Phenotypic plasticity in Periplaneta americana photoreceptors

Environmental stimuli contribute to the proper development and maintenance of sensory receptors and their downstream neural circuits. In visual systems, the effects of such stimulation, or its lack, can range from a failure to establish proper synaptic connections during ontogenesis (Hubel and Wiesel, 1970; Hubel et al., 1977; Jiang et al., 2009) to various forms of synaptic plasticity (Berry and Nedivi, 2016; Pallas, 2017). For peripheral visual systems, numerous short- and long-term activity–dependent modifications have been described over different time scales at both the cellular and network levels in photoreceptors and higher-order visual neurons (Brann and Cohen, 1987; Sokolov et al., 2002; Wagner and Kröger, 2005; Calvert et al., 2006).

Studies of plasticity in invertebrate visual systems have examined developmental changes at the first visual synapse, connections between neurons in the higher-order visual centers (Hertel, 1983; Meinertzhagen, 1989; Barth et al., 1997; Pallas, 2017), short-term light adaptations in the retina (Laughlin, 1989), and illumination-dependent changes at the molecular level in photoreceptors (Bähner et al., 2002; Cronin et al., 2006; Frechter and Minke, 2006). However, little is known about long-term functional adaptations in microvillar photoreceptors. Phenotypic plasticity of the electrophysiological properties of microvillar photoreceptors has primarily been explored in dipterans. Vision of the housefly, Musca domestica, displayed improved absolute sensitivity and contrast sensitivity when it was reared in complete darkness for several days after emergence (Deimel and Kral, 1992). When the fruit fly, Drosophila melanogaster, was exposed to light, its photoreceptor responses were faster and less noisy, with higher information capacity than in dark-reared flies (Wolfram and Juusola, 2004; Voolstra et al., 2017). In addition, long-term changes in the K⁺ current of the sea slug Hermissenda crassicornis photoreceptors were detected after relatively short exposure to light (Yamoah et al., 2005).

Although most research into invertebrate vision has been performed in flies, their photoreceptors are different from many other microvillar photoreceptors. The fly visual system is evolutionarily tuned to operate with relatively high speeds of movement and maneuvering (Weckström and Laughlin, 1995; Frolov et al., 2016). Their compound eyes are characterized by open-rhabdom organization of the ommatidia, with neural superposition taking place in the first optic ganglion, the lamina (Fain et al., 2010). In contrast to hemimetabolous insect species, where photoreceptors must function while they grow during a period of postembryonic development (Frolov et al., 2012), adult fly photoreceptors do not grow. They become functionally mature during the first hours or days after eclosion (Rudolf et al., 2014). Also, the relatively short life spans of flies (Carey, 2001) preclude prolonged light exposure/deprivation experiments. Also, our recent analyses of phototransduction in the cockroach Periplaneta americana, including knockdown of several retinal proteins by RNA interference, have suggested that the phototransduction cascades of flies and cockroaches differ in several important aspects, including the role of Ca²⁺ and expression patterns of light-activated channels (Immonen et al., 2014, 2017; French et al., 2015; Saari et al., 2017).

Here, we investigated photoreceptors of adult P. americana that were reared in uniform bright light or darkness for several months. Electrophysiological recordings from photoreceptors in dissociated ommatidia revealed distinct and opposing physiological adaptations that suggest morphological changes compared with photoreceptors of control animals maintained under normal illumination (12 h light:12 h dark) conditions. These changes are likely to involve structural remodeling of light-sensitive membrane but also affect the timing of phototransduction. We argue that these changes adjust photoreceptor function to different illumination conditions.

American cockroaches, P. americana (Linnaeus), were purchased from Blades Biological and maintained at 25°C under three light regimens: in constant light (CL), in reversed 12-h/12-h illumination conditions with a subjective “night” period matching the actual day (control), and in nearly constant dark (CD). Illumination for the CL group and for the subjective day period of control groups was provided by the built-in incubator light source and corresponded to bright indoor illumination (∼2,000 lux). CL cockroaches were housed in individual transparent cages with a cardboard shelter available. CD cockroaches were housed in a large Plexiglas container with plenty of shelter available and were occasionally and briefly exposed to room light during servicing. Only male cockroaches were used for experiments.

