al., 2022; Steele and O'Tousa, 1990) and the intense metabolic demands that the eye places upon the body (Laughlin et al., 1998). Furthermore, by their very nature as light-receiving cells, photoreceptors experience high metabolic needs when excessively activated and are naturally susceptible to light-induced retinal degeneration (Laughlin et al., 1998; Lee and Montell, 2004). This study aimed to discern the degree to which a variety of different wavelengths impact the recovery of damaged retinae.

White mutant D. melanogaster (w1118, Bloomington stock #3605) were reared on Nutri-Fly Bloomington Formulation (Genesee Scientific, Morrisville, NC, USA) fly media with the addition of 35 ml molasses per 1000 ml media, which also was freely available to flies throughout these experiments. Flies initially were sorted into one of two treatment groups, both of which consisted of 3 day old adults (3 days post-eclosion) at the time of experimentation. In one group (hereafter referred to as 3 day control), freshly eclosed flies were placed under controlled lighting comparable to typical indoor lighting (Ace A19 E26 6W LED bulb, Ace Hardware, Oak Brook, IL, USA; see Table 1 for light intensity) with ambient temperature and humidity for 72 h. The other group (referred to as 3 day damaged) was placed in high light intensity in an incubator at 22°C and ambient humidity. The lighting environment consisted of bright light from a white LED for 12 h per day in addition to a somewhat dimmer constant light source (also from a white LED) that was on for the entire 72 h period (see Table 1 for light intensity values). Following this 72 h period, 3 day old adult D. melanogaster were either used to assess light damage or were placed into secondary treatment groups, each of which lasted for 10 days. Control flies for the latter groups (13 day control) were raised for 10 additional days in the same conditions as described for 3 day old controls. For treatment, 3 day damaged flies were split into the following six groups. One group received treatment identical to that of control animals to serve as a measure of the level of recovery that might be attained under normal rearing conditions (baseline recovery group). The remaining five groups were kept under identical conditions with the exception of the addition of red, near-infrared or infrared light. Specific wavelengths included: 670 nm (ELD-670-524, Roithner LaserTechnik, Vienna, Austria), 750 nm (LED750-03 AU, Roithner LaserTechnik), 810 nm (ELD-810-345, Roithner LaserTechnik), 850 nm (ELD-850-525, Roithner LaserTechnik) and 950 nm (SIR-56ST3FF, Rohm Semiconductor, Kyoto, Japan). All long-wavelength light was presented at comparable intensities (see Table 1 for details) regarding photon count and power for each wavelength. Exposure was kept equally at two 30 min periods per day over the course of the 10 day treatment. Table S1 summarizes details regarding the construction of the LED arrays, and reports temperature values that were recorded via thermal camera.

It has been suggested that the most relevant method of reporting light intensity for a biological setting is in terms of photon count, as individual photons typically trigger a change that is induced by their absorption (Johnsen, 2012). For this reason, our comparisons are based on photons cm⁻² s⁻¹. As many studies define light in terms of power, we also provide values in mW cm⁻² (Table 1). It has been suggested that for light exposure experiments, another important measurement is fluence or energy density (Zein et al., 2018), as it accounts for the total amount of light exposure over the course of a treatment, regardless of the treatment length. Therefore, we also include values for the energy density of the red to infrared lights used for both 30 min time periods as well as for the entire duration of the experiment (Fig. 1, Table 1).

Drosophila melanogaster were prepared for electrophysiological recording as described in Charlton-Perkins et al. (2017) and Rathore et al. (2023). Electroretinogram (ERG) stimuli were delivered as a series of increasing irradiances to define photoresponses as a function of the log of light intensity (referred to as a V–logI curve). The photon flux values used were 4.28×10¹⁰, 8.15×10¹⁰, 2.78×10¹¹, 7.2×10¹¹, 1.75×10¹², 5.42×10¹², 1.04×10¹³, 5.53×10¹³ and 1.07×10¹⁴ photons cm⁻² s⁻¹. Each stimulus consisted of three pulses of 490 nm light (LED490-06, Roithner Lasertechnik) with a 1 s pulse width and 15 s between pulses. A 1 min break between stimuli allowed recovery of photoreceptors. The responses to the three pulses of each stimulus were averaged to create a single measurement of photoresponse for each fly at a given irradiance. Comparisons are based on an intensity of 7.2×10¹¹ photons cm⁻² s⁻¹, as this led to a relatively large photoresponse without saturating. The magnitude of the photoreceptor response was evaluated with a custom-written MATLAB (MathWorks, Inc., Natick, MA, USA) code (Charlton-Perkins et al., 2017; Riazuddin et al., 2012) for the largest difference between values prior to and during light stimulation. If present, the magnitude of the on-transients was recorded also.

