All experiments were performed on adult male cockroaches (Periplaneta americana). Animals were received as adults from Blades Biological Ltd (Cowden, Edenbridge, UK) and were subsequently kept at 25°C under a 12 h day/night rhythm.

Cockroaches were anesthetized for capture with CO₂. Isolation of ommatidia was done under red light conditions. After removing the body and antennae from the head, one eye was carefully sliced off with a sharp razor blade. The retina was dissected with a flattened insect pin and subsequently cut into several pieces. After an incubation time of ~20 min in extracellular solution complemented with 0.2 mg/ml collagenase type 2 (Worthington Biochemical Corp., Lakewood, NJ USA) and 0.2 mg/ml pancreatin (Sigma-Aldrich), retina pieces were gently triturated until ommatidia separated. Individual ommatidia, obtained after trituration with custom-made glass micropipettes, were allowed to settle in the recording chamber on the stage of an inverted microscope (Axiovert 35 M, Zeiss, Germany).

Patch-clamp recordings were performed as described before (Chyb et al., 1999). Briefly, an Axopatch 1-D amplifier (Molecular Devices, USA) was used; data were digitized and recorded on a laboratory computer using PClamp 10 software (Molecular Devices, USA). Recording electrodes had a resistance of 4–11 MΩ that allowed recordings in the whole cell configuration with an access resistance of <30 MΩ. Access resistance was monitored throughout the experiments. Seal resistance was typically greater than 10 GΩ, membrane resistance in darkness was between 250 MΩ and several GΩ. For most measurements a series resistance compensation of at least 80% was applied. However, during quantum bump recordings, no series resistance compensation was applied, because the recorded currents are then very small. If not stated otherwise, no liquid junction potential (LJP) was corrected (−4 mV). Photoreceptors were clamped to a holding potential of −70 mV. Cells were stimulated with a green (525 nm) light emitting diode (LED) via the fluorescence port of the microscope. Quantum bump experiments were sampled at 2 kHz with either 200 Hz (for steady light) or 500 Hz (for flashes) cut-off of low-pass filtering (80 dB/decade Bessel filter). In 14.5 min bump recordings, 1 kHz sampling and 200 Hz filtering was used. All other experiments were sampled at 2 kHz with 500 Hz or 1 kHz cut-off for filtering.

Cockroach bath solution contained (in mM): 120 NaCl, 5 KCl, 4 MgCl₂, 1.5 CaCl₂, 10 N-Tris-(hydroxymethyl)-methyl-2-amino-ethanesulfoncic acid (TES), 25 proline and 5 alanine, pH was adjusted to 7.15 (NaOH). Nominally Ca²⁺-free bath solution (“0” [Ca²⁺]) contained (in mM): 120 NaCl, 5 KCl, 4 MgCl₂, 10 TES, 0.5ĖGTA, 25 proline, and 5 alanine, pH 7.15 (NaOH). Intracellular solution contained (in mM): 140 KCl, 10 TES, 2 MgCl₂, 4 Mg-ATP, 0.4 Na-GTP, and 1 NAD, pH was adjusted to 7.15 (KOH). When 10 mM EGTA was used in the intracellular solution it was loaded with 4 mM CaCl₂ to obtain ~150 nM and ~6 mM free Ca²⁺ and EGTA, respectively (calculated in Webmaxc Extended by Chris Patton, Stanford University, USA). The intracellular solution for recording of reversal potential (E_rev) under bi-onic conditions contained (in mM): 120 CsCl, 15 TEA-Cl, 10 TES, adjusted to pH 7.15 (N-methyl-D-glucamine, NMDG, which is supposed to be impermeable). The extracellular solution for recording of E_rev of monovalent ions contained (in mM): (1) 130 NaCl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for Na⁺.; (2) 130 LiCl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for Li⁺; and (3) 130 KCl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for K⁺. Extracellular solutions for E_rev of divalent ions were (in mM): (1) 10 CaCl₂, 120 NMDG-Cl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for Ca²⁺; (2) 10 MgCl₂, 120 NMDG-Cl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for Mg²⁺; and (3) 10 BaCl₂, 120 NMDG-Cl, 10 TES, 25 proline and 5 alanine, pH 7.15 (NMDG) for measurement of E_rev for Ba²⁺. The presence or absence of Cl⁻ in the solutions did not change the properties of the LIC. All chemicals were purchased from Sigma Aldrich unless stated otherwise.

