Plausible link between circa‘bi’dian activity rhythms and circadian clock systems in the large black chafer Holotrichia parallela

Various environmental factors that are associated with physical and biological features change periodically on the Earth, and organisms have evolved endogenous rhythms with periods that approximate environmental cycles. Many organisms have circadian rhythms driven by the circadian clock with a period of about a day (Dunlap et al., 1999). The suppression of activities during an appropriate time of a day by the circadian clock is an adaptive behavior associated with escape from predation and other environmental dangers (DeCoursey et al., 1997, 2000). Significant deviation from cyclical periods of activity for 24 h decreases both survivorship and reproduction rates (Spoelstra et al., 2016). Endogenous rhythms with periods similar to other environmental cycles such as tidal, lunar and yearly cycles have also been progressively studied (Numata and Helm, 2014; Kaiser et al., 2016). Therefore, the use of an endogenous rhythm with a cycle that corresponds to environmental changes probably allows organisms to appropriately anticipate and prepare their physiological states.

However, rhythmicity associated with cycles that inconsistently approximate cyclical environmental changes has been reported. Two-day rhythms, referred to as circa‘bi'dian rhythms, were recorded in humans and mosquitos. For instance, Kleitman (1963) successfully developed 48 h oral temperature rhythms by enforcing 48 h sleep–wakefulness routines in humans. In an underground bunker, ∼50 h activity cycles were observed in an isolated male subject exposed to constant light conditions, but the cyclical period of the body temperature rhythm was ∼25 h (Aschoff et al., 1967). In the mosquito Culiseta incidens, a ∼46 h flight rhythm was observed in constant dark conditions (Clopton, 1984, 1985). However, occurrences of circabidian rhythms in these two species were infrequent and observed only when organisms were kept for long periods under constant conditions without time cues, or when they were forced to entrain to 2-day cycles (Czeisler et al., 1980; Clopton, 1985). Therefore, it is not known whether circabidian rhythms actually occur in nature.

In the large black chafer Holotrichia parallela (Coleoptera: Scarabaeidae; previously called Lachnosterna morosa), a serious agriculture pest in East Asia, adult populations appear on the ground every 2 days (Yoshioka and Yamasaki, 1983). Measurements of pheromone titers in the pheromone glands of field-collected females showed that a peak occurred every 2 days (Leal et al., 1993). However, this periodicity has only been described in groups, and individual periodicity and the mechanisms behind the rhythm are not known. In contrast to humans and mosquitos, H. parallela populations seem to have 2-day periodicity in the field. If the 2-day periodicity of this species is formed by an endogenous free-running circabidian rhythm under constant conditions, H. parallela represents a good model that could aid our understanding of how an enigmatic rhythm with a period different from an environmental cycle is assembled by a clock mechanism and how it actually works in the field.

In this study, we chronobiologically examined the individual ground emergence rhythms in field-collected H. parallela, and found that this species has a very stable circabidian rhythm under constant darkness. Furthermore, the rhythm was perfectly entrained to every two cycles of 12 h light and 12 h darkness. We also examined phase responses to light pulses, which suggested that the circabidian rhythm is determined by a circadian clock. Finally, we performed mark–recapture studies, which demonstrated that the 2-day periodicity was not as rigid as that observed under solitary laboratory conditions, and the beetles switched the day of appearance in the field. The circadian clock-driven mechanisms might contribute to a half-cycle temporal shift of the circabidian rhythm.

Adult large black chafers, H. parallela (Motschulsky 1854), were collected on the riverbed of the Yamato River (34°35′14″N, 135°30′16″E) in July of 2012 (Fig. 1) and the riverbed of the Yodo River (34°43′41″N, 135°31′31″E) in June to August of 2012–2015 (Figs 2 and 3).