The corneal neutralization technique takes advantage of the innate autofluorescence of Drosophila photoreceptors. As described in Pichaud and Desplan (2001), the active state of the D. melanogaster primary rhodopsin (Rh1), metarhodopsin 1 (mRh1), autofluoresces under illumination of 540–580 nm light. Importantly, in the apical eye region that was imaged, only the outer photoreceptors (R1–R6) fluoresce (as only these photoreceptors express Rh1 – inner photoreceptors R7/8 express other rhodopsins important for color vision) (Franceschini et al., 1981; McCulloch et al., 2022; Pichaud and Desplan, 2001; Wang and Montell, 2007). Drosophila were mounted as described in Charlton-Perkins et al. (2017) and Rathore et al. (2023) for ERGs, with the left eye facing upward. Flies were allowed to dark adapt for 10 min. Each fly and coverslip were then placed into a small Petri dish and were secured with additional dental wax and the dish was filled with distilled water. Photoreceptors were imaged using an Olympus BX51 fluorescence microscope (Olympus Life Science, Tokyo, Japan) with an Olympus LUMPlanFl/IR ×40 objective water lens (Olympus Life Science) and an Olympus 100 W mercury power supply (BH2-RFL-T3; Olympus Corporation, Tokyo, Japan). For consistency, all imaging was performed under identical conditions, with a Retiga-2000R digital camera (Teledyne QImaging, Surrey, BC, Canada) in combination with QCapture (Teledyne QImaging) software under 4×4 binning and with a 70 ms integration time. Eyes were excited with a DAPI filter for 5 s so as to convert available Rh1 to mRh1. Following this, photoreceptors were imaged through a Texas Red filter.

All flies were analyzed blindly. To assess the number of visible photoreceptors, six ommatidia were selected from the best focused portion of the retina. Images were adjusted in batches according to the date of imaging using Photoshop 2023 (Adobe, San Jose, CA, USA) such that the grayscale color value of the background of the image was comparable. Photoreceptors of the selected ommatidia were identified as being R1–R6 on the basis of their position within the trapezoidal organization. This way the number of visible cells of each type was established for each individual. Following this, the relative autofluorescence intensity was assessed in ImageJ (NIH, Bethesda, MD, USA). Specifically, for each selected ommatidium, the area of each rhabdomere was selected and the average fluorescence was recorded. For poorly visible rhabdomeres or those that were not visible or were otherwise indistinguishable from the background, the appropriate region of the rosette was sampled. Photoreceptor fluorescence was corrected for by subtracting the background fluorescence of the image, which was sampled proximally to the ommatidium, immediately behind photoreceptor R4.

To test for the presence of any compounded effects of photobiomodulation and the rhodopsin reconversion pathway, we also implemented a 590 nm light treatment. This is close to the optimum wavelength (∼580 nm; Alloway et al., 2000) for conversion of metarhodopsin back to rhodopsin, and its peak wavelength is below the wavelength band absorbed by COX (Karu et al., 2005). For this treatment, a 590 nm LED (LY G6SP.02-8D7E-36-G3R3-140-R18, ams-OSRAM, Premstätten, Austria) was used as described above: light exposure occurred twice per day for 30 min over 10 days (see Table 1 for details). This allowed us to determine whether neural recovery was hampered by a lack of conversion of metarhodopsin back to rhodopsin, which could also have been promoted by some of our near-infrared exposures.