Ommatidia were dissociated, and whole-cell recordings were performed as described previously (Saari et al., 2017). In brief, data were acquired using an Axopatch 1-D patch-clamp amplifier, Digidata 1550 digitizer, and pClamp 10 software (Axon Instruments/Molecular Devices). Patch electrodes were made from a thin-walled borosilicate glass (World Precision Instruments) and had resistances of 4–9 MΩ. Bath solution contained (in mM) 120 NaCl, 5 KCl, 4 MgCl₂, 1.5 CaCl₂, 10 N-Tris-(hydroxymethyl)-methyl-2-amino-ethanesulfonic acid (TES), 25 proline, and 5 alanine, pH 7.15. Patch pipette solution contained (in mM) 100 K-gluconate, 40 KCl, 10 TES, 2 MgCl₂, 4 Mg-ATP, 0.4 Na-GTP, and 1 NAD, pH 7.15. The liquid junction potential (LJP) was −12 mV. All voltage values cited in the text were corrected for the LJP. The series resistance was compensated by 80%. Membrane capacitance was calculated from the total charge flowing during capacitive transients for voltage steps from −112 to −92/−82 mV.

Light stimulation was performed as described previously (Saari et al., 2017). Stimulus intensity was attenuated with a series of neutral-density filters (Kodak). The highest light intensity achievable in our experiments was 10 (arbitrary units), corresponding to bright room illumination. Light intensities in text and figures are presented with dimensionless numbers; e.g., 5 × 10⁻⁶ intensity is 2 × 10⁶ lower than the maximal intensity.

Only green-sensitive photoreceptors, which showed stable resting potential ≤−45 mV and light responses, were used for analysis. Recordings were performed at room temperature (20–22°C) during the subjective night of the control group.

Relative expression levels of mRNA were determined by real-time quantitative PCR as described previously (French et al., 2015). In brief, total RNA was extracted from 14–20 retinas from each experimental group, using a RNeasy Plus mini kit (Qiagen). mRNA was evaluated using an Experion RNA Analysis kit (Bio-Rad). 50 ng total RNA was used for first-strand cDNA synthesis with ProtoScript II reverse transcription (New England BioLabs). Quantitative PCR was performed using GoTaq qPCR Master (Promega) on a CFX96TM real-time PCR detection system (Bio-Rad). All PCR runs were performed three times. Gene expression levels, PCR efficiency, and the standard error of measurement were calculated using CFX Manager (Bio-Rad). Primer sequences for the specific and reference genes are provided elsewhere (French et al., 2015). Amplification efficiencies of the primers were determined using serially diluted cDNA samples.

To determine the information transfer rate, we used a 60-s stimulus consisting of 30 repetitions of a 2-s Gaussian white noise (GWN) sequence, with mean contrast of 0.36 and a 3-dB cutoff frequency (f_3dB) of 50 Hz. The GWN sequence was preceded by an adapting 0.5-s steady light interval of the same mean intensity to accommodate the initial transient. Data analysis was done in Matlab (MathWorks) as described previously (Frolov, 2015). In brief, a 2-s signal S(t) was obtained by averaging voltage responses to 30 repetitions of the 2-s sequence. S(t) was then converted into S(f) by a fast Fourier transformation (FFT). The noise N(f) was then obtained by subtracting the signal estimate from the original (noise-containing) sequences, converting them to spectra by FFT and averaging all 30 noise spectra. The signal gain of voltage responses |T(f)| was calculated by dividing the cross-spectrum of photoreceptor input (GWN contrast, C(f)) and output (photoreceptor signal) S(f)·C*(f), with * denoting the complex conjugate, by the autospectrum of the input C(f)·C*(f) and taking the absolute value of the resulting frequency response function T(f): T(f) = S(f)·C*(f)/C(f)·C*(f). The Shannon information rate (IR) was calculated as IR = ∫{log₂[|S(f)|/|N(f)|+1]}df within the frequency range from 1 to 50 Hz.

At the initial stage of statistical analysis, the Shapiro–Wilk normality test was applied to data samples to determine if they could be analyzed using parametric statistical methods. Data in the samples that did not pass the normality test were presented using medians and interquartile ranges (25% quartile:75% quartile; e.g., Fig. 7 D). To evaluate differences between such samples, the Mann–Whitney U test (MWUT) was used. In Fig. 7, dependencies of information-processing parameters on frequency are shown as median ± median absolute deviation (m.a.d.). The samples that passed the normality test were analyzed with parametric statistical methods as indicated. Such data are presented as mean ± SD and were compared using a two-tailed unpaired t test with unequal variances. Spearman’s rank order correlation coefficient (ρ) was used in analyses of correlations. In figures, *, P < 0.05; **, P < 0.01. Throughout the text, n stands for experimental group size.