The beetles were individually kept in a polystyrene cylindrical container (11.5 cm height, 7 cm diameter). Two-thirds of the container was filled with culture soil (Sakata Seed Corporation, Yokohama City, Japan), and the side wall of each container was covered with aluminium foil to prevent light exposure to beetles under the soil surface. Activity of the beetles on the ground was individually recorded under 12 h light and 12 h darkness (12 h:12 h L:D) for 10 days and subsequent constant darkness (DD) for 10 days at 25±1°C. The light source was fluorescent light and the intensity was 1.75 W m⁻². During the recording period, a leaf of the Japanese cherry Prunus yedoensis ‘Somei-yoshino’ was provided as food every 5 days or once at the beginning of the experiment, and approximately 25 ml of water was sprayed on the soil surface when it dried. The top of the container was covered with a transparent glass plate. A color image of the surface of the soil was taken every 6 min with a web camera (DC-NCR13U, Hanwha Japan, Tokyo, Japan). According to differences in the total pixel values between two serial photos, beetle movements were detected and plotted on a double-plotted actogram, and the free-running period was determined using a chi-square periodogram (Enright, 1965; Sokolove and Bushell, 1978).

where (x, y) represents pixel coordinates, R_n(x,y) represents red values per pixel, G_n(x,y) represents green values per pixel and B_n(x,y) represents blue values per pixel. We also calculated the threshold value of w based on the amount of noise from all recorded photos for each beetle. w values above threshold indicated that the beetle had moved since the previous photo was taken, and a score of ‘1’ was plotted for that time. w values below threshold were given a score of ‘0’. Activity occurring over a 48 h period was plotted on a double-plotted actogram.

To examine the biological clock cycle driving the circabidian rhythm, the phase responses of the rhythm to light pulses were examined under DD. We used light-transparent sponges instead of soil. Water was added to the sponge through a cotton thread from a water tank set below the container. A cherry leaf was given as food, and the leaf was replaced twice during the recording period in most cases. In other cases, the leaf was only provided at the beginning of the recording period. Activity recordings were conducted for 9–10 days under DD; a 3 h-light pulse was then provided twice in 48 h at different phases of the circabidian time (CbT), and activity was then recorded for 8–10 days under DD. The light intensity measured inside of the sponge pile was 0.626 W m⁻².

An actogram was drawn as the soil activity recordings were conducted. We set the activity onset as a phase reference point of CbT=36. A computer program was written to determine the onset of the activity phase as follows. (1) We first searched for the most active 18 h time zone over 48 h, starting at 10:00 h Japan standard time (JST). Total activity scores were calculated, and ranged from 0 to 180 for each 18 h time zone. The starting time of each 18 h time zone was delayed by 1 h, so the first zone started at 10:00 h JST and ended at 18:00 h JST, the second one started at 11:00 h JST and ended at 19:00 h JST, and so on. Time zones with the highest scores were selected. If two or more high-scoring zones were present, the earlier zone was selected. (2) In the high zone, the first appearance of 15 sets of continuous activity was selected, and the starting time (JST) of the appearance was designated as the activity onset point. The activity onset was determined every 48 h starting at 10:00 h JST. (3) We then evaluated the activity onset point. There was a maximum of five activity onset points before and after the light pulse. The median activity onset time before and after the light pulses was calculated for each beetle, and activity onset times that differed from the median by 6 h or more were regarded as outliers. In the 48 h cycle with outlier points, the first appearance of 15 sets of continuous activity was searched again during a period of the median ±6 h, and the starting time was set as the revised activity onset points. (4) To determine the CbT of the onset of activity, a line with inclination of τ from the chi-square periodogram was drawn at the position where the sum of the squared deviations of the activity onset point from the line was lowest. Based on τ and the position of the line, we calculated the CbT of the activity onset point, and the difference between the onset of the activity phase in the period before versus after the light pulse was calculated in CbT.

Field observations were conducted on the riverbed of Yamato River in Osaka (Fig. S1). Temperature in the field ranged from −2.4 to 36.7°C in 2011 and −3.1 to 37.9°C in 2012.