To test for any possible effects of prolonged depolarizing afterpotential (PDA) (Belusic et al., 2010; Minke, 2012) caused by exposure to white LEDs, photoresponses of 2 day old adult Drosophila were evaluated after exposure to a white LED (Ace A19 E26 6W LED bulb, Ace Hardware, Oak Brook, IL, USA) as well as a custom-built red lamp composed of a halogen lamp (EKE 21V150W, Ushio, Inc., Tokyo, Japan) and red vinyl filter (Oracal 8300 Transparent Vinyl, Orafol Americas Inc., Ellabell, GA, USA) (see Fig. S2 for spectra). Following a random order, either light was applied for 5 min after a 30 min dark adaptation period at an intensity of 1.7×10¹⁵ photons cm⁻² s⁻¹. The photoresponse then was assessed 1 min post-light exposure, after which animals were allowed to dark adapt again for 30 min followed by a second light exposure with the complementary light and photoresponse measurement.

ATP concentration was determined using the Enliten ATP Assay kit (Promega, Madison, WI, USA) as per the manufacturer’s instructions and as described by Park et al. (2006) and Xie et al. (2021) with the following changes. Five flies per sample were dissected in 2.5% TCA. Heads were placed in 50 μl 2.5% TCA and flash frozen in liquid nitrogen for 3 min. Samples were then homogenized on ice with micropestles for 100 downward strokes. Following this, samples were centrifuged at 13,000 g for 15 min at 4°C. Then, 7.5 μl supernatant was collected, transferred to a new vial, and TCA was neutralized with 292.5 μl 1 mol l⁻¹ Tris-acetate buffer. Following the addition of L/rL reagent, luminescence was measured on a BioTek Synergy 4 plate reader (BioTek, Winooski, VT, USA). Measured ATP concentration was used to calculate the picomoles of extracted ATP, per fly head.

To assess putative damage caused by 72 h exposure to bright light, we first performed ERGs to assess the eye's ability to respond to light. Our measurements of control flies (3 day control; Fig. 2A) exhibited the typical waveform (Belusic, 2011) with an on-transient that follows the onset of the stimulus, a sustained photoreceptor response and an off-transient that follows the end of the stimulus. Test flies (3 day damaged) showed a similar but greatly reduced response. Specifically, the maximum photoreceptor response of light-damaged flies was significantly lower (only 2.89±1.03 mV compared with 8.36±1.37 mV of 3 day control flies, means±s.e.m.; Fig. 2B). It was also observed that some light-damaged flies were lacking the on-transient. Specifically, only 71.4% of light-damaged 3 day old flies displayed on-transient responses compared with 100% of the control flies (Fig. 2C). In addition, of those flies that had an on-transient response, light-damaged flies had a lower on-transient amplitude (0.58±0.20 mV) than controls (2.99±0.42 mV) (Fig. 2D). Next we assessed the integrity of photoreceptors by visualizing them through corneal neutralization (Pichaud and Desplan, 2001), a method in which flies are imaged in water, which greatly reduces the refractive power of the facet lenses and hence allows visualization of the underlying photoreceptors. When control flies are imaged this way, the classical trapezoidal shape of photoreceptors R1–R6 is visible (Fig. 3A). At the midline of the eye, the orientation of the trapezoid flips (Dietrich, 1909). Constant exposure to bright light clearly had a major impact on photoreceptors, with affected individuals exhibiting an irregular subset of visible rhabdomeres (Fig. 3B). Based on our blindly scored ommatidia quantification, light-damaged flies had significantly fewer photoreceptors that were visible (only 15.8±8.1% compared with 99.7±0.3% in control flies; Fig. 3C). To quantify these images further, we also evaluated the intensity of the autofluorescence of individual photoreceptors and found that they were significantly brighter (23.86±4.03 a.u.) in undamaged retinae than in damaged retinae (1.47±0.76 a.u.; Fig. 3D).