Patch-clamp recordings were performed from CL, control, and CD photoreceptors between days 100 and 150 into the different light regimens. Basic electrophysiological properties include resting potential, input resistance, and whole-cell capacitance. Of these three, only whole-cell capacitance (C_m) was significantly different between the experimental groups (Fig. 1, A–C). Mean C_m in CL cockroaches was 2.4 times smaller and in CD cockroaches 1.5 times higher than in control (Fig. 1 A). C_m distributions are shown in Fig. 1 B.

Absolute sensitivity to light was estimated by counting quantum bumps (elementary photoreceptor responses) evoked by continuous low-intensity light stimulation. Bump rates were first obtained at different light intensities and then recalculated for the common light intensity corresponding to 5 × 10⁻⁶ light intensity in Fig. 4 B. Absolute sensitivity was strongly reduced in CL and increased in CD photoreceptors in comparison to control (Fig. 1 C). Although a strong positive correlation was found between C_m and absolute sensitivity in control (ρ = 0.73, n = 59, P < 10⁻⁶, Fig. 1 C), no statistically significant correlation was found in either CL or CD groups because of relatively small sample sizes.

Next, we tested if prolonged exposure to CL or dark changed the latency of elementary (quantum bumps) and macroscopic responses. Quantum bumps were evoked in voltage-clamp mode by 1-ms flashes of green light of such intensity as to trigger bumps with a probability of <0.7. Bump latency was determined as an interval between the onset of light and the time quantum bump amplitude reached 10% of its maximum value. Fig. 2 (A–C) shows typical responses of CL, control, and CD photoreceptors, respectively, with stimulus given at 0 ms and red dashed lines indicating median bump latencies. Normalized bump latency distributions are shown in Fig. 2 D. Mean bump latency was significantly smaller in CL than in control and CD photoreceptors (Fig. 2 E).

We also compared mean amplitudes and half-widths of current quantum bumps from the three experimental groups. Although there appeared to be differences in mean amplitudes and half-widths, with the smallest mean amplitude and largest mean half-width observed in CD photoreceptors, these differences were not statistically significant and could in principle be explained by large residual uncompensated capacitance in CD but not CL photoreceptors, which slows clamp speed compared with control and especially CL photoreceptors.

Next, we tested if the differences in photoreceptor capacitance altered the amount of low-pass filtering by the membrane. Indeed, dramatic differences in voltage bumps were observed between CL and photoreceptors of the two other groups (Fig. 3). Fig. 3 (A–C) shows examples of current and voltage bump responses to low-intensity continuous illumination in the same CL, control, and CD photoreceptors. Voltage bumps in CL photoreceptors were much bigger than in control and CD photoreceptors at similar membrane potentials. To compare average voltage bumps, we selected subgroups of photoreceptors so that the average resting potentials were approximately the same for each group (Fig. 3 D). This was done to equalize the effects of voltage-dependent membrane conductances on membrane time constant (τ = C_m·R, where R is total membrane resistance), and therefore, on the amplitude and kinetics of voltage bumps. When the average voltage bumps were normalized, it emerged that voltage bumps in CL rise and decay faster than in control and CD photoreceptors (Fig. 3 E).

Typical examples of light-induced current (LIC) evoked by 4-s pulses of steady light in 10-fold intensity increments from CL, control, and CD photoreceptors are shown in Fig. 4 A. Average dependencies of sustained LIC on light level are shown in Fig. 4 B. Sustained LIC in CL photoreceptors was significantly smaller than LICs in two other groups at all light backgrounds. It should be noted that many CL photoreceptors had such a low absolute sensitivity that only quantum bumps could be evoked in the brightest light. Such cells with effectively zero macroscopic LIC were included neither in the average of Fig. 4 B nor in the statistical group comparisons presented in the figure legend, so the numbers provided represent substantial overestimates of the population-average LIC in CL photoreceptors.