The mark–recapture study was conducted on tree E (Ulmus parvifolia, 2.7 m) in 2011, and trees A (U. parvifolia, 3.4 m) and E in 2012. The distance between trees A and E was 87 m (Fig. S1). We checked the presence of beetles on the trees from 3 May to 8 September in 2011 and from 30 May to 26 September in 2012. During the observation period, the observer went to the trees at sunset every day with a lantern, and H. parallela aggregating on the tree were collected using an insect net and a stepladder. An individual three-digit number was engraved on the elytra of each beetle, and the number of newly captured (unmarked) beetles and marked beetles was recorded separately every night. Beetles aggregating on 10 trees, designated A–J in Fig. S1, were counted at sunset on 28–29 July 2011 with the help of five observers. Beetles on the tree were collected, and were immediately released after the marked number on the elytra was checked and the number of unmarked beetles was counted.

The sex pheromone of H. parallela, which has two components (major component, l-isoleucine methyl ester; minor component, linalool) was prepared (Leal et al., 1992). Ammonia was added to l-isoleucine methyl ester hydrochloride (C₇H₁₅NO₂·HCl; CAS no. 18598-74-8, Wako Pure Chemical Industries, Ltd, Osaka, Japan) to obtain l-isoleucine methyl ester, and l-isoleucine methyl ester and linalool (C₁₀H₁₈O; CAS no. 78-70-6, Nacalai Tesque, Inc., Kyoto, Japan) were mixed at a 4:1 ratio. The pheromone solution (100 µl) was placed in a 500 µl plastic tube, and a piece of filter paper was used as the pheromone source. A 1.5 l plastic bottle was set up under the pheromone tube as a non-lethal trap to collect male adults, and the trap was hung on tree A. From 17 June to 8 October 2010, the number of H. parallela captured in the trap and the number of beetles aggregating on tree A was examined every 8 days except for 27–28 August. The pheromone for this trap was renewed at 12:00 h JST every day. The number of beetles was counted every hour from 06:00 h JST until 06:00 h JST of the following day, and collected beetles in the trap were immediately released after counting was completed. Most beetles were alive, and the number of beetles aggregated on tree A was counted by eye and categorized into the following three groups: −, n=0; +, n=1–9; ++, n=10 or more.

Mann–Whitney U-tests, t-tests (one-sided test), chi-square tests, chi-square goodness of fit tests and two-way ANOVA were conducted by using R software (Ihaka and Gentleman, 1996; http://cran.r-project.org) with an additional package twoway.anova (http://aoki2.si.gunma-u.ac.jp/R/src/twoway_anova.R). For detection of rhythmicity, chi-square periodogram analysis was performed (P<0.05), and a period with the peak value of the variance in the analysis was determined as the rhythm period (Enright, 1965; Sokolove and Bushell, 1978).

Under 12 h:12 h LD conditions, male and female beetles appeared on the ground at the beginning of the dark phase every 2 days (Fig. 1). We occasionally observed that the beetles appeared just below the soil surface and remained there until the light was turned off. After the light was turned off, the beetles then emerged onto the soil, and were observed feeding and walking. The duration on the ground was 6.3±1.6 h (mean±s.d., n=10), and the beetles mostly stayed underground for the rest of the day and all of the following day. In the container, the beetles usually stayed about 5 cm below the soil surface during the rest phase. Even under DD, their emergence rhythm continued, and the free-running period was 47.9±0.3 h (n=7) in males and 47.6±0.3 h (n=3) in females. Holotrichia parallela exhibited a clear endogenous rhythm with a period of approximately 48 h under DD, and it entrained to two cycles of 12 h:12 h LD (Fig. 1).