To assess a possible effect of red and infrared light on the integrity of photoreceptors, we exposed light-damaged flies for 10 days to five different wavelengths that were administered at equal intensity and duration. For comparison, we used two control groups, one consisting of age-controlled flies that were never exposed to bright light (13 day control) and the second containing light-damaged flies that were allowed to recover in the absence of red or infrared light, but otherwise were kept under equal light conditions (baseline recovery). ERGs of all groups showed typical photoresponses, albeit with dramatically different amplitudes (Fig. 4A). Our treatment led to significant differences in the amplitude of the photoreceptor response (Fig. 4B). Notably, baseline recovery flies completely failed to recover their photoreceptor responses, with no significant difference from the post-light damage response of the 3 day old flies (Tukey HSD, P=0.96). Their response was significantly reduced (1.23±0.57 mV) compared with that of 13 day control flies (6.58±1.15 mV). For red and infrared light-treated flies, we found large differences in the magnitude of their photoresponses. Light-damaged flies that were exposed to 670 nm light during recovery had a mean response of 5.56±1.03 mV. While this average is substantially higher, possibly indicating a trend towards improvement, this group of flies exhibited a lot of variance with no significant improvement when compared with baseline recovery flies (Tukey HSD, P=0.14). Notably, the photoreceptor response of these flies was comparable to that of 13 day control flies (Tukey HSD, P>0.999). The 750 nm light-treated flies also showed a lot of variance, but had a slightly higher mean response of 7.99±1.46 mV with significant improvement compared with baseline recovery flies, and no statistical difference from 13 day control flies (Tukey HSD, P=0.991). Treatment with 810 nm light led to similar results as for the 750 nm light-treated flies, with significant improvement of the photoreceptor response (7.11±0.88 mV) when compared with baseline recovery flies and comparable values to those of 13 day control flies (Tukey HSD, P>0.999). Particularly high photoreceptor responses were observed in 850 nm light-treated flies with a mean value of 10.51±1.09 mV. This was not only significantly higher than the response of baseline recovery flies but was also higher on average (albeit not significant) than that of the 13 day control flies that had never experienced light damage (Tukey HSD, P=0.21). Notably, their average photoreceptor responses were comparable (numerically even a touch higher) to those of the much younger 3 day control flies (Tukey HSD, P=0.854) that never experienced light damage (Fig. 2B). The only light-treated group in our setup that did not show an improved photoreceptor response was the 950 nm light group. These 950 nm light-treated flies had a mean response of 0.56±0.26 mV, which is comparable to that of baseline recovery flies (Tukey HSD, P>0.999) and significantly lower than the response of 13 day control flies or any of our other light treatments: 670 nm (Tukey HSD, P=0.034), 750 nm (Tukey HSD, P=0.00007), 810 nm (Tukey HSD, P=0.00073) and 850 nm (Tukey HSD, P<10⁻⁷).

Similar patterns with significant differences between treatment groups also emerged from an analysis of the on-transient (Fig. 4C). We first evaluated flies in terms of the proportion in each group in which an on-transient was visible. This was the case for 90% of 13 day old control flies but only 30% of the baseline recovery flies. After light treatment, on-transients were observed in 77.8% of 670 nm light-treated flies, 100% of 750, 810 and 850 nm light-treated flies and 25% of 950 nm light-treated flies. Of those flies that had an on-transient response, light treatment had a significant effect on its amplitude (Fig. 4D). Specifically, 13 day control and baseline recovery flies had a mean on-transient amplitude of 3.51±0.51 and 1.25±0.90 mV, respectively. The 670, 750 and 810 nm light-treated flies had mean amplitudes of 2.30±0.46, 3.06±0.48 and 2.33±0.39 mV, respectively. These amplitudes were not significantly different from each other. Here too, 850 nm light-treated flies had the highest value, with a significantly higher on-transient amplitude than baseline recovery flies, at 3.85±0.39 mV. The 850 nm light-treated flies also had a statistically significantly higher on-transient amplitude than the 950 nm light-treated flies (Dunn's test, Benjamini–Hochberg corrected, P=0.00915). The 950 nm light-treated flies had a mean on-transient response amplitude of only 0.38±0.11 mV which was significantly lower than that of 13 day control flies but not significantly different to those of baseline recovery flies (Dunn's test, Benjamini–Hochberg corrected, P=0.66).