Although the sustained LIC values in control and CD photoreceptors were not significantly different at intensities 5 × 10⁻¹ and 5 × 10⁻² (Fig. 4 B), at still dimmer intensities of 5 × 10⁻³ and 5 × 10⁻⁴, the sustained LIC recorded from CD photoreceptors was notably higher than LIC in control (Fig. 4 B). These results are consistent with the increased absolute sensitivity of CD photoreceptors. Also consistent with the previous findings, a strong correlation was found between C_m and sustained LIC amplitude at 5 × 10⁻¹ (Fig. 4 C). For the combined CL, control, and CD data, the Spearman’s ρ coefficient was −0.64 (P < 10⁻⁵). Likewise, as can be seen from Table 1, LIC at 5 × 10⁻¹ correlated strongly (ρ = −0.71) with absolute sensitivity at 5 × 10⁻⁶.

Several voltage-activated K⁺ (Kv) currents have been found in P. americana photoreceptors, including a transient IA of unknown molecular origin, and a delayed rectifier (IDR) mainly mediated by Eag channels (Immonen et al., 2017). There were only small differences between Kv currents in the three experimental groups. Fig. 5 A shows a representative Kv current recording from a CL photoreceptor. Fig. 5 B compares conductance–voltage relationships for the IDR in CL, control, and CD photoreceptors. Maximal conductance (G_max) and half-activation potential values were obtained by fitting the relationships with a sigmoidal function. G_max was slightly smaller in CL than in control and CD photoreceptors (Fig. 5 B). These changes in G_max values are consistent with the previously reported moderate positive correlations between C_m and G_max values (Salmela et al., 2012). Half-activation potential values were not different between the groups.

Next, we investigated the effects of chronic light exposure/deprivation on information transfer. A 60-s GWN stimulus was used over a range of light intensities in 10-fold increments. As in the previous patch-clamp studies, dependencies of information rate (IR) on light intensity in control and CD photoreceptors were usually characterized by the presence of a clear IR maximum (IR_max) in relatively bright light, with a sharp IR decrease in still brighter light because of saturation of phototransduction (Frolov, 2016). In contrast, because of low sensitivity to light, voltage responses of CL photoreceptors to GWN usually showed no such IR saturation. Fig. 6 A demonstrates typical voltage responses of CL, control, and CD photoreceptors associated with IR_max. The voltage noise was the highest in the CL photoreceptor and the lowest in the CD photoreceptor.

Consistent with the differences in LIC (Fig. 4), voltage responses to GWN were characterized by the lowest sustained depolarization in CL photoreceptors and the highest in control and CD photoreceptors (Fig. 6 B). When dependencies of IR on background in each photoreceptor were averaged, excluding IR values associated with saturated responses in backgrounds brighter than those eliciting IR_max responses (Fig. 6 C), the following picture emerged. First, at all light levels, control and CD photoreceptors transferred significantly more information than CL photoreceptors. Second, in dim backgrounds, CD photoreceptors were characterized by higher information rates than control photoreceptors. For example, at the intermediate light, 5 × 10⁻³ IR was 8.0 ± 4.0 bits/s in control (n = 21) and 12.3 ± 6.2 bits/s in CD photoreceptors (n = 9, P = 0.03, unpaired t test). This was because of decreased noise in CD relative to control photoreceptors.

Accordingly, IR_max values were observed in relatively dim light in control and CD, and in bright light in CL photoreceptors, with more IR_max values detected in relatively bright backgrounds in control than in CD photoreceptors (Fig. 6 D). The IR_max values were about the same in control and CD but much smaller in CL photoreceptors (Fig. 6 E). However, it should be noted that the mean CL IR_max value could be an underestimate because most IR_max responses were recorded at the highest light intensity technically feasible in our experiments (Fig. 6 D). It is possible that IRs of such photoreceptors did not reach their maxima and might be higher in still brighter light.

We were interested in differences in information processing between the three groups at the peak of their photoreceptor performance. We therefore compared signal gain, signal power, and noise power functions in the frequency domain for the responses associated with IR_max. Median dependencies of signal gain on frequency are shown in Fig. 7 A. The values of 3-dB membrane “corner” frequencies (f_3dB) were obtained by fitting IR_max signal gain functions in each photoreceptor with a first-order Lorentzian function. The values of f_3dB were twice as high in CL as in control and CD photoreceptors (Fig. 7 B). Median frequency-dependencies of signal and noise power for responses associated with IR_max are shown in Fig. 7 C. Consistently with the weaker low-pass filtering, signal power was higher in CL than in control or CD photoreceptors in the higher-frequency region. However, because of the opposite tendencies in the lower-frequency region and a rapid decrease in signal power with frequency, total signal power for any of the three conditions was not statistically different (Fig. 7 D). Consistent with the increased voltage bump noise, the total noise power was significantly higher in CL than in control and CD photoreceptors (Fig. 7 D). Because of high noise, the median signal-to-noise ratio (SNR) function for CL was smaller than the SNR functions for control and CD photoreceptors (Fig. 7 E).