Because the free-running period before the light pulse did not differ significantly between females (47.6±0.3 h, n=28) and males (47.8±0.5 h, n=36, t-test, P>0.05), we plotted the phase shift values of females and males together (Fig. 2). In 61 of the 65 beetles (one individual unsexed), the circabidian rhythm continued after the light pulses, and the free-running period was 47.7±0.4 before the light pulse and 47.7±0.6 h afterwards (paired-sample t-test, P>0.05, n=61). However, the onset of the activity phase was advanced, delayed or unchanged depending on the pulse phase (Fig. 2A–F). We calculated each phase shift value in CbT and plotted the shift value for the phase at which the light pulse was given (Fig. 3A). The CbT is a time scale covering one full circabidian period (∼48 h) during an oscillation under DD, and we set the onset of the activity phase at 36 h in CbT. To examine the dependency of shift values on circabidian phases, we divided a 48 h cycle in CbT into eight periods (t₁–t₈) of 6 h each to compare phase shift values between two consecutive periods (Fig. 3A). Significant differences were detected between the phase shift values of t₃ (median=−1.6 h) and t₄ (median=1.25 h, Mann–Whitney U-test, P=0.004), between t₆ (median=0.4 h) and t₇ (median=−2.6 h, U-test, P=0.002), and between t₇ and t₈ (median=−0.4 h, U-test, P=0.046). At the beginning of t₈, it appeared that the phase shift direction changed from delayed to advanced. In contrast, shift values were small, and a significant difference was not observed between t₁ (median=0.3 h) and t₂ (median=−1.9 h, U-test, P=0.112), between t₂ and t₃ (U-test, P=1.000), between t₄ and t₅ (median=0.4 h, U-test, P=0.262), and between t₅ and t₆ (median=0.4 h, U-test, P=0.800). If the circadian clock cycles twice in a 48 h CbT, a change of phase shift values hypothetically occurs between t₃ and t₄, and between t₇ and t₈. Our data agree with this. Less-responsive periods to light theoretically occur at t_1–2 and t_5–6, and significant differences were not detected in these periods. The lack of a significant difference between t₂ and t₃ and between t₄ and t₅ suggests that phase shift values in t_3–4 are weaker than those in t_7–8 because during t_3–4, H. parallela do not receive light under soil and their sensitivity might be weakened. When phase shift values in the first and second 24 h CbT period were superimposed, phase shift direction appeared similar between the first and second cycles (Fig. 3B). When the 24 h time frame of the superimposed plot was divided into four periods (T₁–T₄), significant differences were not detected between T₁ (median=0.2 h) and T₂ (median=−0.7 h, U-test, P=0.413), but were detected between T₂ and T₃ (median=−1.8 h, U-test, P=0.006), and between T₃ and T₄ (median=0.4 h, U-test, P=0.002).

Circadian-like activity rhythms were observed after the light pulse in four of the 65 beetles (Fig. 2G, Fig. S2). This suggests that H. parallela exhibits an oscillator with a cycle of about 24 h. Because the circadian-like rhythm was observed in response to light pulses given at a variety of phases (5.0, 8.0, 15.5, or 43.0 h in CbT), no specific phase seemed to change from a circabidian to a circadian rhythm.

To understand the emergence times in the field, we counted the number of male H. parallela collected using a pheromone trap set on tree A (Chinese elm Ulmus parvifolia) every hour for 24 h (Fig. S1). Male beetles were mostly trapped within a few hours of sunset from June to October (Fig. 4). A large number of H. parallela were trapped from 25–26 June to 12–13 August, and beetles continued to appear until 7–8 October. In addition to the trapped males, male and female H. parallela appeared on the tree on which the trap was set. These beetles stayed on the tree throughout the night, and mating was often observed within 1 h of sunset. Holotrichia parallela stayed on the tree throughout the night, and their appearance and disappearance were mostly synchronized with sunset and sunrise, respectively (Fig. 4).