Images of photoreceptor arrays under cornea neutralization provide insights into photoreceptor integrity in the treatment groups (Fig. 5). Fig. 5A is a schematic illustration of a single ommatidial array, with the rhabdomere of the centrally located R7/8 photoreceptor in red and the rhabdomeres of the peripherally located R1–R6 in blue. The latter become visible in cornea neutralization as a result of the presence of metarhodopsin. Fig. 5B,C shows representative images of the two control lines, 13 day control and baseline recovery, respectively. The former shows a pristine array with a clearly visible trapezoid organization, while the latter shows major disruption of the array, with only a subset of the rhabdomeres visible. Fig. 4D–F illustrates representative examples of flies that were treated with 670, 750, 810, 850 and 950 nm light. Apart from the 850 nm example, in each of them at least some of the ommatidia failed to show rhabdomeres, and the arrays exhibit various degrees of disruption. To quantify this effect, we first analyzed images with regards to the proportion of visible photoreceptors to the total maximum number of photoreceptors expected in the sampled area (Fig. 6A). The visibility of photoreceptors under equal settings of autofluorescence illumination showed significant differences. Specifically, the two control lines showed either 100% photoreceptor integrity (13 day control) or a reduction to 49.4±11.3% (baseline recovery). In all light treatments, flies had a higher average number of visible photoreceptors than the baseline recovery flies, with significant improvements for 810, 850 and 950 nm light-treated flies (with 86.4±3.2%, 91.7±3.0% and 88.7±4.4% of photoreceptors being visible, respectively). Nevertheless, all treatments showed somewhat lower photoreceptor counts than the 13 day control flies. Significant differences here were observed for 670 nm, 750 nm, and 810 nm treated flies. Specifically, 78.1±7.5% of photoreceptors were visible in 670 nm light-treated flies and 78.9±4.1% photoreceptors were visible in 750 nm light-treated flies. In addition, we quantified autofluorescence brightness (Fig. 6B), which also showed differences across treatment groups. Baseline recovery flies had significantly dimmer photoreceptors than 13 day control flies. Significant improvement in photoreceptor brightness (compared with baseline recovery flies) was observed in 810 and 850 nm light-treated groups.

In flies it is well established that light exposure leads not only to the photoconversion from opsin to the photoproduct metarhodopsin but also to the reconversion from metarhodopsin to opsin (Dolph et al., 1993). While the former is facilitated by blue light (around 480 nm), the latter requires orange light (around 580 nm). The w1118 D. melanogaster used in this experiment are white-eyed and hence lack the red eye color that naturally assists with that reconversion (Stavenga et al., 2017). Furthermore, it is conceivable that our LED-based background illumination had insufficient light levels to fully back-convert the metarhodopsin to rhodopsin, which in itself potentially could lead to unhealthy photoreceptors, which could be mitigated through the application of additional back-converting light. To test for this possibility, we conducted an additional experiment in which we treated flies with 590 nm light, and tested them with ERG as in the previous experiments (Fig. 7A,B). In this experiment, 13 day old control (11.95±1.37 mV) and baseline recovery (1.03±0.466 mV) animals showed comparable responses to our first experiment (Kruskal–Wallis, χ²₂=13.805, P=0.001). The 590 nm light-treated flies had a mean on-transient response amplitude of only 1.85±0.54 mV, which was significantly lower than that of 13 day control flies but not significantly different to that of baseline recovery flies (Dunn's test, Benjamini–Hochberg corrected, P=0.003). Along the same lines, the 590 nm light-treated flies were also comparable to the baseline recovery flies in terms of the proportion of flies that produced on-transients (Fig. S1A) and the size of the observed on-transients (Fig. S1B).

To directly test for the possibility that white LED light could lead to photoreceptors that are stuck in a PDA state, we directly compared the photoresponse after white LED light exposure and compared it with that after red light exposure (Fig. 7C). If the white LED were to elicit a PDA, it would be expected that the white LED pre-exposure treatment would lead to smaller photoresponses than the red light pre-exposure, as photoreceptors would be stuck in the PDA in the former but not in the latter treatment. This, however, was not the case. Instead, irrespective of the order in which lights were presented, the second measurement tended to have slightly stronger responses; however, these differences were not significant (paired t-test, t₉=−1.536, P=0.159).