Quantum bump analysis was performed in MatLab (Mathworks, USA), and was mostly based on the analysis by Henderson et al. (2000). Bumps were detected using a preset amplitude threshold. Due to the relatively large size of bumps the threshold was set to a value that would easily stand out from noise (−4 pA). Fused, double peaked bumps were removed. After detecting the location of bump peaks, bump beginnings and ends were determined from the first points around the peaks that were equivalent to two standard deviations of background noise. Bump parameter distributions were fitted with either the normal distribution, (1)P(p)=12πσ2exp(−(p−m)22σ2), or lognormal distribution, (2)P(p)=exp(−ln(p/tpk)22s2), where σ is the standard deviation, m is the mean, p is the fitted parameter, t_pk is the time-to-peak, and s is the skewness.

If a light response consists of discrete responses, the probability p_n for the occurrence of n responses per time period will follow the Poisson distribution: (3)pn=e−mmnn!, where m is the mean number of responses, i.e., absorbed photons, which can be estimated from the probability of failed absorption (n = 0): (4)m=−ln(p0).

The probability that n or more photons are required to elicit a light response follows the cumulative Poisson distribution: (5)p≥n=∑n=0∞e−mmnn!.

By assuming that the dispersion of response size, in this case the response integral, or charge (Q), follows the Gaussian distribution, the average size produced by a single photon absorption can be estimated via a sum of Gaussian functions weighted according to the Equation (3): (6)p(Q)=∑n=0∞e−mmnn!12π(σ02+nσ2)exp(−(Q−nμ)22(σ02+nσ2))​, where σ₀ is the standard deviation of failed responses and σ is the standard deviation of single photon response Q. μ is the mean Q of a single photon response, estimated from relation Q_a/m, where Q_a is the overall mean Q of failures and successful responses. The Poisson statistics presented in this study are based on (Hecht et al., 1942), (Baylor et al., 1979), and (Henderson et al., 2000).

The permeability ratios under bi-ionic conditions were calculated from E_rev of LICs according to (Hille, 2001): (7)PM:PCs=[Cs+]in[M+]outexp(ErevFRT) and (8)PD:PCs=[Cs+]in4[D2+]outexp(ErevFRT)(exp(ErevFRT)+1), where P stands for permeability, M for monovalent ion, D for divalent ion, F for Faraday constant, R for gas constant and T for temperature.

Data analysis and statistics were performed with Origin 9 (OriginLab Corp., Northampton, MA, USA) and MatLab (Mathworks, USA). The comparison of Spearman's ρ correlation coefficients, sample normality tests (the D'Agostino normality test) and the Mann-Whitney tests (two-tailed; p < 0.05) were performed in Origin 9. Numerical results are presented either as mean ± standard deviation or median followed by interquartile range in brackets (i.e., 1st–3rd quartile).

Bump-like discrete events with variable amplitudes could be elicited in whole-cell voltage-clamp of cockroach photoreceptors with either continuous light (Figure 1A) or short flashes (Figure 1B), as also described in a recent study of cockroach voltage-dependent currents (Salmela et al., 2012). In order to observe these discrete events with continuous light, the light level was first adjusted with neutral density filters to be low enough not to produce any response. The light level was then increased until discrete inward current jumps could be observed (Figure 1A). Further increase in light intensity increased the frequency of these events until clear summation started to occur, indicating that the responses were indeed bumps. Similarly, when short flashes were used, the light level was increased until it was capable of eliciting a response with ~50% success rate. The flash recordings usually consisted of sets of 1 s recordings repeated 100 times. The health of the cell and recording conditions (resting potential and series resistance) was routinely checked between these sets. Using a similar protocol as in Drosophila bump studies (Henderson et al., 2000), the cockroach bumps were analyzed in terms of latency (the time between the flash occurrence and response onset, T_lat), 50% rise time (T₁), 50–100% rise time (T₂), 50% decay time (T₃), and peak amplitude (A) (Figure 1B). The Q of quantum bumps was also calculated.

In order to be considered as single photon absorptions, the cockroach bumps should follow Poisson statistics. First we wanted to see whether the distributions of time intervals between bumps recorded during continuous illumination followed the Poisson prediction (see, e.g., Yeandle and Spiegler, 1973). This could be tested by fitting a single exponential to inter-bump-interval histogram with the reciprocal of the time constant set to the measured bump rate. Figure 2A shows that the cockroach bump intervals indeed follow Poisson distribution. Moreover, if cockroach bumps resulted from the absorptions of at least one photon, the probability of having a successful flash-evoked response should follow Equation (5) with n = 1. To test this we made a series of repeated light flash experiments where the intensity of a light flash was varied by changing its duration (the amount of effective photons per flash was estimated by bump calibrations in dim continuous light). Each set of intensities (100–200 repetitions) produced a successful response at a different rate. These rates, or the frequencies of seeing, were then plotted against flash intensities (Figure 2B). On the assumption that one or more photons are required to evoke a response, the data was well-fitted with Equation (5).