To examine the individual rhythms in the field, we conducted a mark–recapture study on tree E in 2011 and on trees A and E in 2012 (Fig. S1). The first appearance dates of H. parallela were 3 June in 2011 and 4 June in 2012. We set 3 June as a reference date, and named it the ‘first’ day in both 2011 and 2012. Days of emergence were distinguished between odd days and even days. The 1st (3 June), 3rd (5 June), 5th (7 June), 7th (9 June), etc., days were designated as ‘odd days’ and the 2nd (4 June), 4th (6 June), 6th (8 June), 8th (10 June), etc., days were the ‘even days’ for both years. Beetles were marked at the first appearance. Beetles marked on an odd day were categorized into the ‘odd day’ group, and those marked on an even day were placed in the ‘even day’ group.

In 2011 on tree E, the number of beetles in the even and odd day group did not differ significantly (Table 1). The rate of recapture of beetles on the marked tree two times or more was 26.1%, and the number of recaptured beetles was 37. Twenty-four of 37 beetles appeared only on an even day, counted from the first appearance (the marked day), during the observation period. For example, male specimen no. 67 appeared 2, 4, 8, 10, 14, 16, 18 and 22 days after the marked day (Fig. 5). The other 13 beetles reappeared on odd numbered days, counted from the first appearance. Female specimen no. 2 reappeared on day 2, 4, 6, 8, 9, 13, 15, 17 and additional days after the first appearance (Fig. 5). We named the latter behavior ‘temporal switching’, and the occurrence of temporal switching between the odd day and even day groups did not differ significantly in 2011 (Table 2). Of the 13 beetles, nine, three and one switched appearance once, twice and four times, respectively (Fig. S3). The last day of appearance in 2011 was 1 September.

In 2012, the last appearance day was 19 September on trees A and E. More beetles were found on the larger tree A than on tree E, but the rate of recapture of beetles on the tree two times or more did not differ significantly between trees A and E (Table 1, chi-square test, χ₂=3.396E–30, d.f.=1, n=459, P>0.05). In contrast to the 2011 population, the population size in 2012 was significantly larger in the odd day group than in the even day group on tree E, but it was not different on tree A (Table 1 and Fig. 6). In 2012, 33 beetles from two trees exhibited temporal switching, and 29, 3 and 1 beetles switched appearance days once, twice and three times, respectively. For beetles marked on trees A and E, the temporal switching rate was significantly larger in the even day group (Table 2). Most beetles were observed repeatedly only on the tree where the beetle was marked, but some reappeared on different trees. For instance, female specimen no. 330 first appeared on tree A, but it appeared on tree E the second time and tree A at a later time (Fig. 5). This phenomenon was referred to as spatial switching. The spatial switching rate was about 10%, and it did not differ significantly between beetles marked on trees A and E (Table 2; Fig. S3).

Fig. 6 shows seasonal variation in the emergence of beetles. We counted the number of male and female beetles separately each day, and a significant difference in the number of beetles was detected between days but not between sexes (two-way ANOVA, P>0.05 for sex, P<0.01 for day; Table S1). In 2011 at tree E, the odd day and even day groups mostly reappeared on different days in June and July. However, in August and September, when very few beetles appeared, synchronization of appearance days occurred between the two groups (Fig. 6A). In 2012, differences in population size were observed between the odd day and even day groups on tree E (Table 1). Two-day periodic appearances were only obvious in the large population of the odd day group (Fig. 6B). On tree A, an abrupt disappearance of unmarked (newly captured) beetles occurred in the even day group after 20 June (Fig. 6B). There was heavy rain from Typhoon No. 4 and a tropical cyclone that occurred on 19–22 June (Japan Meteorological Agency 2012; http://www.data.jma.go.jp/fcd/yoho/data/typhoon/T1204.pdf). The heavy rains led to the flooding of the ground at the observation site on 22 June (an even day), and no beetles appeared on that day on trees A and E (see inset in Fig. 6B). The beetles appeared again on 23 June. After this submersion, beetles from the even day group exhibited temporal switching (appearing on the odd days), and complete synchronization occurred for 9 days from 23 June to 1 July between the odd day and even day groups on both trees, with the exception of one specimen on 28 June on tree E (Fig. 6B). This explains the high percentage of temporal switching in the even day group in 2012 (Table 2). After the heavy rain, more beetles were newly captured on the odd days than on the even days, suggesting that many of the unmarked beetles also exhibited temporal switching after the submersion.