In addition to contrasting how equally intense exposures to light of different wavelengths affect neural recovery, we also performed a bioassay to assess whether recovery could be related to metabolic processes. Specifically, we quantified ATP concentrations that could be isolated from D. melanogaster heads after selected treatments (Fig. 8). Our results suggest that there was no significant difference in ATP concentration between 13 day control and light-damaged individuals that were recovering in background lighting (baseline recovery), though the latter had slightly lower average values (Tukey HSD, P=0.908). However, with a near-doubling of ATP concentration, there was a significant increase between the baseline recovery flies and both groups of flies that were supplemented with 850 nm (Tukey HSD, P=0.0005) and 950 nm (Tukey HSD, P=0.0009) light treatments.

In our first experiment, we established light exposure parameters that were sufficient for inducing photoreceptor degeneration in D. melanogaster, as in Lee and Montell (2004). In our results, light-induced degeneration manifested as a severely reduced response of photoreceptors to light stimuli (Fig. 2). In addition, an autofluorescence imaging assay indicated reduced visibility of photoreceptors in the light treatment group (Fig. 3), suggesting that this treatment resulted in physical damage of the opsin-rich rhabdomeres of these cells. This could be due to an extended time period of overexcitation of photoreceptors, deprived of dark recovery periods (Lee and Montell, 2004). Light-induced photoreceptor degeneration in Drosophila is thought to be due at least in part to the loss of the primary photopigment rhodopsin, Rh1. For example, it was previously shown that constant light exposure can result in the buildup of stable complexes between Rh1 (the opsin that activates the response) and Arrestin 2 (Arr2; the protein responsible for terminating the response) (Alloway et al., 2000; Kiselev et al., 2000; Wang and Montell, 2007). Under normal conditions, these complexes are transient and are dissociated by the phosphorylation of Arr2 following calcium influx (Wang and Montell, 2007). After overstimulation, these stable complexes undergo endocytosis, which is followed by apoptotic cell death (Kiselev et al., 2000) (for review, see Wang and Montell, 2007). Oxidative stress is also known to play an important role in retinal degeneration (Jarrett and Boulton, 2012), caused by ROS that are byproducts of many normal functions of the cell. Most typically this involves cellular respiration, which is critical in photoreceptors that are known to be metabolically highly demanding (Laughlin et al., 1998). ROS are damaging to the cellular infrastructure and are associated with a decrease in the amount of ATP that is produced, thus resulting in mitochondrial inefficiency and toxicity (Jarrett and Boulton, 2012). Oxidative stress is typical of senescence and induced stress in cells (Gkotsi et al., 2014; Hall et al., 2018).

The fact that too much light can be damaging to photoreceptors is not new. For example, severe damage has been described in the deep sea isopod Cirolana borealis when they were accidentally exposed to bright light during collection (Meyer-Rochow and Nilsson, 1999). Another example is the eyes of the nocturnal net-casting spiders (Dinopids), which are so light sensitive that they deconstruct their rhabdoms every day to prevent light damage (Blest, 1978; Blest et al., 1978). Notable from our experiments on white mutant D. melanogaster is that photoreceptor health did not recover within a 10 day period under control lighting. In addition to the photoreceptor response itself, this was manifested as a loss in prevalence and amplitude of the on-transient interneuron response (Fig. 2C,D). One might expect that a reduced photoreceptor response would potentially coincide with a relatively increased (compensatory) synaptic (on-transient) response (Heisenberg, 1971); however, our data did not show this. This could be indicative of poor health of the photoreceptor cells. Imaging autofluorescence under corneal neutralization likely visualized mRh1 (the photoproduct of Rh1) and as such functioned well as a proxy of the anatomical integrity of the outer photoreceptors of each ommatidium (Pichaud and Desplan, 2001). Indeed, our photoreceptor autofluorescence data align well with this idea; light-damaged flies had fewer visible photoreceptors under corneal neutralization and ommatidia were generally less fluorescent in light-damaged flies. This would suggest that photodamaged retinae had less activated rhodopsin. Taken together, our findings are consistent with light damage leading to reduced levels of Rh1 in the Drosophila retinae, which has already been documented in other studies (Lee and Montell, 2004) and, as we discuss in the next section, serves as a powerful paradigm to evaluate effects of red to infrared light.