We also investigated the expression of genes encoding some proteins important for phototransduction. Quantitative PCR analysis of the mRNA levels for three opsins and two light-activated channels, which were previously identified in the cockroach retina (French et al., 2015; Saari et al., 2017), found that the mean expression of the dominant green-sensitive opsin (GO1) was strongly up-regulated in CD retinas (Fig. 8). This appears to be the photoreceptors’ primary response to light deprivation and contributes to their improved light sensitivity. GO2 and UVO may be less important for cockroach vision in visible light, because their expression decreased during long-term light deprivation. Both cation channels that are required for transduction in cockroach photoreceptors, TRP and TRPL, were slightly down-regulated in CD retinas, suggesting that lack of a functioning transduction cascade reduced their expression. However, the large amount of GO1 still resulted in higher sensitivity and longer-lasting responses. Expression of the Gq protein that mediates TRP and TRPL signaling and Arrestin (Arr) that terminates rhodopsin signaling and halts Gq production were decreased in CD but not in CL photoreceptors. Expression of phospholipase C (PLC), which is activated by the Gq, was unchanged in both experimental conditions. Overstimulation by CL slightly up-regulated GO1, GO2, Arr, and TRPL, resulting in faster and more transient responses but lower absolute sensitivity than control.

In this work, we investigated the effects of long-term exposure to bright light or chronic light deprivation on photoreceptor properties and function in the nocturnal insect P. americana. This is the first study of phenotypic plasticity of microvillar photoreceptors other than dipterans, and its results are only partly consistent with findings in flies (Deimel and Kral, 1992; Wolfram and Juusola, 2004).

Distinct combinations of electrophysiological adaptations were found in CL-exposed and CD-exposed photoreceptors. In comparison to control, CL photoreceptors were characterized by reduced membrane capacitance, sensitivity to light, sustained light-induced and voltage-activated K⁺ currents, latency of phototransduction, sustained depolarization, and maximal information rate. They also exhibited enlarged voltage bumps, voltage bump noise, and membrane corner frequency. All these parameters changed in the opposite directions in CD photoreceptors. However, differences between CD and control photoreceptors were smaller than those between CL and controls. For instance, sustained light-induced current was significantly larger in CD than in control photoreceptors in relatively dim light (Fig. 4), whereas in bright light the relative difference was smaller. Although this could be explained by the relatively small experimental group sizes and the intrinsically high variability in the Periplaneta photoreceptor properties (Heimonen et al., 2006), it is more likely that this reflects the sensitivity-boosting adaptations in CD photoreceptors (see below), which simultaneously increased the total light-induced current in dim light. On the other hand, in bright light, Ca²⁺-dependent light adaptation (Immonen et al., 2014) could have a stronger suppressing effect on light-induced current in CD than in control photoreceptors. These observations are also in line with the finding of higher information rates in CD than in control in relatively dim backgrounds (Fig. 6 C), indicative of shifting operational ranges of photoreceptors in the three experimental groups.

The changes seen in photoreceptors exposed to CL clearly facilitated faster and more broadband signal processing at the expense of sensitivity to light, whereas the light-deprived photoreceptors favored absolute sensitivity by larger rhabdom area and a surge in expression of green opsin. A consequence of the enlarged rhabdom reflected in higher capacitance was the finding of slower voltage responses. These observations are fully consistent with the classic visual ecological paradigm explaining physiological differences in photoreceptor functioning between diurnal and nocturnal species as a result of sensitivity/speed trade-offs (Weckström and Laughlin, 1995; Cronin et al., 2014). However, it should be noted that CL photoreceptors were not able to eliminate the excessive voltage noise, which prevented translation of expanded bandwidth into superior information capacity (see below).

Our results indicate that the differences between the experimental groups originate from two independent sources: extensive changes in the size of the rhabdom and intensive changes in the speed of phototransduction.