Because the percentage of recaptured beetles on trees A and E was not very high, we suspected that the beetles might be appearing on neighboring trees. We examined the occurrence of beetles marked on tree E and unmarked beetles on 10 trees (including tree E) on 28 and 29 July 2011 (Fig. S1; Fig. 7). The total number of appearances for 2 days on the 10 trees was 112, but no beetles were captured at trees B, F and J. Beetles originally marked on tree E were mostly found on tree E (n=14). Regarding other trees, only a single marked beetle was found on trees A (sex, unknown) and G (female), but all other beetles were unmarked.

We chased seven beetles that flew away from tree A around sunrise in June and July of 2012, and found that they dug into the soil in an area of 15 m diameter around tree A (Fig. 8). These observations suggested that most H. parallela individuals repeatedly visited the same tree that was close to their daytime resting place.

The present study revealed that H. parallela exhibited clear endogenous circabidian rhythms with regard to its appearance on the ground under laboratory conditions. Individual appearances in the field also fitted an approximate 2-day cycle, but individuals often switched appearance days, presumably based on environmental conditions.

Circabidian rhythms were observed as ∼50 h periodicity of sleep–wakefulness in humans under conditions free of time cues (Czeisler et al., 1980) and ∼46 h of flight activity rhythms in the mosquito C. incidens under constant conditions (Clopton, 1984). These circabidian rhythms were labile and subsequently returned to circadian or became obscure. Because circabidian rhythms in human subjects are associated with sleep–wakefulness activities observed during desynchronization from body temperature rhythms that oscillate in a circadian manner, researchers concluded that circabidian rhythms resulted from the uncoupling of multi-oscillators under constant conditions (Aschoff et al., 1967). Clopton (1984) basically followed this idea to interpret the rhythms of mosquitos. In these reports, the circabidian rhythm was thought to be expressed only under non-natural conditions as one unique character of the coupled circadian clock system. However, the circabidian rhythm observed in H. parallela clearly differed from labile rhythms. In H. parallela, the 2-day rhythm appeared stable under both LD and DD experimental conditions, and the rhythm was also observed in the field. Their circabidian appearance or activity on the ground might contribute to a reduction in prey risk and conservation of energy for reproduction. A lower appearance level seems disadvantageous because the opportunity to find a mate is diminished by half. Holotrichia parallela might cope with a reduction in mating opportunities by increasing mating efficiency. In fact, the mating period in H. parallela is short compared with that of other congeneric species (Matsumoto, 2010). They might increase mating frequency to copulate with more partners than other congeneric species.

The circabidian rhythm in H. parallela entrained to two cycles of 12 h:12 h LD, with activity every other night. Because beetles stay underground during the day, they are barely exposed to light. Beetles with a free-running period shorter than 48 h probably came up to the surface of the ground before the lights were turned off, thus following an internal clock. Actually, we occasionally observed movement of the ground surface before the lights were turned off on the appearance day, suggesting that beetles came up just below (not on) the soil surface to sample the light. On non-appearance days, light-sampling behavior was not observed. Light-sampling behavior has been shown in some cavern-dwelling bats and the flying squirrel Glaucomys volans (Twente, 1955; DeCoursey, 1986), which leaves the den after arousal to sample light through the sampling porthole. If light was seen, the squirrel returned to the den to take a short nap before venturing out again (DeCoursey, 1986). When the light was off during sampling, the squirrel left the den. This unusual entrainment process is similar to that in H. parallela.

Some H. parallela individuals displayed a free-running period longer than 48 h, which allowed them to come up to the surface of the ground after the light was off and to sample light at dawn to entrain LD cycles. In the field, the beetles actually stayed on the tree throughout the night until dawn (Fig. 4). Under laboratory conditions, H. parallela only stayed on the ground for 6 h after the light was turned off. We think that this resulted from confined solitary conditions without a mate and the inability to fly (Fig. 1).