In contrast to the inability of fly eyes to recover in background lighting (baseline recovery), photoreceptor responses dramatically improved after treatment with 670, 750, 810 and 850 nm light. In stark contrast, flies that were treated with 950 nm light failed to recover. Generally, the effects of near-infrared treatment on on-transients followed a similar pattern to that of the photoreceptor response. Baseline recovery flies had a slightly more robust on-transient response than 3 day damaged flies; however, fewer baseline recovery flies displayed the on-transient response at all. In contrast to our electrophysiology findings, assessment of retina autofluorescence indicated that all red to infrared-treated flies showed some level of recovery. Consistent with the physiological data is that the 850 nm treatment group showed the highest level of recovery, in terms of both the number of visible photoreceptors and the degree of photoreceptor autofluorescence. Perhaps most surprisingly, 950 nm light-treated flies showed impressive recovery in terms of visible photoreceptors (Fig. 6A), with levels of autofluorescence comparable to those of other red and near-infrared-exposed groups, and even the 13 day control group (Fig. 6B). This discrepancy between electrophysiology and optical imaging could relate to these readouts relating to different processes. While the physiology evaluates the response of photoreceptors to light, the optical imaging specifically captures the presence of metarhodopsin. It is possible that for 950 nm light, a disproportionate amount of opsin remains in that state, leading to fluorescence even in light-damaged flies.

Although the study presented here was primarily focused on comparing wavelengths, our data provide some insights into possible underlying mechanisms. For example, in some cases, neural degeneration could be caused by overactive photoreceptors (Lee and Montell, 2004; Wang and Montell, 2007). One reason for such over-activation could be abnormally high metarhodopsin levels. However, our assessment of 590 nm light (Fig. 7A,B) as well as our direct assessment of physiology after exposure to metarhodopsin-to-rhodopsin converting wavelengths of light (Fig. 7C) suggest that this mechanism is unlikely to relate to observed differences. Instead, observed effects may relate to energy metabolism (Laughlin et al., 1998; Lee and Montell, 2004). Our assessment of ATP concentrations suggests that for both 850 and 950 nm light exposure, the availability of ATP nearly doubled, with an increase comparable to that in a study that assessed ATP concentrations after exposure to 670 nm light (Begum et al., 2015). Thus, it is conceivable that the enhanced availability of ATP helped facilitate retina recovery. It remains unclear, and should be subject to further investigation, why the recovery at the electrophysiological level was less for 950 nm light than for 850 nm. Notably, keeping the photon count constant necessarily leads to discrepancies in energy density across treatment groups (Table 1). Thus, the absence of recovery at 950 nm could relate to dosage.

Our study contributes to an exciting recent accumulation of well7-controlled photobiomodulation-related findings that suggest that red to infrared light can influence the physiology of tissue and facilitate recovery after injury. Many studies focus on effects of individual wavelengths but with a multitude of different ways in which light intensity tends to be measured (e.g. illuminance, irradiance, energy, etc.). Unfortunately, these units typically are not directly comparable, as exact light spectra often are not reported. Additionally, many studies utilize different treatment session durations, and differ in how exactly light is applied (e.g. pulsed light exposure versus continuous light exposure; Al-Watban and Zhang, 2004; Ando et al., 2011). Given the non-uniform measurement of light across the literature, a more complete comparison of photobiomodulation studies is important. For this reason, our study aimed at directly comparing several wavelengths of light that were applied with equal duration, application protocol and (in terms of photon count) comparable intensity, and indeed we found major differences in effects. We also were able to establish a photodamage paradigm in white-eyed D. melanogaster that is well suited to dig deeper into underlying molecular mechanisms and to evaluate recovery under different light treatments, which we anticipate will facilitate additional studies in this direction. Taken together, our findings add to the exciting topic of how light affects physiology, a field with many remaining open questions.

We would like to thank Dr Stephen Mergner for his assistance in constructing the LED housing units used in this study, Dr Necati Kaval and the University of Cincinnati Chemical Sensors & Biosensors Core Facility for the use of its plate reader, Amartya Tashi Mitra for providing artwork of a Drosophila silhouette used in the creation of figures, and Dr Annette Stowasser for her support and valued advice for the duration of this study.