We have previously shown that the variability in photoreceptor size of several insect species as approximated by membrane capacitance is strongly linked to variabilities in the absolute sensitivity, amplitude of macroscopic sustained light-induced current, membrane corner frequency, and maximal information rate (Frolov, 2016). Moreover, moderate to strong positive correlations were found for various combinations of these parameters (Table 1). Although correlation does not necessarily mean causation, the common factor of rhabdom size underlies all these parameters. If membrane capacitance is mainly determined by the rhabdom area, this would explain the correlations between capacitance and absolute sensitivity, and between capacitance and sustained light-induced current. On the other hand, information rate depends on the number of microvilli, the photoreceptor’s sampling units, which is also reflected in membrane capacitance, albeit indirectly.

The area of light-sensitive membrane is directly proportional to the number of microvilli and, together with the somatic and axonal membrane, contributes to the capacitance. Axonal membrane is absent in the dissociated ommatidia. In P. americana, using data from previous transmission electron microscopy studies (Frolov et al., 2017) and our unpublished observations indicating that the ommatidium is two tiered, the soma can be approximated by a cylinder of 10 µm in diameter and 100 µm in length, whereas the microvillus has a diameter of 68 nm and average length of ∼3 µm. Disregarding the flanking surfaces and assuming that the photoreceptor contains 30,000 microvilli (a Drosophila estimate, much lower than the P. americana estimate; Frolov et al., 2017) gives ∼3,100 µm² of the somatic membrane and 19,000 µm² of the rhabdomeric membrane.

How much of the membrane area can be captured in capacitive transients in voltage-clamp experiments considering that the photoreceptor is a slender cell containing tens of thousands of even more slender microvilli? To determine potential contribution of incomplete space-clamp to underestimation of whole-cell capacitance, the following calculations were performed. The validity of capacitance measurements in approximating the size of the photoreceptor depends on whether the cell can be considered isopotential. Our recordings were performed on dissociated ommatidia lacking axons. Under such conditions, the photoreceptor can be represented by two compartments: the soma and the rhabdomere. Calculation of length constant for the soma using a relatively low specific membrane resistivity of 1 kΩ·cm² and a normal specific intracellular resistivity of 200 Ω·cm gives a length constant of 350 µm. The microvillus has internal diameter of ∼60 nm. Disregarding the resistivity of the extracellular space and using the specific membrane and intracellular resistivity values above yields a length constant of 27 µm. Considering the very narrow extracellular space between microvilli, and assuming a 5-nm distance between neighboring microvilli and specific intracellular and extracellular resistivity of 300 Ω·cm, yields a microvillus length constant of ∼8 µm. These estimates indicate that when ion channels are closed in the dark at the resting potential, strongly increasing specific membrane resistivity, both the soma and the rhabdomere in P. americana can be considered isopotential.

In Drosophila photoreceptors, voltage-activated K⁺ channels were shown to be expressed in the soma, with delayed rectifier Shab channels found at the microvillar bases (Rogero et al., 1997; Hardie and Raghu, 2001). We previously showed for Periplaneta and several other species that although delayed rectifier K⁺ conductance correlates positively with membrane capacitance, such correlations are usually much weaker than between capacitance and light-induced current. This implies that the delayed rectifier and light-activated channels are situated in different compartments (Frolov, 2016). Here we also showed that the changes in sustained K⁺ conductance are much smaller than the corresponding changes in light-induced current (Fig. 5). This is consistent with both the extra-rhabdomeric localization of the K⁺ channels and the hypothesis that the main changes occur in the rhabdom. It should be noted that the physical dimensions of dissociated ommatidia in all three experimental groups were similar.

The changes in capacitance directly affected two major aspects of information processing: membrane corner frequency and noise power. The former is determined by a product of membrane capacitance and momentary resistance, which mainly depends on sustained K⁺ conductance. As we demonstrated here, changes in capacitance were accompanied by disproportionally smaller changes in the K⁺ conductance. This significantly increased corner frequency in CL photoreceptors and improved transfer of higher-frequency components of the stimulus (Fig. 6 D). However, all possible advantages of such signal power redistribution (Fig. 7 A) were negated by a drastically increased voltage noise (Fig. 7, B and C) because of large voltage bumps (Fig. 3), which are also caused by reduced low-pass filtering. Why then do fly photoreceptors, characterized by relatively small capacitances (Frolov et al., 2016), not demonstrate such high-voltage noise? One reason could be that, in contrast to Periplaneta, fly photoreceptors express relatively few high-conductance TRPL channels. In contrast, we previously found that knockdown of the TRPL gene causes a dramatic reduction in voltage noise in cockroach photoreceptors (Saari et al., 2017). Accordingly, here we found a small increase in the TRPL expression in CL retinas and an increase in voltage noise (Fig. 8).