Holotrichia parallela probably monitor light intensity to determine the timing of emergence on the ground. In a congeneric species, Holotrichia loochooana, adult emergence and mating behaviors were observed in the laboratory under natural light conditions. Emergence on the ground and female calling behaviors occurred at a certain range of light intensity (Kawamura et al., 2001). As H. parallela mating time is limited (Yoshioka and Yamasaki, 1983), it seems important for H. parallela individuals to wait for the start of darkness just below the surface of the ground and to appear quickly on the ground to facilitate population synchronization.

To determine the cycle of the clock that drives the circabidian rhythm, we examined the phase responses of the rhythm to the zeitgeber light. Researchers have demonstrated that phase advances or delays in rhythm in response to a zeitgeber depend on clock phases, and this is a unique characteristic of oscillator-type clocks (Pittendrigh, 1960). Phase-response curves have been constructed for different kinds of biological rhythms, and curve periods of approximately 24 h, 12.4 h and 1 year have been revealed in circadian, circatidal and circannual rhythms, respectively (Pittendrigh and Minis, 1964; Akiyama, 1997; Miyazaki et al., 2005; Satoh et al., 2008). In the circadian clock, clock phases are divided into the subjective day and night periods. In the cricket Gryllus bimaculatus, circadian activity rhythms exhibit little response to light pulses during the subjective day, but a delay in the first half and an advance in the last half of the subjective night were observed (Okada et al., 1991). In H. parallela, we observed two sets of the less-responsive period and more-responsive (delay or advance) period in one circabidian cycle. This result suggests that a circabidian cycle is composed of two cycles of the circadian oscillator. However, phase responses were too obscure to draw a fine curve. To obtain clear phase-response curves a greater number of beetles might be necessary. Meanwhile, four beetles showed circadian-like activity rhythms after light pulses, suggesting that H. parallela are capable of driving activity in a circadian manner. Phase responses to light pulses by the circabidian rhythm and change of the circabidian rhythm to a circadian-like activity rhythm support the idea that the circadian clock generates the circabidian rhythm in H. parallela.

Clopton (1984) proposed a mechanism for the circabidian rhythm in mosquitos, based on a coupled oscillator model. When mosquitos or humans are kept free from timing cues, oscillator coupling between circadian clocks becomes weakened, and a labile oscillator produces a 2-day rhythm that is uncoupled from the rigid 24 h oscillator. The uncoupled labile oscillator lengthens its period and recouples to a rigid 24 h oscillator when its period double. Another mechanism for creating a longer rhythm than intrinsic oscillators has been hypothesized as beats (Bünning, 1962). Beats originate from a difference in period length between oscillators that are not coupled. Semi-lunar periodic phenomena may originate from the cooperation of a daily (about 24 h) and a tidal rhythm (about 12.4 h). However, we think clock mechanisms of H. parallela rhythms are different from oscillator coupling or beats, because the appearance of the circabidian rhythm is rigid, not labile, both under LD and constant conditions, and no oscillator components other than the 24 h oscillator were detected in phase-response experiments. Further rhythmic components with different cycles producing a beat every 48 h are inconceivable.

We suggest the most parsimonious interpretation is that the circabidian output is produced every two oscillations of the circadian clock entraining with 24 h light–dark cycles. The frequency demultiplication hypothesis, in which biological rhythms with a long period are derived from rhythm with a short period through a process of frequency demultiplication, has been supported in the circasemilunar rhythm in Pontomya oceana (Soong and Chang, 2012). The results indicate the presence of the counter-mechanisms in which cycle numbers of the circadian oscillations are counted to make circasemilunar emergence rhythm. To make the circabidian rhythm, two circadian cycles might also be counted in H. parallela.