The stochastic nature of photon absorption will also result in increased probability of having more than one photon absorptions per flash when intensity is increased. Therefore, based on the Poisson prediction, if Q follows Gaussian distribution, the Q distribution of all responses should consist of the sum of Gaussian distributions (Baylor et al., 1979). To test this we applied series of short flashes (containing either 0.3 or 1.5 photons on average) and collected the Q of both failures and successful responses into a histogram (Figures 2C,D). The results were then fitted with Equation (6), σ being the only free parameter. The sum of Poisson-weighted Gaussian functions fitted the data reasonably well; note that the large dispersion of event size caused a significant group overlap (Figure 2D). Together these results strongly support the notion that the recorded bumps conform to Poisson predictions and represent the elementary responses of cockroach photoreceptors to single photons.

We then tested if the occurrence or waveform properties of cockroach bumps change in time by recording responses to 14.5 min-long pulses of dim light. In all cells tested (N = 6) both A and T₃ had a tendency to drift until reaching a stable level, while T₁ and Q remained relatively stable throughout the recording period (Figures 3A,B). Plotting bump peak times together with the running numbers of collected bumps indicated that the bump rate, i.e., light sensitivity, remained roughly the same during the stimulation period (Figure 3C). These results, together with relatively stable resting potential (always below −50 mV) and cell input resistance, suggested that the minor drifting in the observed parameters was not a result of cell rundown. Instead, the results implied that the cells were possibly reaching an equilibrium that might involve mixing of cytosol and pipette solution.

Next, flash-induced bumps were more thoroughly analyzed in terms of parameters presented in Figure 1B. However, this time more accurate quality criteria were used to select the bumps. First, the flash intensity was set low enough to maximize the number of single photon absorptions per flash (<50% success rate). Secondly, due to possible drifting of bump waveform, the distribution of A had to pass the D'Agostino normality test in order to be accepted for analysis. As a result bumps from five photoreceptors were accepted for analysis. Figure 4 shows the average bump calculated from the averages of the five cells (also shown in light gray traces). The distributions of the analyzed bump parameters indicated that they were mostly from a single class of bumps (Figure 5). The T_lat distribution (Figure 5C) was well-fitted with a lognormal function. Median T_lat across all cells (N = 5) was 58 (49–69) ms, and the average t_pk and average half-width (2.35·t_pk·s; from Howard et al., 1984), derived from Equation (2), were 53 ± 11 and 16 ± 6 ms, respectively. Table 1 summarizes the results of statistical analysis for other bump parameters. The non-exponential nature of parameter histograms (Figure 5) suggest that the parameters are not produced by a simple 1st order reaction system (Kirkwood and Lisman, 1994). Such a stochastic system would also produce coefficients of variation (CV), i.e., standard deviation divided by mean, close to unity (Henderson et al., 2000). However, CVs in Table 1 indicate that cockroach bump parameters do not vary as much, but differ from those in Drosophila (Henderson et al., 2000).

Interdependence of bump generation processes can be assessed by examining correlation coefficients between bump parameters (see, e.g., Howard, 1983; Henderson et al., 2000). Table 2 shows the values of Spearman's ρ calculated for two representative cells. Although many different small but statistically significant correlations were found (e.g., between T_lat and T₁) in the five analyzed cells, the only correlations showing consistency were those between A and T₃ (ρ = 0.17–0.47), and T₂ and T₃ (ρ = 0.26–0.46). These results suggest that at least T_lat and bump waveform (including A) are generated by processes independent of each other.

Thus, far, macroscopic LICs in cockroach photoreceptors have been briefly described in two studies from our group (Heimonen et al., 2012; Salmela et al., 2012). In the latter study, the major finding regarding LICs was that the functional variation could stem from differences in phototransduction processes among photoreceptors (Heimonen et al., 2012). However, it remained unclear what are the processes involved, and how the LICs would behave at different light levels. To answer those questions we recorded LICs at several light levels by using both flashes (10 ms) and pulses (10 s) of light. Our previous study had indicated that the electrical properties, such as membrane capacitance, tend to vary a lot among cockroach photoreceptors (Salmela et al., 2012). Therefore, membrane capacitances were also determined for each cell. Figure 6A shows representative LICs recorded from a 493 pF cell. Figure 6B shows the amplitudes of the LIC in the same cell, together with results in two other cells, as a function of relative light intensity. The behavior of LIC amplitudes suggests that the large capacitance cells had higher light sensitivity than their small capacitance counterpart. However, once the light intensities were adjusted to match the number of effective photons (estimated from bump calibrations), the LIC amplitudes fell into a same range (Figure 6C). A rough estimation suggests that the LIC amplitudes increase linearly up to a range of 1–3 nA and 200–300 effective photons, with linear relation of 5.7 pA/effective photon.