Changes in phototransduction manifested in the decrease of mean bump latency by ∼14% in CL photoreceptors and in a small increase in CD photoreceptors (Fig. 2). Decreased latency is associated with a smaller dispersion of quantum bumps in response to a flash of light and therefore with better temporal resolution of a contrast-modulated stimulus by the photoreceptor membrane, and vice versa (Wolfram and Juusola, 2004). Faster phototransduction and voltage responses are strongly associated with diurnal, fast-flying, maneuverable species (Weckström and Laughlin, 1995; Frolov, 2016). As quantum bump latency is thought to be mainly determined by molecular events during the first part of phototransduction cascade (Wolfram and Juusola, 2004), the reduced expression of Gq and Arr (Fig. 8) seems to be consistent with the slightly increased latency in CD photoreceptors.

The effects of rearing Drosophila under normal illumination (a 12-h light/12-h dark cycle) versus in the dark for several generations were previously investigated for changes in photoreceptor properties (Wolfram and Juusola, 2004). Intracellular recordings from photoreceptors were performed between days 2 and 10 after eclosion. In light-reared flies, the following changes were observed in comparison to dark-reared flies: a significant decrease in input resistance without effects on capacitance, accelerated light response, lower sustained depolarization, slightly increased signal and noise power, decreased membrane time constant, and higher information rate. These adaptations can be explained by acceleration of phototransduction and increased membrane leak conductance. In addition, the authors investigated the effects of short-term (2-h) exposure to either light or darkness in both experimental groups before recordings and found that such interventions substantially modify the original photoreceptor phenotypes.

Although some changes reported in Drosophila are similar to our findings, others are not. We did not observe any effect of light/dark rearing on input resistance. In contrast, although membrane capacitance was not altered in Drosophila, it was drastically changed in Periplaneta. Changes in signal and noise power spectra, and in membrane corner frequency, were quite similar, although noise increased more in Periplaneta CL photoreceptors than in light-reared fruit flies. Because of this, information rate decreased in Periplaneta CL but increased in light-reared Drosophila. Phototransduction was accelerated both in CL cockroaches and light-reared flies in comparison to the respective controls. Therefore, although the adaptive changes in Drosophila, e.g., in input resistance, appear to be mainly functional, the changes in Periplaneta associated with C_m seem to be caused by predominantly structural modifications. However, comparing the results of these two studies is problematic because of differing methodologies: in the Drosophila study, rearing under dissimilar conditions for few generations may not produce major changes between the groups (but see Izutsu et al., 2012) if it is the individual history that matters. In the case of fruit flies, this history includes larval and young adult exposure to light or dark. Moreover, the duration of long-term light exposure of the adult photoreceptors was not controlled for, as recordings were performed between days 2 and 10 after eclosion. The relatively short Drosophila life span and practical difficulties with recording from older flies preclude longer experiments like those on mature photoreceptors presented here.

We studied changes in photoreceptor function in P. americana after prolonged exposure to strongly differing illumination conditions. Chronic and drastic alterations in the visual input elicited distinct and opposing patterns of functional adjustments. Our data indicate that most of the long-term light-driven adaptations in Periplaneta can be linked to changes in the size of rhabdom that effectively adjust photoreceptor function to new environmental conditions.

This research was supported by the Natural Sciences and Engineering Research Council of Canada grants RGPIN/5565 (to P.H. Torkkeli) and RGPIN/03712 (to A.S. French).

The authors declare no competing financial interests.

Author contributions: R.V. Frolov: conceptualization, electrophysiology, data analysis, manuscript writing, review, and editing; E.-V. Immonen: electrophysiology; P. Saari: manuscript review and editing; H. Liu: quantitative PCR experiments; P.H. Torkkeli and A.S. French: interpretation of the data, project administration, manuscript writing, and editing.

Richard W. Aldrich served as editor.