Flickering flash signals and mate recognition in the Asian firefly, Aquatica lateralis

Flash recording

All field recording and experiments were performed at the paddy field in the Northern Chita Peninsula, Aichi Prefecture, central Japan, in June and July between 2003 and 2016. The ambient temperature at the firefly’s active period was measured using a thermometer. The flashes were recorded with a digital video camera (NV-GS-400, Panasonic, Japan) mounted on a tripod at a height of 30–50 cm from ground and a distance of 1.0–1.5 m away from the specimen. Isolated specimens were selected for recording to exclude the background light from other nontarget specimens. When another specimen appeared near the target specimen, the video recording was cancelled. When a female copulated during video recording in the field, her flashes until 1 min before copulation were regarded as those of a ‘receptive female’. To record the flashes of a ‘mated female’, the female specimens already mated were prepared in aquariums (because virgin and mated females cannot be distinguished in the field): the eggs were obtained from wild female specimens collected one year before at the same field and reared to adults; immediately after emergence the virgin female was confined in a small container with two cultured males for two nights to facilitate copulation. As the parents of the reared specimens were collected from the observation field (same genetic background), the rearing temperature was almost the same as that of the natural field, the emergence period of the cultured specimens overlapped with that of the natural population, the adult body sizes of the reared and natural specimens were indistinguishable, and the flash pattern of the cultured mated females was indistinguishable from that of the wild (potentially) mated females. Thus, we believe that there was no influence of different rearing environments, i.e., the flash behavior of the cultured mated female specimens is expected to be substantially the same as that of wild mated female specimens. To distinguish them from wild (potentially) mated females, the elytra of cultured mated females were marked with colored ink before placing them in the field, and after three days, the flashes of ink-marked specimens were recorded. Of note, we never observed male attraction and copulation in any of the mated females used for field observation; thus, the mated females were unreceptive.

Waveform analysis

Sequential still images were captured from video files at 30 frames per second using VirtualDub (GPL), and then the light intensities in the images were qualified (8-bit linear gray scaling from black to white at 0–255) using ImageJ software. In this study, we defined ‘flash’ as a luminescent waveform from baseline to baseline and ‘flickering’ as fluctuation above baseline in a single flash. The waveforms containing a saturated signal (255, white) were omitted. The waveforms of the maximum signal value lower than 50 were also omitted because of the difficulty in separating signal and noise. Approximately 10–90 waveforms per individual were analyzed; thus, the effect of the occasional interruption of the flash recording by the specimen’s movement and/or vegetation swinging between the specimen and the video camera is statistically ignorable. FD is defined as the time interval between the beginning and the end of a flash (Fig. S1). Flicker intensity (FI) was defined as

where p, t, and n denote the peak and the trough (local extrema) in the waveform of a flash and the number of toughs in the flash, respectively (Fig. S1). In total, we measured the FD and FI values of 347, 94, and 355 waveforms from 13 sedentary males, 7 receptive females, and 8 mated females, respectively. We did not consider the flash brightness as a factor because the measured value of the light intensity depends largely on the distance between the light source and the detector; thus, the actual brightness of the lantern cannot be practically measured in the field.

e-Firefly

For male attraction experiments, we built an electronic LED device, the e-firefly, to generate patterned flashes with various FDs and FIs using a chip LED (green type, λmax = 568 nm, Everlight Electronics, Taiwan; Figs. S2 and S3) with a microcontroller PIC16F628A (Microchip Technology, USA) (see Figs. S4-S5). An example of the program for the microcontroller is shown in Supplementary Data S1. The brightness was constant in all programs. Flickering frequency ranged between 5–12 Hz, which corresponds to that of sedentary male flashes (approximately 10 Hz)¹⁵. To prevent direct access of the attracted specimen to the light source, the chip LED was covered by a steel net painted green (see Fig. S2). For flying male attraction experiments, when the male landed within a 100-mm distance from the e-firefly, we judged the attraction to be a success; otherwise, it was a failure. For sedentary male attraction experiments, the e-firefly was placed 200–300 mm away from the sedentary male. When the approaching male touched the steel net covering the e-firefly, to warrant a positive approach, we measured the time the male remained on the net. If the male did not move away from the net for more than 2 min, we judged the attraction to be a success (strict criterion for judgment); otherwise, it was a failure.