As it was expected from earlier studies (Heimonen et al., 2006, 2012), the use of 10 s light pulses resulted in highly variable waveforms of LICs across cells. Figure 6D shows examples of two extreme shapes of LICs evoked by the identical 10 s light pulse. This seems to suggest that there could be a link between the LIC response size and the photoreceptor capacitance. In Figure 6E the average plateau responses (time interval between the 3rd and 8th s of a response) of LICs recorded in 21 cells are plotted against relative light intensity. First, it can be seen that the amplitudes span up to seven orders of magnitude of relative light intensities, indicative of high variability in light sensitivity across cells. Secondly, the grouping of cells by their capacitance suggests that capacitance positively correlates with light sensitivity. Again, after adjusting the light intensities according to bump calibrations (this time in terms of effective photons s⁻¹), the amplitudes narrowed down into a tighter group, which indicates that, except for the difference in sensitivity, the cells are functionally similar (Figure 6F). This conclusion was supported by finding that the size of quantum bumps did not seem to correlate with cell capacitance (Spearman's ρ = 0; p = 0.62; N = 5). In conclusion, since capacitance is an indirect measure of membrane area, the sensitivity differences may arise from differences in the size of rhabdomeres.

To see how cockroach LIC is dependent on Ca²⁺ we first tested the effect of a series of different extracellular Ca²⁺ concentrations on cockroach bumps. The test was performed with both continuous light (30 s light pulse) and 1 ms flashes in one cell. The concentration series used consisted of 7 different solutions covering a range from 1.5 mM to nominally Ca²⁺-free (“0” [Ca²⁺]; no added Ca²⁺ plus 500 μM EGTA). The concentrations were chosen following the study by Henderson et al. (2000). Surprisingly, the lowering of extracellular [Ca²⁺] had relatively weak effects on the bump waveform (Figure 7). The bumps induced by a 30 s pulse started to show dramatic effects only at “0” [Ca²⁺] (Figure 7A). With “0” [Ca²⁺] the bump waveforms became substantially slower and prolonged, causing extensive response fusion. Bumps elicited by repeated flashes behaved in a similar manner. The comparison of average bumps (calculated from 65 to 192 bumps) indicated that in Ca²⁺ concentrations from 1.5 mM down to 25 μM the bumps became only slightly smaller and slower (Figure 7B). Although T_lat increased in lower Ca²⁺ concentrations, the changes were relatively small. In “0” [Ca²⁺], the bump shapes were highly unpredictable, varying from smaller rapid events to multi-peaked rectangular-like responses (Figure 7C). Therefore, the analysis of bump waveform parameters was not considered viable in “0” [Ca²⁺]. However, the comparison of averages across three cells suggested that bump A did not drop substantially even in “0” [Ca²⁺] (−35.3 ± 17.7 pA in control vs. −30.0 ± 15.7 pA in “0” [Ca²⁺]; N = 65–185 bumps). The comparison of samples within each cell indicated a statistically significant difference only in one out of three cases. Another observation made in the same cells was that the T_lat dispersion appeared to increase when Ca²⁺ was omitted. The difference in median T_lat (cell 1: 43 (37–49) ms; cell 2: 45 (36–55) ms; and cell 3: 79 (63–96) ms under control conditions vs. cell 1: 53 (44–68) ms; cell 2: 59 (47–71) ms; and cell 3: 104 (89–162) ms in “0” [Ca²⁺]) was statistically significant within all three cells (p < 0.001; N = 65–185 bumps). In addition, the half-widths of lognormal fits of the T_lat distributions indicated that Ca²⁺ removal increased the dispersion of T_lat (13 ± 5 ms in control vs. 23 ± 12 in “0” [Ca²⁺]; N = 3).