Spoelstra et al. (2016) suggested that biological clocks with cycles that do not coincide with environmental cycles are eliminated by natural selection. With regard to biological clocks, the evolution of a cycle that is not associated with an environmental cycle may be difficult. If there are adaptive values associated with a 2-day rhythm, it might be easy to produce circabidian rhythms by modifying the output systems of the circadian clock, which releases a signal every two cycles of the clock.

The 2-day periodicity of individuals was observed in the field. However, its periodicity was not constant, and the absence of a period was often observed. As beetles marked on tree E were seldom found on other trees (Fig. 7), we think that H. parallela individuals rarely go to other trees, but they frequented bushes around the tree on which they were marked. We also observed them eating Poaceae grasses.

In the field, H. parallela occasionally showed temporal switching of the 2-day rhythm. The period 22–23 June 2012 represented a time when beetles were not able to come out on the ground because of heavy precipitation, which resulted in the synchronization of appearances in the odd day and even day groups. For this synchronization, the even day group changed to appear on odd days. In 2012, the majority of beetles started to appear on tree E and were marked from late June after the heavy precipitation, thus, resulting in population size differences between odd and even day groups. Such a clear synchronization was not found in 2011, although there was heavy rain in September. Emergence synchronization seems to depend on environmental changes, and rain is only one factor.

In 2011, some even day beetles showed temporal switching that resulted in a synchronized occurrence with the odd day group in August and September when population size was smaller. It is known that H. loochooana and Dasylepida ishigakiensis, which belong to the same Melolonthinae subfamily as H. parallela, exhibit calling and mating behaviors that occur in the population at a fixed period of the day (Kawamura et al., 2001; Arakaki et al., 2004; Yasui et al., 2007; Fukaya et al., 2009; Tokuda et al., 2010). Yoshioka and Yamasaki (1983) reported that H. parallela mate soon after sunset, and we also confirmed this in the field. Males were caught by pheromone traps within a few hours of sunset, indicating that males quickly fly toward pheromones, and mating subsequently occurs. If copulation occurs during a restricted time of day in these species, sympatric and synchronized population occurrences are needed to increase copulation efficiency. Temporal switching in H. parallela may play a role in the synchronization of sympatric populations. Considering the fact that individual activities recorded in the laboratory never switched to the next day, temporal switching was probably induced by environmental and social cues. Holotrichia parallela uses circabidian rhythms in a flexible manner to change emergence days in order to adjust to the changing environment.

How does the temporal switch occur? If the circabidian rhythm was made by a hypothetical circabidian clock, a phase shift of a half-cycle (∼24 h) of the clock had to occur to switch the appearance days. For an oscillator-type clock, it is hard to make a half-period phase shift at one time (Benstaali et al., 2001), and it usually takes a transient period to complete a full shift. If temporal switching was adaptive for the beetle (e.g. facilitated avoidance of aversive conditions or increased population size), the beetles had to develop some mechanism to make a shift of appearance occur one day at a time. The circabidian rhythm driven by the circadian clock system might facilitate an immediate switch in appearance day. Circabidian output might be activated or suppressed after counting two circadian oscillations. If so, some unknown environmental stimulus may modulate the day-counting system to give an output after one or three circadian oscillations, thus resulting in temporal switching. Here, we propose a novel functional circabidian rhythm that is affected by the circadian clock. In future experiments, we aim to clarify the molecular and neuronal bases of the involvement of the circadian clock in the circabidian rhythm.

We are grateful to Dr Takeshi Matsumoto of the Japanese Society of Scarabaeoidology for his advice on the field study. We would also like to thank Dr Tetsuro Shinada and Dr Tetsuo Tsujioka at Osaka City University and Dr Yasuo Mukai at Local Organization for Promotion of a Japanese Rural Area ‘Nitta’ for their technical advice on pheromone synthesis, image processing analysis and mark–recapture method. This work was supported by a Grant-in Aid for Scientific Research (Houga) 23657060 provided by the Japan Society for Promotion of Science to S.S.