Spectral measurement

The luminescence spectra of e-firefly and A. lateralis were measured using a Flame-S spectrophotometer (Ocean Insight, USA). The living A. lateralis specimens were anesthetized on ice and frozen at − 20 °C until use. The lantern started luminescence by thawing at room temperature, and the spectrum was measured during luminescence (within 5 min).

Statistical analysis

First, we considered a discriminant analysis using a logistic regression model that discriminates between receptive females and others in the observational data. We fitted several models with combinations of FD and FI, quadratic terms of FD and FI (FD², FI²), interaction of FD and FI (FD (times) FI), and temperature (T) as explanatory variables. Based on Akaike’s information criteria (AIC) values and model simplicity, we chose the logistic regression model with FD, FI, FD² and T as explanatory variables. Let (p)(({varvec{x}})) denote the conditional probability that a flash is from a receptive female given ({varvec{x}}=left(mathrm{FD},mathrm{ FI},mathrm{ T}right)) and (widehat{p})(({varvec{x}})) denote its estimate. The coefficients of the logistic regression model are estimated as follows.

[Model for the observational data with temperature (T)]

[Model for the observational data without temperature (T)]

where values in parentheses indicate standard deviations. The same applies hereafter. Temperature (T) is included in the model not because it affects the occurrence of receptive females but because it affects the FD and/or FI of receptive females. The AIC value increased by 30, which is substantial, when temperature was excluded from the model.

Figure 2 shows the FD and FI of each flash from receptive females, mated females and males with the discriminant boundaries of receptive females from others for (p=0.5).

We next considered a discriminant analysis for the experimental data. Let ({q}^{f}({varvec{x}})) denote the conditional probability that a flying male is attracted to a flash of ({varvec{x}}=left(mathrm{FD},mathrm{ FI},mathrm{ T}right)) and lands, and ({widehat{q}}^{f}({varvec{x}})) denote its estimate. Among several models we fit, the smallest AIC value is attained by the logistic regression model with FD, FI and T as explanatory variables, but the AIC is not much different from the model with FD and FI only.

[Model for flying males with temperature (T)]

[Model for flying males without temperature (T)]

For sedentary males, the model with the smallest AIC value includes all the quadratic terms of FI and FD but not temperature. Let ({q}^{s}({varvec{x}})) denote the conditional probability that a sedentary male is attracted to a flash of ({varvec{x}}=left(mathrm{FD},mathrm{ FI},mathrm{ T}right)) and ({widehat{q}}^{s}left({varvec{x}}right)) denote its estimate. The logistic regression model for ({q}^{s}({varvec{x}})) with the best AIC value is given as follows.

[Model for sedentary males]

Figure 3 shows successes and failures of attraction of flying males on the left and sedentary males on the right with estimated discriminant boundaries.

Let us now estimate probabilities that a flying male is attracted and lands or a sedentary male is attracted to a flash when a flash is from a receptive female or when a flash is either from a sedentary male or mated female. The probability that a flying male is attracted and lands when a flash is from a receptive female is a conditional probability and is expressed as follows.

Integrals are taken over the domain (Omega) of ({varvec{x}}=(FD, FI, T)) of all females and males, and (f({varvec{x}})) is the joint density function of ({varvec{x}}.) Because (f({varvec{x}})) is unknown, we use the empirical distribution of the observational data, and conditional probabilities given ({varvec{x}}) are replaced with their estimates by logistic regression models. Let ({{varvec{x}}}_{i}=left(F{D}_{i}, F{I}_{i}, {T}_{i}right), i=mathrm{1,2},dots N) denote the (i) th observation in the observational data. The estimates of probabilities are given as follows:

Thus,

Similarly, we have

The estimated probabilities are shown in Table 1.

Source: Ecology - nature.com