The effects of extracellular Ca²⁺ manipulations were then tested with macroscopic LICs. First, the same set of Ca²⁺ concentrations as above was used with responses to a 10 ms flash in one cell (Figure 8A). As with bumps, the lowering of extracellular Ca²⁺ concentration resulted in only moderately slowed responses (a result of slower bumps and/or the increased dispersion of bump T_lat) until “0” [Ca²⁺] was used. In “0” [Ca²⁺], the LIC amplitude (from ~4 to ~10 nA) and duration increased dramatically. The second test involved using 10 s light pulses at different light levels before and after omission of extracellular Ca²⁺. The use of longer light pulses allowed us to observe changes beyond transient responses, and see how extracellular Ca²⁺ affects light adaptation. Since bump discrimination during continuous stimulation was practically impossible we had to rely on comparing effects using relative light intensities. The omission of extracellular Ca²⁺ resulted in an increase of amplitude and slightly prolonged duration of the LIC when compared to 1.5 mM Ca²⁺ (Figure 8B). The peak-to-plateau transition in the late phase of the LIC was also completely suppressed. However, response delay and the early onset of initial transient were almost as fast in “0” [Ca²⁺] as in 1.5 mM Ca²⁺. The effects were similar in other cells recorded, and the average plateau currents were systematically increased in “0” [Ca²⁺] (Figure 8C). This result suggests that Ca²⁺ influx may be necessary for proper response deactivation.

To learn more about the possible mechanisms involved in the activation of LICs we also attempted to manipulate intracellular Ca²⁺ by buffering with EGTA. Since EGTA is a relatively slow, millisecond scale Ca²⁺ chelator, its presence in the pipette solution should still allow transient increases in Ca²⁺, e.g., through light-gated channels in microvilli (Henderson et al., 2000). Therefore, in principle, if capacitive Ca²⁺ entry (i.e., the activation of InsP₃ or ryanodine receptors) was necessary for response excitation in the cockroach, EGTA should be able to inhibit the activation of the LIC. The effects of 10 mM EGTA were first tested on bumps. Figure 9A shows the behavior of bump A over a 870 s (14.5 min) light stimulation period started right after achieving the whole-cell configuration. Visual examination suggests that after 100 bumps A started to deteriorate. The frequency of the appearance of bumps also decreased with time, which can be seen from the sub-linear relation of bump number with time (Figure 9B).

Flash-induced bumps recorded subsequently in the same cell diminished to barely detectable levels (Figure 9C). This resulted in a relatively poor signal-to-noise ratio, which is why quantitative bump parameter analysis was not carried out in full. T_lat from 72 events were collected, resulting in a median of 84 (66–99) ms. However, although showing otherwise similar behavior with EGTA, bumps recorded in another cell resulted in much shorter T_lat, with median of 38 (34–45) ms (N = 110), making definitive conclusions about the effects of EGTA on transduction impossible. We then tested how a prolonged exposure to intracellular EGTA would affect the waveform of the macroscopic response. This was done by first recording a LIC induced by a relatively bright 10 s light pulse right after formation of the whole-cell recording configuration. After 25 min:s the same light pulse was repeated. The resulting LIC showed a clear drop in transient amplitude, and was reminiscent of a response recorded with lower light intensity in control conditions. This behavior was in line with the drop in quantum efficiency observed with bumps recorded in the presence of EGTA. In the same cell, the effects of extracellular Ca²⁺ removal were also tested (Figure 9E). Two notable observations were made. First, the responses in “0” [Ca²⁺] were very much like those detected without EGTA in the pipette (compare to Figure 8B). Secondly, the responses showed conspicuous inactivation in relatively bright light, suggesting the exhaustion of a transduction component.

The surprisingly large size of cockroach quantum bumps in “0” [Ca²⁺] suggests that the LIC may not be dependent in the influx of Ca²⁺ to such an extent as in the fruitfly photoreceptors. To estimate the Ca²⁺-selectivity of cockroach light-gated channels we determined the relative ionic permeabilities from LIC E_rev recorded under bi-ionic conditions. The permeabilities of extracellular cations were assessed relative to an intracellular cesium solution, which also contained 15 mM TEA-Cl to attenuate voltage-dependent potassium conductances. The approach was largely adopted from the study by Reuss et al. (1997). The E_rev were determined from LICs induced by a 50 ms flash, while the cell was clamped to different voltages in 2 mV increments (Figures 10A–C). The median E_rev measured under control conditions from 10 cells was 8 (5–11) mV (Figure 10D). The calculation of relative mono- and divalent cation permeabilities according to Equations (7) and (8) revealed only slight selectivity over other cation species for Ca²⁺ (Figure 10E). The median permeabilities relative to Cs⁺ for Ca²⁺ and Na⁺ were 13.9 (9.2–17.4) and 0.9 (0.9–0.9), respectively, giving a ~16:1 selectivity for Ca²⁺ over Na⁺.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.